KENYA POVERTY AND GENDER ASSESSMENT 2015/16 Reflecting on a Decade of Progress and the Road Ahead Pov rt & Equit September 16, 2018 © 2018 The World Bank 1818 H Street NW, Washington DC 20433 Telephone: 202-473-1000; Internet: www.worldbank.org Some rights reserved This work is a product of the staff of The World Bank. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this work is subject to copyright. Because The World Bank encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Attribution—Please cite the work as follows: “World Bank. 2018. Kenya Poverty and Gender Assessment 2015/16 - A Decade of Progress and the Challenges Ahead. © World Bank.” All queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. Photo Credits: World Bank Design: Robert Waiharo TABLE OF CONTENTS Abbreviations...................................................................................................................................................................................................................................................................... i Executive Summary........................................................................................................................................................................................................................................................ iii Kenya made progress in reducing poverty and inequality over the past decade................................................................................................................ iii Despite progress in reducing poverty, several challenges remain................................................................................................................................................. x Women are left behind in many areas.............................................................................................................................................................................................................. xii Accelerating poverty reduction .......................................................................................................................................................................................................................... xvi 1. KENYA IN CONTEXT............................................................................................................................................................................................................................................. 1 1.1 Macroeconomic performance over the last decade .............................................................................................................................................................. 2 1.2 Fiscal policy and economic growth.................................................................................................................................................................................................... 8 1.3 A review of some policies over the last decade........................................................................................................................................................................... 11 1.4 Overview of monetary poverty.............................................................................................................................................................................................................. 13 1.5 Overview of non-monetary poverty................................................................................................................................................................................................... 19 1.6 Institutional context, elections and devolution........................................................................................................................................................................... 23 1.7 Perceptions on democracy, governance and political participation.............................................................................................................................. 27 2. THE EXTENT AND EVOLUTION OF POVERTY AND INEQUALITY IN KENYA .......................................................................................................... 31 2.1 Steady but modest progress against poverty 2005/06-2015/16 ..................................................................................................................................... 32 2.2 The incidence of progress, shared prosperity and inequality.............................................................................................................................................. 42 2.3 What explains the trends in poverty reduction? Poverty decomposition exercises............................................................................................. 49 2.4 Poverty profiles – What are characteristics of the poor in Kenya?.................................................................................................................................... 55 3. GENDER AND POVERTY.................................................................................................................................................................................................................................... 59 3.1 A profile of poverty and gender in Kenya ..................................................................................................................................................................................... 60 3.2 Gender gaps in endowments.................................................................................................................................................................................................................. 62 3.3 Gender inequality in economic opportunities............................................................................................................................................................................. 70 3.4 Voice and agency............................................................................................................................................................................................................................................ 80 4. AGRICULTURE AND RURAL POVERTY.................................................................................................................................................................................................. 83 4.1 The decline in rural poverty has been the main driver of poverty reduction nationally................................................................................... 84 4.2 Diversifying away from agriculture improves livelihoods..................................................................................................................................................... 85 4.3 Non-agricultural employment is becoming increasingly important for rural households ............................................................................. 86 4.4 Farm productivity has stagnated while commodity prices have increased.............................................................................................................. 89 4.5 Increased market participation can further reduce rural poverty .................................................................................................................................. 97 4.6 Conclusions......................................................................................................................................................................................................................................................... 100 5. URBANIZATION....................................................................................................................................................................................................................................................... 101 5.1 Urbanization and poverty ......................................................................................................................................................................................................................... 102 5.2 Diagnostic of urban poverty..................................................................................................................................................................................................................... 109 5.3 Urban labor markets...................................................................................................................................................................................................................................... 116 5.4 Urban informal settlements...................................................................................................................................................................................................................... 123 6. EDUCATION AND POVERTY......................................................................................................................................................................................................................... 127 6.1 Kenya’s education sector ........................................................................................................................................................................................................................... 128 6.2 Enrollment .......................................................................................................................................................................................................................................................... 129 6.3 Learning outcomes........................................................................................................................................................................................................................................ 136 6.4 The supply-side................................................................................................................................................................................................................................................. 139 6.5 Teacher incentives and school governance................................................................................................................................................................................... 143 6.6 Summary and policy options................................................................................................................................................................................................................... 146 7. HEALTH AND POVERTY................................................................................................................................................................................................................................149 7.1 Background.................................................................................................................................................................................................................................................... 150 7.2 Health outcomes and uptake through an equity lens ...................................................................................................................................................... 157 7.3 The supply side: Physical inputs, health professionals, and incentives..................................................................................................................... 168 7.4 Summary and policy implications................................................................................................................................................................................................... 174 8. VULNERABILITY, SHOCKS,AND SOCIAL PROTECTION .....................................................................................................................................................177 8.1 Introduction .................................................................................................................................................................................................................................................. 178 8.2 Vulnerability .................................................................................................................................................................................................................................................. 180 8.3 Shocks and coping strategies in 2005/06 and 2015/16..................................................................................................................................................... 186 8.4 The coverage and impact of social protection programs ............................................................................................................................................... 194 References........................................................................................................................................................................................................................................................................253 LIST OF FIGURES Figure 1: Kenya’s economic and poverty progress .......................................................................................................................................................................... iv Figure 2: Share of income by source for rural households ......................................................................................................................................................... vi Figure 3: Non-monetary dimensions of wellbeing........................................................................................................................................................................... viii Figure 4: Regional patterns in poverty..................................................................................................................................................................................................... xi Figure 5: Poverty and vulnerability in Kenya......................................................................................................................................................................................... xiii Figure 6. Gender gaps in Kenya.................................................................................................................................................................................................................... xv Figure 7: Socio-economic indicators of rural Kenya......................................................................................................................................................................... xvii Figure 8: Urbanization remains a challenge for poverty reduction........................................................................................................................................xviii Figure 1.1: Kenya’s GDP growth from 2005 to 2015............................................................................................................................................................................. 3 Figure 1.2: Annual GDP growth for Sub-Saharan Africa and selected countries, per year and between 2005 and 2015........................ 3 Figure 1.3: Contributions to GDP growth................................................................................................................................................................................................... 4 Figure 1.4: Agriculture and GDP growth..................................................................................................................................................................................................... 5 Figure 1.5: Productivity and economic growth ..................................................................................................................................................................................... 5 Figure 1.6: Demand-side contribution to growth between 2005 and 2015........................................................................................................................ 5 Figure 1.7: Contributions to real GDP growth.......................................................................................................................................................................................... 6 Figure 1.8: Contributions to GDP growth, regional comparison ................................................................................................................................................ 7 Figure 1.9: Contributions to real GDP per capita growth ................................................................................................................................................................ 7 Figure 1.10: Sectoral contribution to change in real GDP per capita productivity.............................................................................................................. 8 Figure 1.11: Productivity contribution to real GDP per capita growth ...................................................................................................................................... 8 Figure 1.12: Spending has consistently exceeded revenue collection ...................................................................................................................................... 10 Figure 1.13: Revenue collection has not kept up with spending pressures ........................................................................................................................... 10 Figure 1.14: The evolution of fiscal deficit..................................................................................................................................................................................................... 10 Figure 1.15: Sectoral contribution to growth in total spending .................................................................................................................................................... 11 Figure 1.16: Employment trends ....................................................................................................................................................................................................................... 13 Figure 1.17: Poverty at the US$ 1.20, 1.90, and 3.20 lines.................................................................................................................................................................... 15 Figure 1.18: Cumulative consumption distribution with shock ..................................................................................................................................................... 15 Figure 1.19: GDP sectoral growth simulation of poverty trajectory at international poverty lines, 2005 to 2015 .......................................... 16 Figure 1.20: Overall GDP growth simulation of poverty trajectory at international poverty lines, 2005 to 2015............................................. 16 Figure 1.21: Real sector growth, 2007 to 2015........................................................................................................................................................................................... 16 Figure 1.22: Share of households by sector of household head occupation, 2005 vs. 2015 ....................................................................................... 16 Figure 1.23: Consistent sectoral elasticities for poverty pass-through23.................................................................................................................................. 17 Figure 1.24: Combination of growth and redistribution needed to eradicate poverty in 2030 ................................................................................ 17 Figure 1.25: International comparison of poverty.................................................................................................................................................................................... 18 Figure 1.26: Poverty headcount against GDP per capita..................................................................................................................................................................... 18 Figure 1.27: Poverty rate against depth at international poverty line......................................................................................................................................... 18 Figure 1.28: Poverty headcount at IPL and LMIC, international comparison.......................................................................................................................... 19 Figure 1.29: Poverty gap at IPL and LMIC, international comparison.......................................................................................................................................... 19 Figure 1.30: International comparison of elasticity of poverty reduction................................................................................................................................. 19 Figure 1.31: Elasticity of poverty reduction against GDP per capita............................................................................................................................................. 19 Figure 1.32: Multi-dimensional deprivations, 2015................................................................................................................................................................................. 20 Figure 1.33: Poverty headcount against access to improved water............................................................................................................................................. 20 Figure 1.34: Poverty headcount against access to improved sanitation................................................................................................................................... 20 Figure 1.35: Poverty headcount against HDI............................................................................................................................................................................................... 21 Figure 1.36: Poverty headcount against literacy rates........................................................................................................................................................................... 21 Figure 1.37: Poverty headcount against adult educational attainment, primary................................................................................................................. 21 Figure 1.38: Poverty headcount against adult educational attainment, secondary........................................................................................................... 21 Figure 1.39: Poverty headcount against under-five mortality.......................................................................................................................................................... 22 Figure 1.40: Poverty headcount against child stunting ...................................................................................................................................................................... 22 Figure 1.41: Perception of democracy in sub-Saharan African countries................................................................................................................................. 27 Figure 1.42: Responsiveness of National Assembly members to citizens in sub-Saharan African countries...................................................... 28 Figure 1.43: Major issues for citizens in Kenya that government should address................................................................................................................ 29 Figure 1.44: Perceived involvement in corruption, 2016 (% of respondents)......................................................................................................................... 29 Figure 1.45: Political intimidation or violence during election campaigns.............................................................................................................................. 29 Figure 1.46: Expressing political views in sub-Saharan African countries................................................................................................................................. 29 Figure 2.1: Total fertility rate (women aged 15-49) .............................................................................................................................................................................. 33 Figure 2.2: Trends in absolute, food and extreme poverty, nationally and by area of residence ........................................................................... 34 Figure 2.3: Urban and rural food poverty basket comparison by rank, 2005/06 and 2015/16................................................................................. 35 Figure 2.4: Trends in absolute, food and extreme poverty by province and NEDI classifications .......................................................................... 37 Figure 2.5: Distribution of the poor by province.................................................................................................................................................................................... 39 Figure 2.6: Poverty depth and severity, nationally and by urban/rural strata...................................................................................................................... 40 Figure 2.7: Poverty depth by province and NEDI classification ................................................................................................................................................... 40 Figure 2.8: Proportion of consumption by use, nationally and area of residence ........................................................................................................... 41 Figure 2.9: Differential changes in price indices..................................................................................................................................................................................... 41 Figure 2.10: GICs nationally, by area of residence and NEDI classification .............................................................................................................................. 42 Figure 2.11: Real consumption deciles (2016 prices), nationally and by area of residence........................................................................................... 43 Figure 2.12: Response rates and consumption among urban households............................................................................................................................. 44 Figure 2.13: Annualized consumption growth, nationally, by area of residence and by province .......................................................................... 45 Figure 2.14: Annualized consumption growth compared to benchmark countries......................................................................................................... 46 Figure 2.15: Gini inequality index nationally, by area of residence and by province ........................................................................................................ 47 Figure 2.16: Gini inequality index for select African countries......................................................................................................................................................... 47 Figure 2.17: Atkinson index and P75/P25 inequality index nationally and by area of residence............................................................................... 48 Figure 2.18: Determinants of changes in poverty – Datt-Ravallion decomposition by area of residence........................................................... 50 Figure 2.19: Determinants of changes in poverty – Datt-Ravallion decomposition by province ............................................................................. 51 Figure 2.20: Contribution to poverty reduction........................................................................................................................................................................................ 51 Figure 2.21: Real sector growth 2007–2015................................................................................................................................................................................................. 52 Figure 2.22: Household size and average education level nationally, by province and NEDI classification ....................................................... 57 Figure 2.23: Access to improved sanitation, water and electricity by province, urban/rural, and NEDI/non-NEDI status ........................ 58 Figure 3.1: Male and female poverty rates by age group, 2015/6 .............................................................................................................................................. 61 Figure 3.2: Male and female poverty rates by marital status, 2015/6 ....................................................................................................................................... 61 Figure 3.3: Poverty and household demographic composition, 2015/6................................................................................................................................ 62 Figure 3.4: Regional differences in gender parity in the education sector............................................................................................................................ 63 Figure 3.5: Male and female literacy by county, 2015/6.................................................................................................................................................................... 64 Figure 3.6: Maternal mortality............................................................................................................................................................................................................................ 64 Figure 3.7: Kenya’s demographic transition............................................................................................................................................................................................... 65 Figure 3.8: Household members fetching water, 2015/6................................................................................................................................................................. 66 Figure 3.9: Kenya and comparators gender gaps in land and housing ownership......................................................................................................... 66 Figure 3.10: ICT access by sex and age, 2014, 2015/6............................................................................................................................................................................ 67 Figure 3.11: Financial inclusion, male and female population (15+), 2014.............................................................................................................................. 67 Figure 3.12: Difficulty to come up with emergency funds, male and female population (15+), 2014................................................................... 68 Figure 3.13: Financial inclusion, Kenya and regional comparison, 2014.................................................................................................................................... 68 Figure 3.14: Percent of population employed by category, 2005/6 – 2015/6 ....................................................................................................................... 71 Figure 3.15: Changes in school-to-work transition, 2005/6-2015/6 ............................................................................................................................................ 71 Figure 3.16: Male and female labor force participation, 2015/6...................................................................................................................................................... 72 Figure 3.17: Female labor force participation, Kenya and comparators .................................................................................................................................... 72 Figure 3.18: Geographic variation in male and female labor force participation, 2015/6............................................................................................... 73 Figure 3.19: Correlates of male and female labor force participation, 2015/6 ...................................................................................................................... 73 Figure 3.20: Male and female employment by broad sector, 2015/6.......................................................................................................................................... 74 Figure 3.21: Share of male/female employment by detailed sector, 2015/6.......................................................................................................................... 74 Figure 3.22: Profits of male-, female- and jointly-run household enterprises, 2015/6 ..................................................................................................... 76 Figure 3.23: Gender differences in agricultural employment vs. parcel management, 2015/6 ................................................................................. 77 Figure 3.24: Acceptance of norms that constrain women’s physical mobility....................................................................................................................... 80 Figure 3.25: Share of women (15-49) who experienced physical violence by marital status, 2014.......................................................................... 81 Figure 4.1: Rural poverty headcount and its decline by province ............................................................................................................................................. 84 Figure 4.2: Geographic distribution of the rural poor in Kenya.................................................................................................................................................... 85 Figure 4.3: Share of income from agriculture and non-agricultural sources in rural Kenya........................................................................................ 85 Figure 4.4: Changes in rural non-agricultural economic activities.............................................................................................................................................. 87 Figure 4.5: Female non-agricultural labor allocation........................................................................................................................................................................... 88 Figure 4.7: Share of income from different sources for poor and non-poor households............................................................................................. 88 Figure 4.8: Non-farm economic activity by ISIC classification....................................................................................................................................................... 89 Figure 4.9: Relationship between crop yield and poverty rates at the provincial level in rural Kenya, 2015/16............................................ 90 Figure 4.10: Poverty and crop yield at the county level in rural Kenya, 2015/16.................................................................................................................. 90 Figure 4.11: Relationship between yield decile and poverty rates in rural Kenya, 2015/16.......................................................................................... 90 Figure 4.12: Proportion of cultivated area by crop category in rural Kenya ........................................................................................................................... 91 Figure 4.13: Maize and bean yield in selected African countries.................................................................................................................................................... 91 Figure 4.14: Heterogeneity in crop productivity across provinces in rural Kenya ............................................................................................................... 92 Figure 4.15: Heterogeneity in crop productivity by gender of household head ................................................................................................................ 93 Figure 4.16: Gender differences in input use in rural Kenya.............................................................................................................................................................. 93 Figure 4.17: Trends in input use by farmers (Tegemeo Panel)......................................................................................................................................................... 94 Figure 4.18: Trends of crop prices and overall prices............................................................................................................................................................................. 97 Figure 4.19: There was an observed reduction in subsistence agriculture in rural Kenya between 2005/06 and 2015/16 ..................... 98 Figure 4.20: Relationship between poverty and market participation....................................................................................................................................... 99 Figure 5.1: Urbanization rates in Kenya and other countries, 1950–2050.............................................................................................................................. 102 Figure 5.2: Poverty headcount ratio and number of poor, 2005/6 and 2015/16............................................................................................................... 103 Figure 5.3: Poverty rates and number of poor in urban areas by province, 2005/6 and 2015/16........................................................................... 104 Figure 5.4: Share of urban poor across counties, 2015/16............................................................................................................................................................... 104 Figure 5.5: County-level urban poverty rates and number of urban poor, 2015/16....................................................................................................... 105 Figure 5.6: County-level urban and rural poverty rates, 2015/16................................................................................................................................................ 105 Figure 5.7: Sectoral decomposition of poverty reduction, 2005/6 and 2015/16 ............................................................................................................. 106 Figure 5.8: Share of recent migrants in urban areas in 47 counties, 2014.............................................................................................................................. 108 Figure 5.9: Wealth index by migration status, 2014............................................................................................................................................................................. 108 Figure 5.10: Share of household expenditure in urban Kenya, 2005/06 and 2015/16...................................................................................................... 110 Figure 5.11: Housing units with non-durable structures in urban areas, 2005/06 and 2015/16................................................................................ 111 Figure 5.12: Access to improved water in provinces by urban/rural area, 2005/06 and 2015/16.............................................................................. 111 Figure 5.13: Access to water in urban Kenya, 2005/06 and 2015/16 .......................................................................................................................................... 112 Figure 5.14: Access to improved sanitation in provinces by urban/rural area, 2005/06 and 2015/16.................................................................... 113 Figure 5.15: Access to improved sanitation in urban Kenya, 2005/06 and 2015/16 ......................................................................................................... 114 Figure 5.16: Access to electricity in provinces by urban/rural area, 2005/06 and 2015/16............................................................................................ 115 Figure 5.17: Access to electricity in urban Kenya, 2005/06 and 2015/16.................................................................................................................................. 115 Figure 5.18: Labor force participation rates in urban Kenya, 2005/06 and 2015/16 ......................................................................................................... 116 Figure 5.19: Unemployment rates in urban Kenya, 2005/6 and 2015/16................................................................................................................................. 117 Figure 5.20: Economic sectors of workers in urban Kenya, 2005/6 and 2015/16................................................................................................................. 117 Figure 5.21: Employment in urban Kenya, 2005/06 and 2015/16 ................................................................................................................................................ 118 Figure 5.22: Job types in urban Kenya, 2015/16 ...................................................................................................................................................................................... 119 Figure 5.23: Commuting modes in urban Kenya, 2005/6 and 2015/16..................................................................................................................................... 121 Figure 5.24: Share of accessible jobs within 60 minutes in Nairobi.............................................................................................................................................. 122 Figure 5.25: Job accessibility and per capita household expenditure in Nairobi................................................................................................................. 123 Figure 5.26: Household consumption and rents in Nairobi’s informal settlement and non-informal settlement areas, 2015/16........ 124 Figure 5.27: Housing quality in African informal settlements........................................................................................................................................................... 125 Figure 5.28: Perceived tenure security in African informal settlements..................................................................................................................................... 125 Figure 5.29: Previous residence of urban households........................................................................................................................................................................... 126 Figure 5.30: Probability of households moving to non-informal settlement areas in Nairobi and Mombasa................................................... 126 Figure 6.1: Public expenditure in education, 2000–2015................................................................................................................................................................. 129 Figure 6.2: GERs in pre-primary, primary, secondary, and tertiary, 2000–2016................................................................................................................... 130 Figure 6.3: NERs and GERs by level, poverty, quintile, and locality, 2015/16......................................................................................................................... 131 Figure 6.4: Changes in primary and secondary enrollment, between 2005/06 and 2015/16, by poverty, quintile, and locality........ 132 Figure 6.5: Gross enrollment rates by grade and year........................................................................................................................................................................ 132 Figure 6.6: GERs in primary and secondary education by county, 2015/16......................................................................................................................... 133 Figure 6.7: Net intake rate and transition by poverty, quintile, and locality, 2005/06 and 2015/16....................................................................... 134 Figure 6.8: Primary gross enrollment by provider, location, and quintile, 2005/06 and 2015/16............................................................................ 135 Figure 6.9: Average and median household per-student expenditure on education by level, location, and provider, 2005/06 and 2015/16..................................................................................................................................................................................................................... 136 Figure 6.10: Knowledge of fourth-grade students across Sub-Saharan African countries, early 2010s................................................................. 137 Figure 6.11: Learning outcomes in mathematics in ten-year-old children by socio-economic background, 2014 ..................................... 137 Figure 6.12: Proportion of twelve-year-old children proficient in mathematics and english, percent, 2014..................................................... 138 Figure 6.13: Physical inputs at the school-level by location and type of provider, primary schools, 2012........................................................... 140 Figure 6.14: Student-teacher ratios in public schools, 2004-2015, and students per classroom in primary schools, 2012....................... 141 Figure 6.15: Cross-country comparison of teacher salaries by level............................................................................................................................................. 142 Figure 6.16: Teachers’ subject knowledge and pedagogical skills by country, early 2010s .......................................................................................... 143 Figure 6.17: Absence from school and absence from class by country..................................................................................................................................... 144 Figure 6.18: Absenteeism rates by type of provider and type of teacher, 2012.................................................................................................................... 144 Figure 7.1: Outpatient visits and institutional deliveries by provider, January 2012 to December 2017............................................................ 150 Figure 7.2: Levels and trends in health expenditure by source, 2004-2014.......................................................................................................................... 151 Figure 7.3: Membership and resources of National Hospital Insurance Fund (NHIF), 2006/07-2014/15............................................................ 152 Figure 7.4: Health outcomes in Kenya vis-à-vis benchmark countries and aggregates, latest year available................................................. 155 Figure 7.5: Annual rate of reduction in selected indicators of childhood health, percent, c. 2000 to 2015..................................................... 156 Figure 7.6: TFR (number of births per woman) and under-five mortality rate (deaths per 1,000 live births).................................................. 156 Figure 7.7: TFRs against under-five mortality, countries (2015) and Kenyan counties (2014)................................................................................... 157 Figure 7.8: Levels in trends in registered deaths by cause, 2011–2015.................................................................................................................................... 157 Figure 7.9: Self-reported instances of sickness or injury during last four weeks prior to the survey as percent of population........... 158 Figure 7.10: Under-five mortality (deaths per 1,000 live births) by quintile, mother’s educational attainment, and location................. 159 Figure 7.11: Stunting rate by quintile, mother’s educational attainment, and location, 2003–2014....................................................................... 159 Figure 7.12: Child health outcomes by county, 2014............................................................................................................................................................................. 160 Figure 7.13: Obesity rates (BMI > 30, share of women aged 15-49) by quintile, educational attainment, and locality, 2003–2014......... 160 Figure 7.14: Selected indicators of health services uptake (%), 2000–2015............................................................................................................................. 161 Figure 7.15: Availability of health facilities and distance to nearest health facility in which a doctor would be on duty, 2015/16..... 162 Figure 7.16: Uptake of curative health services and number of curative visits by quintile and locality, 2005/06 and 2015/16.............. 163 Figure 7.17: Uptake of preventive health services during four weeks prior to interview................................................................................................ 163 Figure 7.18: Uptake of preventive health goods, select indicators, by poverty and quintile, 2015/16................................................................... 164 Figure 7.19: Access to health services and uptake by county, 2014, select indicators..................................................................................................... 165 Figure 7.20: Share of births (of surviving children 60 months and younger) by circumstance, 2005/06 and 2015/16................................ 166 Figure 7.21: Share of deliveries by provider, 2009–2014...................................................................................................................................................................... 166 Figure 7.22: Health insurance coverage, health expenditure and incidence of asset sales in response to hospitalization....................... 167 Figure 7.23: Average shares of in-patient health expenditure by funding source (democratic shares per hospitalized individual).......... 167 Figure 7.24: Infrastructure availability in public and private facilities by type of facility and location (select indicators)........................... 168 Figure 7.25: Drug availability by type of facility, provider, and location..................................................................................................................................... 169 Figure 7.26: Number of health professionals per 10,000 population........................................................................................................................................... 170 Figure 7.27: Salaries of nurses and midwives by country, 2005/06-2015/16.......................................................................................................................... 171 Figure 7.28: Salaries of select health workers in Kenya, 2005/06 and 2015/16...................................................................................................................... 171 Figure 7.29: Adherence to clinical guidelines and absence from health facility by country......................................................................................... 173 Figure 8.1: Poverty and vulnerability in Kenya: 2005/06 and 2015/16..................................................................................................................................... 182 Figure 8.2: Vulnerability rates by county: 2015/16................................................................................................................................................................................ 183 Figure 8.3: Vulnerability rates by poverty status: 2015/16................................................................................................................................................................ 183 Figure 8.4: CDFs of the rural and urban population: 2015/16....................................................................................................................................................... 184 Figure 8.5: Vulnerability rates relative to the average: 2015/16.................................................................................................................................................... 185 Figure 8.6: The prevalence of different shocks over consumption quintiles: 2005/06 and 2015/16.................................................................... 188 Figure 8.7: Prevalence of shocks by urban-rural location: 2005/06 and 2015/16.............................................................................................................. 188 Figure 8.8: Incidence of shocks by poverty status, agricultural households only: 2005/06 and 2015/16.......................................................... 189 Figure 8.9: Shock prevalence for agricultural households only: 2005/06 and 2015/16................................................................................................. 190 Figure 8.10: Geographic distribution of different shocks: 2015/16............................................................................................................................................... 190 Figure 8.11: The severity of losses from shocks: 2005/06 and 2015/16...................................................................................................................................... 191 Figure 8.12: Coping mechanisms over the distribution of consumption: 2005/06 and 2015/16.............................................................................. 192 Figure 8.13: Coping strategies by urban-rural place of residence: 2005/06 and 2015/16.............................................................................................. 192 Figure 8.14: Coping strategies by shock type – Rural households only: 2005/06 and 2015/16.................................................................................. 193 Figure 8.15: Expenditure on social safety nets: 2015.............................................................................................................................................................................. 196 Figure 8.16: Number of households receiving cash transfers: 2013 to 2016........................................................................................................................... 197 Figure 8.17: Coverage and share of beneficiaries by county: 2016............................................................................................................................................... 199 Figure 8.18: Share of beneficiary households by county and program: 2016........................................................................................................................ 199 Figure 8.19: CDFs of consumption by cash transfer program.......................................................................................................................................................... 200 Figure 8.20: The impact of grant receipt on the probability that all school-aged children in the household are enrolled...................... 203 Figure 8.21: The impact of grant receipt on the probability that no school-aged child in the household is working................................ 203 Figure 8.22: The impact of grant receipt on the probability that a household is food secure: HSNP counties only..................................... 204 Figure A.1: TFP growth was a key driver of GDP growth................................................................................................................................................................... 206 Figure A.2: As growth in capital accelerated, growth of labor moderated............................................................................................................................ 207 Figure A.3: Stagnating human capital growth resulted in a moderation of human capital per unit of labor................................................. 207 Figure A.4: The increase in labor force resulted in increasing unemployment and declining labor force participation.......................... 207 Figure A.5: County allocation of ordinary government revenues................................................................................................................................................ 210 Figure A.6: Transfers to county governments, 2016–17..................................................................................................................................................................... 210 Figure A.7: Share of transfers to counties.................................................................................................................................................................................................... 211 Figure A.8: Change in allocation of transfers by share of urban population......................................................................................................................... 211 Figure A.9: Development expenditure share of total expenditure............................................................................................................................................. 211 Figure A.10: Absorption rates of county budgets..................................................................................................................................................................................... 212 Figure A.11: Personnel costs by county.......................................................................................................................................................................................................... 212 Figure A.12: Share of county own revenues................................................................................................................................................................................................. 213 Figure A.13: Cumulative annual growth rate of personnel costs.................................................................................................................................................... 213 Figure A.14: Own revenues as a share of actual county expenditure.......................................................................................................................................... 213 Figure A.15: Average annual increase in own-source revenues...................................................................................................................................................... 213 Figure B.1: Map of NEDI counties..................................................................................................................................................................................................................... 216 Figure B.2: Distribution of the log of population density by cluster type.............................................................................................................................. 217 Figure B.3: Occupational sector of household head by area of residence............................................................................................................................. 217 Figure B.4: Source of food consumption by area of residence...................................................................................................................................................... 217 Figure B.5: Household characteristics by area of residence............................................................................................................................................................ 217 Figure B.6: Asset ownership by consumption quintile, Nairobi.................................................................................................................................................... 218 Figure E.1: Number of urban poor and urban poverty rate by county, 2015/16............................................................................................................... 233 Figure E.2: Cash transfer during the last three months in 15 cities, 2013............................................................................................................................... 233 Figure E.3: Expenditure share on housing in urban Kenya.............................................................................................................................................................. 234 Figure E.4: Expenditure share on housing in urban Kenya by county, 2015/16................................................................................................................. 234 Figure E.5: Comparison of health indicators in Kenya, 2000 to 2014........................................................................................................................................ 235 Figure E.6: Number and share of unemployed population in urban area by county, 2015/16................................................................................ 236 Figure E.7: Unemployment rate in urban area by sex and county, 2015/16........................................................................................................................ 236 Figure E.8: Unemployment rate in urban area by the youth and county, 2015/16......................................................................................................... 236 Figure E.9: Comparison of economic sectors in urban Kenya by county, 2015/16.......................................................................................................... 237 Figure E.10: Duration of residence in 47 counties, 2014...................................................................................................................................................................... 238 Figure E.11: Previous residence of recent migrants in 47 counties, 2014.................................................................................................................................. 239 Figure E.12: Previous residence of recent migrants in 47 countries, 2014................................................................................................................................ 240 Figure E.13: Cumulative distribution of the duration of residence in Nairobi and Mombasa...................................................................................... 241 Figure H.1: Consumption levels of vulnerable households, relative to the poverty line: 2015/16.......................................................................... 250 Figure H.2: The prevalence of shocks by poverty and vulnerability status: 2005/06 and 2015/16......................................................................... 250 LIST OF TABLES Table 1: Access to basic services by poverty status........................................................................................................................................................................ v Table 2: Sectoral decomposition of poverty reduction (Ravallion-Huppi)...................................................................................................................... vi Table 3: Monthly earnings in Ksh, by gender..................................................................................................................................................................................... xiv Table 1.1: List of ongoing major projects................................................................................................................................................................................................. 12 Table 1.2: Key monetary poverty Indicators........................................................................................................................................................................................... 14 Table 1.3: Revenue-sharing among counties in Kenya.................................................................................................................................................................... 24 Table 2.1: Absolute poverty headcount rate, nationally, by area of residence.................................................................................................................. 32 Table 2.2: Poor and total populations, nationally, by area of residence and by NEDI classification..................................................................... 32 Table 2.3: Comparison of noncomparable and comparable 2005/06 poverty rates.................................................................................................... 36 Table 2.4: Theil inequality index - decomposition by urban/rural location and province......................................................................................... 49 Table 2.5: Sectoral decomposition of poverty reduction (Ravallion-Huppi)...................................................................................................................... 53 Table 2.6: Sectoral decomposition of poverty reduction (Ravallion-Huppi) - alternative definition................................................................... 54 Table 2.7: Household characteristics by poverty status................................................................................................................................................................... 56 Table 3.1: Primary and secondary enrollment rates and gender parity, 2005/6 and 2015/6................................................................................... 63 Table 3.2: Male and female wage employment by employment status, 2015/6............................................................................................................ 75 Table 3.3: Male and female monthly earnings, in current Ksh, and male-to-female ratio, 2015/6....................................................................... 75 Table 3.4: Descriptive differences between male- and female-run household enterprises, 2015/6................................................................... 76 Table 3.5: Descriptive differences in input use between male and female decision-makers in agriculture, 2015/6................................ 78 Table 4.1: Decomposition of poverty by income classification.................................................................................................................................................. 86 Table 4.2: Determinants of maize yield, FEs model, 2000–10...................................................................................................................................................... 95 Table 5.1: Recent male migration by origin and destination....................................................................................................................................................... 107 Table 5.2: Median nominal wage by economic sector in urban Kenya, 2015/16............................................................................................................ 120 Table 5.3: Average share of accessible jobs in Nairobi..................................................................................................................................................................... 122 Table 5.4: Poverty rates in informal settlement and non-informal settlement areas, Nairobi 2015/16............................................................. 123 Table 7.1: OLS regression of log salary (incl. allowances) on binary indicator of employment in public sector for auxiliary nurses; nurses and midwives; and medical and clinical officers, 2005/06 and 2015/16...................................................................... 172 Table 7.2: Outcomes for select standardized patient cases in Nairobi, urban China, and India............................................................................. 173 Table 7.3: Primary outcomes for standardized patient cases by sector................................................................................................................................. 174 Table 8.1: Profiles of the poor and the vulnerable: 2005/06 and 2015/16........................................................................................................................... 185 Table 8.2: Coping strategies by poverty status: 2015/16................................................................................................................................................................ 194 Table 8.3: Social Protection Programs in Kenya.................................................................................................................................................................................... 196 Table 8.4: Profile of beneficiary households versus non-beneficiary households (by poverty status)............................................................... 201 Table A.1: Poverty trajectory simulation, sectoral and non-sectoral growth...................................................................................................................... 206 Table B.1: Sampling framework ..................................................................................................................................................................................................................... 216 Table B.2: Response rates by county............................................................................................................................................................................................................ 218 Table C.1: Correlates of labor force participation, probit (coefficients).................................................................................................................................. 222 Table C.2: Oaxaca-Blinder decomposition of gender gaps in monthly earnings, summary..................................................................................... 223 Table C.3: Oaxaca-Blinder decomposition of gender gaps in monthly earnings, descriptive statistics ........................................................... 224 Table C.4: Oaxaca-Blinder decomposition of gender gaps in monthly earnings, OLS (coefficients).................................................................. 225 Table C.5: Correlates of household enterprise profits, OLS (coefficients)............................................................................................................................. 226 Table D.1: Determinants of beans yield, FEs Model............................................................................................................................................................................ 230 Table E.1: Nominal monthly salary in urban Kenya............................................................................................................................................................................ 231 Table E.2: Comparison of dwelling characteristics between informal settlement and non-informal settlement areas in Nairobi....... 232 Table E.3: Comparison of access to services between informal settlement and non-informal settlement areas in Nairobi............... 232 Table F.1: GERs and NERs in secondary and primary education by county........................................................................................................................ 242 Table F.2: Determinants of transition from seventh into eighth grade of primary and from primary into secondary............................ 245 Table G.1: Regression results from LPMs – effect of free deliveries in public facilities on uptake by provider (N = 28,154)................. 247 Table G.2: Regression results from LPMs – effect of free deliveries in public facilities on uptake by provider, urban and rural (N = 28,154)..................................................................................................................................................................................................................... 247 Table G.3: Effect of institutional delivery and assistance on neonatal mortality (odds ratios/t-values) (N = 19,080)................................ 248 Table H.1: Coping strategies by poverty status for agricultural households only: 2015/16 ..................................................................................... 251 LIST OF BOXES Box 1.1: The Big 4 policy agenda....................................................................................................................................................................................................................... 9 Box 1.2: The international poverty lines........................................................................................................................................................................................................ 14 Box 1.3: Public expectations from devolution........................................................................................................................................................................................... 25 Box 1.4: Key features of the 2010 Kenyan Constitution....................................................................................................................................................................... 26 Box 2.1: Kenya Integrated Household Budget Survey (KIHBS): A commendable effort.................................................................................................. 33 Box 2.2: Measuring poverty: Computing the poverty lines, the consumption aggregate and classification of peri-urban households.......................................................................................................................................................................................................................... 35 Box 2.3: Nairobi nonresponse rates – dealing with data issues...................................................................................................................................................... 44 Box 2.4: Inequality measures................................................................................................................................................................................................................................ 48 Box 2.5: What does decomposing changes in poverty entail?....................................................................................................................................................... 50 Box 5.1: Definition of urban areas..................................................................................................................................................................................................................... 103 Box 5.2: Decomposition analysis........................................................................................................................................................................................................................ 106 Box 5.3: Job accessibility......................................................................................................................................................................................................................................... 121 Box 5.5: Profile of residents moving to/from informal settlement neighborhoods........................................................................................................... 126 Box 6.1: Free primary education and the quality of education...................................................................................................................................................... 129 Box 6.2: Are private schools more productive?......................................................................................................................................................................................... 139 Box 6.3: Are higher public-sector wages efficient?................................................................................................................................................................................ 145 Box 7.1: Promises and perils of the devolution of health services................................................................................................................................................ 153 Box 7.2: What works to boost skilled birth assistance for safer childbirth?............................................................................................................................. 164 Box 8.1: Concepts of risks, shocks and vulnerability.............................................................................................................................................................................. 179 Box 8.2: Measuring vulnerability using cross-sectional data............................................................................................................................................................ 181 Box 8.3: Measuring the prevalence of and responses to shocks in KIHBS data.................................................................................................................... 187 Box 8.4: Findings from impact evaluations of the OVC and the HSNP programs............................................................................................................... 198 Box 8.5: Evaluating the impacts of Kenya’s cash transfer programs using cross-sectional data and propensity score matching ...... 202 APPENDICES Appendix A: Chapter 1 additional materials ................................................................................................................................................................................................ 206 Appendix B: Chapter 2 additional materials ................................................................................................................................................................................................ 216 Appendix C: Chapter 3 additional materials ................................................................................................................................................................................................ 220 Appendix D: Chapter 4 additional materials ................................................................................................................................................................................................ 229 Appendix E: Chapter 5 additional materials ................................................................................................................................................................................................ 231 Appendix F: Chapter 6 additional materials ................................................................................................................................................................................................ 242 Appendix G: Chapter 7 additional materials ................................................................................................................................................................................................ 246 Appendix H: Chapter 8 additional materials ................................................................................................................................................................................................ 250 ACKNOWLEDGEMENT The World Bank greatly appreciates the close collaboration with the Government of Kenya (GoK), particularly the Kenya National Bureau of Statistics (KNBS) in the preparation of this report. This report was prepared by a team led by Utz Pape (Senior Economist) and Carolina Mejia-Mantilla (Economist)1, with the guidance of Johan Mistiaen (Program Leader) of the Africa Region in the Poverty and Equity Practice. The team consisted of Marina Tolchinsky (Chapter 1), Christine Achieng Awiti (Chapter 1), Nduati Maina Kariuki (Chapter 2), Isis Gaddis (Chapter 3), Habtamu Fuje (Chapter 4), Shohei Nakamura (Chapter 5), Simon Lange (Chapters 6 and 7) and Arden Finn (Chapter 8). The authors received substantive contributions from Haseeb Ali, Paolo Avner, Stephan Dietrich, Yuka Karasawa, Angelo Martelli, Saurabh Naithani, Stephen Okiya, Vera Sagalova, Aaraon Thegeya and the Tegemeo Institute of Agricultural Policy and Development (Egerton University). The report was prepared under the supervision of Diarietou Gaye (Country Director for Kenya, Rwanda, Uganda, and Eritrea) and Pierella Paci (Practice Manager). The peer reviewers were Markus Goldstein, Gabriel Demombynes and Prof. Michael Chege. The report benefitted from excellent comments from G N V Ramana, Ruth Karimi Charo, Frederick Masinde Wamalwa, Emma Mistiaen, Jishnu Das, Richard Chirchir, Mark Pancras and Evelyn Mwangi. 1 With equal contributions. ABBREVIATIONS Currency Equivalents (Exchange Rate Kenyan Shilling Effective as of Sept 28, 2018) US$1.00 = Ksh 100.956 AEZ Agro-Ecological Zone KNUT Kenya National Union of Teachers CBN Cost of Basic Needs LMIC Lower Middle-Income Class CDF Cumulative Density Function LPM Linear Probability Model CEC County Executive Committee LSMS Living Standards Measurement Study CPI Consumer Price Index MP Member of Parliament CRA Commission on Revenue Allocation NEDI North & Northeastern Development Initiative DHS Demographic and Health Survey NER Net Enrollment Rate DPT Diphtheria, Pertussis, and Tetanus NGO Nongovernmental Organization EAC East African Community NHIF National Hospital Insurance Fund ETP Extra Teacher Program NSBDP National School-Based Deworming Programme FAO Food and Agriculture Organization NSNP National Safety Net Programme FBO Faith-Based Organization OCOB Office of the Controller of Budget FPE Free Universal Primary Education ODM Orange Democratic Movement FTSE Free Tuition Secondary Education OPCT Older Persons Cash Transfer GER Gross Enrollment Ratio OVC Orphans and Vulnerable Children GIC Growth Incidence Curve PNU Party of National Unity GoK Government of Kenya PPA Participatory Poverty Assessment HDI Human Development Index PPP Public-Private Partnership HSNP Hunger Safety Net Program PSM Propensity Score Matching ICLS International Conference of Labor Statisticians RCT Randomized Control Trial IDS Institute for Development Studies SDI Service Delivery Indicators IEBC Independent Electoral and Boundaries Commission SPS Social Protection Secretariat ILO International Labour Organization STEM Science, Technology, Engineering and Mathematics IPC Infection Prevention and Control STI Sexually Transmitted Infections IPV Intimate Partner Violence TFP Total Factor Productivity ITN Insecticide-Treated Bed Net TFR Total Fertility Rate KCPE Kenya Certificate of Primary Education TSC Teacher Service Commission KCSE Kenya Certificate of Secondary Education TVET Technical and Vocational Education and Training KDHS Kenya Demographic and Health Survey UFS Urban Food Subsidy KES Kenya Economic Survey UHC Universal Health Coverage KHHEUS Kenya Household Health Expenditure and Utilisation Surveys UNDP United Nations Development Program KICD Kenya Institute of Curriculum Development UNESCO United Nations Educational, Scientific and Cultural Organization KIHBS Kenya Integrated Household Budget Surveys VIP Ventilated Improved Pit KNBS Kenya National Bureau of Statistics WDI World Development Indicators KNOCS Kenyan National Occupation Classification Standard WHO World Health Organization KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead i ii KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead EXECUTIVE SUMMARY KENYA MADE PROGRESS IN REDUCING Households in the bottom 40 percent of the POVERTY AND INEQUALITY OVER THE PAST distribution experienced notable consumption DECADE growth. The annualized consumption growth for T he substantive economic growth of the last decade has brought Kenya into the low middle- income country category in 2014. For the period of the bottom 40 percent, also known as the shared prosperity indicator, was 2.86 percent per year for the period from 2005/06 to 2015/16, and particularly high focus of this report, 2005/06 to 2015/16, growth in for the rural households. Even amongst the poor, those Kenya averaged 5.3 percent, higher than the 4.9 percent at the bottom of the distribution experienced higher observed for sub-Saharan Africa as a whole (Figure 1a). consumption growth: households in the bottom 20 Overall, growth was powered by the service sector, percent of the distribution experienced annualized which now accounts for almost half of the nation’s growth rates of around 3 to 4 percent. This trend was more GDP. The remarkable expansion of telecommunication marked in rural areas, which lead to a more pronounced and mobile-based financial services shifted the poverty reduction amongst rural households compared economic paradigm of Kenya to an extent rarely seen to their urban counterparts. However, compared to its in developing economies. Moreover, the country was regional peers, with the exception of Ethiopia, Kenya capable of bouncing back from the violent political has been less successful in boosting shared prosperity outbreak that followed the 2007 presidential election, (Figure 1d). from the effects of the 2008/09 global financial crisis, and from the harsh drought conditions experienced Consistent with this pro-poor pattern of economic by most of the African Horn in 2011, aggravated by the growth, inequality declined in Kenya. The Gini increase in the international price of oil. index, generally not affected by the upper tail of the distribution, fell from 0.45 in 2005/06 to 0.39 in Poverty incidence declined, benchmarked against 2015/16, indicating that Kenya made considerable both the national and the international poverty lines, progress in reducing inequality. Similar trends are but remains high relative to other lower middle- observed with the Theil index, the 75/25 ratio and the income countries. The proportion of the population Atkinson index. This places inequality in Kenya at a living beneath the national poverty line fell from moderate level when compared to other countries in 46.8 percent in 2005/06 to 36.1 percent in 2015/16, the region (Figure 1e). Now, while the fall in inequality showing a modest improvement in the living standards contributed to poverty reduction, most of the decline is of the Kenyan population, considering the ten year attributable to economic growth rather than a change gap (Figure 1b). Given the high dependence of the in the distribution of resources. This is consistent with agricultural sector on rainfall, the decline was higher in the low coverage of the social protection programs years of good weather and lower in years of drought. in the country, and the fact that fiscal policy as a Similarly, poverty under the international poverty line whole has little incidence on the level of poverty in of US$ 1.90 a day declined from 43.7 percent in 2005/06 Kenya, as shown by a recent fiscal incidence (World to 36.8 percent in 2015/16. At this level, poverty in Bank 2018b). This suggests that a more focused effort Kenya is below the sub-Saharan Africa average and on redistributive policies, such as social protection is amongst the lowest in the East African Community programs and equality of opportunities, can help (Figure 1c). However, it is approximately twice as high accelerate poverty reduction going forward. the average for its middle-income group. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead iii Executive Summary Figure 1: Kenya’s economic and poverty progress a) GDP growth from 2005 to 2016 b) Poverty headcount rate under the national poverty line 10.0 60 8.0 50.5 50 46.8 Proportion of the population 6.0 40 38.8 36.1 32.1 29.4 Percent 4.0 30 2.0 20 0.0 10 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 -2.0 0 National Rural Urban -4.0 GDP per capita growth GDP growth 2005/06 2015/16 c) Poverty headcount rate under the international poverty line d) Shared prosperity indicator 70 12 9.8 60 60 Poverty headcount (% of population) 8 consumption for bottom 40 percent Annualized % change in mean 49 50 41 4.6 4.3 40 37 4 3.5 35 2.9 30 0 20 -0.4 14 10 -4 Kenya Rwanda South Uganda Tanzania Ethiopia 0 (2006 - (2005 - Africa (2005 - (2007 - (2005 - Rwanda Tanzania SSA Kenya Uganda Ghana 2016) 2010) (2005 - 2012) 2011) 2010) 2013 2011 2013 2015 2012 2012 2010) e) Measuring inequality: Gini index f ) Distribution of the poor population across rural and urban areas 0.8 20 0.63 0.6 15 Number of poor (in millions) 0.50 0.42 0.41 0.39 Gini index 0.4 0.38 10 12.6 0.33 14.3 0.2 5 3.8 2.3 0.0 0 Kenya Ethiopia Ghana Rwanda South Tanzania Uganda 2005/06 2015/16 (2015/16) (2010) (2012) (2013) Africa (2011) (2012) (2011) Urban poor Rural poor Source: KNBS; own calculations based on KIHBS 2005/06 and KIHBS 2015/16 and World Bank open data catalogue. iv KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Executive Summary Most of the poverty decline is attributable to the and electricity services is much lower for poor progress observed in rural areas. Poverty declined households (Table 1). In this sense, Kenya should considerably in rural Kenya, from around 50 percent in continue to expand the coverage of this basic services 2005/06 to 38.8 percent ten years later, resulting in a to all segments of the population, while ensuring their decline in the number of rural poor from 14.3 million to quality at the same time. 12.6 million people. This contrasts with the stagnation of poverty in urban areas, where no clear decline in the Off-farm diversification played an important role in reducing poverty poverty headcount is observed as the 2.7 percentage points reduction is not statistically significant from zero. The evidence suggests that off-farm diversification Importantly, the number of urban poor increased during has been important for poverty reduction in this period, with cities concentrating a larger fraction of Kenya. Households whose agricultural income was the poor than they did before (Figure 1f ). This is partly supplemented by non-agricultural activities, mainly explained by the relative increase in food prices, which in small-scale services, account for slightly more than is known to affect the urban poor while benefitting rural a third of the poverty reduction (33.5 percent), the food net-producers. Housing costs have also increased highest share. This highlights the importance of off- in medium- and small-sized towns, reflecting that farm diversification in poverty reduction over the last urban growth has exacerbated a shortage in the supply ten years. While the agricultural sector has not been of affordable housing. It seems that cities, particularly as dynamic as the service or the industrial sector, it secondary cities, are not providing sufficient economic played a notable role in reducing poverty. Households opportunities for individuals to improve their income for which the main source of income is agriculture level and participate in the overall economic progress. (including both crop income, livestock income, and earning of wage workers in the agricultural sector) Poor households remain constrained by demographic account for 27.6 percent of the overall reduction characteristics, low human capital, and low coverage (Table 2). Finally, households engaged exclusively of basic services. Poverty incidence is higher for in non-agricultural activities, including services, households headed by women, the elderly and those manufacturing and construction, contributed with with low educational attainment levels. This suggests about 21 percentage points. that the poor are constrained when accessing income generating opportunities. Moreover, poor households While agriculture remains the main source of tend to be larger, and have higher dependency ratios; income for rural households in Kenya, the share of demographic factors that usually hinder poverty income from non-agricultural employment and non- reduction. In addition, coverage of water, sanitation agricultural employment has increased significantly Table 1: Access to basic services by poverty status 2005/06 2015/16 Significance Significance Non-Poor Poor Non-Poor Poor (wald-test) (wald-test) Access to services             Improved drinking water 70.2% 51.9% *** 80.4% 65.6% *** Improved sanitation 56.4% 37.7% *** 72.2% 47.8% *** Main source light (electricity) 23.0% 4.0% *** 49.9% 18.9% *** HH electricity access 26.5% 4.5% *** 52.0% 20.7% *** HH has garbage collected 10.7% 2.9% *** 21.7% 6.0% *** Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead v Executive Summary Table 2: Sectoral decomposition of poverty reduction (Ravallion-Huppi) Pop. share in Absolute Percentage Source of income period 1 change change Non-agricultural income only 31.64 -2.16 21.19 Agriculture income only 39.79 -2.81 27.63 Mixed - agricultural & non agricultural income 28.57 -3.41 33.51 Total intra-sectoral effect   -8.37 82.33 Population shift effect   -1.68 16.49 Interaction effect   -0.12 1.19 Change in headcount rate   -10.17 100.00 Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16. in the last decade. Income from crops and livestock income in 2005/06 to 21 percent in 2015/16. This as well as wages in the agricultural sector, declined diversification of income, in which households from 64.0 percent in 2005/06 to 57 percent in 2015/16 complement agricultural income with income derived (Figure 2a). Wage income from service employment is from non-agricultural activities (particularly in services the second most important source of income in rural and trading activities), along with an increase in the areas, increasing from 15 percent of rural household share of labor time allocated to non-agricultural Figure 2: Share of income by source for rural households a) Share of income by source for rural households b) Non-agricultural labor allocation in rural Kenya 60 Percent of labor time on non-agricultural activities 8 6 8 9 40 15 21 Percentage 4 7 20 64 57 0 2005/06 2015/16 Kenya Central Coast Eastern North Nyanza Rift Western Agriculture Industry wage Services wage Eastern Valley Transfers Enterprise income 2005 2015 c) GDP growth rates, by sectors d) Growth of mobile payments, by agents and transaction value 14 250,000 400 350 200,000 9 300 150,000 250 Percent 200 4 100,000 150 100 50,000 50 -1 0 0 Mar-07 Dec-07 Sep-08 Jun-09 Mar-10 Dec-10 Sep-11 Jun-12 Mar-13 Dec-13 Sep-14 Jun-15 Mar-16 Dec-16 Sep-17 Jun-18 -6 2007 2008 2009 2010 2011 2012 2013 2014 2015 Agriculture GDP Services Industry Agents (left) Value (KSh billions; right) Source: KNBS; own calculations based on KIHBS 2005/06 and KIHBS 2015/16 and World Bank open data catalogue. vi KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Executive Summary activities between 2005/06 and 2015/16 (Figure 2b), sectors of the economy.2 Similarly, many farmers have have been important to the reduction of poverty in rural shifted to bean production in recent years, as the Kenya. Thus, it is important to support rural households country benefited from favorable bean and maize in their effort to diversify their income. Investments in prices from 2011 to 2016. Farmers that shifted to bean human capital, skills formation, as well as attracting production are less likely to be classified as poor. High non-agricultural economic activities into rural areas, are commodity prices, like those observed between 2010 key areas of actions in which the Government of Kenya and 2016, is generally beneficial for Kenya’s net-selling should focus. farmers. However, this is at the expense of the urban poor, as poor urban households spend a large share of Those deriving income from non-agricultural their income on food and are therefore vulnerable to activities benefitted greatly from the expansion rising food prices. This factor may have contributed to of mobile money (thanks to M-PESA) throughout the divergence in poverty reduction between urban the country, particularly in remote previously and rural areas. uncovered areas. The penetration of mobile phones not only brought market efficiency gains associated Non-monetary wellbeing also improved, some issues still pending to be solved with a reduction in transaction costs (which likely also benefited those engaged in agriculture). Through The progress in the wellbeing of the population as mobile money, it also became a platform for service evaluated by monetary measures was accompanied delivery rather than just a communication tool, by the progress in several non-monetary dimensions changing Kenya’s economic paradigm as some have of poverty. Kenya’s Human Development Index pointed out (Jack and Suri, 2013, 2014; Suri 2017). (HDI), a combination of education, inequality, and life Mobile money, used by 18 million people in Kenya in expectancy indicators, gained 0.07 points in the decade 2017, increased the households’ financial resilience leading to 2015 reaching 0.55. This is the highest HDI in and savings, allowing them to: i) invest productively, ii) the East Africa Community, and a relatively high level move out of agriculture or complement that income given the county’s poverty headcount. with that of other businesses, and iii) improve their consumption levels, while also make risk sharing more Enrollment rates at all levels have increased, driven effectively. The way in which the expansion of mobile by higher enrolment of children from poor families. money transformed Kenya’s economy is an example for The government has invested substantial resources in other African countries, and the factors that enabled its recent years to increase enrollment rates, particularly expansion (including investment in infrastructure, the at the primary level with the introduction of universal regulatory environment and the participation of the primary education in 2003. As a result, primary private sector, among others), can provide important education is nearly universal with a net enrollment of lessons for low and low-middle income countries 85 percent in 2015/16, including 78.8 percent for the around the world. poor (Figure 3a). Enrollment in secondary education increased more gradually, and between 2005/06 and Other factors that likely benefitted Kenyan 2015/16 the net enrollment rate increased by more than households, particularly those in rural areas, are 20 percentage points reaching 42.2 percent (Figure 3b). the penetration of motorbikes (boda bodas), high Similarly, enrollment in tertiary education has increasing commodity prices and increased productivity in rapidly after 2009, and according to the 2015/16 KIHBS, the production of bean crops. Boda bodas helped to the gross enrollment rate is about 15.2 percent. lower the transaction costs of trading agricultural and non-agricultural goods as well as services, enhancing 2 Estimates suggest that in 2008 there were a total of 130,000 motor cycles registered in Kenya. By 2017, this number is likely to have the income rural households engaged in all different reached one million. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead vii Executive Summary Figure 3: Non-monetary dimensions of wellbeing a) Enrollment in primary education, 2015/16 b) Enrollment in secondary education, 2015/16 120 120 90 90 Percent Percent 60 60 30 30 0 0 Total Poor Non-poor Bottom 20% Bottom 40% Top 20% Rural Urban Total Poor Non-poor Bottom 20% Bottom 40% Top 20% Rural Urban By poverty By quintile By locality By poverty By quintile By locality Gross enrollment ratio Net enrollment rate Gross enrollment ratio Net enrollment rate c) Under-five mortality rate (per 1,000 live births) d) Multi-dimensional deprivations, 2015/16 180 80 150 120 60 % of households deprived 90 60 30 40 0 Total Bottom 20% Bottom 40% Top 20% Primary and lower Secondary and higher Rural Urban 20 0 Consumption Adult education Primary School Improved water Improved sanitation Access to electricity By quintile of wealth By mother's By location index highest level of educational attainment 2003 2008/09 2014 e) Average number of curative visits per person per year (total population) f) Distance to health facility where a doctor is available, in kilometers, 2015/16 6 40 5 30 4 3 20 2 10 1 0 0 Top 20% Non-poor Urban Poor Bottom 20% Bottom 40% Top 60% Total Urban Total Rural Rural Bottom 40% T60%, urban B40%, urban T60%, rural B40%, rural By poverty By quintile By locality By quintile By locality By quintile and locality 2005/06 2015/16 Mean Median Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16, WDI data. viii KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Executive Summary Kenyans have experienced significant gains in a among the richest quintile, compared to the poorest range of population health indicators in the last ten quintile, at almost half that level. Rural-urban disparities to fifteen years. Mortality among children below the are pronounced, with a 20 percentage point difference age of five has declined from 114.6 deaths per 1,000 in enrollment (Figure 3b). This is a reflection of low live births in 2003 to only 52 in 2014, a remarkable transition rates between primary and secondary achievement (Figure 3c). Moreover, the gains were school stemming from financial constraints and late widely shared, as the under-five mortality gap between enrollment in primary school. While learning outcomes poor and non-poor children declined. This progress of Kenyan children compare favorably to peer countries, is largely attributed to increased uptake of low-cost, the education system often fails to equip students with high-impact interventions (such as malaria nets) basic skills. Learning assessments suggest that Kenyan and declining fertility. Similarly, Kenya has also made children quickly fall behind the standards set by the substantial gains in reducing child stunting and it now national curriculum: only about half of the children in has one of the lowest stunting rates in the region. As fourth grade master the basic tasks that second-graders of 2014, nearly 1 out of every 4 children under the should be able to accomplish (e.g., read and understand age of 5 is stunted in Kenya, down since 2003, when a paragraph). Regional disparities in learning outcomes 35.6 percent of Kenyan children were stunted. In are pronounced and mirror those in enrollment. addition, improvements in uptake of both curative and Finally, while well-paid and knowledgeable by regional preventive services were also often more pronounced standards, Kenya’s teachers lack pedagogical skills and among the poor. are absent from class too often, suggesting that teacher incentives are not always aligned with student learning. However, Kenyan remain deprived in many of the dimensions. When looking at poverty as a Despite the progress, there are still pronounced multidimensional challenge, along the lines of the socioeconomic gradients in health access and components of the upcoming multi-dimensional some health outcomes warrant action. Children poverty index by the World Bank, households are often from poor families are less likely to be vaccinated deprived beyond the monetary dimension. The most and poor mothers are less likely to give birth with a common type of deprivation is access to services, qualified health provider present. In fact, in all domains notably sanitation and electricity: 40.7 percent of -outpatient care, inpatient care, and preventive care- households lack access to improved sanitation3 and and across almost all age groups, the poor are less likely 64 percent lack access to electricity in 2015/16. Fewer to use health services (Figure 3e exemplifies this point households, around 28.2 percent, are deprived of access by showing the average number of curative care visits to an improved source for drinking water (Figure 3d).4 per person per year). They also often have to overcome greater distances to access health care, particularly Regardless of the positive trends, geographic in rural areas. These gaps in access remain large and and socio-economic disparities in net secondary significant and are a major cause for concern (Figure enrollment remain a challenge and learning 3f ). In addition, maternal mortality ratio remains high assessments suggest that Kenyan children often lag at 510 deaths for every 100,000 live births, close to the behind the curriculum. Net enrollment in secondary average for low-income countries and only somewhat education at 56 percent remains substantially higher lower than the regional average. 3 Improved sanitation is defined as a toilet with a flush, a ventilated improved pit latrine or a latrine with a slab. 4 Improved drinking water sources are defined as a piped water system, public tap, borehole, protected dug well, bottled water or water from rainwater collection vendors. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead ix Executive Summary DESPITE PROGRESS IN REDUCING POVERTY, to healthcare and up-take rates, particularly in terms SEVERAL CHALLENGES REMAIN of children who are treated for illness, vaccination Progress has been slow rates and child-birth delivered by a skilled provider. For example, vaccination rates vary from more than H owever, progress is slow and Kenya is not on track to eradicate extreme poverty by 2030. Even though Kenya has experienced moderate GDP growth 90 percent in the Central region to about 44 percent in Mandera in the Northeast and only 36 percent in West Pokot, part of the NEDI counties. Limited access to in the last decade, transmission of growth into increased healthcare coupled with extremely high fertility rates, consumption of households is low. At 0.57, the country’s results in the highest maternal mortality rates of the elasticity of poverty reduction to economic growth – country. In addition, coverage of improved sanitation how much economic growth translates into poverty and electricity, and to a lesser extent, access to improved reduction – is low, below that of Tanzania, Ghana and water, is lower. While the government has implemented Uganda; and weaker than expected given its level of some measures to improve the connectivity and overall GDP per capita. To eradicate extreme poverty by 2030, wellbeing of the population in these areas, a substantive, an annual poverty reduction rate of 6.1 percent would sustained and cross-sectorial effort is required over the be necessary, despite the fact that in the last decade it medium term. has been 1.6 percent. If the trends observed in the last decade continue, the poverty rate will remain above Another source of spatial inequality is the growing 25 percent in 2030. To accelerate the pace of poverty inequality within cities, as the urban population in reduction, Kenya will require a far more inclusive Kenya increases over time. Within Nairobi, poverty is economic growth coupled with a sharper focus on highly concentrated in informal settlements, where the targeted poverty-reducing policies. living conditions are far worse, not only in comparison to the rest of the city but also in comparison to informal Stark spatial disparities remain settlements in other major African cities. Nearly a third Kenya is characterized by stark regional differences. of informal settlement residents in Nairobi are poor, The wellbeing of the population in the NEDI (North compared to 9 percent of the population living outside & Northeastern Development Initiative) group of informal settlement areas. Mean per capita monthly counties, which includes all counties in the North consumption of informal settlement residents (Ksh Eastern province, lags considerably behind the rest 10,377) is nearly 40 percent lower than that of non- of Kenya. In the NEDI counties, 68 percent of the informal settlement residents (Ksh 16,688), as shown population live in poverty compared to 36.1 percent in Figure 4f. Moreover, the living conditions in informal at the national level (Figure 4a). Moreover, these settlements, in terms of housing, access to services, counties saw little progress between 2005/06 and environmental problems, and health, are extremely 2015/16 and remain prone to food insecurity, as precarious. Informal settlement residents also live far shown by the food poverty and extreme poverty away from jobs, constraining their access to economic indicators (Figure 4b,c). Poor households of the NEDI opportunities. It also remains difficult to move out of a counties also lie far below the poverty line and the informal settlement, exacerbating the spatial poverty prevalence of vulnerability is highest in the counties trap in informal settlements. Mandera, Garissa, Samburu, and Turkana while rates are significantly lower in the central counties, Vulnerability is prevalent particularly in Nyeri, Kirinyaga and Nairobi. Although vulnerability and poverty rates fell over A sustained, multi-sectoral effort is required to raise the last decade, over half of Kenya’s population is the living standards of the population of these areas. currently vulnerable to falling into poverty in the near Educational enrollment rates are much lower for these future. Vulnerability rates5 fell faster in rural areas than counties, particularly in secondary education. In terms 5 Households are considered to be vulnerable if their predicted probability of being below the poverty line at any stage within the next two years is of health services, they present lower rates of access greater than 50 percent. x KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Executive Summary Figure 4: Regional patterns in poverty a) Absolute poverty b) Food poverty 74.0 76.2 80 65 68 80 70.0 68.0 57.6 53 51 55 Proportion of the population 51 Proportion of the population 60 54.2 60 50.6 45 48 45 49.0 47.2 44.2 39 42 40.5 41.4 42.5 38 36.7 32 36 33 40 31.8 31.1 40 32.6 28 30 21.3 20 24.3 22 16.7 16 20 20 0 0 h h rn rn t t DI DI DI DI za za lle ift lle ift l l bi bi rn rn ste rt ra ra as as Ea rn ste rt Ea rn W y W y Ea No te te iro iro Va R Va R NE NE NE NE an an ste ste Ea No nt nt Co Co es es Ny Ny Ce Na Ce Na n- n- No No 2005/06 2015/16 2005/06 2015/16 c) Extreme poverty d) Percentage of children age 12-23 months that received all basic vaccinations 80 60 Proportion of the population 47 50 40 32 27 24 23 23 20 20 20 17 12 14 11 6 7 6 6 3 3 0.6 0 (90,100] (80,90] r h lle ift t rn DI DI za l rn bi ra as ste rt (70,80] Ea n W y Va R ste NE NE an te nt iro Ea No Co (60,70] es Ny Ce n- (50,60] Na (40,50] No (30,40] [20,30] 2005/06 2015/16 e) Access to improved sanitation f ) Household consumption in Nairobi’s informal settlement and non- informal settlement areas, 2015/16 100 95 Consumption .00015 71 78 72 71 Proportion of households, % 75 66 66 68 61 51 58 .0001 Kernel density 49 44 50 51 50 43 41 33 33 36 .00005 19 20 25 0 0 0 10000 20000 30000 40000 50000 l rn t rn l y rn Na a No bi DI DI na ra as lle z iro ste ste te an NE NE nt Co tio Va es Ce Ny n- Ea Ea Per adult-equivalent monthly consumption Na W ft Ri rth No 2005/06 2015/16 Informal settlement Non-informal settlement KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead xi Executive Summary they did in urban areas between 2005/06 and 2015/16, poor, non-beneficiary households. The programs are but the current urban-rural differences are still very effective in fostering food security, improving school large – 43 percent in urban areas, and 57 percent in rural enrolment and reducing the probability of children areas (Figure 5a). Poverty and vulnerability are highly working. Despite recent efforts by the government, correlated, but over one third of non-poor Kenyans are these programs have limited geographical coverage classified as vulnerable. Vulnerability rates vary widely and remain small in scale (Figure 5f ). by county, being highest in the north and east of the country (Figure 5b), and by household characteristics, WOMEN ARE LEFT BEHIND IN MANY AREAS K with high vulnerabilities particularly among those that enyan women are disproportionately affected are engaged primarily in agriculture, and those with by poverty during the core productive and low educational attainment (Figure 5c). Many of these reproductive years, especially if they experienced non-poor but vulnerable households are clustered just a marital dissolution. As in other African countries, above the poverty line, meaning that even a moderate Kenyan women are more likely to live in poor shock could push them below the line. households than men, starting in their mid-20s and continuing until their 50s (Figure 6a). Moreover, women When faced by shocks, many poor and rural who are separated, divorced or widowed are more likely households often resort to coping strategies with to be poor (compared to men), face higher prevalence adverse implications for future wellbeing. The rates of physical violence (compared to other women) overall prevalence of both economic and agricultural and are disproportionately affected by HIV/AIDS. Kenya shocks declined between 2005/06 and 2015/16. is also among the few African countries with gender However, the incidences of certain kinds of shocks inequality in formal inheritance rights – i.e. the Law affecting agricultural households went up. Agricultural of Succession Act. Gender gaps exist also in terms of households were far more likely to report crop losses access to ICT and financial services, though levels of from preventable causes such as crop diseases or access are high by regional standards. pests in 2015/16 than they were in 2005/06 (Figure 5d). The most common response of poor households Girls and women continue to be disadvantaged in after experiencing a shock is to reduce consumption, education and health in some regions. Girls have while for the richest households the most common lower enrollment rates and educational outcomes than response is to use savings (Figure 5e). The inability of boys in Northeastern Kenya and the coast – but boys’ poor households to cope with adverse shocks and disadvantages emerge in parts of Central and Western their limited financial resilience has severe long-term Kenya (Figure 6b). Girls dropping out of secondary school implications, particularly when they are forced to cut are more likely to be married and to have given birth spending on food, education and health, curbing than girls still attending school.6 Despite improvements human capital accumulation. in girls’ education, adult women are twice as likely to be illiterate as adult men (Figure 6c), reflecting historical Kenya expanded its social protection programs, but gender inequalities, which continue to put women at coverage and scale remain limited. Over the last few a disadvantage in terms of labor market opportunities. years, Kenya expanded its social protection programs, Even though maternal mortality declined since 2005, spending about 0.27 percent of GDP in 2015, well Kenyan women face a staggering lifetime risk of 1:42 of below the average of 1.6 percent of GDP in low- and dying due to complications of pregnancy or child birth middle-income countries. The programs are generally (Figure 6d). well targeted: only 23 percent of grant-receiving households had at least one resident member who 6 Secondary drop out is defined as having attended secondary school Form 1-3 during the last school year, but no longer attending school was employed. This is in contrast to 48 percent in poor, during the current school year. Note that there are only few cases of secondary drop outs captured by the KIHBS N=70), which limits the non-beneficiary households, and 54 percent in non- analysis that can be performed. xii KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Executive Summary Figure 5: Poverty and vulnerability in Kenya a) Poverty and vulnerability from 2005/06 to 2015/16 b) Geographic variation of vulnerability in 2015/16 80 70 60 Percentage of population 50 40 30 20 10 0 Poor Vulnerable Poor Vulnerable Poor Vulnerable Vulnerability rate >90% 80.1% to 90% 75.1% to 80% 70.1% to 75% 65.1% to 70% 60.1% to 65% Total Urban Rural 50.1% to 60% 35.1% to 50% 2005/06 2015/16 10% to 35% c) Relative vulnerability rates, by household characteristics d) Shock prevalence for agricultural households only 90 40 80 35 70 30 60 25 Percentage Percentage 50 20 40 15 30 10 20 5 10 0 Agriculture Manufacturing Services Construction Rural Urban No education Primary education Secondary education Tertiary education Female Male Drought or ood Crop disease/pest Livestock death/theft Severe water shortage HH business failure Loss of salaried employment End of regular assistance/aid Large food price rise Large agri. input price rise Dwelling damaged/destroyed Household Household head Agricultural shocks Economic shocks National vulnerability rate 2005/06 2015/16 e) Coping mechanisms in 2015/16, by poverty quintile f ) Number of households receiving cash transfers Used savings 800 43% 600 29% Reduced consumption Thousands 26% 25% 400 Help from family 17% 14% Borrowed 11% 200 9% 8% Sold assets 9% Help institution 11% 0 2013 2014 2015 2016 Quintile 1 Quintile 5 OVC OPCT HSNP PWSD Total Source: Own calculations based on KIHBS 2005/06, KIHBS 2015/16 and Kenya’s Single Registry for Social Protection. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead xiii Executive Summary In the Northeast, women often have a lower Reducing the gender gap can unleash Kenya’s participation in the labor market because of productive potential household work. In 2015/16, Kenya had a female Women’s productivity can be increased by abolishing labor force participation rate of 71 percent for the discriminatory practices in women’s access to core working-age population (15-64 years), compared productive assets. Gender biased legislations, such as with a male labor force participation rate of 77 percent the differential treatment of male and female surviving (Figure 6e). However, there are significant regional spouses under the Law of Succession Act, should differences – female labor force participation is high be eliminated. Savings products with an element of in Central and Western Kenya, but much lower in the illiquidity and soft commitment can increase women’s Northeast (Figure 6f ). Due to traditional gender roles, savings to unlock investments into productive assets. women spent a significant amount of time on unpaid Information campaigns as well as mentoring programs care work within the household. Every child aged 0-5 can help to overcome sectoral segregation locking years reduces women’s probability to be in the labor women into low-productivity jobs. Technological force by over 2 percent. change has the potential to disrupt traditional patterns of sectoral segregation, such as Uber and other ride- There are gender gaps in access to productive hailing services opening up opportunities for women in resources and sectoral segregation. In line with the traditionally male-dominated sectors like transportation. international experience, male wage workers earn 30 percent higher wages and salaries than female Closing the gender gap and creating equal wage workers. This is likely explained by the fact opportunities for boys and girls requires, among that women are disproportionately employed in other interventions, targeted investments in agriculture and services, while men have a higher education and health. Programs subsidizing the share of employment in the industrial sector. Also, direct or indirect cost of education can be effective in profits of male-run household enterprises are about increasing enrollments and educational performance of twice as high as profits of female-run enterprises boys and girls. Increased secondary school enrollment and households for which women are the primary among adolescent girls may also delay fertility decisions. decision-makers in agricultural activities achieve In health, further initiatives to increase access to and lower yields (maize, beans) than other households affordability of reproductive health care services are (Table 3). Only 12 percent of women aged 20-49 years important to reduce maternal mortality, especially in report owning any land on their own, compared with Kenya’s arid and semi-arid regions. Public investments in 39 percent of men. Also, Kenya is among the few services for care can reduce time constraints of women. African countries with gender inequality in formal Scaling up care services for children, however, requires inheritance rights, for example with respect to the innovative approaches, combining public and private Law of Succession Act. sources of funding. Table 3: Monthly earnings in KSh, by gender Male Female Ratio male-to-female Mean 18,276 14,075 1.30 10th percentile 3,000 2,000 1.50 Median 10,000 6,500 1.54 90 percentile th 43,300 35,000 1.24 Source: KIHBS 2015/16. xiv KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Executive Summary Figure 6. Gender gaps in Kenya a) Poverty rates and gender-poverty gap b) Female and male poverty rates by marital status, 2015/6 60 50 50 40 40 Poverty rate 30 Percent 30 20 20 10 10 0 0 Monogamously Polygamously Separated or Widowed or Never 0-4 5-9s 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75+ married or married divorced widower married cohabitating Age Poverty, male Poverty, female Males Females c) Gender parity index for gross primary enrolment rates6 d) Gender parity index for gross secondary enrolment rates7 (1.4,1.5] (1.3,1.4] (1.2,1.3] (1.2,1.3] (1.1,1.2] (1.1,1.2] (1,1.1] (1,1.1] (.9,1] (.9,1] (.8,.9] (.8,.9] (.7,.8] [.7,.8] (.6,.7] [.5,.6] e) Literacy rates, by gender and county f ) Maternal mortality ratio. 1 .8 .6 .4 .2 0 Nairobi Nyeri Mombasa Kiambu Kisumu Nandi Nakuru Machakos Trans Nzoia Kisii Nyandarua Vihiga Uasin Gishu Bungoma Kirinyaga Taita Taveta Siaya Muranga Embu Nyamira Kericho Homa Bay Migori Elgeyo Marakwet Bomet Baringo Kajiado Makueni Lamu Tharaka Nithi Kakamega Kitui Busia Meru Laikipia Tana River Isiolo West Pokot Kwale Samburu Mandera Garissa Wajir Marsabit Turkana Narok Kili (3000,4000] (2000,3000] (1000,2000] (800,1000] (600,800] (400,600] (200,400] [0,200] Male Female Based on 2009 Census. 7 The gender parity index is defined as the ratio of female to male enrollment rates. A value above (below) unity indicates that girls have higher (lower) levels of enrollments. 8 Ibid. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead xv Executive Summary g) Changes in employment, by gender h) Gender gap in labor force participation 80 60 Percent 40 20 0 (40,50] (30,40] (20,30] Male Female Male Female (10,20] (0,10] 2005/06 2015/16 (-10,0] (-20,-10] (-30,-20] (-40,-30] Wage Enterprise Any employment [-50,-40] Source: Own calculations based on KIHBS 2005/06, KIHBS 2015/16, KNBS (2012) and Global Findex 2014. ACCELERATING POVERTY REDUCTION Policies aimed at increasing the adoption of improved agricultural inputs by small farm holders would help Improve the productivity of the agricultural sector and to increase their income and help to further reduce enhance access to markets in rural areas poverty. Extension services programs and educational I ncreasing agricultural productivity remains a potential pathway out of poverty for many households. In Kenya, more productive farmers are less campaigns, together with a competitive inputs markets, are some alternatives. likely to be poor (Figure 7a,b). This correlation between Similarly, agricultural commercialization is also farm productivity and poverty constitutes promising associated with better living conditions. For farmers, evidence that an enhancing agricultural yields a higher degree of commercialization is associated with could lead to a reduction of povertys. However, little higher living standards, as can be observed in Figure progress has been made in terms of raising agricultural 7e. Thus, investments in infrastructure and access to productivity in the last ten years. This is especially true for the production of maize, Kenya’s main food information and communication technologies, so that staple, and commercial crops such as coffee. Increased farmers can more easily reach their clients and can more efficiency in the production of beans appears to be the easily buy the inputs for agricultural production, are an only exception. As a result, agricultural productivity has important policy area to focus in order to accelerate the not been contributing to poverty reduction in rural reduction in poverty. Kenya, a marked difference from the experience of other countries in the region, such as Ethiopia. Policymakers may need to allocate more resources to enhance farmers’ productivity and make sure Technology adoption is the main factor associated that the current spending is efficient and providing with higher productivity, according to analysis using the highest returns. Around 2 percent of total public farm level data. Farmers that applied chemical fertilizer, expenditure was allocated to agriculture in 2016/17, for example, experienced a 20-25 percent increase in even though the sector accounts for 25 percent and maize yield. Moreover, farmers who planted improved 60 percent of the country’s GDP and employment, maize seeds experienced 26-32 percent higher respectively (World Bank 2018). This prevents the productivity compared to those that used traditional country from investing effectively in smallholder low-yield seeds. Despite the yield-enhancing effects of agriculture and provide services to improve basic fertilizer and seeds, the share of farmers adopting these crop yield. There is also a need to asses if the current inputs has not changed much between 2000 and 2010. spending is efficient, taking into account that spending xvi KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Executive Summary Figure 7: Socio-Economic indicators of Rural Kenya a. Maize yield and poverty rate, 2015/16 b) Bean yield and poverty rate, 2015/16 2500 600 Eastern Rift Valley 2000 Rift Valley 500 Yield (kg/hectare) Yield (kg/hectare) Western 1500 Central Central Eastern Nyanza Nyanza 400 1000 Western North Eastern Coast North Coast Eastern 500 300 15 35 55 75 15 25 35 45 55 Poverty rate , % Poverty rate, % c) Poverty and the sale of farm produce in rural Kenya 80 60 Poverty Rate, % 40 20 0 0 0.5 1 Proportion of harvest sold 2005/06 2015/16 Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16, WDI data. on public goods in this context (e.g. research and More and better jobs, along with infrastructure development, extension services, etc.) has been investment, is required in urban areas proven to be more productive than spending on Many workers remain in volatile and low quality jobs private goods (e.g. fertilizer subsidies). In addition in urban areas, despite a decline in unemployment. there is space to reform the input subsidy program by Unemployment rates dramatically dropped in urban ensuring that the program is targeting small farmers areas (Figure 8e), in tandem with an increase in labor and facilitating technology adoption among them. force participation rates. However, a large fraction of the Moreover, investment in irrigation schemes have a urban poor, women, and youth are unemployed.10 The high rate of return9 and could reduce dependence on existing jobs in urban areas are casual and do not offer rainfall. The fact that food security is one of the Big Four long-term security. Nearly 90 percent of construction priority areas outlined by the government (together jobs in Nairobi are casual work, resulting in 41 percent with manufacturing, affordable housing and universal of the poor being casual workers as opposed to 9 healthcare) is a positive sign and the concrete policies percent for the non-poor (Figure 8d). These jobs do not that will be proposed should be scrutinized carefully. provide long term security, and may not conduce to better job opportunities in the future. 10 In Nairobi, for example, more than 20 percent of the poor are 9 World Bank, 2018. unemployed. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead xvii Executive Summary Figure 8: Urbanization remains a challenge for poverty reduction a) Economic sector of urban workers b) Urban unemployment rates in 2005/06 and 2015/16. 100 50 45 45 90 40 80 35 70 30 28 27 27 60 62 24 69 25 22 78 75 73 20 50 80 19 19 19 20 16 15 16 40 15 11 13 13 12 10 10 30 15 10 9 8 8 8 7 7 7 8 9 10 5 20 4 5 9 9 0 10 17 10 10 Non - Poor Non - Poor Non - Poor Non - Poor Poor Poor Poor Poor 15 All All All All 6 8 6 8 0 Poor Non-Poor All Poor Non-Poor All 2005/06 2015/16 All urban Nairobi Mombasa Other urban Agriculture Manufacturing Construction Other Services 2005/06 2015/16 Source: Own calculations based on KIHBS 2005/06, KIHBS 2015/16, DHS 2014 and Cities Baseline Survey 2013. It is important to leverage the potential of main form of motorized transport in Nairobi, a worker urbanization for poverty reduction through more can reach only 4 percent (within 30 minutes) and 25 and better quality jobs. Manufacturing and high value- percent (within 60 minutes) of existing jobs, while added services jobs are still lacking in urban Kenya in Greater Dakar, for example, it allows access to 52 despite the fact that they can play an important role percent of existing jobs within 1 hour of travel. Thus, in providing economic opportunities, especially for the investments to lower the transportation costs and young urban population. The urban poor face several shortening the distances between the individuals and challenges in terms of job availability and accessibility, the economic opportunities is necessary. Moreover, in particularly in informal settlement areas. More and some areas investments in physical infrastructure lag good quality jobs in the manufacturing sector can help behind the needs of the urban population. While the to improve the incomes of the urban poor, if paired with share of urban population with access to improved investments in transport infrastructure and skills. Some sanitation facilities and electricity increased during of the areas of focus should be competitiveness and the last decade, the share of those with improved capabilities. Industrial enclaves can help address some water access dropped in some areas, indicating that of the structural bottlenecks that affect manufacturing urbanization outpaced infrastructure provision. competitiveness and help attract foreign direct investment. Worker capabilities can be enhanced by Broader affordable housing can reduce housing costs prioritizing literacy, numeracy and ICT skills and by in urban areas, relaxing the budget constraints. The improving the training programs in collaboration with high costs in terms of food and housing are curbing the the private sector. As one of the Big 4, manufacturing purchasing power of the less well off. Targeted policies has great potential to help improve the livelihoods of to ensure affordable housing can help them to escape the urban poor. poverty, which will be hopefully part of the set of policies implemented under the Big 4 umbrella. In the Improved connectivity through investments in case of informal settlements, localized interventions are infrastructure and the provision of high quality required to ensure that informal settlements function public services is also crucial. High transportation as a place of opportunity rather than as a poverty trap, costs squeeze the budget of urban households, limiting including better service provision. access to economic opportunities. Using minibus, the xviii KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Executive Summary Improving the provision of education and health attendance and, eventually, student’s outcomes. Along Increasing secondary school enrollment among with greater local oversight, schools could be given the poor requires demand-side interventions. While more resources and greater independence on how to enrollment in secondary has increased among the use them. Increasing the capitation grant, along with poor, significant gaps persist. The evidence presented greater autonomy to school committees to recruit, in this report and numerous academic studies suggest retain, and promote teachers, has the potential to that increasing enrollment in secondary education in improve teacher performance and to lower school Kenya requires primarily demand-side interventions drop-out. The potential of a greater involvement of aimed at loosening the financial constraints that less private providers could also be explored. well-off households face. Cash transfers have already proven effective in increasing enrollment rates. In terms of the provision of universal health coverage Similarly, encouraging on-time primary enrollment and (UHC) one of the ‘Big Four’-priorities, it must be the supporting the primary-to-secondary transition noted that the incidence of catastrophic health could also contribute to raise the enrollment rates at expenditures has decreased recently, which may the secondary level. disincentivize voluntary enrollment going forward. Only around 20 percent of the population are covered Enhancing the quality of education and aligning the by health insurance, with large differences between teachers’ incentives with student learning requires a the poor and the better-off and between rural and series of interventions combined with a close scrutiny urban areas. At the same time, there is evidence that of the recently introduced monitoring and evaluation the incidence of catastrophic health expenditures system. Greater reliance on contract teachers to initially has declined over time and that households rarely fill vacant positions, subsequently moving to an ‘up-or- resort to adverse coping strategies, such as selling out’ promotion system, in which the best-performing their assets, to finance healthcare. This is in line with contract teachers are promoted to public school the removal of user fees in 2013 for a range of public teachers, may have large potential benefits. Contract health services, including birth deliveries, and with the teachers have average levels of subject and pedagogical overall improvement of living standards and health knowledge and lower rates of absenteeism, without amongst Kenyans. The implication maybe that those in being paid a premium. In any case, the system the informal sector have little incentive to voluntarily requires all teachers to be systematically and regularly insure, making it harder for the government to expand evaluated, for benefits to be tied to performance, and a health insurance coverage. credible threat of discontinuation of employment. The effectiveness of the recently introduced monitoring Similarly, given that the poor are more likely to and evaluation systems should be closely monitored. depend on public health services than the rich, While they have the potential to improve teacher effort, recent disruptions in supply during labor disputes it is not clear whether head masters and deputy head between the government and public-sector unions masters are best placed to monitor teacher presence disproportionately affect the less well-off. The string and performance. of recent health worker strikes in the public sector that culminated in major walk-outs at the end of 2016 and The quality of education would benefit from the in mid-2017, resulted in disruptions that likely affect involvement of local stakeholders, particularly the poor disproportionately. Health workers’ salaries in parents, and from enhanced school governance. Kenya remain high by regional standards, despite their Empirical evidence suggests that the local knowledge recent sluggish growth in real terms. This is particularly of stakeholders, particularly parents, may play a key true for health workers in the public sector, which earn role in monitoring teachers at the school-level. Getting a substantial premium, in part because of a lengthy list local stakeholders involved may help improve teacher of allowances that account for a significant portion of KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead xix Executive Summary their total pay. There should be a more open, informed Devolution has the potential to address some of the and transparent debate on adequate remuneration of development challenges, and its implementation public health workers, in order to help prevent these should be carefully monitored so that the necessary disruptions in the future. adjustments can be incorporated. Devolution has the potential to address the wide spatial variation in Finally, the sustainability of health financing, wellbeing across counties and regions and improve particularly of priority programs, should be a the accountability in service delivery. Decentralization priority. A recent report has highlighted funding gaps seems like the right path to address these inequities, in all five priority programs analyzed (World Bank but counties have various degrees of institutional 2018) and despite a falling share, healthcare financing capacity and economic development and must be in Kenya still relies significantly on donors. One provided with the resources required (both human and vehicle to increase revenues is through an increase financial). At the same time, outcomes in all sectors of memberships and contributions to the National should be closely monitored and counties should be Hospital Insurance Fund (NHIF). One alternative to hold accountable for their performance. In the coming increase funding would be by introducing ‘health years, as more data becomes available and enough taxes’ on food and drinks that contain high amounts time has passed for the effects of decentralization to be of saturated fat, sugar, salt, or other unhealthy apparent, more studies and research should focus on ingredients, which would also address the problem of the effects of devolution. rising obesity among urban, better-off Kenyans. The decade-long gap between the two most recent Expand Social Protection programs and provide the household consumption surveys makes it difficult to foundations for devolution to work monitor poverty and analyze the impact of policies. Expanding assistance to vulnerable households While Kenya’s most recent household consumption through existing or new social protection programs survey was implemented in 2015/16, the previous can reduce vulnerability. The effort that has been survey dates back to 2005/06. Without more regular made to coordinate and harmonize social protection data collection, it is very difficult to monitor progress programs, combined with the creation of a registry of in terms of poverty reduction, and to assess the impact beneficiary households means that the country is well of policies and programs. An improved monitoring placed to expand assistance to vulnerable households, system should be put in place, ideally one that provides which would benefit greatly from this potential information at the county level and that can inform the expansion. Furthermore, specialized programs to ongoing devolution process in Kenya. The Government mitigate shocks can reduce vulnerabilities. For example, of Kenya’s plans to establish a continuous household the introduction of emergency cash programs can have survey by the KNBS are a good step in the right direction the potential to offset some of the negative effects to design and implement policies based on evidence. of shocks such as droughts and floods, and protect vulnerable households from resorting to negative scoping strategies with long-term impacts like selling productive assets. xx KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Executive Summary ORGANIZATION OF THE REPORT Chapter 4 analyzes rural livelihoods, and explores T his report is organized as follows. Chapter 1 various factors that might have contributed or provides an overview of macroeconomic drivers of hindered the reduction in rural poverty. More economic growth and its fiscal implications. The trends specifically, it examines the role of diversification into of poverty (under the international poverty line) are non-farm employment to differentiate the contributions compared with other countries alongside indicators of on income-diversification and agricultural income. An non-monetary deprivations to provide an international analysis of rural-urban migration sheds light on the role benchmarking. Kenya’s context is discussed in an analysis of migration for poverty reduction. In the second part, of the political economy, with a focus on the two central the chapter delves into agricultural production and themes of political competition and devolution. This is productivity by analyzing its trends and its potential complemented by an analysis of Kenyans’ perceptions, impact on poverty reduction. The analysis concludes embedded in an international comparison. with a discussion of commodity prices and how they affected rural poverty. Chapter 2 first documents the progress made by Kenya in terms of the monetary measures of poverty, The linkages between urbanization and poverty during the period of focus of this report, 2005/16 to with a particular focus on the challenges faced by 2015/16. It analyzes the trends in terms of the national the urban poor are examined in Chapter 5. It reviews poverty headcount rate, other related indicators (such Kenya’s urbanization trends and examines how the as the depth and severity of poverty) and the incidence geographic patterns of poverty changed during the of food and extreme poverty, as officially defined by the last decade. In so doing, it assesses the contribution Kenya National Bureau of Statistics (KNBS). The chapter of urbanization to poverty reduction in the country. It then turns to examine the incidence of consumption also assesses urban poverty from both a monetary and growth, and how this is reflected in terms of an array a non-monetary perspective, in view of its geographic of inequality indicators. It also examines the factors heterogeneity. Thirdly, the chapter analyzes urban labor behind Kenya’s success in reducing poverty, relying markets to figure out opportunities and challenges on decomposition analysis and the finding of various faced by the urban poor. Finally, it takes a closer look studied on the impact of mobile money in the at informal settlements—mainly in Nairobi—, where wellbeing of the population. The chapter concludes urban poverty is concentrated, showing a stark contrast by providing a profile of the poor, in an attempt to in living conditions between informal settlement identify the factors that may be limiting their economic and non-informal settlement areas and the limited opportunities and overall wellbeing. residential movements between them. A synthesis of what is known about the gender- Recent developments in Kenya’s education sector poverty nexus in Kenya is presented in Chapter 3. and their relationship to poverty and equity are It starts with a basic profile of poverty and gender. analyzed in Chapter 6. It takes stock of the recent Next, following the framework of the 2012 World trends in access to education services as well as Development Report on Gender (World Bank 2011) it their quality and examines the incentives in place to then proceeds to analyze gender gaps in endowments, produce quality education for all. The chapter provides gender inequality in economic opportunities and background information on Kenya’s education system, gender differences in voice and agency. Within each while analyzing current levels and recent trends in of these sections, the chapter also provides a brief access and enrollment and their links to poverty discussion of possible policy options to narrow – and and equity. It then shifts the focus from access and ultimately close – gender gaps and promote a more enrollment to learning outcomes and then analyzes equitable society. inputs into the educational production and their distribution, including physical inputs and the ability of KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead xxi Executive Summary teachers to deliver quality education. Finally, it discusses in access and uptake of health services that persist school governance, especially teacher incentives. today, including significant geo-spatial variation. In addition to the analysis of various data sources11 Finally, it shifts the focus towards providers and inputs including administrative data, household surveys, and into health production, including provider knowledge school assessments, the analysis draws heavily on and physical inputs. The analysis relies on a wide array recent academic studies. of microdata sets and administrative data as well as a review of academic studies. Chapter 7 analyzes levels and trends in health outcomes, uptake of services, and health equity. With the aim of understanding how to address It provides background information on recent vulnerability in Kenya and make sure that the developments and initiatives in Kenya’s health sector, country enters a sustainable path of poverty including the devolution of health service delivery to reduction, Chapter 8 examines and analyze changes the counties, the removal of user fees, health workers in the vulnerability profiles for Kenya in 2005/06 and strikes, and universal health coverage (UHC), one in 2015/16. Moreover, it analyzes and compares the component of the ‘Big Four’-agenda. It documents welfare shocks that affected households in 2005/06 the rapid pace at which Kenya in recent years made and 2015/16, as well as which coping strategies were progress in health outcomes, particularly under-five adopted in the face of these shocks. Finally, the chapter mortality, and in the uptake of certain health goods assess the coverage and effectiveness of Kenya’s social and services. While these improvements have often safety net programs, while also measuring their impact been equitable, the chapter also documents inequities on different measures of household welfare. 11 The report relies on the 2005/06 and 2015/16 Kenya Integrated Household Budget Surveys (KIHBS), the 2012 Service Delivery Indicators (SDI) for education, a facility-based survey of teachers, students, and schools, and the Uwezo data, annual learning assessments. In addition, data from the World Development Indicators (WDI) and the Kenya Economic Survey (KES) was used. xxii KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead CHAPTER I KENYA IN CONTEXT SUMMARY Since 2005, Kenya has experienced resilient economic growth despite several shocks, contributing to a steady, though moderate, reduction in poverty. Economic and political shocks in the past decade have included electoral violence, drought, and an overhaul of the centralized political system. Perceptions of democracy and trust in the government have suffered over the years, following contested elections and corruption concerns. As Kenya begins its next five-year development strategy, a larger emphasis on redistributive policies and the devolution process is necessary to bring the country closer to eradicating poverty by 2030. Kenya’s economic growth has exceeded average growth in sub-Saharan Africa in the past decade. Growth averaged 5.3 percent in the 2005 to 2015 period, primarily driven by the services sector on the supply-side and household consumption on the demand-side. In particular, the mobile phone revolution contributed to an expansion of the financial services sector by increasing access to credit and providing services to previously unbanked households. The country faced two major economic shocks between 2005 and 2015. The first was due to electoral violence in early 2008 that compounded the effects of the global financial crisis. The government’s quick policy actions through a stimulus package helped restore growth in 2009 and 2010. A second shock hit the country in 2011 after an increase in international oil prices, combined with a drought in the Horn of Africa that reduced agricultural production in the region. The government’s development policy has been guided by Vision 2030, Kenya’s long-term development plan. Policies were designed to increase aggregate demand, with a focus on supply-side investments in infrastructure projects such as rail transport and renewable energy. While revenue was volatile between 2005 and 2015, the pace of public spending steadily increased and consistently exceeded revenue collections. This put pressure on the fiscal deficit, which increased from 4.7 percent of GDP in 2005/06 to 8.2 percent in 2015/16. Education spending was the largest beneficiary of social sector spending and had a stable upward pace, largely the result of a free universal primary education (FPE) policy. In the past decade, Kenya has experienced a moderate reduction in poverty. As of 2015, about one third of the Kenyan population lives below the international poverty line of US$ 1.90 a day. Poverty declined from 43.6 percent in 2005 to 35.6 percent in 2015. Poverty reduction has been driven by improvements among the poorest of the poor, and particularly among households engaged in agriculture. Agricultural households remain vulnerable to climate and price shocks, as growth in the sector has a strong impact on household consumption. Kenya compares favorably in monetary and non-monetary poverty with peer countries, but is not yet on the same level as other lower-middle income countries. At the lower-middle income line of US$ 3.20 a day, both the rate of poverty and the depth of poverty are worse in Kenya than in countries having similar levels of wealth per capita. More than two thirds of the Kenyan population lives below the US$ 3.20 a day line. Poor households are often deprived on multiple dimensions, with the most common being access to services such as improved water and sanitation. Kenya lags behind peer countries in access to improved water sources, but performs fairly well on education and health indicators. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 1 Kenya in Context Politics and political institutions in Kenya were until very recently influenced by centralized power residing in the presidency and executive branch. A strong political consensus emerged from the need to devolve powers away from the executive and the central government, with a view to making Kenya’s democracy and development more inclusive. Political and civil society efforts culminated in a constitutional referendum in 2010, which led to a “big bang” political and fiscal decentralization that devolved power to 47 counties created from the former eight provinces. Crucial issues however must still be resolved in order for devolution to have its full impact and for the citizenry to trust the process. So far, inherent disparities between counties determine developmental outcomes even if fiscal allocations are equitable. The disputed presidential elections in 2017 renewed the focus on democratic institutions in Kenya and perhaps on the current shortcomings of devolution as implemented. The Afrobarometer Survey captures key perceptions of Kenyans on democracy, the nature of governance, and on participatory politics. Support for democratic norms and processes remained high even through the political volatility and electoral disputes over the past decade – concluding with the 2017 elections. Perceptions in 2008 however did reflect disillusionment regarding the true extent of democracy. Responses in 2016 show low levels of trust in public officials such as the police, government workers and members of parliament (MPs) who are themselves seen to be involved in corruption to some degree. 1.1 MACROECONOMIC PERFORMANCE 2011). Low-income households were affected the most OVER THE LAST DECADE (19.6 percent overall inflation year-on-year) compared 1.1.1 Resilient economic growth to high-income households (14.5 percent overall inflation year-on-year), given smaller expenditure E conomic growth between 2005 and 2015 remained resilient, despite several challenges. The Kenyan economy recorded an average annual real shares for food and transportation for the latter group. The effects of the shocks in 2011 continued into 2012, causing a dip in annual economic growth to 4.6 percent growth rate of 5.3 percent between 2005 and 2015. before rebounding to 5.9 percent in 2013. Overall growth was volatile, including both years of high growth (6.9 percent in 2007 and 8.4 percent in Real GDP per capita growth mirrored economic 2010) and years of low growth (0.2 percent in 2008). The economy faced two major shocks in this period. First, growth (Figure 1.1). GDP per capita growth rose from electoral violence in early 2008 compounded the initial 2.8 percent in 2005 to 4.0 percent in 2007, then fell to effects of the global financial crisis, reducing annual -2.5 percent in 2008. Low growth in the agriculture economic growth to 0.2 percent. The government sector following post-election violence in 2008 was the took quick policy action through a stimulus package, main driver of the decline in per capita growth in 2008. which contributed to an increase in annual growth to Per capita growth peaked at 5.5 percent in 2009. This 3.3 percent in 2009 and 8.4 percent in 2010. A second can be attributed to a recovery in the agriculture sector, dual shock affected the economy in 2011 when implementation of the government economic stimulus international oil prices increased by 37.4 percent while and a recovery in the tourism sector. Since 2009, per a drought in the Horn of Africa reduced food output.12 capita growth has been moderate, reaching 3.2 percent The escalation in food and fuel prices led to an increase in 2016. in overall inflation (18.9 percent year-on-year as of Q3 12 Kenya Economic Update Edition No. 5. 2 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Kenya in Context Figure 1.1: Kenya’s GDP growth from 2005 to 2015 10.0 8.0 6.0 Percent 4.0 2.0 0.0 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 -2.0 -4.0 GDP per capita growth GDP growth Source: KNBS. 1.1.2 Kenya’s performance vis-à-vis the region 2009 from 5.4 percent the previous year. The growth Although economic growth in Kenya exceeded rate in Kenya however took an upward turn in that average growth in Sub-Saharan Africa, Kenya’s year, reaching 3.3 percent thanks the introduction performance lagged behind that of its peers in of a stimulus package designed to counteract the East Africa (Figure 1.2). In Sub-Saharan Africa, annual shock. Even though the performance was better than growth averaged 4.9 percent between 2005 and 2015, average for Sub-Saharan Africa, Kenya’s growth was 0.4 percentage points lower than growth in Kenya. consistently below that of its East African peers, namely However, Kenya’s growth was, on average, lower than Rwanda, Tanzania and Uganda (9.3, 6.6 and 6.7 percent that of Sub-Saharan Africa between 2006 and 2008.13 respectively). Higher growth in these East African Suffering the effects of the 2008 financial crisis, Sub- countries can be explained by the lower base of their Saharan Africa’s growth dropped to 2.9 percent in economic development compared to Kenya. Figure 1.2: Annual GDP growth for Sub-Saharan Africa and selected countries, per year and between 2005 and 2015 10 12 10.6 8 10 9.3 8 GDP growth (%) GDP growth (%) 6 6.7 6.6 7.0 6 5.3 4.9 4 4 2 2 0 0 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Sub-Saharan Kenya Uganda Tanzania Rwanda Ethiopia Ghana Africa Kenya Sub-Saharan Africa Source: World Bank – Mfmod. 13 Growth in the sub-Saharan region pre-2008 financial crisis was driven by high commodity prices. Since Kenya’s main exports are horticulture, tea and coffee, Kenya did not benefit very much from the commodity price boom. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 3 Kenya in Context 1.1.1 Sectoral supply-side growth growth in the manufacturing subsector slowed down Growth was primarily driven by the services sector, in 2012, following the drought in 2011. The drought fueled particularly by ICT and financial services. The had a moderating effect on the production of hydro- mobile phone revolution increased the number of power, which in turn increased production costs in mobile subscribers to 40.2 million in 2015/16 (from manufacturing due to the use of imported backup 17.4 million subscribers in 2008/09), while Internet thermal-generated power. subscriptions jumped to 26.8 million in 2015/16 (from 1.8 million subscribers in 2008/09).14 The mobile Performance of the agricultural sector was dependent phone revolution also contributed to an expansion of on rainfall. Agriculture, which contributes about 23 the financial services sector, driven by the ability to percent to GDP and employs the bulk of the working provide financial services to previously unbanked population, is also the sector that has contributed the households in the form of the mobile payment least to GDP growth, at 0.8 percent on average (Figure system M-PESA, including credit facilities to mobile 1.3). Performance in this sector was highly correlated phone subscribers. with adequate rain, and years of low rainfall exhibit low growth rates growth rates, such as 2011 when rainfall Kenya’s relatively well-developed financial sector was low and growth was merely 2.4 percent (Figure spurred growth through an increase in access to 1.4).16 Low agricultural production affects food prices.17 credit.15 Credit growth to the private sector averaged Another factor that negatively affected the agricultural 19.6 percent between January 2006 and December sector, mainly after the 2008 financial crisis, was a 2015, comparable to credit growth in regional peers hampered demand for horticultural products in the Uganda and Tanzania. Increased lending across sectors euro area. In addition, Kenya’s loss of competitiveness was broad-based, with households/personal loans and within the East Africa region, a consequence of lower the construction sector having a higher concentration. productivity, has seen regional demand for agricultural Credit to the private sector is an important measure products weaken.18 of the depth of financial systems, and consequently Figure 1.3: Contributions to GDP growth an important driver of short run growth. However, 4 credit growth to the private sector declined to a low of -1.3 percent in 2017. Interest rate caps introduced 3.0 Contribution to GDP growth (%) 3 by legislation complicate this declining credit growth, effectively weakening the private sector. 2 The industrial sector grew by 5.8 percent in 2016, mainly due to construction. The industry sector 1.1 contributed 1.1 percent, on average, to GDP growth 1 0.8 annually between 2005 and 2015. The construction sector recorded an average growth of 10.2 percent, 0 compared to only 3 percent for the manufacturing Agriculture Industry Services subsector. The manufacturing sector experienced a Source: KNBS. slowdown during the period of shocks (2008 and 2011). In 2008, uncertainty due to post poll violence saw 16 It is estimated that a 100mm decline in rainfall would reduce GDP growth by 0.5 percentage points. a decline in output in the manufacturing subsector, 17 For example, the year of the drought recorded annual inflation of 14 with growth slowing to 1.1 percent in 2008 and the percent. This was 7 percentage points higher and almost twice as high as the upper limit of the government target rate of 5 percent +/- 2.5 percentage subsector shrinking by -1.1 percent in 2009. Similarly, points. The high inflation levels were mainly driven by food inflation as the price of foodstuffs such as maize increased due to the drought. 14 Communications Authority of Kenya. 18 The Kenya Economic Update Edition 15 notes that Kenya’s exports to the 15 Beck and Fuchs (2004) note that for a country of its size, Kenya has a East African Region declined from a growth rate of about 29.5 percent in relatively well developed financial sector. 2007 to -8.9 percent in 2013. 4 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Kenya in Context Figure 1.4: Agriculture and GDP Growth Increasing infrastructure spending, coupled 2010 with increased private sector investment, drove 8 investment growth, which in turn supported 7 2007 economic growth. Investment contributed 1.5 percent 2006 6 2011 2013 2015 2005 to GDP growth, second to private consumption. The GDP growth (%) 2004 5 2012 1995 2001 government ramped up spending on investment to 1996 4 ease supply-side constraints, making capital the main 2009 1998 3 2003 1994 contributor to economic growth (Figure 1.5). Examples 2 1999 of infrastructure projects include the Thika Highway, 1991 1 2002 2000 the Northern and Southern bypasses and the Standard 2008 1998 1997 0 Gauge Railway. Infrastructure projects such as the -1 1992 ones undertaken by the government aim to increase -6 -4 -2 0 2 4 6 8 10 12 efficiency and reduce production costs, thereby Agriculture growth (%) creating incentives for domestic production. Source: KNBS. 1.1.2 Demand-side growth analysis Growth in the value of imports, which was much Consumption was the main driver of demand-side faster than growth in the value of exports, widened growth, with household consumption contributing the current account balance.21 The growth rate the largest share to GDP growth.19 With an annual of exports slowed from 19.0 percent in 2005 to 5.5 average growth of 4.1 percent between 2005 and 2015, percent in 2015. As a result, the share of the value of household consumption was the largest contributor to exports declined from 24.8 percent of GDP in 2005 GDP growth. A strong financial services sector, which to 16.7 percent of GDP in 2015. In contrast, the value improved access to credit for households, coupled of imports increased, averaging a growth rate of 9.4 with high remittances, supported consumption percent between 2005 and 2015. Consequently, the growth. Additionally, the government stimulus contribution of net exports to GDP averaged -1.3 program introduced in 2009 led to increased growth percent between 2005 and 2015 (Figure 1.6). in consumption, which contributed 5.7 percentage points to the GDP growth of 8.4 percent in 2010.20 Figure 1.5: Productivity and economic growth Figure 1.6: Demand-side contribution to growth between 2005 and 2015 10 5 4.1 4 8 3 6 Growth (%) 2 Percent 1.5 4 1 0.8 0 2 -1 0 -1.3 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 -2 Private Government Investments Net Capital Labour TFP GDP growth consumption consumption exports Source: KNBS and World Bank. Source: KNBS. 19 Note that final household consumption is calculated as a residual and is likely to include errors and omissions – KNBS. 20 The government stimulus was introduced in 2009 to counter the dual shock of post-election violence and the slowdown in demand for exports due to the global financial crisis. 21 The analysis uses values of exports and imports rather than volumes. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 5 Kenya in Context 1.1.3 Drivers of growth decline in human capital during this latter period was Labor was a key driver of real GDP growth. In the due to an increasing labor force (increase in population five-year period prior to 2005, labor contributed 62.5 age 15+)24, without a matching increase in human percent to growth. Labor’s contribution was the capital levels.25 largest, followed by total factor productivity (TFP) with 26.0 percent, and capital with 11.5 percent. However, TFP was a key driver of growth in countries with lower the trend was reversed in subsequent periods, per capita income. For Rwanda and Tanzania, whose with capital contributing 34.3 percent on average average per capita GDP between 2000 and 2015 was between 2010 and 2015, as the contribution of labor USD 567 and USD 712 respectively, TFP was a key driver declined to 43.4 percent during the same period of growth, contributing an average of 38.4 percent (Figure 1.7).22 Government policy increased spending to growth for Rwanda and 46.1 percent for Tanzania. on infrastructure and led to capital becoming a key Capital was the second most important contributor contributor to growth. to GDP growth in Rwanda, while for Tanzania, labor’s contribution to growth followed the TFP contribution Despite labor contributing the largest share to real to growth. In contrast to Rwanda and Tanzania, TFP was GDP growth, this contribution declined as human on average the lowest contributor to GDP growth for capital per unit of labor declined (Figure 1.7).23 The Ghana (24.4 percent), Kenya (25.7 percent) and Uganda contribution to GDP growth from human capital per (26.3 percent) between 2000 and 2015. While labor was unit of labor averaged 14.5 percent in the five-year the second most important contributor to GDP growth period between 2000 and 2005, but became negative for Kenya and Uganda, capital was the second most averaging -2.0 percent between 2005 and 2010. The important source of growth for Ghana (Figure 1.8). Figure 1.7: Contributions to real GDP growth 70 50 40.7 42.3 40.5 60 40 50 30 25.7 Percent 40 Percent 19.1 20 30 14.5 15.2 10 20 10 0 -2.0 0 2000-2005 2005-2010 2010-2015 -10 2000-2005 2005-2010 Capital Labor TFP Capital stock Labor TFP Human capital per labor Source: KNBS, Barro and Lee, and WDI. 24 Kenya’s youth population (15+) increased significantly in the period between 2000 and 2010. The government has put in place several programs through the Ministry of Sports and Culture that could take 22 The results are derived from the Long Term Growth model, a World Bank advantage of the increase in population/labor to increase growth. analytical tool. 25 Lucas and Mbiti (2012) note that education outcomes did not change 23 This analysis uses Barro and Lee’s definition of human capital, defined as significantly in Kenya even with increased access to education through returns to education per year of schooling. the Free Primary Education programs. 6 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Kenya in Context Figure 1.8: Contributions to GDP growth, regional Figure 1.9: Contributions to real GDP per capita growth comparison 2000 - 2015 2 60 50 Percentage points 40 1 Percent 30 20 10 0 0 Labor Labor Labor Labor Labor TFP TFP TFP TFP TFP Capital Capital Capital Capital Capital Stock 2005-2015 2005-2010 2010-2015 -1 Kenya Tanzania Uganda Ghana Rwanda Productivity Employment rate Participation rate Demographic change Source: KNBS and WDI. Source: KNBS and WDI. 1.1.4 Drivers of per capita GDP Growth26 asset and a binding constraint to development. The Productivity remains a key driver of per capita GDP country could benefit from demographic change growth, with potential for Kenya to reap benefits through an increase in working age population and from the demographic dividends.27 From 2005 to 2015, therefore the potentially larger labor force. However, if productivity contributed 1.75 percentage points, which this demographic dividend is not utilized by more jobs, was 81.1 percent of the total GDP per capita growth. the demographic change can have adverse effects on However, productivity’s contribution to per capita GDP productivity and per capita growth. growth was higher in the first half of the period at 1.60 percentage points, compared to 1.91 percentage Intersectoral reallocation was a key driver of GDP per points in the second half of the period (Figure 1.9). capita productivity as labor moved from agriculture This decline in productivity occurred in spite of an to services. Agriculture makes up about a quarter of the increase in the employment rate. One explanation for economy, contributing 25 percent to GDP. However, the decline in productivity could be due to the quality almost 60 percent of the labor force remains in the of jobs created, as jobs requiring only a low skillset are agriculture sector. Between 2005 and 2010, productivity unlikely to increase productivity substantially. in the agriculture sector declined, contributing -0.22 percentage points to per capita GDP growth. Lewis, The second half of the 2005 to 2015 period took in his structural adjustment model, points out that as advantage of a demographic change in Kenya. more labor (a variable resource) is put to work on land (a During this period, the population aged 15+ grew, fixed resource) – in this case agricultural land – marginal leading to an increase in GDP per capita growth of returns to labor will decrease.28 Since marginal returns 0.66 percentage points (Figure 1.9). However, the to other sectors are high, a wage premium in other demographic change implied that the contribution sectors relative to the agriculture sector can emerge. from the participation rate to GDP per capita between Between 2010 and 2015, the reallocation of labor 2010 and 2015 declined, since the population aged between sectors effectively increased productivity in 15+ increased much faster than the number of jobs the agriculture sector. Its contribution to GDP per capita created. Vision 2030, Kenya’s economic blueprint, growth reached 0.39 percentage points. Consequently, notes that rapid population growth can be both an 28 The Lewis structural change model of growth and development defines a dualistic economy where labor is defined as a variable factor input and land as a fixed factor. Initially, labor is concentrated in the 26 This study uses the shapely decomposition method to analyze the key agriculture sector. However, labor reallocates to the center to work in drivers of per capita GDP growth. the manufacturing sector which is more productive and offers a much 27 Productivity is defined as output per worker and is calculated by dividing higher wage premium. Consequently, productivity increases in both output by the total labor force. agriculture and manufacturing with the reallocation of labor. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 7 Kenya in Context the contribution of productivity in the services sector 1.2 FISCAL POLICY AND ECONOMIC to GDP per capita growth declined to 0.50 percentage GROWTH V points between 2010 and 2015 (Figure 1.10). ision 2030, the country’s development blueprint, outlines government spending policy. Vision 2030 Within the region, productivity has been the main has three key pillars: economic, social and political. driver of GDP per capita growth. Productivity was Government spending falls under the economic pillar.29 a key driver of growth contributing between 80 to The plan is implemented in five-year periods, with the 97 percent to GDP per capita growth between 2005 first period covering 2008 – 2012. Implementation of the and 2015 (Figure 1.11). However, compared to its second period (2013 – 2017) is complete. Preparations for peers, Kenya’s productivity contribution to GDP per the third period are under way, with a focus on the “Big 4” capita growth was the lowest at 80.9 percent, while in priorities of food security, affordable housing, enhanced Rwanda productivity accounted for most of GDP per manufacturing, and UHC (Box 1.1). capita growth at 97.9 percent. Demographic change was the second most important driver of GDP per capita growth, at 20.1 percent for Kenya, 12.2 percent 1.2.1 Revenue vs. expenditure for Uganda, 6.8 percent for Ghana and 3.2 percent for Growth in revenue was erratic, increasing in the Rwanda, an indication that the economy benefitted first half of the 2005 to 2015 period and declining from the increase in the working age population. afterwards. Revenue collection peaked in FY2009/10 However, even as the economy benefitted from the at 21.9 percent of GDP, followed by a decline to 17.2 demographic dividend, growth in job creation did not percent of GDP in FY 2015/16. The main source of match the growth in the working age population, as revenue was income tax, which accounted for almost demonstrated by the declining employment rates. The half of revenue collection (an average of 8.1 percent of employment rate contribution to GDP per capita growth GDP in the ten years prior to FY2015/16). Income tax was negative at -6.3 percent for Kenya, -4.0 percent for comprises personal income tax and corporate income Ghana and -2.1 percent for Rwanda. In contrast, the tax. The second most important source of revenue was employment rate was the second most important VAT, averaging 5.0 percent of GDP in the ten years prior driver of per capita GDP growth for Tanzania explaining to FY2015/16. Other sources of revenue include import 5.3 percent of the GDP per capita growth. and excise duties. Figure 1.10: Sectoral contribution to change in real GDP Figure 1.11: Productivity contribution to real GDP per per capita productivity capita growth 2 2005 - 2015 100 1 Percentage points 90 Percent 0 80 -1 2005-2015 2005-2010 2010-2015 Agriculture Industry 70 Services etc. Intersectoral reallocation e ect Kenya Uganda Tanzania Ghana Rwanda Source: KNBS and WDI. Source: KNBS and WDI. 29 Vision 2030 is the GoK development plan. It is aimed at ensuring Kenya achieves middle income status by the year 2030. 8 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Kenya in Context Box 1.1: The Big 4 policy agenda The GoK has announced four key priorities to advance Vision 2030 over the next five years. Known as the Big 4, these priorities are food and nutrition security, affordable housing, increased share of manufacturing, and UHC. Food and nutrition security. The agriculture sector is a key driver of Kenya’s economy, contributing about 50 percent to GDP. Low productivity in the sector, in combination with a growing population, leads to a structural food deficit and poses risks to food security in the country. The sector is characterized by low yields, particularly in grain crops, and vulnerability to climatic shocks. The government intends to invest in sustainably exploiting national water resources through water towers and river ecosystems and to address the distribution, wastage, storage and value-addition of agriculture commodities. Affordable housing. With an estimated housing shortfall of 2 million units, the housing situation in Kenya is expected to deteriorate as urbanization continues. Each year, 500,000 new residents move to urban areas, often residing in informal settlements. Over the next five years, the Government plans to inject capital into the housing sector and provide affordable housing to 500,000 new households. Policy reforms that lower the costs of construction and increase access to mortgages are further intended to increase the affordability of housing. Enhancing manufacturing. The manufacturing sector holds great potential for high job creation, as witnessed by the impressive poverty reduction in countries in Asia. For this to occur, Kenya’s manufacturing firms need to be competitive both domestically against imports and globally in exports, especially within East Africa. Competitiveness challenges in the sector have resulted in a declining share of manufacturing output in GDP. The government aims to increase the share of the manufacturing sector in GDP from 9 percent to 15 percent in the next five years through reductions in power tariffs for manufacturers. UHC. Kenya is in a favorable position to rapidly expand health coverage given the strong institutional foundations and political will. Health insurance is currently concentrated in the formal sector, where contributions are automatically deducted from salaries. However, 70 to 80 percent of the population remains without health insurance coverage, with most of the uninsured in the informal sector. The government aims to achieve 100 percent universal coverage for all households by reforming and expanding the National Hospital Insurance Fund (NHIF). Source: Kenya Economic Update, April 2018. Official website of the presidency of Kenya, April 2018, www.president.go.ke. In contrast, spending increased over this period, GDP in FY2014/15, a reflection of government policy to consistently outpacing revenues. From FY2005/06 increase infrastructure development in a bid to remove to FY2015/16, the government increased deficit supply-side constraints. However, as the government spending (Figure 1.12).30 Recurrent spending was the began fiscal consolidation, development spending main driver of government expenditure, averaging declined in FY2015/16 to 8.2 percent of GDP. about 17.1 percent of GDP over the period. Wages and salaries were the largest component of recurrent Growth in expenditure was faster than growth in spending, with interest payments picking up during revenue, widening the fiscal deficit (Figure 1.13).32 the latter half of the period to 3.2 percent of GDP in The fiscal deficit has been on an upward trajectory, FY2015/16.31 Development spending nearly doubled widening by 3.5 percentage points from -4.7 percent of from 4.5 percent of GDP in FY2005/06 to 8.7 percent of GDP in FY2005/06, to -8.2 percent of GDP in FY2015/16 (Figure 1.14). The -8.2 percent deficit is more than 30 The pane above the dotted line means that spending is higher than revenue (deficit budget), while any points on the dotted line would mean spending equals revenue collections (balanced budget). 32 On the red dotted line, a percentage point change in spending will equal 31 Domestic interest payments make up the larger share of interest a percentage point change in revenue collection, with both variables payments at 2.6 percent of GDP. measured as a percent of GDP. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 9 Kenya in Context Figure 1.12: Spending has consistently exceeded Figure 1.13: Revenue collection has not kept up revenue collection with spending pressures Source: The National Treasury. Source: The National Treasury and World Bank. Figure 1.14: The evolution of fiscal deficit The education sector has been the largest 2005/06 2007/08 2009/10 2011/12 2013/14 2015/16 beneficiary of social sector spending. Education 0 spending maintained its momentum throughout the ten-year period, mirroring government spending. The -2 largest increase in education expenditure as a share -2.7 of the increase in total government spending was in -4 FY2009/10, after which it slowed to its lowest level in Percent -4.2 -4.7 -4.8 FY2013/14 (Figure 1.15).33 However, even as education -4.9 -6 -5.2 -5.6 spending tended to increase with government -5.8 -6.4 spending, the rate of change in the increases was -8 generally low, an indication that education is not likely -8.2 to have any fiscal risk effects. The stable increase in -9.1 education expenditure is attributable to free primary -10 education, as the government employs more teachers Source: The National Treasury. to cater for increasing demand. double the East African Community (EAC) target of 3.0 percent. The government has embarked on a fiscal Implementation of the new constitution led to consolidation plan that should see the deficit decline in moderated health expenditures at the national the medium term to -3.0 percent of GDP in FY2020/21. level. Growth in government spending has trickled down at a slower pace to the health sector compared 1.2.2 Sectoral analysis in government spending to the education sector (Figure 1.15). Following Growth in government spending was uneven. Growth devolution of the health sector in FY2013/14, the in government spending declined from 22.0 percent in momentum of health expenditures at the national FY2009/10 to 3.0 percent in FY2013/14 (Figure 1.15). level slowed down significantly. Full devolution of the The slowdown in spending in FY2013/14 coincided health sector effectively means local governments with the entrance of a new administration and the are responsible for all health care provision. However, implementation of the 2010 Constitution. However, the national government transfers money to a growth in spending accelerated to 14.0 percent in consolidated fund, without specifically earmarking FY2014/15, but decreased moderately to 12.0 percent the amount that should go to the health sectors at the year after. 33 Momentum is defined as the increase in education spending due to an increase in total government spending. See also Merotto et. al. (2015). 10 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Kenya in Context Figure 1.15: Sectoral contribution to growth in 1.3 A REVIEW OF SOME POLICIES OVER THE total spending LAST DECADE P 25 Growth in government expenditure (percent) olicies were designed to increase aggregate demand, in turn contributing to economic growth 20 in Kenya. Kenya’s vision 2030 set a growth target of 10 percent per annum. While the target growth rate has not 15 been achieved, both supply- and demand-side policies aimed at increasing growth have been a recurrent 10 theme in the budget statements over the last 10 years. This section focuses on infrastructure development, 5 use of renewable energy, domestic production, job creation and income inequality reduction, and analyses 0 2009/10 2010/11 2011/12 2012/13 2013/14 2014/15 2015/16 if these policies could also enhance pro-poor growth. Health Education Infrastructure Other expenditure Source: The National Treasury. Supply-side enhancing infrastructure projects, the devolved units. It is unclear if the decline in health such as rail transport and renewable energy, were spending at the national level is substituted at the prioritized. In 2014, exemptions on import duty for devolved units. railway products as well as import duty on machinery, spares and inputs for direct and exclusive use in the Relative increases in social protection spending development and generation of solar and wind energy are large, due to low base effects. In absolute terms, were introduced. During the same period, imports growth in social protection spending increased to a of railway inputs and machinery increased. A more peak of 13.0 percent in FY2010/11. Momentum of social efficient transportation network – attributable to the protection spending, defined as the growth rate of total construction of the US$ 3.6 billion Standard Gauge spending multiplied by the share of social protection Railway – and a stable supply of electricity are crucial spending, was at less than 1.0 percent. This indicates a in reducing the cost of production and fostering scant allocation to social protection expenditure within competitiveness of the manufacturing sector. the overall budget increases. Kenya has abundant clean energy potential which In contrast, not only did infrastructure spending remains untapped. In order to provide incentives have a relatively high growth rate, it also gained to support local production of clean energy, duty momentum. At its peak, infrastructure spending remission was granted on inputs for the production of increased by 74 percent (FY2009/10), accounting for solar panels in FY2013. Geothermal (290MW) and wind more than half of total growth in spending. This reflects (361MW) energy projects were commissioned (Table government policy priorities, namely the national 1.1). In addition, to encourage usage of environmentally development plan, which emphasizes improving friendly vehicles which aimed at reducing carbon infrastructure. Infrastructure spending momentum emission and noise pollution, battery operated vehicles picked up pace from 3.0 percent in FY2009/10 to 5.0 were exempted from duty. However, there are no data percent in FY2015/16, indicating that as the budget to support an increase in imports of environmentally increases, a larger share of spending is allocated friendly vehicles. towards infrastructure. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 11 Kenya in Context Table 1.1: List of ongoing major projects Project value Project name Type Distance (US$ Millions) Standard Gauge Railway Phase 2A Railway 120 Km 1,500 Lamu Port Southern Sudan and Ethiopia Corridor (LAPSSET) Port, Roads, Rail, Pipeline ... Nairobi Mombasa Expressway Road 473 Km 2,300 Northern Corridor Transport Improvement Project Roads Capacity Project value Public Private Partnerships (PPPs) MW (US$ Millions) Thika Power Thermal 87 146 Triumph Thermal 82 156.5 Gulf Power Thermal 80 108 Orpower Geothermal 150 558 Lake Turkana Wind 300 847 Longonot Geothermal 140 760 Kinangop Wind 61 150 Rabai HFO 90 155 Kipevu HFO 74 85 Mumias Bagasse Co-gen 32 50 Source: PPP Unit, National Treasury; Kenya Railways. Growth in domestic production and industrial growth import duty exemptions were granted on television are not only key in ensuring overall growth in the cameras, digital cameras and video camera recorders economy, but are also important for job creation. In while a 100 percent grant was proposed to investment this regard, policies that enhance domestic production deduction on capital expenditure incurred by film spiked from -0.6 percent in 2012 to 5.6 percent in 2013, producers on purchase of any filming equipment. The the growth was attributable to low base effects and purpose of the exemptions was twofold: i) the film a successful election period. Some pro-poor policies industry has traditionally had a very low performance on the supply side included the removal of the sugar in Kenya, with the introduction of exemptions aimed development levy, duty exemptions for raw materials at creating incentives to promote the industry, and ii) used in the manufacture of sanitary towels, as well as the film industry has potential to create employment duty exemptions on all synthetic yarns, acrylic yarn and for the youth, who are the majority of the population polyester yarn. Additionally, duty was eliminated for in Kenya. Indeed, employment in the modern sector basic commodities that make up the largest share of has increased in recent years (Figure 1.16). A second the consumption basket for poor households. Further, area envisaged as creating potential employment such interventions also increase the competitiveness for the youth was the transport sector, in particular of exports to the region, which in turn has positive the motorcycle taxi. Motorcycles are a relatively new implications for macroeconomic indicators such as mode of transportation in the cities, which create jobs the current account balance and the exchange rate. often for youth as motorcycle taxi drivers. Motorcycles However, exports to the region have declined despite have the advantage of being much faster than motor the export-friendly interventions. vehicles given traffic congestion. In FY2009, duty on motorcycles of between 50cc and 250cc were zero Non-traditional sectors were identified as a potential rated, potentially contributing to an increase in the source of employment. In FY2009 and FY2010, VAT and informal sector employment. 12 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Kenya in Context Figure 1.16: Employment trends While the national poverty lines are critical to analyze 180 poverty dynamics and distribution within the country, they are not comparable across countries. 160 Kenya’s national poverty line is derived from the Cost of Basic Needs (CBN) method.36 The CBN method Index=2009 = 100 140 stipulates a consumption bundle deemed to be 120 adequate for “basic consumption needs”, and then estimates what this bundle costs in reference prices. As 100 basic consumption needs are usually different across 80 countries, the poverty rate measured by the national poverty line is not comparable across countries. 60 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Therefore, this section uses the international poverty Modern sector employment Informal sector employment line defined at US$ 1.90 using 2011 purchasing power Total employment parity (PPP) international dollars (Box 1.2). Chapter 2 Source: KNBS. provides a detailed assessment of poverty trends at the The social pillar of Kenya’s Vision 2030 places national poverty line. emphasis on improved quality of life for all Kenyans. An important condition for a higher standard of living 1.4.1 Monetary poverty trends at the international and therefore quality of life is an increase in income. poverty line Policies to enhance income equality over the period About 1 out of 3 people in Kenya live below the included reduction of taxes on basic commodities international poverty line. The daily consumption typically consumed by poorer households, such as expenditure for 36.8 percent of the population is second-hand clothing. A reduction of import duty below US$ 1.90 in 2011 PPP. For 66.2 percent of the from US$ 0.3 per kg to US$ 0.2 per kg on second-hand population it is below US$ 3.20 in 2011 PPP (Box 1.2). The poverty rate has moderately reduced over the past clothing was implemented in FY2010. Similarly, in decade at both international poverty lines, dropping FY2015, all imported farm inputs used in the processing nearly seven percentage points at the US$ 1.90 line and and preservation of seeds for planting were exempted. three percentage points at the US$ 3.20 line between 2005 and 2011 (Figure 1.17). Poverty reduction has 1.4 OVERVIEW OF MONETARY POVERTY34 been steady over the past decade, except for a shock P overty incidence declined from 46.8 percent in 2005/06 to 36.1 percent in 2015/16, using Kenya’s official national poverty lines. The KNBS released the to consumption in the years following the 2008 global economic crisis (Figure 1.19). most recent poverty statistics in March 2018, based on Increased consumption for the poorest of the poor the KIHBS 2015/16. KIHBS 2015/16 closes an important has driven poverty reduction in the past decade. The data gap, as the previous survey collecting expenditure rate of extreme poverty under the threshold of US$1.20 a day in 2011 PPPs has decreased by 7.3 percentage data to estimate poverty was implemented 10 years ago points since 2005 to reach 13.7 percent in 2015 (Figure in 2005/06.35 1.17). The reduced poverty at the US$ 1.90 international poverty line reflects these improvements. The depth of poverty can be measured by the poverty gap index, representing the average deficit between the 34 This section is derived from the Poverty Special Focus of the Kenya Economic Update, April 2018. total consumption of the poor and the international 35 The KIHBS 2015/16 utilized a two-stage stratified cluster sampling method with the objective of providing data for poverty estimates at poverty line. Using this measure, the depth of poverty national and county levels as well as for urban and rural areas. The sample at the US$ 1.90 line decreased from 16.2 percent of the included 24,000 households from 2,400 clusters distributed to urban and rural strata for each of the 47 counties in Kenya based on the 2009 poverty line in 2005 to 11.6 percent in 2015 (Table 1.2). Census. The survey was implemented over 12 months from September 2015 to August 2016 to take into account seasonal effects. Source: KNBS 36 Ravallion, Martin. 1994. “Measuring Social Welfare With and Without (2018): “Basic Report on Well-Being in Kenya”. Poverty Lines.” The American Economic Review 84 (2): 359–364. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 13 Kenya in Context Table 1.2: Key monetary poverty Indicators37 Poverty headcount Poverty gap 2005 2015 2005 2015 US$ 1.20 2011 PPP poverty line38 21.0 13.7 6.7 3.6 US$ 1.90 2011 PPP poverty line 43.7 36.8 16.2 11.6 US$ 3.20 2011 PPP poverty line 69.2 66.2 33.3 28.4 Source: KIHBS 2005, KIHBS 2015, authors’ calculations. Box 1.2: The international poverty lines The international poverty line is defined in absolute terms as a threshold of being able to purchase a fixed basket of goods that meets basic needs across countries. The concept of an international poverty line was first introduced in the 1990 World Development Report. The objective was to measure poverty in a consistent way across countries, using a poverty line that reflected conditions of poverty in poor countries, while also considering real purchasing power across countries of all incomes. To decide on an international poverty line, the World Bank analyzed data from 33 national poverty lines from both developed and developing countries in the 1970s and 1980s. The threshold of US$ 1 a day was agreed upon and became the first international poverty line. Over the years, the poverty line has periodically been adjusted as new purchasing power parity (PPP) measures became available. The new measures reflected both changes in relative price levels across countries, as well as changes to methodologies. The poverty line increased from US$ 1 a day at 1985 PPPs to US$ 1.08 at 1993 PPPs, then to US$ 1.25 at 2005 PPPs, and finally to its current level of US$ 1.90 at 2011 PPPs. The increase in the international poverty line can be mostly attributed to changes in U.S. dollar purchasing power relative to the purchasing power of the local currencies in the poorest countries. Essentially, the increase in the poverty line says that US$ 1.90 in 2011 real terms would buy about the same basket of goods that US$ 1.25 bought in 2005. The World Bank introduced an additional set of international poverty lines in 2016, taking into account the relationship between national poverty lines and the wealth of the country. These lines are defined as the median national poverty line for each grouping of countries by their GNI per capita, using the World Bank classification of countries as low-income, lower middle-income, upper middle-income and high-income. The World Bank now reports poverty rates for countries using the new lower middle-income and upper middle-income poverty lines. The poverty line for lower middle-income countries is US$ 3.21 per day and for upper middle-income countries, it is US$ 5.48 per day. In addition to these poverty lines, this section also uses a US$ 1.25 2011 PPP poverty line to further distinguish between the poor living below US$ 1.90 and the poorest living below US$ 1.25. To allow for international comparisons, poverty in this section is estimated using the current international poverty line and the lower middle-income class (LMIC) poverty line. Since 2014, Kenya has been classified as a lower middle-income country. Its current GNI per capita of US$ 1,380 puts it at the bottom of the LMIC grouping.39 As the poverty lines are defined using US$ 2011 PPPs, this is converted to the local currency used to measure consumption for both survey years 2005 and 2015. First, US$ 2011 are converted into Kenyan Shilling in 2011 using the PPP estimate for Kenya (35.43). Second, the change in purchasing power per Kenyan Shilling is adjusted for by considering inflation or deflation to the survey period as measured by the national CPI. 37 Poverty estimates in this section are preliminary. The official source for World Bank estimated poverty headcounts is PovcalNet. For the estimation for poverty in this section, the poverty line was adjusted using the 2011 PPP estimate and inflated or deflated to the survey period. The official consumer price index (CPI) used for 2011 was 121.1654. For the KIHBS 2005, the weighted average of the official CPI for the survey period was 73.2557. For the KIHBS 2015 survey period, it was 166.299. Poverty was estimated with a per capita aggregate for consumption expenditure. The aggregate was not spatially deflated and excludes rent, unlike the aggregate used in the Poverty Special Focus of the Kenya Economic Update, April 2018. Thus, poverty estimates in this section differ slightly from those in the Economic Update. 38 The US$ 1.20 line is not an international poverty line. It is included in this section for the purposes of distinguishing the poorest in extreme poverty (Box 1.2). 39 Source: World Bank Open Data Catalogue. 14 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Kenya in Context To further distinguish the poorest of the poor, a poverty line of US$ 1.20 in 2011 PPP is included in this section. This line is based on the share of food consumption in total expenditure. On average, Kenyans spend 63 percent of their total daily consumption on food consumption. Starting with the US$ 1.90 international poverty line as a threshold for total consumption, this translates into daily per capita food consumption of US$ 1.20 in 2011 PPP. Those living below US$ 1.20 a day cannot afford the minimum food consumption calories even if they were to cut out all non-food consumption. As the food share specific to Kenya is used to derive this line, it is not suitable for international comparisons. It is only used in this section to distinguish the poorest in extreme poverty. Well-being has stagnated for households living To estimate the relationship between household between the US$ 1.90 and US$3.20 poverty lines. The consumption and growth at the sector level, the percentage of the population consuming between evolution of poverty from 2005 to 2015 is simulated US$1.90 and US$3.20 increased by 3.9 percentage based on sectoral growth rates, while assuming no points between 2005 and 2015 (Figure 1.17). This is not redistribution beyond that resulting from differences surprising as increases in consumption of the very poor in sectoral growth. Consumption expenditure per have pushed them above the US$ 1.90 poverty line household from KIHBS 2005 is augmented based on while in the same period not as many (net) households the growth rate of the household head’s sector of increased consumption beyond US$ 3.20. Therefore, still economic activity. The poverty rate per sector in KIHBS many households have a certain degree of vulnerability 2015 provides the anchor to estimate the growth- to fall back into poverty measured at the US$ 1.90 level. consumption pass-through parameter of that sector.40 A 10 percent consumption shock would push a fifth of In other words, the pass-through parameter ensures households currently between US$ 1.90 and US$ 3.20 that sectoral GDP growth transmitted to household below the US$ 1.90 a day threshold, raising the poverty consumption growth is consistent with the observed headcount by six percentage points (Figure 1.18). changes in poverty between 2005 and 2015. The pass- Figure 1.17: Poverty at the US$ 1.20, 1.90, and 3.20 lines Figure 1.18: Cumulative consumption distribution with shock 80 100 Poverty headcount (% of population) 80 Percent of population 60 25.5 29.4 60 40 40 22.7 23.1 20 20 21.0 0 13.7 0 100 200 300 400 500 0 2005 2015 Average per capita monthly consumption, US$ 2015 PPP Poverty under US$3.20 USD a day Poverty under US$1.90 USD a day US$ 1.90 Poverty Line Consumption, KIHBS 2015 Poverty under US$1.20 USD a day US$ 3.20 Poverty Line Consumption, 10% shock Source: KIHBS 2005, KIHBS 2015, authors’ calculations. Source: KIHBS 2015, authors’ calculations. 40 Occupations are categorized into three broad categories: (1) agriculture; (2) manufacturing; (3) services. Assumptions about sectoral pass-through parameters for these sector groupings are drawn from the sectoral decomposition of poverty analysis between 2005 and 2015. Parameters are assumed to be constant over years. For households without reported household head occupation, average GDP growth is applied. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 15 Kenya in Context Figure 1.19: GDP sectoral growth simulation of poverty Figure 1.20: Overall GDP growth simulation of poverty trajectory at international poverty lines, 2005 to 2015 trajectory at international poverty lines, 2005 to 2015 80 80 70 70 Poverty headcount (% of population) Poverty headcount (% of population) 60 60 50 50 40 40 30 30 20 20 10 10 0 0 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Poverty, under US$1.20 a day Poverty, under US$1.90 a day Poverty, under US$1.20 a day Poverty, under US$1.90 a day Poverty, under US$3.20 a day Poverty, under US$3.20 a day Source: KIHBS 2005, authors’ calculations. Source: KIHBS 2005, authors’ calculations. through parameter indicates the fraction of sectoral (Figure 1.21). From 2011 to 2015, growth averaged GDP growth that translates into private household 4.1 percent. Most household heads are engaged in consumption. While a large pass-through parameter agriculture, followed by services and then industry suggests that high GDP growth helps to improve (Figure 1.22). Households engaged in agriculture benefit consumption of households, it also flags the risk from the highest pass-through rate, especially for those that high GDP volatility translates into consumption consuming less than US$1.20 a day (Figure 1.23). For volatility, making households vulnerable to shocks that these households, real consumption increases by 0.75 affect GDP growth. percent for each one percent growth in the agriculture sector. The flipside of a high pass-through rate is the Agricultural GDP growth largely translates into vulnerability to shocks. The industrial sector has the consumption growth, exposing agricultural smaller pass-through rate, indicating a protection households to shocks in agricultural GDP. In the against shocks of GDP growth but also implying that years following the slow-down of growth in 2008, households in this sector participate less in sectoral the agriculture sector experienced a strong rebound GDP growth. Figure 1.21: Real sector growth, 2007 to 2015 Figure 1.22: Share of households by sector of household head occupation, 2005 vs. 2015 60 12 10 50 8 Percentage of households 6 40 4 30 2 0 20 -2 10 -4 -6 0 2007 2008 2009 2010 2011 2012 2013 2014 2015 Agriculture Industry Services Agriculture GDP Services Industry 2005 2015 Source: KNBS. Source: KIHBS 2005, KIHBS 2015, authors’ calculations. 16 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Kenya in Context Figure 1.23: Consistent sectoral elasticities for poverty Figure 1.24: Combination of growth and redistribution pass-through23 needed to eradicate poverty in 2030 0.8 3.5 0.7 Annual pace in inequality reduction (Percent) 3 0.6 2.5 Pass-through rate 0.5 2 0.4 1.5 0.3 0.2 1 0.1 0.5 0 Agriculture Industry Services 0 0 2 4 6 8 10 12 Sectoral pass-through, 1.20 Sectoral pass-through 1.90 Sectoral pass-through, 3.20 GDP pass-through, 1.90 Annual growth in household consumption (Percent) Source: Authors’ calculations. Source: KIHBS 2015, authors’ calculations. Kenya is not on track to eradicate poverty by 2030, 1.4.2 Monetary poverty in international comparison and higher and more inclusive growth, as well as pro- Kenya’s poverty rate is below the average in sub- poor policies, are needed. In order achieve a poverty Saharan Africa and is amongst the lowest of its East rate below 3 percent by 2030, the poverty rate must African peers.42 The poverty rate at the US$ 1.90 a day decrease at least 33.8 percentage points. However, line in Kenya is nearly half the poverty rate of Rwanda in Kenya’s annualized poverty reduction rate was 1.6 2013 (60.4 percent). However, it is higher than poverty percent between 2005 and 2015. Assuming this rate is in Uganda (34.6 percent) and Ghana (13.6 percent), maintained for the next 15 years, the poverty rate will both measured in 2012 (Figure 1.25). When considering remain above 25 percent in 2030. To meet the 3 percent GDP per capita in constant PPP terms, poverty in Kenya goal in 2030, an annual poverty reduction rate of 6.1 is in line with expectations given the trend of poverty percent would be necessary. Without any reduction to GDP per capita in sub-Saharan Africa (Figure 1.26). in inequality, real household consumption would Kenya’s ratio of poverty to GDP per capita is close to need to increase on average by 11.3 percent per year that of the sub-Saharan Africa aggregate. Ghana and to achieve this objective. With the observed growth- Uganda both have lower ratios of poverty to GDP per consumption pass-through of 0.25, this would imply capita. However, it is important to note that Kenya has an unrealistically high annual GDP growth of about 45 the most recent estimate for poverty (2015), which may percent. Thus, high growth must be complemented by bias its performance in comparison to countries with stronger inclusive growth, increasing the pass-through older poverty estimates such as Ghana and Uganda parameter, and a reduction in inequality through pro- (both 2012). poor policies (Figure 1.24). 42 Four countries were selected for the international comparison due to geographic proximity, comparable population size and/or level of 41 This figure shows the sector elasticity assumptions for the trajectory of wealth: Ghana (GHA), Rwanda (RWA), Tanzania (TZA), and Uganda (UGA). poverty simulations at the US$ 1.20, 1.90, and 3.20 per day poverty line The aggregate for Sub-Saharan Africa is also included as a regional thresholds. For each threshold simulation, different sectoral elasticities benchmark. Tanzania has a GDP PPP per capita ($2,583) comparable were assumed. The pass-through rate is generally highest for poverty to that of Kenya ($2,926), while Ghana ($3,980) is relatively wealthier. under the US$ 1.20 level, indicating that growth has a larger impact Rwanda ($1,774) and Uganda ($1,687) are both relatively poorer than on consumption of the very poor. The pass-through rate of overall Kenya. In terms of population, Tanzania (55.6 million) and Uganda (41.5 GDP growth, in the US$ 1.90 poverty line simulation, is included as a million) are similar in size to Kenya (48.5 million), whereas Ghana (28.2 benchmark. million) and Rwanda (11.9 million) are notably smaller. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 17 Kenya in Context Figure 1.25: International comparison of poverty Figure 1.26: Poverty headcount against GDP per capita 70 100 90 60 80 Poverty headcount (% of population) Poverty headcount (% of population) 50 70 Rwanda 2013 60 40 50 Kenya 2005 Tanzania 2011 30 40 SSA 2013 30 Uganda 2012 Kenya 2015 20 20 10 10 Ghana 2012 0 0 Rwanda Tanzania SSA Kenya Uganda Ghana 6.0 7.0 8.0 9.0 10.0 11.0 2013 2011 2013 2015 2012 2012 Log GDP per capita, constant PPP Source: KIHBS 2015, World Bank open data catalogue, authors’ calculations. Source: KIHBS 2015, KIHBS 2005, World Bank open data catalogue, authors’ calculations. The depth of poverty at the international poverty When considering Kenya’s LMIC status, poverty is line is consistent with expectations. The relationship relatively high. Poverty in Kenya is higher than the between the poverty headcount and the poverty gap aggregate for LMIC countries, both at the US$ 1.90 in Kenya conforms to the trend for sub-Saharan African and US$ 3.20 lines (Figure 1.28). Ghana provides an countries (Figure 1.27). Kenya’s poverty gap is close to appropriate benchmark as it has a similar GNI per capita that of Uganda (10.3 percent), but is notably higher to Kenya (US$ 1,380). The poverty headcount in Ghana than in Ghana (4.0 percent). The improvement in the at the LMIC line (34.9 percent) is 28.8 percentage points poverty gap since 2005 suggests that many of the poor less than that in Kenya. Poverty in Kenya is also much are close to reaching the US$ 1.90 a day consumption deeper at the LMIC line than it is at the international threshold. This reflects Kenya’s notable reduction in poverty line. The poverty gap at the LMIC line is 27.5 poverty below US$ 1.20 a day since 2005. percent, compared to 11.3 percent at the international poverty line. Kenya’s depth of poverty at the LMIC Figure 1.27: Poverty rate against depth at international line is substantially higher than Ghana and the LMIC poverty line aggregate (Figure 1.29). 45 40 Kenya has a relatively weak relationship between 35 Poverty gap (% of poverty line) poverty reduction and GDP growth. Between 2005 30 and 2015, annualized GDP per capita growth in Kenya 25 Rwanda 2013 was 2.75 percent, while the annualized reduction in the 20 poverty rate was 0.7 percentage points, or 1.58 percent. SSA 2013 Kenya 2005 15 Tanzania 2011 This gives Kenya an elasticity of poverty reduction to 10 Kenya 2015 GDP growth of 0.57, meaning that for every 1 percent Uganda 2012 increase in GDP per year, the poverty rate decreases 5 Ghana 2012 by 0.57 percent. This elasticity is lower than the sub- 0 0 20 40 60 80 100 Saharan aggregate (0.74), as well as Tanzania, Ghana Poverty headcount (% of population) and Uganda (Figure 1.30). Kenya’s ratio of GDP per Source: KIHBS 2015, World Bank open data catalogue, authors’ calculations. capita to elasticity is in line with the sub-Saharan Africa aggregate (Figure 1.31). 18 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Kenya in Context Figure 1.28: Poverty headcount at IPL and LMIC, Figure 1.29: Poverty gap at IPL and LMIC, international international comparison comparison 70 30 60 25 Poverty headcount (% of population) Poverty gap (% od poverty line) 50 20 40 15 30 10 20 5 10 0 0 Kenya 2015 Ghana 2012 LMIC 2013 Kenya 2015 Ghana 2012 LMIC 2013 Poverty rate at US$1.90 PPP per day line Poverty rate at US$1.90 PPP per day line Poverty rate at US$3.20 PPP per day line (LMIC) Poverty rate at US$3.20 PPP per day line (LMIC) Source: KIHBS 2015, World Bank open data catalogue, authors’ calculations. Source: KIHBS 2015, World Bank open data catalogue, authors’ calculations. Figure 1.30: International comparison of elasticity of Figure 1.31: Elasticity of poverty reduction against poverty reduction GDP per capita Rwanda Kenya Sub- Uganda Ghana Tanzania Log GDP per capita, constant PPP Saharan Africa 7.0 7.5 8.0 8.5 0.0 0.0 % change in poverty per 1% change in GDP Rwanda -0.2 -0.2 % change in poverty per 1% change in -0.4 -0.4 GDP per capita PPP -0.6 Kenya -0.6 Sub-Saharan -0.8 Africa -0.8 -1.0 Uganda -1.0 -1.2 -1.2 Ghana -1.4 Tanzania -1.4 -1.6 -1.6 Source: KIHBS 2015, KIHBS 2005, World Bank open data catalogue, authors’ Source: KIHBS 2015, KIHBS 2005, World Bank open data catalogue, authors’ calculations. calculations. 1.5 OVERVIEW OF NON-MONETARY defined as a daily per capita consumption expenditure POVERTY below US$ 1.90 in 2011 PPP, which affects 36.8 percent P oor households are often deprived in multiple of households. In education indicators, nearly one third dimensions. The most common type of deprivation of all households are deprived in adult educational is access to services, notably sanitation and electricity attainment, meaning no adult in the household has (Figure 1.32). 40.7 percent of households lack access completed primary education. Primary school enrollment to improved sanitation43 and 64 percent lack access to is the least common deprivation. Less than one quarter electricity. Fewer households are deprived of access to of all households (23.7 percent) have a child of primary- an improved source for drinking water44 (28.2 percent). school age not currently attending primary school. The second most common deprivation is monetary, 43 Improved sanitation is defined as a toilet with a flush, a ventilated improved pit (VIP) latrine or a latrine with a slab. 44 Improved drinking water sources are defined as a piped water system, public tap, borehole, protected dug well, bottled water or water from rainwater collection vendors. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 19 Kenya in Context Figure 1.32: Multi-dimensional deprivations, 2015 (68 percent) and is in line with its level of poverty 70 (Figure 1.33). Kenya performs much better in access to improved sanitation compared to countries with a 60 comparable poverty headcount (Figure 1.34). 50 % of households deprived Kenya’s performance on human development 40 indicators has improved since 2015, but lags behind 30 Ghana. Kenya’s Human Development Index (HDI), 20 calculated by the United Nations Development Program (UNDP) as a combination of education, inequality, and 10 life expectancy indicators, gained 0.07 points in the past 0 decade to reach 0.55 in 2015. This is the highest HDI in Consumption Adult Primary Improved Improved Access to education school water sanitation electricity the EAC, but still behind Ghana (0.58). Kenya’s level of attainment enrollment human development is relatively high given its poverty Source: KIHBS 2015, authors’ calculations. headcount (Figure 1.35), indicating that Kenya performs Kenya has a relatively high level of access to improved better on non-monetary dimensions of poverty. sanitation compared to international benchmarks, but lags behind in access to improved water. The lack Kenya’s adult literacy rate is among the highest in of improved water sources increases the time burden Africa. In 2015, 84 percent of the population aged 15 for women and children, who generally bear the years and over could read and write in any language, responsibility of fetching water. Though progress has a larger proportion of the population than in a country been made in improving access to improved water like Ghana (71 percent), which has a much lower poverty since 2005, Kenya still lags behind other countries in rate (Figure 1.36). The literacy rate has increased by 11 the international comparison. Only 71.8 percent of percentage points since 2005, reflecting the progress Kenyan households have access to improved water in enrollment in Kenya over the past decade. This is in sources. This is below the level of peer countries like line with results from standardized tests suggesting Ghana, Rwanda and Uganda. Kenya’s rate of improved that Kenyan children have somewhat better learning water is close to the average for sub-Saharan Africa outcomes in primary school than children in other Figure 1.33: Poverty headcount against access Figure 1.34: Poverty headcount against access to improved water to improved sanitation 100 100 Ghana 2012 Access to improved sanitation (% of households) 90 90 Access to improved water (% of households) Uganda 2012 80 Rwanda 2013 80 70 Kenya 2015 70 SSA 2013 60 Kenya 2015 Rwanda 2013 60 Kenya 2005 Tanzania 2011 50 50 Kenya 2005 40 40 30 30 SSA 2013 20 20 Tanzania 2011 Uganda 2012 10 10 Ghana 2012 0 0 0 20 40 60 80 0 20 40 60 80 100 Poverty headcount (% of population) Poverty headcount (% of population) Source: KIHBS 2015, KIHBS2005, World Bank open data catalogue, authors’ Source: KIHBS 2015, KIHBS2005, World Bank open data catalogue, authors’ calculations. calculations. 20 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Kenya in Context Figure 1.35: Poverty headcount against HDI Figure 1.36: Poverty headcount against literacy rates 0.9 100 90 Kenya 2015 0.8 Adult literacy rate (% population 15+) 80 0.7 Uganda 2012 Ghana 2015 70 Rwanda 2013 0.6 Kenya 2015 Ghana 2012 Kenya 2005 Tanzania 2015 Rwanda 2015 60 0.5 50 HDI Uganda 2015 Kenya 2005 0.4 40 0.3 30 0.2 20 0.1 10 0 0 0 10 20 30 40 50 60 70 80 90 0 20 40 60 80 100 Poverty Headcount (% of population) Poverty headcount (% of population) Source: UNDP HDI. Source: KIHBS 2015, KIHBS2005, World Bank open data catalogue, authors’ calculations. countries in the region.45 However, significant gender poverty rate (Figure 1.37). However, Kenya’s rate of gaps in adult literacy continue to exist, reflecting adult primary school completion is lower than in gender inequalities in primary education. Ghana and Tanzania. When considering higher levels of educational attainment, Kenya performs worse (Figure In line with increasing enrollment rates, levels of 1.38). Only 14.4 percent of adults aged 25 and older educational attainment among the adult population have completed secondary education. While this also have increased. Over half (57.8 percent) of all Kenyan marks a substantial improvement over 2005 when only adults above the age of 24 have completed primary 3 percent of Kenyan adults had completed secondary education. This marks a notable increase from 2005 school, it is far below rates found in other countries (44.2 percent). Adult primary educational attainment with comparable levels of poverty.46 is high compared with countries that have a similar Figure 1.37: Poverty headcount against adult educational Figure 1.38: Poverty headcount against adult educational attainment, primary attainment, secondary 100 100 90 90 80 80 Completed secondary education Completed primary education 70 Tanzania 2012 70 (% population 25+) (% population 25+) Kenya 2015 60 Ghana 2012 60 Ghana 2012 50 50 Kenya 2005 40 Uganda 2012 40 30 Rwanda 2013 30 Uganda 2012 20 20 Kenya 2015 10 10 Rwanda 2013 Tanzania 2012 0 Kenya 2005 0 0 20 40 60 80 100 0 20 40 60 80 100 Poverty headcount (% of population) Poverty headcount (% of population) Source: KIHBS 2015, KIHBS2005, World Bank open data catalogue, authors’ Source: KIHBS 2015, KIHBS2005, World Bank open data catalogue, authors’ calculations. calculations. 46 The results might exaggerate differences, as primary education in 45 Sandefur, Justin. 2018. “Internationally comparable mathematics scores Kenya is eight years but only seven and six years in Tanzania and Ghana. for fourteen African countries.” Economics of Education Review 62 (2018): Kenyan primary school children also score higher on standardized tests 267-286. than Tanzanians. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 21 Kenya in Context Kenya’s net school enrollment rates have improved use of insecticide-treated bed nets (ITNs) that protect over the last decade. The net primary school enrollment children from contracting malaria.48 The decline has rate, the proportion of age-eligible children who are been particularly pronounced among children from currently enrolled in primary, is estimated at 84.6 poorer families and those residing in rural areas; in percent in 2015/16. This is lower than expected given fact, differences in mortality between the bottom 40 Kenya’s poverty headcount. Within the EAC, Uganda percent49 and the top 20 percent and rural and urban and Rwanda both have higher net enrollment rates children were not statistically significant in 2014. Kenya’s (NERs). However, the net secondary school enrollment under-five mortality rate is lower than expected given rate in Kenya is now the highest among countries the country’s level of poverty and is among the lowest of the EAC, at 42.2 percent.47 It more than doubled in sub-Saharan Africa (Figure 1.39). since 2005 (21.0 percentage points) and is in line with expectations given Kenya’s poverty level. Increases in Kenya has also made substantial gains in reducing secondary enrollment in recent years are expected to child stunting; it now has one of the lowest stunting boost educational attainment among young adults in rates in the region. Stunting is defined as a height-for- the near future. age z-score that is more than two standard deviations below the median of a reference population.50 As of Under-five mortality has declined rapidly in recent 2015, nearly 1 out of every 5 children under the age years, particularly among the poor, giving Kenya of 4 (24.4 percent) is stunted in Kenya. While this is the one of the lowest under-five mortality rates in the lowest stunting rate among countries of the EAC, it is region. Mortality among children below the age of five still higher than in Ghana. When considering Kenya’s has declined from 114.6 deaths per 1,000 live births level of poverty, the rate of stunting is lower than in 2003 to only 52.4 in 2014. This decline has been expected (Figure 1.40). The prevalence of child stunting driven mostly by the increased provision and uptake has substantially improved since 2005, when 40.1 of low-cost, high-impact measures, particularly the percent of Kenyan children were stunted. Figure 1.39: Poverty headcount against under- five Figure 1.40: Poverty headcount against child stunting mortality 180 70 160 Prevalence of stunting (% of children under 5) 60 140 50 (deaths per 1,000 live births) 120 Kenya 2003 Rwanda 2013 Under 5 child mortality Kenya 2005 100 40 Uganda 2012 Tanzania 2011 80 Uganda 2016 30 SSA 2013 Tanzania 2015 60 Ghana 2012 Ghana 2014 Rwanda 2015 20 Kenya 2015 40 Kenya 2014 20 10 0 0 0 20 40 60 80 100 0 20 40 60 80 100 Poverty headcount (% of population) Poverty headcount (% of population) Source: USAID Demographic and Health Survey (DHS). Source: KIHBS 2015, KIHBS2005, World Bank open data catalogue, authors’ calculations. 48 The share of children under the age of five that sleeps under an ITN increased from only 4.6 percent in 2003 to 54.3 in 2014. 47 The net secondary school enrollment rate is similarly defined as 49 The statement is based on comparisons across quintiles of a wealth index the ratio of secondary school-aged children who are currently that uses assets to proxy the material standard of living, not consumption enrolled in secondary school to the population of all secondary expenditures. school-aged children. 50 The reference used here is that of the World Health Organization (WHO). 22 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Kenya in Context 1.6 INSTITUTIONAL CONTEXT, ELECTIONS Orange Democratic Movement (ODM) was politically AND DEVOLUTION supported by other ethnic groups, particularly the K enya is a presidential-style democratic republic dominant Luo community. The sitting president Mwai based on a multiparty system in accordance with Kibaki of the PNU was initially declared the winner of a constitution passed in 2010. The president of Kenya is a contentious election. The results were immediately both the head of state and the head of government, and challenged by the ODM, citing voter intimidation and leads the executive branch. Legislative powers rest with a other irregularities. The situation was exacerbated bicameral parliament while the judiciary is independent by the Electoral Commission’s own admission of of these two branches. Although democratic processes, inconsistencies in the process.55 The elections damaged particularly elections, are at times accompanied by Kenya’s image as a relatively-stable country with politically-instigated civil unrest and violence, the country politically mature institutions. 56 57 is considered to have a wider democratic space compared to its neighbors.51 Following major institutional reforms The country undertook reconciliatory measures initiated after the Presidential elections of 2007, there is following the political discord. A power-sharing currently a national government and 47 county-level arrangement with intense support of the international governments that exercise executive and legislative community ended the violence and led to the powers at different levels. formation of the Unity government in 2008, comprising the incumbent PNU and the opposition ODM. The The traditional concentration of power in the constitution was altered to create a new position of executive branch has been a source of political Prime Minister for the opposition’s candidate. The grievance. Since independence, there has been a Afrobarometer Survey conducted in 2008 showed 83 “continuous process of centralization of power” as well percent of Kenyans supporting a constitution that limits as concentration of power in the Presidency.52 This the president to two terms in office, and 77 percent resulted in a sweeping mandate that allowed for, at thought the National Assembly and MPs represent the different times, the redrawing of districts to create new people and should therefore make laws even if the offices for the president’s allies. In addition, new power President or Prime Minister did not agree with them.58 centers at the sub-national level were created, such as the Provincial Administration, that answered directly to A constitutional referendum in 2010 created new the executive.53 The executive was also able to hand out checks on executive power. This process also led to public land to patrons and affiliates. the complete separation of the parliament from the executive under a presidential system of government. The 2007 elections were marked by widespread Political decentralization had always found some political violence and a serious challenge to the degree of support within the diverse communities legitimacy of the electoral system. The frontrunner in Kenya and the country did have some features Party of National Unity (PNU) was widely perceived to of regional autonomy at independence. Successive dominate power and access to resources including land leaders – the founder Kenyatta, followed by President and was led by the Kikuyu community.54 The opposition Moi – centralized state power and influenced key 55 From New York Times coverage of the 2007 elections; Africa: Disputed 51 See Op-Ed “Africa’s Powerhouse” by Kimenyi and Kibe, 6th January 2014; vote plunges Kenya in bloodshed, 31st December, 2007.Article by J. online at www.brookings.edu. Gettleman. 52 Sundet, Geir, Scanteam, and Eli Moen. 2009. Political Economy Analysis 56 Civil unrest over two months recorded over 1,000 dead and up to 500,000 of Kenya. Norwegian Agency for Development Cooperation Report internally displaced, as per Human Rights Watch: see Report titled “Ballots 19/2009. to Bullets, Organized political violence and Kenya’s crisis of governance”, 53 Ibid NORAD; see sub-section on the “Increasing concentration of powers 16th March, 2008. in the Executive”, pg.6. 57 See Commentary titled “Kenya: A country redeemed after a peaceful 54 The grievances related to access to and ownership of land in the past are election” by Mwangi Kimenyi, April 2013, online on www.brookings.edu. interlinked with political competition along ethnic lines and these have 58 See Part II on the Afrobarometer Survey; source: Afrobarometer Survey resulted in violent ethnic conflict in multi-ethnic areas. See Sub-section 2009: “Popular attitudes toward democracy in Kenya: A summary of 2.6 on Prospects and Risks regarding Kenya in ODI (2014). Afrobarometer indicators, 2003-2008. Published 6th June 2009. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 23 Kenya in Context decisions such as the formation of the judiciary and of distribute revenues to the county governments based the parliament.59 The strong provision for devolution on a weighted allocation (Table 1.3). in the new constitution was a “key source of public Table 1.3: First revenue-sharing formula among support for the draft of the constitution”.60 counties in Kenya Parameter Percentage weight Devolution of power was at the core of the new Population 45 constitution and has fundamentally changed Poverty index 20 the structure of government in Kenya. This major Land area 8 undertaking aimed to address deeply-entrenched Basic equal share 25 Fiscal responsibility 2 disparities between regions, allow for the regions Source: Brookings Institution (2013). to have greater autonomy, and rebalance power away from a historically strong central government. The formula for the horizontal sharing of revenues 61 The general elections of 2013 marked the official emphasizes fiscal need. The formula provides launch of the decentralization as the 47 newly historically marginalized counties with higher per formed counties elected their governors and county capita transfers than historically privileged counties assemblies, and a new national senate was established Land area and population are proxies for the costs of to represent the counties.62 service delivery. South Africa places a similar emphasis on fiscal need, taking a more sectoral approach, Devolved governance presents considerable however they accomplish this by directly measuring opportunities to Kenya in strengthening local the costs of service delivery in the education and autonomy over resource allocations. As per the health sectors. On the other hand, India’s approach to constitution, it was agreed that 84.5 percent of the revenue-sharing places an emphasis on fiscal capacity country’s revenues are to be allocated to the national as opposed to need. government while 15 percent will be allocated to the 47 county governments.63 The remaining 0.5 percent was The horizontal formula for revenue sharing has designated as an “equalization fund”. The Commission been highly equalizing, re-allocating revenues to on Revenue Allocation (CRA), created in the 2010 marginalized areas of the country. In particular, referendum, recommends the basis for equitable northern parts of the country have benefitted revenue allocation to the National Assembly, including significantly, with Turkana and Mandera receiving the percentage of national revenue to be divided higher benefits. Reallocation is envisaged to spur between the national and county governments as well growth in these areas and to contribute to improving as the distribution by county. This is not an easy task as living standards and regional economic convergence. any specific allocation criterion is bound to favor some The reallocation of revenues has also led to a decrease counties over others and therefore raise questions of revenues previously allocated to urban areas, about the legitimacy of the process. The National incentivizing these areas to improve on own-revenue Assembly accepted the CRA’s recommendation to collections by leveraging existing infrastructure. 59 See World Bank report titled Devolution without Disruption: Pathways to a successful new Kenya. November 2012. Continued disparities in capacities will shape 60 Ibid; Chapter One: Kenya’s devolution in context. both utilization of resource allocations and their 61 See Working Paper 1 on Kenya Devolution (Overview Note on building public participation in Kenya’s devolved government), February 2015, by ultimate impact. The generalized approach based the Center for Devolution Studies, Kenya School of Government. on an equitable allocation formula may work in 62 The country Executive arm is headed by the County Cabinet comprised of up to ten members known as the County Executive Committee principle, but the actual sector-wide utilization of (CEC). Each member of the CEC is in charge of a county department (a “ministry”). This apex body along with most administrative organs have resources depends largely on preexisting capacity already been created at the sub-counties, wards and village-level and at the county-level to effectively utilize the allocated counties recruit key personnel to staff the administrative units. 63 See Op-ed titled “Devolution and resource sharing in Kenya” by Mwangi funds. Differences in human resources, technical S. Kimenyi, on the Brookings Institution online, October 22, 2013. 24 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Kenya in Context abilities and existing infrastructure, among others, Devolving authority to county governments has greatly impact the actual cost of delivering specific given rise to new political dynamics that policymakers services under the management of the county need to address. The political decentralization process governments.64 Policymaking has to capture this vital in some cases resulted in hastily-drawn boundaries factor in resource allocation. which formed new administrative arrangements. Inter- county competition is growing over the ownership and National agencies resist handing over vital services control of national and regional development projects and functions to the counties given human that straddle county borders. This makes border regions resource challenges. The reluctance includes key more prone to violent disputes and reprisals against services such as the management of urban and rural minority residents from rival counties. High impact roads and rural electrification projects. This is due to policy interventions are needed to address disputes the limited administrative and technical capacity to between counties, particularly land claims, as well as handle these functions in certain counties. The central improved efforts towards ethnic inclusion at the county government also deployed County Commissioners, level governments. The latter is already under way on an even before the county governments were fully ad hoc basis in the form of a “County Inclusion Index” by established, who answer only to the Nairobi.65 Some the National Cohesion and Integration Commission.68 public sector agencies and their employees, such as doctors and teachers, are reluctant to be managed The 2017 general election renewed the focus on by local government units that are deemed less the presidency and put pressure on the electoral qualified than their national peers, even if the terms process and its governing institutions. The political and conditions of their services remain the same.66 decentralization achieved through the comprehensive devolution that Kenya has recently undertaken should Political and fiscal decentralization enjoys wide in theory mitigate the political stakes of the country’s political and popular support. There is now widespread presidential elections, among other accomplishments acceptance of – and big expectations (Box 1.3) (Box 1.3). The events of the presidential election in from – the devolution process. The demand for fiscal August 2017, however, demonstrate that this process autonomy is reflected in speed at which new county remains a contentious and ethnically polarizing governments have assumed major responsibilities event. This calls into question the effectiveness of and received greater funding in health, agriculture, new agencies formed under the 2010 referendum, and local roads/infrastructure.67 The share allocated to such as the Independent Electoral and Boundaries counties in 2013-14 was more than twice the minimum Commission (IEBC), which may not have exercised their 15 percent required by the Constitution. powers to the full extent possible.69 Box 1.3: Public expectations from devolution Citizens will get better public services: • Citizens will have better opportunities to participate in governance. • Women will have better opportunities in devolved governments. • Better transparency and accountability mechanisms will be put in place. • Minority communities will have better opportunities. • The process will lead to a more cohesive and peaceful nation. • Vices such as corruption and impunity will be minimized. Source: Based on Figure 4 “Kenyan’s [sic] expectations of Devolution”, Society for Development (2012 figures) in Center for Devolution Studies Working Paper 1 (2015). 64 See “Devolution and resource sharing in Kenya”, Op-Ed by Mwangi Kimenyi, 22nd October 2013; online at www.brookings.edu. 68 This body was created as part of the post-2007 elections’ reconciliatory efforts. A key objective now is to ensure that minorities within the 65 ODI (2014). counties are included in the governance structures and are marginalized Ibid. 66 in development efforts. 67 Center for Devolution Studies Working Paper 1 (2015). 69 ODI (2014). KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 25 Kenya in Context Box 1.4: Key features of the 2010 Kenyan Constitution The demands for constitutional reform in Kenya gathered pace in the 1990s. The impetus for these demands lay primarily within marginalized communities who objected to the centralized nature of power in the presidency. There is a widespread belief in politically disenfranchised communities that devolving powers away from the central government will end bias in resource allocation, among other gains. A referendum in 2005 failed to garner enough support for constitutional change, but a subsequent referendum in 2010 allowed for a groundbreaking redrafting of the constitution. This made way for the first change to the constitution since independence. Key features include: • The country’s first Bill of Rights that states the right of every citizen to basic services such as clean water, decent housing, sanitation and food. • A guarantee in principle to access to public resources irrespective of any community’s lack of influence at the national level. • A new, decentralized system of 47 local counties established that replaced eight provinces and 46 districts. Each county government consists of an Assembly and an Executive which are both directly elected by their constituents. • Dilution in the president’s appointive powers which are now subject to consultations with various commissions and require approval by the National Assembly. • The creation of an Upper House of Parliament called the Senate, where county governments have equal representation. • Establishment of the National Land Commission with powers to allocate land and to repossess illegally-acquired public land. This entity also restricts the ability of the President’s office to allocate public land to individuals and parties as done before. • Article 40 of the constitution sets out principles governing land policy while Article 68 directs the parliament to revise and rationalize existing land laws. Crucially, it stipulates that the manner in which land is converted from one category to another for acquisition must be regulated. • Chapter 11 establishes mechanisms for political and fiscal devolution and directives to allocate 15 percent of the public revenues towards the 47 counties annually. • Chapter 12 of the constitution establishes the Commission on Revenue Allocation to oversee an equitable resource- sharing between the center and the county governments. • A central government funding system that considers counties’ population size, area and poverty levels, and acknowledges that counties have autonomy over the design and details of local spending plans. Source: Online article titled “New constitution means major changes for Kenya”. Voice of America, August 11, 2010. Online at www.voanews.com; online article titled “How Kenya is changing under new constitution” Daily Nation online, Friday, August 28th 2015. Online at www.nation.co.ke; online Country Profile on Kenya and related article titled “Kenya’s new constitution brings political change”. Oxford Business Group, February 2017. Online at oxfordbusinessgroup.com/overview. The IEBC faced allegations of procedural and [therefore was] invalid”.72 The IEBC was observed inconsistencies and weak oversight for the 2017 to have clearly ignored electoral laws and procedures.73 elections.70 The commission had initially declared the An election re-run in October 2017 was boycotted by incumbent President Kenyatta the winner with over 54 the opposition, which demanded reforms to the IEBC.74 percent of the vote. The main challenger Raila Odinga 72 See Al Jazeera Opinion piece titled “Why did Kenya’s Supreme Court from the ODM within the larger National Super Alliance annul the elections?”, by Nanjala Nyabola, 2nd September 2017. Online at coalition challenged the results citing hacking and www.aljazeera.com. 73 Ibid.; the tallying website on which local and international reporting manipulation of the electronic vote-counting system.71 relied was not public as was earlier promised; IEBC conceded that they did not use an electronic transmission system to record ballots The Supreme Court nullified the results a month after and used text messages and photographs of manually filled forms as the elections and determined that the process “was sources of information; and, the forms used for reporting results from different regions were apparently not all available in time for the official not conducted in accordance with the Constitution announcement. The total cost of the elections at USD 500 million makes it one of the most expensive, spending USD 28 per capita in taxpayer money. 70 Article titled “What next in Kenya election crisis?”, by Dickens Olewe, 11th 74 This re-run was won by the incumbent with 98 percent of the votes October 2017. Online at www.bbc.com. while the turnout was recorded at 39 percent and the re-run suspended 71 Article titled “Kenyan opposition leader to challenge election result in 25 constituencies that were opposition strongholds. The Supreme in court”, Reuters/The Guardian, 16th August 2017. Online at www. Court upheld the results, which allows the President to serve another theguardian.com. five-year term. 26 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Kenya in Context 1.7 PERCEPTIONS ON DEMOCRACY, Figure 1.41: Perception of democracy in sub- Saharan African countries GOVERNANCE AND POLITICAL In your opinion, how much of a democracy is Kenya/other today? PARTICIPATION75 (% of respondents) K enyans show a strong preference for their democracy and democratic processes. Citizens largely support the nature of democracy in their country and have Zimbabwe Uganda 6 22 31 27 23 41 17 13 favorable attitudes towards processes linked to the Tanzania 2 11 42 17 functioning of a democratic republic, according to the South Africa 7 43 33 15 2016 Afrobarometer survey. Kenyans have a higher Nigeria 17 47 25 8 regard for their democratic system compared to other Kenya 5 21 48 15 countries in sub-Saharan Africa (Figure 1.41); 63 percent of Kenyans see their country as a “full democracy” or a 0 50 100 “democracy with minor problems.” Kenyans also have a Not a democracy Democracy, with major problems favorable view of the overall environment for electoral Democracy, with minor problems A full democracy Source: Afrobarometer Surveys’ “summary of Results” Kenya, Round 7, 2016, politics in the country. Question 35: “In your opinion how much of a democracy is Kenya today?”; under Question 35 for Uganda and Zimbabwe and Question 40 for Nigeria, South Africa and Tanzania. Remaining respondents in all countries were under the categories The majority of citizens responded positively to “Did not understand question/Democracy” or “Don’t know/refused”. These categories comprised 10 percent or less of total respondents in all countries except several fundamental democratic rights in place Tanzania (27 percent). and supported key features of a functioning democracy. In terms of the freedom of opposition processes. Afrobarometer surveys conducted in Kenya parties or candidates to speak or hold rallies and for in 2003, 2005 and 2008 show that 57 percent of Kenyans, the respondents to state their views or criticize the averaged across the three surveys, regarded their government, over 60 percent of respondents thought country as a “full democracy” with “minor problems”.76 that there was “somewhat more” or “much more” A majority of Kenyans – 68 percent (again, averaged freedom than before. Over 70 percent disapproved – from the three surveys) – also agreed that “many 50 percent “strongly” – of an election where only one political parties are needed to make sure that Kenyans political party is allowed to stand and hold office. A have real choices in who governs them”. Additionally, large majority, 83 percent, disapproved – 63 percent on average, over 88 percent of Kenyans in the surveys did so “strongly” – of the army governing the country rejected military rule as an alternative to electoral as an alternative. Democracy was “preferable to any politics. Democracy “was preferable to any other kind other kind of government” to 67 percent of Kenyans, an of government” for 80 percent of Kenyans in 2003, 75 opinion shared by respondents in other sub-Saharan percent in 2005, and for 79 percent in 2008. African countries: this statement is supported by 81 percent of Ugandans, 75 percent of Zimbabweans, 66 The 2008 survey shows ratings drop on the perceived percent of Nigerians, 64 percent of South Africans, and true extent of democracy, the satisfaction with 57 percent of Tanzanians. democracy, and the quality of the electoral process. Nearly 50 percent of citizens thought that Kenya Views before the devolution in 2010 show was “not a democracy or a democracy with major comparable support for democratic norms and problems”, a 19-point increase since 2005. 42 percent of Kenyans were “fairly satisfied or very satisfied” with 75 Data in this section is based on the latest Afrobarometer Survey’s “summary of Results”, undertaken in Kenya as Round 7 in 2016 the way democracy worked in Kenya, an 11-point (conducted September-October 2016) by the Institute for Development drop from 2005. Only 20 percent of Kenyans in 2008 Studies (IDS). Additionally, previous Summary of Results for Kenya from Round 6, 2014 and Round 5, 2011, and, the report “Popular attitudes claimed that the previous (2007) elections were largely toward Democracy in Kenya: A summary of Afrobarometer indicators, 2003-2008”. Data on sub-Saharan countries is based on Summary of Results from Nigeria, Round 6, 2015; South Africa, Round 6, 2015; 76 Source: Afrobarometer Survey report “Popular attitudes toward Tanzania, Round 6, 2014; Uganda, Round 7, 2017; and Zimbabwe, Round Democracy in Kenya: A summary of Afrobarometer indicators, 2003- 7, 2017. Online at www.afrobarometer.org. 2008”, 26th June 2009. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 27 Kenya in Context free and fair. The drop in positive perceptions from 2005 Kenyans in 2016 listed corruption as the “most to 2008 regarding elections was likely informed by the important problem facing the country” that should disputed 2007 elections and the following civil conflict. be addressed by the government. This was followed by unemployment, crime and security, and management Kenyans hold a neutral view of elected officials. of the economy.78 Concern over corruption has Citizens generally believe that the President, MPs, steadily risen for citizens since 2011 (Figure 1.43). A Members of County Assembly and the County majority of Kenyans stated that ordinary citizens were Governor are doing an acceptable job: in terms of how “very likely” to get away with paying a bribe or using key representatives had performed in their job over a personal connections for a) avoiding payment of year in 2016, 75 percent of Kenyans “strongly approved taxes that they owed to the government (66 percent), or approved” of the performance of the President, b) avoiding paying a traffic fine or going to court (70 45 percent did so of the MPs, and 47 percent of the percent), and c) registering land that did not belong to Members of the County Assemblies. them (73 percent).79 Moreover, 77 percent of Kenyans thought that those who report incidents of corruption The level of responsiveness from elected public “risked retaliation”.80 officials towards their constituents is a concern. When asked whether MPs tried their best to listen to what people have to say, 83 percent of Kenyans Responses around corruption also indicate notably thought that MPs “never did or did so only sometimes,” low levels of trust in public institutions. Some of these while 15 percent thought “often or always”. This is institutions are mandated with addressing corruption comparable to the perceived responsiveness to and redressing grievances, such as the police. Most constituents in other sub-Saharan African countries Kenyans reported some level of involvement in (Figure 1.42). The responsiveness of Members of County corruption by major government institutions (Figure Assemblies in Kenya was thought to be marginally 1.44). Additionally, when asked how well they thought better77 even as Kenyans gave a more balanced view of the current government was fighting corruption, over how they performed in 2016. 70 percent thought “very badly” or “fairly badly”. Figure 1.42: Responsiveness of National Assembly members to citizens in sub-Saharan African countries How much of the time do you think the following try their best to listen to what people like you have to say: Members of Parliament /National Assembly? (% of respondents) Zimbabwe 69 20 Uganda 75 22 Tanzania 87 11 South Africa 80 16 Nigeria 85 10 Kenya 83 15 0 20 40 60 80 100 Never/Only sometimes Often/Always Source: Afrobarometer Surveys’ “summary of results”: Under Kenya Round 7, 2016, Question 54A.; Nigeria Round 6, 2015, Question 59A.; South Africa Round 6, 2015, Question 59A.; Tanzania Round 6, 2014, Question 59A.; Uganda Round 7, 2017, Question 54B.; and Zimbabwe Round 7, 2017, Question 54A. 78 Question 55, Pt.1: In your opinion, what are the most important problems facing this country that government should address? (1st response). 79 Afrobarometer (2016); Questions 48D to 48F: respondents were to choose from a) Not at all likely; b) Not very likely; c) Somewhat likely; 77 Afrobarometer (2016); Question 54A “How much of the time do you think d) Very likely; additionally, there were categories of responses under the following try their best to listen to what people like you have to say? “Missing”, “Refused” (-to answer) and “Don’t know/Haven’t heard”. Members of Parliament.” and Question 54B “How much of the time do 80 Question 47: In this country, can ordinary people report incidents of you think the following try their best to listen to what people like you corruption without fear, or do they risk retaliation or other negative have to say? Members of County Assembly.” consequences if they speak out? 28 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Kenya in Context Figure 1.43: Major issues for citizens in Kenya that Figure 1.44: Perceived involvement in corruption, 2016 government should address (% of respondents) 30 Government workers 3 42 39 8 25 24 20 Police 3 25 37 28 20 19 % of respondents 15 15 13 Members of County Assembly 4 42 33 12 12 10 10 10 10 Members of Parliament 4 40 36 11 6 6 5 4 President's O ce 7 45 26 9 0 2011 2014 2016 0 50 100 Management of the economy Unemployment Percent Corruption Crime and Security None Some of them Most of them All of them Source: Afrobarometer survey, “summary of results”, First Response on Question Source: Based on Questions 44A to 44J (How many of the following people do “In your opinion, what are the important problems facing this country that you think are involved in corruption?), Afrobarometer Survey, “summary of results”, government should address?, Round 5, 2011, Round 6, 2014 and Round 7, 2016 Kenya 2016. Figure 1.45: Political intimidation or violence during Figure 1.46: Expressing political views in sub-Saharan election campaigns African countries During election campaigns in this country, how much do you personally How often do people in this country have to be careful of fear becoming a victim of political intimidation or violence? what they say about politics? (% of respondents) (% of respondents) Zimbabwe 23 76 2016 55 43 Uganda 34 65 Tanzania 45 51 2014 40 59 South Africa 56 44 Nigeria 35 64 2011 56 43 Kenya 22 74 0 50 100 0 20 40 60 80 100 A lot/somewhat A littlebit/not at all Never/Rarely Often/Always Source: Afrobarometer survey, “summary of results”, Round 5, 2011, Question 54; Source: Afrobarometer Survey “summary of results” Kenya Round 7, 2016, Question Round 6, 2014, Question 49; and Round 7, 2016, Question 40. 42A. In your opinion, how often, in this country: Do people have to be careful of what they say about politics? Question 42A. in Uganda and Zimbabwe surveys, Question 51A. in Nigeria, South Africa and Tanzania. Kenyans are also more cautious with respect and associating with political organizations (Figure to political participation. A large proportion of 1.46). The percentage of respondents indicating a respondents are concerned with intimidation or cautionary attitude towards associating with political violence during political campaigns in the country organizations has risen considerably over the past (Figure 1.45). A majority, 74 percent, also thought that decade. Attitudes in 2011 and 2014 indicate fewer they “often or always” had to be careful of what they inhibitions related to joining a political organization.81 say about politics. Citizens show inhibitions on other crucial dimensions of a participatory democracy as compared to other sub-Saharan African countries, 81 According to the Afrobarometer “summary of results” responses, 84 percent of Kenyans thought they were “somewhat free/completely free” seen in responses on expressing political views to join any political organization that they wanted to in 2014 (Question 15B.) and 82 percent thought the same in 2011 (Question 17b.). KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 29 CHAPTER 2 THE EXTENT AND EVOLUTION OF POVERTY AND INEQUALITY IN KENYA SUMMARY Kenya recorded steady progress against poverty between 2005/06 and 2015/16. The proportion of the population living beneath the national poverty line fell from 46.8 percent in 2005/06 to 36.1 percent in 2015/16. Most of the poverty decline is attributable to the progress observed in rural areas, where poverty declined from around 50 percent in 2005/06 to 38.8 percent ten years later. This contrasts with the stagnation of poverty in urban areas, particularly outside Nairobi. As Kenya urbanizes, cities are not providing enough economic opportunities for individuals to improve their income levels and maintain their standards of living. The country also experienced shared prosperity, with substantial consumption growth for households in the bottom 40 percent of the distribution. The annualized consumption growth for the bottom 40 percent has been a satisfactory 2.86 percent per year between 2005/06 and 2015/16, a pattern more pronounced in rural areas. Consistent with this pro-poor pattern of economic growth, inequality declined in Kenya, as confirmed by several inequality measures. While this helped to contribute to poverty reduction, most of the reduction is attributable to economic growth; which means that going forward efforts to reduce inequality can help accelerate poverty reduction. The evidence suggests that off-farm diversification has been important for poverty reduction in Kenya. While a robust agricultural sector and a dynamic services sector contributed to the wellbeing of rural households, most of the poverty reduction is accounted by households whose agricultural income is supplemented by non- agricultural activities (small-scale services). There is compelling evidence that the enabling factor was mobile money. M-PESA increased the households’ financial resilience and savings, allowing them to: i) invest productively, ii) move out of agriculture or complement that income with that of other businesses, and iii) improve their consumption levels. Kenya is characterized by stark regional differences, both in terms of monetary and non-monetary poverty indicators. The wellbeing of the population in the North & Northeastern Development Initiative (NEDI) counties (which includes all counties in the North Eastern province) lags considerably behind the rest of Kenya.82 Moreover, these areas have seen little progress between 2005/06 and 2015/16, remain prone to food insecurity, and present very low levels of educational attainment, access to improved sanitation and, to a lesser extent, access to improved water. While the GoK has implemented some measures to improve the connectivity and overall wellbeing of the population in these areas, substantive, sustained and cross-sectorial efforts will be required moving forward. Poor households remain limited by demographic characteristics, low human capital, and low coverage of basic services. Poverty is associated with female and older household heads, and low levels of educational attainment. This suggests that the poor are constrained when accessing income generating opportunities. Moreover, poor households tend to be larger, and have higher dependency ratios; demographic factors that usually hinder poverty reduction. In addition, coverage of WASH services and household electricity is much lower for poor households. In this sense, Kenya should continue to expand the coverage of this basic services to all segments of the population, while ensuring their quality. 82 NEDI group of counties: Mandera, Lamu, Wajir, Garissa, Tana River, Marsabit, Samburu, Turkana, West Pokot and Isiolo (a map displaying the NEDI counties is included in Appendix B). KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 31 The Extent and Evolution of Poverty and Inequality in Kenya This chapter first documents the progress made by the levels and progress of poverty indicators for the Kenya in terms of the monetary measures of poverty, NEDI counties. during the period on which this report focuses, 2005/16 to 2015/16. It analyzes the trends in terms of the national 2.1.1 Progress in the incidence of poverty poverty headcount rate, other related indicators (such Kenya has seen a steady reduction in the poverty as the depth and severity of poverty) and the incidence rate between 2005/06 and 2015/16 but progress of food and extreme poverty, as officially defined is modest. Over that period and consistent with by the KNBS. The chapter then turns to examine the the overall robust economic growth observed83, the incidence of consumption growth, and how this is country has been able to reduce the share of people reflected in terms of an array of inequality indicators. living below the national poverty line by more than ten It also examines the factors behind Kenya’s success in percentage points. The national poverty headcount reducing poverty, relying on decomposition analysis rate went down from 46.8 percent in 2005/06 to 36.1 and the finding of numerous studies on the impact of percent in 2015/16 (Table 2.1), which corresponds to mobile money in the wellbeing of the population. The an annualized rate of poverty reduction of 2.6 percent. chapter concludes by providing a profile of the poor, in Despite this successful reduction in the incidence of an attempt to identify the factors that may be limiting poverty, the absolute number of poor declined only their economic opportunities and overall wellbeing. marginally, from 16.6 million in 2005/06 to 16.4 million ten years later (Table 2.2). A first look at the absolute 2.1 STEADY BUT MODEST PROGRESS number of the poor in Kenya reveals that the number of AGAINST POVERTY 2005/06-2015/16 R Table 2.1: Absolute poverty headcount rate, nationally, by educing the share of the population living under the area of residence poverty line is an important measure of progress Percentage Annualized 2005/06 2015/16 for any country. This section analyzes how monetary point change change poverty has evolved in Kenya between 2005/06 and National 46.8 36.1 -10.7 -2.6 2015/16, looking closely at the spatial disparities both Rural 50.5 38.8 -11.7 -2.6 in terms of the urban and rural divide and of the marked Urban 32.1 29.4 -2.7 -0.9 Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16 provincial differences. It also pays special attention to Table 2.2: Poor and total populations, nationally, by area of residence and by NEDI classification Population living in poverty Annualized Distribution of poor (%) Percentage (Millions) percentage point change 2005/06 2015/06 change 2005/06 2015/06 National 16.6 16.4 -0.1 100 100 - Rural 14.3 12.6 -1.3 86.2 76.9 -9.3 Urban 2.3 3.8 5.1 13.8 23.1 9.3 Non-NEDI 14.3 13.2 -0.8 85.9 83.1 -2.8 NEDI 2.4 3.2 2.9 14.1 16.9 2.8 Total population (Millions) Annualized Distribution of poor (%) Percentage percentage 2005/06 2015/06 2005/06 2015/06 point change change National 35.5 45.4 2.5 100 100 - Rural 28.4 32.5 1.4 79.9 71.6 -8.3 Urban 7.2 12.9 6.0 20.1 28.4 8.3 Non-NEDI 32.1 39.9 2.2 90.3 88 -2.3 NEDI 3.4 5.4 4.7 9.7 12 2.3 Source: Own calculations based on KIHBS 2005/06 and 2015/16. 83 Except for the economic slow-down that resulted from the events that followed the general elections of 2007 and the slowdown of agricultural production in 2011, described in detail in Chapter 1. 32 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead The Extent and Evolution of Poverty and Inequality in Kenya people living below the poverty line increased in urban many other countries in Africa, where high total fertility and NEDI counties84, from 2.3 to 3.8 million and from 2.4 rates (TFRs) are undermining growth and poverty to 3.2 million respectively, whereas it decreased in rural reduction, as documented by Beegle and Christiaensen and non-NEDI counties. (forthcoming). In the case of Kenya, the average TFR is estimated at 3.9 children per woman in 2014 (Figure Fertility trends in Kenya have not undermined the 2.1), much lower than the 4.85 estimated for Sub- progress against poverty, as has been the case in Saharan Africa. This also means that fertility declined by many countries in Africa. While the small decline in almost one birth per women over the decade leading the number of poor may not appear as major progress, to 2014, a notable accomplishment. in this sense Kenya presents a better outlook than Figure 2.1: Total Fertility Rate (women aged 15-49) a) Kenya b) Benchmark countries most recent DHS year 6 6 5.4 5.4 5.2 5.2 4.9 5 5 4.6 4.6 4.5 4.2 4.2 3.9 4 3.9 Births per woman 4 Births per woman 3.3 3.1 2.9 3 3 2.6 2 2 1 1 0 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Kenya Ethiopia Ghana Rwanda South Tanzania Uganda (2014) (2016) (2014) (2015) Africa (2016) (2016) National Urban Rural (2016) Source: KDHS 2003, KDHS 2009 & KDHS 2014. Source: KDHS 2014, EDHS 2016, GDHS 2014, RDHS 2014-15, SADHS 2016, TDHS 2015-16, UDHS 2016. Box 2.1: Kenya Integrated Household Budget Survey (KIHBS): A commendable effort The analysis of this chapter and most of this report would not be possible without the recent effort by the KNBS to collect the 2015/16 wave of the KIHBS, which comes ten years after the collection of the first wave. Without this effort, it would not be possible to assess with certainty what are the living standards of the Kenyan population along many dimensions, including monetary and non-monetary poverty measures. While both waves are representative at the national, urban/rural, and provincial level, the 2005/06 KIHBS is also representative of Kenya’s 69 districts, and the 2015/16 KIHBS of the 47 counties introduced by the 2010 constitution.85 In addition to reporting statistics by urban and rural areas and by province, this chapter also refers to the NEDI group of counties. These are historically underdeveloped areas and, as will be shown throughout the chapter, lag behind the rest of the Kenya on a wide range of socio-economic indicators. The ten NEDI counties are Mandera, Lamu, Wajir, Garissa, Tana River, Marsabit, Samburu, Turkana, West Pokot and Isiolo (a map displaying the NEDI counties is included in Appendix B). 85 There are two additional differences in the sampling framework of the two waves. Firstly, the 2005/06 survey had 10 households per cluster and an additional 5 replacement households, whereas the 2015/16 KIHBS had 84 However, the number of the poor still grew at a slower pace than the the same number of households per cluster without any replacements. total population, which explains why the proportion of the poor did not Secondly, the 2015/16 KIHBS covered a larger sample: around 21,700 go up. households versus 13,100. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 33 The Extent and Evolution of Poverty and Inequality in Kenya While the reduction in poverty was more pronounced roughly one in ten poor lived in urban areas, by 2015/16 in rural areas, this is where three quarters of the this proportion was close to one in four (Table 2.2). This, poor still live. Poverty incidence in Kenya is still higher in addition to the increase of the absolute number of in rural areas than in urban areas, but it was in rural urban poor, indicates the economics benefits of the areas where the largest decline occurred. During progress observed at the national level are not reaching the ten-year period, rural poverty declined by nearly the poorest households in urban centers, particularly 12 percentage points from 50.5 percent in 2005/06 outside Nairobi, as explored in Chapter 5 of this report. to 38.8 in 2015/16. In contrast, there was little or no progress in urban areas: poverty declined by less Moreover, Kenya has been able to reduce the than 3 percentage points, but the difference is not incidence of food poverty and extreme poverty. statistically different from zero (Figure 2.2a). This Following the KNBS definitions, food poverty is translates into an annualized poverty decline that defined as the share of the population whose food is three times as large for rural Kenya (2.9 percent consumption is below the food poverty line, while versus 0.9 percent). This is explained by an increased extreme poverty is defined as proportion of the diversification of non-farm income sources of rural population whose total consumption (including households, particularly in the services sector, paired food, rent, clothing, energy, health expenditures, with a robust performance of the agricultural sector and education) is below the food poverty line. Both for the better part of the period studied. measures serve as an indication of food security at the household level, and how difficult is for households Poverty is increasingly becoming a concern for to fulfill the minimum caloric requirements. The share Kenya’s urban areas. The distribution of the poor of food-poor people has declined from 44.4 percent population between rural and urban areas changed in in 2005/06 to 32 percent in 2015/16 — a roughly 28 line with the distribution of the total population and percent decline, slightly steeper than the absolute the little progress made in urban areas. While in 2005/06 poverty reduction. Similarly, extreme poverty fell by Figure 2.2: Trends in absolute, food and extreme poverty, nationally and by area of residence a) Absolute poverty b) Food poverty c) Extreme poverty 60 80 60 50.5 50 46.8 50 Proportion of the population Proportion of the population 60 Proportion of the population 38.8 40 36.1 48.3 40 32.1 44.4 29.4 30 40 30 35.1 32.0 29.1 23.0 24.4 19.6 20 20 20 10.7 10 8.6 10 6.0 3.4 0 0 0 National Rural Urban National Rural Urban National Rural Urban 2005/06 2015/16 2005/06 2015/16 2005/06 2015/16 Source: Own calculations based on KIHBS 2005/06 and 2015/16. Note: Lines denote the 95% confidence interval for the statistic. 86 There are two additional differences in the sampling framework of the two waves. Firstly, the 2005/06 survey had 10 households per cluster and an additional 5 replacement households, whereas the 2015/16 KIHBS had the same number of households per cluster without any replacements. Secondly, the 2015/16 KIHBS covered a larger sample: around 21,700 households versus 13,100. 34 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead The Extent and Evolution of Poverty and Inequality in Kenya more than half: from 19.6 percent in 2005/06 to 8.6 the food poverty rate of urban areas as the difference percent in 2015/16 (Figure 2.2 c). In both cases, the between the two years is not statistically significant progress was mainly observed in rural areas. In the (Figure 2.2 b). case of food poverty, it seems there was no change in Box 2.2: Measuring poverty: Computing the poverty lines, the consumption aggregate and classification of peri-urban households Poverty lines The food and absolute poverty lines calculated with the 2015/16 KIHBS follow the Cost of Basic Need (CBN) method outlined in Ravallion (1998). The CBN method defines a consumption bundle required to meet one’s ”basic consumption needs.” The cost of this consumption bundle is then estimated using reference prices for either rural or urban areas. The rural and urban food poverty lines in each survey are determined using the cost of a food basket which meets the 2,250 kilocalorie requirement per adult equivalent. The rural and urban absolute poverty lines are then calculated by adding a minimum allowance for non-food consumption to their respective food poverty lines. While the same methodology had been used in 2005/06 to obtain the food poverty and absolute poverty lines, once the 2015/16 KIBHS was implemented it became evident the changes in the composition and in the relative importance of items within the food consumption basket would require a recalculation of the food poverty line (Figure 2.3). This is not surprising, as ten years later consumer preferences are different and there is larger choice set available to households. To obtain comparable estimates over time, the 2015/16 lines were deflated and revalued at 2005/06 prices. More specifically, the food poverty line is obtained using the 2015/16 basket of food items (and the weights within the basket) at their 2005/06 prices. The non-food component of the line is deflated using the official CPI. Figure 2.3: Urban and rural food poverty basket comparison by rank, 2005/06 and 2015/16 2005/06 2015/16 2005/06 2015/16 1 1 1 1 Beef with bones Beef with bones 2 2 2 2 Sugar 3 Sugar 3 4 4 Fresh cow milk 4 4 5 Fresh cow milk 5 (Unpacketed) (Unpacketed) 6 6 6 6 Bread Bread 7 7 7 7 Sifted maize our 8 Sifted maize our 8 9 9 9 9 Fresh cow milk Fresh cow milk (Packeted) 10 (Packeted) 10 Tomatoes Tomatoes Non-aromatic white rice 13 Non-aromatic white rice 13 Loose maize grain 14 14 Loose maize grain Kale (sukuma wiki) Kale (sukuma wiki) 17 17 18 18 Source: KNBS 2018 There is a minimal difference at the national level between the 2005/06 poverty rates resulting from the noncomparable lines (the 2005/06 poverty line) and comparable lines (using the 2015/16 basket of food items at 2005/06 prices). The absolute and extreme poverty rates calculated using comparable poverty lines are just 0.2 and 0.1 percentage points higher, respectively, than when calculated using the original 2005/06 poverty lines (Table 2.3). Nationally, food poverty is 1.4 percentage points lower due to the drop in the urban food poverty line, which also results in a reduction in the urban extreme poverty rate from 8.3 percent to 6.0 percent. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 35 The Extent and Evolution of Poverty and Inequality in Kenya Table 2.3: Comparison of noncomparable and comparable 2005/06 poverty rates 2005/06 2005/06 Extreme poverty rate (%) (Noncomparable) National 46.6 46.8 Absolute poverty rate (%) Rural 49.7 50.5 Urban 34.4 32.1 National 45.8 44.4 Food poverty rate (%) Rural 47.2 48.3 Urban 40.4 29.1 National 19.5 19.6 Extreme poverty rate (%) Rural 22.3 23.0 Urban 8.3 6.0 Source: Own calculations based on KIHBS 2005/06 and 2015/16. Consumption aggregate The consumption aggregate in both surveys was constructed using the approach outlined in Deaton & Zaidi (2002). The food aggregate uses a recall period of 7 days and comprises food consumption from four sources, namely: purchases, own production, own stock and gifts. Prices were imputed using the cluster-level median for each item since a household may have consumed but not purchased an item and household-level prices may contain outliers. The non-food component of the aggregate includes consumption of energy, education, transport and clothing among other item groups. Housing rent is also included in the non-food component, however only for urban households, wherein the rent is imputed for households that own their dwelling. Over-the-counter medication (items such as cough syrup, painkillers and anti-malaria medicine) is the only form of health expenditure included the non-food aggregate. Lastly, in each survey in order to account for spatial and temporal food price differences, a household-level price deflator based on a Paasche price index was created. Spatial adjustment occurs as the cluster median prices are referenced to the overall rural or urban median prices. Temporal adjustment occurs as each cluster is surveyed in a 2-week period within a year and these prices are then referenced to the median price for the entire survey period. This adjusts for differences in the cost-of-living within urban and rural areas after it is applied to the nominal food and total aggregates. Peri-urban classification Peri-urban households were classified as rural households in the 2005/06 KIHBS survey for the purpose of generating a consumption basket used to create the food poverty lines as well as for the spatial price deflator and the calculation of poverty rates. However, after Kenya’s 2009 Population Census, the KNBS established that the urban category should include peri-urban households. For this report, and after a careful analysis of the characteristics of the peri-urban households in the KIHBS 2015/16, we classify peri-urban households as being rural (as in the 2005/06 KIBHS). As seen in the Appendix B, the socio- economic conditions of these households are closer to their rural counterparts than their core urban counterparts. Thus, using the urban poverty line to identify if these households are poor would not be appropriate and would result in an underestimation of the welfare of these households. 36 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead The Extent and Evolution of Poverty and Inequality in Kenya 2.1.2 Regional patterns in poverty and poverty fact that the economic progress observed during this reduction period is not reaching all areas of the country, and it While poverty fell in every province, there are large validates the recent effort of the government to invest spatial differences in the poverty levels and changes these regions. In addition, female headed households across the different provinces of Kenya. Figure 2.4a in NEDI counties exhibit higher poverty rates (absolute, shows the striking provincial variation in the poverty food and extreme poverty) than in the rest of Kenya. incidence across the different provinces of Kenya: while 70 percent of the population in the North Eastern Food and extreme poverty are highly heterogeneous Province live in poverty, that is true for only 16.7 of the across different provinces. While Nairobi enjoys a population in Nairobi. Moreover, this former province food poverty rate that is close to being only half of the barely saw any progress over the period of focus of this national average (16.1 percent, down from 20 percent study, with poverty declining from 74 to 70 percent in 2005/06) and it has almost eliminated extreme between 2005/16; representing the lowest annual poverty, the North Eastern Province performs drastically reduction rate for all provinces (around 0.6 percent per worse with half of the population being food poor and year). On the contrary, the Eastern and Coast provinces one in four in extreme poverty. Interestingly, these exhibited the largest reductions in the incidence of two extreme cases (the worst- and best- performing absolute poverty (of around 18.8 and 17.1 percentage provinces) have the lowest rates of progress in the points), with annual reduction rates of 4.5 and 3.5 country (Figure 2.4b and c). percent respectively. These two provinces account for around 43 percent of the poverty decline in the country. It is clear that the NEDI counties are prone to food insecurity. Food poverty and, particularly, extreme Poverty incidence in the NEDI counties is significantly poverty, are remarkably high in NEDI counties when higher than in the rest of the country. Remarkably, the compared to the rest of the country. For 55.4 percent poverty rate amongst the NEDI counties in 2015/16 is of the population in these counties food expenditure more than double that of the rest of the country, 68.0 is not sufficient to reach the minimum caloric percent versus 32.6 percent (Figure 2.4a). Moreover, requirement (compared to 29.5 percent for the non- progress has been slow: while the non-NEDI poverty NEDI counties Figure 2.4b). Also, as shown in Figure 2.4c headcount rate fell by 3 percent annually, it only fell for 31.8 percent of the population, even if they devoted by 1.1 percent in the NEDI counties. This reflects the their entire budget into food, this would still not suffice Figure 2.4: Trends in absolute, food and extreme poverty by province and NEDI classification a) Absolute poverty 74.0 76.2 80 70.0 68.0 57.6 Proportion of the population 54.2 60 50.6 49.0 47.2 40.5 41.4 42.5 44.2 36.7 40 31.8 31.1 32.6 24.3 21.3 16.7 20 0 Coast North Eastern Central Rift Valley Western Nyanza Nairobi Non-NEDI NEDI Eastern 2005/06 2015/16 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 37 The Extent and Evolution of Poverty and Inequality in Kenya b) Food poverty 80 65.2 68 52.7 51.5 55.4 51.4 Proportion of the population 60 47.9 45.3 45.4 37.9 42.3 38.8 32.1 35.8 32.9 40 28.5 29.5 22.4 20 16.1 20 0 Coast North Eastern Central Rift Valley Western Nyanza Nairobi Non-NEDI NEDI Eastern 2005/06 2015/16 c) Extreme poverty 80 Proportion of the population 44.6 49.6 60 40 26.7 31.8 24.0 22.7 23.3 19.6 20.2 16.9 20 12.1 13.6 11.1 6.0 6.6 6.1 6.1 2.9 2.8 0.6 0 Coast North Eastern Central Rift Valley Western Nyanza Nairobi Non-NEDI NEDI Eastern 2005/06 2015/16 Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16. (compared to 6.1 percent for non-NEDI). Moreover, the decline in the Eastern province, which as mentioned progress in these counties has been slower than in had a stellar performance in terms of poverty reduction. the rest of the counties. Not being able to attain the Looking at the distribution is worth noting that, given nutrition requirements has severe consequences on its high poverty incidence, the North Eastern province health, productivity and the accumulation of human concentrates a higher share of the poor (close to 7 capital among children, which results in poverty traps percent) compared to the share of the total population that are difficult to overcome. (3.5 percent). The majority of the poor reside in the Rift Valley, It is clear that the national poverty estimates mask followed by the Nyanza and the Western province. stark spatial disparities across the different regions. One third of the poor population resides in the Rift Historically, provincial disparities have been marked in Valley, the most populated province of Kenya, followed Kenya, partly explained by climatic and agro-ecological by Nyanza, accounting for 15 percent of the poor, and differences that affect agricultural productivity, partly the Western province, with 12.7 percent. Overall, as by differences in infrastructure and access to public seen in Figure 2.5, the distribution has not changed services (as will be shown later in the chapter), and much in the past ten years, except for a substantial partly by the differences in political representation 38 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead The Extent and Evolution of Poverty and Inequality in Kenya Figure 2.5: Distribution of the poor by province 2005/06 2015/16 3.6% 4.5% 11.4% 10.4% 14% 14.2% 4.9% Coast 6.8% North Eastern Eastern 11.9% Central 14.5% 17.7% 12.7% Rift Valley Western 7.5% Nyanza Nairobi 8.2% 25.8% 32% Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16. and participation in the decision-making process as poverty line. During the period of focus of this study, discussed in Chapter 1 (World Bank 2008). Making inequality amongst the poor declined nearly by half sure that all regions are part of the economic from 8.2 to 4.5 percent. development process and benefit from it will be an important part of sustaining the poverty reduction As with the poverty headcount rate, urban areas effort moving forward. saw less progress in terms of the depth and severity of poverty. Analyzing the poverty gap and poverty 2.1.3 Poverty depth and severity: How far are the severity for the urban and rural population, once again poor below the poverty line and how much inequality amongst the poor is there? it is observed that the decline is steeper amongst rural households. The gap went down from 18.2 to 11.0 Both the depth and severity of poverty have declined percent over the last ten years, while in urban areas the in Kenya. The depth of poverty is represented by how far, on average, the poor fall below the poverty line, and decline was only 1.7 percentage points from 10.6 to 8.9 is expressed as a percentage of the poverty line value. percent. Similarly, rural severity halved from 9.2 percent This is also known as the poverty gap and serves to in 2005/06 to 4.7 in 2015/16, a level similar to that measure the intensity of poverty in a given population. observed in urban areas (Figure 2.6). In short, in terms Between 2005/06 and 2015/16, this measure fell from of how far the poor are below the poverty line and 16.7 percent to 10.4 percent for Kenya as a whole (Figure how much inequality exists amongst the poor, rural 2.6). In other words, if transfers could be perfectly and urban households currently look quite similar. The targeted, it would take a transfer of roughly 10.4 percent same cannot be said of NEDI and non-NEDI counties, (KSh 407) of the poverty line to each poor individual to where a striking contrast arises. Poverty depth in NEDI eradicate poverty. Another alternative indicator is the countries is a staggering 28.7 percent in 2015/16 (Figure poverty gap squared – or severity of poverty – which 2.7), significantly higher than in non-NEDI counties, describes inequality amongst the poor by placing a meaning that the effort needed to lift households out greater weight on individuals who are further from the of poverty in these areas will be considerable. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 39 The Extent and Evolution of Poverty and Inequality in Kenya Figure 2.6: Poverty depth and severity, nationally and by urban/rural strata 20 18.2 16.7 15 11.0 10.6 10.4 10 9.2 8.9 8.2 4.5 4.7 5.0 5 3.9 0 2005/06 2015/16 2005/06 2015/16 2005/06 2015/16 2005/06 2015/16 2005/06 2015/16 2005/06 2015/16 Depth Severity Depth Severity Depth Severity National Rural Urban Source: Own calculations based on KIHBS 2005/06 and 2015/16. Figure 2.7: Poverty depth by province and NEDI classification 80 Proportion of the poverty line 60 40 38.6 33.6 28.7 25.7 20.8 20 18.1 17.7 18.9 16.7 14.8 14.0 12.0 10.7 10.2 8.5 9.0 8.4 5.5 6.4 3.4 0 Coast North Eastern Central Rift Western Nyanza Nairobi Non- NEDI Eastern Valley NEDI 2005/06 2015/16 Source: Own calculations based on KIHBS 2005/06 and 2015/16. 2.1.4 Consumption patterns rural households than urban households. Nonetheless, The share of consumption spent on food has consistent with a lower level of wellbeing, rural increased for households across Kenya. Despite the households allocate more of their consumption to food reduction in poverty, the average share of consumption than urban households. devoted to food has risen by 3.3 percentage points, from 51 in 2005/6 to 54.3 in 2015/16 percent nationally Share of consumption on rent (mainly for urban (Figure 2.8). A contributing factor to this phenomenon households), education and energy increased is that food prices increased at a faster rate compared marginally. The share of consumption spent on rent to non-food prices during that period. As depicted in for urban households87 also increased slightly – from Figure 2.9, while the cumulative inflation (based on 14.1 to 15.1 percent (Figure 2.8). While the increase is the overall CPI) over this period was 134 percent, food not alarming, the housing deficit in urban Kenya is well inflation was significantly higher at 219 percent. The documented, and for the majority of poor households relative increase in food prices likely benefited net-food the housing conditions in which they live, and the producer households and hurt urban households in service accessibility do not correspond to the prices the lower part of the distribution (as will be explored paid (World Bank 2018b). in Chapters 5 and 6 of this report, respectively), which 87 As determined by the KNBS the consumption aggregate for rural helps to explain why poverty declined faster among households does not include rent. 40 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead The Extent and Evolution of Poverty and Inequality in Kenya Figure 2.8: Proportion of consumption by use, nationally and area of residence National 2005/06 Rural 2005/06 Urban 2005/06 30.4% 26.7% 34.3% 39.4% 51.0% 4.9% 5.3% 61.8% 6.6% 5.6% 6.6% 6.9% 6.5% 14.1% National 2015/06 Rural 2015/06 Urban 2015/06 24.3% 22.3% 25.9% 46.6% 7.3% 6.0% 5.0% 54.3% 6.6% 63.8% 7.0% 7.4% 8.4% 15.1% Food Rent Education Energy Others Source: Own calculations based on KIHBS 2005/06 and 2015/16. Figure 2.9: Differential changes in price indices 350 2 319.2 290.0 1.9 300 260.3 1.8 CPI Index (base year (2005 = 100) 239.5 250 Food / overall price ratio 223.3 1.7 203.0 233.8 1.6 200 219.9 168.5 206.3 159.1 193.1 1.5 137.8 182.6 150 167.0 1.4 107.5 112.0 140.7 146.4 100.0 128.5 1.3 100 110.6 106.0 100.0 1.2 50 1.1 0 1 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Food Overall Food / Overall price ratio Source: Own calculations based on KNBS 2017. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 41 The Extent and Evolution of Poverty and Inequality in Kenya 2.2 THE INCIDENCE OF PROGRESS, SHARED 2.2.1 Incidence of progress PROSPERITY AND INEQUALITY Overall, households in the bottom of the distribution, W hile poverty is an important measure of how Kenyan particularly the bottom 20 percent, have experienced living standards have improved, understanding substantial growth in real consumption over the if economic progress has reached all segments of the last ten years. Growth incidence curves (GICs), which population and how the distribution of consumption display annualized consumption growth over the entire has changed over time is also important. This section distribution of the population, reveal that economic takes a closer look at which parts of the distribution have growth in Kenya has been pro-poor from 2005/06 to benefitted the most from economic progress experienced 2015/16 (Figure 2.10a). The lower tail of the distribution, by the country between 2005/06 and 2015/16, focuses on particularly below the 20th percentile, experienced the consumption growth if the bottom 40 percent88 and annualized growth rates of around 3-4 percent. These analyzes changes in consumption distribution in rural and growth rates decline monotonically towards the upper urban areas. tail of the distribution, reaching 2.86 percent at the Figure 2.10: GICs nationally, by area of residence and NEDI classification a) National b) Rural 5 Annualized % change in real consumption Annualized % change in real consumption 4 4 3 3 2 2 1 1 0 0 -1 0 20 40 60 80 100 0 20 40 60 80 100 Share of population ranked, percent Share of population ranked, percent c) Urban d) Non-NEDI Annualized % change in real consumption Annualized % change in real consumption 4 4 3 3 2 2 1 1 0 0 -1 -1 0 20 40 60 80 100 0 20 40 60 80 100 Share of population ranked, percent Share of population ranked, percent e) NEDI 8 Annualized % change in real consumption 7 6 5 4 3 2 1 0 0 20 40 60 80 100 Share of population ranked, percent Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16. 88 This group is the focus of the World Bank’s Group goal of shared prosperity. 42 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead The Extent and Evolution of Poverty and Inequality in Kenya 40th percentile and 2 percent for the 70th percentile of when looking at the 40th percentile of the distribution the population. The rates become negative at the very (Figure 2.10b). Pro-poor consumption growth is also top of the distribution, but this might be related to the observed in urban areas, although the growth rates fact that the 2015/16 KIBHS suffered from very high are less spectacular when compared to their rural nonresponse rates in households at the top of the counterparts. Given that the nonresponse problems of distribution in Nairobi, as explained in detail in Box the 2015/16 KIBHS mainly affected households at the 2.3. Given this nonresponse issue, the data are likely top quintile of the consumption distribution (See Box underestimating the consumption levels and thus 2.3), the conclusion that economic growth benefitted the growth rates for the top two deciles in Nairobi the bottom of the distribution in urban settings still (see Figure 2.11a). However, this issue does not remains true. affect the bottom part of the distribution and given the rather steep decline of the GIC up to the 80th In NEDI counties, households at the lower end of percentile, it is clear that economic progress over the distribution also experienced a much higher the past ten years has benefitted the poor, and even consumption growth. Looking at GIC for NEDI counties among the poor, it has disproportionally benefitted separately, it is worth mentioning that households in the poorest of the poor. the bottom of the distribution experienced substantial annualized real consumption growth. Growth for the While consumption growth in rural areas was higher 10th percentile was close to 8 percent while at the 40th for the poor, consistent with the impressive decline percentile, it was around 3.5 percent. Nonetheless this of poverty incidence, all households along the was not translated into a substantial poverty decline, distribution experienced consumption growth since as expected, given how far below the national poverty 2005/06. Despite varying performance, no percentile line are the poor in these counties. Moreover, it is only in rural areas experienced negative real consumption for these counties that we do not observe a decline growth, and the average annualized change is roughly in real growth at the very top of the distribution, and 1.5 percent p.a. for rural households. The highest growth consumption growth for households at the very top rates took place for the poorest ten percent of the was above the average (which is represented by the population at around 4 percent, while this rate halves horizontal red line in Figure 2.10e). Figure 2.11: Real consumption deciles (2016 prices), nationally and by area of residence a) National b) Urban c) Rural 10 10 10 9 9 9 8 8 8 7 7 7 Declie Decile Decile 6 6 6 5 5 5 4 4 4 3 3 3 2 2 2 1 1 1 0 5,000 10,000 15,000 20,000 25,000 30,000 35,000 0 15,000 30,000 45,000 60,000 0 5,000 10,000 15,000 20,000 Mean consumption per decile (2016 KShs) Mean consumption per decile (2016 KShs) Mean consumption per decile (2016 KShs) 2015/16 2005/06 2015/16 2005/06 2015/16 2005/06 Source: Own calculations based on KIHBS 2005/06 and 2015/16. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 43 The Extent and Evolution of Poverty and Inequality in Kenya Box 2.3: Nairobi nonresponse rates – dealing with data issues The KIHBS 2015/16 survey had an irregularly elevated level of nonresponse among households in Nairobi: only 3 out of 4 households (76.9 percent) in the capital successfully completed the questionnaire, whereas the response rate was between 81.9 and 96.5 percent in the rest of the country (see Appendix B)89. The high nonresponse rate (both at the item and household level), coupled with the non-replacement of unsuccessful interviewed households, likely caused the survey to not accurately capture the upper end of the consumption distribution. Survey response probabilities usually fall with rising incomes/consumption and if this is not adequately addressed in the sampling strategy, reported mean consumption and inequality measures are likely to be underestimated. Fortunately, the nonresponse can generally be expected to leave all poverty measures widely unaffected (Korinek, Mistiaen, and Ravallion 2006). Figure 2.12 below shows the response rate and median consumption by county for all urban households, where each scatter point is weighted by the proportion of total urban households the county represents. The linear trend line shows that in counties with higher median consumption, response rates tended to be lower and it is expected that the same occurs at the household level. Thus, most likely, the nearly 25 percent nonresponse rate in Nairobi was concentrated among wealthier households. Detailed analysis of asset ownership patterns by consumption quintile provides further evidence for the hypothesis that the missing data stems disproportionally from the upper tail of the distribution (Appendix B). For all of considered assets (house, fridge, sofa, car and washing machine) ownership falls dramatically between the 2005/06 and 2015/16 surveys within the top quintile (and in some instances, within the top two quintiles), which is unlikely to occur. According to the data, house ownership fell from 21.4 percent in 2005/06 to 8.8 percent in 2015/16 in the top quintile and car ownership declined 36.8 to 22.7 percent (Appendix B). Unfortunately, it is then likely that the consumption level for the top two deciles in Nairobi is underestimated. Thus, the staggering decline of almost 60 percent in the real consumption of the 10th decile (wealthiest ten percent of the population of Nairobi), as well as the 10 percent decline for the 9th decile is likely overstated. As mentioned, while the poverty estimates are likely unaffected, this does affect the inequality estimates. For that reason, the national and urban inequality measures most likely will overestimate the reduction in inequality, despite the fact that inequality did decline over the period of interest, as consumption growth was more prominent among the poorest households both in rural and urban areas. Household survey data in emerging countries is widely Figure 2.12: Response rates and consumption among known to underestimate top levels consumption urban households and inequality (Assouad, Chancel, and Morgan 2018). 12000 Nairobi Embu Nonresponse, both item and household nonresponse, is Tharaka Nithi Median consumption per county (2016 Kshs) Kirinyaga a crucial factor contributing to this challenge (Medeiros, Meru Nyeri Mombasa 10000 de Castro, and de Azevedo 2016). In countries like Brazil, Kitui Machakos Makueni Baringo India and South Africa, tax records have been used to i) Nakuru Kiambu Lamu Nyamira Trans Nzoia Kericho Kwale Laikipia verify that household survey data was indeed not properly UAsin Gishu Samburu Kisumu 8000 Migori capturing the income and consumption levels of the West Pokot Elgeyo Marakwet Kakamega Nyahururu top part of the distribution, and more importantly, ii) to Busia Homa Bay estimate more accurate inequality estimators through a 6000 Marsabit Garissa Vihiga Tana River combination of imputation and reweighting techniques. It Mandera Wajir is important to further study if similar techniques could be 4000 implemented in the case of Nairobi and Kenya in general, Turkana 70 75 80 85 90 95 in order to obtain a more accurate measure of inequality Response rate (%) - urban in the country. Source: Own calculations based on KIHBS 2015/16. 89 Unfortunately, these data are not available for the 2005/06 KIBHS. 44 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead The Extent and Evolution of Poverty and Inequality in Kenya 2.2.2 Shared prosperity Eastern and Coast provinces – also saw the greatest Kenya is making satisfactory progress in fulfilling the increases in consumption amongst the bottom 40 shared prosperity goal: promoting the consumption percent. The respective annualized rates at 4 and 4.5 growth of the bottom 40 percent of the distribution. percent were above the national average of 2.86 for The annualized growth rate of Kenya’s bottom the period 2005/06 to 2015/16. Nairobi, an entirely 40 percent of the population was 2.86 percent urban province, saw the lowest consumption growth for the period between 2005/06 and 2015/16. for the bottom 40 percent, at an annualized growth Consistent with the GICs shown in the previous rate of 1.3 (Figure 2.13) (this is not at all affected by the section, consumption growth amongst the rural nonresponse issue). Growth and economic progress in bottom 40 percent was 2.5 times higher than for the Kenya was less broad-based in the urban areas, which urban counterpart (2.4 percent versus 0.9 percent). might help to reduce the urban-rural gap but is not Diversification of income sources off-farm, together consistent with an outlook in which cities are centers of with high food prices during this period, benefitted progress for everyone. rural households more than urban households in this distribution bracket. The resulting rural shared While Kenya’s shared prosperity growth indicator is prosperity premium (calculated as the difference low when compared to other sub-Saharan African between the growth rate of the bottom 40 countries over comparable periods, economic percent and the average growth rate for the whole progress has been concentrated in this lower distribution) is estimated at around 1 percentage segment of the distribution. While the annualized point (Figure 2.13) This number is likely to be close to real consumption growth for the bottom 40 percent the shared prosperity premium for Kenya as a whole in Kenya was 2.9 between 2005/16 and 2015/16, most over this period.90 countries in the region have experienced higher growth amongst households in this segment. It reached 4.6 In all provinces in Kenya real consumption growth for percent for Rwanda between 2005 and 2010, 3.51 in the bottom 40 percent was positive and higher than Uganda over the period 2005 to 2012 in Uganda and for the top of the distribution. However, there were an astonishing 9.76 percent for Tanzania between 2007 marked differences across provinces. Those provinces and 2011 (Figure 2.14). However, economic growth in that saw the largest reduction poverty – mainly the Kenya has been markedly pro-poor, and the estimated Figure 2.13: Annualized consumption growth, nationally, by area of residence and by province 5 4.5 4.0 Annualized % change in mean consumption 4 3 2.9 2.8 2.9 2.6 2.6 2.4 2.4 2.2 2 1.9 1.8 1.5 1.7 1.6 1.2 1.1 1.2 1.3 1.3 0.9 1.0 1.0 1.0 1.1 1.1 1 0.8 0 National Urban Rural Coast North Eastern Central Rift Western Nyanza Nairobi -1 Eastern Valley -2 -3 Bottom 40% Total pop. Premium Source: Own calculations based on KIHBS 2005/06 and 2015/16. Note: due to the non-response issues in Nairobi, not all categories report all three indicators. 90 If taking the average consumption growth of 1.1 at face value, the shared prosperity premium is 1.8. However, given that the consumption growth of the top deciles is underestimated (because of the nonresponse rates in Nairobi), the true premium should be much lower than that. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 45 The Extent and Evolution of Poverty and Inequality in Kenya Figure 2.14: Annualized consumption growth compared to benchmark countries 12 Annualized % change in mean consumption 9.8 10 9.1 8 6 4.6 4.3 4.4 4 3.4 3.6 3.5 2.9 1.8 2 1.1 1.2 1.4 0.7 0.7 0 -0.4 -0.9 -2 -1.8 -4 Kenya Rwanda South Africa Uganda Tanzania Ethiopia (2006 -2016) (2005 - 2010) (2005 - 2010) (2005 - 2012) (2007 -2011) (2005 - 2010) Bottom 40% Total pop. Premium Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16 shared prosperity premium (again defined as the improvement for an indicator that is usually very stable difference between the growth rate of the bottom 40 over time. This suggests that redistribution contributed percent and the average growth rate for the whole positively to the substantial poverty reduction distribution) of one percentage point is higher than observed in Kenya’s rural areas during this period. In in all benchmark countries except for Rwanda (with a terms of provincial heterogeneity, inequality declined premium of 1.2 percentage points). faster in the Coast province (from 0.43 to 0.38), in the Central region (from 0.38 to 0.34), and to a lesser extent 2.2.3 Inequality indicators in the North Eastern and Rift Valley provinces (Figure 2.15). The level of inequality in Kenya as measured by Inequality in Kenya declined between 2005/06 and the GINI index is moderate and comparable to that of 2015/16, as confirmed by different measures. The nonresponse problem identified for Nairobi does Tanzania, Uganda and Ghana, and is much lower than affect the precision of some of the measures of South Africa’s index of 0.63 (Figure 2.16). inequality at the urban level using the KIBHS 2015/16, and thus, at the national level. However, the collection Alternative measures of inequality confirm an of the evidence presented in this section indicates improvement in Kenya’s distribution. The Atkinson that inequality in Kenya has declined at the national index, which at high levels of the inequality aversion level since 2005/06, in line with the pro-poor pattern parameter gives more weight to the lower consumption of economic growth described by the incidence levels making the measure less sensitive to issues at the curves of Section 2.2.1 and contributing to the poverty top of the distribution, confirms that the Consumption reduction observed. distribution has improved. At an inequality aversion parameter (∑=1), the Atkinson index declined from 0.3 The decline in the Gini index indicates an in 2005/16 to 0.23 in 2015/16 (Figure 2.17b). This means improvement in the distribution of resources in that in 2015/16, Kenya should be willing to forgo Kenya. The Gini index, which is generally not heavily 23 percent of its consumption to achieve a uniform affected by the upper tail of the distribution (Cowell consumption distribution. Another measure that is not and Flachaire 2002), fell from 0.45 in 2005/06 to 0.39 affected as much by the nonresponse issue is the ratio in 2015/16, indicating that Kenya made considerable of consumption at the 75th and 25th percentile. Under progress in terms of reducing inequality (Figure 2.15). The this measure inequality also declined, albeit the drop Gini index in rural areas (unaffected by the nonresponse is less pronounced and the levels and changes in rural issue) declined from 0.37 to 0.33, a significant and urban areas resemble each other. The consumption 46 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead The Extent and Evolution of Poverty and Inequality in Kenya Figure 2.15: Gini inequality index nationally, by area of residence and by province 0.6 0.45 0.46 0.44 0.43 0.42 0.39 0.38 0.38 0.38 0.38 0.38 0.4 0.37 0.37 0.35 0.35 0.34 0.35 0.34 0.33 0.33 0.33 Gini Index 0.31 0.2 0.0 National Urban Rural Coast North Eastern Central Rift Western Nyanza Nairobi Eastern Valley 2005/06 2015/16 Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16 Figure 2.16: Gini inequality index for select African countries 0.8 0.63 0.6 0.50 0.41 0.42 Gini index 0.39 0.38 0.4 0.33 0.2 0.0 Kenya Rwanda South Africa Uganda Tanzania Ethiopia Ghana (2015/16) (2013) (2011) (2012) (2011) (2010) (2012) Source: World Bank Poverty & Equity Databank. level of the 75th percentile went from being 2.7 times in the past ten years (Table 2.4). Moreover, in 2015/16, higher than that of the 10th percentile in 2005/06 to 2.5 differences across rural and urban households help times higher in 2015/16 (Figure 2.17a). explain about one fourth of the overall inequality. Thus, about three quarters of the inequality can be attributed Inequality in Kenya is primarily explained by to differences within rural and urban households. differences within urban and rural areas (and within Interestingly, these fractions have remained constant provinces), rather than by differences between these over time. Nonetheless, as will be seen later on, the groups. Analysis of the Theil index allows for a better urban-rural divide in non-monetary living conditions understanding of the nature of inequality and how it has and access to services is large, with rural areas lagging changed over time. More specifically, it helps determine behind in access to WASH services and electricity in how much of the inequality in the country is rooted particular. The analysis of the contribution to inequality within particularly groups and how much is attributed from differences within and across provinces produces to differences between these groups. Consistent with comparable results, and it is mostly inequality within all measures described so far, under the Theil index, provinces that helps explain inequality in Kenya. inequality went down by one third from 0.42 to 0.28 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 47 The Extent and Evolution of Poverty and Inequality in Kenya Box 2.4: Inequality measures While poverty measures absolute deprivation with respect to a given threshold, inequality is a relative measure of poverty indicating how little some parts of a population have relative the entire population. In the context of monetary poverty, equality can be defined as an equal distribution of consumption / income across the population. This means that each share of the population owns the same share of consumption / income. The Lorenz Curve compares graphically the cumulative share of the population with their cumulative share of consumption / income. A perfectly equal consumption / income distribution is indicated by a diagonal. The other extreme is complete inequality where one individual owns all the consumption / income. These two (theoretical) extremes define the boundaries for observed inequality. The Gini coefficient is the most commonly used measure for inequality. A Gini coefficient of 0 indicates perfect equality while 1 signifies complete inequality. In relation to the Lorenz Curve, the Gini coefficient measures the area between the Lorenz Curve and the diagonal. The Theil Index measures inequality based on an entropy measure. A parameter α controls emphasis to measure inequality for higher incomes (larger α) or lower incomes (smaller α). The Theil index with parameter α=1 is usually called Theil T while using α=0 is called Theil L or log deviation measure. Relative and absolute consumption / income differences can be used to compare inequality dynamics over time. Usually, percentiles are used to compare incomes of different groups. For example, p90/p10 is the ratio (for relative incomes) or difference (for absolute incomes) of the average consumption in the 90th and 10th percentile. Given the nonresponse issues in Nairobi, we opt for the p75/p25 ratio, which is the average consumption ratio in the 75th and 25th percentiles. Finally, the Atkinson index introduces value judgements about the degree of inequality aversion prevalent in the society, which is expressed by the choice of an inequality aversion parameter. The higher this parameter, the more emphasis is placed on the lower tail of the distributions and the changes experienced there. Source: World Bank’s Poverty Handbook. Figure 2.17: Atkinson index and P75/P25 inequality index nationally and by area of residence a) Atkinson index (∑=1) b) P75/P25 ratio 0.4 3 2.7 2.5 2.5 0.30 2.4 0.3 0.29 2.2 2.1 2 0.23 P75/P25 ratio 0.21 Index 0.2 0.19 0.17 1 0.1 0.0 0 National Urban Rural National Urban Rural 2005/06 2015/16 2005/06 2015/16 Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16 48 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead The Extent and Evolution of Poverty and Inequality in Kenya Table 2.4: Theil inequality index - decomposition by urban/rural location and province By urban / rural By province 2005/06 2015/16 2005/06 2015/16 Between group 0.11 0.07 0.10 0.05 Within group 0.30 0.21 0.32 0.23 Total 0.42 0.28 0.42 0.28 Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16. 2.3 WHAT EXPLAINS THE TRENDS IN have a large impact in terms of poverty reduction, in POVERTY REDUCTION? POVERTY addition to other benefits in terms of human capital DECOMPOSITION EXERCISES accumulation explored in detail on Chapter 8 of this U nveiling the main drivers behind the observed report. It is nevertheless likely that the nonresponse changes in the poverty incidence of Kenya is an problem experienced in Nairobi during the collection important objective of this report. This section makes of the KIHBS 2015/16 survey is overestimating the use of several decomposition exercises to help shed some contribution of the redistribution effect. light on what were some of the main factors behind the ten-percentage point decline observed during the The contribution of economic growth to poverty ten-year period focus of this report. More specifically, it reduction is more marked in rural areas. As shown examines the role of growth versus redistribution, the in Figure 2.18a, poverty reduction is mainly driven by economic growth, accounting for three quarters of the progress in urban / rural areas and provinces and the almost twelve percentage point reduction in the share population shift amongst them, the relative importance of the rural population living beneath the poverty line. of the household’s sources of income (in the sectorial The results of the growth-inequality decomposition sense), and the role of mobile money.91 for urban areas show that the decline in inequality (redistribution effect) drove the entirety of the reduction 2.3.1 The role of growth and redistribution in poverty between 2005/06 and 2015/16.92 However, While both economic growth and redistribution these results are affected by the nonresponse problem contributed to Kenya’s poverty reduction, the former in Nairobi. To partially address this problem, the same helps explain almost two thirds of the decline decompositions have been conducted in a sample observed. Consistent with the common view that excluding the top twenty percent of households and overall economic growth is usually accompanied by are presented in Figure 2.18b. This scenario shows that an increase in the living standards of the population, it is likely that economic growth did contribute to the growth accounts for almost 60 percent of the poverty reduction of poverty in urban areas. reduction observed in Kenya for the period 2005/06 to 2015/16. The remaining 40 percent (the interaction Consistent with the pattern observed at the terms explains less than 1 percent), is attributable to the national level, the growth effect was larger than the redistribution effect (Figure 2.18a). This is an important distributional effect in each of the Kenyan provinces. result, as further efforts to improve redistribution and Economic growth accounts for the majority of the decline further reduce inequality would likely accelerate observed in the provinces (except for Nairobi), with the poverty reduction for Kenya in the medium term. magnitude ranging from 5.5 percentage points (almost Redistributive policies such as the expansion of three quarters of the overall reduction) in Rift Valley to social protection programs at a national level, would 17 percentage points in the Coast province (which 91 For the role of mobile money, we mostly review the extensive recent 92 Actually, the results point out that the redistribution effect alone would economic literature on the link between access to this financial services have reduced poverty by 10.8 percentage points between 2005/06 and and poverty reduction. For most of these studies it is possible to identify 2015/16, resulting in a poverty incidence of 21.3 percent in 2015/16 a causal link either by the use of randomized control trial (RCT) or the rather than the observed 29.4 percent, had it not been because the robust econometric identification strategies. growth effect hindered poverty reduction. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 49 The Extent and Evolution of Poverty and Inequality in Kenya Box 2.5: What does decomposing changes in poverty entail? In this chapter the results of two decomposition methods are presented. The first method is the Datt-Ravallion approach, which isolates the growth and redistribution effects associated with the decline in poverty over the period of analysis. Conceptually, this decomposition is based on the idea that that a measure of monetary poverty can be expressed as the product of mean consumption and a parameterized Lorenz curve. Keeping the Lorenz curve constant gives the distribution neutral growth that would drive the average increase in consumption across the population, for instance, raising the levels of consumption of all households by the same rate. The other part is derived from holding the mean consumption constant (a mean-preserving redistribution) to capture the change in the shape of the consumption distribution driven by, for instance, a faster growth in the consumption of the poorest relative to the consumption growth of the richest (Datt and Ravallion 1992). The second is the Ravallion and Huppi (1991) decomposition method, that quantifies how much poverty reduction among mutually exclusive groups or movement between these groups accounts for national poverty reduction. More specifically, the analysis decomposes changes in poverty over time into “intra-group effects” (poverty changes within sectors, within provinces, or within urban and rural areas, while assuming no changes in the distribution of the population across groups), “inter-group effects” (allowing for changes in the distribution of the population between groups keeping poverty rates constant) and an “interaction” term that can be interpreted as a measure of the correlation between the population shifts and the intra-group changes in poverty. Under both methods, a counterfactual scenario is used and estimates are made as to what would have happened to poverty had the counterfactual scenario occurred. By defining a counterfactual scenario, the changes that have been important to overall poverty reduction can be quantified, be it a distribution-neutral consumption growth, the amount of poverty reduction that took place within a sector (as if the distribution across sectors had not changed), or the amount of poverty reduction that took place as a result of people moving between groups. Figure 2.18: Determinants of changes in poverty – Datt-Ravallion decomposition by area of residence a) All households b) Without top 20% of urban households 10 10 Percentage point change in headcount rate Percentage point change in headcount rate 5 5 8.1 0 0 -1.5 -2.8 -6.3 -9.9 -8.7 -5 -5 -8.7 -10.8 -4.3 -10 -3.0 -10 -0.9 -3.0 -15 -15 National Rural Urban National Rural Urban Growth Distribution Growth Distribution Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16. corresponds to nearly the entirety of the reduction, as offsetting the growth effect. Thus, for these two redistribution contributes mere 0.1 percentage points provinces, the decline in poverty has been hindered by to the overall decline, Figure 2.19). In the case of two the changes in the consumption distribution, despite provinces, the North Eastern (poorest) and Central the fact that inequality declined in both provinces, as (second-wealthiest) provinces, the distributional effect measured by the Gini index (Figure 2.15). actually contributed to an increase in poverty, partially 50 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead The Extent and Evolution of Poverty and Inequality in Kenya Figure 2.19: Determinants of changes in poverty – Datt-Ravallion decomposition by province 15 10 5 8.8 2.6 Percentage point change in 0 2.1 -5.5 headcount rate -6.5 -8.9 -7.7 -5 -8.5 -13.4 -2.0 -2.9 -10 -3.2 -17.0 -15 -17.0 -0.1 -1.8 -20 -25 Coast North Eastern Central Rift Western Nyanza Nairobi Eastern Valley Growth Distribution Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16. 2.3.2 The role of poverty reduction within prices through most of the period analyzed and greater geographical areas versus internal migration commercialization of agricultural production, increased Unsurprisingly, poverty reduction amongst rural the wellbeing of households engaged in agricultural households accounted for almost all the poverty production. Moreover, the fact that rural households reduction observed at the national level. The Ravallion- experienced an increased off-farm diversification of Huppi decomposition exercises allow a decomposition income activities (as showed later in this section) also of Kenya’s change in poverty over time into changes helped to reduce poverty in rural Kenya. amongst urban households and rural households, assuming no migration between the two sectors, Internal migration, specifically rural-urban migration, as well as changes due to population shifts among is usually associated with economic progress, them (internal migration). Unsurprisingly, poverty access to better job opportunities and better living reduction amongst rural households accounted for conditions. Rural-urban migration is an inherent aspect 87.6 percent of poverty reduction observed in Kenya of the economic development process all around the during the period 2005/06 to 2015/16 (Figure 2.20a). A world and can in principle support poverty reduction. robust performance of the agricultural sector after the When migration is driven by “pull” forces that, for economic slow-down of 2008, together with high food instance, attract migrants from rural areas looking for Figure 2.20: Contribution to poverty reduction a) Rural/urban b) Provincial - 4.2% Coast -7.0% 3.3% 3.4% 14.8% North Eastern 14.3% 1.2% Eastern Rural 13.7% Central Urban 5.1% Rift Valley Population shift e ect Western Interaction e ect 13.7% 28.9% Nyanza Nairobi Population shift e ect 87.6% 17.4% Interaction e ect 7.9% Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 51 The Extent and Evolution of Poverty and Inequality in Kenya better and higher paid economic activities, as well 2.3.3 Structural pattern of poverty reduction as better returns to their endowments, it is usually The agricultural sector contributed to poverty accompanied by poverty reduction (Kenya Urbanization reduction. Several studies have determined that Review, World Bank 2016). However, migration can growth in the sectors in which the poor are employed also be motivated by “push” factors, where migrants is more poverty reducing than growth in other sectors are escaping from conflict, political turmoil, natural (Loayza and Raddatz 2010; Christiaensen, Chuhan-Pole, disasters, or a particular shock affecting their place of and Aly Sanoh 2013). In line with these findings, although residence. In these cases, internal migration can result the agricultural sector has not been as dynamic as the in the physical, social and human capital depletion, and services or the industrial sector (Figure 2.21), it played can lead to a higher incidence of poverty. a notable role in the reduction of poverty in Kenya in the decade leading up to 2015/16. When looking at the For Kenya, internal migration contributed moderately contribution of different sectors to poverty reduction, to poverty reduction throughout the ten-year period each household is first attributed to the sector from being studied. According to the analysis of the two which it draws at least 50 percent of its total income. waves of the KIHBS, the population shift among rural- Those households which do not rely on any one sector urban households did contribute to poverty reduction as their main source of income (meaning no source of in Kenya throughout the period of analysis. Migration income accounts for more than 50 percent) are classified between rural and urban areas accounted for about as diversified. Households for which the main source 14 percent of the overall national reduction in poverty of income is the agricultural sector (including crop (Figure 2.20a). Despite the fact that migration between income, livestock income, and earnings of wage workers urban and rural areas is prevalent in Kenya, this modest in the agricultural sector) account for around 33.85 percent contribution may be partly explained by migration of the overall national poverty reduction (Table 2.5). This selection or the fact that those who migrate are usually contribution stems from the fact that agriculture remains better off or have higher levels of physical or human a source of livelihood for around 60 percent of the labor capital, as is discussed in Chapter 5 of the report based force, and the robust growth of the sector observed on data from the DHS. throughout the period analyzed, thanks to high food and commodity prices. However, agricultural productivity In terms of the provincial contribution to poverty remains low, particularly the production of grains, which reduction, the Eastern province accounted for hinders the transformative potential of the sector to boost almost one third of the overall poverty reduction. As the incomes of poor households. expected, the extent to which the different Kenyan Figure 2.21: Real sector growth 2007–2015 provinces contributed to the overall decline in poverty varies with the progress experienced by the province, 14 as well as changes in the share of the national population and the share of the poor population 10 living there. The Eastern province, for which poverty incidence declined from 50.6 percent in 2005/06 to 6 Percent 31.8 percent in 2015/16, is responsible for over one fourth of the overall poverty reduction (28.9 percent). 2 Other important contributors were the Rift Valley, the -2 most populated province, and the Coast Province, which experienced a large decline in poverty. On the -6 other hand, Nairobi and the North Eastern Province 2007 2007 2007 2007 2007 2007 2007 2007 2007 contributed only 3.4 and 1.2 percent of the decline, Agriculture GDP Services Industry respectively (Figure 2.20b). Source: World Bank 2018b 52 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead The Extent and Evolution of Poverty and Inequality in Kenya Similarly, households deriving their income from classified into agricultural, non-agricultural, and mixed the services sector account for almost 30 percent of households, where agricultural and non-agricultural Kenya’s poverty reduction. The expansion of modern are defined as deriving at least 90 percent of the services, particularly financial intermediation and total income from either sector, and mixed being mobile communication (as a result of the introduction everything else in between. Households depending of innovative solutions like M-PESA), has stimulated overwhelmingly on agriculture income account the demand for more traditional services, and the for 27.63 percent of poverty reduction, while non- sector as a whole is responsible for the bulk of the agricultural households account for almost 21.19 economic growth of Kenya. Between 2006 and 2013, percent (Table 2.6). Interestingly, the contribution the sector accounted for 75 percent of GDP growth.93 of diversified households was around 33.51 percent, Not surprisingly, households that report earning the showing that an important factor for poverty majority of their income from services (comprising reduction has been the ability of households engaged those in wage employment and the majority of in agriculture (and sometime deriving the majority of those engaged in a non-agricultural enterprise) help their income from this activity) to complement their explain about 29 percent of the decline in the national incomes through non-agricultural activities. The ability poverty rate. Moreover, the inter-sectoral movement of agricultural households to engage in activities such of households across these classifications accounts for as petty trading, kiosk retailing, operating taxis and 16.4 percent of the decline (Table 2.5), suggesting that running local enterprises, reduces their vulnerability to the structural transformation in the country, mainly climatic and price shocks, and increases their ability to from the agricultural sector towards the services sector, generate income. has aided poverty reduction. 2.3.4 The role of mobile money The evidence suggests that off-farm diversification Access to mobile phones in Kenya increased has been important for poverty reduction in Kenya. dramatically over the last 15 years, transforming Given that close to 70 percent of those households the economic paradigm. In 1999, the Kenya-based engaged in agriculture have additional sources of mobile service provider Safaricom projected that income from non-agricultural activities, additional the total mobile phone market would reach three decompositions analyzing the diversification of million subscribers by 2020 in Kenya. However, by income sources were undertaken. Households were 2009 Safaricom alone had over 14 million subscribers Table 2.5: Sectoral decomposition of poverty reduction (Ravallion-Huppi) Pop. share in period 1 Share of total change Source of income Absolute change (percent) (percent) Agriculture 49.18 -3.32 33.85 Industry wages 5.84 -0.20 2.05 Service wages 24.49 -1.76 17.91 Non-agr. enterprise 7.30 -1.12 11.41 Transfers 5.99 -0.38 3.87 Diversified 7.19 -1.18 11.99 Total intra-sectoral effect -7.95 81.08 Population shift effect -1.61 16.40 Interaction effect -0.25 2.52 Change in headcount rate -9.81 100.00 Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16. 93 According to the World Bank’s Country Economic Memorandum (2016). KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 53 The Extent and Evolution of Poverty and Inequality in Kenya Table 2.6: Sectoral decomposition of poverty reduction (Ravallion-Huppi) - alternative definition Pop. share in period 1 Share of total change Source of income Absolute change (percent) (percent) Non-agricultural income only 31.64 -2.16 21.19 Agriculture income only 39.79 -2.81 27.63 Mixed - agricultural & non-agricultural income 28.57 -3.41 33.51 Total intra-sectoral effect   -8.37 82.33 Population shift effect   -1.68 16.49 Interaction effect   -0.12 1.19 Change in headcount rate   -10.17 100.00 Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16. (Aker and Mbiti 2010). As in many countries in increased per capita consumption levels and lifted Africa, mobile phones represented the first modern around 2% of Kenyan households out of poverty (Jack telecommunication infrastructure of any kind, and Suri 2016).94 particularly for those in rural areas. Moreover, the adoption among firms, mainly in urban centers, More effective risk-sharing has been a key factor appears to be correlated with the poor quality of the in improving financial resilience. Through mobile landline services. While it was clear that the reduction in money, households are able to share risk more the communication costs would bring efficiency gains efficiently with relatives, friends and other associates, in all economic markets, as information about prices helping them to smooth consumption over time and and quantities reaches agents faster than before and increase their savings. Mobile money users report better market coordination is possible (Jensen, 2007; having access to more credit and emergency-related Aker, 2008; Aker, 2010; Klonner and Nolen, 2008), transfers than nonusers, suggesting that both explicit the impact of mobile phones has gone well beyond credit and informal insurance arrangements can be that thanks to introduction of mobile money (more more effectively sustained (Suri and Jack, 2013). M-PESA specifically M-PESA). As a growing body of rigorous users are more likely to receive and send (internal) academic literature has shown recently, mobile phone remittances (Jack, Ray, and Suri 2013), and in case of a penetration in Kenya has shifted the economic paradigm, negative shock, M-PESA users receive a larger amount constituting a platform for service delivery rather than of funds from a wider network than non-users (Jack and being a simple communication tool. Suri 2014). Beyond the economic efficiency gains from Mobile money has also contributed to increased improved communication, mobile phones, through employment, savings, and productive investment. mobile money, have been shown to increase per The introduction of mobile money has been shown to capita consumption and reduce poverty among have a positive effect on local employment (Plyler et users. As of 2014, 97 percent of Kenyans reported al., 2010; Mbiti and Weil 2011), and contributes to an having an M-PESA account and by 2015 there were improved business environment and access to trade 65,569 registered M-PESA agents in the country. This credit (Plyler, Haas, and Nagarajan 2010; Beck et al. 2015). service, which allows individuals to transact money Using mobile money appears to increase savings not only without having access to a formal bank account, among the “unbanked” but also among those with access has expanded the economic possibilities of the population, by increasing their financial resilience to regular banking services (Morawczynski and Pickens and savings, and allowing them to move out of 2009), and creates new savings incentives for smallholder agriculture and into business. A recent study shows farmers (Kikulwe, Fischer, and Qaim 2014). In turn, that through these mechanisms access to M-PESA 94 The authors use US$ 1.25 per day as a measure of extreme poverty. 54 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead The Extent and Evolution of Poverty and Inequality in Kenya households are able to make productive investments: 2.4 POVERTY PROFILES – WHAT ARE mobile money has allowed users to accumulate greater CHARACTERISTICS OF THE POOR amounts of capital (Plyler, Haas, and Nagarajan 2010), IN KENYA? and allowed farmers to engage in more commercially- oriented farming, increase sales of harvested products, P rofiling the characteristics of the poor is helpful in identifying what factors are limiting their economic opportunities. Moreover, comparing poor and non- and realize higher profits per acre of production (Kikulwe, Fischer, and Qaim 2014). poor households along different dimensions, such as demographics, human capital, economic activities Gender outcomes have also improved due to and asset ownership, can pin down specific policy mobile money. The introduction of mobile money, in actions that may help raise their living standards and conjunction with M-PESA has helped to increase female overcome poverty. empowerment in Kenya. This was particularly true in rural areas, where the technology made it easier for Household living in poverty have older household women to obtain remittances from relatives (other than heads and are more likely to be headed by a their husbands) and friends increasing their financial woman. Poor households tend to have slightly older independence (Morawczynski and Pickens 2009). household heads. The average age in 2015/16 among Similarly, the effects of M-PESA on consumption and poor households was 47 years versus 42 for non-poor poverty accrue particularly to women: the magnitude households (see Table 2.7). This age gap between poor of the effect for female-headed households was more and non-poor households has remained constant than twice as high as the average effect (Suri and Jack since 2005/06. Female headed households are more 2016). More specifically, the evidence suggests that likely to be poor, even after all other characteristics of mobile money allowed women to increase their savings the household are taken into account in a multivariate and smooth consumption, and induced changes in regression analysis (see column Significance -Model- in occupation choice. Financial inclusion helped them to Table 2.7). This is particularly true for households headed graduate from subsistence agriculture (into sales/small by widows and divorcées (or separated). As explored in business) and to reduce their reliance on multiple part- detail in Chapter 3, marital rupture frequently entails a time occupations. loss of economic means for women. In addition to that, the proportion of households headed by a woman has More recently, the M-PESA platform has facilitated increased slightly between 2005/06 and 2015/16 for targeted development interventions in education both poor and non-poor households. This is important and health. Recognizing that the transition from as age and gender reflects the economic opportunities primary school to high school is costly and often leads of the household head, which matter significantly for to dropout, an intervention encouraging parents of the total income of most households. primary school leavers to open a mobile banking account through the M-PESA platform substantially Poor households tend to have a larger size and increased financial savings and high school enrolment higher dependency ratios. In terms of demographic (Jack and Habyarimana 2018). The M-PESA platform characteristics, poor households have 1.75 household has also been used for provision of conditional cash members more (a considerable gap) and a larger transfers and vouchers covering the full cost of dependency ratio95 than non-poor households (see giving birth in a medical facility, which appears to be Table 2.7). While household size decreased by about highly effective in increasing institutional deliveries one person for both poor and non-poor households among poor rural women (Grépin, Habyarimana, and between 2005/6 and 2015/16, the dependency ratio Jack 2017). did not decline for poor households. Regression analysis 95 Which imply a lower share of working age male and female members aged 15-65. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 55 The Extent and Evolution of Poverty and Inequality in Kenya Table 2.7: Household characteristics by poverty status 2005/06 2015/16 Non- Significance Significance Non- Significance Significance Poor Poor Poor (Wald-test) (Model) Poor (Wald-test) (Model) Demographic                 Age of head 42.6 47.9 *** *** 42.1 47.1 *** *** Female head 27.4% 31.3% *** *** 31.2% 35.7% *** *** HH size 4.4 6.2 *** *** 3.5 5.2 *** *** Dependency ratio 36.7% 47.6% *** *** 33.1% 47.4% *** ** Urban 30.9% 16.9% *** *** 39.9% 27.5% *** *** Education                 Ave. years sch. (15+) 7.7 5.3 *** *** 9.1 6.1 *** *** HH head levels               No education 14.3% 32.7% *** Reference 8.8% 27.7% *** Reference Some or complete 43.4% 51.3% ***   42.5% 52.3% ***   primary Some or complete 39.0% 15.9% *** ** 41.6% 19.5% ***   secondary Some or complete tertiary some or 3.2% 0.1% *** *** 7.1% 0.4% *** *** complete Sources of income                 Non-agricultural income 55.5% 44.2% *** Reference 71.1% 68.1% *** Reference only Agriculture income only 12.6% 17.4% ***   2.5% 4.4% *** ** Mixed - agricultural & 31.9% 38.5% *** *** 26.5% 27.4%   *** non agricultural income Access to services                 Improved drinking water 70.2% 51.9% *** *** 80.4% 65.6% ***   Improved sanitation 56.4% 37.7% *** *** 72.2% 47.8% *** *** Main source light 23.0% 4.0% *** *** 49.9% 18.9% *** * (electricity) HH electricity access 26.5% 4.5% *** *** 52.0% 20.7% *** *** Assets                 HH owns radio 58.1% 51.2% *** *** 51.8% 40.6% *** *** HH owns cell phone 27.9% 5.8% *** *** 90.1% 77.8% *** *** HH owns kerosene stove 53.0% 23.2% *** *** 42.3% 22.9% *** *** HH owns charcoal jiko 62.6% 40.3% *** *** 61.7% 44.2% *** *** HH owns mosquito net 40.0% 25.7% *** *** 68.8% 66.1% ** *** HH owns fridge 5.5% 0.5% *** *** 8.2% 0.7% *** *** HH owns sofa 56.8% 29.2% *** *** 62.3% 40.2% *** *** Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16. Note: Column wald test shows significance values from a wald test of differences between the means. Column Model shows significance values from a log-linear probability model (LPM) (with log of consumption as the dependent variable) controlling for all variables shown along with province dummies. *, **, and *** indicate significance level at 10%, 5%, and 1%. Robust errors used. 56 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead The Extent and Evolution of Poverty and Inequality in Kenya Figure 2.22: Household size and average education level nationally, by province and NEDI classification a) Household size b) Average years of schooling (age 15 and older) 7 14 6.2 6 12 11.3 5.2 5 4.7 10 9.8 4.5 4.3 4.4 9.0 Number of people Number of years 4.0 3.9 3.9 3.9 8.7 4 8 8.3 7.9 7.8 7.6 8.1 7.6 7.6 7.1 5.1 3.2 3.0 6.8 6.5 6.4 6.8 5.8 6.5 3 6 2 4 3.6 2.5 2.1 1.4 1 2 0 0 l t rn l y rn za bi DI DI rn na ra as t l l n Ce n l y rn Na a n- i DI DI lle rn No irob na ra ra as iro ste te lle an NE NE z ba r nt ste Co tio ste te an NE NE Va Ru nt ste Co tio es Va Na Ce Ny Ur n- Ea es Ea Ny Na Ea W ft Ea Na No W ft Ri rth Ri rth No 2005/06 2015/16 2005/06 2015/16 No Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16 confirms that a larger share of dependents within a the country when it comes to education outcomes, household is associated with lower consumption, with only 2.54 and 3.64 average years of schooling even after considering all other relevant household respectively (Figure 2.23b). Improving the education characteristics. When looking at the evolution of these outcomes of the poor is necessary for them to access dimensions at the provincial level (and under the NEDI better income-generating economic activities and and non-NEDI classifications), it is clear that the lack of enhance their consumption levels. a demographic transition in the North Eastern province and in the NEDI counties has slowed down the pace of The proportion of households that solely depend poverty reduction in these regions (Figure 2.23a). on agriculture for their income is much lower than before, but still remains higher amongst the poor. As expected, poverty is associated with lower levels Back in 2005/06, 17.4 percent of poor households of educational attainment. In 2015/16, the average depended exclusively on agriculture for their income years of schooling (of household members 15 years (12.6 percent for non-poor households), but ten years old and above) for non-poor households is three years later this proportion declined to 4.4 percent (2.5 higher than for poor households. Similarly, only 19.9 percent of non-poor, see Table 2.7), illustrating the percent of household heads in poor households have off-farm diversification that propelled the consumption completed at least secondary education, compared to growth among the Kenyan population. As expected, 48.7 percent for their non-poor counterparts (see Table the regression analysis points out that household 2.7). Of greater concern, the gap in the educational consumption increases with diversification and attainment between poor and non-poor households decreases with agricultural dependency. Interestingly, increased between 2005/06 and 2015/16, suggesting the latter was not the case for 2005/06, showing that that poor households still face notable barriers to the structural pattern of poverty has evolved since then. access Kenya’s the education system, as will be discussed in Chapter 6. Regression analysis shows that Access to basic services tends to be lower amongst every additional year of education at the household poor households. Although between 2005/06 and level (for those 15 and older) increases consumption 2015/16 access to sanitation, water and electricity by 2.9 percent, consistent with the idea that individuals improved for the poor, they continue to have much with higher levels of education have higher paid jobs. lower access rates than non-poor households. Not surprisingly, the North Eastern Province and the While three out of four non-poor households have NEDI counties lag considerably behind the rest of access to improved sanitation, only one in two KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 57 The Extent and Evolution of Poverty and Inequality in Kenya poor households do. Moreover, while access for both Ownership of basic assets is limited for poor groups increased since 2005/6, progress was more households with the exception of mobile phones: pronounced amongst non-poor households. In the almost 80 percent of poor households have one. case of improved water, the coverage rate for the In general, poor households are characterized by poor is 66 percent, also considerably lower than for limited ownership of assets when compared to non- non-poor at 80 percent. In terms of electricity, access poor households. They are less likely to own a radio remains low for non-poor households at 52 percent, (41 percent versus 52 percent), a stove (23 versus 42 and it is even lower for the poor, with a coverage rate of percent) and a refrigerator (1 percent versus 8 percent), 21 percent (Table 2.7). Access to improved sanitation is to provide some examples (see Table 2.7). One notable remarkably low in the North Eastern province and NEDI exception is the ownership of mobile phones: between counties when compared to the rest of the country. At 2005/06 and 2015/16 ownership of mobile phones by the same time, access has worsened dramatically in poor households increased from 6 to 78 percent. This is the Western province: while in 2005/06 65.8 percent relevant, given the importance of mobile phones and, of the province’s population had access to improved more specifically, of mobile money in transforming sanitation, this number was only 40.9 in 2015/16 calling the livelihoods of Kenyan households discussed in a for an urgent policy action on this front (Figure 2.23a). previous section of this chapter. Figure 2.23: Access to improved sanitation, water and electricity by province, urban/rural, and NEDI/non-NEDI status a) Improved sanitation b) Improved water 99 100 97 100 95 85 84 82 71 78 78 76 Percentage of households, % Percentage of households, % 72 71 75 68 70 75 66 66 68 65 64 66 61 58 51 58 44 50 51 49 50 50 43 41 33 36 33 19 20 25 25 0 0 t DI t n l l za l za rn Ny n DI DI rn bi rn No obi l l rn rn ey ey ra as ra ra DI na na as ba r iro NE te te ste ste an an ste NE NE ste Ru nt nt Co all all Co NE tio tio ir Ur es es Ce Ce Na Na Ny tV tV n- n- Ea Ea Ea Ea Na Na W W No Rif Rif rth rth No No 2005/06 2015/16 2005/06 2015/16 c) Access to electricity 100 91 81 Percentage of households, % 75 62 52 50 46 44 36 29 22 23 25 17 16 17 0 DI Ea t n l l za Ny rn DI No bi rn l rn y ra ra na as ba lle iro NE te ste an NE Ru ste nt Co io Va Ur es Ce Na n- t Ea Na W ft Ri th r No 2005/06 2015/16 Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16 58 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead CHAPTER 3 GENDER AND POVERTY SUMMARY Kenyan women are poorer than men during core productive and reproductive years, especially if they experienced a marital dissolution. As in other African countries, Kenyan women are more likely to live in poor households than men starting in their mid-20s and continuing until their 50s. Moreover, women are more likely than men to reside in a poor household if they are separated/divorced (31 vs. 24 percent, p<0.01) or widowed (38 vs. 25 percent, p<0.01). Women who have experienced a marital dissolution also face higher prevalence rates of physical violence than other women and are disproportionately affected by HIV/AIDS. In education, girls continue to be disadvantaged in parts of Kenya but there is also an emerging pattern of boys’ underperformance. Kenya has achieved significant increases in primary and secondary enrollments of boys and girls since the early 2000s, which has been accompanied by a trend towards greater gender parity. At the subnational level, girls have lower enrollment rates than boys in Northeastern Kenya and the coast – but boys’ disadvantages emerge in parts of Central and Western Kenya. Similar patterns are evident for learning outcomes, where advantages for girls are even stronger. Despite these improvements in girls’ education, adult women are twice as likely to be illiterate as adult men, reflecting historical gender inequalities in the education sector. Kenyan women face daunting health challenges and bear the brunt of care work within the household. Despite some decline in maternal mortality since 2005, Kenyan women face a staggering lifetime risk of 1 in 42 of dying due to complications of pregnancy or child birth, which is high even by regional standards. Maternal mortality is most severe in the Northern/Northeastern parts of Kenya, areas with extremely high fertility rates and poor access to reproductive health care. And due to traditional gender roles, women spend a significant amount of time on unpaid care work (for children, elderly and the sick or disabled) within the household. Women are much less likely than men to own property and gender biases linger in parts of Kenya’s legal system. Only 12 percent of women aged 20-49 years report owning any land on their own, compared with 39 percent of men. Kenya is among the few African countries with gender inequality in formal inheritance rights – i.e. with respect to the Law of Succession Act. Gender gaps exist also in terms of access to ICT and financial services, though levels of access are high by regional standards. In 2015/6, 71 percent of working-age women participated in the labor force, compared with 77 percent among men. There has been a significant increase in male and, particularly, female employment over the past decade. For men, this increase was driven by a rise in wage employment, while for women it reflects both rising wage employment but also increasing employment in (farm and non-farm) household enterprises. Female labor force participation is linked to religious norms, education, marital status and the presence of young children in the household. Among women, being of Muslim or other non-Christian religion reduces the probability of participating in the labor force by about 30 percent (relative to being Catholic). Women who are widowed, separated/divorced or polygamously married are significantly more likely to participate in KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 59 Gender and Poverty the labor force than women who are monogamously married. Every child aged 0-5 years reduces women’s probability to be in the labor force by over 2 percent. On the other hand, education, even at the primary level, increases women’s probability to participate in the labor force. Gender inequality in earnings is substantial and cuts across all segments of the labor market. Male wage workers earn 30 percent higher wages/salaries than female wage workers and profits of male-run household enterprises are about twice as high as profits of female-run enterprises. Similarly, households where women are the primary decision-makers in agriculture achieve lower yields (e.g. for maize and beans) than households where men are the primary decision-makers. Gender inequality in earnings reflects a variety of different factors, including gender gaps in access to productive resources and sectoral segregation by gender (i.e. women being disproportionately engaged in low-paying sectors). Gender equality is central to Kenya’s vision of becoming a approach to measuring poverty, the primary source “middle-income country providing a high-quality life to all of information are household surveys and the key its citizens by the year 2030.” (Government of Kenya 2007). indicator is a money‐metric measure of welfare based No society can develop sustainably without transforming on consumption (or income) data collected for the the distribution of opportunities, resources, and choices household as a whole. This approach masks within- for males and females so that they have equal power to household differences in consumption along gender, age shape their own lives and contribute to their families, and other dimensions. communities, and country. We use a lifecycle approach to obtain a better This chapter provides a synthesis of what is known about understanding of gender differences in poverty in Kenya, the gender-poverty nexus in Kenya. It starts with a basic even with the existing constraints (i.e. poverty status being determined at the household-level). A recent profile of poverty and gender in section 3.1. Following the collaborative effort between UN Women and the World framework of the 2012 World Development Report on Bank (Munoz Boudet et al. 2018) analyzes whether Gender (World Bank 2011), it then proceeds to analyze life events – such as the transition from childhood to gender gaps in endowments (section 3.2), economic adolescence, adulthood, and elder years; or marriage, opportunities (section 3.3), and voice and agency (section divorce and widowhood – affect men and women 3.4). At the end of each section, the chapter also provides differently in terms of their probability to live in poor a short discussion of possible policy levers to narrow – households The study further develops a demographic and ultimately close – gender gaps and promote a more taxonomy that categorizes households by the number equitable society.96 and sex of adult household members (e.g. 2 adults of opposite sex, single adult male/female households, etc.) to examine the relationship between poverty and 3.1 A PROFILE OF POVERTY AND GENDER the household’s demographic composition in a way IN KENYA that goes beyond a comparison of male- and female- O ne of the key challenges towards an understanding of poverty and gender is that poverty is typically measured at the household level. In the standard headed households. Following this approach and using the data of the 2015/6 KIHBS, this section presents a profile of poverty and gender in Kenya. 96 Underlying this chapter are several data sources, including the KIHBS of 2015/6 and 2005/6, DHS of 2014, 2008/9 and 2003, Global Findex database for 2014 (Demirguc-Kunt et al. 2015), and other country-level databases (e.g. WDI – World Bank 2017b; Women, Business and the Law – World Bank 2015b). These data sources provide a rich information base to analyze gender gaps in different sectors and their link to poverty in Kenya, but there are still important data gaps. Appendix C1 of this chapter hence provides suggestions for possible tweaks to the KIHBS instrument that would help to fill key gaps in data and knowledge about gender inequality in Kenya. 60 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Gender and Poverty 3.1.1 Gender differences in poverty through the and van de Walle 2018). Conversely, among the never lifecycle married population (which here includes children), As in other African countries, Kenyan women are female poverty rates are lower than male poverty rates. more likely to live in poor households than men starting in their mid-20s and continuing until their 50s Figure 3.2: Male and female poverty rates by marital status, 2015/6 (Figure 3.1). Women are hence poorer than men during core productive and reproductive stages of life. This 50 46 pattern suggests that care responsibilities for children 43 40 38 combined with constraints in economic opportunities 38 34 are major vulnerability factors for women. 29 31 30 28 Percent 24 25 Figure 3.1: Male and female poverty rates by age group, 2015/6 20 60 10 50 0 Poverty rate (%) 40 Monogamously Separated Widow Never Polygamously married or living married or divorced or widower married together 30 Males Females 20 Source: KIHBS 2015/6. Note: Cross-tabulation of individuals’ poverty status (assigned at the household- level) and individual-level characteristics (marital status). All differences between 10 males and females are statistically significant at 1 percent, except for the polygamously married population (significant at 10 percent). 0 0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75+ 3.1.2 Gender differences in poverty using a Age group demographic taxonomy of households Poverty, male Poverty, female There is a large gender difference in poverty Source: KIHBS 2015/6. Note: Cross-tabulation of individuals’ poverty status (assigned at the household- amongst households with only a single adult (and level) and individual-level characteristics (age, sex). possibly children). Households comprising one adult female are twice as likely to be poor (35 percent) than Gender differences in poverty also emerge from a households comprising one adult male (18 percent) comparison of male and female poverty rates – i.e. (Figure 3.3). This reflects, among other things, that the probability of living in a poor household – by women living on their own are much more likely than marital status (Figure 3.2). Gender gaps are relatively men to care for children. Poverty rates are highest, at 40 small among the married population, though still percent, among households comprising only children/ favoring men. Women, however, are much more likely seniors and among households comprising 2 adults than men to be poor if they are separated/divorced of same sex or 3+ adults, typically multigenerational (31 vs. 24 percent, p<0.01) or widowed (38 vs. 25 households. Almost 42 percent of Kenya’s poor live percent, p<0.01). These findings are consistent with in these multigenerational households. Poverty rates other studies showing that, for African women, marital are somewhat lower (35 percent) for households rupture frequently entails a loss of economic means and comprising 2 adults of opposite sex, which are in most support that are acquired through, and conditional on cases nuclear families. However, due to their prevalence marriage—including access to productive assets (such in the population, these households still account for as land) and the marital home (Kevane 2004; Djuikom about 40 percent of Kenya’s poor population. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 61 Gender and Poverty Figure 3.3: Poverty and household demographic composition, 2015/6 Poverty rates by demographic household composition Distribution of the poor population by demographic household composition 50 40 40 40 35 35 30 Percent 20 18 10 1 adult female, 13% 0 Only 1 adult 1 adult 1 adult 2 adults Only male female male, 1 of same children 2 adults of same sex or 1 adult male, children 1 adult adult sex or 3+ and/or 3+ adults, 42% 1 adult female, 40% and/or male female adults seniors seniors Source: KIHBS 2015/6. Note: Household taxonomy is based on the number of male and female adults (18-64 years), irrespective of the number of children (<18 years) and elderly (65+ years). Left: Share of population below the poverty line by demographic taxonomy. Right: Distribution of poor population by demographic taxonomy. 3.2 GENDER GAPS IN ENDOWMENTS Girls also perform better than boys in Math, English and T he focus of this section is gender differences Kiswahili, especially in earlier grades of primary school in endowments. This includes human capital (Uwezo 2016). endowments – education and health – but also time “Previously the community preferred withdrawing availability and access to physical and financial assets. a girl child from school during times of economic Gender gaps in endowments not only matter in their own stress. After the introduction of free primary right, but also contribute to gender inequality in economic education, the situation has changed and all opportunities (highlighted in section 3.3) and are hence children have equal opportunity to attend critical for poverty reduction efforts. school.” (Namwitsula community, Busia) 3.2.1 Education Gender gaps in the education sector, however, Kenya has achieved significant increases in primary differ markedly across regions (Figure 3.4). Gross and secondary enrollments since the early 2000s, enrollment rates are higher for girls than for boys especially among girls. The 2005 Participatory Poverty in parts of Central and Western Kenya, but in most Assessment (PPA) already provided a glimpse of this other areas – especially the Northeast and Coast – the societal transformation, as illustrated below by a traditional patterns of higher enrollments among boys quotation from a community in Busia.97 Ten years on, still hold. In terms of learning outcomes (here math the trend towards higher girls’ enrollment is clearly proficiency) girls’ advantages are more widespread – visible. Between 2005/6 and 2015/6, gender parity in consistent with the results at the national level – but gross enrollments, defined as the ratio of female to show a broadly similar geographic pattern. These male enrollment rates, increased at the primary (from regional differences, which may reflect differences 0.95 to 0.97) and secondary (from 0.89 to 0.95) levels. across regions in broader development, female labor And since girls are less likely than boys to attend school force participation, religious and social norms, are over-aged (for the level at which they are enrolled), currently not well understood and would merit further NERs are even higher for girls than for boys (Table 3.1). in-depth analysis. 97 The communities interviewed for the 2005 PPA often commented that a greater emphasis on girls’ education came in the wake of Kenya’s FPE policy introduced in 2003. However, Lucas and Mbiti (2012a) argue that while FPE boosted primary school completion rates of girls and boys, it had larger effects on boys. These results suggest that FPE was not the primary driver for greater gender parity in Kenya’s school. 62 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Gender and Poverty Table 3.1: Primary and secondary enrollment rates and gender parity, 2005/6 and 2015/6 Primary Net Gross   Female Male Gender parity Female Male Gender parity   index index 2005/6 0.82 0.81 1.01 1.16 1.22 0.95 2015/6 0.85 0.84 1.01 1.06 1.09 0.97 Secondary Net Gross   Female Male Gender parity Female Male Gender parity   index index 2005/6 0.21 0.19 1.09 0.39 0.44 0.89 2015/6 0.44 0.41 1.08 0.73 0.77 0.95 Source: KIHBS 2005/6 and 2015/6. Note: The gender parity index is defined as the ratio of female to male enrollments. Figure 3.4: Regional differences in gender parity in the education sector Gender parity index Gender parity index Gender parity index Gross Primary Enrollment Rates Gross Secondary Enrollment Rates Uwezo - Math proficiency (1.4,1.5] (1.3,1.4] (1.2,1.3] (1.6,1.8] (1.1,1.2] (1.4,1.6] (1.2,1.3] (1,1.1] (1.2,1.4] (1.1,1.2] (.9,1] (1,1.2] (1,1.1] (.8,.9] (.8,1] (.9,1] (.7,.8] (.6,.8] (.8,.9] (.6,.7] (.4,.6] [.7,.8] [.5,.6] [.2,.4] Source: KIHBS 2015/6 and Uwezo 2014 data. Note: The gender parity index is defined as the ratio of female to male enrollment/proficiency rates. A value above (below) unity indicates that girls have higher (lower) levels of enrollments/proficiency. Girls and boys, when they drop out of school, about fertility and schooling are typically made jointly often do so for different reasons. Girls dropping out (see Ozier 2016; Duflo, Dupas, and Kremer 2015a). of secondary school are more likely to be married and to have given birth, than girls still attending While girls are enrolled in greater numbers in Kenyan school.98 Asked directly about the main reason why schools than ever before, adult women continue to a household member stopped attending secondary be disadvantaged in educational attainment and school (before secondary completion), “school cost” is literacy compared with adult men. At the national the most commonly cited reason for boys, followed by level, illiteracy is almost twice as high among women ”lack of interest.” For girls, the reason most commonly aged 15+ (18 percent) than among adult men aged mentioned is ”pregnancy”, followed by ”school cost”. 15+ (10 percent), and no county, except Nairobi, has However, the causality between pregnancy and achieved gender parity in literacy among this age group dropping out of school may run both ways, as decisions (Figure 3.5). This reflects historical gender inequalities in 98 Secondary dropout is defined as having attended secondary school the education sector, which continue to put women at Form 1-3 during the last school year, but no longer attending school a disadvantage in terms of labor market opportunities. during the current school year. Note that there are only few cases of secondary dropouts captured by the KIHBS N=70), which limits the analysis that can be performed. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 63 Gender and Poverty Figure 3.5: Male and female literacy by county, 2015/6 100 80 60 Percent 40 20 0 Nairobi Nyeri Mombasa Kiambu Kisumu Nandi Nakuru Machakos Trans Nzoia Kisii Nyandarua Vihiga Uasin Gishu Bungoma Kirinyaga Taita Taveta Siaya Muranga Embu Nyamira Kericho Homa Bay Migori Elgeyo Marakwet Bomet Baringo Kajiado Makueni Lamu Tharaka Nithi Kakamega Kitui Busia Meru Laikipia Tana River Isiolo West Pokot Kwale Samburu Mandera Garissa Wajir Marsabit Turkana Narok Kili Male Female Source: KIHBS 2015/6. Note: Respondents who report being able to read or write in any language or attended secondary school or above are considered as literate. Population aged 15+ years. 3.2.2 Health and fertility Women are disproportionately affected by the HIV/ Kenyan women face a staggering lifetime risk of 1 AIDS epidemic. While prevalence rates have declined in 42 of dying due to complications of pregnancy from about 7 percent in 2006 to just over 5 percent or child birth. While the maternal mortality ratio has in 2016 (of the total population aged 15-49 years), declined from 728 to 510 (maternal deaths per 100,000 women make up more than 60 percent of the share live births) between 2005 and 2015, it remains high of the population (15+) living with the disease (World by regional standards (Figure 3.6a). Geographically, Bank 2017b). Moreover, the 2008 Kenya Demographic maternal mortality is highest in the Northern/North and Health Survey (KDHS; the latest to include HIV/AIDS Eastern parts of Kenya (Figure 3.6b). These areas testing) shows that widows and divorced/separated of high maternal mortality also perform poorly in women are at particularly high risk, with prevalence rates terms of the share of live births being delivered by a that at the time were more than five times (widows) or skilled provider or in a health facility (Muraguri 2015), twice (divorced/separated women) as high as those of suggesting that lack of access and/or poor affordability the total female population (KNBS et al. 2010). Similar of reproductive care services play an important role demographic patterns have been observed for other (see chapter 7 on health). countries in Africa (Djuikom and van de Walle 2018). Figure 3.6: Maternal mortality a) MMR - 2015 model estimates, Kenya and comparators b) MMR - 2009 census estimates by county 600 547 510 398 400 353 319 290 200 138 (3000,4000] (2000,3000] (1000,2000] (800,1000] 0 (600,800] (400,600] SSA Kenya Tanzania Ethiopia Ghana Rwanda South (200,400] [0,200] Africa Based on 2009 Census. Source: WHO et al. (2015) and KNBS (2012). Note: The maternal mortality ratio (MMR) is defined as the number of maternal deaths per 100,000 live births. 64 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Gender and Poverty Kenya entered the demographic transition earlier noted that women and girls were disproportionately than most other African countries, but the fertility engaged in fetching water for the family and in care decline has somewhat slowed over the past two work – as illustrated below by quotations from two decades. Kenya’s TFR fell from about 8 births per communities from Kilifi. The KIHBS 2015/6 confirms woman in the 1960s to just over 5 births in the late these intrahousehold differences in labor allocations 1990s, a rate of decline that outpaced most other (see Figure 3.8). African countries at the time (Figure 3.7a). Fertility continued to decline throughout the 2000s to reach “Water collection is a responsibility of women about 4 births per woman in 2015, albeit at a slower and girls but in times of water scarcity, men are pace than observed, for example, in Ethiopia and also involved. Men are affected by lack of water Rwanda. From a geographic perspective there is huge in that it stops them from going to work. It is variation in fertility across regions. In 2014, counties socially frowned on for a man to fetch water. like Wajir or West Pokot still had a TFR above 7 births Women have to wait in long queues and do per woman, similar to Kenya’s national average during not have enough time to attend to household the 1980s or present-day Niger, the country with the chores and run their bossiness’s [sic]. In severe highest fertility in the world. At the other end of the water crisis children do not go to school so spectrum, counties like Kiambu, Kirinyaga, Nairobi as to look for water. It also denies them an or Nyeri have a fertility rate of around 2.5 births per opportunity to play. Women usually carry woman, only slightly higher than current-day Mexico water on their heads which they find tedious.” (Figure 3.7b). (Manjengo-Mariakani community, Kilifi) 3.2.3 Time use “Men and women play different roles when Gender differences in time use, related to social family members get sick. The women nurse norms about the division of labor inside the family, the patient by washing them, preparing their are among the most pertinent factors that distinguish meals and feeding them. The men mostly the lives of men and women in Africa (Blackden and provide money to cater for medical costs.” Wodon 2006). In the 2005 PPA, almost every community (Miyani community, Kilifi) Figure 3.7: Kenya’s demographic transition a) TFR, Kenya and comparators, 1960-2015 b) TFR by Kenyan county and comparators, 2014 Total fertiliy rate 8 9 8 6 7 4 6 2 5 4 0 Wajir West Pokot Turkana Samburu Garissa Tana River Migori Trans Nzoia Mandera Homa Bay Marsabit Bungoma Isiolo Baringo Kwale Busia Vihiga Kajiado Kakamega Lamu Bomet Siaya Keiyo-Marakwet Nandi Kericho Kitui Nakuru Laikipia Kisii Uasin Gishu Kisumu Nyandarua Nyamira Tharaka Machakos Makueni Taita Taveta Mombasa Meru Embu Murang'a Nyeri Nairobi Kiambu Kirinyaga Niger Mexico Narok Kili 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 Year Kenya Tanzania Uganda Ethiopia Rwanda Kenyan counties Reference countries Source: World Bank 2018c and DHS 2018 StatCompiler (data for 2014) KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 65 Gender and Poverty Figure 3.8: Household members fetching water, 2015/6 3.2.4 Physical and financial assets Kenyan women are less likely than men to own land and housing property. 12 percent of women aged 7% 20-49 years report owning any land on their own, 9% compared with 39 percent of men – a gender gap of 27 percentage points. The gender gap in sole ownership is even larger for housing – 32 percentage points (Figure 26% 58% 3.9a). Since women are much more likely than men to report joint property ownership, gender gaps are much smaller if joint ownership is taken into consideration, but remain in favor of men (Figure 3.9b). Kenya’s gender gaps in property ownership are similar in magnitude to Boys Girls Men (15+) Women (15+) those found in Tanzania, but significantly larger than, for example, in Ethiopia, where there has been an Source: KIHBS 2015/6. emphasis on joint land registration (Melesse, Dabissa, Though Kenya lacks nationally representative data and Bulte 2018). to document these gender differences in time use, case studies confirm that women spend a significant Kenya is among the few African countries with amount of time on unpaid work. A report by Action gender inequality in formal inheritance rights Aid (2013) collected information on time use patterns (World Bank 2015b). As in other African countries, (based on diaries) across three sites in Kenya. The property rights of women in Kenya are shaped by legal study finds that (per day) women spend on average 99 pluralism, which includes vestiges of colonial, modern minutes on collecting fuel or water, and 359 minutes constitutional, customary and religious laws (Deere and on unpaid care work, together almost a full working Doss 2006; Harrington and Chopra 2010). While Kenya’s day (7 hours and 38 minutes). Men, conversely, report 2010 Constitution contains detailed articles in relation spending only 38 minutes on collecting fuel or water, to equality and non-discrimination, gender biases and 167 minutes on unpaid care work, together 3 hours linger in subordinate statutes. In particular, the Law of and 25 minutes. While these data are based on a small Succession Act distinguishes explicitly between male sample and not nationally representative, they give a and female surviving spouses (Republic of Kenya 2015; sense of the time scarcity of Kenyan women. World Bank 2015b). Figure 3.9: Kenya and comparators gender gaps in land and housing ownership a) Sole ownership b) Sole and joint ownership 60 60 Male ownership rate (%) - female ownership rate (%) Male ownership rate (%) - female ownership rate (%) 40 40 20 20 0 0 -20 -20 Dem Rep Congo Dem Rep Congo Mozambique Mozambique Côte d’Ivoire Côte d’Ivoire Burkina Faso Burkina Faso Sierra Leone Sierra Leone Zimbabwe Zimbabwe -40 Comoros Comoros -40 Tanzania Tanzania Namibia Namibia Ethiopia Ethiopia Rwanda Lesotho Rwanda Lesotho Senegal Senegal Uganda Uganda Burundi Burundi Gambia Gambia Zambia Zambia Guinea Guinea Nigeria Nigeria Malawi Malawi Liberia Liberia Ghana Ghana Kenya Kenya Benin Benin Niger Niger Chad Chad Togo Togo Mali Mali Housing Land Housing Land Source: KDHS 2014 and Gaddis, Lahoti, and Li 2018. Note: Self-reported property ownership in population aged 20-49 years. 66 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Gender and Poverty Gender gaps exist also in terms of access to ICT and lower on various indicators of financial inclusion (Figure financial services, though levels of access are high 3.11) and are more likely to report difficulties in coming by regional standards. Women are less likely to own up with emergency funds (Figure 3.12). However, a phone or to have a subscription to a mobile money access to financial services is still higher in Kenya than transfer platform than men, and the gender gaps in its comparator countries, apart from South Africa increase with age (Figure 3.10). Similarly, women score (Figure 3.13). Figure 3.10: ICT access by sex and age, 2014, 2015/6 a) Mobile phone ownership by sex and age b) Subscription to mobile transfer platform by sex and age 100 100 90 90 80 80 Phone ownership (%) 70 Subscriptions (%) 70 60 60 50 50 40 40 30 30 20 20 10 10 0 0 75+ 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75+ 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 Age Age Phone, male Phone, female M-money transfer, male M-money transfer, female Source: KIHBS 2015/6. Figure 3.11: Financial inclusion, male and female population (15+), 2014 Saved to start, operate, or expand a farm or business Saved at a nancial institution Saved any money in the past year Outstanding mortgage Mobile account Debit card in own name Borrowed to start, operate, or expand a farm or business Borrowed from a nancial institution Borrowed any money in the past year Account at a nancial institution Account 0 20 40 60 80 Percent Male Female Source: Global Findex 2014 (Demirguc-Kunt et al. 2015). KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 67 Gender and Poverty Figure 3.12: Difficulty to come up with emergency funds, male and female population (15+), 2014 Male Female Very Not at all Very possible, possible, possible, 23.0% 17.8% 23.9% Not at all possible, 26.3% Not very possible Somewhat 24.0% Somewhat Not very possible, possible, possible, 34.8% 28.0% 21.2% Source: Global Findex 2014 (Demirguc-Kunt et al. 2015). Figure 3.13: Financial inclusion, Kenya and regional comparison, 2014 Account at a nancial institution, male and female population (15+) 70 60 50 40 Percent 30 20 10 0 South Africa Kenya Ghana Sub-Saharan Uganda Ethiopia Tanzania Africa Male Female Source: Global Findex 2014 (Demirguc-Kunt et al. 2015). 3.2.5 Policies to reduce gender gaps in endowments In the education sector, recent data on enrollments This section lays out possible policy levers to reduce and educational performance paint an uneven gender gaps in endowments. Given the cross- picture – with girls’ and boys’ advantages in different sectoral nature of the analysis, this naturally cannot be parts of the country. Further efforts to improve girls’ exhaustive. Moreover, most gender gaps do not have school enrollment, retention and attainment are still instant solutions but require fundamental changes needed in many parts of Kenya, where gender gaps in social norms about women’s and men’s roles and in the education sector continue to favor boys. But at abilities. The objective of the section is hence rather the same time, emerging boys’ underachievement, modest – to reflect on the previous analysis from a especially in educational performance, also requires policy perspective and to bring in additional empirical attention and should be further analyzed and addressed evidence on what works to close gender gaps in before the pattern becomes deeply entrenched endowments, especially from the growing impact (building, for example, on experiences in Caribbean evaluation literature. countries, which have experienced similar patterns, see Plummer 2010; USAID 2016). 68 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Gender and Poverty Several studies from Kenya suggest that programs preventative health products, empirical evidence from subsidizing the direct or indirect cost of education multiple developing countries, including Kenya, shows can be effective in increasing enrollment and that demand is highly price-sensitive. This suggests educational performance of boys and girls. Based that subsidies are a policy option to boost adoption, on a randomized evaluation across 328 public primary especially if targeted to women (Meredith et al. 2013). schools in Western Kenya, Duflo, Dupas, and Kremer In terms of HIV/AIDS, Duflo, Dupas, and Kremer (2015a) (2015a) show that a program that subsidized the cost show that the government’s HIV curriculum, which of upper primary education by providing free school emphasizes abstinence until marriage, does not reduce uniforms significantly reduced school drop outs for sexually transmitted infections (STIs) (nor teenage boys and girls. Similarly, Kremer, Miguel, and Thornton pregnancy). A joint program, where the HIV curriculum (2009) and Friedman et al. (2016) find that a merit- is combined with the education subsidy highlighted based scholarship program targeting adolescent girls earlier, reduced the prevalence of STIs among girls, but in Western Kenya increased academic test scores the education subsidy in isolation was more effective in among girls from treatment schools. The scholarships lowering dropouts and teen pregnancies. These results were awarded to the highest-scoring 15 percent of highlight the complexities of individual decision- grade-6 girls in the program schools in each district making around schooling and engagement in different and included financial grants to cover school fees forms of casual versus committed relationships, which and supplies and public recognition at an awards each carry different propensities for STI and early ceremony. Interestingly, the scholarship program had pregnancy. Policies targeting any one of these issues positive spillover effects on boys (who were ineligible should therefore be carefully evaluated for unintended for the scholarship) and on girls with low pretest scores consequences. (who were unlikely to win the scholarship). Public investments in services for care can reduce Increased secondary school enrollment among time constraints of women. Scaling up care services adolescent girls may also delay fertility decisions. for children, however, requires innovative approaches, The education subsidy program highlighted above for combining public and private sources of funding. IFC its impact on reducing dropouts also reduced teenage (2017) shows examples of employer-provided child pregnancies (Duflo, Dupas, and Kremer 2015a). care (including case studies of Safaricom and an Similarly, Ozier (2016), using a regression discontinuity agroprocessing company in Kenya), and discusses approach, shows that secondary school enrolment what policies and regulations the public sector can lowers teenage pregnancies among women. put in place to support private child care provision. Wattanga (2015) discusses an innovative initiative of In health, further initiatives to increase access to and the Nairobi City County to use social impact bonds affordability of reproductive health care services are to fund 97 new early childhood education centers in important to reduce maternal mortality, especially poor parts of the city. in Kenya’s arid and semi-arid regions. Examples of such efforts are the recent government-supported Further empirical work would be needed to “Linda Mama” program providing free maternity better understand how different types of public services. An evaluation of a pilot program in central infrastructure provision affect time use and the Kenya further demonstrates that post-natal follow ups, intra-household allocation of labor. A desktop review where community health workers visited or called new from Asia (ADB 2015) argues that improved access to mothers three days after delivery and administered a water reduces the time women spend fetching water, simple checklist, led to earlier utilization of postnatal but that this often leads to an increase in time spend care and better recognition of potential complications on other unpaid activities, such as caring for children. from pregnancy (McConnell et al. 2016). In terms of Investments in sanitation were found to reduce the KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 69 Gender and Poverty amount of time needed each day to find a place to and use their newly acquired bank accounts to protect defecate and reduce the burden of caring for family their income from such demands. The notion that members who fell sick due to poor sanitation. Access social pressure to share resources affects women’s to electricity was found to increase the number of decisions to save and invest is further supported by hours women spend on paid work, partly due to an two other studies. Jakiela and Ozier (2016) show in a lab increase in the number of waking hours. Transportation experiment that Kenyan women adopt an investment infrastructure reduced travel time for women but strategy that aims to conceal their initial endowments also added new time demands. More research along from relatives, even though this strategy reduces their these lines for Kenya and other African countries expected earnings. Schaner (2017) shows that offering would be important to understand how infrastructure ATM cards for newly-opened bank accounts (which investments affect women’s time constraints and intra- increases the liquidity of savings) increased account household labor allocation, particularly in rural areas. use (especially the number of transactions) of male- and jointly-owned accounts, but not of female-owned A review of Kenya’s legal landscape could help to accounts. This is consistent with the idea that women ensure the consistency of various laws on property prefer savings instruments with lower levels of liquidity, ownership and inheritance with the progressive as this protects their savings from the demands of principles of Kenya’s 2010 constitution. Gender spouses and other family members. In addition, Dizon, biased legislation, such as the differential treatment of Gong, and Jones (2017) show that accounts with soft male and female surviving spouses under the Law of commitment can help to increase women’s savings. Succession Act, should be eliminated. There is evidence Their study offered “labeled” mobile money (M-PESA) from other African countries that land formalization accounts to vulnerable women, who were existing programs promoting joint registration of both spouses users of M-PESA and already had an account. The can potentially improve outcomes for women and initiative encouraged the women to use the “labeled” narrow gender gaps (O’Sullivan 2017). Rwanda’s land account for emergency purposes and specific saving tenure regularization program, for example, which goals (to help mental accounting), and sent weekly registered married women as co-owners of land SMS reminders (nudges) related to their savings goals, by default, significantly improved documentation but did not affect financial access (since all women of informal land rights among married women (Ali, already had an existing M-PESA account). The program Deininger, and Goldstein 2014). However, at the same was found to increase women’s mobile money savings, time, women who are not legally married saw an without crowding out other types of savings. It also erosion of rights, which highlights the complexities of led to a substitution away from informal-risk sharing these interventions. arrangements, but did not reduce the women’s capacity to manage risks. Several recent studies from Western Kenya suggest that savings products with an element of illiquidity 3.3 GENDER INEQUALITY IN ECONOMIC and soft commitment can increase women’s savings OPPORTUNITIES T (O’Sullivan 2017). Dupas and Robinson (2013) show his section turns to gender inequality in economic that interest-free bank accounts with large withdrawal opportunities. It starts with a brief description of fees increased savings of female market vendors, while trends in male and female labor market indicators over no such effects were observed for male bicycle-taxi the past decade, and a portrayal of the current situation drivers. A potential explanation for the high take- based on the 2015/6 KIHBS data. The section then reviews up rates of accounts by women – despite (de facto) gender gaps in three broad segments of the labor market: negative interest rates – is that women face pressures in wage employment, non-farm household enterprises to share their income with family members and friends and in agriculture. 70 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Gender and Poverty Figure 3.14: Percent of population employed by category, 3.3.1 Labor market trends and current situation 2005/6 – 2015/6 There has been a striking increase in male and, 80 73 particularly, female employment over the past 65 63 decade. Male employment increased from 63 percent 60 51 in 2005/6 to 73 percent in 2015/6, while female 48 employment increased from 51 percent to 65 percent 40 Percent 40 38 38 39 over this period (Figure 3.14). For men, this increase 30 was driven by a rise in wage employment, while for 20 21 14 women it reflects both rising wage employment and also increasing employment in household enterprises, 0 which here includes both farm and non-farm Male Female Male Female 2005/06 2015/16 enterprises.99 Wage Enterprise Any employment Source: KIHBS 2005/6 and 2015/6. Rising employment has transformed the school- Note: Working-age population (15-64 years). Comparable employment definition. to-work transition of Kenyan youth. There is an increasing number of adolescents, male and female, between 2005/6 and 2015/6. Nonetheless, young who are working while still in school. Moreover, the women continue to be significantly more likely than share of the population below the age of 35 who are young men to be neither employed nor in school neither employed nor in school significantly declined (Figure 3.15). Figure 3.15: Changes in school-to-work transition, 2005/6-2015/6 School-to-work transition School-to-work transition Males - 2005/6 1 Females - 2005/6 1 .8 .8 .6 .6 .4 .4 .2 .2 0 0 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 Employment only Employment and school School only Neither Employment only Employment and school School only Neither School-to-work transition School-to-work transition Males - 2015/6 Females - 2015/6 1 1 .8 .8 .6 .6 .4 .4 .2 .2 0 0 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 Employment only Employment and school School only Neither Employment only Employment and school School only Neither Source: KIHBS 2005/6 and 2015/6. Note: Comparable employment definition. 99 See Appendix C2 for a discussion of the comparability of the 2005/6 and 2015/6 KIHBS labor modules. This section uses a definition of employment and labor force participation that is broadly consistent with the labor statistics standards adopted by the 13th International Conference of Labor Statisticians (ICLS) in 1982. The changes adopted by the 19th ICLS in 2013, which reduce the definition of employment to work performed for pay or profit (thus excluding subsistence agriculture) are not yet incorporated in the KIHBS 2015/6 instrument and hence are not considered in this section. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 71 Gender and Poverty In 2015/6, Kenya had a female labor force Within Kenya, there are significant regional participation rate of 71 percent for the core working- differences – female labor force participation is high age population (15-64 years), compared with a male in central and Western Kenya, but much lower in labor force participation rate of 77 percent (Figure the Northeast. Male labor force participation, though 3.16). The labor force comprises those who are (i) also somewhat lower in the Northeast, varies less. employed and at work, those who are (ii) employed As a result, gender gaps in labor force participation but temporarily absent from work, and those who (measured here as the absolute gap, in percentage are (iii) unemployed.100 In the case of Kenya, both points) are most pronounced in the Northeast, unemployment and temporary absence only account followed by the coast and the areas bordering Tanzania for a small share of the labor force. In terms of regional comparisons, Kenya’s female labor force participation (Figure 3.18), where women are much less likely than rate is higher than the Sub-Saharan African average, men to participate in the labor force. These areas but lower than in most other East African countries, map closely with Kenya’s arid and semi-arid lands, except for Uganda (Figure 3.17). where livestock rearing, particularly of cattle, makes Figure 3.16: Male and female labor force participation, 2015/6 Male share employed, unemployed and out of the labor force (%) Female share employed, unemployed and out of the labor force (%) 100 100 90 90 80 80 70 70 60 60 50 50 40 40 30 30 20 20 10 10 0 0 Employed - at work Employed - absent Employed - at work Employed - absent Unemployed Out of the labor force Unemployed Out of the labor force Source: KIHBS 2015/6. Note: Working-age population (15-64 years). Figure 3.17: Female labor force participation, Kenya and comparators 100 86.1 80 77.0 79.5 74.5 70.0 66.4 62.8 60 Percent 47.8 40 36.6 20 0 Lower middle South Africa Sub-Saharan Uganda Kenya (KIHBS Ghana Ethiopia Tanzania Rwanda income Africa 2015/6) Source: KIHBS 2015/6 and World Bank 2017b. Note: Population aged 15+. 100 To be counted as unemployed, a person must meet the following three criteria: (i) not be presently employed, (ii) available to work and (iii) actively looking for a job (see Appendix C2). 72 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Gender and Poverty Figure 3.18: Geographic variation in male and female labor force participation, 2015/6 Female labor force participation (%) Male labor force participation (%) Gender gap in labor force participation (40,50] (80,90] (30,40] (80,90] (20,30] (70,80] (70,80] (10,20] (60,70] (60,70] (0,10] (50,60] (50,60] (-10,0] (40,50] (40,50] (-20,-10] (30,40] (30,40] (-30,-20] (20,30] (20,30] (-40,-30] [10,20] [10,20] [-50,-40] Source: KIHBS 2015/6. Note: Population aged 15-64. The gender gap is defined as the absolute difference between male and female labor force participation rates, a value greater (lower) than zero indicates higher (lower) male labor force participation. an important contribution to local livelihoods. On the another non-Christian religion reduces the probability other hand, male and female labor force participation of participating in the labor force by about 30 percent rates are similar in most parts of Central and Western (relative to being Catholic). Women who are widowed, Kenya. separated/divorced or polygamously married are significantly more likely to participate in the labor force Multivariate analysis points to the salience of religious than women who are monogamously married (or living norms, education, marital status and the presence of together). Every child aged 0-5 years reduces a woman’s young children for women’s participation in the labor probability to be in the labor force by over 2 percent. force. Following Klasen and Pieters (2015) we estimate Living in urban areas reduces a woman’s likelihood to the probability of being in the labor force for men and be in the labor force by another 7 percent, perhaps a women conditional on different socio-demographic reflection of greater difficulties in combining child care variables. The results are summarized in Figure 3.19 with labor market activity in urban areas, where most (marginal effects significant at 10 percent at least) – economic opportunities are outside of agriculture. On the complete set of coefficients is reported in Table the other hand, education, even at the primary level, C.1, Appendix C.4. Among women, being of Muslim or increases a woman’s probability to participate in the Figure 3.19: Correlates of male and female labor force participation, 2015/6 (Marginal e ects, only if signi cant at 10 percent at least) Urban Number of children aged 6-14 years Number of children aged 0-5 years Head's education = 4, other Head's education = 3, university Head's education = 2, secondary Head's education = 1, primary (own) education = 4, other (own) education = 3, university (own) education = 2, secondary (own) education = 1, primary Religion = 5, None Religion = 4, Other Religion = 3, Muslim Religion = 2, Protestant/other Christian Marital status = 5, never married Marital status = 4, widow or widower Marital status = 3, separated or divorced Marital status = 2, polygamously married Age in years -0.35 -0.3 -0.25 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 Male Female Source: KIHBS 2015/6. Note: Marginal effects after probit estimation (see Table C.1, Appendix C.4). The figure only shows marginal effects significant at the 10 percent level at the minimum. Reference categories as follows: Head’s/own education – no schooling; Religion – Catholic; Marital status – monogamously married. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 73 Gender and Poverty labor force. Interestingly, we do not find any effects Male and female employment in Kenya shows the on women’s labor force participation of the level of traditional patterns of sectoral segregation. Women education of the household head, suggesting that are disproportionately employed in agriculture and income effects are not very important. Religious norms, services, while men have a higher share of employment education and the presence of children are much in the industrial sector (Figure 3.20). Further analysis less important factors for labor force participation at the detailed industry level shows large differences of men. Marital status plays a role, with men who are across sectors in the female intensity of employment separated/divorced or have never been married being (Figure 3.21). The highest female intensities of significantly less likely to participate in the labor force employment are found in the sectors “activities of relative to monogamously married men (essentially household as employers” (i.e. domestic personnel), the opposite pattern as was found for women). These “accommodation and food services” (i.e. the hotel and results are consistent with traditional norms of married restaurant industry) and “human health and social men being the main breadwinner for their families. work”. On the other side of the spectrum, the lowest female intensities are found in “transportation and storage”, “construction”, and “mining and quarrying”. Figure 3.20: Male and female employment by broad sector, 2015/6 Male Female 42.45% 39.7% 42.45% 54.78% 4.77% 17.85% Agriculture Industry Services Agriculture Industry Services Source: KIHBS 2015/6. Figure 3.21: Share of male/female employment by detailed sector, 2015/6 Crop and animal production, hunting and related service activities Wholesale and retail trade; repair of motor vehicles and motorcycles Construction Manufacturing Education Transportation and storage Other service activities Accommodation and food service activities Activities of households as employers, undi erentiated Administrative and support service activities Human health and social work activities Sorted Public administration and defense; compulsory social security by total Professional, scienti c and technical activities employment Mining and quarrying (in descending Information and communication order) Financial and insurance activities Arts, entertainment and recreation Water supply; sewerage, waste management and remediation activities Real estate activities Electricity, gas, steam and air conditioning supply Activities of extraterritorial organizations and bodies 0 20 40 60 80 100 Percent Male Female Source: KIHBS 2015/6. 74 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Gender and Poverty 3.3.2 Wage employment Table 3.3: Male and female monthly earnings, in current Ksh, and male-to-female ratio, 2015/6 Far fewer women than men are in wage employment. Ratio male- Male Female Of the total employed population, almost 50 percent to-female of men are paid employees (denoted in this section as Mean 18,276 14,075 1.30 wage earners) in their primary job, compared with only 10 percentile th 3,000 2,000 1.50 30 percent of women. On the other hand, more women Median 10,000 6,500 1.54 than men work as own-account or contributing family 90th percentile 43,300 35,000 1.24 workers (Table 3.2).101 Source: KIHBS 2015/6. Note: Unconditional earnings (cash and in-kind) of wage earners. Not normalized for working hours. There are large gender gaps in monthly earnings among wage earners, with mean wages/salaries to these characteristics) and an interaction between being 30 percent higher for men than for women. the endowment and coefficient effects.102 As shown These gender gaps are even larger at the bottom in Table C.2, Appendix C.4, the average difference of the earnings distribution (and up to the median), between log earnings of male and female employees where men earn about 50 percent more than women in the regression is 0.37, which corresponds to about (Table 3.3). 45 percent higher wages for male wage workers (consistent with Table 3.3 above).103 The endowment Differences in characteristics between male and effect explains about 43 percent of this difference, female wage earners – in terms of age, education, while 65 percent are explained by the coefficient usual number of working hours, industry, occupation effect. In addition, there is a negative interaction and urban-rural location – explain about half of effect, of -8 percent. the gender gap in monthly earnings. We use the Oaxaca-Blinder decomposition to disaggregate Exploring the endowment effect in detail shows that the gender difference in average monthly earnings the largest advantage of male wage workers is their into an endowment effect (reflecting differences in overrepresentation in industries with relatively high characteristics between male and female wage workers), wage premiums (Table C.3, Appendix C.4). In addition, a coefficient effect (reflecting differences in returns males wage workers benefit from being, on average, Table 3.2: Male and female wage employment by employment status, 2015/6 Population aged 15-64 years, primary job Male Female Total Paid employee (outside household) 47.7 27.4 37.9 Paid employee (within household) 2.2 2.3 2.2 Working employer 1.1 0.7 0.9 Own-account worker 35.9 51.7 43.5 Member of producer cooperative 0.1 0.1 0.1 Contributing family worker 11.3 16.1 13.6 Apprentice 0.7 0.9 0.8 Volunteer 0.4 0.3 0.3 Other (specify) 0.7 0.6 0.6 Total 100 100 100 Source: KIHBS 2015/6. Note: Column percentages. 102 The decomposition is implemented in Stata using the oaxaca command 101 While the KIHBS makes a distinction between own-account and with survey settings and default options (see Jann 2008). contributing family workers, the criteria to distinguish between these 103 Since the dependent variable is log-transformed, we follow Halvorsen types of workers in the context of small, family-run enterprises are often and Palmquist (1980) in calculating the percentage difference in earnings not clearly defined (Beegle and Gaddis 2017). as (exp(0.37)-1)*100=44.8 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 75 Gender and Poverty slightly older than female wage workers, (35 years vs. as high as profits of female-run enterprises. Jointly 33 years) and working longer hours per week (52 vs. run enterprises come out in-between (Figure 3.22).104 46 hours). Female wage workers, on the other hand, The unconditional gender gap is similar in magnitude benefit from having slightly higher levels of education to what is reported in Hallward-Driemeier (2013) based and being overrepresented in occupations with higher on an analysis of household enterprise modules for 19 wage premiums. These two effects, however, are just Sub-Saharan African countries. at the margin of statistical significance and cannot Figure 3.22: Profits of male-, female- and jointly-run compensate the other endowment advantages of household enterprises, 2015/6 male workers. .4 The sizable coefficient effect suggests that men also .3 benefit from more favorable returns to characteristics – but further disaggregation of this effect does not Density yield additional insights. First, apart from age (where 2 women benefit from greater returns to experience), none of the disaggregated coefficient effects is .1 statistically significant in isolation. Second, most of the male advantage in the overall structural effect 0 reflects differences in the regression intercept for male 2 4 6 8 10 12 Pro t (Natural log) and female wage workers, which potentially captures Male Female Joint gender-based discrimination in the labor market, but Source: KIHBS 2015/6. Note: Monthly profits (winsorized) in current Ksh (figure shows natural log). also unobservable factors, and is therefore difficult to interpret (Table C.4, Appendix C.4). Compared with male-run enterprises, female-run household enterprises are less likely to be in industry, 3.3.3 Non-farm household enterprises less likely to be formally registered and tend to Gender gaps in earnings carry over to the non-farm employ fewer paid non-household workers. They household enterprise sector, with average profits of are also less likely to be in urban areas and are more male-run household enterprises being about twice concentrated in poor households (Table 3.4).105 Table 3.4: Descriptive differences between male- and female-run household enterprises, 2015/6 Male-run Female-run By sector (%) Agriculture 2.0 1.9 Industry 15.2 9.6 Services 82.8 88.5 Share registered (%) 15.3 9.2 Labor input Average number of household or unpaid workers 1.2 1.1 Average number of paid non-household workers 0.4 0.1 Share in urban areas (%) 51.5 45.4 Share in poor households (%) 15.2 20.0 Source: KIHBS 2015/6. 104 See Appendix C3 for details on the classification of enterprises as male- female or jointly run. 105 The KIHBS 2015/6 data only collect very limited information at the enterprise-level – i.e. its sector, whether the enterprise is registered with the Registrar of Companies, and male and female labor inputs. For this reason, we do not perform a full decomposition analysis in this section. 76 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Gender and Poverty Differences in profits between male- and female-run 3.3.4 Agriculture household enterprises do not, however, primarily Even though women make up 56 percent of the reflect differences in the distribution of enterprises total population employed in agriculture, they are across sectors, formal registration, labor input or the primary decision maker on only 39 percent urban-rural location. Regressing log profits on a of agricultural plots (Figure 3.23). This reflects, to dummy variable that captures whether the enterprise is some degree, gender differences in land ownership female-run shows that profits of female-run enterprises documented in section 3.2.4, though land ownership are about 52 percent lower than those of male-run and land use rights do not necessarily fall together enterprises (i.e. a coefficient of -0.73, see Table C.5, (Slavchevska et al. 2017; Doss, Kieran, and Kilic 2017; column 1, Appendix C.4). Controlling for enterprise Gaddis, Lahoti, and Li 2018). characteristics (sector, urban, registration and labor input) reduces this to 43 percent lower profits for There are significant differences in input use and female-run enterprises (i.e. a coefficient of -0.57, see cropping choices between male and female farmers Table C.5, column 3, Appendix C.4). In other words, even (Table 3.5). Parcels managed by men are larger, more after controlling for these differences in characteristics, likely to be irrigated and more likely to use fertilizer than female-run enterprises achieve much lower profits parcels managed by women. Likewise, households than male-run enterprises.106 Unfortunately, the KIHBS where the primary decision-maker in agriculture is male data do not make it possible to investigate the role spend significantly more on labor and non-labor inputs of other key enterprise characteristics, such as capital than households where the primary decision-maker is intensity or access to finance, that are often found to female.107 Interestingly, female primary decision-makers contribute to performance gaps between male and appear to be more diversified, cultivating a slightly larger female entrepreneurs (Hallward-Driemeier 2013). number of crops. This is, however, entirely driven by a Figure 3.23: Gender differences in agricultural employment vs. parcel management, 2015/6 a) Agricultural employment b) Agricultural parcel management 43.6% 38.5% 56.4% 61.5% Male Female Male Female Source: KIHBS 2015/6. Note: Agricultural employment shows the male-female composition of the total population employed in agriculture. Parcel management shows the male-female composition of the primary decision-makers (regarding input use and cropping activities) on agricultural parcels. 107 The KIHBS 2015/6 asks for each parcel which household member makes decision on input use and cropping activities and this is used to determine the primary-decision maker. However, many inputs are collected at the household- (e.g. labor and non-labor input cost) or 106 In this respect, the Kenyan results differ from those reported in Hallward- crop-level (i.e. use of improved seeds). To analyze gender differences, Driemeier (2013) for 19 Sub-Saharan African countries, where controlling we distinguish between household where the primary decision-maker for whether the enterprise is registered reduced the gender gap by is male vs. female based on the share of the household’s agricultural land about half. that is being managed by male vs female household members. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 77 Gender and Poverty Table 3.5: Descriptive differences in input use between male and female decision-makers in agriculture, 2015/6 Parcel-level Household-level Male Female Significance Male Female Significance decision- decision- level of decision- decision- level of maker maker gender gap maker maker gender gap Land size (ha) 0.58 0.49 *** 0.80 0.66 *** Share irrigated 0.06 0.04 *** 0.08 0.05 *** Share using inorganic fertilizer 0.62 0.58 *** 0.65 0.60 *** Total input cost (KSh/year) n.a. 9,337 6,291 *** Total labor cost (KSh/year) n.a. 5,256 4,014 *** Number of cash crops n.a. 0.43 0.33 *** Number of food crops n.a. 1.95 2.23 *** Total number of crops n.a. 2.39 2.55 *** Source: KIHBS 2015/6. Note: *** denotes p<0.01 greater number of food crops, as households having trade through small/large traders and millers, while a man as the primary decision-maker cultivate more households where women are the primary farmers cash crops. Overall, these results are consistent with are more likely to sell through consumers, neighbors World Bank (2013a), which provides a more in-depth and cooperatives. Also, a greater proportion of women analysis of gender differences in agriculture using data (22 percent) than men (7 percent) reported that their collected under the Kenya Agricultural Productivity and spouse kept the revenue from crop sales (dried maize), Agribusiness Project. even in cases where women were managing the production of the crop. The study further showed that Households where the primary decision-maker in female farmers are less likely than male farmers to seek agriculture is female achieve, on average, yields advice from extension service providers. that are 15 percent lower for maize and 8 percent lower for beans than households where the primary- 3.3.5 Policies to reduce gender gaps in economic decision maker is male. Maize and beans are the opportunities two most common food crops. Decomposing these The preceding analysis has shown that Kenyan gender differences in yields using the Oaxaca-Blinder women’s decision to participate in the labor force is decomposition method (as described in section 3.3.2) strongly influenced by cultural and religious norms and a regression model similar to the one used in and there is some evidence that gender norms can chapter 4 on agriculture, shows that gender differences be transformed through programs targeting young in endowments (especially household size, as a proxy adolescents (e.g. Lundgren et al. 2013 for Nepal). for household labor availability, use of certified seeds, However, more rigorous empirical evidence is needed and spending on non-labor inputs) explain more than to understand if such programs work in conservative, 70 percent of the gap in maize yields, but only about 20 percent of the gap in beans yields.108 traditional societies, like Northeastern Kenya, where gender gaps in labor force participation are most Gender differences also emerge with respect to prominent. An ongoing evaluation by the Africa Gender trading channels, decision-making power over crop Innovation Lab of the CHOICES program in Somalia income and the use of agricultural extension services. (P165258) will provide further insights in a culturally As highlighted in World Bank (2013a), households similar context.109 where men are the primary farmers are more likely to 109 This is a program designed to transform attitudes and behaviors of 108 The fact that the KIHBS data do not allow assigning household-level very young adolescent girls and boys aged 10-14 years towards greater inputs to specific crops may also play role in explaining the difference in gender equality, which are perceived as markers of future labor market results for maize and beans. decisions. 78 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Gender and Poverty Social protection programs need to pay attention teams. Moreover, the Africa Gender Innovation Lab to the specific vulnerabilities of women who went is evaluating the effectiveness of coding boot camps through a marital dissolution, especially if they are with a focus on gender under the Kenya Industry and also caring for children. As shown in this section, Entrepreneurship Project (KIEP, P161317). A recent female labor force participation is strongly linked gender assessment of the oil and gas sector in Kenya to marital status and the number young of children provides additional recommendations tailored to living in the household. Moreover, section 3.1 showed extractive industries, where women’s participation has that women who went through a marital dissolution traditionally been low (Cardno 2018). (divorce, death of their spouse) are significantly poorer than their male counterparts. The results also reinforce Business training programs hold some promise to the need to invest in care services to allow women with enhance the performance of female-run household children to participate in the labor force. enterprises, though more rigorous empirical evidence would be needed to support the effectiveness of While patterns of sectoral segregation are highly a specific curriculum. Reviews of business training persistent, a few studies suggest that information programs in developing countries have found that, interventions and, possibly, mentoring programs in general, the effectiveness of trainings differs across hold promise. Findings from an evaluation in Western study contexts and curriculums, and is often worse for Kenya of the national vocational training program women than for men (McKenzie and Woodruff 2014). show that information interventions, which emphasize However, a recent evaluation of the International Labour the discrepancies in expected earnings for graduates Organization’s (ILO’s) “Get Ahead” business training of traditionally male-dominated trades (e.g. mechanic) program, has found that the program significantly vs. female dominated trades (e.g. seamstress) can increased the sales and profits of female market encourage women to enroll in male-dominated vendors three years after the intervention (McKenzie professions (Hicks et al. 2016). A study from Uganda and Puerto 2017). The study, which was conducted in of female entrepreneurs who managed to succeed four counties in Western and Eastern Kenya, also did in male-dominated sectors highlights further the not find any evidence of negative spillover effects on importance of mentoring relationships and role models non-treated businesses, as markets as a whole appear (Campos et al. 2015). to have grown in terms of customers and sales volumes as a result of the intervention. Technological change has the potential to disrupt traditional patterns of sectoral segregation. New Empirical evidence from across Africa suggests business models, such as Uber and other ride-hailing that providing access to formal savings products services, can open up opportunities for women in is a promising approach to improve labor market traditionally male-dominated sectors like transportation outcomes of women (Campos and Gassier 2017). In (IFC 2018). Ongoing World Bank activities explore the Kenyan context, Suri and Jack (2016) show that the options to increase women’s participation in Science, rollout of the country’s mobile money system M-PESA Technology, Engineering and Mathematics (STEM) induced women to move out of agriculture into the occupations and may provide additional guidance non-farm enterprise sector, thereby contributing to over the lifespan of this assessment. Specifically, the a reduction in poverty, which was more pronounced project “Women in STEM – Infrastructure” (P166990) among female-headed households. In a similar vein, seeks to collate practical strategies on the recruitment, Dupas and Robinson (2013) find that better access retention and promotion of women in STEM to formal savings products increased productive occupations, specifically in infrastructure sectors, and investments of female entrepreneurs in Western Kenya. to develop a compendium of good practices for project KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 79 Gender and Poverty More research is needed on how to reduce gender 3.4.1 Women’s mobility gaps in agricultural productivity. There is still a lack of Social norms for acceptable behavior often constrain empirical evidence on what interventions are effective women’s physical mobility, i.e. their ability to move in closing these gaps. A few studies suggest that freely beyond the household. In 2014, 22 percent of programs that strengthen women’s property rights over Kenyan women and 19 percent of Kenyan men agreed land and tenure security can increase investment and with the statement that a husband is justified in hitting productivity among female farmers (Goldstein and Udry or beating his wife if she goes out without telling him. 2008; Ali, Deininger, and Goldstein 2014). In addition, an Acceptance of social norms that limit women’s mobility ongoing research project of the Africa Gender Innovation is strongly linked to poverty, with 36 (31) percent of Lab on agricultural labor constraints of female farmers Kenyan women (men) in the poorest quintile agreeing (P166082) might provide useful information, as studies with the above statement, compared with 9 (13) from other Sub-Saharan African countries have shown percent in the wealthiest quintile (Figure 3.24a). Yet, consistently that female farmers’ lack of access to labor social norms are changing rapidly, as the share of the is a key determinant of the gender gap in agricultural population who agreed with the above statement fell productivity (O’Sullivan et al. 2014). by about half between 2003 and 2014 (DHS 2018). Kenyans are also less likely to agree with the above 3.4 VOICE AND AGENCY statement than the population in most other African G ender gaps in endowments and economic opportunities are in many cases a reflection of women’s lack of agency. Agency is the ability to make countries (Figure 3.24b). Constraints on women’s physical mobility curb their decisions about one’s own life and act on them to achieve labor market opportunities and life choices. They not desired outcomes (World Bank 2015a). Differences only directly affect women’s preferences for seeking between men and women’s ability to make these choices, employment outside the home, but also limit women’s usually to the detriment of women, exist in all countries access to education, markets, banks and social networks and cultures. This section zooms in on two expressions and thus affect labor market behavior indirectly of agency –women’s mobility and their freedom from (Chakravarty, Das, and Vaillant 2017). Salon and Gulyani gender-based violence (Klugman et al. 2014). (2010), using data collected in informal settlements in Nairobi in 2004, show that working women are less Figure 3.24: Acceptance of norms that constrain women’s physical mobility a) Share of women/men accepting wife beating b) Share of women/men accepting wife beating, Kenya and by residence and quintile, 2014 comparators if wife goes out without telling her husband if wife goes out without telling her husband 100 40 Percent of population (15-49) Percent of population (15-49) 80 30 60 20 40 10 20 0 0 Guinea Chad Mali Sierra Leone Dem Rep Congo Burundi Ethiopia Niger Gambia Tanzania Congo Uganda Senegal Burkina Faso Zambia Liberia Comoros Cote d'Ivoire Cameroon Nigeria Zimbabwe Rwanda Kenya Gabon Togo Ghana Angola Namibia Lesotho Benin Mozambique Malawi Second Highest Lowest Urban Fourth Rural Middle Total Residence Wealth quintile Male Female Male Female Source: KDHS 2014 (KNBS et al. 2015) and DHS STATcompiler (DHS 2018). Note: Population aged 15-49 years. 80 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Gender and Poverty likely to travel outside their home settlement for work Figure 3.25: Share of women (15-49) who experienced physical violence by marital status, 2014 and, if they do commute, they are less likely than men 70 to use motorized transportation. Mobility constraints 64 can also further increase the time women spend on 60 domestic tasks and hence contribute to time poverty. 50 47 For example, some communities visited for the 2005 40 Percent PPA reported that it was “inappropriate” for women to 32 30 use bicycles or wheelbarrows for fetching water. 24 25 20 12 3.4.2 Gender-based violence 10 In 2014, 45 percent of women aged 15-49 years 0 Never Married/ Divorced/ reported having (ever) experienced physical married living separated/ together widowed violence, while nearly half of all ever-married women Past 12 months Ever experienced at least one form of intimate partner Source: Zumbyte 2018 based on KDHS 2014. violence (IPV, i.e. emotional, physical, or sexual IPV).110 Gender-based violence is a serious violation of women’s to experience IPV, compared to those married younger voice and agency and can lead to reduced mobility, than 15 years, which highlights the importance of less access to economic opportunities and long-term eliminating child marriages. Women who have been in physical and mental health issues – for the women more than one union have about four times the odds themselves, but also their children. of IPV compared with women who have been in just one union, which shows again the vulnerable position Women who have ever been married, and especially of women who underwent a marital dissolution. those who have gone through a marital dissolution, Women who are employed for cash are twice as likely are more likely to have experienced physical violence to experience IPV than women who are not working. than women who have never been married – which In terms of her partner’s characteristics, the risk of a reflects that violence is often perpetrated by current women experiencing IPV declines with the education or former spouses (Figure 3.25). In addition, there is level of her spouse, but increases if her spouse has a strong regional variation – with the highest rates of history of alcohol abuse. Perhaps surprisingly, women’s physical violence being reported in Nyanza, Nairobi education does not have a significant association with and Western regions. the experience of IPV. However, there is evidence from other studies that education may positively affect Multivariate analysis shows that the risk of a women’s attitudes towards domestic violence. For woman experiencing (physical) IPV is linked to example, the evaluation of a merit-based scholarship her age at marriage, whether she remarried and program targeting adolescent girls (discussed in her employment status – though her partner’s section 3.2.1) found that the program led adolescent characteristics play an important role as well. girls to reject the legitimacy of domestic violence Women who marry older than 25 years are less likely (Friedman et al. 2016). 110 This section draws on Zumbyte (2018). It reports standardized measure of gender-based violence (see KNBS et al. 2015 for details). The reported incidence of physical violence declined from 47 percent in 2003 to 39 percent in 2008/9, and then increased to 45 percent in 2014 (DHS 2018). This uneven trend, which may partly reflect reporting behavior, merits further investigation. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 81 CHAPTER 4 AGRICULTURE AND RURAL POVERTY SUMMARY Rural Kenyan households experienced a remarkable decline in poverty over the last decade, independent of their source of income. The proportion of the rural population living below the national poverty line declined from 50.5 percent (14.3 million people) in 2005/06 to 38.8 percent (12.6 million) in 2015/16. Households with diversified income from both agricultural and non-agricultural sources accounted for most of the poverty reduction, followed by agriculture and non-agricultural households. Improved infrastructure, including mobile network access, has raised the welfare of rural households, particularly of those with diversified income. In recent years, mobile network coverage improved substantially in rural areas, enhancing the efficiency of labor and agricultural markets. Improved coverage made it possible for mobile networks to not only serve as a communication tool but also constitute a platform for service delivery in rural areas. This has especially benefited households that rely on both agricultural and non-agricultural income, suggesting that off-farm diversification has been important for poverty alleviation efforts in Kenya over the last decade. Although productivity growth in the production of many crops has been stagnant, increased agricultural productivity remains a potential pathway out of poverty for many households. Little progress has been made in terms of raising productivity in the agriculture sector, especially concerning the production of maize, Kenya’s main food staple, and commercial crops such as coffee. Increased efficiency in the production of beans appears to be the only exception. As a result, agricultural productivity has not been contributing to poverty reduction in rural Kenya, a marked difference from the experience of other countries in the region such as Ethiopia. Nevertheless, more productive farmers are less likely to be poor in Kenya. This correlation between farm productivity and poverty constitutes promising evidence that an improvement in agricultural yields could lead to a reduction of poverty. Agricultural commercialization has helped to improve the livelihoods of Kenya’s farmers. Between 2005/06 and 2015/16, the country’s level of agricultural commercialization increased, and agricultural households sold a higher share of their production. Given that agricultural yields have been stagnant, better access to markets, as a result of infrastructure investments and better access to information and communication technologies, is the likely cause for higher levels of commercialization in the sector. Since farmers that sell a higher share of their products exhibit a lower incidence of poverty, agricultural commercialization is likely having a positive contribution to poverty reduction in Kenya. High commodity prices and increased productivity in the production of bean crops have also contributed to an improvement in the welfare of agricultural households. Many Kenyan farmers have shifted to bean production in recent years, as the country benefited from favorable bean and maize prices in 2011-16. Data suggest that farmers that shifted to bean production were less likely to be classified as poor. However, the increase in crop prices is generally beneficial for Kenya’s net-selling farmers at the expense of the urban poor, as poor urban households spend a large share of their income on food and are therefore sensitive to rising food prices. This may have contributed to the large divergence in poverty reduction between urban and rural areas. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 83 Agriculture and Rural Poverty 4.1 THE DECLINE IN RURAL POVERTY HAS The large drop in rural poverty along with a BEEN THE MAIN DRIVER OF POVERTY stagnating urban poverty rate cannot solely REDUCTION NATIONALLY explained by rural-urban migration. The migration R ural poverty alleviation has been driving Kenya’s of large numbers of poor rural households to urban overall progress in reducing poverty over the last areas can lead to a decline in rural poverty without decade. The country’s national poverty rate declined an actual improvement in livelihoods. Moreover, an from 46.6 percent in 2005/06 to 36.1 percent in inflow of poor households to urban areas can raise 2015/16, driven by a substantial decline in rural poverty, the urban poverty rate. However, while the share of from 50.5 percent to 38.8 percent in the same period Kenya’s population living in rural areas declined by 8 (Figure 4.1). By contrast, urban poverty declined by only percent between 2005/06 and 2015/16, households 2.7 percentage points, from 32.1 percent in 2005/06 to that migrated to urban areas were not from the 29.4 percent in 2015/16. bottom part of the distribution, as will be further explored in the Chapter 5. As a result, factors other While rural poverty has declined across Kenya, the than migration must explain the country’s progress in rate of poverty reduction has varied significantly reducing rural poverty. across provinces. Rural poverty headcount rates varied substantially across provinces in 2005/06, from 31 Some provinces with low rural poverty rates still percent in Central to 74 percent in North Eastern (Figure constitute a large proportion of the rural poor 4.1). Between 2005/06 and 2015/16, the rate of poverty population due to their large population size. For reduction ranged from 23 percentage points in Coast example, while the Rift Valley does not have the to statistically non-significant 3 percentage points in highest rate of rural poverty, due to its large land North Eastern, which remained the province with the size and relatively dense population, the province highest rural poverty rate at 71 percent in 2015/16. accounts for a third of the rural poor population By contrast, Central had the lowest poverty rate at 24 (Figure 4.2). Similarly, a considerable share of the percent in the same period, followed by Eastern at 32 rural poor reside in the Eastern, Western and Nyanza percent and Nyanza at 36 percent. Although Eastern provinces. These three provinces, along with Rift and Nyanza still suffer from poverty levels above the Valley, account for almost 78 percent of the rural poor 2005/06 average, they reduced their poverty rates by population in Kenya. an impressive 20 percentage points and 13 percentage points, respectively, between 2005/06 and 2015/16. Figure 4.1: Rural poverty headcount and its decline by province 80 74 71 71 60 55 52 Percent of population 51 49 50 48 42 42 39 40 36 31 32 24 20 0 Kenya Central Coast Eastern North Eastern Nyanza Rift Valley Western 2005/06 2015/16 Source: Authors’ calculation using KIHBS 2005/06 and KIHBS 2015/16. 84 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Agriculture and Rural Poverty Figure 4.2: Geographic distribution of the Figure 4.3: Share of income from agriculture and non- rural poor in Kenya agricultural sources in rural Kenya Central, Western, 6.7% 8 6 14.9% Coast, 9 8 8.7% 15 21 4 7 Eastern, 13.7% 64 57 Rift Valley, North Eastern, 33.6% 7.0% 2005/06 2015/16 Nyanza, Enterprise income Transfers Services wage 15.5% Industry wage Agriculture Source: Authors’ calculation using KIHBS 2015/16. Source: Authors’ calculation using KIHBS 2005/06 and KIHBS 2015/16. Note: Agriculture income includes return from crop, livestock, and agriculture wage incomes. On the other hand, the Central and Northeastern Although poverty declined among all rural provinces each account for only 7 percent of the total households, independent of their income source, the rural poor population. Despite the prevalence of a rise in welfare among households with diversified high poverty headcount in Northeastern, the province incomes contributed the most to poverty reduction. accounts for only a small share of the total rural poor Kenya’s rural poverty reduction of 11.7 percentage in the county as it is sparsely populated (Figure 4.3). points between 2005/06 and 2015/16 was mainly The Central and Coast provinces have small rural driven by households that continued to derive their populations and the former also has a relatively lower income from just one source (either agriculture or non- poverty rate. As a result, they account for only a small agricultural activities), contributing 10.4 percentage fraction of Kenya’s rural poor population. points (Table 4.1). The poverty rate fell by a mere 0.8 percentage points for households that changed their 4.2 DIVERSIFYING AWAY FROM source of income (e.g., from exclusively agricultural AGRICULTURE IMPROVES LIVELIHOODS income to mixed or non-agricultural income) in the W hile agriculture remains the main source of same period. While, the remainder 0.5 percentage point income for rural households, the share of drop in the poverty rate was attributed to the interaction income from non-agricultural employment has effect, i.e. resulting from, for instance, a population shift increased significantly in the last decade. As a share into a sector that is greatly contributing to poverty of agricultural household income in rural areas, income reduction. Among the different groups of income from crops and livestock as well as wages declined sources, households with diversified incomes—while from 64 percent in 2005/06 to 57 percent in 2015/16 only representing one-third of the rural population— (Figure 4.3). Wage income from service employment is contributed 40 percent of the 10.4 percentage point the second most important source of income in rural decrease in rural poverty in 2006.16, followed by solely areas, increasing from 15 percent of rural household agricultural households at 31.4 percent and exclusively income in 2005/06 to 21 percent in 2015/16, whereas non-agricultural households at 17.6 percent. the share of wage income in industry increased by a mere 3 percentage points in the same period. The share of rural household income from non-farm enterprises and transfers has remained at basically the same level since 2005/06. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 85 Agriculture and Rural Poverty Table 4.1: Decomposition of poverty by income classification 2006 50.48 Headcount rate 2016 38.76 Source of income Pop. share in period 1 Absolute change Percentage change Non-agricultural income only 18.2 .2.1 17.6 Agriculture income only 48.0 .3.7 31.4 Mixed: agriculture and non-agriculture income 33.8 .4.7 40.0 Total intra-sectoral effect .10.4 89.0 Population shift effect .0.8 6.7 Interaction effect .0.5 4.4 Change in headcount rate .11.7 100.0 Note: Agricultural income includes income from wages from agricultural employment, inferred income from the value of crop sales plus the value of own crop consumption, and income from livestock. A household with only agricultural income is defined as having a share of income of more than 90 percent from agriculture. A household with only non-agricultural Income is defined as having a share of income of less than 10 percent from agriculture. Households with incomes in between are defined as mixed. 4.3 NON-AGRICULTURAL EMPLOYMENT IS activities). As a result, the share of rural households’ BECOMING INCREASINGLY IMPORTANT income from non-agricultural sources increased from FOR RURAL HOUSEHOLDS an average of 35 percent in 2005/06 to 42 percent 4.3.1 Households are allocating more time to non- in 2015/16, with the biggest gains in household agricultural activities income in the provinces of Western (39 percent) and W hile agriculture remains the primary sector of employment for rural households, labor time allocated to non-agricultural activities increased Coast (12 percent). The poverty rate among households that depend between 2005/06 and 2015/16. Rural households in solely on agricultural work is higher compared all provinces, except for Coast, spent an average of less to those engaged in non-agricultural activities. than 45 percent of their labor time on non-agricultural Households engaged in non-agricultural activities on a activities in 2015/16, up from below 40 percent in full-time or part-time basis are less often poor compared 2005/06 (Figure 4.4). By contrast, rural households in to households that focus exclusively on agriculture, the province of Coast allocated an average of only 52 a trend that was already visible in 2005/06 but has percent of their time to non-agricultural activities in strengthened since (Figure 4.6). While it is difficult to 2015/16, up from 40 percent in 2005/06. The increase establish a causal relationship, the strong correlation in labor time spent on non-agricultural activities varied between off-farm diversification and lower poverty between provinces, from an increase of 4 percentage rates is suggestive of the fact that households that points in the province of Nyanza to 15 percentage complement agricultural income with non-agricultural points in the province of Northeastern. Also, there was activities are better prepared to face an adverse virtually no change in the allocation of labor time to agricultural shock such as a drought or low prices, and non-agricultural activities in Nyanza. Compared to smooth consumption. At the same time, households 2005/06, fewer households are exclusively agricultural with higher levels of education are less likely to depend (allocating more than 75 percent of labor to agricultural exclusively on agricultural employment. 86 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Agriculture and Rural Poverty Figure 4.4: Changes in rural non-agricultural economic activities a) Non-agricultural labor allocation in rural Kenya 60 53 50 41.5 40 39 Percent of households 40 37 37 35 34 33 31 31 31 31 30 29 28 24 20 10 0 Kenya Central Coast Eastern North Eastern Nyanza Rift Valley Western 2005 2015 b) Distribution of employment by time spent on agricultural and non-agricultural activities 70 63 60 53 50 Percent of households 40 30 20 18 15 13 12 13 12 10 0 [0,0.25] (0.25,0.5] (0.5,0.75] (0.75,1] Proportion of a household's total employment hours spent on agriculture 2005-2006 2015-2016 c) Proportion of income earned from non-agricultural sources 70 62 60 54 50 50 Percent of households 45 42 43 40 38 40 35 37 34 33 35 35 31 33 30 20 10 0 Kenya Central Coast Eastern North Eastern Nyanza Rift Valley Western 2005/06 2015/16 Source: Authors’ calculation using KIHBS 2005/06 and KIHBS 2015/16. Note: Female employment in non-agricultural activities as a share of total female employment has increased significantly in the Coast (by 20 percentage points) and North Eastern (by 19 percentage points) provinces (Figure 4.5). There has been little change in the remaining provinces since the previous survey was conducted in 2005/06, and the share of female employment in total non-agricultural employment even decreased in Rift Valley, Eastern, and Western. This suggests that most of the increase in non-farm employment has been concentrated among men in Kenya, which can potentially have adverse consequences for intra-household equality. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 87 Agriculture and Rural Poverty Figure 4.5: Female non-agricultural labor allocation Figure 4.6: Rural poverty rate by the proportion of total employment in agriculture 60 70 60 50 48 43 50 40 Poverty rate, % 40 Percent 30 29 29 28 25 27 30 23 23 25 24 23 22 21 20 20 18 20 10 10 0 0 0 0.2 0.4 0.6 0.8 1 Kenya Central Coast Eastern North Nyanza Rift Western Proportion of labor hours spent engaged in agriculture Eastern Valley 2005 2015 2005/06 2015/16 Source: Authors’ calculation using KIHBS 2005/06 and KIHBS 2015/16. Source: Authors’ calculation using KIHBS 2005/06 and KIHBS 2015/16. 4.3.2 Wage employment within the service sector has Moreover, wholesale and retail trade is the most increased significantly important non-agricultural industry in terms of Income from wage employment in the services sector employment (Figure 4.8). About 17 percent of rural represents the largest share of non-farm household Kenyan households had at least one family member income in the Kenya (Figure 4.7). While it has increased who worked in wholesale and retail trade in 2015/16, for both poor and non-poor households since 2005/06, a more than threefold increase compared to 2005/06. it constitutes a larger share of the income of non-poor Similarly, employment in transport and communication households. In rural Kenya, the share of wage income also increased threefold, from 2 percent to 6 percent of from the services sector in total household income rural households having one family member employed increased from 15 percent in 2005/06 to 21 percent in the industry in the same period. However, there in 2015/16, reducing the share of agricultural income. was only a slight increase in the employment rate However, agricultural income still remains the most in community, social, and personal services (which important income source for both poor (64 percent) mainly includes public and private sector employment and non-poor (53 percent) households. in education, health, and administration) between Figure 4.7: Share of income from different sources for poor and non-poor households 9 7 7 4 9 8 9 8 11 15 19 24 4 7 4 7 69 64 61 53 Non-poor Poor Non-poor Poor 2005/06 2015/16 Agriculture Industry wage Services wage Transfers Enterprise income Source: Authors’ calculation using KIHBS 2005/06 and KIHBS 2015/16. 88 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Agriculture and Rural Poverty Figure 4.8: Non-farm economic activity by ISIC classification a) Participation of households in non-farm employment by industry b) Proportion of salaried non-farm income by industry 20 19 50 50 Percent of total salaried employment income, % 17 40 15 Percent of households 13 34 33 30 10 7 20 18 6 6 5 12 12 5 10 10 3 3 10 9 10 2 0 0 Mining & Construction Wholesale & Transport, Community, Mining and Construction Wholesale and Transport, Other service Manufacturing Retail Trade Storage and Social and manufacaturing retail trade+food storage & activities Communication Personal Services Service activities communication 2005/06 2015/16 2005/06 2015/16 Source: Authors’ calculation using KIHBS 2005/06 and KIHBS 2015/16. 2005/06 and 2015/16. Also, employment rates rose in groundnut varieties increased household crop income the mining and manufacturing, construction, trade, by US$130.$254 and improved the chance of a and transportation and communication industries household escaping poverty by 7.9 percentage points during the same period, even though their individual (Kassie, Shiferaw and Muricho, 2011).112 Gains were shares have remained relatively low.111 greater for households with a relatively smaller farm size and for more educated households. Finally, a quasi- 4.4 FARM PRODUCTIVITY HAS STAGNATED experimental study by Davis et al. (2012) demonstrated WHILE COMMODITY PRICES HAVE the importance of learning about improved farming INCREASED practices among small-scale farmers in East Africa. 4.4.1 Higher productivity is associated with lower It showed that farmer field schools contributed poverty rates to increased crop productivity, resulting in higher S everal studies of African countries show a causal link between improved agricultural productivity and reduced poverty rates. A meso-level study of household income and an improvement in farmer welfare. While the productivity and income of female-headed households increased significantly, village-level data in Madagascar shows that communes they increased only marginally for male- that adopted agricultural technologies at a higher rate, headed households. Moreover, the effects were and subsequently had higher crop yields, enjoyed lower concentrated among households with little formal food prices, had higher real wages for unskilled workers, education, presumable because these households and exhibited better welfare indicators (especially lower had the most to gain from such training programs. extreme poverty rates, Minten and Barrett 2008). This This section presents some indicative evidence suggests that an increase in agricultural productivity that this causal relation between agricultural can raise incomes for surplus farmers, reduce prices for productivity and poverty reduction. However, it consumers, and increase employment opportunities should be noted that they represent correlations, and wages for unskilled workers. Similarly, another not necessarily causal, between higher crop yield study in Uganda found that adopting improved and increased household welfare. 111 KIHBS 2005/06 uses ISIC Revision 2 to classify employment by subsector, whereas KIHBS 2015/16 uses ISIC Revision 4. Appropriate steps have 112 The authors of the studies eliminated selection bias on observable been taken to ensure correspondence of industrial classification. differences between adopters and non-adopters. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 89 Agriculture and Rural Poverty Figure 4.9: Relationship between crop yield and poverty rates at the provincial level in rural Kenya, 2015/16 a) Maize yield and poverty b) Bean yield and poverty 2300 600 Rift Valley Eastern 2100 550 1900 Rift Valley Yield (kg/hectare) Yield (kg/hectare) 1700 Western 500 1500 Central 450 1300 Nyanza Central Eastern 1100 400 Nyanza North Western 900 Eastern 350 Coast North Coast 700 Eastern 500 300 15% 25% 35% 45% 55% 65% 15% 25% 35% 45% 55% Poverty Rate Poverty Rate Source: Authors’ calculation using KHIBS 2015/16. Figure 4.10: Poverty and crop yield at the county level in rural Kenya, 2015/16 a) Maize yield and poverty b) Bean yield and poverty 3500 1400 3000 1200 2500 1000 Yield (kg/hectare) Yield (kg/hectare) 2000 800 1500 600 1000 400 500 200 0 0 0% 10% 20% 30% 40% 50% 60% 70% 0% 10% 20% 30% 40% 50% 60% Poverty Rate Poverty Rate Central Coast Eastern Nyanza Central Eastern Nyanza Rift Valley Western Linear (Kenya) Rift Valley Western Linear (Kenya) Source: Authors’ calculation using KIHBS 2005/06 and KIHBS 2015/16. Figure 4.11: Relationship between yield decile and poverty rates in rural Kenya, 2015/16 a) Maize b) Beans 60% 55% 50% 45% 40% 35% Poverty Rate Poverty Rate 30% 25% 20% 15% 10% 5% 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Yield decile Yield decile Central Eastern Nyanza Rift Valley Western Central Eastern Nyanza Rift Valley Western Source: Authors’ calculation using KIHBS 2015/16. 90 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Agriculture and Rural Poverty The evidence suggests that an improvement in farm instance, the poverty rate among households in the productivity could potentially reduce poverty in lowest maize yield decile is 49 percent, compared to Kenya. While agriculture was not the main driver of only 22 percent for those in the highest (10th). poverty reduction in rural Kenya between 2005/06 and 2015/16, an increase in crop yield could significantly 4.4.2 Stagnating productivity means that there is an reduce poverty, as agricultural productivity is strongly unmet potential for rural farmers and negatively correlated with poverty rates at the Almost 85 percent of Kenya’s cultivated land was provincial, county, and household level. Provinces with devoted to growing maize and beans in 2015/16. higher maize and bean yields have generally lower Bean production increased significantly in cultivated poverty rates (Figure 4.9). Similarly, a comparison of areas: from 27 percent of total crop areas in 2005/06 to counties within a given province shows that counties 37 percent in 2015/16 (Figure 4.12). However, there were with higher farm productivity have much lower poverty only minor changes in the share of land allocation for rates (Figure 4.10). all other crop categories. Approximately half of Kenya’s total crop area was devoted to maize production for Most farm households with high crop yields appear both years. The remainder of this section will focus on to have escaped poverty in Kenya. In each Kenyan maize and bean yields, the two most commonly grown province, households in a higher yield decile tend to staple crops in Kenya. have lower poverty rates (Figure 4.11). In Rift Valley, for Figure 4.12: Proportion of cultivated area by crop category in rural Kenya a) KIHBS 2005/06 b) KIHBS 2015/16 2%2% 3% 4% 3% 2% 4% 5% 37% 48% 51% 27% 6% 6% Maize & cereal Tubers & roots Maize & cereal Tubers & roots Beans, legumes & nuts Fruits & vegetables Beans, legumes & nuts Fruits & vegetables Tea & co ee Other cash crops Tea & co ee Other cash crops Other crops Other crops Source: Authors’ calculation using KHIBS 2015/16. Figure 4.13: Maize and bean yield in selected African countries a) Maize yields in selected African countries, 2005–16 6000 5000 4000 Kg/hectare 3000 2000 1000 0 2005 2016 Burundi Kenya Malawi Rwanda South Africa Uganda Tanzania Ethiopia KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 91 Agriculture and Rural Poverty b) Bean yield in selected African countries (2005-2016) c) Bean yield in Kenya since 2000 1800 1,668 1,667 700 1,589 1,530 1600 1,487 1,471 650 1,434 1,416 1,401 1,382 1,347 1400 1262 600 1199.9 1,225 1,185 1,169 1,147 1200 1,069 1,043 550 1,035 1,002 1,001 1,000 Kg/hectare Kg/hectare 997 947.7 968 967 1000 916 891 500 877 846 791 780 762 800 718 700 450 659 622 588 615 585 567 557 600 534 508 484 400 413 370 400 350 200 300 0 1995 2000 2005 2010 2015 2020 Kenya South Africa Tanzania Ethiopia 2000-2009 2010-2016 Source: Author’s calculation based on FAO data. In recent years, Kenya’s agricultural productivity Tegemeo panel household data survey, collected has been low and stagnant compared to that of between 2000 and 2010,113 and both waves of the neighboring countries, except for bean crops. Since KIHBS household data (Figure 4.14). In contrast, bean 2005, maize yields in Kenya have been stagnant at a yield increased by approximately 50 percent between relatively low level compared to many of its neighbors, 2010 and 2016, according to the FAO (Figure 4.13).114 according to cross-country yield data from the Food and Agriculture Organization (FAO) (Figure 4.13). Other There are important differences in yield levels across countries, such as Ethiopia, Malawi, Rwanda, and provinces. Maize yield is multiple times higher in Rift Uganda, have experienced varying levels of productivity Valley than in North Eastern, the latter which has a growth. The level of maize yield in South Africa, which high and persistent poverty rate (Figure 4.14). Maize is indicative of capital- and input-intensive farms, yield is also low in Coast, which is likely explained by illustrates the tremendous potential for Kenyan farmers the high share of non-agricultural employment in the to increase their crop productivity and raise their living province. By contrast, heterogeneity of bean yield is standards. The stagnation in maize productivity over less pronounced, with Eastern and Rift Valley provinces the period 2005-2016 seems to be confirmed by the having relatively higher yields than Kenya’s other provinces. Figure 4.14: Heterogeneity in crop productivity across provinces in rural Kenya a) Maize a) Beans 2500 1000 2,268 2,150 648 2000 1,738 1,683 750 1,532 1,493 Yield (kg/hectare) Yield (kg/hectare) 1,532 1,474 577571 1500 559 1,301 502 535 531 1,260 1,215 485 1,191 885 500 435 383 346 353 388408 383 380 1000 780 767 594 250 500 0 0 Kenya Central Coast Eastern North Nyanza Rift Western Kenya Central Coast Eastern North Nyanza Rift Western Eastern Valley Eastern Valley 2005/06 2015/16 2005/06 2015/16 Source: Authors’ calculation using KHIBS 2005/06 and KIHBS 2015/16. 113 Tegemeo data are of households in some parts of Kenya and not in the entire country. 114 This trend in however not observed in the KIHBS data. 92 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Agriculture and Rural Poverty Nationally, female headed households have lower while in bean cultivation, this difference amounts to productivity in both beans and maize crops (Figure over 15 percent. However, there is heterogeneity across 4.15). Female headed households have 10 percent lower provinces, with statistically insignificant differences maize yields compared to male headed households, observed in maize cultivation in the Rift Valley, Eastern, Figure 4.15 : Heterogeneity in crop productivity by gender of household head a) Maize a) Beans 2500 700 2,189 2,144 582 2,005 600 544 546 2000 1,927 519 469 1,325 500 455 1,645 444 381 429 1,490 1,439 382 1500 1,383 368 384 377 Kg/hectare Kg/hectare 1,282 1,250 1,300 400 328 1,219 300 1000 700 590 200 421 500 100 0 0 Kenya Central Coast Eastern North Nyanza Rift Western Kenya Central Coast Eastern Nyanza Rift Western Eastern Valley Valley Male Female Male Female Source: Authors’ calculation using KHIBS 2005/06 and KIHBS 2015/16. Figure 4.16: Gender differences in input use in rural Kenya a) Land area cultivated by households b) Total input costs (excluding labor) per acre 2.9 11,273 3 12,000 2.4 2.3 2.5 2.1 10,000 1.9 2.0 1.8 7,591 2 8,000 1.6 6,040 1.4 1.4 5,554 KSh./acre 5,218 Acres 1.5 1.2 1.2 5,262 6,000 4,403 1.1 1.1 4,318 4,317 3,587 3,043 1 4,000 2,516 2,893 1,886 0.5 2.000 0 0 Kenya Central Coast Eastern Nyanza Rift Western Kenya Central Coast Eastern Nyanza Rift Western Valley Valley Male Female Male Female c) Inorganic fertilizer spent per acre d) Labor costs per acre 8000 6,961 10000 7,500 9000 7000 5,392 8000 6000 7000 5,675 3,976 5000 5,591 KSh./acre 6000 KSh./acre 3,380 3,425 3,546 4000 3,038 5000 2,748 3,967 2,435 4000 3,463 3000 2,095 3,048 1,726 3,070 1,615 2,751 1,353 3000 2,446 2000 1,719 2,114 1,596 1,907 2000 367 314 1,307 1,286 1000 1000 0 0 Kenya Central Coast Eastern Nyanza Rift Western Kenya Central Coast Eastern North Nyanza Rift Western Valley Eastern Valley Male Female Male Female ource: Authors’ calculation using KHIBS KIHBS 2015/16. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 93 Agriculture and Rural Poverty and Western provinces and in bean cultivation in D). More specifically, to investigate the determinants of Western and Eastern provinces. The differences in crop yield, we apply a fixed effects model. In this model, productivity are partly explained by differences in we start with a basic specification where logarithm of the use of yield enhancing inputs. Farm households per acre yield is regressed on fixed effects of household headed by women use inputs less intensively than and a vector of household characteristics. male-headed households, as they spend less on yield- enhancing inputs such as inorganic fertilizer (Figure Technology adoption is the main factor associated 4.16). While these households also have slightly lower with improvements in maize yield. Households that labor costs, reflecting lower labor inputs, differences are applied chemical fertilizer, for example, experienced a only statistically significant in Eastern. 20.25 percent increase in maize yield. Moreover, farmers who planted improved maize seeds experienced 4.4.3 Improved technologies are the key drivers of 26.32 percent higher productivity compared to agricultural productivity those that used traditional low-yield seeds. However, The adoption of improved farming technologies farmers who used both chemical fertilizer and planted and practices can increase agricultural productivity improved maize seeds did not appear to have higher and reduce rural poverty among small farmers. This maize yield relative to those who applied these section examines what factor are associated with high inputs individually. While the application of chemical crop productivity at the household level in Kenya using fertilizer is positively associated with higher bean the Tegemeo panel dataset115 for 2000-10 (see Appendix yield, the yield increase is negligible. Figure 4.17: Trends in input use by farmers (Tegemeo Panel) a) While the share of farm households that apply chemical fertilizer b) The share of households that use improved seeds for maize increased is high, it has increased only moderately since 2000 in the most recent survey round 100 100 79.1 80 76 75 80 70 72 67.6 65.7 68.1 60 60 Percent Percent 40 40 20 20 0 0 2000 2004 2007 2010 2000 2004 2007 2010 Source: Author’s calculation based on Tegemeo Panel Household Survey (2000-2010). 115 The Tegemeo Rural Household Panel Data cover the years 2000, 2004, 2007, and 2010. The data were collected in 22 rural districts across the country. Stratified simple random sampling was used to create the sample of households. After assigning agro-ecological zones (AEZ) to each rural division, 2-3 divisions were selected in each AEZ based on their population size. Villages within selected divisions and households within selected villages were picked through a blind equal chance ballot. A total of 1,446 sampled households were interviewed in 2000, 1,397 in 2004, 1,342 in 2007, and 1,304 in 2010. The rate of household attrition was 9.8 percent between 2010 and 2000. Households that were overlooked during the interview process were not replaced and efforts were made to interview them in subsequent surveys. 94 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Agriculture and Rural Poverty Despite the yield-enhancing effects of fertilizer, There is a positive relationship between the adoption the share of households that applied chemical of improved seeds and maize yield. The application fertilizer did not increase much between 2000 and of certified seeds is strongly associated with maize 2010. In the Tegemeo panel, which only covers parts productivity. However, the opposite is true for bean of Kenya, more than 70 percent of farmers apply productivity, a result attributable to the small number fertilizer on their maize plots. However, the share of farmers that use certified seeds for beans (less than 10 percent) compared to maize (close to 70 percent). of farmers that use fertilizer has not changed much since 2000 (Figure 4.17a). By contrast, the share of An analysis of the relationship between crop yield farmers that use improved seed varieties increased and plot size shows that, even after controlling by more than 10 percent between 2000 and 2010, to for technology adoption and other household almost 80 percent of maize farmers in 2010 (Figure characteristics, smallholder farmers are more 4.17b). It is worth noting however, there is very productive than large farmers. Columns 2 and 3 of limited use of improved seeds for other crops. Table 4.2 show the relative productivity of maize Table 4.2: Determinants of maize yield, FEs model, 2000–10 (1) (2) (3) 0.21 *** 0.20 *** Fertilizer adoption (Yes=1) (3.77) (4.16) 0.26*** 0.28*** Improved seed adoption (Yes=1) (6.30) (6.89) .0.00 0.00 0.00 Distance to extension services (.0.21) (0.05) (0.90) 0.05 0.07* 0.06 Cooperative/Group membership (1=yes) (1.32) (1.89) (1.63) Cropped land quartile (the lowest quartile is the reference group): 0.00 0.00 .0.17*** .0.19*** 2nd quartile (.3.80) (.4.22) .0.38 *** .0.41*** 3rd quartile (.9.62) (.9.91) .0.69 *** .0.68*** 4th quartile (.14.20) (.12.93) Effectiveness of fertilizer on improved seed 0.00 No improved seed x Fertilizer used (0.83) 0.00*** Improved seed x Fertilizer used (3.25) 5.64*** 5.79*** 6.43*** Constant (18.07) (19.08) (18.44) Observations 4897 4897 3996 Source: Author’s calculation based on Tegemeo Panel Household Survey (2000-2010). Note: Standard errors in parentheses: * p < 0.1, ** p < 0.05, *** p < 0.01. Note that the dependent variable is logarithm of yield (kg/acre). A vector of household characteristics including: gender, age, age squared and education of household head, household size and dependency ratio. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 95 Agriculture and Rural Poverty farmers with plot size in the upper three quartiles and initiatives can help bridge the productivity gap. An compared to those with plots in the lowest quartile. increase in agricultural productivity, as demonstrated The production of large maize farmers in the highest in the previous section, could significantly reduce landholding quartile is 69 percent lower per acre poverty among farm households. That is why the compared to those in the lowest quartile. This inverse announcement of having food security and agricultural relationship between plot size and maize yield is productivity as one of the main four priorities (the Big 4) persistent, and the productivity gap increases from of the GoK is welcome news. 17 percent to 38 percent and 69 percent as plot size quartile ranking increases from 2nd to 3rd and 4th Policymakers may need to allocate more resources quartile, respectively. Similarly, small bean farmers are to enhance farmers’ productivity and make sure more productive than large bean farmers (see Table D.1 that the current spending is efficient and providing in Appendix D). the highest returns. The recently published Kenya Economic Updates noted that only 2 percent of total The inverse relationship between plot size and public expenditure was allocated to agriculture in maize yield is not unique to rural Kenya, as it has 2016/17, even though the sector accounts for 25 been observed in several developing countries and percent and 60 percent of the country’s GDP and confirmed by various studies.116 The relationship employment, respectively. This prevents the country’s is contrary to economic theory, which states that from investing effectively in smallholder agriculture factor productivity must be equal across farms, as and provide services to improve basic crop yield such land would be sold or leased from farmers with lower as extension services, improved seeds and seedlings, marginal productivity to farmers with higher marginal irrigation, etc. There is also a need to assets if the productivity.117 Some of the most common and current spending is efficient, taking into account that plausible explanations for this inverse relationship spending on public goods in this context (e.g. research relate to market imperfections. First, smallholder and development, extension services, etc.) has been farmers face an imperfect labor market and continue proven to be more productive than spending on to excessively use labor on their small plots. Second, private goods (e.g. fertilizer subsidies). In addition an imperfect insurance and crop market forces risk- there is space to reform the input subsidy program by averse small farmers to work more hours than optimal ensuring that the program is targeting small farmers to secure enough food from their plots.118 and facilitating technology adoption among them. Moreover, investment in irrigation schemes have a 4.4.4 Policies that promote investments in high rate of return119 and could reduce dependence productivity-enhancing technologies are vital for farmers on rainfall. Currently, only 2 percent of Kenya’s total arable land is irrigated, compared to 6 percent in Investment in productivity-enhancing technologies Sub-Saharan Africa, and most of the country’s crop such as fertilizer, improved seeds, and agricultural production is rainfed. extension services, as well as irrigation, is critical to increase the productivity and welfare of Kenya’s 4.4.5 An increase in grain prices since 2005 may have farmers. There is a huge potential to facilitate poverty helped reduce rural poverty reduction through increase in agriculture income (crop In the absence of major crop-enhancing productivity and wage income). The fact that Kenya’s current level investments, higher crop prices can reduce the of crop productivity is lower compared to that of its poverty rate among farm households. Due to a lack of neighboring countries, signals that public investment farmgate price data to analyze changes in crop prices, market price data are used as a proxy. An analysis of Barrett et al. (2010) summarizes a list of studies, including Chayanov 116 market price data reveals that crop prices had been (1926) and Sen (1962), that have noted this inverse relationship. In addition, a recent study by Ali and Deininger (2015) also found similar increasing at a similar rate as general prices through results in Rwanda. the period 2005 to 2011. Figure 4.18 shows the nominal Barrett et al. 2010. 117 Barrett et al. 2010; Ali and Deininger 2015. 118 119 World Bank 2018b. 96 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Agriculture and Rural Poverty Figure 4.18: Trends of crop prices and overall prices a) Maize prices, 2006–16 6,000 5,000 4,000 3,000 2,000 1,000 0 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 Nyanza Coast Nairobi Rift Valley KIHBS implied price change* *Price increase of the average price across all provinces in the first year (2006) using price inflator implied by the numerical change in the poverty line in KIHBS between 2005 and 2015. b) Bean prices, 2006–16 10,000 9,000 8,000 7,000 6,000 5.000 4.000 3.000 2,000 1,000 0 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 * Price increase of the average price across all provinces in the rst year (2006) using price in ator Nyanza Coast * Price increase of the average price across all provinces in the first year (2006) using price inflator Source: Author’s calculation using FEWS NET data. price and the estimated trend if crop prices would have 4.5 INCREASED MARKET PARTICIPATION followed the overall inflation pattern.120 Data show CAN FURTHER REDUCE RURAL POVERTY I that maize and bean prices were significantly above mproving access to markets for rural households has their estimated trends in 2011-15, which coincides been a key policy goal for Kenyan policymakers. with the period directly prior to KIHBS 2015/16. As Easier access to markets allows rural households to higher crop prices generally tend to benefit rural areas improve their productivity by facilitating access to (which produce crops) at the expense of urban areas agricultural inputs (such as fertilizer and improved (which consume crops), these trends in crop prices may seeds), and by enabling them to sell their production explain why rural poverty rates have declined more more easily and at more competitive prices. The Kenya dramatically than poverty rates in urban areas. This Vision 2030 calls for investment in rural infrastructure assertion, however, is based on the strong assumption to improve accessibility to all of the country’s villages. that increases in grain prices are passed onto farmers in This section explores the extent to which market the form of higher farmgate prices. participation among farm households and production 120 This “average trend” is constructed by first calculating the average prices for markets are associated with poverty. of maize and beans in 2006 and applying the general inflation (living cost adjustment in KHIBS) between 2005/06 and 2015/16 to reconstruct expected linear trend in average prices of maize and bean. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 97 Agriculture and Rural Poverty Previous studies show that commercialization has cash crops such as coffee, tea, fruits, and vegetables improved the welfare of households in rural Kenya. produced by smallholder farmers are sold in the market. Rao and Qaim (2011) found that smallholder farmers that participated in supermarket supply channels The low level of commercialization in Kenya’s witnessed a substantial gain in income and improved agricultural sector reflects the prevalence of their welfare. However, institutional support is needed subsistence farming instead of cultivating specialized to realize the benefits of market orientation and crops for the market. In 2015/16, about 60 percent of connect farmers to consumers. Another study found households did not sell any of their produce in the that participation in farmer cooperative organizations market, while only 4 percent of households sold all their also increased the income and welfare of Kenyan crop production and engaged is purely commercial farmers. Cooperative membership increases income agriculture (Figure 4.19). The low level of agricultural by facilitating access to better input and output prices commercialization in Kenya may be due to limited as well as helping farmers adopt new technologies access to land and/or markets. (Fischer and Qaim 2012).121 Finally, Barrett (2008) underscores that reducing the costs of intermarket Nonetheless, there has been a clear trend toward commerce and improving the access of poorer market orientation in Kenya’s agricultural sector, as a households to improved technologies and productive higher share of farmers sells their own produce. This assets are central for smallholder farmers to participate is demonstrated by the upward shift in the cumulative in markets and escape poverty. distribution function of the proportion of own crops consumed between 2005/06 and 2015/16 (Figure 4.19). Kenyan rural households are less likely to produce Although the proportion of households that either major staple crops, such as maize and beans, to sell consume or sell all their farm produce remained almost in the market. 40 percent of all maize produced in unchanged, there is an increase in the proportion of Kenya by small farmers is consumed by households households that sell more of their production. For themselves, an indication of a moderately high level example, the proportion of households that consumed of subsistence among small farmers. Similarly, around more than 50 percent of their produce decreased from 45 percent of beans and legume is consumed, while 70 percent in 2005/06 to approximately 62 percent in 55 percent is sold in the market. As expected, most 2015/16. Figure 4.19: There was an observed reduction in subsistence agriculture in rural Kenya between 2005/06 and 2015/16 a) Proportion of own crops sold b) Proportion of crop production sold 100 CDF of share of crop production sold 1.0 80 0.8 60 Percent 0.6 40 0.4 20 0.2 0 0.0 Maize & Tubers & Beans, Fruits & Tea & 0 0.2 0.4 0.6 0.8 1 cereals roots legumes & vegetables co ee nuts Proportion of own crop sold 2005/06 2015/16 2005/06 2015/16 Source: Authors’ calculation using KIHBS 2005/06 and KIHBS 2015/16. 121 98 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Agriculture and Rural Poverty Figure 4.20: Relationship between poverty and market participation a) Poverty headcount rate by crop category in rural Kenya 60 55 48 50 50 45 37 38 40 37 33 Poverty Rate, % 30 30 25 20 10 0 Maize & other cereals Tubers & roots Beans, legumes & nuts Fruits & vegetables Tea & co ee 2005/06 2015/16 Note: Major crop grown by households, based on share of cropped land, is used to classify households into these groups. b) Poverty and the sale of farm produce in rural Kenya 70 60 50 40 Poverty rate, % 30 20 10 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Proportion of harvest sold 2005/06 2015/16 Source: Authors’ calculation using KIHBS 2005/06 and KIHBS 2015/16. The type of crop grown and the level of agricultural than the 27 percent among households engaged in commercialization are highly associated with purely commercial agriculture. poverty outcomes. Poverty incidence is much higher among households that are predominantly engaged Farmers that produce beans and legumes have in the production of staple crops such as maize and escaped poverty at a higher rate than those that beans (Figure 4.20).122 Farmers that cultivate mainly produce maize and serials. Poverty among beans and cash crops such as fruits, vegetables and tea and leghum producers declined from 50 percent in 2005 to coffee have lower poverty rates. The poverty rate 36 percent in 2015. The arable area devoted to bean is about 45 percent among maize and other cereal cultivation increased significantly in Kenya between producers, compared to only 25 percent among fruit 2005/06 and 2015/16. Bean prices also increased more and vegetable producers and 30 percent among coffee than maize prices in the same period. Moreover, bean and tea producers. Moreover, the poverty rate is much yields appear to have increased as well, according higher among households engaged in subsistence to FAO data, which suggests that bean farming is agriculture. About 56 percent of households that becoming increasingly popular and may have improved consume all of their produce are poor, much higher livelihoods of poor agrarian households. 122 The major crop is defined based on the proportion of households’ land dedicated to each crop. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 99 Agriculture and Rural Poverty Kenyan policymakers can reduce poverty by constitutes promising evidence that an enhancing reducing the level of subsistence farming, increasing agricultural yields could lead to a reduction of poverty. agricultural commercialization, and helping farmers However, little progress has been made in terms of access markets. Farm households that sell a larger share raising agricultural productivity. This is especially of their produce, tend to have lower poverty levels. In true for the production of maize, Kenya’s main food order to increase commercialization, access to markets staple, and commercial crops such as coffee. Increased where farmers can buy vital inputs to grow their crops efficiency in the production of beans appears to be the and sell their output, both at competitive prices, should only exception. As a result, agricultural productivity has be priority for the sector. not been contributing to poverty reduction in rural Kenya, a marked difference from the experience of 4.6 CONCLUSIONS other countries in the region such as Ethiopia. W hile agriculture remains the main source of income for rural households in Kenya, the share of income from non-agricultural employment and non- Technology adoption is the main factor associated with higher productivity, according to analysis agricultural employment has increased significantly in using farm level data. Farmers that applied chemical the last decade. Income from crops and livestock as fertilizer, for example, experienced a 20-25 percent well as wages in the agricultural sector, declined from increase in maize yield. Moreover, farmers who 64.0 percent in 2005/06 to 57 percent in 2015/16. Wage planted improved maize seeds experienced 26-32 income from service employment is the second most percent higher productivity compared to those that important source of income in rural areas, increasing used traditional low-yield seeds. Despite the yield- from 15 percent of rural household income in 2005/06 enhancing effects of fertilizer and seeds, the share of to 21 percent in 2015/16. This diversification of income, farmers adopting these inputs has not changed much in which households complement agricultural income between 2000 and 2010. Policies aimed at increasing with income derived from non-agricultural activities the adoption of improved agricultural inputs by small (particularly in services and trading activities) has been farmholders would help to increase their income and key to the reduction of rural poverty in Kenya. While help to further reduce poverty. Extension services agriculture remains the primary sector of employment programs and educational campaigns, together with for rural households, labor time allocated to non- a competitive inputs markets, are some examples. agricultural activities increased between 2005/06 and 2015/16. It is important to support rural households Similarly, agricultural commercialization is also in their effort to diversify their income. Investments in associated with better living conditions in the case human capital, skills formation, as well as encouraging of Kenyan farmers. Between 2005/06 and 2015/16, non-agricultural economic activities in rural areas, are the country’s level of agricultural commercialization key areas of actions in which the GoK should focus. increased, and agricultural households sold a higher share of their production. Moreover, a higher degree Although the productivity of many crops has been of commercialization is associated with higher stagnant for the ten years, increased agricultural living standards. Thus, investments in infrastructure productivity remains a potential pathway out and access to information and communication of poverty for many households. In Kenya, more technologies, are an important policy areas to further productive farmers are less likely to be poor. This reduce poverty in Kenya. correlation between farm productivity and poverty 100 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead CHAPTER 5 URBANIZATION KEY MESSAGES Poverty is increasingly becoming an urban phenomenon in Kenya, which will require poverty alleviation efforts that focus on urbanization and urban poverty. The rural poor population decreased by more than 1 million during the last decade, as the rural poverty rate fell from 50.5 percent in 2005/6 to 38.8 percent in 2015/16. By contrast, the poor population in urban areas increased by 1.5 million, with only a marginal decline in the urban poverty rate. As a result, the share of the urban poor increased from 13.8 percent of the total poor population in 2005/06 to 23.1 percent in 2015/16. The share of the urban poor and the inadequate living standards of poor households have remained relatively constant during the last decade. Urban poverty rates have remained at roughly the same level in most provinces since 2005/06. Excluding Nairobi, Kenya’s urban poverty rates have been similar to poverty rates in rural areas. While the share of the urban population with access to improved sanitation facilities and electricity has increased in all provinces, the share of households with access to improved water has dropped in some provinces, suggesting that the country’s water infrastructure is struggling to cope with the pace of urbanization. Moreover, the gap in access to basic services between the poor and non-poor remains wide in urban areas. Rising food and housing costs have constrained household finances in urban areas. While poor urban households allocate half or more of their budgets to food, food expenditures of non-poor households have also increased in urban areas—a likely result of the increase in the relative price of food. In addition, the share of private spending on housing has increased in medium- and small-sized towns, reflecting increased urbanization and a rise in the cost of living. These financial constraints have limited household spending on transport and other services, lowering households’ access to economic opportunities and the ability to generate capital. Urban unemployment has dropped dramatically in recent years, but many workers remain in insecure jobs. Unemployment rates have dropped throughout urban areas in tandem with increasing labor force participation rates. However, a large portion of the urban poor, women, and the youth are unemployed. In Nairobi, for example, more than a fifth of the poor are unemployed. There has been an increase in construction jobs in urban areas, which has raised the income of lower-income households. Nevertheless, a large portion of workers in Kenya are in insecure positions as casual workers. There is also a lack of manufacturing jobs in urban areas. Poverty—both monetary and non-monetary—is still concentrated in Nairobi’s informal settlement neighborhoods. In Nairobi, which is home to nearly two thirds of Kenya’s population that lives in informal settlements, nearly one-third of residents in informal settlement neighborhoods are poor, while only 9.1 percent of residents in non-informal settlement areas are poor. The gap in living standards, such as housing quality, access to services, environmental challenges, and health, is wide between poor and non-poor households. Residents in informal settlement areas also live far away from jobs, which can further lower their economic performance, and they have limited opportunities to move out of informal settlement neighborhoods, creating spatial poverty traps. Therefore, it is imperative for Kenyan authorities to leverage urbanization for poverty reduction while addressing urban-specific poverty challenges. First, the government needs to accelerate infrastructure projects and target the urban poor to accommodate an increasing urban population. Second, job creation in urban Kenya should be a priority, given the large number of unemployed and casual workers in the economy. Third, economic opportunities in cities need to be extended to the rural poor. This will require an in-depth analysis of internal migration patterns. Finally, informal settlement neighborhoods need to be economically integrated to ensure that they function as places of opportunities instead of poverty traps. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 101 Urbanization 5.1 URBANIZATION AND POVERTY Figure 5.1: Urbanization rates in Kenya and other countries, 1950–2050 B etween 2005/06 and 2015/16, Kenya’s population increased by about 10 million while the number of poor people in rural areas decreased by more than 60 SSA Tanzania Percentage of urban population 1 million. By contrast, the number of poor people in Kenya urban areas increased by 1.5 million in the same period. 40 Ethiopia The share of the urban poor in the country’s poor Uganda population increased from 13.8 percent in 2005/06 to 23.1 percent in 2015/16. The number of poor urban 20 households has not only been increasing in Nairobi, which accommodates 19.6 percent of the country’s urban poor, but in all of Kenya’s provinces. Nearly 0 one in four of the country’s poor people live in urban 1950 1960 1970 1980 1990 2000 2010 2020 2030 2040 2050 areas. While poverty rates in both urban and rural areas Year Source: United Nations, Department of Economic and Social Affairs, Population continue to decrease, the urban poor is benefitting the Division (2014). most from economic growth. However, urban poverty reduction has been marginal during the last decade. While Nairobi’s population has been rapidly Moreover, poverty headcount ratios in urban areas are growing, the number of people living in medium- similar to those in rural areas when Nairobi is excluded. sized cities has also dramatically increased. Nairobi Therefore, it is not clear if the process of urbanization accommodates more than 3 million people, or one- has been an engine for poverty reduction in Kenya in third of Kenya’s urban population. The next biggest recent years. city, Mombasa, hosts about 10 percent of the country’s urban population. Moreover, the number 5.1.1 Urbanization and poverty trends of people living in medium- or small-sized cities About 28.0 percent of Kenya’s population currently increased dramatically from 2.7 million in 1999 to 8.3 lives in urban areas, and the country’s rate of million in 2009. Of the 25 largest urban areas, 10 are urbanization is similar to that of other East African within the Nairobi metropolitan area, accounting for countries. Urbanization rates (i.e., the share of the about 40 percent of the urban population and more population living in urban areas) increased from 20.1 than one-third of Kenya’s GDP. Given the different percent in 2005/06 to 28.4 percent in 2015/16.123 characteristics between Nairobi, Mombasa, and Kenya’s urban population is projected to rise to 22 the rest of Kenya’s cities, this chapter treats them million by 2030, accounting for 33 percent of the total separately in many analyses. population.124 While Kenya has been urbanizing at a similar pace to that of other East African countries, some Despite a slight decline in the urban poverty countries, such as Tanzania, have had higher urbanization headcount ratio, Kenya’s urban poor population rates (Figure 5.1). Additionally, the level of urbanization in increased during the last decade. While the urban Kenya is much lower than the average of Sub-Saharan poverty rate declined from 32.1 percent to 29.4 percent Africa, suggesting that Kenya is under-urbanized given between 2005/06 and 2015/16, it declined from 39.1 its middle-income country.125 Given its current GDP, percent to 36.1 percent in the same period when about 40 percent of Kenyans should be living in urban Nairobi is excluded (Figure 5.2a).126 As a result, the urban areas, according to a correlation analysis of GDP per poverty rate without Nairobi (36.1 percent) was not capita and urbanization rates across countries. statistically different from the rural poverty rate (38.8 percent) in 2015/16. Kenya’s urban poor population 123 KIHBS. 124 United Nations, 2014. 126 The difference in urban poverty rates (whether Nairobi is included or not) 125 World Bank 2016. was not statistically significant between 2005/6 and 2015/16. 102 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Urbanization Box 5.1: Definition of urban areas This report classifies the census term ‘core urban’ as urban areas and includes ‘peri-urban’ in rural areas. Kenya’s census in 1999 and 2009 divided the country into core urban, peri-urban, and rural areas. According to the census definition,127 an urban area refers to: “a built-up and compact human settlement with a population of at least 2,000 people defined without regard to the local authority boundaries. It is normally a trading, market and service centre that provides goods and services to both the resident and surrounding population and is therefore sometimes referred to as an urban center.” The 2009 census further distinguishes core urban and peri-urban areas. A core urban area is defined as, “the central built-up area of an urban center with intense use of land and highest concentration of service functions and activities”; peri-urban is “the area beyond the central built-up area that forms the transition between urban and rural areas.” This approach has also been adopted by the United Nations128 and the World Bank.129 Figure 5.2: Poverty headcount ratio and number of poor, 2005/6 and 2015/16 a) Poverty headcount ratio b) No of poor/non-poor 60 50 45 51 50 47 40 Poverty headcount ratio (%) 35 40 39 36 30 32 29.0 Millions 29 25 18.9 30 19.9 20 14.0 20 15 10 16.4 9.1 16.6 10 12.6 5 4.9 14.3 2.3 3.8 0 0 2005 2015 2005 2015 2005 2015 Urban Rural National Urban Rural National 2005 2015 Number of poor Number of non-poor Source: Staff calculation based on KIHBS 2005/6 and 2015/16. increased by 1.5 million (about 65.0 percent) between urban households increased in all provinces between 2005/06 and 2015/16, from 2.3 million to 3.8 million 2005/06 and 2015/16 due to population growth (Figure (Figure 5.2b). By contrast, the rural poor population fell 5.3b). While there was a slight increase in the poor from 14 million to 13 million in the same period.130 urban population in Nairobi in the same period, the city’s non-poor population increased dramatically, from The decrease in the urban poverty rate, along 2.2 million in 2005/6 to 3.7 million in 2015/16. with an increase in the urban poor population, is observed in most of the country’s provinces. Except The populous counties surrounding Nairobi for in Nyanza, where the urban poverty rate increased accommodate a larger number of the urban poor from 37.7 percent to 44.1 percent, urban poverty while having relatively low urban poverty rates. headcount ratios fell in all provinces between 2005/06 In 2015/16, county-level urban poverty rates varied and 2015/16 (Figure 5.3a). For example, Nairobi’s urban widely, from 16.7 percent in Nairobi, which hosts poverty rate fell from 21.3 percent to 16.7 percent in one-fifth of the country’s urban poor, to 76.2 percent the same period. Nevertheless, the number of poor in Turkana (Figure 5.4).131 Counties with lower urban 127 KNBS 2012. 128 2014 129 2016 130 Throughout this chapter, poverty is measured based on the absolute poverty line, unless otherwise noted. 131 See Appendix E for more information on county-level poverty rates. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 103 Urbanization Figure 5.3: Poverty rates and number of poor in urban areas by province, 2005/6 and 2015/16 a) Poverty rate 80 74 70 68 60 51 49 50 44 41 40 38 38 36 32 31 32 30 29 30 29 26 21 20 17 10 0 Coast North Eastern Eastern Central Rift Valley Western Nyanza Nairobi All provinces 2005/6 2015/16 b) Number of poor 14 12 10 9.1 8 Millions 6 4.9 4 3.7 2 3.8 1.6 2.2 2.3 1.3 1.1 0.6 0.8 0.0 0.1 0.2 0.5 0.3 1.0 0.2 0.2 0.5 0.5 0.4 0.6 0.1 0.2 0.1 0.2 0.1 0.4 0.4 0.2 0.2 0.3 0.4 0.6 0.7 0 05/06 15/16 05/06 15/16 05/06 15/16 05/06 15/16 05/06 15/16 05/06 15/16 05/06 15/16 05/06 15/16 05/06 15/16 Coast North Eastern Eastern Central Rift Valley Western Nyanza Nairobi All provinces Poor Non-poor Source: Staff calculation based on KIHBS 2005/6 and 2015/16. poverty rates are clustered around Nairobi, and a Figure 5.4: Share of urban poor across counties, 2015/16 sizable number of the urban poor is concentrated in Nairobi these counties because of their high density (Figure Others 20% 20% 5.5). By contrast, sparsely populated counties in the Bungoma northern part of Kenya have higher urban poverty rates 2% Machakos along with a smaller proportion of the country’s urban 2% Kakamega poor population. 2% Mombasa 8% Kisii 2% Homa Bay Most counties with lower urban poverty rates also 2% Garissa Turkana have lower rural poverty rates (Figure 5.6). However, 2% 8% Kisumu despite the clear linear correlation between urban and 3% Kiambu rural poverty rates, some counties deviate from the Uasin Gishu 4% Kajiado 8% 4% Nakuru Kili trend. The counties of Kitui, Bomet, and Samburu have Mandera 4% 5% 6% much lower urban poverty rates relative to rural poverty Source: Staff calculation based on KIHBS 2015/16. rates. By contrast, some counties have higher urban Note: Counties are ordered based on their share of the urban poor in the total number of urban poor in Kenya. poverty rates relative to rural poverty rates, including Meru, Nyeri, Homa Bay, Siaya, Isiolo, and Vihiga.132 132 No clear relationship is observed between urbanization and urban poverty rates and the size of urban population and urban poverty rates (not reported). 104 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Urbanization Figure 5.5: County-level urban poverty rates and number of urban poor, 2015/16 a) urban poverty rates b) Number of urban poor LEGEND LEGEND Number of poor (thousands) Urban poverty rate (%) 2 to 30 15 to 22 30 to 83 22 to 34 83 to 176 34 to 47 176 to 321 47 to 64 321 to 745 64 to 76 c) Gap between rural and urban poverty rates LEGEND Gap (ppt) -40 to -32 -32 to -5 -5 to 2 2 to 10 10 to 27 Source: Staff calculation based on KIHBS 2015/16. Note: Panel (A) shows urban poverty headcount ratios; Panel (B) shows the number of urban poor; and Panel (C) shows the difference in poverty rates between urban and rural areas. Figure 5.6: County-level urban and rural poverty rates, 2015/16 100 90 Samburu Turkana 80 Rural poverty headcount ratio (%) Mandera Busia 70 Garissa Marsabit Wajir 60 Tana River Kili West Pokot Bomet Kwale 50 Kitui Laikipia Kajiado Isiolo Elgeyo Marakwet Vihiga 40 Baringo Migori Kisii Nyamira Kericho Bungoma Siaya 30 Embu Machakos Homa Bay 20 Kirinyaga Meru Kiambu Nyeri 10 Nairobi Mombasa 0 0 10 20 30 40 50 60 70 80 90 100 Urban poverty headcount ratio (%) Source: Staff calculations based on KIHBS 2015/16. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 105 Urbanization The correlation between urban and rural poverty the underperformance of poverty alleviation efforts in headcount ratios at the county-level implies that cities and towns, which is a cause for concern. cities and towns are not realizing the full of their economic potential. On one hand, the narrow gap in Population shifts from rural to urban areas urban and rural poverty headcount ratios may be due contributed moderately to poverty reduction to the poverty-reducing function of cities and towns. between 2005/06 and 2015/16. The results of a The gap in economic opportunities and living standards decomposition analysis suggest that the transition between urban and rural areas have induced many in of people from rural to urban areas accounted for poor rural areas to migrate to urban areas, narrowing the only 12.1 percent of the fall in poverty during the last gap. However, those with relatively good endowments decade (Figure 5.7). Instead, most poverty reduction often migrate from rural to urban areas. Rural poverty was due to poverty alleviation efforts within rural areas, headcount ratios may also be due to spillover effects, as there has been little progress in eliminating poverty as cities and towns often bring economic benefits to within urban areas, including Nairobi.133 Moreover, a surrounding rural areas. On the other hand, the narrow province-level analysis of poverty in Kenya showed gap in poverty headcount ratios between urban (except that urbanization has not substantially contributed to for Nairobi) and rural areas may also be an indication of poverty reduction, except for in the province of Coast. Figure 5.7: Sectoral decomposition of poverty reduction, 2005/6 and 2015/16 All provinces Coast North Eastern Eastern Central Rift Valley Western Nyanza 2.0 1.1 0.5 1.0 0.2 0.2 0.1 0.1 0.0 0.0 0.1 0.0 - 0.3 - 0.5 - 0.2 -2.0 -1.3 -1.2 -2.4 -4.0 -3.9 -3.8 -6.0 - 6.6 -7.0 -7.1- 6.8 -8.0 -10.0 - 9.7 -10.5 -10.2 -10.3 -10.3 -10.1 -12.0 -14.0 -16.0 -18.0 -17.1 -18.4 -18.5 -20.0 -18.7 Di Intra-sectoral e ect Population-shift e ect Interaction e ect Source: Staff calculations based on KIHBS 2015/16. Box 5.2: Decomposition analysis The decomposition analysis estimates the poverty rate based on 1) the intra-sectoral effect, 2) the population shift effect, and 3) the remaining part as a residual, following Ravallion and Huppi (1991). First, the intra-sectoral effect estimates changes to the overall poverty rate if the urban/rural population share remained constant while the level of poverty within urban/rural areas changed. Second, the population shift effect estimates changes to the overall poverty rate if the poverty rate in rural/urban areas remained constant while there was a change in the share of the urban/rural population. However, the population shift neither considers the effect of migration (since rural residents with a high likelihood to be economically successful tend to migrate from rural to urban areas) nor spillover effects (i.e., that urban economies benefit nearby rural villages). Nevertheless, this still offers useful insight into understanding the linkages between urbanization and poverty reduction. 133 Poverty reduction within urban areas accounted for only 6.0 percent (3.0 percent in Nairobi and 3.0 percent in other urban areas) of the intra- sectoral effect, while rural areas accounted for 94.0 percent. 106 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Urbanization 5.1.2 Urban-rural linkages River, Lamu, Marsabit, Mandera, Nyamira, Samburu, Rural to urban migration accounted for 26 percent and Kajiado have received migrants from other areas of the recent internal migration of Kenya’s male (Appendix E).135 population. About 20 percent of working-age men in Kenya moved to their residence within the last four The extent to which urbanization has contributed years.134 Among them, about 26 percent were migrants to poverty reduction depends to a large degree on from rural to urban areas: 11 percent had moved to who migrated to/from urban areas. The direct effect large cities (either Nairobi, Mombasa, or Kisumu) and is unclear if urban areas attracted mostly non-poor another 15 percent had moved to other urban areas. migrants from rural areas. Moreover, the effect is not As a result, a significant portion of men has recently necessarily clear even if poor rural households migrated moved into their current residence in urban areas. to urban areas and managed to escape poverty. The Among the current male urban population (between direct effect of internal migration on poverty reduction 15 and 54 years old), around 30 percent of men moved is only clear if households that migrated to urban areas to their current residence within the last four years, and would have remained poor had they stayed in rural another 16.0 percent of men moved to their current areas. An understanding of the selection mechanism residence between four and eight years ago (Appendix that determines whether or not a household migrates E). Internal migrants are concentrated in or near major based on their current and prospected economic cities (Figure 5.8). status is therefore crucial in gauging the impact of urbanization on poverty reduction.136 Table 5.1: Recent male migration by origin and destination Destination While recent rural migrants living in Nairobi and Nairobi / Mombasa are more well-off than rural residents, Other Mombasa / Rural urban there is no clear difference observed in other urban Kisumu Nairobi / areas. When male individuals are ranked based on Mombasa / 15% 5% 6% the Composition of Wealth index, which measures Origin Kisumu the assets held by households, few recently settled Other urban 4% 14% 9% migrants in Nairobi and Mombasa are ranked in Rural 11% 15% 21% the bottom 40 percent of the population based on Source: Staff calculation with DHS 2014. wealth (Figure 5.9). This wealth gap between recently Note: Numbers show the share of internal migration during the last four years (only men between 15 and 54 years old). migrated rural households living in urban areas and rural residents could potentially be due to a better Meanwhile, a large portion of the population ability to generate assets while living in cities. However, has also moved from urban to rural areas, partly the most likely explanation is migration selection—that offsetting rural-to-urban migration. Urban-to-rural rural residents with already high levels of human and migration accounted for about 15 percent of Kenya’s physical capital have moved to Nairobi and Mombasa. internal migration during the last four years. The Outside of Nairobi and Mombasa, the distribution share of the country’s working-age men who recently of the wealth index does not differ much between migrated from urban to rural areas was relatively high rural-to-urban migrants and rural residents, which is in Muranga, Taita Taveta, Turkana, and Vihiga. These consistent with the small gap in poverty rates between provinces have accommodated many migrants from urban (excluding Nairobi) and rural areas. Nairobi, Mombasa, and Kisumu, while Nyandarua, Tana 135 The temporary nature of migration is also reflected in the 20 percent of urban households that own land outside of the city (probably agricultural land in rural areas). Also, around 20 percent of households in informal settlement areas (and 15 percent of households in non-informal 134 The analysis is based on the DHS 2014. The analysis focuses on the settlement areas) also own a second home outside of the city. internal migration of male populations because of the lack of information 136 Unfortunately, the latest KIHBS does not contain any migration-related about female migrants. Since men tend to be more mobile than women, questions, which makes it difficult to assess the welfare impacts of actual migration rates of the total population would be lower than what internal migration. Therefore, the DHS 2014 and the Cities Baseline Survey is reported here. 2013 are used for this analysis (see Appendix A for data descriptions). KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 107 Urbanization Figure 5.8: Share of recent migrants in urban areas in 47 counties, 2014 a) Within 4 years (urban) b) Within 8 years (urban) c) Within 4 years (rural) d) Within 8 years (rural) LEGEND Share (%) 0 - 10 10 - 20 20 - 30 30 - 40 40 - 50 50 - 60 Source: Staff calculations based on the 2014 DHS. Note: Share of men who moved to their current residence in urban (panels a and b) and rural areas (panels c and d) within a specific time period. Major cities and roads are also shown. Figure 5.9: Wealth index by migration status, 2014 a) Urban residents from rural areas b) Nairobi residents from rural areas 80 80 70 70 60 60 50 50 Percent Percent 40 40 30 30 20 20 10 10 0 0 1 2 3 4 5 1 2 3 4 5 Wealth index Wealth index <4 yrs 4-8 yrs >8 yrs <4 yrs 4-8 yrs >8 yrs Non-migrants Rural Non-migrants Rural 108 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Urbanization c) Mombasa residents from rural areas d) Other urban residents from rural 80 80 70 70 60 60 50 50 Percent Percent 40 40 30 30 20 20 10 10 0 0 1 2 3 4 5 1 2 3 4 5 Wealth index Wealth index <4 yrs 4-8 yrs >8 yrs <4 yrs 4-8 yrs >8 yrs Non-migrants Rural Non-migrants Rural Source: Staff calculation based on the 2014 DHS. Note: Graph shows the distributions of male migrants from rural areas by quintile in the Composition of Wealth index. Rural non-migrant males are also shown for comparison. Domestic remittances are another aspect of 5.2.1 Monetary dimension urbanization that affect poverty reduction. A majority Poor urban households spend a large portion of their of urban residents sent cash to their family members income on food and housing, and consumption (mostly in rural areas) during the last three months.137 patterns have changed little during the last decade. Despite the wide consumption gap between residents In 2015/16, urban households spent an average of that live in informal settlement and non-informal 46.6 percent of their monthly expenditure on food, settlement areas, households that live in informal 15.1 percent on housing, 6.3 percent on utilities, 7.7 settlement areas send remittances at a higher rate percent on transport, 7.4 percent on education, and than households in non-informal settlement areas. This 16.9 percent on other goods and services (Figure probably reflects the high proportion of migrants living 5.10a).139 This expenditure pattern was similar between in informal settlement areas, enduring inadequate living households in Nairobi (Figure 5.10b), Mombasa conditions to support rural family members or relatives. (Figure 5.10c), and other urban areas (Figure 5.10d). More than Ksh4,000 is on average sent from urban to Moreover, poor households allocate a large share of rural areas every three months, which translates into their income to food (53.0 percent) and only a small a 10 percent increase in per capita consumption for fraction to transport (3.2 percent), which limits their receiving households.138 job accessibility and potentially lowers their (and the city’s) economic performance. Poor households 5.2 DIAGNOSTIC OF URBAN POVERTY also spend more on utilities (8.9 percent) than other T he share of non-poor households’ budgets dedicated to food increased in the last decade. In addition, housing costs constrained household finances in medium- and households, probably because they are more likely to rely on services with higher unit costs.140 Finally, poor households in Nairobi and Mombasa allocate a larger small-sized towns, reflecting higher rates of urbanization portion of their expenditure to education than non- and a rise in the cost of living. While access to basic services poor households. improved among urban households, including the urban poor, the gap in access between the poor and non-poor households remains wide. 139 Housing rents were imputed for owners. 140 The previous poverty assessment (World Bank 2009, p.57-58) mentions that toilet facilities were more expensive than food in Nairobi informal 137 The analysis is based on the Cities Baseline Survey 2013. settlements: toilet facilities cost Ksh5 per visit per family member, 138 Meanwhile, about 20 percent of urban households received cash regardless of the nature of the visit. It also mentions that informal from other family members during the last three months. Most cash dwellers pay approximately eight times more for water than their non- transfers came from urban areas, though the average amount of money informal settlement counterparts in Nairobi, as the latter pay a standard transferred was much smaller compared with urban-to-rural transfers. rate of Ksh120 shillings for up to 10,000 liters of water. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 109 Urbanization Figure 5.10: Share of household expenditure in urban Kenya, 2005/06 and 2015/16 a) All urban b) Nairobi 14 13 17 17 11 25 24 14 19 18 5 24 24 5 7 4 3 3 7 7 11 8 8 10 7 7 9 8 11 7 7 4 8 10 7 9 9 9 6 6 12 11 11 12 14 6 6 7 7 6 15 15 15 6 16 16 14 14 17 17 56 53 56 46 47 49 38 39 43 43 34 35 P NP All P NP All P NP All P NP All 2005/6 2015/16 2005/6 2015/16 Food Housing Utilities Transport Education Other c) Mombasa d) Other urban 16 13 16 15 16 14 18 18 25 24 23 26 6 4 6 5 3 8 8 4 2 3 3 5 5 7 9 8 9 6 9 6 6 10 6 7 6 10 7 7 7 6 7 10 7 14 9 9 7 17 14 14 7 15 15 15 17 10 10 17 57 55 49 48 50 50 45 49 49 40 41 43 P NP All P NP All P NP All P NP All 2005/6 2015/16 2005/6 2015/16 Food Housing Utilities Transport Education Other Source: Staff calculation based on KIHBS 2005/6 and 2015/16. Note: Based on per adult equivalent monthly consumption (spatially deflated). P: poor based on absolute poverty line; NP: non-poor. Urban households’ food expenditure has been rising 2015/16, the share increased from 10.1 percent to 14.8 at a concerning rate. Household spending on food percent in the other urban areas in the same period, increased as a share of total spending in all urban reflecting urban growth in these areas. The share of areas, except for poor households who already allocate housing expenditures in total household spending a large portion of their budgets to food. The rising varied within and across counties. For example, the share of food expenditure is probably due to the rise median expenditure share on housing was 7.0-8.0 in food prices. Food prices, measured by the CPI, have percent in Baringo, Bungoma, and Wajir counties, increased more than the prices of non-food items in whereas the median share reached nearly 20.0 percent Kenya since 2005.141 in Kajiado. Urban households allocate a larger share of their budgets to housing in counties with a larger Urban households in populous counties allocate urban population. Poor households also spend a lot a larger share of their budgets to housing than on housing in counties where non-poor households households in less populous counties. While the allocated a large share of their income to housing, share of the budget dedicated to housing remained although they often avoid paying high rents by living at the same level in Nairobi between 2005/06 and in informal settlement areas. 141 Kenya’s CPI is calculated based on price information collected in Nairobi and 13 other urban centers. 110 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Urbanization 5.2.2 Non-monetary dimension years. The number of housing units made from durable Despite the increase in housing costs, the majority of material increased by only 150,000 units during the urban residents still live in housing of substandard same period. Nearly 1 million non-durable housing quality. Most of the newly added housing units in units currently accommodate the country’s urban urban areas between 2005/06 and 2015/16 were poor. To increase the supply of affordable housing (one constructed with non-durable materials (Figure 5.11). of the Big 4 policy priorities announced by the new The number of housing units with either walls, roofs, or administration), it is imperative for the government to floors—or all of these parts—made from non-durable address supply-side bottle necks—such as high land materials increased by 2.1 million during the last ten costs—and constraints to housing demand—such as the underdeveloped mortgage markets.142 Figure 5.11: Housing units with non-durable structures in urban areas, 2005/06 and 2015/16 While the share of households with access to 3.5M improved water remains high in urban areas, it 0.4 has decreased in some provinces during the last decade.143 Kenya’s constitution guarantees access to 1.0 basic services, such as water, sanitation, and a clean environment, as a basic right for all Kenyans. The share of 1.4M the population with access to improved water increased 0.3 0.3 2.1 0.8M in rural areas between 2005/06 and 2015/16. By 0.4M 0.2 contrast, the share of urban households with access to 0.8 0.1 0.2 0.1 0.4 improved water slightly decreased from 94.8 percent to 0.2 2005 2015 2005 2015 90.8 percent in the same period (Figure 5.12).144 Access to Urban Urban poor improved water fell significantly in Coast (98.3 percent to 1 part non-durable 2-parts non-durable 3-parts non-durable 87.1 percent) and Eastern (94.4 percent to 76.6 percent), Source: Staff calculation based on KIHBS 2005/6 and 2015/16. Note: Non-durable housing units contain one or more of walls, roof, and floors an indication that water provision has not kept pace made of non-durable materials. with urbanization in these provinces. While the urban Figure 5.12: Access to improved water in provinces by urban/rural area, 2005/06 and 2015/16 98 100 99 100 94 95 92 92 87 89 90 89 91 91 90 82 82 80 77 74 73 75 74 % of households with access 70 68 65 63 57 58 58 60 50 51 51 52 50 47 47 48 44 40 30 20 10 0 Coast North Eastern Eastern Central Rift Valley Western Nyanza Nairobi All provinces 2005/6 Rural 2015/16 Rural 2005/6 Urban 2015/16 Urban Source: Staff calculation based on KIHBS 2005/6 and 2015/16. Note: The main source of drinking water is classified as improved for piped water within a dwelling; piped water outside a dwelling; public tap or standpipes; tube well or borehole with pump; protected dug well; protected spring; tankers or vendor; and bottled water. There are no rural area in Nairobi. 142 World Bank 2018b. 143 The following types of main sources of drinking water are classified as ‘improved’: piped water within the dwelling; piped water outside the dwelling; public tap or standpipes; tube well or borehole with a pump; protected dug well; protected spring; tankers or vendor; and bottled water. 144 The difference is statistically significant at the 1.0 percent level. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 111 Urbanization population more than doubled between 2005/06 and Mombasa (Figure 5.13c), the majority of the poor rely on 2015/16, its rural population remained roughly the standpipes (58.2 percent and 71 percent, respectively). same, and the share of rural households with access Access to improved water is worse in Mombasa than in to improved water increased by 10 percentage points. Nairobi: only 16.6 percent of its residents have access The Joint Monitoring Program145 by the WHO and the to water taps, and the share of the population with United Nations Children’s Fund has confirmed this unimproved water access increased among both poor downward trend in the share of urban households with and non-poor residents. It is important to extend access access to improved water since the 1990s. to basic services to poor households while maintaining affordability and cost recovery, as they spend a larger The gap in access to improved water resources share of their income on utilities. between poor and non-poor urban residents is wide and has not converged since 2005/06. While about The share of households with access to improved 92 percent of non-poor households have access to sanitation facilities has profoundly increased in improved drinking water, this figure is only 86 percent urban areas.146 Between 2005/06 and 2015/16, the for poor households (Figure 5.13a). Moreover, only 3.4 share of Kenyan households with access to improved percent of poor households have access to private sanitation increased from 66.2 percent to 90.9 percent taps within a dwelling. In Nairobi (Figure 5.13b) and in urban areas (Figure 5.14a). While all provinces Figure 5.13: Access to water in urban Kenya, 2005/06 and 2015/16 a) All urban b) Nairobi 3 5 8 3 4 3 1 1 12 14 9 7 6 7 8 17 12 18 13 21 27 22 21 18 26 20 21 24 24 56 58 42 34 36 40 33 36 34 32 31 31 27 30 36 23 26 33 28 23 32 24 19 22 3 3 4 4 P NP All P NP All P NP All P NP All 2005/6 2015/16 2005/6 2015/16 Piped within dwelling Piped outside dwelling Public tap/standpipe Other improved Not improved c) Mombasa d) Other urban 12 18 7 11 17 12 13 19 17 36 37 10 38 13 26 13 12 26 17 18 21 60 14 15 18 18 34 71 54 26 51 35 60 40 1 31 35 28 15 5 31 5 3 5 21 21 13 5 5 13 18 13 1 3 2 11 3 5 P NP 2 All P NP All P NP All P NP All 2005/6 2015/16 2005/6 2015/16 Piped within dwelling Piped outside dwelling Public tap/standpipe Other improved Not improved Source: Staff calculation based on KIHBS 2005/6 and 2015/16. Note: P: poor based on absolute poverty line; NP: non-poor. 146 Improved sanitation includes the use of flush toilets, VIP latrines, and 145 WHO/UNICEF, 2014. covered pit latrines. 112 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Urbanization Figure 5.14: Access to improved sanitation in provinces by urban/rural area, 2005/06 and 2015/16 a) Improved sanitation (lower standard) 100 90 80 73 70 68 % of households with access 62 60 60 55 56 50 47 43 42 41 43 44 40 33 32 30 25 26 28 21 18 20 17 15 15 11 12 11 10 5 4 7 6 6 6 7 3 4 0 Coast North Eastern Eastern Central Rift Valley Western Nyanza Nairobi All provinces 2005/6 Rural 2015/16 Rural 2005/6 Urban 2015/16 Urban b) Improved sanitation (higher standard) 100 95 94 95 91 90 87 87 87 80 78 79 75 74 70 69 70 70 66 66 % of households with access 65 64 57 60 59 60 54 51 50 48 47 42 42 44 40 36 37 32 30 27 20 18 15 10 0 Coast North Eastern Eastern Central Rift Valley Western Nyanza Nairobi All provinces 2005/6 Rural 2015/16 Rural 2005/6 Urban 2015/16 Urban Source: Staff calculation based on KIHBS 2005/6 and 2015/16. Note: Lower-standard improved sanitation includes flush toilets, VIP latrines, and covered pit latrines (Panel A), and higher-standard improved sanitation includes only flush toilets and VIP latrines (Panel B). There are no rural areas in Nairobi. improved access to sanitation during this period, the Similar to water access, the type of sanitation access provinces of Coast (57.4 percent to 86.6 percent), Rift in urban areas is linked to a household’s level of Valley (58.9 percent to 87.3 percent), and North Eastern consumption. 54.5 percent and 12.1 percent of non- (42.1 percent to 78 percent) made remarkable efforts poor residents used flush toilets and VIP latrines, to catch up with other provinces. However, these respectively (Figure 5.15a) in 2015/16. By contrast, only achievements are less impressive if compared to a 19.5 percent and 12.1 percent of poor households relied higher standard of improved sanitation, which includes on the same sanitation facilities in the same period. flush toilets and VIP latrines but excludes covered pit Nevertheless, poor residents managed to catch up with latrines (Figure 5.14b). The share of urban households their non-poor counterparts over the last decade. For with access to higher-standard improved sanitation example, the share of poor households with uncovered increased from 43.6 percent in 2005/06 to 59.7 percent pits dramatically dropped between 2005/06 and in 2015/16. Yet, less than one-third of households 2015/16, from 43.3 percent to 11.5 percent in Nairobi, in Western and Nyanza have access to this type of and from 77.3 percent to 23.3 percent in Mombasa. Still, sanitation. Not surprisingly, Nairobi has the highest only 29.7 percent of households (11.2 percent among share of households with access to higher-standard poor households) in Kenya’s other urban areas used improved sanitation (73 percent). flush toilets in 2015/16 (Figure 5.15d). KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 113 Urbanization Figure 5.15: Access to improved sanitation in urban Kenya, 2005/06 and 2015/16 a) All urban b) Nairobi 2 4 5 1 2 2 5 4 1 2 2 4 7 3 13 16 22 20 11 25 20 22 30 27 24 31 8 3 3 0 43 18 9 36 12 0 20 46 12 43 7 7 3 69 74 70 26 10 62 13 55 48 47 0 43 40 5 20 24 13 P NP All P NP All P NP All P NP All 2005/6 2015/16 2005/6 2015/16 Flush toilet VIP latrine Covered pit latrine Uncovered latrine Others c) Mombasa d) Other urban 4 0 1 0 1 1 1 4 5 1 2 10 6 10 12 9 23 16 34 12 28 14 31 44 37 11 37 40 25 11 11 77 30 50 14 9 32 20 14 14 19 66 37 12 61 10 41 37 16 37 5 32 28 30 6 23 12 11 11 2 P NP All P NP All P NP All P NP All 2005/6 2015/16 2005/6 2015/16 Flush toilet VIP latrine Covered pit latrine Uncovered latrine Others Source: Staff calculation based on KIHBS 2005/6 and 2015/16. Note: ‘others’ include... P: poor based on absolute poverty line; NP: non-poor. Access to electricity improved dramatically in urban the urban poor are electricity users (Figure 5.17a). The Kenya over the last decade. The proportion of urban situation in Nairobi is relatively better, as 77.2 percent households that use electricity as their main source of of the city’s poor households have electricity access lighting increased from 61.8 percent in 2005/06 to 80 (Figure 5.17b). Compared to Nairobi, poor residents percent in 2015/16 (Figure 5.16). There was a significant are less likely to have access to electricity in Mombasa increase in electricity users in Eastern (44.2 percent to (Figure 5.17c) and other urban areas (Figure 5.17d). 78.2 percent), Central (59.3 percent to 84.8 percent), Rift Valley (47.7 percent to 75.1 percent), North Eastern Health conditions are not necessarily better in urban (24.8 percent to 58.8 percent), Western (17.4 percent to than in rural areas. This is especially true for living 50.4 percent), and Nyanza (31 percent to 59.5 percent). conditions in urban informal settlements. For example, Nairobi also increased its share of households that use households in urban informal settlement areas electricity from 84.9 percent to 90.7 percent in the score lower than households in rural areas on health same period. indicators covering early childhood mortality, child malnutrition, and the prevalence of childhood illness.147 Yet, poorer households still have limited access to Nevertheless, the health gap between urban informal electricity in urban areas. Poorer households are settlement areas and rural areas has been shrinking. more likely to rely on non-electric sources of lighting Still, while Kenya’s under-five mortality rate has dropped (Figure 5.17). While nearly 90.0 percent of non-poor significantly since 2000 (from 151.0 in 2002 to 79.8 in urban households use electricity, only 54.2 percent of 147 Mberu et al. 2016. 114 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Urbanization Figure 5.16: Access to electricity in provinces by urban/rural area, 2005/06 and 2015/16 100 91 90 85 85 78 80 80 75 % of households with access 73 70 62 58 59 59 60 60 50 50 48 44 44 40 31 30 25 17 18 17 19 20 15 15 11 10 10 5 2 3 1 2 2 4 0 0 Coast North Eastern Eastern Central Rift Valley Western Nyanza Nairobi All provinces 2005/6 Rural 2015/16 Rural 2005/6 Urban 2015/16 Urban Source: Staff calculation based on KIHBS 2005/6 and 2015/16. Note: There are no rural areas in Nairobi. Figure 5.17: Access to electricity in urban Kenya, 2005/06 and 2015/16 a) All urban b) Nairobi 13 11 8 9 20 15 27 23 38 46 41 74 87 89 92 91 80 85 73 77 62 54 59 26 P NP All P NP All P NP All P NP All 2005/6 2015/16 2005/6 2015/16 c) Mombasa d) Other urban 18 22 17 20 27 32 38 46 59 53 77 87 83 62 73 78 80 68 54 62 41 47 23 13 P NP All P NP All P NP All P NP All 2005/6 2015/16 2005/6 2015/16 Source: Staff calculation based on KIHBS 2005/6 and 2015/16. Note: ‘others’ include... P: poor based on absolute poverty line; NP: non-poor. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 115 Urbanization 2012), it is still higher in Nairobi’s informal settlements 5.3.1 Employment status than in any other areas of the country, including rural Poor households are less likely to be active in the areas (56.0 in 2014). Moreover, although the prevalence labor market despite an increase in labor force of diarrhea in Nairobi’s informal settlements dropped participation rates between 2005/06 and 2015/16. from 30.8 percent of the population in 2002 to 20.2 In urban Kenya, the labor force participation rate percent in 2012, it is still higher than in any other area, rose from 69.7 percent in 2005/06 to 76.9 percent in including rural areas (15.7 percent in 2014). The TFR in 2015/16 (Figure 5.18). However, this was mostly due to Nairobi’s informal settlements (4.0 in 2000 and 3.5 in increased participation by the non-poor, as labor force 2012) is lower than in rural areas, but it is still higher participation rates among the poor dropped in Nairobi than in any other urban area. Kenya’s high rate of natural and Mombasa. population growth has contributed to the expansion of urban informal settlements. While Kenya’s unemployment rate decreased significantly in the past decade, it remains relatively 5.3 URBAN LABOR MARKETS low in Nairobi and among women and the youth. W hile the urban unemployment rate fell dramatically in the last decade, a large portion of women, the youth, and the urban poor—particularly in Nairobi—remains The urban unemployment rate declined from 19.3 percent in 2005/06 to 10 percent in 2015/16 (Figure 5.19). In particular, the unemployment rate among the unemployed. An increase in urban construction jobs has urban poor dropped by an impressive 13 percentage provided job opportunities with relatively high earnings points in the same period—from 27.8 percent to 14.7 for the poor. However, these jobs are often unreliable. percent. Nairobi has a higher unemployment rate (12.7 For example, nearly 90.0 percent of construction jobs in percent) than other urban areas (8.4 percent). Despite Nairobi are classified as casual work. As a result, about 40.0 the overall reduction in unemployment, more than percent of poor households in Nairobi are casual workers. a fifth of the poor in Nairobi remains unemployed. Moreover, Nairobi’s low job accessibility is likely imposing Unemployment rates vary widely across counties, a severe burden on poor households and informal with the highest rates in Nairobi. There were nearly settlement residents in their search for well-paid and/ 250,000 unemployed workers in Nairobi in 2015/16. or formal jobs, worsening inefficiencies in labor markets. The unemployment rate is also higher among women Linkages need to be improved between workers and jobs and the youth in many counties. in urban areas to improve employment conditions for Kenya’s most vulnerable groups. Figure 5.18: Labor force participation rates in urban Kenya, 2005/06 and 2015/16 85 81.5 80.3 79.7 79.4 Labor force participation rate (%) 80 76.9 77.3 77.6 74.8 75.1 74.5 75 73.2 73.0 72.7 70.7 69.3 69.7 69.8 69.4 70 69.0 67.4 68.1 67.1 66.8 65 62.3 60 P NP All P NP All P NP All P NP All All urban Nairobi Mombasa Other urban 2005/6 2015/16 Source: Staff calculation based on KIHBS 2005/6 and 2015/16 Note: P: poor based on absolute poverty line; NP: non-poor. 116 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Urbanization Figure 5.19: Unemployment rates in urban Kenya, 2005/6 and 2015/16 50 44.5 45 40 Unemployment rate (%) 35 30 27.8 27.1 27.4 23.6 25 21.5 19.3 18.8 20.2 18.8 20 16.2 16.1 14.7 15 12.7 13.0 12.2 10.0 11.4 10.2 8.7 8.0 8.4 8.4 10 6.5 5 0 P NP All P NP All P NP All P NP All All urban Nairobi Mombasa Other urban 2005/6 2015/16 Source: Staff calculation based on KIHBS 2005/6 and 2015/16. Note: P: poor based on absolute poverty line; NP: non-poor. Although a majority of urban workers are in the female workers are more likely to work in the service service sector, there has been an impressive increase sector (82.0 percent versus 64.0 percent) and less likely in the share of construction jobs among the poor to engage in construction activities (1 percent versus in Nairobi. Service sector jobs are dominant in urban 17 percent). Kenya: 72.5 percent of urban residents worked in services in 2015/16 (Figure 5.20). By contrast, construction jobs The composition of employment types has been only accounted for 10.1 percent of all urban jobs in the stable in urban areas during the last decade. In same period. However, the share of construction jobs 2015/16, around 62.0 percent of workers were paid among the urban poor in Nairobi increased from a mere employees, 1.0 percent were working employers, 33.0 8.4 percent in 2005/06 to 22.4 percent in 2015/16 (Figure percent were own-account workers, and the remaining 5.20b). Poor workers also transitioned from agriculture 3.0 percent were classified as other employment types to construction in other urban areas (Figure 5.20d). (Figure 5.21). This overall employment composition has There are only a limited number of manufacturing jobs changed little since 2005/06, except that the poor are in urban Kenya, accounting for 9.3 percent of all jobs no longer less likely to work as paid employees. in 2015/16.148 Compared with their male counterparts, Figure 5.20: Economic sectors of workers in urban Kenya, 2005/6 and 2015/16 a) All urban b) Nairobi 62 69 73 78 75 80 63 75 74 80 80 80 15 7 7 10 22 4 5 8 9 10 11 8 5 6 9 10 9 17 10 13 12 13 12 15 9 12 6 8 6 8 2 3 3 3 3 3 P NP All P NP All P NP All P NP All 2005/6 2015/16 2005/6 2015/16 Agriculture Manufacturing Construction Other services 148 Appendix E shows the distribution of economic sectors by county. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 117 Urbanization c) Mombasa d) Other urban 63 60 74 75 71 78 76 78 79 79 89 86 5 13 5 4 8 9 4 8 14 10 11 6 7 14 7 3 6 27 7 7 10 9 16 20 8 7 7 10 10 13 0 1 1 3 1 2 P NP All P NP All P NP All P NP All 2005/6 2015/16 2005/6 2015/16 Agriculture Manufacturing Construction Other services Source: Staff calculation based on KIHBS 2005/6 and 2015/16. Note: P: poor based on absolute poverty line; NP: non-poor. Figure 5.21: Employment in urban Kenya, 2005/06 and 2015/16 a) All urban b) Nairobi 7 5 3 3 4 4 4 1 2 2 13 8 19 25 33 33 25 26 30 28 27 34 31 0 33 5 5 4 2 1 4 1 1 0 1 2 80 63 63 62 64 67 67 67 69 61 60 52 P NP All P NP All P NP All P NP All 2005/6 2015/16 2005/6 2015/16 Paid employee Working employer Own-account worker Others c) Mombasa d) Other urban 7 4 5 3 3 3 10 12 12 9 10 16 21 31 32 26 36 37 22 36 38 26 29 6 0 36 6 0 4 3 0 1 1 4 1 3 64 63 66 65 59 59 59 61 55 54 57 45 P NP All P NP All P NP All P NP All 2005/6 2015/16 2005/6 2015/16 Paid employee Working employer Own-account worker Others Source: Staff calculation based on KIHBS 2005/6 and 2015/16. Note: P: poor based on absolute poverty line; NP: non-poor. 118 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Urbanization A large portion of the urban poor—particularly in earn on average 18−29 percent more than workers with Nairobi—is classified as casual workers. In 2015/16, primary or no education. Also, the average earnings of 74.8 percent of urban workers had full-time jobs, while workers with higher education are more than double 7.4 percent and 14.3 percent were employed part- that of workers with primary or no education. Finally, time or as casual workers, respectively (Figure 5.22). workers with no written employment contract earn less Poor workers are more likely to be insecure, as about than half that of workers with written contracts. 27 percent of them were casual workers in 2015/16. Additionally, around 40 percent of poor workers worked Women, workers with little or no education, and in construction in the same period. Since nearly 90 workers from poor households earn relatively well percent of construction jobs in Nairobi are considered in the construction sector. The median monthly casual, 40.6 percent of poor workers have casual jobs, labor income for urban workers in the manufacturing much higher than 9.4 percent of non-poor workers. sector (Ksh14,000) is higher than in services (Ksh12,000) and construction (Ksh12,000) (Table 5.2). 5.3.2 Labor income Agriculture income (median of Ksh5,500 per month) is Labor incomes tend to be higher in urban areas for substantially lower than in either of these sectors. Poor older workers, men, workers with more education, workers who work in construction earn more than and laborers with written employment contracts. workers employed in the service sector (Ksh10,000 There is a wide gender pay gap in Kenya: female workers versus Ksh7,500 per month). Non-poor workers earn earn 44−54 percent less than men, even when age, a similar wage across the manufacturing, services, and education, the nature of work, and working hours are construction sectors (around Ksh15,000 per month). statistically controlled for.149 The difference in income In addition, women and workers with little or no between men and women is even bigger in urban areas education earn relatively well with construction jobs than at the national level (see Chapter 3). Moreover, compared to jobs in the service sector. However, only workers that have completed secondary education a tiny share of women works in construction. Figure 5.22: Job types in urban Kenya, 2015/16 11 14 9 13 13 11 15 15 27 3 4 22 1 24 3 7 3 5 4 1 5 4 7 41 5 6 1 5 30 6 29 9 8 2 29 4 79 82 79 81 75 76 59 57 62 53 55 47 P NP All P NP All P NP All P NP All All urban Nairobi Mombasa Other urban Full time Part time Seasonal Casual worker Others Source: Staff calculation based on KIHBS 2015/16. Note: P: poor based on absolute poverty line; NP: non-poor. 149 The description is based on a regression analysis with a natural logarithm of monthly labor income (with housing allowance). Appendix E summarizes the results for urban areas in Kenya. Excluding housing allowances and/or controlling for hours worked did not change the findings (not reported). KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 119 Urbanization Table 5.2: Median nominal wage by economic sector in urban Kenya, 2015/16 All urban Nairobi Mombasa Other urban Agriculture 5,500 4,200 Manufacturing 14,000 15,000 15,000 10,800 All workers Services 12,000 12,000 13,000 11,600 Construction 12,000 13,000 15,000 12,000 Agriculture 3,000 3,000 Manufacturing 9,500 12,000 11,500 8,000 Poor workers Services 7,500 8,000 9,000 7,000 Construction 10,000 10,500 16,000 9,000 Agriculture 6,000 6,000 Manufacturing 15,000 15,000 15,000 14,250 Non-poor workers Services 14,400 15,000 15,000 14,000 Construction 15,000 15,000 15,000 13,000 Source: Staff calculation with KIHBS 2015/16. Although manufacturing jobs only account for 12.0 Improving job accessibility will be key to achieving percent of all jobs in Nairobi, they offer relatively good functioning labor markets in Kenyan cities. A worker’s salaries for workers with little or no education. Workers labor performance is dependent on job accessibility— earn on average higher wages in the manufacturing the available number of job opportunities that can sector than in services, and manufacturing jobs pay be accessed within a certain travel time (Box 5.3). especially well for less educated workers. Limited job accessibility imposes high job-search costs, hindering an efficient matching between workers 5.3.3 Job accessibility and jobs, which lowers the benefits of agglomeration Urban households’ mode of transportation remained economies. In Kenya, job accessibility is especially low largely unchanged between 2005/06 and 2015/16. for workers in Nairobi.151 Using a minibus, the main form There was only a slight decrease in the share of workers of motorized transport in Nairobi, a worker can reach that commuted by foot and a small increase in the 4 percent (within 30 minutes), 10.8 percent (within share of minibus commuters (Figure 5.23a). Workers 45 minutes), and 23.9 percent (within 60 minutes) of from more well-off households tend to commute by existing jobs (Table 5.3). This level of job accessibility minibus, while poorer workers are more likely to walk is lower than that of comparable cities. For example, to work. Only households in the top 20 percent of the in the metropolitan area of Buenos Aires in Argentina, income distribution commute with their own cars. an urban area that has four times the population of In Nairobi, 39.4 percent of workers walk to work; 38.5 Nairobi, accessibility figures using public transportation percent use minibuses; 5.2 percent commute with their are 7.0 percent, 18 percent, and 34 percent for the own cars; and the remaining 18.0 percent use other same time thresholds.152 In addition, in Greater Dakar transport options (Figure 5.23b). About 75 percent of in Senegal, an urban area roughly equivalent to the size households in the bottom 20 percent of the income of Nairobi with a population above 3 million, the share distribution walk to work. The share of workers that of accessible jobs within 1 hour is 52.0 percent—more commute by minibus is much smaller in other urban than twice the level in Nairobi.153 areas than in Nairobi (Figure 5.23c). Women’s mobility tends to be more restricted. For example, women in Nairobi’s informal settlements are less likely to travel 151 Nakamura and Avner (2018) measured job accessibility in Nairobi by combining various datasets. Table 5.3 shows a calculated job accessibility outside their settlements for work, and if they do, they index at the 1km2 grid cells: the share of accessible jobs by (A) foot, (B) minibus, or (C) car within 60 minutes in Nairobi. are less likely to use motorized transport.150 152 Quirós, 2015. 150 Salon and Gulyani 2010. 153 Stokenberga, 2017. 120 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Urbanization Figure 5.23: Commuting modes in urban Kenya, 2005/6 and 2015/16 a) All urban b) Nairobi 5 8 6 9 9 7 11 10 10 13 0 1 0 3 0 1 5 14 14 1 9 6 2 0 2 4 12 10 8 6 5 3 1 0 5 3 15 3 3 6 1 10 3 3 4 3 2 2 2 1 1 2 1 29 23 5 30 25 33 38 5 38 42 4 7 5 2 78 2 1 75 2 79 70 48 56 43 37 35 39 42 48 P NP All P NP All P NP All P NP All 2005/6 2015/16 2005/6 2015/16 Walk Bicycle Matatu Bus Commuter train Employer provided Private vehicle Other c) Mombasa d) Other urban 9 13 14 9 0 18 15 15 0 18 15 15 17 16 0 2 1 3 2 2 1 0 2 19 2 1 2 19 2 1 5 4 18 7 4 3 3 3 5 31 29 31 16 11 32 19 36 36 7 3 6 3 7 4 3 3 69 69 70 57 54 49 47 49 48 40 44 40 P NP All P NP All P NP All P NP All 2005/6 2015/16 2005/6 2015/16 Walk Bicycle Matatu Bus Commuter train Employer provided Private vehicle Other Source: Staff calculation based on KIHBS 2005/6 and 2015/16. Note: P: poor based on absolute poverty line; NP: non-poor. Box 5.3: Job accessibility Economically dense cities can spur a country’s economic growth through agglomeration economies—the economic benefits from a concentration of firms and people in cities. This requires an economy that can efficiently match workers and firms (Duration and Puga 2004, Bertaud 2014). Crowded, disconnected, and costly African cities, however, restrict economies of agglomeration by lowering workers’ job accessibility (Lall, Henderson, and Venables 2017). Living farther away from potential employment opportunities increases the job search costs of workers, reducing their chances of finding well-paid and/or formal jobs. Disadvantaged workers may be disproportionally challenged by limited job accessibility, as demonstrated by the spatial mismatch hypothesis (Kain 1968, Gobillion and Selod 2014, Andersson et al. 2014; Aslund, Osth, and Zenou 2010). For example, an RCT recently conducted in Addis Ababa in Ethiopia found that providing a transport subsidy to disadvantaged job-seekers increased their chances of finding better jobs (Abebe et al. 2017). Job accessibility is commonly measured by the number of jobs a candidate can access within a certain travel time (Avner and Lall 2016, Quirós and Mehndiratta 2015). An accurate measurement of job accessibility in a city requires data on (i) the spatial distribution of jobs, (ii) the spatial distribution of households, and (iii) transport networks. Specifically, job accessibility in Nairobi is measured with data from the Nairobi Personal Travel Survey 2013 (JICA 2013), the Cities Baseline Survey (see Appendix B), and minibus network data compiled by the Digital Matatus Project (Willams et al. 2015; Nakamura and Avner 2018) (Figure 5.24). KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 121 Urbanization Figure 5.24: Share of accessible jobs within 60 minutes in Nairobi a) By foot b) By minibus Share of jobs accessible Share of jobs accessible by foot within an hour (%) by matatu within an hour (%) No data 16 - 20 No data 40 - 50 0-4 20 - 24 0 - 10 50 - 60 4-8 24 - 28 10 - 20 60 - 70 8 - 12 28 - 32 20 - 30 70 - 80 12 - 16 32 - 36 30 - 40 80 - 90 36 - 40 90 - 100 c) By car Share of jobs accessible by cars within an hour (%) No data 40 - 50 0 - 10 50 - 60 10 - 20 60 - 70 20 - 30 70 - 80 30 - 40 80 - 90 90 - 100 Source: Nakamura and Avner 2018. Table 5.3: Average share of accessible jobs in Nairobi access about 20 percent fewer number of jobs than Walking Minibus Cars more well-off households—even if they use the same (1) (2) (3) method of transportation. This gap in job accessibility Within 30 minutes 1.8% 3.9% 43.7% comes mainly from the fact that poorer households Within 45 minutes 4.0% 10.8% 71.8% tend to live in informal settlement neighborhoods that Within 60 minutes 7.3% 23.9% 88.7% lack efficient transport networks. Source: Nakamura and Avner 2018. Note: Numbers are the average share of jobs that Nairobi residents can reach by foot (column 1), minibus (column 2), and car (column 3) within 30, 45, and The labor behavior of workers in Nairobi is well 60 minutes. correlated with their level of job accessibility. Poor residents in informal settlement areas have Households with low consumption levels and/or live a lower level of job accessibility in Nairobi than in informal settlement neighborhoods tend to walk more well-off residents (Figure 5.25). For example, to work when their job accessibility level is high. For households in the bottom 20 percent of the example, a 1.0 percentage point increase in the share consumption distribution can reach 7.4 percent and of accessible jobs by foot in 60 minutes is associated 25.6 percent of the city’s jobs within 60 minutes by foot with a 3.6 percent higher chance of commuting by foot and by minibus, respectively. By contrast, these figures by workers that live in informal settlements. People also are 9.5 percent and 30.8 percent for households in the tend to spend less time commuting in neighborhoods top 20 percent. This means that poorer households can with good job accessibility. In addition, women are 122 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Urbanization Figure 5.25: Job accessibility and per capita household expenditure in Nairobi a) By foot b) By minibus 12 40 35 10 Share of accessible jobs (%) Share of accessible jobs (%) 30 8 25 6 20 15 4 10 2 5 0 0 1 2 3 4 5 1 2 3 4 5 Consumption quintile Consumption quintile All Informal settlement Non-Informal settlement All Informal settlement Non-Informal settlement Source: Nakamura and Avner 2018. Note: Share of jobs in Nairobi that are accessible within 60 minutes by (A) foot and (B) minibus for households with different consumption quintiles and informal settlement neighborhood status. more likely to participate in labor markets when they difference in the consumption levels between enjoy better job accessibility. A 1.0 percentage point households that live in Nairobi’s informal settlement increase in the share of accessible jobs by foot and by and non-informal settlement areas. Mean per capita minibus in 30 minutes is associated with a 0.84 and monthly consumption of residents that live in informal 0.59 percent point higher chance of women joining the settlement areas (Ksh 10,377) is nearly 40 percent lower labor force, respectively. than that of residents in non-informal settlement areas (Ksh16,688) (Figure 5.26). Moreover, nearly 30 percent 5.4 URBAN INFORMAL SETTLEMENTS of residents in informal settlement neighborhoods A ccording to the 2009 census, around 60 percent of are poor, compared to 9.1 percent of residents in non- urban households live in informal settlements, and informal settlement areas (Table 5.4). A quarter of the 62 percent of Kenya’s population that lives in informal labor force in informal settlements are also unemployed. settlements lives in Nairobi. The latest KIHBS estimates that Finally, mean monthly housing rents paid by residents 30 percent of households that live in informal settlements in informal settlements (Ksh2,819) are only one-third of are poor. By contrast, only 9.1 percent of Nairobi’s residents the rents paid by residents in non-informal settlement that live outside of informal settlement neighborhoods areas (Ksh8,524), reflecting the low standard of living in are classified as poor. Moreover, there is a stark difference informal settlement neighborhoods.154 in living standards between informal settlement and Table 5.4: Poverty rates in informal settlement and non-informal settlement urban areas. Nevertheless, living non-informal settlement areas, Nairobi 2015/16 conditions are even worse in rural areas and informal Percent 95% CI settlements in other African countries. Data also show that Poverty headcount ratio in informal 29.2 [23.2, 35.3] settlements there are limited options for most residents to move out of Poverty headcount ratio in informal settlement neighborhoods. 9.1 [6.0, 12.2] non-informal settlements Poverty headcount ratio in the city 16.7 [13.6, 19.8] Poor households are concentrated in Nairobi’s Source: Staff calculation with KIHBS 2015/16. Note: 95.0 percent confidence intervals in square brackets. informal settlements, where the cost of living is relatively low but housing conditions are 154 An inter-generational implication comes from a recent study by Abuya, Ciera, and Kimani-Murage (2012), finding that mother’s education is a substandard. More than half of Nairobi’s residents strong predictor of child nutrition status in Nairobi informal settlements. live in informal settlement neighborhoods, which Becquer et al. (2010) find that child mortality rate is higher for recent migrants in Nairobi informal settlements. Mberu et al. (2014) analyze account for 62 percent of Kenya’s total population the panel data collected in two Nairobi informal settlements between 2006 and 2009. By measuring poverty based on a composite of various that lives in informal settlements. There is a large indicators, they find the transient nature of poverty. In particular, Muslim and Kikuyu people are found to be more likely to fall into poverty. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 123 Urbanization Figure 5.26: Household consumption and rents in Nairobi’s informal settlement and non-informal settlement areas, 2015/16 .00015 a) Consumption b) Rent .0004 .0003 .0001 Kernel density Kernel density .0002 .00005 .0001 0 0 0 10000 20000 30000 40000 50000 0 10000 20000 30000 40000 50000 Per adult-equivalent monthly consumption Monthly rent Informal settlement Non-Informal settlement Informal settlement Non-Informal settlement Source: Staff calculation based on KIHBS 2015/16. 5.4.1 Living standards households have their garbage collected on a regular There is a stark difference in living conditions basis (in contrast to 85.5 percent in non-informal between informal settlement and non-informal settlement areas). settlement areas. Most dwellings in Nairobi’s informal settlements are made from non-durable materials, Residents in informal settlement areas are also more and 84 percent of houses have only one room (Table likely to face environmental challenges. In Nairobi, E.2). Houses in informal settlement areas also tend more than half of households in informal settlements to be structured with walls of either corrugated iron (about 59 percent) report flooding and garbage dumps sheets (54.3 percent) or stone, cement, or bricks (35.7 as problems.155 By contrast, only 28.0 percent and 23.0 percent), roofs of corrugated iron sheets (87.7 percent), percent of residents in non-informal settlement areas and cement floors (79 percent). Housing structures in report flooding and garbage dumps as problems, non-informal settlement neighborhoods are of better respectively. quality, as 85.7 percent of dwellings have walls made of stone, cement, or bricks. Roofs of corrugated iron Residents in Nairobi’s and Mombasa’s informal sheets are less common in non-informal settlement settlement neighborhoods live in worse housing areas (47.5 percent). conditions, have access to less services, and have less secure tenure than residents in other African Residents in informal settlement areas have less countries’ informal settlements. Compared with access to basic services than residents in non-informal informal housing in Ghana’s capital of Accra and settlement areas in Nairobi. In Nairobi’s informal Ethiopia’s capital of Addis Ababa, houses in the informal settlements, only 29.1 percent of households have settlements of Nairobi and Mombasa are more likely private taps (in contrast to 88.2 percent of households to have cement or tile floors but less likely to be in non-informal settlement areas), and 58.0 percent of made from permanent materials (Figure 5.27a). While households rely on either public taps or standpipes almost all households have access to electricity in the (Table E.3). Whereas 87.5 percent of households in non- informal settlements of Accra and Addis Ababa, many informal settlement areas use flush toilets, this figure of their counterparts in Nairobi and Mombasa lack such is only 43.9 percent in informal settlement areas. access (Figure 5.27b). In addition, a large proportion of In informal settlement neighborhoods, about 17.0 informal settlement residents feel tenure insecurity in percent of households have no access to electricity Nairobi’s informal settlements (Figure 5.28), reflecting (compared to 3.7 percent of households in non- the prevalence of forced evictions in that city. informal settlement areas), and only 47.4 percent of 155 The Cities Baseline Survey 2013. 124 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Urbanization Figure 5.27: Housing quality in African informal settlements a) Dwelling structure 99 99 99 99 98 91 92 92 88 88 88 84 77 74 65 59 58 55 55 54 47 49 43 44 34 32 28 21 11 7 6 8 5 7 9 7 5 2 3 5 More than three rooms Durable walls Durable roof Durable oor Accra All Accra Owned Accra Renting Accra Rent-free Addis All Addis Owned Addis Kebele Addis Other rentals Nairobi Rental Mombasa Rental b) Access to basic services 97 97 99 97 97 96 99 96 99 100100 97 70 64 61 60 55 55 50 47 44 49 45 44 49 43 33 32 26 17 18 15 20 18 19 8 9 5 4 4 Piped water Flush toilet Private toilet Electricity Accra All Accra Owned Accra Renting Accra Rent-free Addis All Addis Owned Addis Kebele Addis Other rentals Nairobi Rental Mombasa Rental Source: Nakamura and Yoshida 2018. Nairobi and Mombasa based on the Cities Baseline Survey 2013. Figure 5.28: Perceived tenure security in African informal 5.4.2 Residential mobility to/from informal settlements 52 settlements While residential mobility is quite high in Nairobi and Mombasa, a large share of households has 34 been living in informal settlement neighborhoods 26 for many years. About 47 percent of residents in 24 informal settlement areas in Nairobi and 34 percent 16 15 in Mombasa were born or are currently living in 8 informal settlements. The average years of residence 5 3 in the current informal settlement neighborhoods 0 Feel insecure (%) of the remaining households are four years in both Accra all Addis all Accra owned Addis owned Accra renting Addis kebele Accra rent-free Addis other rentals Nairobi and Mombasa. The overall distribution of Nairobi rental Mombasa rental the duration of residence is similar between informal Source: Nakamura and Yoshida 2018. Data for Nairobi and Mombasa are based on settlement and non-informal settlement areas in each the Cities Baseline Survey 2013. Note: Bars for Accra and Addis indicate the share of households that answered yes to city. A recent analysis of the census in 1999 and 2009 both of the following questions: 1) Do you think somebody will ask you to move out in the next 12 months?; and 2) Do you think you will be able to resist in the situation? also suggests that moving from rural areas to urban Bars for Nairobi and Mombasa indicate the share of households that answered yes to the question “Do you feel you have secure tenure for your unit, structure, or informal settlements is not necessarily temporary.156 dwelling? By “secure” I mean that no one could just come and force you to leave without an official legal process in which you would participate.” 156 Bird, Montebruno, and Regan 2017. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 125 Urbanization Box 5.5: Profile of residents moving to/from informal settlement neighborhoods An analysis of the previous residence of households in non-informal settlement neighborhoods shows that there has been limited movement between informal settlement and non-informal settlement areas. Figure 5.29 describes the share of current residents in informal settlement and non-informal settlement areas who were born in their current neighborhoods or moved from informal settlement areas, non-informal settlement Figure 5.29: Previous residence of urban households areas, other cities, rural areas, or abroad. More than 40 Mobility 100 percent of Nairobi’s residents were born in their current neighborhoods. Moreover, a majority of the city’s 80 residents in non-informal settlement areas who were not born in their current neighborhoods moved from 60 Percent other non-informal settlement areas in Nairobi. Only 157 40 a small portion of current residents in non-informal settlement areas had moved directly from informal 20 settlements or rural neighborhoods. Many residents in 0 informal settlement areas had previously lived in other Urban Informal Urban Informal Urban Informal Urban Informal formal settlement formal settlement formal settlement formal settlement informal settlement neighborhoods in Nairobi, and a Nairobi Mombasa Kisumu Other cities portion of them had lived in non-informal settlement Not moved From Informal settlement From rural From urban formal Other country From other city areas. Similar to Nairobi, residential movement from informal settlement to non-informal settlement areas Source: Cities baseline survey 2013. Note: Bars indicate the share of households from each type of previous residence. has been very limited in Mombasa. While the likelihood of residents moving out of to move from informal settlement to non-informal informal settlement neighborhoods is low, educated settlement areas, and moving directly from rural to households have a better chance of moving from non-informal settlement areas is also rare (Figure informal settlement to non-informal settlement 5.29). Yet, years of schooling is positively correlated areas. The location of a household’s previous residence with a household’s chance of moving to non-informal is a strong predictor of its future location (Nakamura settlement neighborhoods, and it is especially strong and Karasawa 2018). Households are highly unlikely for female-headed households. (Figure 5.30).158 Figure 5.30: Probability of households moving to non-informal settlement areas in Nairobi and Mombasa a) Nairobi b) Mombasa 1 1 0.9 0.9 0.8 0.8 Predicted probability Predicted probability 0.7 0.7 0.6 0.6 0.5 0.4 0.5 0.3 0.4 0.2 0.3 0.1 3 6 9 12 15 18 21 3 6 9 12 15 18 Years of education Years of education Previously lived in non-Informal settlement Previously lived in non-Informal settlement Previously lived in Informal settlement Previously lived in Informal settlement Source: Nakamura and Karasawa 2018 based on the Cities Baseline Survey 2013. Note: Y-axis indicates predicted probability of a household moving from informal settlement to non-informal settlement areas. 90 percent confidence intervals are also shown. 156 Beguy, Bocquier, and Zulu (2010) analyzed data from Kibera between 2003 and 2007, and they found that a majority of in-migrants are young, 158 The education level of the household head is clearly related to the 60 percent come from rural areas, and 40 percent come from other parts chance of the household moving to non-informal settlement areas. of Nairobi. Among out-migrants, 44 percent moved to rural areas, 32 percent moved to other informal settlements in Nairobi, and 19 percent moved to non-informal settlement areas in city. While households with access to electricity and home ownership were less likely to leave informal settlements, water access did not seem to influence decision-making. 126 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead CHAPTER 6 EDUCATION AND POVERTY SUMMARY Education is central to achieving the goals of eliminating extreme poverty and boosting shared prosperity. Quality education is a key ingredient to sustainable social and economic development. High levels of education are often associated with improved economic opportunities, including improved access to jobs and higher lifetime income. At the country level, economic benefits include increased rates of economic growth through gains in productivity and a greater capacity to adopt new technologies. In addition, education is positively associated with healthier life choices and increased voice and agency, the ability to make decisions and act on them. Education is not only instrumental in promoting development. It is also by itself an end of development. This chapter assesses recent developments in Kenya’s education sector and their relationship to poverty and equity. It takes stock of the recent trends in access to education services as well as their quality and explores their links to poverty and equity. It further examines the inputs into the education sector and the incentives in place for teachers to produce quality education for all. Enrollment rates have increased since 2005/06, but geographic disparities remain, poor children are substantially less likely to attend post-primary education, and learning assessments suggest that Kenyan children often lag behind the curriculum. The GoK has in recent years invested substantial resources to increase enrollment rates, particularly at the primary level. As a result, enrollment rates at almost all levels show robust gains and primary education is nearly universal. However, enrollment in secondary and tertiary education remains substantially higher among the better-off and geographic disparities are pronounced. While learning outcomes for Kenyan children compare favorably to peer countries, Kenya children quickly fall behind the standards set by the national curriculum: only about half of the children in fourth grade master the basic tasks that second-graders should be able to accomplish (e.g., read and understand a paragraph). While well-paid and knowledgeable by regional standards, Kenya’s teachers lack pedagogical skills and are absent from class too often, suggesting that teacher incentives are not always aligned with student learning. The chapter identifies three, intertwined policy priorities: First, enrollment in post-primary education among poor children should be increased. Second, the trade-off between fiscal costs and the provision of quality inputs, most importantly teachers, needs to be addressed in a sustainable manner. Third, teacher incentives and school governance need to be strengthened. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 127 Education and Poverty 6.1 KENYA’S EDUCATION SECTOR The Kenya Institute of Curriculum Development (KICD) T he Kenyan education system currently follows was put in charge of the curriculum and setting student an 8-4-4 structure (excluding pre-primary), with standards. The new constitution also empowered the Teacher Service Commission (TSC), a central agency the use of nation-wide, standardized tests that and now a constitutional commission (World Bank determine student progression. Students are eligible 2014c).161 The TSC has far-reaching authority to govern to start first grade when they are at least six years of age teacher training, recruitment, placement, and promotion, at the start of the school year in January. A full course as well as disciplinary control. Parents are in practice of primary education in Kenya comprises eight grades free to choose which school their children attend. (also called “standards” in Kenya), followed by four years of secondary (“forms”), and four years of tertiary While public expenditure in education accounted for education.159 To earn a primary school certificate, a large share of overall government expenditure in students must take the national primary school exit the past, its importance has declined recently. Real exam, the Kenya Certificate of Primary Education (KCPE), public education expenditure has more than doubled upon completion of grade eight. Almost all students between 2000 and 2015, with most of the increase who complete grade eight take the KCPE. Students are realized over the early 2000s and a smaller portion admitted to a government secondary school based between 2010 and 2015 (Figure 6.1a). As a proportion of on their scores on the KCPE, district-specific quotas, GDP, public education expenditure increased over the and school preferences that students express prior first half of the 2000s, from 5.2 percent to a peak of 7.2 to taking the exam, with more prestigious national percent by 2005 (Figure 6.1b), but was only 5.4 percent schools admitting only the top-scoring students from on average between 2010 and 2015, close to the each district (Lucas and Mbiti 2014). At the end of the regional average. Similarly, while education accounted fourth year of secondary, students sit for the Kenya for around one fourth of Kenya’s overall government Certificate of Secondary Education (KCSE) examination, expenditure between 2000 and 2009 (compared to an the entrance requirement for Kenyan universities. average of 17 percent in the region), it has declined to around 16 percent in 2015, again, a level more typical Kenya’s education system blends substantial for Sub-Saharan African countries (Figure 6.1c). centralization with parental school choice. The central government, through dedicated agencies, sets In the 2000s, Kenya successively abolished school the curriculum and national standards, administers fees for public primary and secondary education, the KCPE and the KCSE, and oversees teacher training, resulting in a sharp increase in enrollment and recruitment, retention, and promotion. Central control a significant shift in demand towards private over the education system has further increased in provision. In 2003 and 2008, respectively, the recent years. Kenya’s new constitution, adopted in 2011, GoK introduced FPE and Free Tuition Secondary only devolved limited responsibilities to the newly- Education (FTSE). The former was associated with founded counties, notably pre-primary education a substantial increase in enrollment, increasing and childcare facilities as well as certain parts of the student-to-teacher ratios, and a significant shift in vocational education system.160 This contrasts with the demand towards private provision among better- experience of other sectors, such as the health sector, off households (Lucas and Mbiti 2012b), presumably where counties assumed considerable responsibilities. out of concern over a deterioration in the quality of education in public schools (Bold et al. 2014) (Box 159 The current curriculum is due to be replaced and the structure of the school system will likely be changed from an 8-4-4 into a 3-6-3-3-3, 6.1). Despite the introduction of FTSE, there are still with two years of pre-primary, six years of primary, and six years of secondary education. substantial fees associated with public secondary 160 The Constitution of Kenya provides under its 2nd schedule that upon education (Matata, 20015; see also p. 136). devolution, Technical and Vocational Education and Training (TVET) institutions shall be under the responsibility of the national government whereas the village polytechnics, craft centres and farmers training centres, and, by extension, similar institutions that train operators in See 161 http://www.klrc.go.ke/index.php/constitution-of-kenya/176- vocational trades and skills shall be under the responsibility of the chapter-thir teen-the -public-ser vice/par t-3-teachers-ser vice - counties. See http://www.education.go.ke/index.php?catid=0&id=19, commission/406-237-teachers-service-commission for some accessed 4/24/2018. background, accessed 20/3/17. 128 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Education and Poverty Figure 6.1: Public expenditure in education, 2000–2015 a) Real expenditure (2000=1) b) % of GDP c) % of total government expenditure 4 8 30 3.5 7 25 6 3 20 5 2.5 4 15 2 3 10 1.5 2 1 5 1 0.5 0 0 2000 2002 2004 2006 2008 2010 2012 2014 2000 2002 2004 2006 2008 2010 2012 2014 2000 2002 2004 2006 2008 2010 2012 2014 Kenya Ethiopia South Africa Kenya Sub-Saharan Africa Kenya Sub-Saharan Africa Source: Own calculations based on WDI data. Box 6.1: Free primary education and the quality of education FPE is credited with expanding access to education, and there is no evidence that it adversely affected test scores. One study finds that the introduction of free education in 2003 significantly increased access for students from disadvantaged backgrounds (Lucas and Mbiti 2012b). Yet there is no evidence that it reduced KCPE test scores. A likely explanation is an increase in the share of students attending private primary education institutions. In particular, the authors find evidence for sorting across school types by socioeconomic status, with students from better-off backgrounds showing a higher propensity to attend private schools in the wake of the reform. A similar finding is reported by Bold et al. (2014). 6.2 ENROLLMENT in secondary school has increased substantially since the early 2000s, but remains at levels much lower than 6.2.1 Overall trends in enrollment primary enrollment (Figure 6.1b). E nrollment in primary is nearly universal and enrollment in pre-primary and secondary education has increased steadily since the early Kenya has made significant progress in closing gender gaps in enrollment. Between 2005/06 and 2000s. Various data sources suggest a gradual increase 2015/16, gender parity in gross enrollment, defined in enrollment in pre-primary since 2000 (Figure 6.2a). As as the ratio of female to male GERs, increased both in mentioned before, the removal of tuition fees in public primary and secondary education, from 0.95 to 0.97 for primary schools in 2003 resulted in a significant increase the former and from 0.89 to 0.95 for the latter. However, in the Gross Enrollment Ratio (GER) at this level, from an regional variation in gender gaps is pronounced: while already moderately high level (Figure 6.1b) (Lucas and gross enrollment is higher for girls than for boys in parts Mbiti 2012b).162 More recently, the ratio has declined of Central and Western Kenya, the reverse is true for the but this is mainly due to a decrease in the number northeast and coastal areas. Chapter 3 provides details. of students that are not of primary school age. This is reflected by the fact that Net Enrollment Rates (NERs) Comparison of 2015/16 KIHBS data and the WDI from increased moderately (Figure 6.2a).163 Gross enrollment the World Bank suggest that enrollment in tertiary education has been increasing rapidly after 2009. Comparison of the two series indicate an increase by 162 The GER is defined as the ratio of students enrolled in a specific level of education -regardless of age- expressed as a percentage of the more than ten percentage points, from around four population in the age group corresponding to this level of education. percent in 2009 (WDI) to around 15 percent in 2015/16 163 The NER is defined as the percentage of children in the age group that officially corresponds to primary schooling who attend primary school. (KIHBS, only counting undergraduate students in KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 129 Education and Poverty Figure 6.2: GERs in pre-primary, primary, secondary, and tertiary, 2000–2016 a) Pre-primary b) Primary 100 140 KIHBS, 2015/16 120 80 100 60 80 Percent Percent 40 60 KIHBS, 2015/16 40 20 20 0 0 2000 2005 2010 2015 2000 2005 2010 2015 Kenya, WDI Kenya, KES KIHBS 2015/16 SSA Kenya, WDI Kenya, KES KIHBS 2015/16 SSA c) Secondary d) Tertiary 80 16 70 14 KIHBS, 2015/16 60 12 50 10 Percent Percent 40 8 30 6 KIHBS, 20 2015/16 4 10 2 0 0 2000 2005 2010 2015 2000 2005 2010 2015 Kenya, WDI Kenya, KES KIHBS 2015/16 SSA Kenya, WDI Kenya, KES KIHBS 2015/16 SSA Source: Own calculations based on WDI, KES 2017 (2012-2016 data) and KES 2013 (2008-2012 data), and KIHBS 2015/16. Note: Data on gross enrollment tabulated across different volumes of the KES were not always consistent. See also notes to Figure 6.3. universities) (Figure 6.2d). Alternative series also point the administrative data reported in the KES. While nearly to a rapid increase in recent years. Not counting middle- all children from the richest quintile of the population level colleges, the 2017 KES suggests an increase in the eventually enroll in secondary, gross enrollment is number of university students by more than 50 percent only 45 percent among children from the poorest between 2013/14 and 2016/17. quintile. Enrollment in tertiary education is negligible among young adults from households in the lower two 6.2.2 Enrollment by poverty status and locality quintiles but close to 45 percent in the top quintile. Differences in enrollment are pronounced at the secondary and tertiary level. Overall GERs in pre- NERs are substantially lower than GERs, and a primary, primary, secondary, and tertiary education as larger socio-economic gradient - suggests a greater estimated from the 2015/16 KIHBS are 95, 107, 75, and propensity to enroll late among the poor. Whereas 15 percent respectively (Figure 6.3). These estimates are the GER counts all students enrolled for a given level substantially higher for pre-primary164 and somewhat of education, the NER counts only those students who higher for primary and secondary, when compared to are in the usual age group for that level. Estimates of 164 Authority over pre-primary education has been delegated to the overall NERs are 66 and 85 percent for pre-primary counties with the implementation of the new constitution. Hence, one reason for larger discrepancies may be an undercounting of students and primary education, respectively, and 42 percent in pre-primary in administrative data, as data producers differ and are more diverse. 130 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Education and Poverty Figure 6.3: NERs and GERs by level, poverty, quintile, and locality, 2015/16 a) Pre-primary b) Primary 120 120 90 90 Percent Percent 60 60 30 30 0 0 Poor Urban Total Rural Non-poor Top 20% Bottom 20% Bottom 40% Poor Urban Total Rural Non-poor Top 20% Bottom 20% Bottom 40% By poverty By quintile By locality By poverty By quintile By locality Gross enrollment ratio Net enrollment rate Gross enrollment ratio Net enrollment rate c) Secondary d) Tertiary 120 120 90 90 Percent Percent 60 60 30 30 0 0 Poor Urban Total Rural Non-poor Top 20% Bottom 20% Bottom 40% Poor Urban Total Rural Non-poor Top 20% Bottom 20% Bottom 40% By poverty By quintile By locality By poverty By quintile By locality Gross enrollment ratio Net enrollment rate Gross enrollment ratio Net enrollment rate Source: Own calculations based on KIHBS 2015/16. Note: 95-percent confidence intervals are indicated. The relevant age brackets are 3-5 years of age at the beginning of the school year for pre-primary, 6-13 years for primary, 14-17 years for secondary, and 18-21 years for tertiary. 95-percent confidence intervals are indicated. in secondary education.165 As with the GERs, there are The decrease in gross enrollment is most likely a result significant differences by poverty status, quintile, and of a reversion to the long-term trend, as many over-age locality that increase throughout the different levels of pupils enrolled or re-enrolled in primary school in the the education system. Larger discrepancies between wake of the 2003 reform. GERs and NERs in pre-primary and primary education among the bottom quintiles suggest that a larger share Over the same period, enrollment in secondary of children enrolled in Kenya at these levels are over- education increased significantly for most children. age. Late enrollment will be further discussed below. Both GERs and NERs show significant improvements in access to secondary education between 2005/06 Net enrollment in primary has increased moderately and 2015/16. Gross enrollment increased by more than between 2005/06 and 2015/16. KIHBS data suggest 30 percentage points while net enrollment increased that the GER in primary has declined by ten percentage by more than 20 percentage points. The increase was points between 2005/06 and 2015/16, from 117 comparable for both children from poor and non- percent to 107 percent. NERs increased over the same poor families but more pronounced in urban areas time yet only modestly, by three percentage points. and among children from the top 20 percent of the expenditure distribution (Figure 6.4b). 165 NERs are not well-defined for tertiary enrollment as it is not clear at what age students should be enrolled in universities. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 131 Education and Poverty Figure 6.4: Changes in primary and secondary enrollment, between 2005/06 and 2015/16, by poverty, quintile, and locality a) Change in primary enrollment (percentage points) b) Change in secondary enrollment (percentage points) 10 50 40 0 Percent 30 Percent 20 -10 10 -20 0 Poor Urban Total Rural Non-poor Top 20% Bottom 20% Bottom 40% Poor Urban Total Rural Non-poor Top 20% Bottom 20% Bottom 40% Poverty Quintile Locality Poverty Quintile Locality Gross enrollment ratio Net enrollment rate Gross enrollment ratio Net enrollment rate Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16. Note: 95-percent confidence intervals are indicated. GERs are high throughout the first seven standards of The drop in enrollment after grade seven is explained primary but drop significantly by the time students to a large extent by low transition rates among reach the final grade of primary (standard 8) and children from poor families. The drop in GERs between then on into secondary. GERs in Kenya are more than seventh and eighth grade of primary and the final grade 100 percent for the first seven standards of primary of primary and the first of secondary is driven by lower (Figure 6.5). However, they drop to 70 percent in form transition rates among the poor: overall, the transition one, the first grade of secondary, suggesting low rates rate drops from 90.5 percent at the end of grade six of transition from primary into secondary (see next to 83.6 percent at the end of grade seven and 73.6 subsection). It is worth noting that the drop in gross percent at the end of grade eight as children transition enrollment in going from primary to secondary was into secondary education. But among children from significantly more pronounced in 2005/06 compared families in the bottom 40 percent, transition rates at to 2015/16, i.e. transition rates have been increasing in these last two transitions are only 81.1 and 65.1 percent, recent years. respectively. In contrast, the primary-to-secondary transition rate among children from families in the top Figure 6.5: Gross enrollment rates by grade and year 60 percent is still 78.6 percent. 160 Low transition rates into secondary among the poor 120 mostly result from financial constraints. As will be shown below, secondary school attendance, even Percent 80 attendance of public secondary schools, often cost a significant fraction of the poverty-line consumption 40 level while primary school attendance is nearly free. The high cost of school attendance is also cited by many 0 dropouts as the main reason for not attending school. STD 1 STD 2 STD 3 STD 4 STD 5 STD 6 STD 7 STD 8 FORM 1 FORM 2 FORM 3 FORM 4 The important role of financial constraints in keeping children from poor families from attending secondary 2005/06 2015/16 education are also confirmed in additional analyses Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16. Note: 95-percent confidence intervals are indicated. of transitions; differences in the age at the time of the 132 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Education and Poverty transition and in physical access to schools explain at in Kwale to 111 percent in Nairobi. While the levels are most a minor fraction of the difference in transition rates generally much lower, GERs in secondary education into secondary between poor and non-poor students exhibit a similar geographical pattern across counties (Appendix F: Chapter 6 additional materials). However, as GERs in primary education (Figure 6.6b).167 it is less clear what causes the drop in transition rates between the seventh grade of primary and the first 6.2.3 Late enrollment and the transition into grade of secondary. And there is no evidence for an secondary education improvement in the transition rate between 2005/06 Net intake into the first grade of primary remains and 2015/16 at this stage of the education system. low. As mentioned above, NERs in primary education One possible explanation is that schools hold back in Kenya are significantly lower than gross enrollment or discourage students that are projected to perform rates, indicating that a significant fraction of children poorly in the KCPE to boost their mean scores.166 More evidence on this is needed to effectively tackle student are not ”on time” for their respective grade.168 In drop-out and grade repetition at this stage. 2015/16, the net intake rate, defined as the number of children that start first grade as a share of the number Enrollment rates in primary and secondary education of children of official school entrance age, was only vary substantially across counties. Enrollment in 31 percent (Figure 6.7a). It is significantly lower for primary education is nearly universal in the more children from families in the bottom 20 percent of the densely populated counties around Nairobi and near distribution (25.3 percent) and somewhat lower in rural Lake Victoria while most counties in the North and areas (29.7 percent). There is no evidence of significant Northeast lag behind (Figure 6.6a). Overall, the GER improvements between 2005/06 and 2015/16: none of varies from 60 percent in Garissa to close to 126 percent the differences in the net intake rate between 2005/06 in Makueni. The secondary GER varies from 34 percent and 2015/16 are significantly different from zero. Figure 6.6: GERs in primary and secondary education by county, 2015/16 a) Primary b) Secondary (120, 140) (120, 140) (100, 120) (100, 120) (80, 100) (80, 100) (60, 80) (60, 80) (40, 60) (40, 60) (20, 40) (20, 40) (0, 20) (0, 20) Source: Own calculations based on KIHBS 2015/16. 166 There is some indirect evidence that teachers in Kenya have incentives to ”teach to the top” of the achievement distribution, disregarding weaker students (Duflo, Dupas and Kremer 2011). In brief, this is one of the few 167 Appendix F.1 provides Chapter 6 additional materials provides detailed configurations under which tracking, the practice of separating students results at the county-level. by academic ability, would raise achievement for all students – which is 168 GERs in some official reports differ from those reported in the WDI. The what one study finds in a randomized evaluation. Incentives to teach 2017 KES (Kenya National Bureau of Statistics 2017) puts the GER in 2016 to the top could be instilled in teachers if their objective function is to at 104.1, compared to an NER of 89.2, suggesting that approximately maximize their school’s average test score in national examinations. every sixth child enrolled in primary is not of primary-school age. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 133 Education and Poverty Figure 6.7: Net intake rate and transition by poverty, quintile, and locality, 2005/06 and 2015/16 a) Net intake rate into the first grade of primary 60 40 Percent 20 0 Total Bottom 20% Bottom 40% Top 20% Rural Urban By quintile By locality 2005/06 2015/16 b) Transition rate from primary into secondary education 100 80 60 Percent 40 20 0 Total Bottom 20% Bottom 40% Top 20% Rural Urban By quintile By locality 2005/06 2015/16 Source: Own calculations based on KIHBS 2015/16. Note: 95-percent confidence intervals are indicated. Late enrollment in primary education and grade While transition rates into secondary education repetition remains common in Kenya. Late enrollment, have improved in line with higher enrollment combined with a repetition rate of around nine percent rates in secondary, access remains rationed. The in primary education, results in a share of over-age transition rate from the last grade of primary to the children in primary that rises from 50.8 percent in grade first of secondary improved substantially, from 53 one to 64.7 percent in grade eight. Older children face percent in 2005/06 to 74 percent in 2015/16. The higher opportunity costs when studying as they are increase is pronounced in all subgroups, but it is more likely to find gainful employment. Hence, late particularly large for the urban poor (Figure 6.7b). enrollment will be expensive in terms of foregone On the other hand, there is still further room for income. Leading explanations for late enrollment improvement, particularly among the poor. Grade include supply-side constraints, credit constraints, promotion throughout secondary education is high, and insufficient ”school-readiness” (e.g. poor health) suggesting that most students that start secondary (Glewwe and Jacoby 1995). A better understanding of education will also complete it. the causes of late enrollment as well as the associated costs in the Kenyan context are necessary to tackle the issue. 134 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Education and Poverty 6.2.4 Choice of provider including concerns about negative effects on economic Enrollment in private primary education is mobility. There is evidence that private schools produce increasing, particularly in urban areas and among the better learning outcomes (see Box 6.2). better-off. By 2015/16, more than one in five children enrolled in primary were enrolled in a private school 6.2.5 Is education affordable? – up from less than one in ten in 2005/06 (Figure 6.8). Low incomes remain an impediment to enrollment, Private school enrollment has increased in both rural particularly at the secondary level. Evidence from and urban areas and both among poor and non-poor Kenya’s Cash Transfer for Orphans and Vulnerable students. To wit, gross enrollment among children in Children (OVC) program, which provides flat transfers the bottom 40 percent of the population has more than to the caregiver of an orphan or a vulnerable child doubled, from a GER of 5.1 percent in 2005/06 to 12.4 below the age of 18 (Kenya CT-OVC Evaluation Team percent in 2015/16. However, the trend is particularly 2012), suggests that the transfer increases enrollment pronounced among children from urban, better-off for children aged twelve and above by 7.8 percentage families: while three in five enrolled children in this points, nine percent over the baseline mean.170 Another group attended private primary schools in 2005/06, (somewhat dated) study suggests that subsidizing four in five did so in 2015/16. These trends are well- school uniforms, often a significant cost factor documented elsewhere and have been linked to the associated with schooling, increases enrollment and introduction of free public primary education in 2003 learning outcomes (Evans, Kremer, and Ngatia 2008). (Lucas and Mbiti 2014; Bold et al. 2014).169 While primary education is universally affordable, Differences between the poor and the non-poor in secondary education often remains prohibitively uptake of private education raise equity concerns. expensive. Households are spending less than one tenth Private provision is the preferred option among better- of the poverty line per student on items associated with off families in urban areas but plays a less important role primary enrollment (e.g., books, uniforms, and tuition), among other population groups, particularly in rural suggesting that primary education is affordable even areas. To the extent that private schools provide better- for the poor (Figure 6.9a). Private education is more quality education, this finding raises equity concerns, expensive, at around 20-25 percent of the poverty line Figure 6.8: Primary gross enrollment by provider, location, and quintile, 2005/06 and 2015/16 140 120 100 80 Percent 60 40 20 0 2005/06 2015/16 2005/06 2015/16 2005/06 2015/16 2005/06 2015/16 2005/06 2015/16 Total Rural, bottom 40% Urban, bottom 40% Rural, top 60% Urban, top 60% Public Private Source: Own calculations based on KIHBS 2005/06 and 2015/16. Note: 95-percent confidence intervals are indicated. 169 There is no comparable trend for secondary education: enrollment 170 However, there was no evidence for the program’s effect on learning in private secondary schools is still moderately low, at only about ten outcomes and it should also be noted that the program targets percent. And while there is a strong increase in enrollment in public disadvantaged groups. While enrollment was high at baseline, it is higher schools across all subgroups, enrollment in private schools has been still in other population groups, limiting the potential for treatment almost stable. effects in a scaled-up version. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 135 Education and Poverty Figure 6.9: Average and median household per-student expenditure on education by level, location, and provider, 2005/06 and 2015/16 a) Primary b) Secondary 80 80 60 60 Percent Percent 40 40 20 20 0 0 Rural Urban Rural Urban Rural Urban Rural Urban Public Private Public Private 2005/06 2015/16 Median 2005/06 2015/16 Median Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16. Note: 95-percent confidence intervals are indicated. for the median child enrolled in an urban area and 6.3 LEARNING OUTCOMES T somewhat more in rural areas. Secondary education is he knowledge and skills acquired by students much more expensive: median household expenditure are a key dimension of the education system. per child enrolled in a public school is close to 50 Assessments of education systems often focus on percent of the poverty line. enrollment and attainment, which tend to be easier to measure than actual skills. However, what matters for High costs are also the leading reason respondents long-run prosperity are a population’s cognitive skills, cite for non-attendance among drop-outs. When not mere attainment (Hanushek and Woessmann 2015). asked why children that are age-eligible are currently In other words, what matters will be a combination of not enrolled, respondents for children that have never both enrollment and the quality of education children attended a school171 tend to cite parental objection receive. This section provides an assessment of learning (34 percent), the need to work or help at home (21 outcomes in Kenya. percent), as well as children’s age (18 percent).172 Reasons differ for those that have attended school Kenyan students lag substantially behind students at some point but were not enrolled at the time of in Europe, North America, and East Asia. While there the interview: almost two in five respondents cite are few international comparisons in which Kenya high costs associated with school.173 Taken together, participated, some authors have tried to make different evidence on the reported costs of education, regional assessments comparable. For instance, one experimental evidence from interventions that address recent study uses various approaches to link results financial constraints, and reported reasons for drop- from student assessments conducted in Southern out all point to high costs of secondary education as a and Eastern Africa174 to those from the Trends in constraint to higher rates of enrollment. International Mathematics and Science Study, which mostly covers developed countries. Findings suggest that learning outcomes for grade-six students in Kenya are comparable to grade-four students in New Zealand (Sandefur 2018). The average score of Kenyan sixth- 171 The respondent in the KIHBS survey was the child if it was at least ten graders in 2003 would likely place them in the bottom years old, and a guardian for those below the age of ten. five percent of the international ranking. 172 Respondents were allowed to state up to two reasons for being out of school at the time of the interview. 174 These assessments were carried out by the Southern and Eastern Africa 173 Pregnancy is a leading reason for girls that dropped out before Consortium for Monitoring Educational Quality. The data used in the completing secondary education. See Chapter 3. study by Sandefur (2018) was collected in 2000 and 2007. 136 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Education and Poverty However, Kenyan fourth-graders are more 2017). These findings are broadly consistent with knowledgeable than their peers in other Sub-Saharan results from the Uwezo surveys, household-based African countries. Recent results from standardized assessments of children between the ages of six and 16 tests conducted in seven Sub-Saharan African countries that have been conducted since 2009. indicate that fourth-grade students in Kenya in primary schools perform better in literacy and numeracy tasks Learning outcomes vary substantially by socio- than in other countries in the region (Figure 6.10). economic background. The Uwezo assessment is a rapid assessment administered at the household-level Kenyan fourth-graders lag substantially behind (i.e., including out-of-school children). Test items are Kenya’s official curriculum. Only one in two fourth- based on the grade-two curriculum. In other words, graders can identify words and only one in four can eight-year-olds that were enrolled on time would read a paragraph (Figure 6.10). Similarly, only seven be expected to answer all questions correctly by the out of ten are capable of ordering numbers and only standards of the Kenyan curriculum. However, the one in four can complete a simple sequence. Based on 2014-round data suggest that only half of all ten-year- the same data, one study shows that fourth-graders in olds are proficient in grade-two mathematics, in that Kenya after three and a half years of actual education only half were able to demonstrate comprehension have acquired on average only around two and a half of the most difficult topic (division) (Figure 6.11). years of effective education (Bold, Filmer, Molina, et al. Comparisons by household wealth reveal large Figure 6.10: Knowledge of fourth-grade students across Sub-Saharan African countries, early 2010s a) Literacy b) Numeracy 100 100 80 80 60 Percent 60 Percent 40 40 20 20 0 0 Read Read Identify Read a Read a Read Read Identify Read a Read a a letter a word words sentence paragraph a letter a word words sentence paragraph Kenya, 2012 Nigeria, 2013 Tanzania, 2014 Togo, 2013 Kenya, 2012 Nigeria, 2013 Tanzania, 2014 Togo, 2013 Uganda, 2013 Average Uganda, 2013 Average Source: Based on Bold, Filmer, Marin, et al. 2017 and their analysis of SDI data. Figure 6.11: Learning outcomes in mathematics in ten-year-old children by socio-economic background, 2014 100 80 60 40 20 0 Total None Some Some Post Bottom 2 3 4 Top 20% primary secondary secondary 20% By mother's education By quintile of asset index Nothing Counting Numbers Values Addition Subtraction Multiplication Division Source: Own calculations based on data from the 2014 Uwezo survey. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 137 Education and Poverty differences in learning outcomes: while almost two On average, girls perform better than boys in math, thirds of all children in households in the top quintile English, and Kiswahili, especially in earlier grades of are proficient in mathematics, the proportion is only primary. Female advantage in learning outcomes is a little more than one third in the bottom quintile. more likely to be observed in Western and central Similarly, parental education is also highly correlated Kenya. But the pattern in this case is not as clear-cut with learning outcomes in children. Only one in three as for overall enrollment or learning outcomes (see children out of those whose mothers have no formal chapter 3 for details). education are proficient with the standard-two math curriculum by age ten. In contrast, three out of five Private schooling in Kenya is associated with more ten-year-olds whose mothers have some secondary learning and lower operating costs. The empirical education are proficient at that age and four out of five evidence for productivity differentials between of those whose mothers have attended post-secondary public and private education providers is generally education. Results for English and Swahili are similar mixed and likely to be context-specific. 176 For both qualitatively and quantitatively. Kenya, however, differences in learning outcomes between public and private schools are unlikely Learning outcomes vary substantially across to be fully explained by selection of more able counties, following the patterns observed for students into private schools (Box 6.2). One study enrollment. Higher levels of proficiency are evident also finds that most private providers operate at in more densely-populated counties in the center of lower overall costs than public providers (Bold, Kenya while low levels of proficiency are observed in Kimenyi, Mwabu, and Sandefur 2013), a result that the northwest, the northeast and in large counties in is consistent with much lower wage levels for private the east (i.e., Garissa and Tana River) (Figure 6.12). As one school teachers (see next section). would expect, proficiency is spatially highly correlated with enrollment.175 Secondary education in Kenya is associated with large gains in skills and other desirable outcomes. While gender gaps in learning outcomes often favor A recent study exploits the fact that the probability girls, there is considerable variation across counties. of completing secondary education increases Figure 6.12: Proportion of twelve-year-old children proficient in mathematics and english, percent, 2014 a) English b) Mathematics (80, 100) (80, 100) (60, 80) (60, 80) (40, 60) (40, 60) (20, 40) (20, 40) (0, 20) (0, 20) Source: Based on Bold, Filmer, Marin, et al. 2017 and their analysis of SDI data. 176 For instance, a study for Indonesia finds that public school graduates score higher on national exit exams than their privately schooled peers 175 The correlation coefficient between GERs in primary and secondary (Newhouse and Beegle 2006) while another study for Colombia finds and proficiency in mathematics (English) is 0.76 (0.68) and 0.64 (0.78), that private schooling in Colombia was associated with higher scores on respectively. All correlations are significant at the one-percent level. achievement tests (Angrist, et al. 2002). 138 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Education and Poverty Box 6.2: Are private schools more productive? Greater productivity of private education providers, combined with greater uptake among the better-off in recent years, would raise equity concerns. Do Kenya’s private schools offer better-quality education? The question is clearly of great relevance given that enrollment in private primary schools is significantly higher among non-poor children. Estimating the causal effect of school type on learning outcomes is challenging. There is some indirect evidence (presented later in this chapter) that suggests that teacher incentives in the private sector are better aligned with student learning while teachers do not differ in terms of subject knowledge and knowledge of pedagogy. However, establishing the direct causal effect of private school attendance on learning outcomes is difficult because of self- selection of students from better-off families into private schools. Two recent studies suggest that private primary education is more productive than public primary. One study aims to identify the causal effect of private school attendance on grade-eight test scores through aggregation at the district-gender-year level (Bold, Kimenyi, Mwabu, and Sandefur 2013): if changes in mean test scores over public and private school students in a district are correlated with changes in the share of students in private schools, this should reflect differences in the quality of schools, not selection. The authors argue that private provision is associated with a one-standard deviation increase in learning outcomes, a large effect by any standard. However, focusing on eighth-grade students may be problematic when their school trajectories are not observed. Another study employs a battery of controls and different econometric methods to estimate the causal effect of private school attendance in grade two to grade four (Wamalwa and Burns 2017). They find that the private-school premium ranges from 0.13 to 0.18 standard deviations in math and from 0.21 to 0.27 standard deviations in languages. sharply around the admission cut-off177 (Ozier 2016). 6.4 THE SUPPLY-SIDE Comparing students just below and just above the 6.4.1 Physical inputs cut-off (i.e. of similar aptitude), the study finds that completing secondary education increases the adulthood performance on vocabulary and reasoning A significant fraction of schools in Kenya lack appropriate infrastructure; this is true for both urban and rural areas and for private and public tests by around 0.6 standard deviations, a very large improvement.178 Another study finds no evidence institutions. To be effective, schools should provide for differences in the productivity across secondary environments conducive to learning, which includes schools in Kenya (Lucas and Mbiti 2014). Taken together, basic infrastructure, appropriate sanitary facilities, these findings suggest that admitting more students furniture, and adequate learning materials. The 2012 SDI to secondary schools would initially increase human data contain information about classroom conditions, capital. As enrollment rates are high among children learning materials (e.g., pencils, books, and boards), from better-off families, increasing admission into and general school infrastructure. Results indicate secondary schools would also be equitable. However, that almost one third of the classrooms do not meet as noted before, demand-side constraints, specifically the minimum visibility requirement (where the cut- the high costs of attending secondary, also play a major off is 300 lux), with little variation by location or type role in preventing children from poorer families from of provider (Figure 6.13). Similarly, more than one in attending secondary. four schools did not have a clean toilet. Differences between urban and rural public schools as well as 177 The general admission cut-off for secondary schools is set at a score private (for-profit, nongovernmental organization of 250 points in the KCPE. However, elite secondary schools, including in the public sector, usually have higher admission cut-offs (Lucas and [NGO] run, or faith-based) and public schools were Mbiti 2014). small and statistically insignificant. 178 In addition, men who completed secondary education were found to be less likely to be in low-skill self-employment while women were found to have a lower risk of becoming pregnant as teenagers. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 139 Education and Poverty Figure 6.13: Physical inputs at the school-level by location and type of provider, primary schools, 2012 a) Sufficient light b) Clean toilets c) Textbooks per student (English classes) 100 100 80 80 80 60 60 60 Percent Percent Percent 40 40 40 20 20 20 0 0 0 Total Rural, private Rural, public Urban, private Urban,public Total Rural, private Rural, public Urban, private Urban,public Total Rural, private Rural, public Urban, private Urban,public Source: Own calculations based on 2012 SDI data. Note: 95-percent confidence intervals are indicated. The private category includes for-profit and non-profit schools (i.e., NGO- and FBO-run schools). Less than one half of all students had textbooks colleges and universities as public school teachers. during English classes. Basic teaching equipment – Once hired, promotions, transfers, and disciplinary such as blackboards and chalk – is usually available in measures are decided through the TSC and are based Kenyan classrooms. In addition, almost all students were heavily on formal, objective criteria, such as educational found to have pencils and exercise books. However, qualifications and tenure. Reports about long queues the same was not true for textbooks: only 47 and 39 for jobs in the public sector suggest that the overall percent of students in urban and rural public schools, employment conditions are attractive. In some cases, respectively, had a textbook at the time of the visit. The contract teachers are hired to work in public schools proportion was significantly higher in private schools. through a local Parent-Teacher Association (PTA), a local school committee. Teacher training requirements vary However, the lack of textbooks and other similar between primary and secondary school teachers, with physical inputs is unlikely to be a major impediment more demanding training requirements for the latter to learning in the Kenyan context. Randomized (Teachers Service Commission 2007). experiments conducted in Kenya and elsewhere do not support the premise that more physical inputs such as The average student-to-teacher ratio in public textbooks or flipcharts improve learning outcomes in primary schools decreased slightly since 2011, albeit Kenya (Moulin, Kremer, and Glewwe 2009; Glewwe et al. from a high level. As a result of increased enrollment 2004).179 Another study finds that school infrastructure and the hiring freeze established in the late 1990s, and the availability of teaching resources (such as Kenya’s student-to-teacher ratio in public primary blackboards) were uncorrelated with student learning schools increased between 2004 and 2011, from 42 outcomes (Martin and Pimhidzai 2013). to 55. With the end of the hiring freeze in 2010 and a shift in demand towards private provision, it has since 6.4.2 Teachers decreased to a level of 48 students per teacher (Figure 6.14a) There are more teachers per student enrolled Both public school teachers and locally hired teachers in public secondary schools, roughly one for every 30 staff Kenya’s schools. As pointed out before, the TSC, students, with little change in recent years. a central agency, hires graduates of teacher training 179 However, the authors of one study point to the mismatch between the input provided and student needs: textbooks were written in English, The number of students per classroom in fourth the third language for most of the students in their study. Hence, most students were not able to use them effectively. Therefore, it cannot be grade in public schools varies substantially across ruled out that more appropriate inputs would have positive effects on schools. The average number of students per teacher student learning. 140 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Education and Poverty Figure 6.14: Student-teacher ratios in public schools, 2004-2015, and students per classroom in primary schools, 2012 a) Trend in student-to-teacher ratios in public schools b) Fourth-grade students per classroom by location and provider, primary schools, 2012 60 54 55 70 50 52 51 49 50 48 47 48 44 60 50 42 50 40 31 31 30 32 40 30 30 30 30 28 30 30 20 20 10 10 0 0 Rural, Urban, Rural, Urban, Total 2004 2006 2008 2010 2012 2014 public public private private Primary Secondary Average 10th percentile 90th percentile Source: Own calculations based on KES, various instalments, (panel (a)) and 2012 SDI data (panel (b)). Note: 95-percent confidence intervals are indicated in panel (b). The private category includes for-profit and non- profit schools (i.e., NGO- and FBO-run schools). hides considerable variation across schools. Ten percent Teachers in Kenya earn high wages by regional of schools have 57 students or more per classroom standards. Teacher salaries, in conjunction with job while another ten percent have no more than eleven security and other benefits, determine the quality students (Figure 6.14b). Only a minor fraction of the of prospective teachers in the long-run and may variation, around 17 percent, is explained by variation determine teacher effort in the short-run.181 Teacher across counties. One reason private schools may have salaries, expressed as a ratio to per capita GDP, have become more popular in recent years is that they offer been declining in Kenya between 2005/06 and 2015/16, lower number of students per classroom: while public from 3.9 to 2.0 in the case of primary teachers and from primary schools on average have a ratio of 37 students 7.8 to 3.6 for secondary teachers.182 This places primary per classroom, private school classrooms have on school teachers in Kenya in line with their colleagues average only 20 students per classroom. in Uganda (2.1) but above their colleagues in Ghana and Nigeria (1.4 and 1.0, respectively) (Figure 6.15a1).183 Students may benefit from additional teachers; For reference, the OECD average is only 1.3 times GDP yet careful attention must be paid to contextual per capita for primary school teachers (with 15 years of factors. A recent study that assesses the effect of experience). The comparison for teachers in secondary additional teachers180 in Kenya shows that students’ education is similar: with a ratio of 3.6 times GDP per test scores only increased in classes taught by locally- 181 Adequate wages, benefits and job security are necessary to attract talented and motivated teachers. Whether wages will have an hired contract teachers, not in classes taught by public immediate effect on teacher effort, however, will typically depend on the institutional framework (de Ree, et al. 2015). school teachers (Duflo, Dupas, and Kremer 2015b). 182 Findings based on the KIHBS data are broadly in line with TSC pay scales. The authors show that contract teachers had lower The collective bargaining agreement between the TSC and the Kenya National Union of Teachers (KNUT), signed in October 2016, suggests absence rates and argue that public school teachers that teacher salaries are high compared to average incomes. Before tax, the ratio of annual earnings to GDP varies between 1.73, the minimum may have put in less effort in response to additional for entry grade primary school teachers, and 10.51, the maximum for teachers provided. However, a governance program chief principals, the highest pay grade (KNUT 2016). These figures are for rural areas that are not subject to hardship allowances. They include that empowered parents within school committees house and leave allowances (paid once a year to all teachers serviced by the TSC) but exclude other allowances that teachers may be eligible had attenuated those negative effects on public school for (including responsibility allowance, commuter allowance, hardship allowance, transfer allowance, and special education allowance). teachers’ behavior. This suggests that the incentives 183 Surveys used differ in terms of how primary and secondary school and supervision that teachers face play a key role in teachers can be identified: the Kenyan surveys use the three-digit Kenyan National Occupation Classification Standard (KNOCS) (codes their performance. 370 and 252), the Nigerian and Ghana surveys use the four- and three- digit 1988 ISCO standard (codes 2331 and 2320, and 233 and 232, 180 The study analyzes the Extra Teacher Program (ETP), which provided respectively), and the Ugandan survey uses the 2008 ISCO (codes 2341 funds to randomly selected schools to hire an additional teacher. and 2330, respectively). KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 141 Education and Poverty Figure 6.15: Cross-country comparison of teacher salaries by level a1) Primary level, ratio of GDP per capita b1) Primary level, int. US$ per month 5 4,000 4 3,000 3 2,000 2 1,000 1 0 0 Kenya, Kenya, Ghana, Nigeria, Uganda, Kenya, Kenya, Ghana, Nigeria, Uganda, 2005/06 2015/16 2009/10 2015/16 2013/14 2005/06 2015/16 2009/10 2015/16 2013/14 Mean Median Mean Median a2) Secondary level, ratio of GDP per capita b2) Secondary level, int. US$ per month 10 10,000 8 8,000 6 6,000 4 4,000 2 2,000 0 0 Kenya, Kenya, Ghana, Nigeria, Uganda, Kenya, Kenya, Ghana, Nigeria, Uganda, 2005/06 2015/16 2009/10 2015/16 2013/14 2005/06 2015/16 2009/10 2015/16 2013/14 Mean Median Mean Median Source: Own calculations based on WDI, KES 2017 (2012-2016 data) and KES 2013 (2008-2012 data), and KIHBS 2015/16. Note: Data on gross enrollment tabulated across different volumes of the KES were not always consistent. See also notes to Figure 6.3 capita, teachers in Kenya earn similar relative salaries by a fall in the average age of teachers and in their to those in Uganda (2.9) and considerably more than probability of working in the public sector,184 with the their counterparts in either Ghana (1.9) or Nigeria (1.6) remainder is potentially accounted for by slow growth (Figure 6.15a2). in nominal wages. As a consequence of slow nominal wage growth Subject knowledge of Kenyan primary school and changes in the composition of the workforce, teachers is high by regional standards, but average teachers’ average salaries have been declining in real scores suggest that teachers are often struggling with terms. Estimates from the KIHBS suggest that average the curriculum they are supposed to teach. Teachers salaries for both primary and secondary school teachers require a deep understanding of the subjects they have been declining by more than half in real terms, teach as well as pedagogical skills in order for learning although they remain higher than in other countries in 184 Based on OLS regressions of log salaries on age, gender, a public-sector dummy, and a dummy for 2015/16, it was found that real salaries for the region (Figure 6.15b1-b2). Inflation may be to blame both primary and secondary teachers increased by about 2.4 percent with each year of age. The average age declined by 2.5 and 1.2 years, for some of the decrease but not all of it. Decomposing respectively, where only the former estimate was found to be statistically the decrease suggests that about two thirds of the significant. The public-sector premium in these regression was very high, in excess of 100 percent for both primary and secondary, in line with decline in real wages for primary and about one third recent results in the academic literature (Barton, Bold, and Sandefur 2017). In the KIHBS sample, employment in the public sector decreased of the decline for secondary teachers can be explained by 25 percentage points for primary school teachers and by ten percent for secondary school teachers. 142 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Education and Poverty to take place. Compared to teachers in other Sub- 6.5 TEACHER INCENTIVES AND SCHOOL Saharan African countries for which comparable data GOVERNANCE A are available, Kenyan primary school teachers score gap exists between what teachers in Kenya are high on subject knowledge tests (Figure 6.16). Still, the capable of doing and what they actually are average fraction of correct answers on the English and doing in practice. An important factor for the effective mathematics tests, which were based on the fourth- delivery of education services is whether teachers’ grade curriculum, were only 64.6 and 80.6 percent, incentives are aligned with student learning. The respectively. There is no evidence for differences in previous section suggests that teachers in Kenya often teacher subject knowledge by type of provider. have better subject knowledge than teachers in other countries in the region. They also tend to be better At the same time, most Kenyan teachers are paid in real terms. However, this section argues that unfamiliar with basic pedagogy. On average, teachers while teacher incentives are well-aligned with student answered only 36 percent of the questions related to learning, there is still a gap when it comes to teaching basic pedagogy correctly. In this domain, teachers in practice. In line with better learning outcomes in Kenya achieved scores similar to their colleagues in children that attend private institutions, this gap seems Tanzania and Senegal but still better than teachers in to be more pronounced in the public sector. Madagascar, Mozambique, Togo, or Uganda. The lack of pedagogical skills among Kenyan teachers is also As in other countries of Sub-Saharan Africa, evident in observational studies of teacher-student absenteeism among teachers in Kenya is rampant. interactions: individual seat work and purely teacher- High rates of teacher absenteeism and large centered activities (e.g. instructions, demonstrations, discrepancies between time spent teaching and the lesson reviews) take up most of the time of a typical scheduled teaching time are common across Sub- lesson in Kenya (Ngaware, Oketch, and Mutisya 2014). Saharan Africa and have been found to negatively Teacher-led recitations, including highly-ritualized affect student learning (Bold, Filmer, Marin, et al. 2017). choral responses by students, are often the dominant While Kenyan teachers appear knowledgeable by form of teacher-student interactions in Kenyan primary regional standards, they are more likely to be absent schools (Pontefract and Hardman 2005). Teachers rarely from class than teachers in Madagascar, Niger, Nigeria ask open questions that would require students to and Togo (Figure 6.17). It is interesting to note that only explain their reasoning or expand on a thought, and around 16 percent of teachers were absent from school explicit feedback is rare. while more than 40 percent were absent from class. Figure 6.16: Teachers’ subject knowledge and pedagogical skills by country, early 2010s 90 80 70 60 50 Percent 40 30 20 10 0 Language Math Basic Pedagogy Kenya, 2012 Madagascar, 2016 Mozambique, 2014 Senegal, 2010 Tanzania, 2014 Togo, 2013 Uganda, 2013 Average Source: Own calculations based on SDI database. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 143 Education and Poverty Figure 6.17: Absence from school and absence from class by country a) Absence from school b) Absence from class Uganda, 2013 23.8 Uganda, 2013 52.5 Togo, 2013 18.4 Togo, 2013 33.6 Tanzania, 2014 14.4 Tanzania, 2014 46.7 Nigeria, 2013 13.7 Nigeria, 2013 19.1 Niger, 2015 16.6 Niger, 2015 27.0 Mozambique, 2014 45.0 Mozambique, 2014 56.0 Madagascar, 2016 30.7 Madagascar, 2016 37.9 Kenya, 2012 15.5 Kenya, 2012 42.4 0 10 20 30 40 50 0 10 20 30 40 50 60 Percent Percent Source: Own calculations based on SDI database. Note: The vertical line indicates the average. This suggests that teachers are not so much involved second teacher in public schools was absent during in other income-generating activities outside of school an unannounced visit, the rate was ten percentage but have poor incentives to actually teach while in points lower in schools run by NGOs or faith-based school. High absenteeism rates are also consistent organization. Teacher absenteeism is less pronounced with earlier estimates for the Busia and Teso districts among private-for-profit schools at only 14.8 percent. (Glewwe, Ilias, and Kremer 2010). Importantly, that Head or deputy head teachers in public schools had study demonstrates that high average absenteeism is the highest absenteeism rates at 72.4 percent, followed the result of many teachers being absent occasionally, by public school teachers (43.6), contract teachers in not of few teachers being absent all the time. public schools (39.5), and contract teachers in private schools (33.9) (Figure 6.18b). It should be noted that Absenteeism rates are lower among teachers the difference between public school (government) working for private providers and contract teachers. teachers and contract teachers in private and public Absenteeism rates across all types of teachers are much schools remains large after controlling for age, gender, higher for teachers working in the public sector than and educational attainment. Together with evidence for either teachers working for nonprofits or private- presented above on learning outcomes by type of for-profit schools (Figure 6.18a). While almost every provider and teacher salaries, this finding raises further Figure 6.18: Absenteeism rates by type of provider and type of teacher, 2012 a) By type of provider b) By type of teacher and provider 60 80 72.4 47.7 50 36.5 60 40 43.6 39.5 33.9 Percent Percent 40 30 14.8 20 20 10 0 Head or Government Contract Contract 0 deputy teacher teacher teacher Public NGO/Faith-based Private-for-pro t head teacher (public) (public) (any private) (public) Source: Own calculations based on SDI 2012. Note: 95-percent confidence intervals are indicated. 144 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Education and Poverty questions regarding the differences in educational is no evidence so far that the reform has improved production between private and public providers teacher effort due to lack of adequate data. A new (Box 6.3). assessment of teacher presence and practice along the lines of the SDI survey would be needed to shed A recent reform aims at improving teacher more light on this question and is currently scheduled performance through closer monitoring by superiors. for 2019/20. However, empirical studies of incentive Until recently, the probability of being employed as a schemes that relied on monitoring of teachers teacher and of being promoted depended largely on through headmasters suggest that they can be formal qualifications and grades in teacher training as ineffective. An inputs-based incentive intervention in well as time spent in the queue, i.e. years past since Kenyan preschools in which teachers were eligible for graduation from teacher training (Wanzala 2016; Bold, attendance bonuses had no effect on absenteeism or Kimenyi, and Sandefur 2013). In 2016, a performance most measures of teacher pedagogy (Chen et al. 2001). evaluation system (Teacher Performance Appraisal The authors attribute this result largely to the fact that Development tool or TPAD) was introduced by the headmasters were administering the incentive scheme. TSC with the support from the Global Partnership for In a similar fashion, another study finds no effect of Education.185 Teachers are now meant to be evaluated prizes given for good teaching on teacher absence in by their superiors, with criteria including the preparation an experiment in which the task of allocating prizes falls of lesson plans, the extent to which the syllabus is to school committees, some of which were controlled followed, as well as attendance and observance of by headmasters (De Laat, Kremer, and Vermeersch effective time use (Kiplang’at 2016). 2008). Finally, one study for Uganda finds evidence that headteachers were less likely than other school-board While there is no evidence on the effects of this members to hold teachers to account (Barr and Zeitlin particular reform to date, studies suggest that 2011). It is also worth noting that headmasters and monitoring by superiors within schools is often their deputies have had higher absenteeism rates in ineffective. While the TSC reports improvements, there the 2012 SDI data (Figure 6.18b). Box 6.3: Are higher public-sector wages efficient? The analysis in this chapter suggests that public school teachers in Kenya earn significantly higher wages than their counterparts in private schools. In theory, a public-sector premium could be efficient if it reflects a compensating differential, if it results in positive selection, or if it succeeds in eliciting higher levels of motivation (i.e., if efficiency wages are offered). However, the evidence presented here is not consistent with a public-sector premium as an efficient reward for talent and effort. On average, public school teachers have similar knowledge in terms of subject content and pedagogy, yet were less likely to be teaching during unannounced visits. In addition, there is some evidence for a greater effectiveness of private provision in producing educational outcomes. A recent study suggests that higher salaries earned by public school teachers in Kenya reflect inefficient rents. The hiring freeze on public school teachers ended in 2010 when the GoK recruited 18,000 new public school teachers. A recent quasi-experimental study exploits this natural experiment (Barton, Bold, and Sandefur 2017). The empirical strategy employed allows the authors to credibly rule out differences in observable and unobservable teacher characteristics (positive selection) as an explanation for the public-sector premium. The authors find that applicants that obtained jobs in the public-sector but had otherwise identical characteristics earned a wage premium of KSh 10,000, more than 100 percent. At the same time, the authors find no evidence for an effect of these jobs on motivation. Hence, both compensatory and efficiency wages are unlikely explanations for the observed wage premium, leaving pure rents as the only plausible explanation. 185 See https://www.globalpartnership.org/blog/transforming-teaching- kenya. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 145 Education and Poverty Parents can play a key role in monitoring teachers. issues in the education sector of Kenya: (1) improving Academic studies provide insights of how enhancing access to quality education for the poor, particularly school governance can overcome some of the most at the secondary level, (2) managing the trade-off pressing constraints to effective learning in Kenya. between increasing costs and the provision of quality Parents can play an important role in improving inputs, particularly teachers, and (3) strengthening school governance and teacher incentives, including teacher incentives and school governance. reductions in absenteeism rates, if their mandate and ability to monitor teachers is strengthened (Duflo, Increasing secondary school enrollment among Dupas, and Kremer 2015b). Raising the stakes for the poor requires demand-side interventions. parents in their children’s academic performance, for Completion of secondary school is associated with a instance, by providing financial support to successful range of positive outcomes during adulthood. But while students, can strengthen their incentives to monitor enrollment in secondary has increased among the teachers (Friedman et al. 2016). poor, significant gaps persist. Further analysis presented in this chapter and academic research suggest that While the hiring of more contract teachers in public increasing enrollment in secondary education in Kenya schools seems to be a promising idea to reduce costs requires primarily demand-side interventions. Cash without hurting the quality of education, the political transfers have already proven effective in increasing economy of taking this idea to scale is challenging. enrollment rates in Kenya. Contract teachers were more efficient in a limited experimental study (Duflo, Dupas, and Kremer 2015b). 186 In the medium term, greater reliance on contract However, a scaled-up program implemented by the GoK teachers to initially fill vacant positions should that aimed at hiring 18,000 additional contract teachers be combined with close monitoring of the recent – almost one per public primary school – faced several overhaul of teacher hiring and retention practices. implementation constraints. The scaled-up version also Contract teachers earn lower salaries, have average altered the political economy in a way that lowered its levels of subject and pedagogical knowledge, and effect on test scores: one study found that the program lower rates of absenteeism. Hence, greater reliance had similar effects on test scores as in the original study on contract teachers to initially fill vacant positions when it was run by an NGO. But no discernable effect would seem to be a way to supply additional teachers was found when it was run by the government (Bold, at low cost. Moving to an ‘up-or-out’ promotion system Kimenyi, Mwabu, Ng’ang’a, et al. 2013).187 in which the best-performing contract teachers are promoted to public school teachers may have large 6.6 SUMMARY AND POLICY OPTIONS potential dynamic benefits (Duflo, Dupas, and Kremer T aken together, findings in this chapter suggest 2015b). Ultimately, a system in which teachers start three, intertwined policy challenges. The analysis their careers as contract teachers and receive tenure presented in this chapter identified three main policy conditional on performance puts in place incentives and improves the selection of teachers into tenured positions.188 However, the threat of a discontinuation 186 Recent experimental evidence from India supports the notion: contract of employment would have to be credible. The teachers improved learning outcomes substantially (Banerjee, et al. 2007, Muralidharan and Sundararaman 2013). See also Bruns, Filmer, & effectiveness of recently introduced monitoring and Patrinos (2011) for a review of non-experimental studies of the effect of contract teachers. evaluation systems should be closely followed. While 187 The authors show that local capture in the form of a larger share of they have the potential to improve teacher effort, it positions filled with relatives of individuals involved in the hiring process was more common in the government arm. In addition, there was a lack is not clear whether head masters and deputy head of top-down accountability: district-level employees of the MoE failed to conduct monitoring visits and to report back to the central government, masters are best placed to monitor teacher presence resulting in higher rates of teacher absenteeism and delays in payments. and performance. Finally, the program also faced resistance from Kenya’s teacher unions, which demanded that newly-hired contract teachers be eventually 188 Combining their empirical estimates with assumptions about the steady turned into civil servants. Contract teachers in the government arm state-share of contract teachers, Duflo, Dupas, and Kremer (2015b) seemed to have anticipated this outcome, which in turn changed their reckon that such a promotion system might increase test scores by 0.18 incentives and adversely affected their performance. standard deviations. 146 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Education and Poverty School governance might benefit from greater ensure equitable access to private provision and to involvement of local stakeholders, particularly ensure minimum quality standards through monitoring parents. Empirical evidence suggests that the local (Romero, Sandefur, and Sandholtz 2017). knowledge of stakeholders, particularly parents, may play a key role in monitoring teachers at the school- Recent events show that such policies are likely to level. Putting local stakeholders in charge of monitoring be met with solid resistance from teachers’ unions. and evaluating teachers may help improve teacher While the above proposals, an overhaul of the hiring attendance and, thus, students’ test scores. Moreover, process, an increase in the mandate of local school it is important to pay close attention to how incentives committees, a shift of financial resources, and a more are structured so that stakeholders themselves have a active approach to private involvement, have the strong interest in improving learning outcomes. Parents potential to increase both access to and quality of will likely need training and information to effectively education at modest costs, attention to the details of undertake monitoring. It remains to be seen in this their implementation are of utmost importance. In context whether the creation of Boards of Management addition, recent experience suggests that the political in 2013 resulted in significant improvements. economy of reforms along those lines is difficult. Greater reliance on contract teachers and private Along with greater local oversight, schools could provision has already been met with solid resistance be given more resources and greater autonomy to on the part of teachers’ unions (Bold, Kimenyi, Mwabu, use them. Under the current system, schools receive Ng’ang’a, et al. 2013). capitation grants, fixed per-student payments. But these have been pegged at very low levels and have never There are several areas that require further analysis, been adjusted for inflation. In addition, they cannot be including late-enrollment in primary, the transition used to pay for salaries. Increasing the capitation grant, from primary into secondary education, and follow- along with greater autonomy to school committees to up assessments of service delivery. Late enrollment, recruit, retain, and promote teachers, has the potential particularly among the poor, remains a concern. to improve teacher performance and to lower school More research is required to inform policies aimed dropout rates. at increasing the net intake into the first grade of primary. And while the transition rate from primary The potential of a greater involvement of private into secondary has improved between 2005/06 and providers should be explored. In recent years, 2015/16, it is still comparatively low among the poor. enrollment in private primary schools has been High costs associated with secondary education still increasing. Evidence reviewed in this chapter suggests constitute a barrier to access but a more detailed that private primary schools in Kenya are more understanding of this would shed some light on productive and often operate at lower costs than public the formulation of policies aimed at boosting timely schools (Bold, Kimenyi, Mwabu, and Sandefur 2013). access to education among disadvantaged children. There is also some evidence that teachers in private In addition, this report has heavily relied on data from schools have higher levels of motivation and that greater the 2012 SDI, which has proved valuable in assessing enrollment in private schools has freed up resources in different dimensions of the education system in Kenya. the public sector. Given the growing importance of But these data are somewhat dated now. A new private provision of education in Kenya documented dataset would help to understand whether recent in this chapter, current oversight arrangements and reforms and initiatives have made a difference. Finally, regulations should be reviewed and strengthened. The there are several areas of the education system whose government has a vital role to play in markets with both link to poverty remains unexplored. These include early public and private providers, particularly in providing childhood development, technical education and information to parents (Andrabi, Das, and Khwaja 2017). tertiary education. It also needs to have the capacity to craft contracts that KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 147 CHAPTER 7 HEALTH AND POVERTY SUMMARY Kenyans have experienced significant and equitable gains in a range of population health indicators over the past ten to 15 years. Driven by increased uptake of low-cost, high-impact technologies and declining fertility, under-five mortality has fallen by more than 50 percent between 2003 and 2014.189 The proportion of children under five that are chronically malnourished has declined by almost ten percentage points over the same time period (DHS STATcompiler, 2018). While still high and estimated with considerable margins of error, maternal mortality, the number of maternal deaths per 100,000 livebirths, likely also declined over this time-period.190 Improvements in uptake and outcomes were often more pronounced among the poor. For instance, while the children of the poorest 20 percent of Kenyans191 were almost 50 percent more likely to die before their fifth birthday than the children of the richest 20 percent in 2003, the gap had declined to only a little more than ten percent by 2014. New challenges for the Kenyan health sector are quickly emerging. While progress has been robust in many domains, this chapter will argue that there are still pronounced socioeconomic and geographic disparities in health access and outcomes that warrant action. At the same time, new challenges such as the increasing burden of non-communicable diseases, concerns about the sustainability of healthcare financing,192 and disruptive labor disputes have added to the challenges Kenya’s health sector is facing. Recent reforms and policy initiatives have the potential to address some of these challenges. The devolution of health service delivery in 2013, the removal of user fees in public facilities for basic services and deliveries in the same year, and the more recent focus on UHC as one of the “Big Four” priorities demonstrate steps the GoK’s commitment to equitable access to quality health services. However, implementation presents challenges, outcomes should be carefully monitored, and adjustments should be made as needed. For instance, early reports suggest that a lack of coordination between the central and county governments in the months following devolution had adverse effects on service delivery. Removal of user fees for deliveries has mainly led to a shift in demand from private provision to public provision among urban, better-off women, suggesting that supply-side policies are an important ingredient in the policy mix that will eventually allow the GoK to reach its goal of UHC. 189 190 191 192 189 One study suggests that the increase in uptake of ITNs alone accounts for close to 80 percent of the decrease in infant mortality between 2003 and 2008. 190 The maternal mortality ratio is based on sibling death histories and is estimated in the case of Kenya with high levels of uncertainty: the 95-percent confidence bands in this case range from 398 to 614 deaths in 2003 and from 254 to 471 deaths in 2014. Hence, the difference in the estimates is not significant at conventional levels of statistical significance. 191 As these rates are based on data from the KDHS, the variable used to construct quintiles is an asset index, not consumption expenditures. 192 A recent World Bank report on the financing of priority programs, including immunization campaigns and programs to address HIV/AIDS, malaria, tuberculosis, and reproductive health, shows that these programs face funding gaps despite being heavily reliant on resources provided by development partners. The report estimates that closing the combined funding gap for these programs would require the GoK to increase health spending by more than 50 percent (World Bank 2018a). KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 149 Health and Poverty 7.1 BACKGROUND consists of county referral hospitals. They are primary and secondary hospitals that provide both outpatient 7.1.1 Kenya’s health sector: key characteristics and and inpatient care. The fourth tier, the national referral recent developments facilities that offer highly specialized care, is used for I mprovements in the quality and efficiency in health service delivery can help the poor to move out of poverty and protect the non-poor from falling training and support research (Ministry of Health 2013). Both public and private providers play an important into poverty. Broad access to quality health services role in Kenya’s health system. A large and increasing provides the foundation for healthy societies. Yet there part of the population — and especially among the is also a strong empirical relationship between health poor — relies on public health services (Figure 7.1). and equitable economic growth that operates at For instance, seven out of ten episodes of outpatient various levels, with causality running in both directions: care among the poorest 40 percent are provided by improved living standards can often lead to better government facilities, compared to only five in ten health, and better health also improves the material among the richest 20 percent. The private sector, which standard of living. In addition, an efficient health system, includes both for-profit providers and non-profits (faith- including adequate protection from catastrophic out- based organizations and NGOs), still plays a significant of-pocket health expenditure, secures livelihoods and thus protects individuals from falling into poverty. role, particularly in urban areas and among better-off Kenyans (section 7.2.3). Kenya’s health care system is currently organized around six levels of care that fit into four tiers of The GoK remains strongly committed to improving care, based on the scope and complexity of services health care delivery. The GoK’s Vision 2030 stipulates offered. The basic unit is the community unit staffed by a two-pronged approach to building an efficient and community health workers. These comprise the first tier. high-quality health care system: (i) the devolution Primary care facilities, dispensaries and health centers, of funds and management of health care from the comprise the second and third level, both part of the central government to counties and (ii) a shift of second tier. They provide basic preventive and curative expenditures from curative to preventive care services. care including health services for childbirth. Health It also recognizes the need for additional efforts to tackle centers also provide basic inpatient services, including HIV/AIDS, malaria, and tuberculosis as well as lowering deliveries. The top three levels are hospitals that focus infant and maternal mortality. More recently, the GoK’s more on curative care and rehabilitation. The third tier “Big Four” agenda includes UHC as one of four pillars. Figure 7.1: Outpatient visits and institutional deliveries by provider, January 2012 to December 2017 a) Health facility attendance (outpatient visits, millions) b) Institutional deliveries by month, thousands, Jan 2012 - Dec 2017 80 12 70 10 60 8 50 40 6 30 4 20 2 10 0 0 Jan - 12 Apr - 12 Jul - 12 Oct - 12 Jan - 13 Apr - 13 Jul - 13 Oct - 13 Jan - 14 Apr - 14 Jul - 14 Oct - 14 Jan - 15 Apr - 15 Jul - 15 Oct - 15 Jan - 16 Apr - 16 Jul - 16 Oct - 16 Jan - 17 Apr - 17 Jul - 17 Oct - 17 Jan - 12 Apr - 12 Jul - 12 Oct - 12 Jan - 13 Apr - 13 Jul - 13 Oct - 13 Jan - 14 Apr - 14 Jul - 14 Oct - 14 Jan - 15 Apr - 15 Jul - 15 Oct - 15 Jan - 16 Apr - 16 Jul - 16 Oct - 16 Jan - 17 Apr - 17 Jul - 17 Oct - 17 Private/faith-based/NGOs Ministry of Health Private/faith-based/NGO Ministry of Health Source: Own calculations based on data from Kenya’s District Health Information System (DHIS 2). 150 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Health and Poverty Figure 7.2: Levels and trends in health expenditure by source, 2004-2014 a) Public health expenditure (percent of GDP) b) Total health expenditure (percent of GDP) by financing source 8 6 South 6 1.7 Africa 4 0.8 1.9 1.2 Percent Percent 4 1.4 Kenya 2.8 2.7 2.0 Tanzania 2.0 2 1.8 2 Ghana Uganda 2.3 1.9 1.5 1.6 1.4 0 0 1,000 10,000 2001/02 2005/06 2009/10 2012/13 2015/16 GDP per capita (2011 PPPs, log scale) Public Private Donors Other Source: Own calculations based on CEQ database (http://commitmentoequity.org/) and KNBS 2016 (panel (a)) and Ministry of Health 2017 (panel (b)). Policies to prevent the transmission of communicable to Kenya’s regional peers. Across countries, public diseases, particularly malaria and HIV/AIDS, have health spending as a share of GDP tends to increase by been main priorities for the Gok in recent years. Kenya a little less than a tenth of a percentage point for every adopted several measures to fight malaria since the ten-percent increase in GDP per capita (Figure 7.2a). In early 2000s. ITNs have been distributed free of charge Kenya, it was around two percent of GDP per capita to children under five and pregnant women since in 2015/16, only slightly higher than in neighboring 2006 and to all age groups since 2010 (WHO 2018). Uganda and Tanzania and in line with Kenya’s level of Intermittent preventive treatment in pregnancy, a full economic development. therapeutic course of antimalarial medicine given to all pregnant women at routine antenatal care visits, was While total health expenditure as a share of GDP adopted in 2001, and residual spraying of insecticides increased moderately between 2005/06 and was adopted in 2003.193 In its National Malaria Strategy 2012/13, it declined more recently. Between 2012/13 2009-2017, the Kenyan Ministry of Public Health and and 2015/16, total health expenditure (including Sanitation announced a new goal of universal coverage public health expenditure, private sources, and donor with long-lasting ITNs for populations at risk by 2013 contributions) declined by about nine percent in real (Ministry of Public Health and Sanitation, 2009). Similarly, terms and by 1.7 percentage points of GDP (Figure the battle against HIV/AIDS has featured prominently in 7.2b). This translated into a decline in per capita terms national strategies.194 by almost one third, from KSh12,000 to less than KSh8,000 per person (in 2015 constant prices). Lower At about two percent of GDP, the level of public health private funding accounts for around half of the decline expenditure in 2015/16 is in line with countries at in real per capita health expenditure. similar levels of economic development and is similar While the ratio of donor financing to total health 193 Both ITNs and residual spraying have been demonstrated to be highly expenditure has decreased since 2009/10, Kenya’s effective against malaria, including in RCTs in Kenya (Guyatt, et al. 2002). health sector is still highly reliant on donors. The 194 Kenya’s Vision 2030 document identifies HIV and AIDS as “one of the greatest threats to socio-economic development in Kenya” and envisions government’s share in total health expenditure was 37 a Kenya free of HIV infections, stigma, and AIDS-related deaths. The GoK, through the National AIDS Control Council, has developed the percent in 2015/16, up from 29 percent in 2005/06. The Kenya AIDS Strategic Framework 2014/2015 to 2018/2019 to provide share of private expenditure declined between 2001/02 guidance on the country’s priorities in HIV programming and increase the effectiveness of the national response, which stipulates a reduction and 2005/06 and then stagnated at around 40 percent in new HIV infections by 75 percent by 2030. The framework will build on and succeed the Kenya National AIDS Strategic Plan 2010–2013 (KNASP over the subsequent ten years. Donors have more than III) and aims to contribute to achieving goals defined in the Vision 2030 (National AIDS Control Council 2015). doubled their share over the course of the 2000s, from KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 151 Health and Poverty 16 percent in 2001/02 to 35 percent in 2009/10. But The devolution of health service provision to the their share has since declined to 23 percent in 2015/16. counties has the potential to improve accountability. However, some priority programs still rely heavily on An important provision of the new Constitution donor funds. For instance, almost three quarters of total adopted in 2011 was the devolution of health services expenditure on HIV/AIDS was financed by donors in to the newly-created 47 counties. The health service fiscal year 2013/14 (World Bank 2018a). delivery function was formally transferred to counties in August 2013.195 One third of the total devolved Coverage through the NHIF has increased rapidly budget in 2013/14 was earmarked for health and in recent years. Membership in Kenya’s NHIF, a most health workers became employees of counties. state entity with the mandate to provide social However, health policy, the management of national health insurance, is mandatory for all formal sector referral hospitals, and capacity building remain the employees (public and private) and voluntary for responsibility of the national government (GoK 2010). those in the informal sector. Contributions are Nevertheless, this was a radical departure from the calculated on a graduated income scale for the formal highly centralized form of governance that had been sector and at a fixed rate for the informal sector. The in place since independence and which was often benefits package, which has been expanded in 2016 seen as resulting in both political and economic (Healthy Nation 2017), includes medical consultation, disempowerment as well as an unequal distribution of lab work, drug administration and dispensing, dental resources (World Bank 2012) (Box 7.1). health care, radiology examinations, nursing and midwifery services, surgical services, radiotherapy, The introduction of free basic health services in and physiotherapy (NHIF 2015). According to the public facilities, including the free provision of 2017 KES, membership has increased from less than deliveries, in June 2013, resulted in increased uptake two million in 2006/07 to more than six million in and a shift in demand towards public provision of 2014/15 (Figure 7.3). Estimates based on the 2015/16 basic health services. In June 2013, the GoK moved to KIHBS data suggest that 8.1 million Kenyans were abolish all user fees in public dispensaries and health covered through the NHIF in 2015/16 while only 0.5 centers and made deliveries in all public facilities free million had different health coverage (section 7.2.3). of charge.196 This policy, perhaps in conjunction with Figure 7.3: Membership and resources of National Hospital Insurance Fund (NHIF), 2006/07-2014/15 a) Membership (millions) b) Resources (KSh billions) 7 20 6 15 5 4 10 3 2 5 1 0 0 2006/07 2008/09 2010/11 2012/13 2014/15 2006/07 2008/09 2010/11 2012/13 2014/15 Informal sector Formal sector Receipts Bene ts Contributions net of bene ts Source: Own calculations based on KES 2017. 195 Gazette Supplement No. 116, Legal Notice 137 of August 9, 2013. 196 User fees, introduced in Kenya and in many other developing countries in the late 1980s, increasingly came to be seen as a barrier to access. In response, many African countries introduced partial or total elimination of user fees in the 2000s (Meessen et al. 2011). 152 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Health and Poverty Box 7.1: Promises and perils of the devolution of health services Devolved government presents an opportunity to address the diversity of local public health challenges. Kenya is a diverse country with ten major and more than thirty minor ethnic groups and marked by large spatial disparities. Needs in terms of health services vary widely. For instance, many rural areas still need to catch-up in providing basic health services while more developed urban centers are facing a rapidly progressing epidemiological transition. With these stark differences, it makes little sense to provide the same mix of services across counties. Counties might be more accountable and provisions for increased transparency and participation might help to keep a lid on corruption. On the other hand, massive changes in processes imply severe challenges. Capture of resources by local elites and low capacity to absorb resources in some counties were seen as major risks associated with devolution (World Bank 2012). In addition, a recent assessment suggests significant variation in the degree to which counties are ready to take on full responsibility for health service delivery, as measured by health care accessibility and counties’ ability to generate revenue (Barker, et al. 2014). Nevertheless, political pressure from the newly elected county governments led to a bulk transfer of functions, irrespective of the counties’ level of preparedness. More recently, an analysis based on data from national health accounts generated for twelve pilot counties still found high levels of out-of-pocket spending. The analysis also found large variation in both per capita spending on health in 2013/14 and 2014/15 and the share accounted for by county government health expenditure (Maina, Akumu and Muchiri 2016). Early reports often pointed to disruptions in service provision due to a low level of preparation, but more empirical evidence is required to shed light on the effects of devolution on health outcomes and equity in access. A low level of preparedness to provide health services effectively on the part of some counties resulted in problems such as the disruptions in staff salary payments and delays in procurement of essential medicines and medical supplies (Tsofa, et al. 2017). One area in which improvements should be monitored is the allocation of health professionals across counties. Section 7.3.2 takes a closer look at health professionals in Kenya. It documents both a general shortage as well as a maldistribution of nurses in the public sector across counties prior to devolution. However, a redistribution of resources and authority across counties has the potential to address at least the second problem. Outcomes in this regard should be closely monitored. the devolution of responsibilities to the counties, has health workers have increasingly resulted in disruptions resulted in a relative shift in demand away from private to service provision. For instance, strikes starting in late- provision and towards public provision that is evident 2016 and mid-2017 paralyzed operations in public for outpatient visits and institutional deliveries alike health facilities. Outpatient visits and deliveries in (Figure 7.1). However, the directive, which took effect public facilities have both declined dramatically during immediately, reportedly took many health professionals these episodes. While the former seem to have been in the public sector by surprise: for instance, there were skipped altogether, an increase in deliveries in private several reports of overcrowding and stock-outs at facilities has partly compensated for the decrease of public maternity hospitals (Cherondo 2013). those in public facilities (Figure 7.1). Recent studies, including one for Kenya, find adverse effects of labor Labor disputes between the government and public- strikes of health workers on health outcomes, including sector unions197 have resulted in disruptions in the higher rates of child mortality and reduced vaccination supply of health services. The frequency of labor strikes rates (Friedman and Keats 2017). increased in recent years and walkouts of public-sector 197 The new constitution included a bill of rights that gave every Kenyan worker the freedom to join a trade union and compelled every employer to recognize employees’ trade unions. This allowed medical doctors, who previously had no right to unionize, to form a union for the first time. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 153 Health and Poverty 7.1.2 Population health and the demographic and Total fertility fell more rapidly than expected given epidemiological transitions Kenya’s growth in GDP per capita. The TFR, the average Except for maternal mortality, most indicators number of births per woman, has declined from 5.2 in of population health are close to what would 2000 to only 3.9 in 2015 (Figure 7.6a). While the former be expected given Kenya’s level of economic TFR was higher than expected based on Kenya’s GDP development and are typically better than regional per capita, the latter is well in line with other countries averages. In Kenya, mortality rates in children and at comparable levels of GDP per capita. Similarly, under- adults are usually lower than in typical low-income five mortality was much higher than one would expect countries yet higher than in typical middle-income based on GDP per capita in 2000 but is now lower than countries (Figure 7.4). This also holds for the stunting expected (Figure 7.6b). rate, the percentage of children whose height-for-age falls at least two standard deviations below the average The fertility decline has resulted in more favorable of a healthy reference population.198 The one exception conditions at birth. Birth outcomes depend on birth is the maternal mortality ratio, which, with an estimated spacing, i.e. the length of time between a birth and 510 deaths for every 100,000 live births, is close to the a subsequent pregnancy, as well as the age of the average for low-income countries and only somewhat mother at birth.200 The lower number of births per lower than the regional average.199 woman in Kenya has been associated with an increase in the average birth-to-pregnancy interval, from 30.2 Child health outcomes have improved substantially months in 2003 to 35.7months in 2014,201 and a small since 2000. Exceptionally large reductions in increase in the average age at birth from 26.3 years in under-five mortality have been observed in several 2003 to 26.5 years. countries in Sub-Saharan Africa. In Kenya, under-five mortality decreased by close to seven percent per Kenya’s counties differ markedly in terms of their year between 2003 and 2014. It is worth noting that position in the demographic transition. With a TFR of this is rate of reduction is far higher than the rate 6.1 births per woman and an under-five mortality rate implicit in the international MDG target, a two-thirds of 77.8 deaths per 1,000 births, Garissa is at a similar reduction and, hence, an annual rate of reduction of stage of the demographic transition as Burundi about 4.3 percent. As in other countries, reductions (Figure 7.7). With a TFR of 2.7 births per woman and in child mortality (mortality between the ages of an under-five mortality rate of only 31.6 deaths, one and four) were larger than reductions in infant Kiambu, on the other hand, is at a similar stage as mortality (ages zero to one; Figure 7.5a and Figure the Dominican Republic. It is also worth noting that 7.5b). The share of neonatal deaths, deaths that occur a few counties, including West Pokot, Wajir, Turkana, during the first month of life, in under-five deaths and Samburu, seem to have high fertility rates given has increased, suggesting that further declines in observed rates in under-five mortality. As falling under-five mortality will likely require a different mortality in children typically precedes falling fertility mix of public health interventions. Stunting among rates, a further drop in fertility in these counties may children below the age of five and the incidence of be imminent. underweight also decreased, although reductions were more in line with the (implicit) MDG target (Figure 7.5c and Figure 7.5d). 200 For Kenya, one study has found that “the length of the preceding birth 198 The indicator reflects a process of failure to reach linear growth potential interval is a major determinant of infant and early childhood mortality: in resulting from suboptimal health and/or nutritional conditions. On urban informal settlements in Nairobi between 2003 and 2009 (Fotso, et a population basis, high levels of stunting are associated with poor al. 2013). socioeconomic conditions and increased risk of frequent and early 201 The share of pregnancies that started within 24, 18, and six months of the exposure to adverse conditions such as illness and/or inappropriate most recent birth declined by 7.4, 5.5, and 1.1 percentage points. feeding practices. 199 See also chapter 3 for a more detailed discussion of maternal mortality. 154 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Health and Poverty Figure 7.4: Health outcomes in Kenya vis-à-vis benchmark countries and aggregates, latest year available a) Under-five mortality rate (per 1,000 live births) b) Infant mortality rate (per 1,000 live births) Middle income, 2016 Middle income, 2016 Rwanda, 2016 Rwanda, 2016 South Africa, 2016 South Africa, 2016 Kenya, 2016 Kenya, 2016 Uganda, 2016 Uganda, 2016 Tanzania, 2016 Tanzania, 2016 Ghana, 2016 Ghana, 2016 Low income, 2016 Low income, 2016 Sub-Saharan Africa, 2016 Sub-Saharan Africa, 2016 0 20 40 60 80 100 0 10 20 30 40 50 60 c) Adult mortality rate, women (per 1,000 women between d) Adult mortality rate, men (per 1,000 men between the ages of 15 and 60) the ages of 15 and 60) Middle income, 2015 Middle income, 2015 Rwanda, 2015 Rwanda, 2015 Kenya, 2015 Kenya, 2015 Tanzania, 2015 Ghana, 2015 Ghana, 2015 Tanzania, 2015 Low income, 2015 Low income, 2015 Sub-Saharan Africa, 2015 Sub-Saharan Africa, 2015 Uganda, 2015 Uganda, 2015 South Africa, 2015 South Africa, 2015 0 100 200 300 400 500 0 100 200 300 400 500 e) Maternal mortality ratio (modeled estimates, f ) Low height-for-age (stunting), percent of children per 100,000 live births) under five South Africa, 2015 Ghana, 2014 Middle income, 2015 Middle income, 2016 Rwanda, 2015 South Africa, 2008 Ghana, 2015 Kenya, 2014 Uganda, 2015 Uganda, 2011 Tanzania, 2015 Sub-Saharan Africa, 2016 Low income, 2015 Tanzania, 2011 Kenya, 2015 Low income, 2016 Sub-Saharan Africa, 2015 Rwanda, 2010 0 100 200 300 400 500 500 0 10 20 30 40 50 Source: Own calculations based on WDI data. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 155 Health and Poverty Figure 7.5: Annual rate of reduction in selected indicators of childhood health, percent, c. 2000 to 2015 a) Child mortality (ages 1-4) b) Infant mortality (ages 0-1) Uganda Uganda Tanzania Tanzania Rwanda Rwanda Kenya Kenya Ghana Ghana Ethiopia Ethiopia 0 2 4 6 8 10 12 0 2 4 6 8 10 12 Percent Percent c) Stunting d) Underweight Uganda Uganda Tanzania Tanzania Rwanda Rwanda Kenya Kenya Ghana Ghana Ethiopia Ethiopia 0 2 4 6 8 0 2 4 6 8 Percent Percent Source: Own calculations based on DHS data (Ethiopia, 2000, 2016; Ghana, 2003, 2014; Kenya, 2003, 2014; Rwanda, 2000, 2014/15; Tanzania, 2004/05, 2015/16; Uganda, 2000/01, 2011). Note: 95-percent confidence intervals are indicated. Moderate and extreme stunting is defined as a height-for-age z-score below two standard deviations against the WHO reference scale. Dotted lines indicate implicit MDG targets.202 Figure 7.6: TFR (number of births per woman) and under-five mortality rate (deaths per 1,000 live births) a) TFR, 2015 b) Under-five mortality rate, 2016 8 160 7 Kenya 140 Kenya 2000 6 2000 120 Kenya 2005 5 100 4 80 Kenya Kenya 3 60 2016 2015 2 40 Kenya 1 20 2010 0 0 500 5,000 50,000 500 5,000 50,000 Source: Own calculations based on KDHS 2014 and WDI data. 202 MDGs 1C and 4 called for reductions in the proportion of children below the age of five that are underweight and in under-five mortality one half and by two thirds, respectively, over the course of 25 years. These targets translate into reductions by 3.4 and 4.3 percent on an annual basis.” 156 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Health and Poverty Figure 7.7: TFRs against under-five mortality, countries (2015) and Kenyan counties (2014) 9 West Pokot 8 Wajir Turkana 7 Samburu 6 Garissa Mandera 5 4 Migori 3 2 Nairobi All Kenya Mombasa 1 0 0 20 40 60 80 100 120 140 160 Kenyan counties Other countries Linear (Other countries) Source: Own calculations based on KDHS 2014 and WDI data. A large share of registered deaths in Kenya still (Figure 7.8a).204 Nevertheless, the death rates for stem from communicable diseases – malaria and various communicable diseases have often been pneumonia are the leading causes of death – but flat or decreasing, while the number of registered deaths from non-communicable diseases are on deaths from non-communicable diseases has been the rise. Malaria and pneumonia were the leading on the rise (Figure 7.8b): deaths caused by cancer and causes among registered deaths in Kenya in 2015,203 heart diseases are up 36 and 32 percent, respectively, followed by cancer, HIV/AIDS, and tuberculosis while deaths from tuberculosis and malaria are down (Kenya National Bureau of Statistics, 2016). Malaria, by nine and 22 percent. The pattern suggests that pneumonia, HIV/AIDS, and tuberculosis combined the epidemiological transition in Kenya is rapidly – all of them communicable diseases – account progressing, a process associated with newly emerging for one third of non-accident, registered deaths challenges in the realm of public health. Figure 7.8: Levels in trends in registered deaths by cause, 2011–2015 a) Registered deaths by cause, 2015 b) Change between 2011-2015 (percent) Cancer Heart disease Malaria 10.3% Road tra c accidents Cancer 45.92% Tuberculosis Other 11.2% Road tra c accidents Anemia Heart disease 7.8% Menengitis Other Pneumonia Pneumonia HIV/AIDS HIV/AIDS 5.6% Anemia Tuberculosis 5.1% Other accidents Menengitis Other accidents 4.2% 2.2% 2.9% 1.9% 2.7% Malaria -40. -20 0 20 40 Source: Own calculations based on KES 2016. 204 To the extent that deaths in remote areas are less likely to be registered, the share of deaths due to non-communicable diseases in administrative 203 2015 estimates are provisional. data is likely to be an underestimate. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 157 Health and Poverty 7.2 HEALTH OUTCOMES AND UPTAKE significant in all subpopulations, supporting the notion THROUGH AN EQUITY LENS that health outcomes have significantly improved since T he present section investigates the relationship 2005/06. between health outcomes, service uptake, health expenditures, and poverty. As noted above, Gaps in under-five mortality have declined. Under- health outcomes across a given population are five mortality has declined among all sectors of the usually strongly correlated with poverty. Causality, population (Figure 7.10). However, declines were more while usually difficult to establish in a specific context, pronounced among the bottom 20 percent (as based typically runs in both directions: low incomes are both on a wealth index constructed from household assets). a result of and a cause of poor health. This section In fact, the advantage in survival of children from the exploits the 2005/06 KIHBS and the recent 2015/16 top 20 percent vis-à-vis children from the bottom 20 KIHBS, as well as DHS data, to investigate links between percent, which was still close to 50 percent in 2003, poverty and levels and trends in health outcomes, declined to about ten percent. Surprisingly, the under- uptake of services, and expenditures. five mortality rate in rural areas is now lower than in urban areas (although the difference is not statistically 7.2.1 Health outcomes significant at conventional levels).205 Kenyans feel sick less often than a decade ago, especially the poor. While self-reported incidence of There is still a considerable socioeconomic gradient sickness or injury are likely unreliable, their evolution in stunting rates. Overall, stunting rates among is still informative about broad patterns of change children below the age of five have been falling in disease burdens. The KIHBS data suggest that the between 2008/09 and 2014: while more than one in number of individuals that reported having been three children were either moderately or extremely sick or having been injured in the last four weeks has stunted in 2008/09, only one in every four fell into these decreased from 27.4 to 21.5 percent between 2005/06 categories in 2014 (Figure 7.11). Gains were driven as and 2015/16 (Figure 7.9). This pattern is statistically much by improvements among poor children as by Figure 7.9: Self-reported instances of sickness or injury during last four weeks prior to the survey as percent of population 30 20 Percent 10 0 Total Bottom 40% Top 60% Rural Urban Bottom 40%, rural Top 60%, urban By quintile By location By quintile and location 2005/06 2015/16 Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16. Note: 95-percent confidence intervals indicated. 205 One reason for the narrowing of the gap may be high levels of under- five mortality in urban informal settlements. While under-five mortality rates have also decreased in urban informal settlements, they remain high both compared to rural and urban, non-informal settlement areas (Kimani-Murange, et al. 2014). 158 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Health and Poverty Figure 7.10: Under-five mortality (deaths per 1,000 live births) by quintile, mother’s educational attainment, and location 180 150 120 90 60 30 0 Total Bottom 20% Bottom 40% Top 20% Primary and Secondary Rural Urban lower and higher By quintile of wealth index By mother's highest level By location of educational attainment 2003 2008/09 2014 Source: Own calculations based on 2003, 2008/09, and 2014 KDHS. Note: 95-percent confidence intervals indicated. Figure 7.11: Stunting rate by quintile, mother’s educational attainment, and location, 2003–2014 50 40 30 20 10 0 Total Bottom 20% Bottom 40% Top 20% Primary and Secondary Rural Urban lower and higher By quintile of wealth index By mother's highest level By location of educational attainment 2003 2008/09 2014 Source: Own calculations based on 2003, 2008/09, and 2014 KDHS. Note: 95-percent confidence intervals indicated. improvements among better-off children.206 Overall, five mortality rates are below ten deaths per 1,000 the marked difference in socioeconomic gradients live births in the counties of Laikipia in the center and betweeen under-five mortality and stunting rates Kajiado in the South. But they are in excess of 100 suggests a closing of the gap in mortality but not deaths per 1,000 in the Southwestern county of Migori in morbidity. (Figure 7.12). The stunting rate reveals a very different pattern: it is particularly high in the Northwestern Child health outcomes vary substantially across county of West Pokot (46.3 percent) and in Kitui counties and by domain, suggesting that counties (45.0) in the center. Stunting rates tend to be lower they face different public health challenges. Under- in the Lake Victoria region, an observation that has been noted elsewhere (Priebe and Gräb 2009). These 206 It is worth noting at this point that a recent government program, the Kenya’s National School-based Deworming Programme (NSBDP), may disparities in outcomes suggest that counties face well have addressed the problem of child malnutrition. Stunting reflects the accumulated effects of malnutrition and the program will not be in very different public health challenges. full effect until 2017. (Ministry of Education, Science and Technology and Ministry of Health 2015). Deworming of children, in turn, has been shown to result in weight gain (Croke, et al. 2016) and other short- and long- term benefits (Miguel and Kremer 2004, Baird, et al. 2016, Bleakley 2007). KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 159 Health and Poverty Figure 7.12: Child health outcomes by county, 2014 a) Under-five mortality (deaths per 1,000 live births) b) Incidence of low height-for-age (% of children below five) (100,120) (80,100) (70,80) (45,50) (60,70) (40,45) (50,60) (35,40) (40,50) (30,35) (30,40) (25,30) (20,30) (20,25) (10,20) (15,20) Source: Own calculations based on 2014 KDHS. Figure 7.13: Obesity rates (BMI > 30, share of women aged 15-49) by quintile, educational attainment, and locality, 2003–2014 50 40 30 20 10 0 Total Bottom 20% Bottom 40% Top 20% Primary and Secondary Rural Urban lower and higher By quintile By mother's highest level By location of educational attainment 2003 2008/09 2014 Source: Own calculations based on 2003, 2008/09, and 2014 KDHS. Note: 95-percent confidence intervals indicated. Obesity rates have increased dramatically among 7.2.2 Access and uptake better-off Kenyans. In line with a shift of the disease Most indicators of uptake of health services in burden towards non-communicable diseases, obesity Kenya show robust improvements since the early rates have increased steadily between 2003 and 2014 2000s. The proportion of children sleeping under (Figure 7.13). Among women between 15 and 49, for an ITN has increased from only six percent in 2003 to which detailed data are available from the DHS, the rate 54.3 percent in 2014 (Figure 7.14a). This is a significant has increased from six percent to ten percent. But while achievement with important implications for health the increase is notable across the entire population, it outcomes. A study finds that increased ownership of is particularly pronounced among better-off Kenyans: ITNs in endemic malaria zones explains 79 percent one in five women among the top 20 percent of the of the decline in infant mortality between 2003 and population (based on the DHS wealth index) were 2008 (Demobynes and Trommlerová 2016). There were obese in 2014, up from only 13 percent in 2003. marked increase in both the proportion of children with symptoms of ARI who were taken to a health provider and the proportion of births attended by 160 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Health and Poverty Figure 7.14: Selected indicators of health services uptake (%), 2000–2015 a) Share of children under five sleeping under an ITN b) Share of children below five with symptoms of ARI taken to health provider 100 100 80 80 60 60 40 40 20 20 0 0 2000 2005 2010 2015 2000 2005 2010 2015 Kenya Ethiopia Rwanda Tanzania Uganda Kenya Ethiopia Rwanda Tanzania Uganda c) Percentage of births attended by skilled health staff d) Percentage of children between 12- 23 months immunized against measles 100 100 80 80 60 60 40 40 20 20 0 0 2000 2005 2010 2015 2000 2005 2010 2015 Kenya Ethiopia Rwanda Tanzania Uganda Kenya Ethiopia Rwanda Tanzania Uganda Source: Own calculations based on WDI. skilled health staff, particularly between 2008/09 and A major challenge in Kenya is to ensure equity in 2014 (Figure 7.14b and Figure 7.14c). This suggests service delivery across space, which also includes a greater uptake of services related to child and delivery of basic health services to population groups maternal health. However, there is clearly further room that reside in sparsely populated areas. Geographic for improvement. For instance, close to two out of five access is a necessary condition for uptake and thus births in Kenya in 2014 were not attended by qualified for effective health service provision. The physical health professionals. Immunization rates have been high accessibility of health services is lower among the poor generally, although they have apparently registered and the rural population. In 2015/16, 60 percent of the a drop by 20 percentage points around the year 2012 population lived in a community with a health facility. (Figure 7.14d).207 Immunization against diphtheria, However, while this was the case for four out of five pertussis, and tetanus (DPT) has also decreased, from Kenyans in urban areas, only one in two in rural dwellers 94 to 87 percent, between 2012 and 2013. lived in such a community (Figure 7.15a). There are also 207 While it is unclear what caused this drop, one issue that resurfaces pronounced differences in the distance to the nearest sporadically is misinformation about the intended outcomes of immunization campaigns and the safety of vaccines (Larson 2015). Another health facility by locality and expenditure quintile. The candidate explanation is the influx of Somali refugees in 2011, among which the rate of immunizations was reportedly very low (Polonsky, et typical (median) Kenyan travels nine kilometers to the al. 2013). An outbreak of measles in the Dadaab refugee camp in Garissa was reported between June and November 2011 (Navarro-Colorado, et al. next health facility at which a doctor can be consulted. 2014). Finally, there were some reports pointing to stock-outs of vaccines and other medical supplies in the wake of devolution. The distance is shorter for urban dwellers, at only one KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 161 Health and Poverty Figure 7.15: Availability of health facilities and distance to nearest health facility in which a doctor would be on duty, 2015/16 a) Population share in community with health facility b) Distance to health facility where a doctor is available (km) 100 40 80 30 60 Percent Percent 20 40 10 20 0 0 Total B40% T60% Rural Urban B40%, rural T60%, rural B40%, urban T60%, urban Total B40% T60% Rural Urban B40%, rural T60%, rural B40%, urban T60%, urban By quintile By locality By quintile and locality By quintile By locality By quintile and locality Mean Median Source: Own calculations based on KIHBS 2015/16 Note: 95-percent confidence intervals indicated. Items extracted from the community questionnaire. kilometer, and greater among rural dwellers, at 15 rate as Americans (Das, Hammer, and Leonard 2008). kilometers (Figure 7.15b). More pronounced differences The decrease in the number of visits is evident across between urban and rural areas than among the bottom different subgroups and is in line with improvements in 40 percent and the top 60 percent suggest that the the overall health outlook. main challenge is remoteness, not low incomes. An increase in the uptake of preventive health services has been driven by an increase among Uptake of curative care is high and has increased the poor and the rural population. The propensity between 2005/06 and 2015/16, particularly among to seek preventive health services four weeks prior the poor. 83.9 percent of Kenyans will consult with a health service provider when they become sick or to the survey interview has increased from 2.5 to suffer an injury, up from 70.3 percent in 2005/06 (Figure 4.2 percent (Figure 7.17). The increase is driven by 7.16a).208 These proportions are exceptionally high a higher propensity among the poor and the rural and suggest that an increasing share of Kenyans have population. There is no statistically significant change both physical access to some type of provider and the in the uptake of preventive health services among means to obtain some treatment. Differences between urban dwellers. With the exception of preventive care the poor and the rich and between urban and rural among children, differences between the poor and areas are small and declining. 78 percent among the the non-poor are evident at each age and for each poorest 20 percent will consult with a health provider type of medical service. It is worth mentioning that in case they have a medical concern, compared to 87 the poor and the rural population rely more heavily percent among the richest 20 percent. on public health services. Government health services are sought in about three in every five (61.7 percent) The average number of curative visits per year has outpatient visits in Kenya, with roughly similar shares declined. The average number of curative visits has for government hospitals (21.8), health centers (18.0), declined from 4.9 visits per person and year to 3.5 in and dispensaries (21.9) 2015/16 (Figure 7.16b).209 Kenyans frequent medical providers less often than Indians and at about the same Poor households have a lower probability of owning a bed net, and are less likely to have been vaccinated 208 According to the Kenya Household Health Expenditure and Utilisation Surveys (KHHEUS), the rate at which Kenyans sought treatment in the against measles. Overall, almost three in four Kenyans case of sickness increased from 77.2 percent in 2003 to 83.3 percent in (73.1 percent) live in a household that owns a bed net 2007 and 87.3 percent in 2013 (Ministry of Health 2014). 209 Data from the KHHEUS are not consistent with this. According to that (Figure 7.18a). However, the share drops to only around source, the number of visits per person and year increased from 1.9 visits in 2003 to 2.6 in 2007 and 3.1 in 2013 (Ministry of Health 2014). two in three among the bottom 20 percent of the 162 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Health and Poverty Figure 7.16: Uptake of curative health services and number of curative visits by quintile and locality, 2005/06 and 2015/16 a) Uptake of curative health services (proportion of sick/injured b) Average number of curative visits per person per year during last four weeks seeking treatment) (total population) 100 6 80 4 60 Percent Percent 40 2 20 0 0 Total B40% T60% Rural Urban B40%, T60%, Total B40% T60% Rural Urban B40%, T60%, rural urban rural urban By quintile By locality By quintile By quintile By locality By quintile and locality and locality 2005/06 2015/16 2005/06 2015/16 Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16. Note: 95-percent confidence intervals indicated. Figure 7.17: Uptake of preventive health services during Vaccination rates210 fall sharply, from more than 90 four weeks prior to interview percent in the Central region to about 44 percent in 100 Mandera in the Northeast and only 36 percent in West 80 Pokot (Figure 7.19b). Percent 60 While the majority of pregnant women receive 40 antenatal care at least once in almost all counties, 20 a far smaller share delivers with the assistance of a 0 skilled provider. The share of women that receive antenatal care prior to birth is higher than two thirds Urban Total T60% Rural B40% T60%, rural B40%, rural in all counties with the exception of Mandera (50.5 By quintile By quintile percent) and Wajir (57.6) (Figure 7.19c). However, the By locality and locality percentage of births attended by skilled health workers 2005/06 2015/16 varies considerably across counties, with higher uptake Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16. Note: 95-percent confidence intervals indicated. in more densely populated counties in the center and lower uptake in less densely populated counties in the population. Similarly, children from poor families are north and in the east (Figure 7.19d). Bringing high- less likely to have been vaccinated against measles. quality assistance during deliveries to mothers should Close to four in five children (78.7 percent) have remain a main priority over the next years (Box 7.2). received a measles vaccination at least once by the time they are between twelve and 23 months (Figure 7.18b). However, that share falls to only 64.3 percent Preventive health goods and services that address among children in the bottom quintile. communicable diseases have been shown to be highly cost-effective. Kenya adopted several measures Geographic disparities in access to health care and to fight infectious diseases over the last years, including uptake are pronounced. Geospatial variation in access the distribution of ITNs (WHO 2018) and deworming of and uptake is often pronounced but depends on the school children in endemic zones. These programs are domain. More than three in five children with fever are usually delivered free of charge and have been shown taken to a health provider on average. But the estimates 210 The indicator includes DPT (three shots), BCG, polio (four shots), and range from little more than two in five in Garissa to measles. A child is registered as having received the vaccination if the vaccination is marked on the child’s health card and/or the mother more than four in five in Muranga (Figure 7.19a). recalls the vaccination. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 163 Health and Poverty Box 7.2: What works to boost skilled birth assistance for safer childbirth? Increasing the share of deliveries in health facilities will not automatically improve birth outcomes. Current global health policies often emphasize institutional deliveries as a pathway to achieving reductions in newborn mortality in developing countries, a priority also in Kenya. For instance, recent interventions to improve maternal and child health outcomes in countries like India and Rwanda were based on the assumption that increasing the share of institutional deliveries will have positive health effects. However, evaluations of these programs show that while they clearly increased uptake, they often had no effect on neonatal mortality (Okeke and Chari 2016). In Kenya, there are no statistically significant correlations between skilled assistance and institutional delivery and neonatal and infant mortality conditional on maternal and child background characteristics (Appendix G.2).211 Demand-side interventions that aim to incentivize uptake of institutional deliveries should be met with skepticism, both because of recent evidence for their lack of effectiveness and out of respect for the choices patients make. The quality of obstetric care may differ significantly across localities and facilities and a large part of this variation may be perfectly observable to prospective users.212 It is also worth noting in this context that the removal of user fees in mid-2013 had no discernable effect on uptake in rural areas. This suggests that the options available to the poor and rural communities are simply not seen as bringing sufficient advantages. Supply-side interventions hold some potential for greater uptake with improved birth outcomes. While the share of deliveries in tier-two facilities in Kenya has increased over recent years, two in three deliveries in the public sector still take place in hospitals (Figure 7.21). To boost skilled attendance, it is paramount to increase the share of deliveries in these lower-level facilities. But doing so requires an improvement in the quality of services these facilities currently offer. For instance, a 2012 study found that many of them lacked electricity, clean water, and basic medicines (Section 7.3.1), making them ill-equipped for deliveries. Devolution may have improved this state of affairs but this remains to be seen. Figure 7.18: Uptake of preventive health goods, select indicators, by poverty and quintile, 2015/16 a) Proportion of individuals that live in a household that owns at b) Children betweeen 12 and 23 months that received at least one least one bednet measles vaccination 100 100 80 80 60 60 Percent Percent 40 40 20 20 0 0 Poor Poor Total Total Non-Poor Non-Poor Bottom 20% Bottom 20% Bottom 20% Bottom 20% T60%, rural T60%, rural By poverty By quintile By poverty By quintile Source: Own calculations based on KIHBS 2015/16. Note: 95-percent confidence intervals indicated. 211 These estimates are merely correlations and thus only suggestive; the empirical identification of causal effects of institutional delivery requires a more elaborate strategy. Omitted variables are a concern. For instance, maternal characteristics that matter for both uptake and pregnancy outcomes may not be fully captured in the analysis. Another concern is that women who experience complications during pregnancy may be more likely to deliver in a health facility (Okeke and Chari 2016). Finally, causal effects may be very different across facilities that differ in the quality of maternal care they can provide. In Kenya, health worker strikes, which resulted in a lower proportion of institutional births, were associated with higher rates of neonatal and infant mortality (Friedman and Keats 2017), suggesting that those who opt for institutional deliveries do so because birth outcomes are usually better. 212 To wit, one 2009 study of 25 health facilities in informal settlement areas in Nairobi concludes that only two met the criteria for comprehensive emergency obstetric care and that “[t]he quality of emergency care services in Nairobi informal settlements is poor and needs improvement” (Ziraba, et al. 2009). 164 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Health and Poverty Figure 7.19: Access to health services and uptake by county, 2014, select indicators a) Percentage of children under age five with fever for whom advice or b) Percentage of children age 12-23 months that received all treatment was sought basic vaccinations (90,100) (90,100) (80,90) (80,90) (70,80) (70,80) (60,70) (60,70) (50,60) (50,60) (40,50) (40,50) (30,40) (30,40) (20,30) (20,30) No data c) Percentage of pregnant women receiving antenatal care d) Percentage of live births delivered in health facility from skilled provider (90,100) (90,100) (80,90) (80,90) (70,80) (70,80) (60,70) (60,70) (50,60) (50,60) (40,50) (40,50) (30,40) (30,40) (20,30) (20,30) Source: Kenya Demographic and Health Survey (KDHS) report (KNBS and others 2015). Note: Panel (a): Excludes pharmacies, shops, markets, and traditional practitioners. Estimates based on less than 25 cases not reported. Unweighted estimates reported if based on 25-49 cases. Panel (c): If more than one source of ANC was mentioned, only the provider with the highest qualifications was considered in the calculation. to be highly cost-effective. Not only do they effectively The share of deliveries that occur in hospitals and prevent malaria and worm infections in beneficiaries, the share attended by a doctor, a nurse, or a midwife they also protect those around them from infections. have increased, but a significant socioeconomic Access to these health goods should remain free of gradient persists. Among surviving children below charge. Evidence from several RCTs suggests that the age of five, the share of those that were born in a uptake is highly price-elastic (Abdul Latif Jameel hospital increased from a little more than one in four Poverty Action Lab 2011). Even a highly subsidized to almost one in two (Figure 7.20a). The increase was price can have adverse effects on demand vis-à-vis free particularly pronounced in urban areas, and there provision. In addition, there is no evidence that free are gaps between the bottom 40 percent and the distribution results in low rates of utilization (e.g., for top 60 percent. The 2008 KDHS reckons that the ITNs). The general finding extends to other preventive most common reason for not delivering in a facility health goods, such as slippers (which prevent worm was distance or lack of transport (42 percent) (KNBS infections in endemic zones) (Meredith et al. 2013) et al. 2010). High costs were a deterrent in only 16.9 and water purification (Ashraf, Berry, and Shapiro 2010; percent of all deliveries, with limited variation across Kremer et al. 2011), and have been found in a range of wealth quintiles. low- and middle-income settings. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 165 Health and Poverty Figure 7.20: Share of births (of surviving children 60 months and younger) by circumstance, 2005/06 and 2015/16 a) Delivered in a hospital b) Attended by a doctor, a nurse, or a midwife 100 100 80 80 60 60 Percent Percent 40 40 20 20 0 Urban 0 Total T60% Rural B40% T60%, urban B40%, urban T60%, rural B40%, rural Urban Total T60% Rural B40% T60%, urban B40%, urban T60%, rural B40%, rural By quintile By locality By quintile and locality By quintile By locality By quintile and locality 2005/06 2015/16 2005/06 2015/16 Source: Own calculations based on KIHBS 2005/06 and 2015/16. Note: 95-percent confidence intervals indicated. Recent years have seen a shift from demand for However, there is little evidence for an effect of the private provision of health services to the public reform on the overall propensity to deliver in a formal sector, particularly for deliveries. One would health facility (Appendix G.1). clearly expect greater uptake of public services vis- à-vis private services in the wake of the June-2013 7.2.3 Health insurance uptake and out-of-pocket decision to waive user fees for basic health services expenditure and, perhaps, an increase in overall uptake. Using While coverage is still low, membership in Kenya’s administrative data (as in Figure 7.1), the ratio of public NHIF has increased since 2006/07. 6.1 million Kenyans care visits to private care outpatient visits increased were registered members of the NHIF in 2015/16, up from an average of 2.7 between June 2012 and May from 1.8 million in 2006/07 (Figure 7.3), representing 2013 to 3.2 in the subsequent twelve months. The an increase from four to 13 percent of the population. ratio for deliveries, which were included by the policy, Among members, the share of informal workers increased from 2.0 to 3.1 over the same time period.211 increased from eleven to 41 percent in 2015/16. Figure 7.21: Share of deliveries by provider, 2009–2014 Coverage through NHIF accounts for the large 70 majority of health insurance arrangements in Kenya. 60 Estimates based on the 2015/16 KIHBS data suggest 50 that 8.1 million Kenyans were covered through the NHIF in 2015/16 while only 0.5 million had different 40 Percent health coverage.212 However, these estimates are 30 substantially lower than the number of beneficiaries 20 reported elsewhere, suggesting that underreporting 10 may be an issue.213 0 2009 2010 2011 2012 2013 2014 Health insurance coverage is still low among the Any provider Private home or en route to provider poor and among the rural population. According Government hospital Health center or dispensary Private to the 2015/16 KIHBS, the coverage rate varies Source: Own calculations based on 2014 KDHS. substantially by poverty and locality. While only 7.5 211 Note that the responsibility for health service delivery was devolved at around the same time, August 2013, and that it is thus may be difficult 212 Estimates based on the 2013 KHHEUS suggest that approximately one in to disentangle the effects these respective reforms have had on relative five Kenyans had health insurance, with around 88.4 percent of insurance demand. However, given reports about the problems that some counties arrangements accounted for by the NHIF (Ministry of Health 2014). faced initially as well as the more pronounced shift observed for services 213 According to one June 2017 report, for instance, the Chief Executive of included in the June 2013 policy, it seems more likely that the shift in the NHIF stated that the NHIF had 6.5 million contributing members and demand is driven by abolition of user fees. 24 million beneficiaries (Murumba 2017), three times the KIHBS estimate. 166 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Health and Poverty percent of the poor were covered by some form of 6.7 percent of total in-patient expenditure on average health insurance, the corresponding share among the and only 7.3 percent resorted to this strategy (Figure non-poor was 25.4 percent (Figure 7.22a). And while 7.23). Medical cover, either through one’s employer or almost every third Kenyan in urban areas is covered, the private arrangements, are moderately important only share is only around 14 percent for rural residents. among the richest 40 percent. The share of health expenditure in total household Figure 7.23: Average shares of in-patient health expenditure by funding source (democratic shares per expenditures is moderately low across the hospitalized individual) population. Poor households spend around seven 100 percent of their expenditure on in-patient or out- patient care but only around one third of a percent on 80 health insurance premiums (Figure 7.22b). In contrast, 60 households in the fifth quintile of the expenditure Percent distribution spend only around six percent of their 40 total income on health and around one percent on health insurance. 20 Only six percent of Kenyans that are hospitalized 0 Total Poorest 20% 2 3 4 Richest 20% resort to asset sales, a coping strategy that is Other transfers Fund raising/family contribution Employer or own medical cover important in other developing countries. In Kenya, Sale of assets Loan Own funds only about six percent of households resorted to asset Source: Own calculations based on KIHBS 2015/16. Note: 95-percent confidence intervals indicated in panel (a). sales and four percent resorted to borrowing (including loans without interest) to cope with a health shock that requires hospitalization. This is much lower than Estimates of the share of the population experiencing shares reported for other low- and middle-income high or “catastrophic” health expenditures vary countries (Kruk, Goldmann, and Galea 2009). While substantially across sources. According to data asset sales are more common among the poor and from the WHO’s Global Health Observatory,214 which almost exclusively a rural phenomenon, even among are based on the 2005/06 KIHBS, less than six and the poorest 20 percent, asset sales account for only two percent of the population experienced health Figure 7.22: Health insurance coverage, health expenditure and incidence of asset sales in response to hospitalization a) Health insurance coverage b) Average health expenditure shares by type of service 100 10 80 8 60 6 Percent Percent 40 4 20 2 0 0 Urban Total T60% Rural B40% Top 20% W/out health Poor With health Bottom 20% Bottom 40% Urban Total Rural Non-poor Top 20% Bottom 20% Bottom 40% By poverty By quintile By locality By poverty By quintile By locality Health insurance Out-patient care In-patient care Source: Own calculations based on KIHBS 2015/16. Note: 95-percent confidence intervals indicated in panel (a). 214 See http://www.who.int/gho/en/. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 167 Health and Poverty expenditures in excess of ten and 25 percent of their While these estimates should be interpreted total household expenditure in 2005/06. This would still carefully, they suggest that it may prove difficult place Kenya into the 42nd and 61st percentile of the cross- to expand coverage to Kenya’s poor and informal country distribution, respectively. However, estimates workers though voluntary health insurance. Out- based on the KHHEUS suggest that 15.5 percent of the of-pocket health expenditures and adverse coping population experienced health expenditures in excess mechanisms are rarely observed in Kenya likely of ten percent in 2007, a substantially larger fraction. It because the poor are willing to forgo treatments they seems likely that this discrepancy is due to differences need. The analysis of uptake (section 7.2.2) suggests in survey design, including the items covered and, that this is part of the explanation. On the other hand, potentially, recall periods. it may also reflect the success of recent measures to lower out-of-pocket expenditures, particularly Despite notable differences in health insurance through the removal of user fees in public facilities. coverage across socioeconomic groups, the share of This, in turn, would suggest that the poor may have the population at risk of impoverishment will likely only limited incentives to voluntarily seek coverage have declined. Joint analysis of the 2007 and 2013 through the NHIF. KHHEUS suggests that the incidence of catastrophic health expenditures has declined. The share of 7.3 THE SUPPLY SIDE: PHYSICAL INPUTS, households that spent more than ten percent of total HEALTH PROFESSIONALS, AND INCENTIVES expenditure or more than 40 percent of non-food expenditure on health was 15.5 and 11.4 percent in 7.3.1 Physical inputs 2007 and 12.7 and 6.6 percent in 2013, respectively. Households rarely resort to adverse coping strategies to pay for inpatient care. Importantly, a decline in the A 2012-study found that Kenyan health facilities often lacked access to a regular supply of electricity and clean water. Only around two out of risk of impoverishment through health expenditures five health facilities had reliable access to electricity in is in line with significant improvements in population 2011, regardless of whether these were dispensaries, health in recent years and the removal of basic user health centers, or hospitals (Figure 7.24a). While the fees in public health facilities.215 share was higher for facilities run by faith-based Figure 7.24: Infrastructure availability in public and private facilities by type of facility and location (select indicators) a) Regular supply of electricity b) Has clean water 98 96 100 100 84 78 62 80 80 57 49 50 50 51 43 60 37 60 Percent Percent 38 41 35 34 40 40 20 20 0 0 MoH/LA Urban Total MoH/LA FBO/NGO Urban Rural Total FBO/NGO Rural Hospitals Hospitals Health centers Dispensaries Health centers Dispensaries By type of facility By provider By location By type of facility By provider By location Source: Own calculations based on 2012 SDI data. Note: 95-percent confidence intervals indicated. 215 Future work will analyze trends in catastrophic out-of-pocket health expenditures. While possible, this task is complicated by changes in the design of the KIHBS surveys. 168 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Health and Poverty organizations (FBOs) and NGOs vis-à-vis facilities incentives of health professionals. The effective run by the MoH or local authorities (LA), it was still delivery of health services relies crucially on a sufficient only around one half. Clean water was available supply of well-trained and motivated professionals, only in every other dispensary but in three out of including doctors, nurses, and midwives. To the extent four health centers and almost all hospitals (Figure that patients can observe the quality of care they are 7.24b). MoH-run facilities and facilities in rural areas receiving, they will also be more likely to take up services were less likely to have a regular supply of electricity if the quality is high. However, “boots on the ground” and clean water. are only a necessary condition for the health services to be effective. Increasingly, research documents that In 2012, almost half of all drugs on a basic list216 were motivation and incentives play a crucial role (Das and either not available or expired. Drug availability was Hammer 2014). This subsection investigates levels particularly low in dispensaries but somewhat better and trends in the number of health workers, their in hospitals (Figure 7.25). Differences between public remuneration, subject knowledge, and, to the extent and private non-profits and between rural and urban that data is available, their actual performance. Given areas were again discernable but not as pronounced as the greater reliance among the poor on public health for access to services. This is in line with another study services and recent shifts in demand away from private of drug safety in Nairobi, which finds no association provision and towards public provision, the subsection between drug quality and ownership, size, or location will pay special attention to differences in these of the facility from which the drug was obtained indicators across providers. (Wafula et al. 2017). Kenya continues to suffer from a shortage of health 7.3.2 Health professionals professionals. A widely-used threshold for the The quality of health services depends on the number of doctors, nurses, and midwives combined number, training, and practice of personnel in the is 22.8 per 10,000 population.217 In 2013, Kenya had health sector, as well as on the motivation and two physicians per 10,000 population and less than Figure 7.25: Drug availability by type of facility, provider, and location 80 67 56 54 60 54 51 49 50 50 Percent 40 20 0 Total Dispensaries Health Centers Hospitals MoH/LA FBO/NGO Rural Urban By type of facility By provider By location Source: Own calculations based on 2012 SDI data. Note: 95-percent confidence intervals indicated. 216 This indicator is defined as the number of drugs of which a facility has one as a proportion of all the drugs on the list. The drugs had to be unexpired and had to be observed by the enumerator. The drug list 217 The ratio of 2.28 well-trained health workers per 1,000 population has contains tracer medicines for children and mothers identified by the been put forward by the WHO (WHO 2006). It is the density estimated WHO following a global consultation on facility-based surveys. See to be necessary to achieve 80 percent coverage of deliveries by skilled Martin and Pimhidzai 2013. birth attendants. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 169 Health and Poverty Figure 7.26: Number of health professionals per 10,000 population a) Physicians b) Physicians, nurses and midwives China, 2012 19.4 South Africa, 2015 58.8 South Africa, 2015 7.7 China, 2012 37.9 India, 2014 7.3 India, 2011 24.4 Kenya, 2013 2.0 Uganda, 2010 14.2 Uganda, 2010 1.2 Kenya, 2013 10.7 Ghana, 2010 1.0 Ghana, 2010 10.2 Rwanda, 2010 0.6 Rwanda, 2010 7.3 Tanzania, 2012 0.3 Tanzania, 2012 4.6 Ethiopia, 2010 0.2 Ethiopia, 2010 2.6 0 5 10 15 20 25 0 20 40 60 Source: Own calculations based on WDI data. Note: The dotted vertical line in the right panel indicates a widely-used threshold of 22.8 per 1,000 doctors, nurses, and midwives (see footnote 219). nine nurses and midwives (Figure 7.26).218 These profession, and provide motivation. A major concern estimates suggest that Kenya suffers from the type of in Kenya and other Sub-Saharan African countries is severe shortage in health professionals typical of Sub- the emigration of health workers, often referred to Saharan Africa. They also suggest that this shortage as “brain drain” and directly linked to the shortage of is particularly pronounced for nurses and midwives. health professionals (Yonga, Muchiri, and Onyino 2012). They are in line with survey-based estimates,219 but One study finds that a ten-percent increase in pay in not with the number of registered health personnel Ghana, a country with sizable outward migration tabulated in the 2017 KES. of health workers, decreased attrition by about one percentage point (or 12.5 percent) (Antwi and The general shortage of health professionals is Phillips 2013). While remuneration is also argued to aggravated by an uneven distribution across be important for the motivation of the providers, counties. One study found that the number of nurses there is little empirical evidence for this assertion. in the public sector (but excluding nurses deployed in It seems plausible that how health workers are paid national referral hospitals) just prior to the devolution – whether through fixed salaries, fees-for-service, or of services varied across counties, from only 0.8 to capitation payments – matters as much or more for twelve per 10,000 population (Wakaba et al. 2014). The provider motivation and effort. same study also finds a positive correlation between nurse density and per capita health expenditure, as Real salaries of nurses and midwives in Kenya are well as a positive correlation between nurse density similar to those in Ghana and higher than in Nigeria and immunization rates. An important question that and Uganda.220 In Kenya, nurses and midwives earn remains to be answered in this context is to what extent a monthly salary of about US$976 (in 2011 PPPs and devolution resulted in a convergence in the density of including allowances) compared to US$936 in Ghana, health professionals across counties. US$690 in Nigeria, and only US$410 in Uganda (Figure 7.27b). Salaries relative to GDP per capita are an Remuneration of health workers is often argued to be indicator of the relative earning opportunities within important to increase retention, attract talent to the the economy for a given profession: Kenyan nurses and midwives earn on average 3.8 times GDP per capita in 218 Estimates are based on the WHO’s Global Health Workforce Statistics and disseminated through the World Bank’s WDI. 2015/16, the highest among the comparison group 219 Comparison between survey-based estimates for 2005/06 and 2015/16 (Figure 7.27a). also suggest that the relative number of health professionals of these occupations has increased over time: the number of auxiliary nurses, nurses and midwives, and medical and clinical officers increased from 220 The focus here is on nurses and midwives as the classifications used are 2.3, 3.9, and 2.3 per 10,000 to 2.3, 7.5, and 4.3, respectively. comparable despite different systems of occupational classification. 170 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Health and Poverty Figure 7.27: Salaries of nurses and midwives by country, 2005/06-2015/16 a) As ratio of GDP per capita b) In 2011 international dollars (PPPs) 7 1,400 $936 4.93 $976 6 1,200 $817 5 3.77 1,000 $690 4 2.88 2.89 800 3 1.86 600 $410 2 400 1 200 0 0 KEN0506 KEN1516 UGA1314 GHA0910 NGA1516 KEN0506 KEN1516 UGA1314 GHA0910 NGA1516 Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16, the 2009/10 Ghana Socioeconomic Panel Survey, the 2013/14 Uganda National Panel Survey, and the 2015/16 Nigeria General Household Survey. Note: 95-percent confidence intervals are indicated. Between 2005/06 and 2015/16, salaries of Kenya’s and nurses and midwives, on the contrary, have still health professionals have declined in relative terms. experienced increases in real salaries, albeit only by Salaries of auxiliary nurses declined from a ratio of 4.6 a modest 2.7 and 1.5 percent per year, respectively to 3.9 times GDP per capita while salaries of nurses (Figure 7.28b). and midwives declined from 5.1 to 3.8 times GDP per capita. Medical officers and clinical officers, who on The public-sector premium for health workers has average earned more than ten times GDP per capita increased. In Kenya, there is evidence of a public sector in 2005/06, earned an average of 5.2 times GDP per premium in 2015/16 but not in 2005/06 (Table 7.1). capita in 2015/16. This would translate into a decline This is in line with an increase in the demad for public in real salaries by about 2.6 percent per year. However, provision, but also with an increase in the bargaining due to the low number of observations, the difference power of public-sector unions in recent years.221 is not statistically different from zero. Auxiliary nurses Public-sector premia may be efficient if they constitute Figure 7.28: Salaries of select health workers in Kenya, 2005/06 and 2015/16 a) Ratio of salaries to GDP per capita, Kenya, b) Monthly salaries (int. dollars, 2011 PPPs), Kenya, 2005/06 and 2015/16 2005/06 and 2015/16 18 3,000 11.15 16 $1,755 2,500 14 12 2,000 10 $1,349 1,500 8 $1,003 $841 $976 5.09 5.17 6 4.60 $768 3.85 1,000 3.77 4 500 2 0 0 Auxiliary nurses Nurses and Medical and Auxiliary nurses Nurses and Medical and midwives clinical o cers midwives clinical o cers 2005/06 2015/16 2005/06 2015/16 Source: Own calculations based on KIHBS 2005/06 and 2015/16. Note: 95-percent confidence intervals are indicated. 221 However, there might be an equity consideration. For instance, civil- servants in Kenya obtain hardship allowance if they are posted in remote areas. These may be necessary to attract qualified professionals into these areas and to thus ensure service availability in these places. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 171 Health and Poverty Table 7.1: OLS regression of log salary (incl. allowances) on binary indicator of employment in public sector for auxiliary nurses; nurses and midwives; and medical and clinical officers, 2005/06 and 2015/16 2005/06 2015/16 (1) (2) (3) (1) (2) (3) All Rural Urban All Rural Urban 0.162 0.386 0.088 0.816*** 0.896*** 0.745*** Public sector (0.199) (0.444) (0.234) (0.159) (0.230) (0.175) Observations 82 18 64 128 43 85 R-squared 0.276 0.437 0.210 0.316 0.499 0.283 Source: Own calculations based on KIHBS 2005/06 and 2015/16 data. Note: Significance level: 1% (***), 5% (**), and 10% (*).Robust standard errors reported in parentheses. Outliers (error terms of +/- 5.5 standard deviation) were removed. All regressions include controls for age, age squared, gender, locality (rural or urban), and type of worker (auxiliary nurse, nurse or midwife, and medical or clinical officer). Results are obtained from unweighted OLS regressions. Re-running the regressions with sample weights did not alter the results qualitatively. compensating differentials (e.g. if public-sector health even doctors asked only three out of five relevant history professionals have to be compensated for working in and examination questions (Figure 7.29a). And while remote areas or for superior skills) or if they elicit greater 70 percent of all diagnoses were correct, only around effort and thus better outcomes for patients. However, half of the recommended treatments were correct and they may also reflect taxpayer-funded rents. As health complete. Moreover, there was significant variation workers typically earn above-average wages, these across domains. Malaria, for instance, is diagnosed in only would be regressive. They would also drive up the around 30 percent of the cases. costs of an urgently-needed expansion of the public- sector workforce. More research is required to settle the Case management of health providers in Nairobi question of whether these premia are efficient. compares favorably with middle income countries such as India and China. The use of standardized or While Kenya’s health workers are often more likely “mystery” patients, which are individuals recruited from to adhere to clinical guidelines when presented with the local community and extensively trained to present vignettes than health workers in other Sub-Saharan the same clinical condition to multiple healthcare African countries, even medical doctors typically providers in a study sample, is gaining acceptance as a ask only three out of five relevant history and gold standard for the measurement of clinical practice. examination questions. The SDI surveys conducted Only one such study has been conducted in Kenya and in several countries in Sub-Saharan Africa over the last was limited to a sample of providers in Nairobi (Daniels years include assessments of provider competence et al. 2017), but it is possible to tentatively compare and knowledge.222 These are administered through results with similar studies conducted in urban China medical vignettes across common tracer conditions, and India. The results suggest that the quality of care acute diarrhea in children, pneumonia, diabetes patients receive in this setting depends on the condition mellitus, tuberculosis, and malaria. Compared to their they present: Indian and Chinese providers managed counterparts in other countries in the region, Kenyan angina better but did not provide ORS for children with doctors and nurses are as likely or somewhat more likley diarrhea (Table 7.2). In addition, Kenyan providers were to adhere to clinical guidelines in their hypothetical significantly more expensive. Mystery patients also treatment of the cases presented to them. However, spent more time waiting to see the provider but also more time with the provider. 222 The evaluation of the medical providers is done using seven standardized cases. These cases are based upon common pathologies and are adjusted to the local context using national treatment guidelines. See http://siteresources.worldbank.org/AFRICAEXT/Resources/SDI_ In Kenya’s primary health care system, compliance instruments_Kenya.pdf. The indicator “Adherence to clinical guidelines” is defined as the unweighted average of the share of relevant history taking with infection prevention and control (IPC) questions and the share of relevant examinations performed for each of the following five case study patients: (i) malaria with anemia, (ii) diarrhea practices, which have been shown to effectively with severe dehydration, (iii) pneumonia, (iv) pulmonary tuberculosis, protect patients from infections, varies widely and (v) diabetes. 172 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Health and Poverty Figure 7.29: Adherence to clinical guidelines and absence from health facility by country a) Adherence to clinical guidelines b) Absence from health facility 70 60 Uganda, 2013 46 50 Togo, 2012 40 40 Percent 30 Nigeria, 2013 29 20 10 Kenya, 2012 27 0 Madagascar, 2016 27 Togo, 2013 Togo, 2013 Uganda, 2013 Uganda, 2013 Kenya, 2012 Kenya, 2012 Nigeria, 2013 Nigeria, 2013 Tanzania, 2014 Tanzania, 2014 Madagascar, 2016 Madagascar, 2016 Tanzania, 2014 14 0 10 20 30 40 50 Doctors Nurses Percent Source: SDI database. Table 7.2: Outcomes for select standardized patient cases in Nairobi, urban China, and India Total Time Time with Antibiotics Case and location Common price (US$ waiting in provider (never N preferred case management checklist equivalent clinic (mins) (mins) necessary) PPP) Diarrhea (in an 18-month-old child) Nairobi 0.73 51.08 4.45 0.20 6.63 0.32 40 Urban China 0.00 1.03 1.13 0.17 2.73 0.43 42 India 0.18 9.97 1.57 0.14 1.22 0.63 389 Unstable angina (in an adult) Nairobi 0.10 53.59 8.12 0.25 12.51 0.60 42 Urban China 0.63 2.13 4.09 0.18 4.92 0.08 40 India 0.41 9.94 3.56 0.25 1.67 0.20 323 Source: Daniels et al. 2017. across domains. A recent study223 of compliance with knowledge, training, or the availability of supplies (Das IPC practices in primary health care in Kenya found and Hammer 2014). Differences between private and that, out of more than 100,000 interactions, mean public facilities were minor. compliance was only around one third (Bedoya et al. 2017). It also varied widely across domains: health Absence rates were high prior to devolution. Absence care workers followed IPC practices for injection and rates, which are often interpreted as a proxy for provider blood draw safety during 87.1 percent of the relevant effort, were high at the time of the SDI survey in 2012, interactions and waste segregation of needles during with a national average of 27.5 percent of health 81.9 percent; in contrast, for the segregation of medical workers absent from the facility (Figure 7.29b). Absence waste (other than needles and syringes) IPC practices rates were higher for public providers compared to were followed only 5.4 percent of the time, and for private and non-profit: 29.2 percent versus 20.9 percent proper hand hygiene only 2.3 percent. Patient safety (Martin and Pimhidzai 2013). Interestingly, among is driven by behavioral norms rather than technical public providers the absenteeism rate was higher in urban facilities (37.6 percent) versus rural facilities (28.3 223 Data collection in this study was conducted across 1,680 health workers during outpatient interactions with more than 14,000 patients at close to percent) even though the difference is not statistically 1,000 facilities in 2015. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 173 Health and Poverty significant at conventional levels.224 However, the and, to a somewhat lesser extent, in stunting rates, SDI survey was conducted before delivery of health has been unprecedented. Also, progress in outcomes services was devolved to the counties, which may have and uptake of health services has been mostly pro- resulted in changes in the staffing of facilities. poor. This chapter notes that these improvements are largely due to demographic trends, improved Private and public providers do not differ in terms standards of living, and an increase in the uptake of of case management, but private provision is more low-cost, preventive goods with proven impact. A expensive and patients spend less time waiting to see large body of evidence strongly suggests that these a provider. Data from the SDI suggest few differences interventions should continue to be provided free of in the ability of public and private providers to diagnose charge. Two cases in point are ITNs and deworming and treat common conditions. Similarly, Daniels et medicines. al. (2017) find that the main differences between public and private provision in Nairobi are in the time At the same time, the poor and some remote regions patients spend waiting, the time they spend with the still face challenges in accessing quality healthcare provider, the checklist providers apply, and the total services. Children from poor families are less likely to price patients pay for treatment (Table 7.3). In terms be vaccinated and their mothers are less likely to give of case management, differences were not statistically birth in the presence of a qualified health provider. In significant except for tuberculosis, for which public fact, in all domains – outpatient care, inpatient care, providers had a higher likelihood of recommending the and preventive care – and across almost all age groups, right course of treatment. the poor are less likely to use health services. They also often have to overcome greater distances to access 7.4 SUMMARY AND POLICY IMPLICATIONS health care. These gaps remain large and significant K enya has experienced remarkable improvements in the health indicators of its population over the last fifteen years. The decline in under-five mortality and are a major cause for concern. Addressing these gaps and maintaining the momentum achieved will require further strengthening of primary care. Table 7.3: Primary outcomes for standardized patient cases by sector Significant at Public Private one-percent level? Preferred management Asthma: inhaler or bronchodilator 0.79 0.82 Child diarrhea: ORS 0.62 0.78 Unstable angina: referral, ECG, or aspirin 0.14 0.07 Tuberculosis: sputum test 0.79 0.36 Yes. Basic statistics (selection) Time waiting in clinic (mins) 94.70 26.53 Yes. Time with provider (mins) 4.21 8.64 Yes. Checklist 0.25 0.44 Yes. Total price (KSh) 141.54 563.05 Yes. Medications (selection) Antibiotics (never necessary) 0.47 0.50 Observations 55 111 Source: Adopted from Daniels et al. 2017, table 4. 224 In multivariate regressions, the only consistently significant pattern the authors find is a positive partial correlation between the number of staff and the absenteeism rate: controlling for location, type of cadre, available infrastructure, as well as ownership and type of facility, the absenteeism rate is at least 20 percentage points higher if the number of health workers increases (p. 50). This is consistent with a greater propensity to free-ride in larger groups. 174 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Health and Poverty While health insurance coverage is low, the incidence provision and towards public provision. But there is of catastrophic health expenditures has likely no evidence that this reform increased institutional decreased recently. Only around 20 percent of the deliveries overall. Rather, the share of deliveries population are covered by health insurance, with large increased at a similar rate as before the reform, while differences between the poor and the better-off and those that would have delivered in private facilities are between rural and urban areas. Among those that are now more likely to do so in public facilities. covered, a large majority are beneficiaries of the NHIF. At the same time, there is evidence that the incidence Because the poor are more likely to depend on of catastrophic health expenditures has declined public health services than the rich, the recent and that households rarely resort to adverse coping disruptions caused by labor disputes between the strategies to finance healthcare. This is in line with the government and public-sector unions affect the removal of user fees in 2013 for a range of public health poor disproportionately. A string of recent health services, including deliveries, as well as improvements worker strikes in the public sector that culminated in in living standards and overall population health. Poor major walk-outs in 2016 and 2017 have disrupted the Kenyans in the informal sector may therefore have little public health service provision, affecting the poor incentive to voluntarily insure, making it harder for the disproportionately. While substituting private for public government to expand coverage. provision during strikes is an option for the better-off, the costs associated with private provision are likely Devolution has the potential to address some of prohibitive for the poor. This chapter finds that despite the most pressing concerns, including a shortage only sluggish growth in real terms, health workers’ of health workers in some localities. This chapter salaries in Kenya remain high by regional standards. has documented wide geo-spatial variation in uptake One way of bringing more transparency to the public of health services and outcomes as well as variation debate about adequate remuneration would be to in inputs prior to devolution, particularly health simplify wage scales by scrapping at least some of professionals. For instance, a general shortage of these allowances that almost all public-sector workers health workers was aggravated prior to devolution are entitled to, while adjusting base salaries accordingly. by a maldistribution across counties. By making county governments accountable and providing the The sustainability of health financing, particularly resources needed to address pressing concerns at their the financing of priority programs, should also be a level, decentralization seems the right way to address priority moving forward. A recent World Bank report these inequities. However, more analysis on its effects has highlighted funding gaps in five priority health and challenges is required. So far, little data has been programs (Immunization, HIV/AIDS, Tuberculosis, produced and counties also initially struggled with the Malaria and Reproductive Health) (World Bank 2018a). increased responsibilities. Healthcare financing in Kenya still relies significantly on donors, despite the fact that the ratio of donor The recent removal of user fees for a number of financing to total health expenditure has been services, including deliveries, has resulted in a shift in declining recently. One vehicle to increase revenues demand from private to public provision, but there is through an expansion in the coverage of the NHIF, is no evidence that it positively affected the share which would increase total member contributions. of deliveries in formal health facilities. While public Also, the government could consider introducing health services account for the majority of healthcare “health taxes” on food and drinks that contain high provision in Kenya, the private sector has also played amounts of saturated fat, sugar, salt, or other unhealthy a significant role in the past. However, the removal of ingredients. This would also address the problem of user fees in public facilities for some services, notably rising obesity among urban, better-off Kenyans. deliveries, has shifted demand away from private KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 175 CHAPTER 8 VULNERABILITY, SHOCKS, AND SOCIAL PROTECTION SUMMARY Although vulnerability and poverty rates fell over the last decade, just over half of Kenya’s population is currently vulnerable to falling into poverty in the near future. Vulnerability rates fell faster in rural areas than they did in urban areas between 2005/06 and 2015/16, but the current urban-rural differences are still very large – 43 percent in urban areas, and 57 percent in rural areas. Poverty and vulnerability are highly correlated, but over one third of non-poor Kenyans are classified as vulnerable. Many of these non-poor-but-vulnerable households are clustered just above the poverty line, meaning that even a moderate shock could push them below the line. Vulnerability rates vary widely by county, and are highest in the north and east of the country. The county vulnerability map looks similar to the county poverty map. The prevalence of vulnerability is highest in Mandera, Garissa, Samburu, and Turkana. Rates are significantly lower in the central counties, particularly in Nyeri, Kirinyaga and Nairobi. Vulnerability is largely concentrated in certain groups of households, particularly those that are engaged primarily in agriculture, and those which have a head with low educational attainment. Fewer than half of Kenyan households had agriculture as the main sector of employment in 2015/16, yet this group contained 57 percent of the vulnerable population. Similarly, although the share of households headed by someone with a primary education or lower was 64 percent in 2015/16, this group made up nearly 80 percent of all vulnerable households. The overall prevalence of both economic and agricultural shocks declined between 2005/06 and 2015/16. However, the incidences of certain kinds of shocks affecting agricultural households went up. Agricultural households were far more likely to report crop losses from preventable causes such as crop diseases or pests in 2015/16 than they were in 2005/06. The relative stabilization of food prices was evidenced by a significant decline in the share of households reporting an economic shock in the form of a large food price increase. There has been a reduction in households resorting to coping strategies with adverse implications for future wellbeing. Nevertheless, the poor and those in rural areas were still more likely to resort to these mechanisms. The share of the poorest households that sold productive assets in response to experiencing a shock fell from almost one third to under one sixth. The most common response of the poorest households after experiencing a shock was to reduce consumption, while for the richest households the most common response was to use savings. Rural households were more likely to use multiple coping strategies for non- agricultural shocks than they were for agricultural shocks. There has been an expansion of social protection programs in Kenya, but overall coverage remains low relative to existing needs. The effort that has been made to coordinate and harmonize social protection programs, combined with the creation of a registry of beneficiary households means that the country is well placed to expand assistance to vulnerable households. An assessment of the KIHBS 2015/16 data shows that social protection programs have generally had positive impacts. Three main findings about the existing programs are that a) they are well-targeted; b) they have had positive effects on school enrolment, have reduced the probability of children working, and have reduced food insecurity; and c) poor non-beneficiaries would benefit greatly from expanding such programs. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 177 Vulnerability, Shocks, and Social Protection 8.1 INTRODUCTION for OVC, the Hunger Safety Net Program (HSNP), T here is a close relationship between poverty and the Persons with Severe Disability Cash Transfer and vulnerability in the developing world. Many (PWSD). Resources allocated to these four programs households that are considered non-poor because has grown significantly, and coverage increased from they are just above the poverty line may experience around 240 000 households in 2013 to around 770 000 a negative shock which could cause them to fall households in 2016. The Kenyan government spends into poverty.225 Similarly, poor households often find around 0.7 percent of GDP on social protection in themselves in poverty traps in which they are both general, and around 0.3 percent of GDP on safety nets more likely to experience negative shocks and are in particular (Álvarez and Van Nieuwenhuyzen 2016). less able to cope with these shocks. In Kenya in This is substantially lower than the sub-Saharan African particular, a number of studies have documented average of 1.4 percent of GDP spent on social safety nets, how weather-related shocks negatively affect both and a developing world average of 1.6 percent (World the income from sales of crops and the welfare of rural Bank 2015c). Overall coverage of the programs, at households engaged in rain-fed agriculture (Wineman about 6 percent of households, is low compared to the et al. 2016; Christiaensen and Subbarao 2005). These existing need for social protection. Therefore, although studies have found that droughts, rather than floods, the expansion of the programs is commendable, there are the predominant weather shock affecting welfare is still a large scope for further coverage increases. across the different regions of the country. In general, poor households in Kenya are 78 percent more The Kenyan government is planning further likely to experience a negative shock than non-poor expansions of the NSNP, as well as complimentary households (Government of Kenya 2012). The specific social interventions. Given this commitment, it is vulnerabilities that are faced by women, and by female- important to identify where poverty and vulnerability headed households are likely to mirror those that were are concentrated, as well as what impact the existing described in relation to poverty in Chapter 3. In general, cash transfer programs are having on household women who went through a marital dissolution, and welfare and on reducing the risks faced by these who often need to take care of young children are households. It is also important to evaluate the more likely to be in poverty. In addition, women are prevalence and intensity of the shocks experienced disproportionately affected by HIV/AIDS in Kenya. by poor and vulnerable households, as well as which coping strategies are used to mitigate the negative Social protection programs are an important policy welfare impacts of these shocks. tool that can be used to raise poor households above the poverty line, and to reduce the vulnerabilities This chapter has three main aims, all of which fall faced by poor and non-poor households. Coverage of under the overarching aim of understanding how to these programs in Kenya is relatively low, but there has address vulnerability and make poverty reduction been a concerted effort from government to increase sustainable in the long run. First, it will construct and efforts to improve social protection. This has been analyze changes in the vulnerability profiles for Kenya particularly evident in the expansion of cash transfers in 2005/06 and in 2015/16. Second, it will analyze and which are targeted at the poorest and most vulnerable compare the welfare shocks that affected households in people in the country. The National Safety Net 2005/06 and 2015/16, as well as which coping strategies Programme (NSNP) was established in 2013 in order were adopted in the face of these shocks. Third, it will to harmonize the four cash transfer programs in the assess the coverage and effectiveness of Kenya’s social country in an integrated system. These consist of the safety net programs, while also measuring their impact Older Persons Cash Transfer (OPCT), the Cash Transfer on different measures of household welfare. 225 These shocks can be idiosyncratic (for example the death of an employed member of the households), or they could be covariate (for example a drought). 178 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Vulnerability, Shocks, and Social Protection Box 8.1: Concepts of risks, shocks and vulnerability Risks: Risks are potentially adverse events that could cause welfare losses. They are distinguished from shocks which are the actual realizations of these risks and losses. Risks can be major impediments to households escaping poverty over time (World Bank 2013c). Poor households may be more likely than non-poor households to be exposed to risks because of a relative lack of ex ante options such as insurance and income diversification (World Bank, 2007). There are also some risks which are commonly and widely distributed within Kenya across socio-economic groups, for example external shocks such as natural disasters, or conflicts (Carter et al. 2007). Shocks: Shocks are sudden adverse events that may cause material or human capital damages. The economic impacts of shocks can have devastating effects on household resources, and are important determinants of poverty dynamics (Dercon 2004). Households that are poorly-equipped to handle shocks are more likely to fall into poverty (or remain in chronic poverty) as a result of experiencing a shock. Shocks have been found to increase poverty rates by up to 4 percent in Mexico (Rodriguez-Oreggia et al. 2013), to be the main driver of a 9 percent increase in poverty in the Philippines (Datt and Hoogeveen 2003), while households in Ethiopia reported between 13 percent and 28 percent lower consumption levels several years after suffering a shock (Dercon, Hoddinott, and Woldehanna 2005). There is a wide range of natural and man- made shocks, which may affect one household at a time (idiosyncratic shocks) or many households at the same time, typically within close geographic proximity to one another (covariate shocks). While idiosyncratic shocks can be singularly devastating, covariate shocks can be even more difficult to cope with, as households may not be able to rely on the traditional networks formed by other households in the area (World Bank 2007). Vulnerability: Vulnerability combines the concepts of risk and poverty, and reflects the probability that a household will be poor in the future. It is a forward-looking measure that takes into account the risks of a household experiencing a shock that would push it into poverty in the future. Vulnerable households also include those that are expected to remain in poverty in the near future, even if they do not experience shocks. A non- poor household may be vulnerable to poverty if the household faces a high risk of suffering shocks in the future. Vulnerability can vary geographically and across households, depending on the structure of risks and the resources available to cope with shocks. Efficient risk management reduces the vulnerability to poverty. As noted in World Bank (2007), the ability of households to reduce or prevent vulnerability depends on three broad areas. The first is the severity and frequency of risks facing households. The second is the level of household resources which can include financial assets as well as physical capital such as land and livestock. The third is access to social networks (family, friends, neighbors, community associations, markets, etc.) and public social protection programs. The first requirement for efficient ex ante risk management is an accurate identification of risks. Based on this, households decide on their risk portfolio by adopting different forms of formal or informal insurance mechanisms. In cases where a risk is realized and a shock occurs, ex post risk management tools are required to compensate for losses. Efficient risk management leads to resilience to shocks and decreases the vulnerability to poverty. A clear understanding of the profiles and causes of vulnerability matters for the design of interventions that aim to prevent rather than alleviate poverty. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 179 Vulnerability, Shocks, and Social Protection 8.2 VULNERABILITY insurance for vulnerable households might be a more efficient way of preventing households from becoming 8.2.1 Defining and measuring vulnerability poor in the future. H ouseholds are typically exposed to a large range of potential idiosyncratic and covariate shocks that can cause substantial income fluctuations if Household welfare is not a static concept, and vulnerability to poverty acknowledges this by realized. Households in risky environments have combining the concepts of poverty with risk developed various ex-ante and ex-post risk-coping exposure and risk management. The classification of strategies to reduce income fluctuations, or to insure vulnerability is driven by two components. The first is consumption against these income fluctuations. As a household’s expected level of welfare, for example many poor households have limited or no access to consumption or income. The second is the expected formal insurance and credit, they rely on informal level of variation of welfare in the future. Several coping strategies. Some of these mechanisms that types of vulnerability can be drawn from these two allow households to mitigate the impacts of shocks components. On the one hand, there are vulnerable include transfers and remittances, asset liquidation, households that are currently classified as poor and income diversification and migration (Morduch 1999; which are expected to remain poor in the future. These Barnett, Barrett, and Skees 2008). However, these households are often categorized as being chronically instruments are incomplete. Large covariate shocks poor. On the other hand, vulnerable households that such as natural disasters can overwhelm the capacity of are currently non-poor but face large income risks and these instruments, partly because households located are likely to drop into poverty at some point in the within the area of incidence of the shock may be unable future may be classified as being in transitory poverty. to support each other. If this occurs, then households For example, a small-scale farmer who cultivates cash may be forced to reduce consumption, and take other crops may not be recorded as poor after a season measures such as withdrawing children from school with normal weather conditions. However, under less or selling productive assets. These actions can have favorable weather conditions the following season, the long term, possibly irreversible, impacts on household farmer may enter poverty. Therefore, this farmer could members in general, and children in particular (Jacoby be classified as non-poor today, but as vulnerable to and Skoufias 1997; Carter and Maluccio 2003) in terms being in poverty in the future. of human capital accumulation and future productivity. Dynamic assessments of poverty are challenging from Today’s poor may not be tomorrow’s poor, and efforts an empirical perspective. The ideal is to make use of to reduce poverty in the future need to be targeted longitudinal data that cover fluctuations of households at the poor today but also at non-poor households in and out of poverty. These kinds of datasets are still that can be prevented from slipping into poverty. not routinely available in many developing countries. In A lack of options to manage risks may mean that the absence of longitudinal data that track households variance of household consumption over time remains over time, it becomes difficult to quantify the future risk high, particularly in risky environments (Günther and of poverty. Several procedures have been proposed Harttgen 2009). In these cases, a household’s current to overcome this challenge. This chapter follows the poverty status is not necessarily a good indicator of the approach outlined in Chaudhuri, Jalan, and Suryahadi poverty status in future years. Separating out the parts (2002) to classify vulnerability to poverty using cross- of poverty that are structural versus those that are the sectional data. The approach estimates the two result of risks to shocks has important implications from components of vulnerability: the predicted level of a policy perspective. While social assistance programs consumption and the predicted level of variation of a may be more appropriate for poverty alleviation, household’s consumption, and is outlined in Box 8.2. 180 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Vulnerability, Shocks, and Social Protection Box 8.2: Measuring vulnerability using cross-sectional data Vulnerability combines the concepts of poverty, exposure to risks, and risk management in order to predict the probability that a household or individual will be poor in the future. It can be analyzed at various levels, such as the country, household or individual level. In contrast to poverty, vulnerability is a forward-looking measure that reflects the probability of poverty in the future. Therefore, each vulnerability measure that uses consumption as the welfare indicator is very closely related to consumption smoothing and the capacity to cope with shocks (Klasen and Waibel 2015). As the future is uncertain, measuring vulnerability is more complex than measuring cross-sectional poverty. Quantifying a household’s vulnerability is subject to various challenges. Ideally, longitudinal data that capture welfare dynamics and exposure to shocks are used, which help to accurately determine welfare dynamics inherent to vulnerability concepts. However, if longitudinal data are not available, several methods have been proposed to estimate household’s vulnerability to poverty (Chaudhuri, Jalan, and Suryahadi 2002; Günther and Harttgen 2009). This chapter follows the methodology outlined in Chaudhuri, Jalan, and Suryahadi (2002), which can be summarized in four steps: 1. In the first step, the main correlates of the household’s consumption level are identified to assess the strength of the relationships between different characteristics and household welfare. That is, household consumption is regressed on a set of independent variables which include household composition and demographics, livelihoods, and regional and geographic control variables. 2. In the second step, the relationship between the household characteristics and the risk of welfare shocks is estimated. The variation in household consumption that is not explained by the estimation model in step 1 includes the household’s risk of shocks. This variation is explored to test which characteristics are associated with the risk of welfare shocks. 3. Based on step 1 and 2 a household’s future level of consumption and variation of consumption is predicted. 4. Households that are determined to have a probability of being in poverty at any stage over the next 2 years of over 50 percent are classified as vulnerable (following Günther and Harttgen 2009). Furthermore, poor households that are predicted to be poor may be classified as chronically poor, whereas non-poor households that face significant risks of welfare fluctuations are classified as transient poor. It should be borne in mind that the approach is based on several assumptions about the distribution of risks, and that due to the data limitations these indicators of vulnerability need to be interpreted with caution. The main limitation is that in the presence of cross-sectional data that only covers one year, it has to be assumed that household’s variation of consumption is constant over time. That is, the indicators are unable to account for large but rare shocks that do not occur in every year. Other important assumptions include the absence of measurement error in consumption reports, and assumptions on the distribution of risks and the validity of ordinary least square estimates (see Klasen and Povel 2013) for a more detailed discussion). The vulnerability threshold is defined relative to uses the methodology proposed by Günther and the predicted probability that a household will be Harttgen (2009) and defines the “near future” as being in poverty in the near future. This classification is within the next two years. Households are considered complicated by the fact that there are no clear rules to be vulnerable if their predicted probability of being about what time period constitutes the “near future”, and below the poverty line at any stage within these two what the predicted probability should be. This chapter years is greater than 50 percent. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 181 Vulnerability, Shocks, and Social Protection 8.2.2 Profiles of vulnerability in Kenya 2005/06 and within two years. The map of vulnerability reflects the 2015/16 same patterns that were seen in the poverty map in The fall in the vulnerability rate was larger than Chapter 2, though the rates of vulnerability are always the fall in the poverty rate between 2005/06 and higher than the rates of poverty. This has important 2015/16, and this reduction was largely driven by policy implications, as the reduction in poverty may not rural areas. Vulnerability rates are higher than poverty be sustained if the sources and potentially damaging rates because there are significant numbers of Kenyans coping strategies of shocks are not addressed. who are non-poor but are vulnerable to falling into poverty in the near future. More than two thirds of Although there is a strong link between poverty and Kenyans were classified as vulnerable in 2005/06, vulnerability, not all poor are vulnerable, and not all and this had reduced to just over half in 2015/16. non-poor are non-vulnerable. In 2015/16, close to 80 As can be seen in Figure 8.1, poverty rates in urban percent of the poor were also vulnerable. This means areas fell from 34 percent to 29 percent, while poverty that about one fifth of the poor population in Kenya rates in rural areas fell from 50 percent to 39 percent. was expected to be consistently non-poor during The vulnerability rate in urban areas fell by almost 9 the following two years. Conversely, 37 percent of percentage points, which represented a much smaller the non-poor were classified as vulnerable. Figure 8.3 drop than the corresponding 17 percentage point disaggregates the share of the poor and non-poor that decrease in rural areas. are vulnerable by rural and urban location. The share of the poor that are vulnerable is not significantly different Vulnerability rates vary widely by county, but are between the two groups – 79 percent in rural areas and highest in the north and east of the country. The 76 percent in urban areas. There is, however, a large county with the highest vulnerability rate is Mandera, difference in the share of the non-poor population that in which almost all households have a greater than is likely to be poor in the near future. 41 percent of the 50 percent predicted probability of experiencing rural non-poor are classified as vulnerable, compared poverty within the next 2 years. Other counties to just over one quarter of the urban non-poor. Many with similarly high vulnerability rates are Garissa, of these non-poor-but-vulnerable households had Samburu, Turkana and Busia. Counties in the center consumption levels that were clustered just above the of Kenya generally have the lowest vulnerability rates, poverty line, as seen in Appendix H. As expected, the with fewer than one fifth of the population in Nyeri, figure shows that the density of vulnerable households Kirinyaga and Nairobi expected to experience poverty decreases as distance from the poverty line increases. Figure 8.1: Poverty and vulnerability in Kenya: 2005/06 and 2015/16 80 73 70 68 Percentage of population 60 56 52 51 50 47 50 42 40 36 39 34 30 29 20 10 0 Poor Vulnerable Poor Vulnerable Poor Vulnerable Total Urban Rural 2005/06 2015/16 Source: Own calculations based on KIHBS 2005/06 and 2015/16. 182 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Vulnerability, Shocks, and Social Protection Figure 8.2: Vulnerability rates by county: 2015/16 100 90 80 70 Vulnerability rate 60 50 40 30 20 Vulnerability rate 10 >90% 80.1% to 90% 75.1% to 80% 0 70.1% to 75% Nyeri Kirinyaga Nairobi Tharaka-Nithi Meru Embu Murang'a Machakos Lamu Mombasa Taita-Taveta Nyandarua Kiambu Nakuru Baringo Makueni Nyamira Nandi Siaya Kericho Homa Bay Kisumu Laikipia Bungoma Trans Nzoia Vihiga Migori Kakamega Kwale Kisii Uasin Gishu Bomet Kajiado Kitui Elgeyo-Marakwet Isiolo Wajir West Pokot Tana River Marsabit Busia Turkana Samburu Garissa Mandera Narok Kili 65.1% to 79% 60.1% to 65% 50.1% to 60% 35.1% to 50% 10% to 35% Source: Own calculations based on KIHBS 2015/16. Figure 8.3: Vulnerability rates by poverty status: 2015/16 100 90 21 24 80 70 59 60 72 Percent 50 79 76 40 30 41 20 28 10 0 Poor Non-poor Poor Non-poor Rural Urban Vulnerable Non-vulnerable Source: Own calculations based on KIHBS 2015/16. However, there are still a fair number of households poverty line. The dashed horizontal lines correspond to that are classified as vulnerable even though their the rural and urban poverty rates of 38.8 percent and consumption levels are 2 to 3 times the poverty 29.4 percent, respectively. The difference in vulnerability line. The fact that so many non-poor-but-vulnerable rates between the areas is greater than the difference in households are clustered just above the poverty line poverty rates. 56.5 percent of the rural population was means that even a moderate idiosyncratic or covariate classified as vulnerable, compared to 43.1 percent of the shock is likely to push these households into poverty. urban population. This 13.4 percentage point difference in vulnerability is larger than the 9.4 percentage point Even though vulnerability rates fell faster in rural difference in poverty between the areas. areas than in urban areas, the differences in 2015/16 were still very large. The cumulative density functions The share of households that had services as the main (CDFs) in Figure 8.4 show the cumulative proportion of sector of employment increased, and the poverty the rural and urban populations against consumption rate for this group decreased. The first section of Table levels relative to the poverty line. The vertical line on 8.1 shows different shares of the population, the poor, the x-axis corresponds to consumption exactly at the and the vulnerable by the main sector of employment KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 183 Vulnerability, Shocks, and Social Protection Figure 8.4: CDFs of the rural and urban population: The rapid pace of urbanization of close to 1 2015/16 percentage point per year was reflected in the 1 .9 composition of the poor and vulnerable. 80 percent Proportion of the population .8 of Kenyans lived in rural areas in 2005/06, and this share .7 dropped to 71.6 percent in 2015/16. The composition .6 .5 of and changes in the urban and rural poor and .4 vulnerable populations over the time period tracked .3 each other closely. The share of the poor and the .2 vulnerable who lived in urban areas increased by about .1 0 8 percentage points. In 2015/16 almost one quarter of 0 50 100 150 200 250 300 the poor and the vulnerable could be found in urban Consumption as % of the poverty line areas, compared to 15 percent in 2005/06. Therefore, the Rural Urban Rural poverty rate: 38.8%. Rural vulnerability rate: 56.5% nature of poverty and vulnerability became increasingly Urban poverty rate: 29.4%. Urban vulnerability rate: 43.1% urbanized, as the rise in the national share of the urban Source: Own calculations based on KIHBS 2015/16. population was met by a similar rise in the share of the of the household.226 There was a decline in the overall poor and vulnerable who lived in urban areas. share of households with agriculture as the main sector of employment – this proportion dropped to less than The fall in the share of the population who had no half in 2015/16. The relative decline in the share of education was not fully reflected in changes in the agriculture was replaced by a combination of increases composition of the poor and vulnerable. Overall there in services and construction. Even though the overall was a steep fall in the proportion of household heads share of services grew by 5 percentage points, the with no education, with bulk of this difference being share of all the poor who lived in households in which taken up by the secondary and tertiary education services was the main sector of employment dropped categories. Almost 80 percent of the vulnerable in by 2 percentage points. The changes for employment 2015/16 had a primary education or less, a very similar in the construction sector went the opposite way, figure to 2005/06, despite the fact that the overall share with 8 percent of the poor working in construction in of no education fell from 22 percent to 16 percent. By 2015/16, compared to 3 percent in 2005/06. 2015, more than a third of household heads had at least a secondary level of education. This proportion Changes in the profile of vulnerability were close was significantly higher than the corresponding to changes in the overall composition of the shares in the poor and vulnerable, which were around employment sectors of households. Although the 20 percent. share of the poor in services declined between 2005/06, the share of the vulnerable in households with services There was a small increase in the share of female- as the main employment sector increased by almost headed households over the period, but these 6 percentage points. There was little change in the households were not more likely to be vulnerable than vulnerability profile of agricultural households over male-headed households in 2015/16. Interestingly, the period. In both 2005/06 and 2015/16, the share of female-headed households were a little more likely to the vulnerable in agriculture was 9 percentage points be in poverty than male-headed households, but the higher than the share of agricultural households in the gender shares of vulnerability match the overall shares general population. shown in the first two columns of the table. There was, however, a small change over time. In 2005/06 the share of vulnerable households that were female- 226 In both 2005/06 and 2015/16 the main sector of employment was headed was lower than the overall share of female- derived from the occupation codes of the head of the household. 184 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Vulnerability, Shocks, and Social Protection Table 8.1: Profiles of the poor and the vulnerable: 2005/06 and 2015/16   Total Poor Vulnerable   2005/06 2015/16 2005/06 2015/16 2005/06 2015/16 Main employment sector           Agriculture 56.5 48.4 62.5 59.1 65.5 57.2 Manufacturing 5.1 5.6 4.3 5.1 5.2 4.8 Household Services 35.0 39.0 30.2 28.1 25.3 31.0 Construction 3.5 7.1 3.0 7.8 4.0 7.0 Location Rural 79.9 71.6 85.1 76.9 85.1 76.8 Urban 20.1 28.4 14.9 23.1 14.9 23.3 Education       No education 22.4 15.8 32.0 27.7 26.3 23.1 Household head Primary education 47.2 47.7 51.7 53.4 54.1 55.4 Secondary education 28.5 32.3 16.1 18.6 19.1 20.4 Tertiary education 1.9 4.2 0.2 0.4 0.5 1.2 Gender Female 26.3 29.6 28.4 32.9 23.5 29.4 Male 73.7 70.4 71.6 67.1 76.5 70.6 All 46.6 36.2 68.4 51.7 Source: Own calculations based on KIHBS 2005/06 and 2015/16. headed households (23.5 percent versus 26.3 percent). Figure 8.5: Vulnerability rates relative to the average: 2015/16 In 2015/16 the proportions were far closer to one 90 another (29.4 percent versus 29.6 percent). 80 70 Unconditional vulnerability rates were significantly 60 Percentage higher than the average for agricultural households, 50 40 as well as those living in households in which 30 the head had no education. Figure 8.5 shows the 20 unconditional vulnerability rates associated with 10 Agriculture Manufacturing Services Construction Rural Urban No education Primary education Secondary education Tertiary education Female Male different characteristics, relative to the average national vulnerability rate of 51.7 percent. Triangles above the dashed line have higher-than-average vulnerability rates, while the squares below the dashed line Household Household head National vulnerability rate represent lower-than-average vulnerability rates. There Source: Own calculations based on KIHBS 2015/16. are some useful insights that can come from plotting these different correlates together, not least the fact There were no statistically significant differences in that doing so can potentially uncover their ordering of the vulnerability rates of households engaged in importance. However, as noted in Dang, Lanjouw, and manufacturing and services. 61 percent of agricultural Swinkels (2017), the major caveat in presenting data this households227 were classified as being vulnerable, way is that is that there will be overlap between different compared to 41 percent of households that were groups (for example, those with lower education levels primarily engaged in services. This is partly explained may be more likely to live in agricultural households). by the fact that regular weather and price shocks 227 Agricultural households are defined as households in which the head’s main sector of employment/activity is agriculture. These make up around 40 percent of households in Kenya. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 185 Vulnerability, Shocks, and Social Protection make agricultural income highly variable. The stark 8.3 SHOCKS AND COPING STRATEGIES IN differences in vulnerability rates according to the 2005/06 AND 2015/16 U education level of the household head are evident in nderstanding the kinds of shocks that households the figure. Nearly 90 percent of households with a head face, where different shocks are concentrated, and who had no education were vulnerable, compared to what strategies households employ to cope with shocks the national average of 51.7 percent. Less than one has important implications for tackling vulnerability and third of households with a head who had a secondary for guiding the design and expansion of social protection level of education were vulnerable in 2015/16. A programs. KIHBS 2005/06 and KIHBS 2015/16 contain separate analysis of agricultural households in the 23 modules that ask respondents wide-ranging questions arid and semi-arid counties (not shown) showed that about the prevalence and welfare effects of shocks, and it is these households are not statistically significantly more to these that the chapter now turns. likely to be vulnerable than agricultural households in the other 24 counties. 8.3.1 Incidence and types of shocks 2005/06 and 2015/16 Although the unconditional vulnerability rates The overall prevalence of both economic and for male and female-headed households were agricultural shocks appears to have declined. Half both close to the overall average, female-headed of the poorest quintile of households experienced households were statistically significantly more likely an economic shock as measured in 2005/06, with the to have been classified as vulnerable in 2015/16. corresponding proportion in 2015/16 being 30 percent. The vulnerability rate for female-headed households In fact, in 2015/16 the richest 20 percent of households was 54.2 percent, almost four percentage points were the most likely quintile to report having higher than the corresponding rate for male-headed experienced an economic shock in the last 5 years. households of 50.6 percent. These differences for vulnerability reflect similar differences for poverty rates The probability of a household reporting an by the characteristics of the household head that were agricultural shock did not vary over the 2015/16 reported in Chapter 2. consumption distribution. Interestingly, the decreasing prevalence of agricultural shocks shown In summary, the overall decrease in vulnerability by the blue bars in Figure 8.6 had become far flatter was driven by changes in rural areas, but there is a by 2015/16, mainly because of fewer shocks for the wide geographic variation in vulnerability in Kenya. poorest households. The prevalence of “other shocks” Households in the north and east of the country are far was higher in 2015/16 than it was in 2005/06. This may more likely to be vulnerable than households in other be explained to a certain extent by the fact that illness areas. Households in which the primary economic shocks were not directly asked in the later survey and activity is agricultural-based are significantly more households inserted these shocks into the catch-all likely to be vulnerable than households engaged in “other” category in the shocks module. other activities, as are households in which the head has a low level of educational attainment. A significant The urban-rural difference in the prevalence of number of the non-poor are vulnerable to falling into economic shocks disappeared between surveys, poverty in the near future. One potential trigger for while the difference in agricultural shocks remained poverty entry is the experience of a shock. Profiling the large. As can be seen in Figure 8.7, more than 60 prevalence, severity and coping strategies associated percent of urban households reported experiencing with different kinds of shocks is the focus of the next an economic shock in the 2005/06 KIHBS. This declined section of this chapter. to 33 percent of urban households in the 2015/16 186 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Vulnerability, Shocks, and Social Protection Box 8.3: Measuring the prevalence of and responses to shocks in KIHBS data Questions about the prevalence, impact and responses to shocks were asked in similar ways in KIHBS 2005/06 and KIHBS 2015/16. There were, however, a number of differences that are worth highlighting if comparisons across the surveys are to be made. The respondent answering the household questionnaire in both surveys was asked a series of questions about recent shocks to household welfare over the last 5 years. Although this is a fairly long time period, the respondent was also asked exactly how long ago the shock took place, so it is possible to get some finer details about the timing of each shock. Households were asked about 23 different kinds of shocks in 2005/06, and 27 different kinds of shocks in 2015/16. The responding household member was asked to rank the three most severe shocks to have hit the household over the last 5 years, and was then asked to estimate the value lost due to the shock, whether the shock was idiosyncratic or covariate, and what strategies household members used to mitigate the negative effects of the shock. In both surveys, respondents were presented with 25 different kinds of responses to shocks, and were able to rank the response strategies in order of importance. One drawback is that we are not able to identify multiple experiences of the same shock over the last five years. For example, it is possible that a household may have experienced a drought more than once in the five years prior to being interviewed. In order to simplify some of the analysis, shocks are grouped into categories in a similar way to what is done in a former Kenya poverty and inequality assessment by the World Bank (2008). Economic shocks: Business failure (non-agricultural); loss of salaried employment or non-payment of salary; end of regular assistance, aid or remittances from outside the household; large rises in the price of food; a large rise in agricultural input prices; breakup of the household; dwelling unit was damaged or destroyed. Agricultural shocks: Drought or flood; crop disease or crop pest; livestock died or were stolen; household experienced a severe water shortage. Other shocks: Bread winner was jailed; robbery, burglary, or assault; eviction; ethnic clashes or conflict; other shocks not listed. Health (2005/06 only): Chronic or severe illness or accident; deaths of economically active household members; household member diagnosed with HIV. Chronic or severe illness was the second most prevalent shock in 2005/06 after drought/flood. Unfortunately, this question was not asked in KIHBS 2015/16, and so we do not include health-related shocks for these years, as the comparison to 2005/06 would not be valid. KIHBS – a number that was not significantly different Separating the experiences of shocks into the to that in rural areas. Unsurprisingly, rural households poor and non-poor populations reveals that the were far more likely to report having experienced an only significant difference was in the prevalence of agricultural shock in 2005/06 and in 2015/16 compared agricultural shocks. The higher poverty rate associated to urban households. Nevertheless, the incidence of with agricultural households shown in Table 8.1 is agricultural shocks in rural households fell from 53 percent consistent with this finding. There were no significant in the first period to 46 percent in the second period. difference in the prevalence of economic or health KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 187 Vulnerability, Shocks, and Social Protection Figure 8.6: The prevalence of different shocks over consumption quintiles: 2005/06 and 2015/16 a) Economic shocks by b) Agricultural shocks by c) Other shocks by consumption quintile consumption quintile consumption quintile 60 60 60 50 50 50 40 40 40 Percent Percent Percent 30 30 30 20 20 20 10 10 10 0 0 0 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 Quintiles Quintiles Quintiles 2005/06 2015/16 2005/06 2015/16 2005/06 2015/16 Source: Own calculations based on KIHBS 2005/06 and 2015/16. Figure 8.7: Prevalence of shocks by urban-rural location: 2005/06 and 2015/16 2005/06 2015/16 70 70 60 60 50 50 40 40 Percent Percent 30 30 20 20 10 10 0 0 Economic Agricultural Health Other Economic Agricultural Other shock shock shock shock shock shock shock Urban Rural Urban Rural Source: Own calculations based on KIHBS 2005/06 and 2015/16. shocks between the poor and the non-poor in While the previous figure was focused on the urban- 2005/06, and the proportion of poor and non- rural grouping of households, Figure 8.8 restricts the poor households experiencing economic shocks lens to agricultural households only. The scale of the were almost identical in 2015/16. Figure H.2 in the shock reduction for agricultural households between appendix extends the poor and non-poor distinction the two KIHBS surveys was not nearly as significant to vulnerable households. The results show that as the corresponding figures for non-agricultural vulnerable households were actually more likely households. to report having experienced an agricultural shock than poor households were. This brings into focus the The share of agricultural households reporting importance of insuring against these kinds of events, having experienced a drought or flood, some form of as they could serve as a trigger for a household’s entry crop failure, or livestock loss actually increased over into poverty in the near future. the time period. This change is particularly pronounced for the prevalence of crop diseases and pests, which Paying closer attention to agricultural households went from 7.5 percent of households to 25 percent of specifically reveals some interesting differences. households. This has important policy implications, as 188 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Vulnerability, Shocks, and Social Protection Figure 8.8: Incidence of shocks by poverty status, agricultural households only: 2005/06 and 2015/16 2005/06 2015/16 70 70 60 60 50 50 40 40 Percent Percent 30 30 20 20 10 10 0 0 Economic Agricultural Health Other Economic Agricultural Other shock shock shock shock shock shock shock Poor Non-poor Poor Non-poor Source: Own calculations based on KIHBS 2005/06 and 2015/16. the prevalence of crop pests can be prevented, while counties. More than 70 percent of households reported the effects of a drought can be countered through having experienced an economic shock in Kisii, Kitui and more extensive irrigation programs. The increase in the Narok, while under 5 percent of households reported prevalence of experiencing a drought or flood was not experiencing economic shocks in Wajir and Samburu. statistically significant, though the increase in livestock loss was. The proportion of households reporting a There is a fairly high correlation between the extent severe water shortage was the same in both periods, at of economic shocks and the extent of agricultural around 10 percent. shocks at the county level. The counties of Kitui, Narok and Nyamira all have agricultural shock prevalence Greater price stability led to a reduction in the rates of over 80 percent. Households in Kitui and Narok economic shocks experienced by agricultural appear to be particularly prone to experiencing shocks, households. One third of these households reported as they are in the top three counties for both economic experiencing a large rise in the price of food in and agricultural shocks. They stand in stark contrast to KIHBS 2005/06, and this fell to about one quarter of Samburu which has an agricultural shock prevalence households in KIHBS 2015/16.228 Likewise, there was rate of 68 percent, but an economic shock prevalence a small fall in the prevalence of shocks reported as a rate of 4.7 percent. large increase in agricultural input prices. The extent of other economic shocks in agricultural households The prevalence of “other” shocks is more evenly was relatively small, though about 7 percent reported spread around the country, with much lower rates experiencing a business failure in both surveys. than economic and agricultural shocks. No county has more than 40 percent of households reporting The prevalence of economic shocks as reported in having experienced “other” kinds of shocks, though KIHBS 2015/16 was concentrated in counties in the rates are a little above 30 percent in Lamu, Mandera, southern half of Kenya. In general, there was a very West Pokot, Tharaka-Nithi and Bungoma. Fewer than 2 wide range in the reporting of economic shocks across percent of households report experiencing other kinds of shocks in Bomet and Garissa. Both of these counties, 228 Food price inflation was in fact the most commonly reported shock experienced by households in 2005/06. This particular shock also had a in fact, have a very low prevalence of economic shocks consistent prevalence rate across the distribution of consumption, with (under 7 percent) and are also amongst the counties the top quintile almost as likely to report large food price increases as the bottom quintile. This is in contrast to shocks like droughts or floods, reporting the lowest rates of agricultural shocks. which affected the bottom quintile about 4 times more than the top quintile according to KIHBS 2005/06. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 189 Vulnerability, Shocks, and Social Protection Figure 8.9: Shock prevalence for agricultural households only: 2005/06 and 2015/16 50 35 30 25 20 15 10 5 0 Drought Crop Livestock Severe HH business Loss of End of Large food Large agri. Dwelling or ood disease/pest death/theft water failure salaried regular price rise input price rise damaged shortage employment assistance /destroyed /aid Agricultural shocks Economic shocks 2005/06 2015/16 Source: Own calculations based on KIHBS 2005/06 and 2015/16. Figure 8.10: Geographic distribution of different shocks: 2015/16 a) Economic shocks b) Agricultural shocks c) Other shocks Proportion of households Proportion of households >70% >70% 60% to 70% 60% to 70% 50% to 59.9% 50% to 59.9% Proportion of households 40% to 49.9% 40% to 49.9% 30% to 39.9% 30% to 39.9% 30% to 39.9% 20% to 29.9% 20% to 29.9% 20% to 29.9% 10% to 19.9% 10% to 19.9% 10% to 19.9% <10% <10% <10% Source: Own calculations based on KIHBS 2015/16. 8.3.2 Severity of shocks in 2005/06 and 2015/16 Non-response rates about the extent of losses differ The respondent answering the household markedly between the datasets. This may result in questionnaire was asked about how much value some complications when comparing the loss data was lost as a result of different shocks. In this chapter in KIHBS 2005/06 and KIHBS2015/16. The nature of we calculate the severity of the loss from a shock as this non-response is that a respondent said that the the size of the loss in proportion to total household household experienced a particular shock but did not consumption. This amount is then benchmarked provide an amount for the value lost because of the against the richest quintile. For example, a value of 5 for shock. In 2005/06 this non-response rate was about 4 the poorest quintile should be interpreted as meaning percent for economic and agricultural shocks, while in that the severity of losses as a share of consumption 2015/16 it was about 18 percent for economic shocks, was 5 times higher for the poorest quintile than for the and 27 percent for agricultural shocks. Understandably, richest quintile.229 it was difficult for respondent to put a value on losses from health shocks – the non-response rate for these was close to 50 percent in 2005/06.230 229 Loss severity is only calculated for households that reported experiencing 230 The severity of losses from “other” shocks is not included here, as many of a shock. This means that although the number of households in each these categories do not give the respondent the option of providing a quintile will be equal, the number of households experiencing a shock value of the economic loss (examples include the death of the household will not be the same across quintiles. head, or the jailing of a household member). 190 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Vulnerability, Shocks, and Social Protection Figure 8.11: The severity of losses from shocks: 2005/06 and 2015/16 2005/06 2015/16 Poorest Poorest 2 2 Quintiles Quintiles 3 3 5 5 Richest Richest 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 Scale of loss relative to richest quintile Scale of loss relative to richest quintile Economic shocks Agricultural shocks Health shocks Economic shocks Agricultural shocks Source: Own calculations based on KIHBS 2005/06 and 2015/16. The relative severity of agricultural shocks was very in the future, and for whether a household becomes large for poor households according to both KIHBS trapped in chronic poverty, or is able to transition out 2005/06 and KIHBS 2015/16. The poorest quintile of poverty relatively quickly (Dercon 2004). experienced losses that were almost 9 times greater than the richest quintile, relative to household In both 2005/06 and 2015/16 the most common consumption in 2005/06. This had reduced somewhat coping strategy used by the poorest quintile was by 2015/16, but was still very high at about 6.5. The to reduce consumption, while the most common severity of losses from agricultural shocks for quintiles coping strategy employed by the richest quintile 2, 3 and 4 relative to quintile 5 were very similar in was to use savings. Figure 8.12 shows that 44 percent both datasets. of the poorest households reduced consumption as one way of coping with shocks, while 40 percent In contrast, the relative impact of economic shocks sought help from non-household family members in increased for the poorest quintile compared to the 2005/06.231 Households in the richest quintile were far richest quintile. In 2005/06 losses for the bottom less likely to seek help from an institution in 2005/06 quintile were just under 4 times as severe as losses for than households in the poorest quintile, though this the top quintile. This increased to almost 5.5 times in difference had largely disappeared by 2015/16. 2015/16. This result, combined with what was shown in Figure 8.6 suggests that even though poor households The proportion of households in the poorest experienced fewer shocks in 2015/16 compared to ten quintile selling productive assets in response to years previously, the outcomes of the shocks that did shocks fell from 31 percent in KIHBS 2005/06 to 14 occur were more severe in the later time period. percent in KIHBS 2015/16. The richest households were far more likely to use savings as a coping 8.3.3 Coping strategies for dealing with shocks mechanism than any other strategy. The proportion The ways that households respond to shocks may of households that borrowed money in order to help have implications for future wellbeing. For example, mitigate the negative effects of a shock was low in poor households may sell productive assets, reduce both time periods, with a small positive gradient over consumption, or withdraw children from school in the consumption distribution. response to a shock in order to meet immediate needs. 231 The figure shows all the strategies that households employed when This has important consequences for welfare dynamics faced with shocks. The patterns look very similar if only the main coping mechanism used by households is considered. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 191 Vulnerability, Shocks, and Social Protection Figure 8.12: Coping mechanisms over the distribution of consumption: 2005/06 and 2015/16 2005/06 2015/16 56% Reduced consumption, 44% Using savings, 33% Help from family, 40% Using savings, 33% 34% Sold assets, 31% 30% 29% 25% Reduced consumption, 44% Help institution, 25% 19% 17% Help from family, 40% 14% Borrowed, 13% Borrowed, 13% 9% Sold assets, 31% 7% 8% Help institution, 25% Quintile 1 Quintile 5 Quintile 1 Quintile 5 Source: Own calculations based on KIHBS 2005/06 and 2015/16. Rural households were far more likely to sell assets 2015/16, while about one third used savings. In both in response to a shock than urban households 2005/06 and 2015/16 rural households were more were. As shown in Figure 8.13, this was the case in likely to employ multiple coping strategies than urban both surveys, though the overall use of the strategy households were. was lower in 2015/16. This is in line with the previous figure showing that asset sales were more commonly The coping strategies used by rural households used for the poorest households, given the relatively changed significantly between 2005/06 and 2015/16. higher concentration of poverty in rural areas. Not only Figure 8.14 restricts the analysis to rural households are rural households poorer and more vulnerable than only, and shows how these households responded urban households, but they are forced to deplete their to agricultural shocks versus other shocks. There were assets more regularly as well. This suggests that there is some interesting changes between both KIHBS surveys. possible scope for the introduction of some emergency In 2005/06 rural households were more likely to take cash programs which have the potential to offset some up additional work in response to an agricultural shock of the negative effects of shocks such as droughts and than to other kinds of shocks. In 2015/16 this situation floods. Around one fifth of urban and rural households had reversed. In contrast, in KIHBS 2005/06, about one sought help from non-resident family members in third of rural households used savings to deal with Figure 8.13: Coping strategies by urban-rural place of residence: 2005/06 and 2015/16 2005/06 2015/16 80 80 70 70 60 60 50 50 Percent Percent 40 40 30 30 20 20 10 10 0 0 Solid assets Worked more Borrowed Help institution Help family Reduced consumption Used savings Spiritual help Other strategies Multiple strategies Solid assets Worked more Borrowed Help institution Help family Reduced consumption Used savings Spiritual help Other strategies Multiple strategies Urban Rural Urban Rural Source: Own calculations based on KIHBS 2005/06 and 2015/16. 192 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Vulnerability, Shocks, and Social Protection both agricultural and non-agricultural shocks. In KIHBS The main form of borrowing in response to shocks 2015/16 about 42 percent of rural households used was from relatives, and there was no difference in the savings as a mechanism to deal with non-agricultural probability of this strategy being used by poor versus shocks, but only about 26 percent responded to non-poor households. There were very low levels of agricultural shocks with this strategy. Rural households borrowing from formal institutions or from moneylenders, were also far more likely to use multiple coping while around 7 percent of households borrowed from strategies to deal with non-agricultural shocks than relatives. Around one fifth of both poor and non-poor they were to deal with agricultural shocks in 2015/16. households turned to non-resident family members for support following a shock. Table H.1 in the appendix Poor households deplete productive assets more presents coping strategies for agricultural shocks only, regularly than non-poor households in response and shows that the non-poor were again more likely to shocks. Table 8.2 shows a finer level of detail for than the poor to use savings as a coping strategy. Selling the coping strategies employed by households than animals was also still the main form of asset depletion for poor households. There were lower rates of borrowing was the case in earlier figures. Differences between in response to agricultural shocks compared to the non- poor and non-poor households that are statistically agricultural shocks. Poor households reduced both food significant at the 10 percent level and below are and non-food consumption with more regularity than shown in bold. Consistent with the patterns in Figure non-poor households in response to agricultural shocks. 8.12, the non-poor were more likely to have used Both the poor and the non-poor were quite a lot less savings in an attempt to mitigate the negative effects likely to borrow from relatives after agricultural shocks of shocks. Non-poor households were also more likely compared to other shocks. to have started a business, and to have borrowed from a formal institution in response to shocks, but these Although the percentage point differences between two strategies were actually relatively rarely used in the kinds of coping mechanisms used by poor 2015/16. The most common kind of asset depletion versus non-poor households are not particularly for poor households came in the form of selling large, poor households are nevertheless more likely livestock. 13 percent of poor households used the to use strategies that may have adverse dynamics sale of livestock as a coping strategy, compared to 9 implications for welfare. Social protection programs may percent of non-poor households. 20 percent of poor reduce the use of these kinds of strategies, conditional households reduced food consumption in response to on being well-targeted. The next section of this chapter the occurrence of a shock, compared to 16 percent of investigates the coverage and impact of some of the non-poor households. main cash transfer programs in Kenya. Figure 8.14: Coping strategies by shock type – Rural households only: 2005/06 and 2015/16 2005/06 2015/16 50 50 40 40 30 30 Percent Percent 20 20 10 10 0 0 Solid assets Worked more Borrowed Help institution Help family Reduced consumption Used savings Spiritual help Other strategies Multiple strategies Solid assets Worked more Borrowed Help institution Help family Reduced consumption Used savings Spiritual help Other strategies Multiple strategies Urban Rural Urban Rural Source: Own calculations based on KIHBS 2005/06 and 2015/16. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 193 Vulnerability, Shocks, and Social Protection Table 8.2: Coping strategies by poverty status: 2015/16 All (%) Poor (%) Non-poor (%) Difference Used savings 36.5 32.2 38.3 *** Send children to relatives 0.9 1.1 0.9 Sold assets 2.5 2.6 2.4 Sold farmland 0.7 0.7 0.7 Rented farmland 0.9 1.0 0.9 Sold animals 10.5 13.4 9.3 *** Sold more crops 3.9 3.6 4.1 Worked more 14.0 13.5 14.2 HH member started work 1.0 1.5 0.9 ** Started business 4.7 2.4 5.6 *** Children worked 0.4 0.6 0.3 ** Migrated for work 4.4 3.8 4.6 Borrowed from relative 7.2 7.4 7.1 Borrowed from moneylender 1.9 1.6 2.0 Borrowed from formal institution 1.5 0.5 1.9 *** Help from church 2.4 3.4 2.0 *** Help from local NGO 0.2 0.5 0.1 *** Help from Intl. NGO 1.0 1.9 0.7 *** Help from government 2.5 4.7 1.6 *** Help from family member 20.4 21.9 19.8 ** Reduced food consumption 17.4 20.3 16.1 *** Consumed less preferred food 11.2 10.7 11.4 Reduced non-food consumption 16.3 16.8 16.0 Spiritual help 9.9 9.6 10.0 Other coping strategy 7.3 6.5 7.6 Source: Own calculations based on KIHBS 2015/16. Note: *** p<0.01, ** p<0.05, * p<0.1. 8.4 THE COVERAGE AND IMPACT OF SOCIAL components which may include: social assistance PROTECTION PROGRAMS through cash transfers to those who need them, 8.4.1 Social protection systems in Kenya especially children; benefits and support for people Social protection systems help individuals and of working age in case of maternity, disability, work households cope with shocks, find jobs, improve injury or for those without jobs; and pension coverage productivity, invest in the health and education of for the elderly. Assistance may be provided through their children, and protect the aging population. contributory social insurance, tax-funded social benefits, Social protection coverage is made up of several social assistance services, public works programs and 194 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Vulnerability, Shocks, and Social Protection other schemes guaranteeing basic income security Africa, 40 countries operated an unconditional cash (see, for example, World Bank 2015c). This section of transfer program in 2014, which is about twice as many the chapter has a narrower focus on social assistance as in 2010 (World Bank 2015c). Every African country programs that are defined as noncontributory benefits has at least one social safety net program. The average provided either in cash or in kind and intended to number of programs per country on the continent is support the poor or vulnerable. Other components of 16 – ranging from 2 in the Republic of Congo to 48 in social protection such as contributory social insurance Chad (World Bank, 2018). (pensions and health insurance) are important pillars of social protection, but often play a less important Even though the number of social safety net role in the welfare of the poorest households in programs has increased significantly, their coverage developing countries. is often limited and programs remain fragmented within countries. The combined coverage of programs Social safety net programs may have in Africa is less than 10 percent of the population. transformational effects over time. Some As part of the effort to enhance the efficiency and programs are directly targeted towards mitigating coordination of safety net programs, many countries are the immediate negative impacts of shocks such strengthening coordination among programs, and are as droughts, while others are aimed at changing investing in shared systems to reduce the duplication of structural characteristics (for example cash transfers that efforts and cost inefficiencies. Delivery platforms such are conditional on children remaining in school). These as social registries, management information systems, programs help to increase the chances of households and shared payment systems promote administrative escaping poverty, and then remaining non-poor in the cost savings and facilitate planning and coordination. future. The pathways through which these programs Social registries are currently used in 23 countries and affect outcomes include those at the individuals and are being developed in an additional 13 countries on household level, the local economy level, and the the continent (World Bank 2018c). macro economy level (Alderman and Yemtsov 2012). Kenya, like many other countries in sub-Saharan Unconditional cash transfer programs can be Africa, has expanded its social protection programs, expected to reduce current and future poverty in at but there is still a significant gap between coverage least two ways. First, the receipt of income transfers and needs. About 0.27 percent of the country’s GDP raises the disposable income of participant households was spent on social safety net programs in 2015.232 This and therefore alleviates consumption deficits. Second, was well below the average of 1.6 percent of GDP under favorable conditions, regular and reliable in low- and middle-income countries, as shown in transfers raise permanent household income leading Figure 8.15. Of the benchmark countries shown in the to an increase in human capital investment thus figure, only Tanzania had a lower overall expenditure raising the productivity of participant households level in 2015. The majority of Kenya’s spending on (Barrientos 2013; Fiszbein and Schady 2009). These social safety nets was made up of unconditional favorable conditions are inclusive growth and basic cash transfers, with a smaller share going to school service provision. feeding schemes, and a very small amount being allocated to fee waiver programs. Positive empirical evidence on the welfare effects of social protection programs has helped to spur strong There are currently four major public cash transfer support in these programs over the past decade. It is programs in Kenya. These are detailed in Table 8.3 estimated that nearly one billion individuals in low- and and include: Cash Transfer for OVC; OPCT Programme; middle-income countries are reached by antipoverty 232 This number excludes expenditure on unconditional food and in-kind transfer programs (Barrientos 2013). In sub-Saharan assistance. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 195 Vulnerability, Shocks, and Social Protection Figure 8.15: Expenditure on social safety nets: 2015 4 3.49 3.5 3 2,5 Percentage of GDP 2.12 2 1.8 1.6 1.5 1 0.80 0.57 0.5 0.27 0.07 0 Tanzania Kenya Uganda Ethiopia Low-middle SSA Rwanda South Africa income CCT UCT School feeding Public works Fee waivers Other SSN Source: The State of Social Safety Nets, World Bank 2015c. Note: CCT=conditional cash transfer; UCT=unconditional cash transfer; SSN=social safety nets. Table 8.3: Social Protection Programs in Kenya Core objectives Amount Targeting Coverage • Retention of OVC within families / KSh 4,000 paid • PMT (Extremely poor; Cash Transfer communities every two OVCs; HH not enrolled in months another CT program) National for OVC • Human capital development • Civil registration • Poverty reduction among the KSh 4,000 paid • 65 years and above elderly population every two • Poor and vulnerable months • HH members not enrolled OPCT Programme National in other CT program, not receiving pension, not employed • Strengthen capacities and KSh 4,000 paid • HH member with severe livelihoods of households whose every two disability Persons with Severe members have disabilities months • Poor Disability Cash • Poverty reduction of households • HH members not enrolled National Transfer (PWSD) whose members have disabilities in other CT program, not receiving pension, not employed • Reduce extreme hunger and KSh 5,400 paid • PMT (extremely poor) Mandera, HSNP vulnerability every two • Community-based Marsabit, months targeting Turkana, Wajir. Persons with Severe Disability Cash Transfer; HSNP.233 established to provide a common operating framework The fact that the programs were originally operated for the government’s cash transfer programs. As part of independently of one another by different departments efforts to develop a harmonized social safety net, the and ministries led to a lack of coordination. In 2013, Social Protection Secretariat (SPS) unified the social the Kenya NSNP was established as part of the assistance program information in a single registry. The government’s initiatives to improve social protection objective of this single registry system is to consolidate delivery in the country. In particular, the NSNP was information from the different cash transfer programs in a single platform. 233 The Urban Food Subsidy program (UFS) is an additional public cash transfer program in the country, but is not included in this chapter. The program currently covers around 10 000 households in the Mombasa constituencies of Mvita, Likoni, Changamwe and Kisauni. 196 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Vulnerability, Shocks, and Social Protection 8.4.2 Coverage of the four main cash transfer programs to regions with high levels of poverty, or by programs using eligibility criteria that disproportionally benefit While the number of beneficiary households poor household. increased from about 240 000 in 2013 to almost 770 000 in 2016, coverage remains limited. This Households in counties in the Northern and corresponds to a national coverage rate of these four Eastern parts of Kenya receive proportionally programs of about 6.4 percent of households. The more assistance with respect to their population OPCT is currently the largest program in terms of size. Figure 8.17 displays the spatial distribution of household coverage, reaching approximately 350 000 households benefitting from any of these programs in households in 2016. This reflects significant growth 2016. The map on the left shows the absolute number from the coverage of about 50 000 households in of beneficiary households per county, and the map 2013. Coverage of the OVC grew more slowly than on the right shows the relative share of households the OPCT, and stood at around 260 000 households in per county receiving transfers. Population shares for 2016, making it the second largest program in terms the second map were calculated using the county of coverage. It is followed the HSNP and the PWSD population numbers in KIHBS 2015/16. which had 2016 coverage levels of 100 000 and 50 000 beneficiary households, respectively. The four counties in which the HSNP operates are also the four with the highest number of recipients of Figure 8.16: Number of households receiving cash transfers: 2013 to 2016 all assistance. According to the single registry dataset, over 57 000 households in Turkana were cash transfer 800 beneficiaries. The corresponding numbers in Mandera, 700 Wajir and Marsabit were 38 500, 30 300 and 29 800, 600 respectively. The counties with the fewest absolute 500 number of recipients were Lamu, Laikipia and Isiolo Thousands 400 which all had coverage of under 7 000 households. 300 200 On average a little over 6 percent of Kenyan households received cash benefits from one of the 100 four programs. However, the share of beneficiary 0 2013 2014 2015 2016 households varies significantly across counties. Just OVC OPCT HSNP PWSD Total under half of households in Marsabit were registered Source: Own calculations from Kenya’s Single Registry for Social Protection. beneficiaries in 2016, 44 percent of households were beneficiaries in Wajir, 35 percent in Mandera, and just under one quarter in Turkana. Fewer than 3 percent of If social assistance programs aim to alleviate poverty, households in Nairobi and Mombasa were beneficiaries then they should be targeted at poor households. of one of the cash transfer programs. Almost 22 500 An efficient program satisfactorily solves the trade- households in Nairobi were cash transfer beneficiaries, off between minimizing exclusion errors (poor but its large population size meant that this translated households that are not beneficiaries) and inclusion into a very low coverage rate.234 In general, as shown errors (non-poor households that are beneficiaries). 234 The share of female beneficiaries of the four cash transfer programs by As discussed in the previous chapters, poverty is not county is shown in Figure H.3 in the appendix. Overall, 64 percent of registered beneficiaries in 2016 were female. This ranged from around evenly distributed within Kenya. Therefore, an efficient three quarters of beneficiaries in Kwale, Vihiga, Kitui and Makueni, to 40 social safety net design would be expected to result in percent and 48 percent in Wajir and Marsabit, respectively. The average registered recipient was 57 years old, and this was driven up by the stronger program prevalence in regions with relatively average age of OPCT recipients of 67 years old. The share of female- headed households receiving grants as measured in KIHBS 2015/16 is higher poverty rates. This can be achieved by confining shown in the final row of Table 8.4. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 197 Vulnerability, Shocks, and Social Protection in the second panel of Figure 8.17, the within-county at 9 percent of households, but is under 5 percent for coverage rates are highest in the north and east of the 42 of the 47 counties. The variation in OPCT coverage country, which is also where poverty rates are highest. rates is a little smaller than OVC coverage rates, though the difference in OPCT coverage between Samburu at The distribution of OPCT beneficiary coverage is 9.5 percent and Nairobi at 0.7 percent is considerable. generally quite consistent at the county level, while The largest spatial variation is observed for the HSNP, OVC coverage rates are relatively higher in the eastern which currently only operates in Turkana (16 percent), parts of Kenya. Figure 8.18 shows the disaggregated Marsabit (33 percent), Wajir (27 percent) and Mandera share of beneficiary households for each of the four (20 percent). transfer programs. OVC coverage is highest in Isiolo, Box 8.4: Findings from impact evaluations of the OVC and the HSNP programs The utilization of RCTs for evaluation purposes has become a common feature in the implementation of cash transfer programs in Sub-Saharan Africa. If implemented appropriately, RCTs provide rigorous evaluation results with causal inference that can help to inform policy decisions. Yet, problems with the randomization of programs, changes in policy design, and attrition can contaminate the power of RCTs to produce reliable estimates. In Kenya, the OVC and the HSNP were quantitatively evaluated through RCTs, which has led to a substantial amount of empirical evidence on the effects of cash transfer programs. The OVC was among the first public cash transfers in Sub-Saharan Africa that was formally approved after the evaluation of the pilot program between 2007 and 2009. In a baseline and a follow-up survey, 2 255 households were interviewed covering a broad range of welfare and human capital indicators. After two years of program implementation, the OVC transfer program was found to increase beneficiaries’ consumption expenditures and to reduce poverty headcount levels by 13 percentage points. The program was also found to have increased food expenditure and food diversity, and to have had a positive impact on secondary school enrolment (Ward et al. 2010). In addition to the evaluation report, a number of academic research articles have used the OVC impact evaluation data to test program impacts on human capital (Kenya CT-OVC Evaluation Team 2012), health behavior (Handa et al. 2014), spill-over effects of cash transfers to non-beneficiaries (Thome et al. 2013), productive and labor market effects (Asfaw et al. 2014), and food security and coping strategies (Tiwari et al. 2016). The HSNP was evaluated between 2009 and 2012 in three survey waves capturing the welfare and human capital developments of approximately 2 500 households (Merttens et al. 2013). The results indicate positive impacts of the program on reducing poverty and increasing consumption, particularly food consumption. There was also a positive impact on the number of livestock owned by recipient households. The program impacts were evaluated during an exceptionally severe drought, which suggests that transfers helped households to compensate for losses and prevented them from applying disruptive coping strategies in response to the drought such as sales of productive assets. The impacts on human capital were small (health expenditures) or absent (education and nutritional status of children). As both programs transferred a constant sum to all beneficiary households during the evaluation period, smaller and poorer households were found to experience more significant impacts in general. Despite these positive findings, it should be noted that the implementation of the evaluations was not problem-free, which could have implications for the validity of the findings. In both surveys, significant attrition rates may have affected the external validity of findings. In addition, program randomization was only implemented in a few districts (OVC), which leads to a lower power of estimates. In addition, the evaluation plan of the HSNP was not followed strictly, as eight sub-locations were excluded from the evaluation in the second follow up. 198 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Vulnerability, Shocks, and Social Protection Figure 8.17: Coverage and share of beneficiaries by county: 2016 Beneficiaries Share of Beneficiary HHs Number of bene ciary HHs Share of HHs 40 001 to 60 000 40% to 50% 30 001 to 40 000 30% to 39.9% 20 001 to 30 000 20% to 29.9% 15 001 to 20 000 15% to 19.9% 10 000 to 15 000 10% to 14.9% <10 000 <10% Source: Own calculations from Kenya’s Single Registry for Social Protection and KIHBS 2015/16. Figure 8.18: Share of beneficiary households by county and program: 2016 Share of OVC beneficiary HHs Share of OPCT beneficiary HHs Share of HHs Share of HHs 30% to 40% 30% to 40% 20% to 29.9% 20% to 29.9% 15% to 19.9% 15% to 19.9% 10% to 14.9% 10% to 14.9% 5% to 9.9% 5% to 9.9% <2% 2% to 4.9% <2% Share of HSNP beneficiary HHs Share of PWSD beneficiary HHs Share of HHs Share of HHs 30% to 40% 30% to 40% 20% to 29.9% 20% to 29.9% 15% to 19.9% 15% to 19.9% 10% to 14.9% 10% to 14.9% 5% to 9.9% 5% to 9.9% <2% 2% to 4.9% <2% Source: Own calculations from Kenya’s Single Registry for Social Protection and KIHBS 2015/16. Note: the HSNP covers Turkana, Marsabit, Wajir and Mandera only. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 199 Vulnerability, Shocks, and Social Protection 8.4.3 A profile of cash transfer beneficiaries There are significant differences across a wide range Although coverage rates of the four main cash of characteristics between beneficiary households, transfer programs are lower in KIHBS 2015/16 than poor, non-beneficiary households, and non-poor in the single registry dataset, there is enough power non-beneficiary households. Simply comparing to allow us to profile recipient households and to adult equivalent levels of consumption would give estimate the impacts of the transfers.235 Figure 8.19 a somewhat misleading picture of the differences shows CDFs of consumption as a percentage of the between these groups of households, in part, because poverty line for recipient households of the HSNP, as mentioned, this measure of welfare incorporates OVC, OPCT, and non-recipients. Households receiving the receipt of the grant. Table 8.4 shows that there the HSNP are far poorer, on average than all other are striking differences in the kinds of activities these households in the figure. About 65 percent of HSNP- groups were engaged in, the levels of assets, and the receiving households were below the poverty line, educational attainment levels of household heads. even after the cash transfer is taken into account. This is in contrast to a poverty rate of about 50 percent Households that received one of the four main for households receiving the OPCT and households cash transfers were more likely to be engaged receiving the OVC transfer. The poverty rate for non- primarily in agriculture than both poor and non- poor households that did not receive grants. 70 Figure 8.19: CDFs of consumption by cash transfer program percent of beneficiary households were engaged in 1 agriculture, while 22 percent were in services. This is .9 less than half of the proportion of the non-poor, non- .8 beneficiary households that were employed in services. .7 Share of population .6 Unsurprisingly, given the results that were presented .5 previously in the chapter, the large majority of .4 beneficiary households lived in rural areas (87 percent). .3 .2 .1 Beneficiary households have, on average, higher 0 consumption expenditure levels than poor, non- 0 50 100 150 200 250 300 350 400 450 500 beneficiary households, but have lower levels of Consumption as percentage of the poverty line HSNP recipients OVC recipients asset accumulation. The difference in consumption OPCT recipients Non-recipients levels is generated by the fact that household welfare Source: Own calculations from KIHBS 2015/16. is measured inclusive of grant receipt, and because 57 percent of beneficiary households are poor, whereas recipient households was about 25 percent, indicating 100 percent of households in column 2 are poor. that there are possibly large coverage gaps that could Focusing on adult equivalent consumption expenditure be addressed. The median recipient household has a alone paints a misleading picture of the overall welfare consumption level that is at 88 percent of the poverty of beneficiary households. One example of this is the line, compared to the median poor household that is at fact that the asset index for beneficiary households 76 percent of the poverty line. This difference is largely is just over half of what it is for poor, non-beneficiary driven by the fact that these welfare levels are inclusive households, and less than a third of what it is for non- of grant income. poor, non-beneficiary households.236 236 The asset index used in this table is a share index which is calculated by first multiplying an indicator variable (for example: household owns a fridge) by the proportion of households that do not own the asset (for example: proportion of households that own a fridge). This ensures that 235 Coverage of all four programs in the KIHBS 2015/16 dataset is about less common assets receive a relatively higher weight in the index. These 2.2 percent of households in Kenya. This lower rate of coverage is products are then summed over each component at the household unsurprising, given the geographic concentration of the programs level to generate the share index. The components of the index are: compared to the sampling methodology of KIHBS, and given that the refrigerator, washing machine, microwave, kettle, computer, radio, KIHBS weights are not stratified on grant receipt. bicycle, car, cellphone, television, sofa, and kerosene stove. 200 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Vulnerability, Shocks, and Social Protection Table 8.4: Profile of beneficiary households versus non-beneficiary households (by poverty status) 1 2 3 1 vs 2 1 vs 3 Beneficiary Poor, non- Non-poor, household beneficiary non- household beneficiary household Employment Sector Agriculture 69.5% 56.4% 37.6% *** *** Manufacturing 6.4% 4.8% 6.7% Services 22.1% 30.1% 48.5% *** *** Construction 2.0% 8.7% 7.2% *** *** Location Household Rural 86.6% 72.1% 59.7% *** *** Welfare Poor 57.3% 100% 0% *** *** AEQ consumption (KSh) 3 852 2 913 9 717 *** *** Asset index 0.5 0.9 1.7 *** *** Composition Household size 4.9 5.2 3.5 *** *** 1 or more employed 22.6% 47.9% 53.9% *** *** Education No education 67.1% 25.5% 8.1% *** *** Household head Primary education 26.0% 53.8% 42.6% *** *** Secondary education 6.2% 20.2% 42.1% *** *** Tertiary education 0.7% 0.4% 7.2% *** Gender Female 54.5% 34.8% 30.9% *** *** Source: Own calculations from KIHBS 2015/16. Note: *** p<0.01, ** p<0.05, * p<0.1. The limited connection to labor markets amongst beneficiary households. Both of these categories had beneficiary households is very clear. Only 23 percent significantly larger households than non-poor, non- of grant-receiving households had at least one resident beneficiary households (3.5 people, on average). member who was employed. This is in contrast to 48 percent in poor, non-beneficiary households, and 54 Another non-monetary dimension in which percent in non-poor, non-beneficiary households. beneficiary households are far worse off than other Much of this difference is driven by the fact that households is in the educational attainment of the households that receive the OPCT tend to be older, household head. More than two thirds of beneficiary and contain more members who no longer work, or households were headed by someone who reports not are unable to work. The average beneficiary household having completed any education. One quarter of these size was 4.9, and this was slightly lower than poor, non- households have a head with primary education, and KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 201 Vulnerability, Shocks, and Social Protection Box 8.5: Evaluating the impacts of Kenya’s cash transfer programs using cross-sectional data and propensity score matching There is no experimental data contained in KIHBS 2015/16, and there are not repeated observations of the same individuals or households over time. As such, a suitable cross-sectional estimator needs to be found. This chapter uses a propensity score matching (PSM) estimator to uncover the welfare effects of Kenya’s cash transfer programs. There are three outcomes of interest: 1) Whether in households with school-aged members all of the children are enrolled in school, 2) Whether in households with school-aged members none of the children are working, 3) Whether a household is food secure. The two central assumptions that need to be met for this estimation strategy to be credible are the existence of unconfoundedness, given the propensity score (unobserved factors are not influencing selection), and that there is common support over the propensity scores for both beneficiary (treated) and non-beneficiary (control) households. The average treatment effect on the treated (ATT) is defined using potential outcomes notation as follows: where is the dummy variable indicating household beneficiary status, is the outcome of interest for household i when = 1, and is the potential outcome of the same household had it not been a beneficiary of one of the grants. The vector of observable characteristics contains a set of variables that influence both grant beneficiary status and household welfare, and are used to estimate an ATT in which and are equal, conditional on the propensity score : In this chapter a lasso shooting algorithm is used to select the components of by choosing the union of variables that are significant predictors of outcome and treatment. Once the propensity scores are obtained, the ATT is estimated using nearest neighbor matching, population weights, and robust Abadie- Imbens standard errors. just 6 percent have a head who has a secondary level of that the KIHBS 2015/16 is a single cross-section, and that education. Among poor, non-beneficiary households, assignment to grant beneficiary status is not random. one quarter have a head with no education, while a little As such, the results should be interpreted with care. over half have a head with a primary level of education. This is in stark contrast to non-poor, non-beneficiary The cash transfer programs had a significant and households. Only 8 percent of these households have positive impact on child enrolment, and this effect a head who reports having no education, while 43 was particularly strong for the OVC grant. Figure 8.20 percent have a primary education, 42 percent have shows the PSM estimates of the effect that three of the a secondary education, and 7 percent have a tertiary four main cash transfer programs had on the probability education level. that all school-aged children in the household were enrolled.237 The sample used in the estimation of these 8.4.4 The impact of cash transfer receipt on results was restricted to households that contained household welfare children who were within the compulsory schooling The next part of the chapter attempts to evaluate age range of 6 to 14 years old. the impact of Kenya’s cash transfer programs on a number of welfare outcomes. This exercise is not 237 Effects of the PWSD cash transfer are not estimated due to the small number of households reporting having received this grant in the KIHBS intended to be an exhaustive impact evaluation, given 2015/16 data. 202 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Vulnerability, Shocks, and Social Protection Figure 8.20: The impact of grant receipt on the probability impact on increasing the probability that no child in the that all school-aged children in the household are enrolled household was working. Once again, the overall impact of grant receipt lies in between the impacts of the individual programs, and stands at almost 4 percentage points. The other three cash transfer programs all had positive impacts that were statistically significant at the 1 percent level, with OPCT receipt having the largest effect on increasing the probability that no child in the household was working, or close to 7 percentage points. 0 1 2 3 4 5 Figure 8.21: The impact of grant receipt on the probability Percentage point impact that no school-aged child in the household is working Any grant receipt HSNP receipt OVC receipt OPCT receipt Source: Own calculations from KIHBS 2015/16. Note: Households that were beneficiaries of one of the cash transfer programs were less likely to contain school- aged children who were working. The variable of interest in Figure 8.21 is the probability that no school-aged child in a household was reported to be working in the 2015/16 KIHBS. A positive effect in this framework corresponds to a higher likelihood that no child in the household is working. 93.5 percent of households containing school-aged children had all of those children enrolled in 2015/16. This enrolment rate is consistent with the findings that were presented in Chapter 6. Nevertheless, it appears that grant receipt was enough to increase these 0 1 2 3 4 5 6 7 8 Percentage point impact already high rates. The impact of a household being Any grant receipt HSNP receipt OVC receipt OPCT receipt the beneficiary of any of the cash transfer programs Source: Own calculations from KIHBS 2015/16. was an increase in the probability of all children being enrolled of about 1.4 percentage points.238 As expected, the program with the largest impact was the OVC cash The final outcome variable of interest is the transfer, which had a positive effect of 3 percentage probability that a household was food secure – with points. The HSNP also had a small positive effect of the HSNP reflecting a small but positive impact. The around 1 percentage point, while the OPCT’s effect was analysis underlying the results in Figure 8.22 is based close to zero, with the 95 percent confidence interval only on data from the four counties in which the HSNP including zero. was operating.239 The definition of food security used in this chapter is not based on caloric intake or on food The unconditional proportion of households that did expenditure, but is rather based on a number of self- not contain a school-age child who was working was reported food adequacy questions in the KIHBS 2015/16 87 percent. The positive impacts of the cash transfer household questionnaire. A household is defined as programs in the figure therefore had a significant food insecure if, in the last 12 months before being interviewed, members missed meals because of a lack 238 The treated group in the “any grant receipt” results includes all households with school-aged children that received a cash transfer. of money/resources, or the household ran out of food The corresponding control group was all households with school-aged because of a lack of money/resources, or household children that did not receive a cash transfer. For the program-specific effects (for example the OVC), the treated group was all households members were hungry or did not eat at all because of with school-aged children who received the OVC, and not any other cash transfer. The corresponding control group was all households a lack of money/resources, or any household members with school-aged children that did not receive any cash transfer. Practically, this meant that households with school-aged children who went without food for a whole day because of a lack of received the HSNP, OPCT or PWSD cash transfers were excluded from the control group, in order to limit the confounding effects that these money/resources. other programs may have had on the outcome variable. This logic was extended to the OPCT and HSNP effects as required. 239 These are Turkana, Marsabit, Wajir and Mandera. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 203 Vulnerability, Shocks, and Social Protection Figure 8.22: The impact of grant receipt on the probability rate in the HSNP counties is only 14 percent. Given that a household is food secure: HSNP counties only this extremely low level of food security, there is a moderately positive impact of the HSNP on food security in the four northern counties. On average, receipt of the HSNP increased the probability that a household would self-report as being food secure by about 2.5 percentage points. Taken together, the results of this section of the chapter suggest that social assistance programs in Kenya are well targeted, and that they are having 0 1 2 3 4 positive impacts on a number of measures. However, Percentage point impact Any grant receipt HSNP receipt overall coverage remains low relative to existing needs. The effort that has been made to coordinate and Source: Own calculations from KIHBS 2015/16. harmonize social protection programs, combined with Households in the four HSNP-receiving counties the creation of a registry of beneficiary households have much lower rates of food security than the means that the country is well placed to expand other 43 counties. In all of Kenya except for the four assistance to vulnerable households, which would HSNP counties, 52 percent of households self-report benefit greatly from this potential expansion. that they are food secure, while the corresponding 204 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead APPENDICES Appendices APPENDIX A: CHAPTER 1 ADDITIONAL MATERIALS A.1. Tables Table A.1: Poverty trajectory simulation, sectoral and non-sectoral growth GDP sectoral growth simulation Overall GDP growth simulation  Year Poverty rate, Poverty rate, Poverty rate, Poverty rate, Poverty rate, Poverty rate, US$1.20 a day US$1.90 a day US$3.20 a day US$1.20 a day US$1.90 a day US$3.20 a day 2005 21.0 43.7 69.2 21.0 43.7 69.2 2006 20.2 42.9 68.9 20.4 43.1 69.0 2007 18.9 41.8 68.5 19.0 42.1 68.6 2008 19.7 42.5 68.6 19.0 42.1 68.6 2009 20.0 42.7 68.5 18.7 41.6 68.4 2010 18.0 40.9 67.8 17.4 40.5 67.8 2011 17.2 40.3 67.4 16.5 39.6 67.3 2012 16.6 39.7 67.1 15.8 39.1 67.0 2013 15.5 38.8 66.7 14.9 38.4 66.7 2014 14.6 37.9 66.4 14.3 37.4 66.4 2015 13.6 36.9 65.9 13.6 36.7 66.1 Source: Author’s calculations based on KIHBS. A.2. Drivers of growth – Diagnostic Figure A.1: TFP growth was a key driver of GDP growth Output growth Estimated TFP growth 10 10 8 8 Growth rate (percentage points) Growth rate (percentage points) 6 6 4 4 2 2 0 0 2000 2002 2004 2006 2008 2010 2000 2002 2004 2006 2008 2010 -2 -2 -4 -4 Source: KNBS and World Bank. 206 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices Figure A.2: As growth in capital accelerated, growth of labor moderated Capital growth Labor growth 10 10 8 8 Growth rate (percentage points) Growth rate (percentage points) 6 6 4 4 2 2 0 0 2000 2002 2004 2006 2008 2010 2000 2002 2004 2006 2008 2010 -2 -2 -4 -4 Source: KNBS and World Bank. Figure A.3: Stagnating human capital growth resulted in a moderation of human capital per unit of labor Human capital growth Level of human capital per labor growth 10 10 8 8 Growth rate (percentage points) Growth rate (percentage points) 6 6 4 4 2 2 0 0 2000 2002 2004 2006 2008 2010 2000 2002 2004 2006 2008 2010 -2 -2 -4 -4 Source: KNBS and World Bank. Figure A.4: The increase in labor force resulted in increasing unemployment and declining labor force participation 15 0.72 0.70 Unemployment rate, percent 10 0.68 Participation rate 0.66 5 0.64 0.62 0 0.60 2000 2002 2004 2006 2008 2010 Unemployment rate Participation rate Source: KNBS and World Bank. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 207 Appendices A.3. An analysis of devolution and fiscal transfers in Kenya 1. The context and legal framework of devolution in decision-making and public service delivery, while Kenya’s transition to a devolved system of ensuring the equitable sharing of national resources. government began with the promulgation of the Constitution of Kenya (2010) on August 27, 2010. Strong economic rationale underpins these The Constitution created a new sub-national level objectives. Decentralization allows for improvement of government including 47 counties, each with in service delivery through better preference an elected governor, county executive, and county matching, as local governments have an informational assembly. County governments are interdependent advantage over the national government in terms of with the national government, which consists of the local household preferences and demand for public National Executive, Parliament, and Judiciary. Each services. As local governments internalize the costs county has a voice in the National Parliament through and benefits of local public service provision, they the Senate, an upper house that includes within its improve the delivery of services. Additionally, citizens membership 47 directly elected county representatives, exercise better control over their locally elected as well as within the National Assembly. representatives, whom they are better able to identify and hold accountable through elections. Powers granted by Chapter 11 of the Constitution give county governments the power to govern Under Kenya’s devolved government, key functions themselves, raise revenues, make local laws and were transferred from the national to county elect local officials. The Constitution recognizes the governments through the fourth schedule of the right of communities to manage their own affairs, and constitution. Amongst the key functions transferred gives powers of self-governance at a local level to to county governments are agriculture, county health enhance the participation of individuals in decision- services, provision of county transport services, pre- making. County assemblies make laws necessary for primary education, water and sanitation, control of the effective performance of the county government, pollution and conservation of the environment, and and exercise oversight over the county executive. This development of trade. A transition authority was constitutes a major reorganization of governance established to transfer functions from the national to from the previous centrally-led government, giving county governments, with the transfer beginning in counties significant autonomy over their local needs February 2013. and service delivery priorities, while at the same time increasing local accountability. However, the national Under Article 187 of the Constitution, arrangements government continues to maintain a key policy and have been put in place to provide resources regulatory role. necessary for county governments to perform devolved functions. The Constitution provides that if The objectives of devolution are outlined in a function or power is transferred from a government Article 174 of the Constitution. The key political at one level to a government at another level, then objectives of the Constitution are the separation of arrangements shall be put in place to ensure that power between national and county governments, the resources necessary for the performance of the and decentralization of state organs, while ensuring function or exercise of the power are transferred in checks and balances for the accountable exercise of line with the “finance follows function” principle. The power. Further, the Constitution recognizes cultural Constitution provides for a minimum unconditional diversity, and provides for the protection of minorities transfer of 15 percent of shareable revenues to counties, and marginalized groups. The economic objectives of and the allocation of these revenues across counties is the Constitution are autonomy and local participation determined by a revenue allocation formula. 208 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices Additionally, regulatory and intergovernmental The devolution program timeline was accelerated bodies have been established to ensure oversight faster than originally envisaged under intense over devolution. Amongst key bodies created is the bargaining amongst stakeholders. The first transfer of CRA and the Office of the Controller of Budget (OCOB). functions was performed in February 2013, where a The CRA is mandated to make recommendations on suite of functions formerly performed by former local the basis for sharing of revenues between the national authorities were transferred to the counties. However, government and county governments, as well as the originally envisaged three-year period of transfer, the equitable sharing of revenues across county during which functions were to be transferred to governments, while the OCOB is established to counties in line with growth of capacity, was truncated. oversee and report on implementation of the budgets The first full year of the revenue- sharing cycle was of both the national and county governments. completed in 2013/14. Intergovernmental economic and technical committees are also established with key oversight The rapid pace of devolution brought with it roles across levels of government. implementation challenges due to capacity constraints and coordination difficulties amongst Despite its attractive features, devolution can worsen different levels of government. An insufficient policy economic outcomes. The provision of public services and legal framework to guide the implementation may be dependent on economies of scale, whereas of the constitution resulted in slow implementation. devolution to small-scale local governments can Additionally, an unclear unbundling of devolved increase costs and lower efficiency. Further, functions functions resulted in gaps in service provision in that have externalities across multiple counties may some cases, and in duplication of efforts in other result in inefficient allocation of resources when cases. Further, insufficient human capital was a major devolved, and therefore are better executed at the constraint in the provision of services, extenuated national level. In addition, devolution can obstruct the by the major transitional challenge in reorganization redistribution role of the central government. of the existing civil service to the new government structure. Citizens were also ill-informed about II. Historical background of devolution their ability to participate and contribute to local Elections in 2013 marked the official launch governance, and existing legislation was weak and of devolution with the selection of county ineffective in promoting public participation. officials, and the national senate. This was a complex undertaking that included election of In their first full year of financing, county the president, national assembly, the senate, governments received funding from transfers and women’s representatives, governors, and county assemblies. The transition to a devolved system own-source revenues. In the first full year of funding, of government was guided by the Taskforce on an equitable share of KSh 190 billion was transferred Devolved Government, which was established after to county governments, in addition to an equalization promulgation of the Constitution and guided the fund allocation of KSh 4.3 billion.1 This represented 21 formulation of devolution laws. Subsequently, key percent of shareable government revenues during the devolution laws were enacted, county government year. An additional total of KSh 16.6 billion was also structures were operationalized, and functions and transferred in donor-funded conditional grants. Own- resources were allocated to county governments. source revenues during the year were KSh 26 billion. 1 The Equalization Fund was officially launched in March 2016, to allow establishment of the Fund’s guidelines and administrative structures. Prior to this date there was no disbursement, and allocations were deposited in an account held with the Central Bank of Kenya. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 209 Appendices Across counties, the horizontal sharing of revenues the Constitution. Transfers in 2013/14 amounted was guided by a formula that included parameters to 21 percent of ordinary government revenues, that were a proxy for the cost of service delivery 6 percentage points higher than the mandated (population and land area), fixed administrative costs minimum of 15 percent of shareable government (basic equal share), and poverty (the poverty gap). revenues. In subsequent years, transfers have all Nairobi received the highest horizontal allocation (KSh maintained a level above 20 percent of annual 9.5 billion), while Lamu received the lowest allocation government revenues (Figure A.5).2 (KSh 1.5 billion). The equalization formula has had the intended The formulae for the equitable allocation of effect, with under-developed regions of the country resources amongst county governments, as well receiving significant allocations. Turkana, which as the vertical division of revenues between has the highest poverty rate in Kenya (79.4 percent) the national and county governments, have had an equitable share allocation of KSh 11 billion subsequently been revised as mandated by the in 2016/17, which is second only to the allocation Constitution. The first revision of criteria for both to Nairobi.3Mandera, which has the second-highest horizontal and vertical sharing of revenues was poverty rate (77.6 percent) also has a significant mandated after three years, occurring in 2015. The allocation of KSh 9.7 billion, which ranked fourth in revision of the horizontal criteria did not include major the country as of 2016/17. Allocation of these funds changes, with the inclusion of a development factor is expected to have a significant impact on service and adjustment of weights to existing parameters. delivery and living standards. Subsequent revisions are mandated every three years for the vertical criteria and every five years for the Further, the Equalization Fund has provided horizontal criteria. additional transfers to marginalized areas. The marginalization policy created a county development III. Performance of devolution in Kenya index, which was a composite index constructed Fiscal transfers to county governments have from indicators measuring health care, education, occurred more rapidly than envisaged under infrastructure and poverty within a county. Figure A.5: County allocation of ordinary government Figure A.6: Transfers to county governments, 2016–17 revenues 24 16,000 14,000 22 12,000 Percent of government revenues Millions KSh 20 10,000 8,000 18 6,000 16 4,000 2,000 14 - Nairobi Mandera Bungoma Kitui Meru Makueni Mombasa Marsabit Nyeri Kericho West Pokot Embu Kirinyaga Taita Isiolo Narok 12 10 2013/14 2014/15 2015/16 2016/17* Conditional allocation Equitable share Source: National Treasury, CRA. Source: OCOB. 2 This calculation is based on current year ordinary government revenues. The Constitution mandates a base of last audited government revenues. As a percentage of audited government revenues, transfers have maintained a level above 30 percent of total government revenues. 3 While Nairobi still as the highest allocation, its share decreased by 4.7 percentage points between 2010/11 and 2013/14. 210 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices Figure A.7: Share of transfers to counties 12 10 8 6 4 Percent 2 0 -2 -4 Nairobi Kakamega Nakuru Kiambu Kisii Wajir Meru Garissa Migori Homa Bay Marsabit Trans Nzoia Nyeri Bomet Kericho Kajiado Nyandarua Nyamira Vihiga Kirinyaga Taita Taveta Tharaka-Nithi Lamu Narok -6 Change in share, 2010/11 to... Source: CRA, OCOB. Additionally, the index was complemented with an Figure A.8: Change in allocation of transfers by share of urban population analysis on historical and legislative discrimination. 3 Based on the index, fourteen counties were identified 2 as marginalized, which were concentrated in northern Change in share of transfers, 2010/11 1 and eastern parts of the country. to 2013/14 (Percent) 0 -1 However, the redistribution of revenues has led to a relative shift in resources away from urban -2 areas. Areas with high shares of rural population are -3 receiving higher transfers (Figure A.8), and in turn are -4 allocating a higher share of resources to development -5 expenditure, which should stimulate regional growth -6 - 20 40 60 80 100 and lead to economic convergence over time (Figure Share of urban population A.9). Lower transfers to highly urbanized areas are an Source: CRA, OCOB. incentive to grow own-source revenues from already established revenue bases, although the reallocation Figure A.9: Development expenditure share of total expenditure of transfers can have the adverse effect of lowering the 70 quality of service delivery. For example, development 60 expenditures in Nairobi and Nakuru are less than 20 percent of total expenditures, whereas development 50 expenditures in Turkana and Mandera are above 50 40 Percent percent of total expenditures. 30 20 Moreover, absorption rates of transfers vary 10 significantly across counties. Initially low absorption rates4 of development budgets have shown a trend 0 Taita Taveta Tharaka-Nithi Nyamira Lamu Kisii Meru Mombasa Murang’a West Pokot Busia Kwale Embu Makueni Tana River Narok Kili of improvement over time, while absorption rates of recurrent budgets have remained consistently high Source: OCOB. 4 Absorption rates are defined as the ratio of actual expenditures relative to approved budget amounts. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 211 Appendices (Figure A.10) Low development budget absorption indicating an increase in the dependence of counties rates were the result of an array of factors, including on transfers to finance county activities. Own-source low technical and implementation capacity, unrealistic revenue collections vary significantly across counties, budgeting and delays in the transfer of funds relative reflecting in part the priorities of the previous to expected schedules. government structure: Nairobi finances 44 percent of its expenditures with own revenues, while West Pokot, Additionally, recurrent expenditure accounts for Turkana, Garissa, Wajir, Tana River and Mandera are able over two-thirds of total expenditure, with the to finance less than 2 percent of their expenditures largest share of recurrent expenditure accounted with own-source revenues (Figure A.14). There is also a for by personnel costs. Wages account on average for wide disparity in the capacity of counties to raise own over 60 percent of recurrent expenditure in counties revenues, with seven counties showing a decrease (Figure A.11) with some counties reaching levels as in average own revenue collections since 2013/14 high as three-quarters of their recurrent spending on (Figure A.15). wages. Additionally, counties have grown their wage bills on average by 19 percent annually between 2013 Devolution provides favorable conditions to and 2017, with Nyamira and Turkana growing wages improve public service delivery outcomes. Trends by as much as 70 percent and 90 percent per annum in public service delivery outcomes have improved over this period (Figure A.13) . While significant over the last ten years, with access to better water wage bill growth is indicative of increasing capacity, and sanitation facilities, electricity and education and such rapid increases may lead to crowding out of health facilities. Additionally, the overall poverty rate in development projects. the country has decreased by 9.8 percentage points, from 45.9 percent in 2005 to 36.1 percent in 2015. Own-source revenues have not increased Local decision-making in a devolved government substantially with devolution. County own-source gives counties a foundation to better living standards revenues as a share of actual expenditures show a of households across the country. decreasing trend over time (Figure A.12), thereby Figure A.10: Absorption rates of county budgets Figure A.11: Personnel costs by county 100 80 90 70 Percent recurrent expenditure 80 60 70 50 60 40 Percent 50 30 40 20 30 10 20 0 10 Tharaka Nithi Machakos Kakamega Embu Kirinyaga Taita/Taveta Homa Bay Nyeri Kiambu Baringo Siaya Nyamira Isiolo Meru Kericho Mombasa Nyandarua Bungoma Turkana Kwale Trans Nzoia Lamu Marsabit 0 2013/14 2014/15 2015/16 2016/17 Recurrent Development Total Source: OCOB. Source: OCOB. 212 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices Figure A.12: Share of county own revenues Figure A.13: Cumulative annual growth rate of personnel costs 18 100 16 90 80 Percent, 2013/14 to 2016/17 14 70 Percent of actual expenditure 12 60 50 10 40 8 30 6 20 10 4 0 Nyamira Tana River Wajir Marsabit Bomet Machakos Taita/Taveta Migori Busia Trans Nzoia Makueni Vihiga West Pokot Nakuru Baringo Uasin Gishu Kitui Kakamega Murang’a Tharaka Nithi Embu Laikipia Nyeri Narok 2 0 2013/14 2014/15 2015/16 2016/17 Source: OCOB. Source: OCOB. Figure A.14: Own revenues as a share of actual Figure A.15: Average annual increase in county expenditure own-source revenues 50 80 45 70 40 60 35 30 50 25 40 20 30 15 20 10 5 10 0 0 -10 Nairobi Kisumu Machakos Nyeri Average Kericho Murang’a Embu Baringo Migori Nandi Busia Lamu Kwale Kisii Tharaka Nithi Vihiga Elgeyo/Marakwet West Pokot Garissa Tana River Narok Kili -20 Nandi Kwale Taita Isiolo Kitui Siaya Migori Kisumu Mombasa Narok Vihiga Average Nakuru Muranga Homa Bay Garissa Makueni Machakos Marsabit Turkana West Pokot Kakamega Nyandarua Kili Source: OCOB. Source: OCOB. IV. International perspectives on fiscal transfers Kenya’s horizontal revenue sharing formula places International principles on fiscal transfers reflect an emphasis on fiscal need. The formula for the fiscal need, capacity and effort. These principles horizontal sharing of revenues includes six parameters: support the view that needs vary across different a basic share (26 percent), population (45 percent), segments of society, and that those with greater land area (8 percent), poverty index (26 percent), needs should receive greater support. Further, the fiscal responsibility (2 percent) and a development provision of public services should be adjusted for the factor (1 percent). Population has the largest weight capacity of different parts of a country to generate and is a proxy for the expenditure needs of a given their own revenues, either through their natural county. Additionally, land area accounts for higher endowments of resources, or their ability to leverage costs associated with delivering services to larger existing infrastructure. Also importantly, rewarding geographical areas. The basic share accounts for the subnational governments based on their efficiency fixed costs of running county governments, which are creates incentives for better performance while assumed to be similar to some extent across all county reducing the potential for moral hazard problems. governments. The poverty gap and development factor reinforce the redistributive elements of the formula. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 213 Appendices Similar to Kenya’s approach, South Africa’s revenue The emphasis of India’s formula has changed over sharing formula also places an emphasis on time. Population, land area and fiscal capacity have fiscal need. South Africa’s formula includes six consistently been maintained in India’s revenue- components, namely: education (48 percent), sharing formula, while other factors such as tax effort, health (27 percent), a basic component (16 infrastructure index, fiscal discipline, demographic percent), an institutional component (5 percent), change and forest cover have been implemented in poverty (3 percent) and economic output (1 different formulas based on the recommendations of percent). Analogous to Kenya, the basic component India’s Finance Commissions. is distributed based on each province’s share of the national population, while the institutional Fiscal effort is accounted for in Kenya with a component is divided equally amongst provinces. fiscal responsibility factor. The fiscal responsibility parameter is meant to reward implementation of South Africa measurement of needs takes a sectoral sound economic and budgetary practices. This is similar approach. The education and health components to the fiscal discipline factor in India’s formula, which account for 75 percent of total allocations, and was implemented in recommendations between the implicitly reflect government priorities. Further, Eleventh and Thirteenth Finance Commissions. these components use parameters that measure the actual cost of delivering services in each V. Looking forward: Opportunities and respective sector: the education component uses challenges the school-age population and school enrolment Devolution has been a complex endeavor, which has rate, while the health component uses a risk- occurred at a rapid pace. While significant progress adjusted capitation index and the number of visits has been made, challenges to implementation remain. to primary health care clinics. Key forward-looking opportunities include: i) Continued improvement of capacity within However, Kenya’s approach proxies fiscal need counties. While counties have shown substantial with population and land area. Taken together, progress in improving capacity, gaps remain in population and land area account for 53 percent of county human capital, infrastructure, processes horizontal transfers, a weight more than three times (including budgeting and absorption) and that assigned to South Africa’s basic population- coordination. Closing these gaps presents a based component. The share of South Africa’s significant opportunity for counties to improve equally distributed institutional component is also efficiency and deliver higher quality public services significantly smaller than Kenya’s basic share. to households. India’s revenue sharing formula places an emphasis ii) Increasing own-source revenues and lowering on fiscal capacity, in contrast to Kenya which does dependence on transfers. Own-source revenue not consider fiscal capacity. Parameters included collections remain low, and vary significantly in the India formula are population (17.5 percent), across counties. Counties have an opportunity income distance (50 percent), land area (15 percent), to increase revenues by widening local tax bases demographic change (10 percent) and forest cover and increasing tax effort. Common and generally (7.5 percent). The highest weight is attributed to accepted sources of subnational revenues include fiscal capacity in terms of income distance, which is property taxation, fees and charges, licenses, some measured by the shortfall between actual per capita types of business taxation, motor vehicle taxes and income of a state compared with the state with the licenses and business or sales taxes.5 highest per capita income. 5 See Smoke 2012. 214 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices iii) Review of criteria for the horizontal sharing of are measured more directly within priority revenues. International comparisons show that decentralized sectors, or in India where priority is population and land area, which account for over given to fiscal capacity. half the share of horizontal transfers in Kenya and which proxy the costs of service delivery, Managing wage bills. Wage bills comprise a are also used in other countries with varying significant share of county recurrent costs, and are weights. However, there are also key differences growing significantly. Managing rising wage costs in approaches, such as the sectoral approach in prudently will lower the risks of increasing recurrent South Africa where the costs of service delivery expenditures crowding out public investment. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 215 Appendices APPENDIX B: CHAPTER 2 ADDITIONAL MATERIALS B.1. Map of NEDI counties Figure B.1: Map of NEDI counties NEDI Non-NEDI Source: World Bank. B.2. Characteristics of peri-urban households Peri-urban clusters have a population density much percent of peri-urban household heads are engaged closer to that of rural clusters. The median population in agriculture as are 64.1 percent of rural household density of the clusters classified as peri-urban in the heads, yet this number is only 6.0 percent for those KIHBS 2015/16 survey in the 2009 Population Census living in core urban areas. was 537 individuals per square kilometer. This number is much closer to the median population density of Moreover, the proportion of food consumption the clusters classified as rural in the KIHBS 2015/16 that comes from own production for peri-urban (297 individuals per square kilometer) than to that of households resembles that of rural households. urban clusters, at 8,235 people per square kilometer. Given the high proportion of household heads The distribution of the natural log of population working in agriculture for peri-urban households, it densities also displays a similar trend (Figure B.2). is not surprising that roughly one fifth of the value of food consumption comes from own production. Moreover, for both peri-urban and rural households For rural households, the same figure is around the most popular occupational sector of household 27.8 percent, whereas only 2.5 percent of food heads is agriculture. Figure B.3 below shows the consumption in core urban households is obtained employment sector of the household head, 43.3 through own production (Figure B.4). Table B.1: Sampling framework  Rural Peri-urban Core urban Number of clusters 1,386 282 694 Mean population density (persons per sq. km.) 520 1,507 19,032 Median population density (persons per sq. km.) 297 537 8,235 Source: Author’s calculations based on KIHBS. 216 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices Figure B.2: Distribution of the log of population density by Figure B.3: Occupational sector of household head by area cluster type of residence 100 Proportion of employed household heads, % 80 69.8 64.1 60 43.3 41.1 40 26.5 20 12.2 12.0 7.0 8.6 6.0 4.2 5.1 0 Agriculture Manufaturing Services Construction Rural Peri-urban Core urban Source: 2009 Kenya Population Census. Source: Own calculations based on KIHBS 2015/16 Peri-urban housing characteristics resemble those and peri-urban counterparts. More than 80 percent of rural households rather than those of core urban of core urban households have access to electricity, households. The vast majority of peri-urban and whereas this number is less 40 percent in peri-urban rural households own their dwellings, compared to urban areas and 20 percent for households in rural only one in 6 core urban households. Similarly, core areas (Figure B.5). As can be seen, the conditions urban households have a significantly higher level of of peri-urban households resemble those of rural access to services and infrastructure (such as water, households, particularly when compared to those of sanitation, and waste management) than their rural core urban households. Figure B.4: Source of food consumption by area of Figure B.5: Household characteristics by area of residence residence 96.6 100 100 90.9 85.2 Proportion of food consumption by source, % 85.6 80.6 73.7 73.9 80 69.9 Proportion of households, % 80 63.5 65.3 60 60 57.2 48.6 44.5 38.9 40 40 16.9 20.0 27.8 20 21.9 20 3.6 1.6 8.2 8.0 9.0 6.7 4.8 0 2.5 2.9 Owns Access to Access to Access to Garbage 0 house improved improved eletricity collected Purchases Own stock Own production Gifts / other water sanitation Rural Peri-urban Core urban Rural Peri-Urban Core Urban Source: Own calculations based on KIHBS 2015/16. Source: Own calculations based on KIHBS 2015/16. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 217 Appendices Table B.2: Response rates by county Number of house- Number of house-  County Response rate County Response rate holds (000s) holds (000s) Mombasa 88.5% 397 West Pokot 90.8% 119 Kwale 89.6% 174 Samburu 95.7% 61 Kilifi 92.2% 326 Trans Nzoia 93.3% 210 Tana River 90.8% 56 Uasin Gishu 90.6% 270 Lamu 94.8% 30 Elgeyo Marakwet 92.6% 99 Taita Taveta 92.7% 102 Nandi 93.5% 202 Garissa 82.9% 78 Baringo 91.0% 152 Wajir 89.1% 69 Laikipia 90.6% 135 Mandera 91.5% 111 Nakuru 86.5% 578 Marsabit 80.5% 62 Narok 95.2% 223 Isiolo 95.0% 34 Kajiado 81.9% 250 Meru 95.4% 393 Kericho 91.0% 211 Tharaka Nithi 93.1% 107 Bomet 93.5% 179 Embu 94.6% 164 Kakamega 95.2% 392 Kitui 90.4% 236 Vihiga 95.2% 144 Machakos 92.9% 328 Bungoma 93.7% 321 Makueni 95.0% 233 Busia 90.8% 177 Nyandarua 93.3% 191 Siaya 93.3% 246 Nyeri 96.5% 271 Kisumu 93.0% 284 Kirinyaga 91.4% 198 Homa Bay 92.1% 224 Muranga 93.3% 323 Migori 91.0% 233 Kiambu 88.1% 600 Kisii 95.4% 291 Turkana 87.9% 246 Nyamira 94.0% 179 Nairobi 76.9% 1.503 Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16. Figure B.6: Asset ownership by consumption quintile, Nairobi Owns house 25 21.4 20 18.3 Proportion of households, % 15 12.8 10.5 10 8.8 8.5 8.1 7.3 7.6 6.9 6.8 4.9 5 2.6 2.8 1.3 1.0 0 Q1 Q2 Q3 Q4 Q5 NP P ALL 2005/06 2015/16 218 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices Owns fridge 80 61.0 60 Proportion of households, % 42.0 40 23.7 21.0 19.0 18.3 18.3 16.5 20 10.0 4.7 5.1 5.1 1.9 2.1 2.3 1.5 0 Q1 Q2 Q3 Q4 Q5 NP P ALL 2005/06 2015/16 Owns sofa 100 82.0 80 75.9 73.9 71.9 70.0 73.7 70.6 68.1 70.0 Proportion of households, % 66.6 65.9 57.6 58.0 60.0 60 55.6 52.8 40 20 0 Q1 Q2 Q3 Q4 Q5 NP P ALL 2005/06 2015/16 Owns car 40 36.8 Proportion of households, % 30 22.7 20 11.9 10.3 9.9 0.0 6.0 6.7 6.0 0.0 0.0 0.0 0.8 1.3 0.0 0.0 0 Q1 Q2 Q3 Q4 Q5 NP P ALL 2005/06 2015/16 Owns washing machine 50 38.5 40 Proportion of households, % 31.5 30 13.5 10.6 20 9.7 6.7 6.3 1.2 0.6 1.0 2.4 2.3 1.1 0.0 0.0 0.0 0 Q1 Q2 Q3 Q4 Q5 NP P ALL 2005/06 2015/16 Source: Own calculations based on KIHBS 2005/06 and KIHBS 2015/16. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 219 Appendices APPENDIX C: CHAPTER 3 ADDITIONAL MATERIALS C.1. Review of the KIHBS 2015/6 data This chapter relies heavily on the analysis of the . Possibly: Collect information on which KIHBS 2015/6. Below is a list of possible tweaks to household member(s) own(s) the enterprise. the KIHBS instrument that could potentially enhance . Collect enterprise-level information on access the usefulness and quality of the data in the future, to finance, value of capital stock, business especially with respect to understanding gender gaps expenses, etc. to allow for more detailed in economic opportunities. analysis of enterprise productivity. º Questionnaire Q1A – Household Members . Review question N06 (e.g. distinction between Information – Section C (Education) paid and unpaid household members may be blurred; unpaid apprentices/volunteers are rare . C09: Review skip pattern of response categories V (too old to attend school). and may not require separate categories). º Questionnaire Q1A – Household Members º Questionnaire Q1B – Household Level Information – Section D (Labor) Information – Section K (Agricultural Holdings) . Incorporate new labor statistics definitions . Collect information on which household adopted by the 19th ICLS. See http://www.ilo. member(s) own(s) the parcel (after K08). org/wcmsp5/groups/public/---dgreports/- . Possibly: Collect information on which --stat/documents/normativeinstrument/ household member(s) provided labor on wcms_230304.pdf each parcel during the last 12 months (or last season, as per the reference period for . Review screening questions (D2_01-D2_06) agricultural production). Refer to upcoming for length. Examples/illustrations do not guidelines on agricultural labor data of the necessarily have to be included in the main question. Living Standards Measurement Study (LSMS) group. . Harmonize reference periods used for questions on wages/salaries (D26, D27) and hours worked º Questionnaire Q1B – Household Level (D18-D20) in the primary job (to facilitate Information – Section L (Agricultural Output) normalizing wages/salaries for hours worked). . Collecting agricultural output at the crop- parcel level would provide an opportunity º Questionnaire Q1B – Household Level to analyze gender differences in productivity Information – Section N (Household Enterprises) (comparing male- and female managed plots . Collect information on which household within households) and improve the analysis member(s) manage(s) the enterprise or of agricultural productivity (currently, inputs is(are) most familiar with it (ask respondent to cannot be linked to crops). Data on the crop specify ID(s), potentially allowing for multiple disposition and sales should remain at the managers). crop-level. 220 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices C.2. Classifying household enterprises as male-, female- or jointly run Household enterprises are classified as male-, female- paid household members is greater, equal, or or jointly-run based on question N06 of section N smaller than the number of female paid household (household enterprise module) of the household members (and the same applies to steps (2), (3) questionnaire of the KIHBS 2015/6. Question N06 and (4) below). asks how many male and female (i) paid household • If there are no paid household members engaged members, (ii) unpaid household members, (iii) unpaid in the income generating activity but 1+ unpaid proprietors/directors, (iv) unpaid apprentices, (v) proprietor(s)/director(s), the enterprise is classified unpaid volunteers, and (vi) paid non-household as female-, male- or jointly run based on the members are engaged in the income generating male-female composition of unpaid proprietors/ directors. activity (see below). Based on the respondent’s answers to these questions, enterprises are classified • If there are neither any paid household members as follows: engaged in the income generating activity nor any unpaid proprietor(s)/director(s), but 1+ unpaid • If 1+ paid household member(s) are engaged in household member(s), the enterprise is classified the income-generating activity, the enterprise is as female-, male- or jointly run based on the classified as female-, male- or jointly run based on male-female composition of unpaid household the male-female composition of paid household members. members. Enterprises with male and female paid • Enterprises engaging no household members household members are always classified as jointly (paid or unpaid) and no unpaid proprietor(s)/ run, irrespectively of whether the number of male director(s) are not classified. ! KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 221 Appendices C.3. Additional tables and figures Table C.1: Correlates of labor force participation, probit (coefficients) Note: Probit estimation with survey settings. Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1 Reference categories as follows: Head’s/own education – no schooling; Religion – Catholic; Marital status – monogamously married. 222 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices Table C.2: Oaxaca-Blinder decomposition of gender gaps in monthly earnings, summary lnwage coef. Overall % of total difference group_1 (male) 9.291 group_2 (female) 8.920 difference 0.372 endowments 0.159 43% coefficients 0.242 65% interaction -0.029 -8% Endowments % of total endowment effect age 0.025 16% hours 0.044 28% education -0.016 -10% industry 0.160 101% occupation -0.046 -29% location -0.009 -6% Coefficients % of total coefficient effect age -0.037 -15% hours -0.053 -22% education -0.139 -57% industry 0.069 28% occupation -0.004 -2% location -0.035 -15% _cons 0.442 183% interaction % of total interaction effect age -0.002 6% hours -0.007 23% education 0.001 -4% industry -0.061 211% occupation 0.038 -131% location 0.001 -4% Source: KIHBS 2015/6. Note: See Tables C.3 and C.4 for details. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 223 Appendices Table C.3: Oaxaca-Blinder decomposition of gender gaps in monthly earnings, descriptive statistics Variables (1) (2) Variables (1) (2) Male Female Male Female Dependent vaiable O/Q - Education, Health, Social 0.0785 0.0612 in (wage) 9.291 8.92 Security (0.0240) (0.0329) (0.00558) (0.00632) Independent vairable R/T - other Services 0.127 0.219 Age (0.00642) (0.0104) Age in years 34.81 33.36 0.0601 0.204 (0.199) (0.224) (0.00466) (0.0121) Hours Occupation Usual working hours 51.67 46.01 1. Legislators, admnistrators and 0.0221 0.0196 managers (0.369) (0.544) (0.00349) (0.00445) Education 2. Professionals 0.0592 0.0711 Primary or post-primary 0.439 0.412 (0.00435) (0.00693) (0.0101) (0.0143) 3. Technicians and associated 0.105 0.139 Secondary or college 0.458 0.462 professionals (0.103) (0.0136) (0.00613) (0.00848) University graduate or post 0.0731 0.0886 4. Secretariat, clerical services 0.0309 0.0599 graduate and related workers (0.00761) (0.00976) (0.00394) (0.00615) Other 0.00116 0.000789 5. Service workers, shop and 0.0853 0.161 (0.000312) (0.000680) market sales workers Idustry (0.00514) (0.0100) B - Mining 0.0145 0.00343 6. Skilled farm, fishery, wildlife 0.0515 0.0642 (0.00413) (0.00144) and related workers C- - Manufacturing 0.0879 0.042 (0.00357) (0.00576) (0.00731) (0.00666) 7. Craft and related trade 0.0879 (0.0352 workers D/E/F - Utilities, Construction 0.178 0.0199 (0.00568) (0.00645) (0.00889) (0.00428) 8. Plant and machine operators 0.119 0.00829 G - Trade 0.0882 0.0817 and assemblers (0.00570) (0.00776) (0.00658) (0.00219) H - Transport 0.113 0.00826 10. Armed forces 0.00337 0.000115 (0.00586) (0.00200) (0.001331) (7.10e-05 I - ICT 0.0321 0.0691 Location (0.00331) (0.00680) Urban 0.494 0.512 K/L - Finance, Real Estate 0.0104 0.00979 (0.0151) (0.0173) (0.00282) (0.00331) M/N - Professional, 0.00808 0.0154 Observations 7,562 4,088 Adminisrative Services (0.00235) (0.00351) Source: KIHBS 2015/6. Note: Standard errors in parentheses. Only variables included in the regression - reference categories not shown (see Table C.4). Ln(wage) denotes log monthly earnings (winsorized). 224 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices Table C.4: Oaxaca-Blinder decomposition of gender gaps in monthly earnings, OLS (coefficients) Variables (1) (2) Variables (1) (2) Male Female Male Female Age R/T Other Services 0.0577 -0.0159 Age in years 0.0163*** 0.0174*** (0.0588) (0.0535) (0.00116) (0.00153) Occupation Hours 2. Professionals 0.970*** 1.099*** Usual working hours 0.00658*** 0.00774*** (0.0746) (0.0917) (0.000937) (0.00110) 3. Technicians and associated 0.811*** 0.731*** Education professionals Primary or post-primary 0.236** 0.373*** (0.0595) (0.0753) (0.0687) (0.0795) 4. Secretarial, clerical services 0.663*** 0.755*** and related workers Secondary 0.530*** 0.660*** (0.0700) (0.0710) (0.0699) (0.0846) 5. Service workers, shop and 0.169*** 0.208*** University graduate or post 1.231*** 1.488*** market sales workers graduate (0.0521) (0.0521) (0.0933) (0.111) 6. Skilled farm, fishery, wildlife 0.0643 -0.0597 Other -0.0829 0.428*** and related workers (0.195) (0.122) (0.0547) (0.0786) Industry 7. Craft and related trades 0.238*** 0.374** B - Mining 0.166 0.825*** workers (0.198) (0.197) (0.0702) (0.170) C- Manufacturing 0.426*** 0.380*** 8. Plant and machine operatins 0.395*** 0.00362 and assemblers (0.0596) (0.144) (0.0450) (0.240) D/E/F - Utilities, Construction 0.490*** 0.651*** 10. Armed forces 1.381*** 1.798*** (0.0477) (0.176) (0.178) (0.105) G - Trade 0.342*** 0.196** 1. Legislators, administrators 1.132*** 1.159*** (0.0670) (0.0806) and managers H - Transport 0.489*** 0.451** (0.101) (0.137) (0.0559) (0.190) Location I - Accomodation 0.280*** 0.106 Urban 0.427*** 0.496*** (0.0716) (0.0916) (0.0328) (0.0402) J - ICT 0.668*** 0.472** Constant 7.144*** 6.702*** (0.107) (0.205) (0.0938) (0.118) K/L - Finance, Real Estate 0.675*** 0.238 (0.160) (0.145) Observations 7,562 4,088 M/N - Professional, 0.351*** 0.458*** R-squared 0.561 0.616 Administrative Services (0.0545) (0.0746) O/Q - Education, Health, Social 0.383*** 0.242*** Security (0.0601) (0.0738) Source: KIHBS 2015/6. Note: OLS regression with survey settings. Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Dependent variable is log monthly earnings (winsorized). Reference categories as follows: Education – no schooling; Industry – agriculture; Occupation – elementary occupations. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 225 Appendices Table C.5: Correlates of household enterprise profits, OLS (coefficients) Variables (1) (2) (3) raw gender gap with controls with all controls (except labor input) Female-run enterprise -0.732*** -0.623*** -0.567*** (0.0565) (0.0507) (0.0487) Industry A - Agriculture -0.132 -0.170 (0.178) (0.179) B - Mining -0.265) -0.312 (0.196) (0.202) C - Manufacturing -0.139 -0.165** (0.0866) (0.0804) D/E/F - Utilities, construction 0.511** 0.384* (0.212) (0.210) H - Transport 0.165 0.222** (0.109) (0.104) I - Accomodation (0.188) 0.0256 (0.118) (0.117) J - ICT 0.118 0.127 (0.230) (0.216) K/L - Finance, real estate 0716*** 0.738*** (0.197) (0.204) M/N - Professional, administration services 0.649** 0.663** (0.299) (0.298) O/Q - Education, health, social security 0.631*** 0.498** (0.188) (0.184) R/T - Other services -0160* -0.153* (0.0951) (0.0924) Urban 0.823*** 0.782*** (0.0658) (0.0650) Enterprise is registered 0.443*** 0.298*** (0.0905) (0.0797) Number of households and/or unpaid workers 0.0885* (0.0504) Number of paid non-households workers 0.199*** (0.0306) Constant 8.929*** 8.404*** 8.262*** (0.0460) (0.0564) (0.0821) Observations 4,125 4,125 4,125 R-squared 0.080 0.255 0.255 Source: KIHBS 2015/6. Note: OLS regression with survey settings. Standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Dependent variables is log monthly profits (winsorized). Reference category for Industry is ‘trade’. Jointly-run enterprises excluded from sample. 226 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices C.4. Comparability of the 2005/6 and 2015/6 KIHBS labor modules To compare wage and enterprise employment across the KIHBS 2005/6 and 2015/6, the following definitions are used: KIHBS 2005/6 KIHBS 2015/6 Questionnaire Categorized as Questionnaire Categorized as e05: During the past 7 days, how Employment - In the last d02_1: worked (at least one hour) as an employee Employment - many hours was NAME employed wage 7 days, has for wage, salary, commission or any payment in wage for a wage, salary, commission or (if hrs>=1) [NAME] ...... kind; including doing paid domestic work or farm (if yes) any payment in kind? work? e06: During the past 7 days, how Employment - d02_2: worked (at least one hour) on your own Employment - many hours did NAME work on enterprise account or as an employer in a business enterprise, enterprise any enterprise belonging to a (if any hrs>=1) for example, as a trader, shopkeeper, barber, (if any yes) member of household, including dressmaker, carpenter, taxi driver. helping for no pay? d02_3worked (at least one hour) on your own e07: During the past 7 days, how account or as an employer on a farm owned or many hours did NAME work on rented, whether in cultivating crops or in other the household farm, in a field or farm maintenance tasks, or have you cared for herding livestock? livestock belonging to you or a member of your household? d02_4: helped (for at least one hour) in a business enterprise /agricultural activity or cared for livestock belonging or run by this household? d02_5: worked (at least one hour) as an intern or an apprentice? d02_6: worked (at least one hour) as a volunteer? For cross-sectional analysis using the 2015/6 (without comparison to 2005/6), a slightly wider definition of employment is used: KIHBS 2015/6 Questionnaire Categorized as In the last 7 days, d02_1: worked (at least one hour) as an employee for wage, salary, commission or any payment in kind; Employment - at has [NAME] ...... including doing paid domestic work or farm work? work (if any yes) d02_2: worked (at least one hour) on your own account or as an employer in a business enterprise, for example, as a trader, shopkeeper, barber, dressmaker, carpenter, taxi driver. d02_3worked (at least one hour) on your own account or as an employer on a farm owned or rented, whether in cultivating crops or in other farm maintenance tasks, or have you cared for livestock belonging to you or a member of your household? d02_4: helped (for at least one hour) in a business enterprise /agricultural activity or cared for livestock belonging or run by this household? d02_5: worked (at least one hour) as an intern or an apprentice? d02_6: worked (at least one hour) as a volunteer? d04: Even though [NAME] did not do any of these activities in the last 7 days, does he/she have ....... Employment - that he/she would definitely return to? MULTIPLE absent (if any A-F) A paid job ............................................. A A business ............................................. B An own family/farming activity .......................C An unpaid job ..........................................D Apprentice/intern .......................................E Volunteer ................................................F No Activity ............................................. G d05: Why was [NAME] absent from work during the last 7 days? (not absent if d05==07) VACATION/HOLIDAYS ................................01 A business ............................................. B ILLNESS, INJURY, TEMPORARY DIABILITY...03 An unpaid job ..........................................D TEMPORARY SLACK WORK ......................05 Volunteer ................................................F OFF SEASON ...............................................07 EDUCATION OR TRAINING ...........................09 TEMPORARY CLOSURE ...............................11 OTHER SPECIFY ..........................................13 d06: Do you have an agreement or contract to return to the same job after this absence, or if it is your (not absent if own/family business, is the business still operational ? d06==NO & d07>=3) d07: After how long will [NAME] return to work? LESS THAN 1 MONTH ..............................01 1 MONTH TO LESS THAN 3 MONTHS ......02 3 MONTHS AND ABOVE ...........................03 NOT SURE WHEN TO RETURN...............04 NOT RETURNING ....................................05 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 227 Appendices Unemployment is computed as follows in 2015/6: • Unemployment = NOT employed AND job search effort (except ‘registering dispute’, other passive, none – see d11) AND available (d13<=2) ! ! ;: ^!.-Z+,-!($!f(%)#,V!$') ! /,+00%.%U+#%()!(.!-)#-$[$%0-0!+0!Z+,- " " " 228 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices APPENDIX D: CHAPTER 4 ADDITIONAL MATERIALS D.1. Empirical approach: Crop yield analysis To rigorously investigate the determinants of crop yield, we apply a fixed effects model. In this model, we start with a basic specification where logarithm of per acre yield () is regressed on fixed effects of household , a vector of household characteristics , human capital endowment of the household head , and technology adoption indicators as follows: (1) …where stands for time of survey (2000, 2004, 2007 and 2010), and is the error term. captures household size and dependence ratio; includes the head’s gender, age, age squared, and years of completed education; and includes dummy for application of chemical fertilizer, improved maize seed, and membership to cooperative or group through which farmers access agricultural technologies. In some specifications interaction between dummies of fertilizer application and the use of improved seed is included to assess the join effects using both technologies. This analysis is conducted for maize, beans, tea and coffee. Two addition specifications of Equation 1 are also implemented to investigate the ‘inverse plot size productivity puzzle’ and the differential impact of fertilizer application by farmers with different plot size. First, we add three dummies indicating the quartile of plot size as in Equation 2 below. Reserving the lowest quartile as a reference group, this estimation provides the conditional difference in crop yield among farmers with larger plots (quartile 2.3) and those with smaller plots (the lowest quartile). (2) …where is an indicator function which is equal to one when land quartile, is equal to j (=2, 3, 4) and zero otherwise. The next specification (Equation 3) is intended to analyze the effectiveness of fertilizer application in enhancing productivity for smallholder and large farmers. By introducing interaction between a dummy for the application of fertilizer and four dummies for plot size quartiles, we capture the percentage improvement in yield ( ) for households in the four plot size quartiles.249 249 In this specification, fertilizer is not included by itself as an explanatory variable. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 229 Appendices Table D.1: Determinants of beans yield, FEs Model (1) (2) Fertilizer used per plot 0.00*** 0.00 (2.60) (1.31) Distance to extension services 0.00 0.00 (0.41) (0.46) Belong to Cooperative/Group membership 0.09 0.11** (1.56) (1.97) Cropped land quartile (the lowest quartile is the reference group): 0.00 2nd quartile .0.41*** (.7.34) 3rd quartile .0.66*** (.10.01) 4 quartile th .0.86*** (.11.71) Constant 4.20*** 4.45*** (8.57) (9.34) Number of Households Number of Observations 3784 3784 Note: Standard errors in parentheses: * p < 0.1, ** p < 0.05, *** p < 0.01. Note that the dependent variable is logarithm of yield (kg/acre). 230 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices APPENDIX E: CHAPTER 5 ADDITIONAL MATERIALS Table E.1: Nominal monthly salary in urban Kenya (1) (2) (3) (5) (6) 0.093*** 0.114*** 0.0893*** 0.0804*** Age (0.006) (0.006) (0.006) (0.006) -0.000*** -0.001*** -0.000*** -0.000*** Age (squared) (0.000) (0.000) (0.000) (0.0000) -0.472*** -0.542*** -0.483*** -0.448*** Female (0.019) (0.022) (0.020) (0.020) Education: primary (base) 0.289*** 0.259*** 0.179*** Education: secondary (0.022) (0.022) (0.024) 1.176*** 1.139*** 0.860*** Education: higher (0.025) (0.025) (0.027) -0.616*** -0.665*** -0.477*** -0.383*** Economic sector: agriculture (0.046) (0.043) (0.039) (0.044) 0.056 -0.031 0.054 -0.049 Economic sector: manufacturing (0.039) (0.037) (0.033) (0.033) Economic sector: services (base) -0.027 -0.269*** -0.082** 0.130*** Economic sector: construction (0.038) (0.036) (0.032) (0.045) Contract: written (base) -0.545*** Contract: verbal (0.029) -0.699*** Contract: implied (0.103) -0.509*** Contract: none (0.024) 7.299*** 9.508*** 7.494*** 7.426*** 8.120*** Constant (0.112) (0.013) (0.123) (0.112) (0.121) County fixed effects Yes Yes Yes Yes Yes Adj-R2 0.361 0.054 0.180 0.375 0.424 Obs. 7081 7523 7523 7081 6125 Source: Staff calculation with KIHBS 2015/16. Note: Standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 231 Appendices Table E.2: Comparison of dwelling characteristics between informal settlement and non-informal settlement areas in Nairobi Informal Non-informal Diff settlement settlement (3) (1) (2) Housing tenure: owned 0.117 0.060 0.057 Housing tenure: rent-paying tenant 0.846 0.883 -0.036 Housing tenure: rent-free tenant 0.037 0.057 -0.020 Number of rooms: 1 0.840 0.615 0.224*** Number of rooms: 2 0.084 0.166 -0.082*** Number of rooms: 3 0.059 0.123 -0.064** Number of rooms: 4 0.012 0.062 -0.050*** Number of rooms: 5 or more 0.005 0.034 -0.029** Wall: mud 0.088 0.000 0.088*** Wall: other non-durable 0.005 0.000 0.005 Wall: corrugated iron sheets 0.543 0.132 0.411*** Wall: wood 0.007 0.012 -0.005 Wall: stone, cement, bricks 0.357 0.857 -0.500*** Roof: grass, thatch, makuti, mud 0.004 0.000 0.004 Roof: corrugated iron sheets 0.877 0.475 0.401*** Roof: concrete 0.112 0.500 -0.388*** Roof: tiles 0.000 0.025 -0.025*** Roof: other 0.008 0.000 0.008 Floor: earth, sand, dung 0.126 0.008 0.118*** Floor: wood, bamboo 0.000 0.007 -0.007** Floor: tiles 0.048 0.110 -0.062** Floor: cement 0.790 0.775 0.016 Floor: other 0.036 0.100 -0.064*** Source: Staff calculation with KIHBS 2015/16. Note: Standard errors in parentheses. * p < 0.1, ** p < 0.05, *** p < 0.01 Table E.3: Comparison of access to services between informal settlement and non-informal settlement areas in Nairobi Informal settle- Non-informal Diff ment settlement (3) (1) (2) Water: private tap within dwelling 0.105 0.459 -0.354*** Water: private tap outside dwelling 0.186 0.423 -0.237*** Water: public tap/standpipe 0.580 0.065 0.514*** Water: other improved water 0.104 0.044 0.061*** Water: non-improved 0.024 0.009 0.016 Toilet: flush toilet 0.439 0.875 -0.436*** Toilet: VIP latrine 0.058 0.008 0.049*** Toilet: covered pit latrine 0.382 0.116 0.266*** Toilet: uncovered pit latrine 0.074 0.000 0.074*** Toilet: other 0.046 0.000 0.046*** Electricity 0.833 0.963 -0.130*** Garbage collection 0.474 0.855 -0.381*** Source: Staff calculation with KIHBS 2015/16. Note: * p < 0.1, ** p < 0.05, *** p < 0.01 232 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices Figure E.1: Number of urban poor and urban poverty rate by county, 2015/16 5000 80 4500 70 4000 60 Population (thousands) 3500 50 Urban poverty rate (%) 3000 2500 40 2000 30 1500 20 1000 10 500 0 0 Urban poor Non-poor Urban poverty rate Source: Staff calculation based on KIHBS 2015/16. Note: Counties are ordered from the highest poverty rate (left) to the lowest poverty rate (right). Absolute poverty line is used. Figure E.2: Cash transfer during the last three months in 15 cities, 2013 Direction of cash transfer (% household sent) Direction of cash transfer (average amount sent) 60 5,000 Average amount of money (Ksh) 4,000 Percentage of households 40 3,000 2,000 20 1,000 0 0 Non- peri-urban Peri-urban Non- peri-urban Peri-urban Within the same city Other city Within the same city Other city Rural Abroad Rural Abroad Direction of cash transfer (% household received) Direction of cash transfer (average amount received) 10 500 8 400 Percentage of households Percentage of households 6 300 4 200 2 100 0 0 Non- peri-urban Peri-urban Non- peri-urban Peri-urban Within the same city Other city Within the same city Other city Rural Abroad Rural Abroad Source: Staff calculation based on the 2013 Cities Baseline Survey. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 233 Appendices Figure E.3: Expenditure share on housing in urban Kenya a) Urban households b) Poor and non-poor Median expenditure share on housing among poor (%) 20 20 Median expenditure share on housing (%) 15 15 10 10 5 5 8 10 12 14 5 10 15 20 Log of urban population Median expenditure share on housing among non-poor (%) County Fitted line County Fitted line Source: Staff calculation based on KIHBS 2015/16. Figure E.4: Expenditure share on housing in urban Kenya by county, 2015/16 a) Urban households b) Urban poor households Homa Bay Homa Bay Baringo Baringo Bungoma Bungoma Vihiga Busia Wajir Nyandarua Siaya Wajir Turkana Isiolo Kilifi Nyamira Nyamira Kilifi Tana River Turkana Kitui Siaya Elgeyo Marakwet Kisumu Tharaka Nithi Elgeyo Marakwet Isiolo Taita Taveta Mandera Vihiga Nyandarua Tharaka Nithi Taita Taveta Tana River Makueni Machakos Busia Migori Kisumu Mandera Nyeri West Pokot Nakuru Kwale Bomet Nakuru Kericho Makueni Trans Nzoia Marsabit Migori Embu Kirinyaga Uasin Gishu Marsabit Laikipia Machakos Garissa Samburu Kericho West Pokot Nyeri Garissa Nairobi Nandi Mombasa Mombasa Kirinyaga Embu Narok Uasin Gishu Meru Kwale Samburu Narok Trans Nzoia Nairobi Muranga Lamu Kakamega Kakamega Kiambu Meru Nandi Laikipia Lamu Muranga Kisii Kiambu Bomet Kisii Kajiado Kajiado Kitui 0 10 20 30 40 50 0 10 20 30 40 50 Expenditure share on housing (%) Expenditure share on housing (%) Source: Staff calculation based on KIHBS 2015/16. 234 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices Figure E.5: Comparison of health indicators in Kenya, 2000 to 2014 Under-five mortality rate 151.0 117.0 115.0 93.0 95.0 79.8 56.0 57.0 52.0 22.0 Nairobi Rural urban Nairobi National Nairobi Rural urban Nairobi National peri-urban peri-urban 2000 2003 2012 2014 Prevalence of diarrhea 30.8 20.2 15.8 17.0 16.0 15.7 15.6 13.9 14.3 15.2 Nairobi Rural urban Nairobi National Nairobi Rural urban Nairobi National peri-urban peri-urban 2000 2003 2012 2014 Total fertility rate 5.4 4.9 4.5 4.0 3.9 3.3 3.5 3.1 2.7 2.7 Nairobi Rural urban Nairobi National Nairobi Rural urban Nairobi National peri-urban peri-urban 2000 2003 2012 2014 Source: Mberu et al. 2016. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 235 Appendices Figure E.6: Number and share of unemployed population in urban area by county, 2015/16 14 300 12 250 10 200 Percent Thousands 8 150 6 100 4 2 50 0 0 Number of unemployed All age Source: Staff calculation based on KIHBS 2015/16. Note: Counties are ordered from the highest unemployment (left) to the lowest (right). Figure E.7: Unemployment rate in urban area by sex and county, 2015/16 20 18 16 14 12 Percent 10 8 6 4 2 0 All age All age (men) All age (women) Source: Staff calculation based on KIHBS 2015/16. Note: Counties are ordered from the highest unemployment (left) to the lowest (right). Figure E.8: Unemployment rate in urban area by the youth and county, 2015/16 25 20 15 Percent 10 5 0 All age Youth Source: Staff calculation based on KIHBS 2015/16. Note: Counties are ordered from the highest unemployment (left) to the lowest (right). 236 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices Figure E.9: Comparison of economic sectors in urban Kenya by county, 2015/16 a) Ordered by agriculture 100 90 80 70 60 Percent 50 40 30 20 10 0 Agriculture Manufacturing Construction Other services b) Ordered by manufacturing 100 90 80 70 60 Percent 50 40 30 20 10 0 Manufacturing Construction Other services Agriculture c) Ordered by services 100 90 80 70 60 Percent 50 40 30 20 10 0 Other services Construction Manufacturing Agriculture Source: Staff calculation based on KIHBS 2015/16. Note: Counties are ordered from the highest share of (a) agriculture, (b) manufacturing, and (c) services. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 237 Appendices Figure E.10: Duration of residence in 47 counties, 2014 Narok Uasin Gishu Kajiado Nyamira Homa Bay Nairobi Kisumu Laikipia Kili Baringo Kiambu Tana River Makueni Kakamega Nyeri Nakuru Siaya Meru Muranga Kwale Trans-nzoia Embu Busia Nandi Kitui Bungoma Isiolo Machakos West Pokot Kisii Bomet Elgeyo Marakwet Turkana Mombasa Migori Taita Taveta Wajir Lamu Samburu Nyandarua Tharaka Vihiga Garissa Kirinyaga Marsabit Mandera Kericho -40 -20 0 20 40 60 % of men in the county Rural to urban: <4 years Rural to urban: 4-8 years Urban to rural: <4 years Urban to rural: 4-8 years Source: Staff calculation based on the 2014 DHS. Note: Share of men who stay in the current residence less than 4 years or 8 years by previous residence (either urban or rural areas). Counties are ordered from the largest share of rural to urban migrants during the last 8 years (top) to the lowest share (bottom). 238 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices Figure E.11: Previous residence of recent migrants in 47 counties, 2014 a) Current urban residents b) Current rural residents Mombasa Kirinyaga Bungoma Machakos Kirinyaga Bomet West Pokot Tharaka Meru Bungoma Nakuru Uasin Gishu Trans-nzoia Baringo Makueni Meru Machakos Trans-nzoia Laikipia Garissa Mandera Uasin Gishu Kili Nyandarua Nakuru Homa Bay Elgeyo Marakwet Migori Kisii Elgeyo Marakwet Nandi Tharaka Isiolo Nandi Laikipia Kericho West Pokot Taita Taveta Kakamega Baringo Narok Kisii Kiambu Busia Kwale Bomet Migori Kisumu Marsabit Kitui Narok Embu Garissa Nyeri Wajir Wajir Kili Busia Nairobi Samburu Nyeri Kitui Embu Kisumu Kakamega Nyandarua Kajiado Makueni Siaya Nyamira Tana River Kajiado Muranga Lamu Turkana Homa Bay Kiambu Siaya Lamu Tana River Isiolo Nyamira Taita Taveta Kwale Turkana Vihiga Muranga Samburu Vihiga Marsabit Mandera 0 20 40 60 80 100 0 20 40 60 80 100 % of men arrived during the last 8 years % of men arrived during the last 8 years Nairobi/Mombasa/Kisumu Other towns Countryside Abroad Nairobi/Mombasa/Kisumu Other towns Countryside Abroad Source: Staff calculation based on the 2014 DHS. Note: Share of previous residence among men who moved in the current residence during the last 8 years in urban areas (panel a) and rural areas (panel b). Counties are ordered from the largest share of countryside (top) to the lowest share (bottom). KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 239 Appendices Figure E.12: Previous residence of recent migrants in 47 countries, 2014 a) From urban b) From NMK Share (%) Share (%) 0 - 10 0 - 10 10 - 20 10 - 20 20 -30 20 -30 30 - 40 30 - 40 40 - 50 40 - 50 50 - 60 50 - 60 60 - 70 60 - 70 70 - 80 70 - 80 80 - 90 80 - 90 90 - 100 90 - 100 a) From rural b) From abroad Share (%) Share (%) 0 - 10 0 - 10 10 - 20 10 - 20 20 -30 20 -30 30 - 40 30 - 40 40 - 50 40 - 50 50 - 60 50 - 60 60 - 70 60 - 70 70 - 80 70 - 80 80 - 90 80 - 90 90 - 100 90 - 100 Source: Staff calculation based on the 2014 DHS. Note: Share of previous residence (a) from any urban area, (b) from either Nairobi, Mombasa, or Kisumu, (c) from rural area, and (d) from abroad, among men who moved in urban areas during the last 8 years 240 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices Figure E.13: Cumulative distribution of the duration of residence in Nairobi and Mombasa a) Nairobi b) Mombasa 100 100 90 90 80 80 70 70 60 60 Percent Percent 50 50 40 40 30 30 20 20 10 10 0 0 0 2 4 6 8 10 12 14 16 ≥18 20 or 0 2 4 6 8 10 12 14 16 ≥18 20 or never moved never moved Duration of residence in the neighborhood (years) Duration of residence in the neighborhood (years) Informal settlement Non- Informal settlement Informal settlement Non- Informal settlement Source: Cities baseline survey 2013. Note: Households with duration of residence in the current neighborhood less than 30 years are shown for the purpose of presentation. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 241 Appendices APPENDIX F: CHAPTER 6 ADDITIONAL MATERIALS F.1. GERs and NERs in secondary and primary education by county Table F.1: GERs and NERs in secondary and primary education by county Primary Secondary NER GER NER GER Mombasa 88.2% 102.4% 59.9% 95.4% Kwale 71.0% 103.7% 16.8% 34.0% Kilifi 76.5% 104.5% 25.7% 50.7% Tana River 70.5% 93.9% 31.2% 62.7% Lamu 80.0% 100.7% 31.4% 60.4% Taita Taveta 90.4% 110.9% 49.3% 74.7% Garissa 41.7% 59.8% 22.4% 52.5% Wajir 56.1% 77.6% 22.2% 41.7% Mandera 59.1% 79.3% 28.1% 59.5% Marsabit 55.3% 69.3% 24.4% 40.6% Isiolo 76.1% 95.7% 33.3% 56.8% Meru 88.9% 114.2% 38.2% 73.1% Tharaka Nithi 92.6% 122.6% 40.9% 72.0% Embu 94.5% 119.2% 47.4% 68.8% Kitui 92.1% 116.7% 37.8% 73.6% Machakos 96.0% 117.6% 55.6% 99.9% Makueni 95.4% 125.6% 45.2% 75.8% Nyandarua 91.2% 108.1% 54.8% 87.4% Nyeri 96.8% 116.6% 67.5% 102.3% Kirinyaga 94.3% 111.9% 66.8% 92.5% Muranga 93.7% 114.1% 56.8% 80.1% Kiambu 90.8% 104.7% 67.8% 100.1% Turkana 49.0% 71.7% 13.7% 36.9% West Pokot 68.9% 100.3% 22.4% 57.8% Samburu 60.7% 77.8% 17.9% 38.4% Trans Nzoia 89.8% 114.8% 39.0% 77.6% Uasin Gishu 89.7% 113.9% 44.0% 82.8% Elgeyo Marakwet 89.8% 119.9% 32.4% 76.5% Nandi 87.7% 118.7% 33.1% 71.8% Baringo 84.9% 111.7% 30.2% 68.4% Laikipia 77.0% 95.3% 45.2% 74.2% Nakuru 92.5% 105.8% 45.7% 80.1% Narok 77.0% 105.5% 22.4% 42.5% Kajiado 81.3% 101.5% 44.9% 80.1% Kericho 94.1% 118.2% 41.4% 73.3% 242 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices Primary Secondary NER GER NER GER Bomet 93.1% 119.1% 35.8% 70.6% Kakamega 89.8% 118.3% 41.3% 69.6% Vihiga 90.4% 112.3% 48.1% 79.3% Bungoma 85.9% 111.6% 46.3% 74.9% Busia 83.4% 116.1% 25.9% 57.3% Siaya 85.0% 109.1% 37.8% 60.6% Kisumu 90.0% 110.5% 41.6% 76.3% Homa Bay 85.0% 111.0% 35.4% 60.4% Migori 80.1% 108.3% 37.0% 71.1% Kisii 91.2% 111.7% 52.2% 92.8% Nyamira 88.0% 106.8% 54.3% 98.3% Nairobi 89.9% 100.9% 64.0% 111.0% Source: Own calculations based on KIHBS 2015/16. F.2. The transition from primary into secondary ii. Poorer households live in areas in which access to education secondary schools is limited (e.g. children would have to travel larger distances). This appendix provides further evidence regarding iii. Poorer children are initially enrolled when they are the transition from primary into secondary. Gross already older than their richer peers. At the time enrolment ratios drop at the transition from standard they complete primary school, their opportunity seven to standard eight and, even more pronounced, at costs of attending secondary exceeds the the transition from primary education into secondary expected benefits. education. This drop in GERs is more pronounced for children from families in the bottom 40 percent of the consumption distribution, which are on average The analysis is based on LPMs. Simple linear 13.5 percentage points less likely to transition from regression models were estimated with an indicator primary into secondary education than children from variable of a successful transition (from grade seven families in the top 60 percent. This appendix presents to grade eight of primary and from grade eight of further analysis of this phenomenon based on the primary into secondary). Explanatory variables include 2015/16 KIHBS. age of the child, a binary indicator for girls, a binary indicator for rural location, log per capita expenditure, Several non-exclusive hypotheses can be tested with the number of children in the household that attend the data at hand. Affordability is an obvious candidate primary and secondary, respectively, and a binary explanation for this large gap. However, the gap may indicator for attending grade eight of primary in the also be explained by factors that are correlated both previous school year. More sophisticated models with household consumption expenditure and the include cluster-fixed effects in order to control for the transition such as the age at the time of the transition physical accessibility of schools. or geographical remoteness. With the data at hand, the following hypotheses can be tested empirically: Physical access plays a minor role in explaining lower transition rates among the poor. On average i. Secondary education is more expensive than over the pooled sample, a ten-percent increase in primary (which is free for all practical purposes) the per capita expenditure is initially associated with and unaffordable for poorer households. a one-percentage point increase in the probability KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 243 Appendices of transitioning into the next grade (column (1)). associated with greater opportunity costs. It is found Controlling for age, gender, and rural locality and that one additional year in age is associated with a then including PSU-level fixed effects, the estimated 2.4-3.0 percentage point decline in the probability coefficient drops to 0.8 and 0.7 percentage points, of transitioning. It is also worth noting that among respectively, but remains statistically significant at the children enrolled in the eighth grade of primary, one- percent level (columns (2) and (3)). This suggests those in the bottom 40 percent of the consumption that locality and, thus, physical access, plays a minor distribution are on average 0.84 years older than role in explaining lower transition rates among the children in the top 60 percent. This suggests that only poor. Splitting the sample into children that attended around one fifth of the gap between the bottom 40 seventh grade of primary in the previous school year percent and the top 60 percent can be explained and those that attended the final year of primary, it by differences in age at the time they reach the final is found that a ten-percent increase in per capita grade of primary. expenditure is associated with an increase in the probability of transitioning from seventh into eighth Further support for the notion that costs associated and from eight grade of primary into the first grade with secondary play a major role comes from the of secondary by about 0.6 and 1.4 percentage points, estimated effects of the number of other children in respectively (columns (5) and (6)). However, only the the household that attend primary and secondary, later estimate is significant at conventional levels. respectively. While primary school attendance of other household members has no statistically significant One additional year in age lowers the probability effect on the probability of transitioning from primary of a successful transition in secondary by about into secondary, secondary school attendance of one three percentage points. Older children at the additional household member lowers the probability time of transition may be able to earn higher wages of a successful transition by almost 20 percentage in the labor market because of greater physical points (column (6)). readiness. Hence, continuing their education may be 244 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices Table F.2: Determinants of transition from seventh into eighth grade of primary and from primary into secondary (1) (2) (3) (5) (6) (6) 7 primary to th 8 primary to th Pooled 8th primary 1st secondary Age -0.024*** -0.030*** -0.030*** -0.014 -0.029* (0.005) (0.008) (0.008) (0.013) (0.015) Girl 0.008 -0.011 -0.011 -0.052 0.011 (0.015) (0.024) (0.024) (0.045) (0.050) Rural 0.001 (0.034) Log p.c. expenditure 0.101*** 0.079*** 0.069** 0.069** 0.060 0.139* (0.013) (0.016) (0.033) (0.033) (0.056) (0.072) # of HH members in -0.075** -0.007 -0.196** secondary education (0.037) (0.064) (0.099) # of HH members in -0.009 -0.010 0.099 primary education (0.034) (0.063) (0.069) Grade 8 in previous year -0.113*** -0.083*** (0.018) (0.018) Cluster-fixed effects? Yes. Yes. Yes. Yes. Observations 4,328 4,328 4,328 4,328 2,456 1,872 R-squared 0.038 0.054 0.506 0.508 0.585 0.800 Transition rate 0.792 0.792 0.792 0.792 0.836 0.736 Source: Own calculations based on KIHBS 2015/16. Note: Significance level: 1% (***), 5% (**), and 10% (*). Standard errors clustered at the PSU-level reported in parentheses. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 245 Appendices APPENDIX G: CHAPTER 7 ADDITIONAL MATERIALS G.1. Did the provision of free maternity services increase deliveries in health facilities? On June 1, 2013, the GoK initiated a policy of free provision of maternity services in all public facilities. The directive would take effect immediately and reportedly took many health professionals in the public sector by surprise: there were several reports of overcrowding and stock-outs at public maternity hospitals (Cherondo 2013). Prior to the reform, uninsured mothers were required to pay at least KSh 3,000 (about US$ 35 at 2013 exchange rates) for a normal birth and often considerably more. The 2014 KDHS were employed to analyze the effect of the reform on private vs. public uptake as well as the share of deliveries in any type of formal health facility.250 The timing of birth was exploited jointly with observations on birth by provider (if any) before and after June 2013. Both LPMs and logit models were used to model the choice of provider (public, private, or either). The LPMs were estimated with county-fixed effects, controlling for mother’s age at birth and its square, locality (rural or urban), and mother’s level of education in some specifications. It was found that a model that includes the interaction between month of birth (centered on the month in which the policy took effect) and a binary indicator for the reform was more appropriate than the alternative of either only the binary indicator for the reform or both variables. Table G.1 and Table G.2 report results from weighted estimations (using sample weights) but are robust to unweighted regression. In the preferred specification, the sample was restricted to two births that took place up to two years prior to the policy change and up to 18 months after the change. Overall, results suggest that the June-2013 decision resulted in a shift in demand from private to public provision; the overall effect on the proportion of births taking place in either private of public facilities is small. Table G.1 reports main regression results from LPMs for the three different outcome variables in columns (1) through (3). Logit results were qualitatively similar and for brevity they are not reported here. The estimated coefficient on the main variable of interest, the interaction between treatment and month of birth, in column (1) suggests a positive effect on uptake of public provision: each month post-reform is associated with an increase in the share of deliveries in public facilities by about eight tenths of a percentage point. Column (2) suggests that the reform lowered the propensity to deliver in the private sector by around four tenths of a percentage point per month. Both effects are significant at the one- percent level. Finally, column (3) indicates that the combined effect on the share of deliveries in either public or private facilities is small and statistically significant only at the ten-percent level. This suggests that the largest effect of the policy change was a shift in demand from private to public provision among those that would have given birth in a facility anyways. 250 The authors are aware of only one study that investigates the effect of this policy change in Kenya, Njuguna, Kamau, & Muruka (2017), and which finds a positive effect on the overall number of institutional deliveries. The study also finds a shift from private to public provision. However, the study lacks a clear identification strategy with regard to the first finding. 246 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices Table G.1: Regression results from LPMs – effect of free deliveries in public facilities on uptake by provider (N = 28,154) (1) (2) (3) (4) (5) (6) Public Private Any Public Private Any Reform X month (centered) 0.008*** -0.005*** 0.003* -0.001 -0.000 -0.001 (0.002) (0.001) (0.002) (0.003) (0.002) (0.003) Reform X month X primary 0.009** -0.002 0.006 (0.004) (0.003) (0.004) Reform X month X secondary or higher 0.014*** -0.012*** 0.002 (0.005) (0.004) (0.004) 0.093 0.148 0.248 0.094 0.149 0.249 Source: Own calculations based on 2014 KDHS data. Note: Significance level: 1% (***), 5% (**), and 10% (*). Standard errors clustered at the PSU-level reported in parentheses. All regressions include further controls (see text). Regressions in columns (4)-(6) include separate linear time trends for individuals with different levels of education. More educated mothers in urban areas were the most likely to switch from private to public provision. Who are those that switch? Columns (4) through (6) present results from models in which the treatment- month-interaction was further interacted with educational attainment, a proxy for poverty. This shows that better educated mothers switched from private to public provision. There is no evidence for an effect on the propensity to deliver in any formal facility in column (6). Table G.2: Regression results from LPMs – effect of free deliveries in public facilities on uptake by provider, urban and rural (N = 28,154) Urban Rural (1) (2) (3) (4) (5) (6) Reform X month (centered) -0.019* 0.013 -0.006 0.002 -0.002 -0.001 (0.009) (0.008) (0.008) (0.003) (0.002) (0.004) Reform X month X primary 0.030** -0.017 0.013 0.005 0.001 0.005 (0.011) (0.009) (0.009) (0.005) (0.002) (0.005) Reform X month X secondary or higher 0.037** -0.031** 0.006 0.005 -0.004 0.002 (0.011) (0.010) (0.008) (0.006) (0.004) (0.005) R-squared 0.057 0.125 0.124 0.127 0.078 0.207 Source: Own calculations based on 2014 KDHS data. Note: Significance level: 1% (***), 5% (**), and 10% (*). Standard errors clustered at the PSU-level reported in parentheses. All regressions include further controls (see text). Regressions in columns (4)-(6) include separate linear time trends for individuals with different levels of education. There is no effect in rural areas – the observed changes in the characteristics of mother across providers is driven entirely by urban residents – suggesting that cost is not a binding constraint to institutional delivery in Kenya. Splitting the sample by urban and rural dwellers, one finds that the change in the characteristics of mother across providers is only observed among mothers living in urban areas (Table G.2). This suggests that it is physical access or transport costs that are keeping prospective mothers from seeking deliveries in formal health facilities, not provider fees. The results may have additional implications that should be explored further: distributional consequences and the potential for improved oversight. First, while the policy change did not increase uptake overall and not among the poor, it may still be pro-poor insofar as the transfer also benefits poor households that would have delivered in a public facility even in the absence of the policy change. However, a fraction of the transfer is also captured by better-off mothers that would have delivered in private facilities in the absence of the policy KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 247 Appendices change. How the benefits are distributed and, thus, whether the policy change was pro-poor has not been explored but could potentially be calculated based on assumptions about the costs of deliveries in the private and the public sector prior to the reform. Second, greater use of public services by better educated individuals is sometimes argued to be associated with increasing demand for quality, improved oversight, and monitoring. Better educated individuals may be more empowered to demand quality services. Since this oversight would constitute a public good, the poor may stand to benefit from this increased demand. G.2. Effect of institutional delivery and skilled assistance on infant and neonatal mortality The share of neonatal deaths, i.e., death within the first month of life, accounts for an increasing share of all child deaths. While infant and child mortality rates decreased in recent years, the increase was more pronounced for the latter (Figure 7.5). The share of neonatal deaths in total under-five deaths increased from three in ten to four in ten between 2003 and 2014. Progress in reducing under-five mortality in Kenya further will thus depend on finding ways to effectively address neonatal mortality. While it is often assumed that births that are assisted by skilled health professionals, doctors, nurses, or midwifes, and deliveries in formal health facilities are safer, there is at best very mixed evidence to support this claim (Box 7.2). The DHS data provide information on deaths by age and circumstances around birth that can be exploited to estimate the effect of institutionalized delivery and assistance on mortality. Data from the 2014 KDHS was used to investigate the link between mortality and assistance at the time of birth. Logit models were estimated that relate child deaths within the first month of life (neonatal mortality) and the first twelve months of life (infant mortality) to binary indicators for assistance through a doctor, a nurse, or a midwife (vis-à-vis no assistance or assistance by somebody else) as well as the place of delivery (government hospital, public health center, public dispensary, mission hospital, or private hospital as opposed to private home).251 The sample was restricted to births that took place within the last five years but at least one month or twelve months ago for neonatal and infant mortality respectively. Births of multiples were excluded. Controls are the gender of the child, dummies for the month of birth and the order of birth, as well as dummies for educational attainment of the mother, household wealth quintile (based on the DHS asset index), age of the mother and age squared, urban/rural locality and county fixed effects. Table G.3: Effect of institutional delivery and assistance on neonatal mortality (odds ratios/t-values) (N = 19,080) (1) (2) Assisted by doctor, nurse, or midwife 0.91 (-0.48) Delivery in formal health facility 1.03 (0.14) Source: Own calculations based on 2014 KDHS data. Note: Significance level: 1% (***), 5% (**), and 10% (*). Standard errors clustered at the PSU-level reported in parentheses. All regressions include further controls (see text). An analysis based on observational data and retrospective reports is also subject to important limitations. Major concerns include selection of better-off mothers into formal health facilities or into deliveries assisted by skilled professionals and adverse selection into assistance by health professionals with higher levels of formal qualifications and higher-level facilities (Okeke and Chari 2016). The former problem may potentially be addressed by including controls such as household wealth, maternal education, etc. But there is no obvious way to address the latter problem with the data at hand. Note that the two will have opposite effects on estimates. 251 Other categories, such as “en route to provider” were discarded. 248 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices Another concern in the Kenyan context is to differentiate between the effects of assistance and the effects of institutionalized birth. Both indicators are highly correlated in Kenya, that is, assisted births typically only take place in formal facilities.252 This makes it impossible to disentangle their effects. Both variables were used in separate estimations. The null hypotheses of no effect of assistance by a doctor, a nurse, or a midwife and no effect of delivering in formal health facility could not be rejected at the 95-percent level of confidence. Results from logit estimates for neonatal mortality are reported in Table G.3. Results for infant mortality were similar qualitatively and were thus omitted for brevity. Neither assistance during delivery by a doctor, or midwife, nor delivery in a formal health facility was associated with a lower risk of neonatal mortality. 252 Only around two percent of the births in the dataset were either assisted but did not take place in a facility or were delivered in a facility without the assistance of a doctor, a nurse, or a midwife. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 249 Appendices APPENDIX H: CHAPTER 8 ADDITIONAL MATERIALS Figure H.1: Consumption levels of vulnerable households, relative to the poverty line: 2015/16 Source: Own calculations from KIHBS 2015/16. Figure H.2: The prevalence of shocks by poverty and vulnerability status: 2005/06 and 2015/16 2005/06 2015/16 60 60 50 50 40 40 Percent Percent 30 30 20 20 10 10 0 0 Economic Agricultural Health Other Economic Agricultural Health Other shock shock shock shock shock shock shock shock Total Non-poor Poor Vulnerable Total Non-poor Poor Vulnerable Source: Own calculations from KIHBS 2005/06 and KIHBS 2015/16. 250 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead Appendices Table H.1: Coping strategies by poverty status for agricultural households only: 2015/16 (%) (1) (2) Used savings 24.5 21.0 26.2 *** Send children to relatives 0.4 0.8 0.2 *** Sold assets 0.9 0.8 0.9 Sold farmland 0.4 0.4 0.4 Rented farmland 0.7 0.7 0.7 Sold animals 10.5 12.2 9.7 *** Sold more crops 2.7 1.9 3.1 *** Worked more 10.7 10.7 10.7 HH member started work 0.4 0.7 0.3 * Started business 2.1 1.4 2.4 ** Children worked 0.3 0.6 0.1 ** Migrated for work 1.7 1.8 1.6 Borrowed from relative 3.4 4.5 2.8 *** Borrowed from moneylender 0.6 0.7 0.5 Borrowed from formal institution 0.6 0.1 0.9 *** Help from church 0.6 1.3 0.2 *** Help from local NGO 0.1 0.2 0.0 * Help from Intl. NGO 0.9 1.7 0.5 *** Help from government 2.2 4.7 1.0 *** Help from family member 5.1 6.9 4.1 *** Reduced food consumption 13.7 17.9 11.6 *** Consumed less 6.3 6.8 6.1 Reduced non-food consumption 9.3 12.4 7.9 *** Spiritual help 5.3 4.8 5.5 Other coping strategy 4.8 4.5 5.0 Source: Own calculations from KIHBS 2015/16. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 251 REFERENCES Abdul Latif Jameel Poverty Action Lab. 2011. “The Price is Wrong.” J-PAL Policy Bulletin, April. Action Aid. 2013. Making Care Visible. Women’s Unpaid Care Work in Nepal, Nigeria, Uganda and Kenya. Johannesburg: Action Aid. ADB (Asian Development Bank). 2015. Balancing the Burden? Desk Review of Women’s Time Poverty and the Infrastructure in Asia and the Pacific. Manila, Philippines: ADB. Afrobarometer. 2009. “Popular Attitudes toward Democracy in Kenya: A Summary of Afrobarometer Indicators, 2003– 2008.” Aker, Jenny C., and Isaac M. Mbiti. 2010. “Mobile Phones and Economic Development in Africa.” Journal of Economic Perspectives 24 (3): 207–32. Alderman, H., and R. Yemtsov. 2012. “Productive Role of Safety Nets.” World Bank Social Protection and Labor Discussion Paper 1203. Ali, Daniel A., and Klaus Deininger. 2015. “Is There a Farm Size–Productivity Relationship in African Agriculture? Evidence from Rwanda.” Land Economics 91 (2), 2015: 317–343. Ali, Daniel A., Klaus Deininger, and Markus Goldstein. 2014. “Environmental and Gender Impacts of Land Tenure Regularization in Africa: Pilot Evidence from Rwanda.” Journal of Development Economics 110: 262–275. Álvarez, L. G., and H. Van Nieuwenhuyzen. 2016. Evaluation of the Kenya Hunger Safety Net Programme Phase 2: Study on Fiscal Space for Social Protection in Kenya. Oxford: Oxford Policy Management. Andrabi, Tahir, Jishnu Das, and Asim Ijaz Khwaja. 2017. “Report Cards: The Impact of Providing School and Child Test Scores on Educational Markets.” American Economic Review 104 (6): 1535–1563. Antwi, James, and David C. Phillips. 2013. “Wages and Health Worker Retention: Evidence from Public Sector Wage Reforms in Ghana.” Journal of Development Economics 102: 101–115. Asfaw, S., B. Davis, J. Dewbre, S. Handa, and P. Winters. 2014. “Cash Transfer Programme, Productive Activities and Labour Supply: Evidence from a Randomised Experiment in Kenya.” Journal of Development Studies 50 (8): 1172–1196. Ashraf, N., J. Berry, and J. M. Shapiro. 2010. “Can Higher Prices Stimulate Product Use? Evidence from a Field Experiment in Zambia.” American Economic Review 100 (5): 2383–2413. Assouad, L., L. Chancel, and M. Morgan. 2018. “Extreme Inequality: Evidence from Brazil, India, the Middle East and South Africa.” WID World Working Paper 2018/4. Barnett, B., C. Barrett, and J. Skees. 2008. “Poverty Traps and Index-based Risk Transfer Products.” World Development 36 (10): 1766–1785. Barr, Abigail, and Andrew Zeitlin. 2011. “Conflict of Interest as a Barrier to Local Accountability.” CSAE Working Paper WPS/2011-13 Barrett, Christopher B. 2008. “Smallholder Market Participation: Concepts and Evidence from Eastern and Southern Africa.” Food Policy 33 (4): 299–317. Barrett, Christopher B., et al. 2010. “Reconsidering Conventional Explanations of the Inverse Productivity–Size Relationship.” World Development 38 (1): 88–97. Barrientos, A. 2013. Social Assistance in Developing Countries. Cambridge: Cambridge University Press. Barton, Nicholas, Tessa Bold, and Justin Sandefur. 2017. “Measuring Rents from Public Employment: Regression Discontinuity Evidence from Kenya.” Center for Global Development Working Paper 457. Beck, T., and M. Fuchs. 2004. “Structural Issues in the Kenyan Financial System.” World Bank Policy Research Working Paper 3363. Beck, T., Haki Pamuk, Ravindra Ramrattan, and Burak Uras. 2015. “Mobile Money, Trade Credit, and Economic Development: Theory and Evidence.” CEPR Discussion Paper DP10848. Bedoya, Guadalupe, Amy Dolinger, Khama Rogo, Njeri Mwaura, Francis Wafula, Jorge Coarasa, Ana Goicoechead, and Jishnu Das. 2017. “Compliance with Infection Prevention and Control Practices in Primary Care Settings: A Cross- Sectional Study in 935 Kenyan Health Facilities.” Bulletin of the World Health Organization 95: 503–516. Beegle, K., and Luc Christiaensen, eds. Forthcoming. World Bank Accelerating Poverty Reduction in Sub-Saharan Africa. Washington, DC: World Bank. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 253 References Beegle, K., and Isis Gaddis. 2017. “Informal Employment in Africa: What Do We Know and What Can We do To Know More?” Paper presented at the 61st ISI World Statistics Congress in Marrakech. Blackden, Mark, and Quentin Wodon. 2006. “Gender, Time Use, and Poverty in Sub-Saharan Africa.” World Bank Working Paper 73. Bold, T., D. Filmer, E. Molina, and J. Svensson. 2017. “The Lost Human Capital: Teacher Knowledge and Student Achievement in Africa.” CEPR Discussion Papers 12956. Bold, T., D. Filmer, G. Marin, E. Molina, C. Rockmore, S. Brian, and W. Wane. 2017. “What Do Teachers Know and Do? Does It Matter? Evidence from Primary Schools in Africa.” World Bank Policy Research Working Paper 7956. Bold, Tessa, Mwangi Kimenyi, Germano Mwabu, Alice Ng’ang’a, and Justin Sandefur. 2013. “Scaling-up What Works: Experimental Evidence on External Validity in Kenyan Education.” Center for Global Development Working Paper 321. Bold, Tessa, Mwangi Kimenyi, Germano Mwabu, and Justin Sandefur. 2013. The High Return to Private Schooling in a Low-Income Country. London: International Growth Centre. ———. 2014. “Can Free Provision Reduce Demand for Public Services? Evidence from Kenyan Education.” World Bank Economic Review 29 (2): 293–326. Bold, Tessa, Mwangi S. Kimenyi, and Justin Sandefur. 2013. “Public and Private Provision of Education in Kenya.” Journal of African Economies 22 (AERC Supplement 2): ii39-ii56. Campos, Francisco, and Marine Gassier. 2017. “Gender and Enterprise Development in Sub-Saharan Africa. A Review of Constraints and Effective Interventions.” World Bank Policy Research Working Paper 8239. Campos, Francisco, Markus Goldstein, Laura McGorman, Ana Maria Munoz Boudet, and Obert Pimhidzai. 2015. “Breaking the Metal Ceiling: Female Entrepreneurs Who Succeed in Male-Dominated Sectors.” World Bank Policy Research Working Paper 7503. Cardno. 2018. “Final Report: Gender Assessment in the Oil and Gas Sector in Kenya.” Prepared for the Kenya Petroleum Technical Assistance Project. Cardno – Shaping the Future. Carter, M., and J. Maluccio. 2003. “Social Capital and Coping with Economic Shocks: An Analysis of Stunting of South African Children.” World Development 31 (7): 1147–1163. Carter, M., P. Little, T. Mogues, and W. Negatu. 2007. “Poverty Traps and Natural Disasters in Ethiopia and Honduras.” World Development 35 (5): 835-856. Center for Devolution Studies, Kenya School of Government. 2015. “Building Public Participation in Kenya’s Devolved Government.” Devolution Working Paper 1. Chakravarty, Shubha, Smita Das, and Julia Vaillant. 2017. “Gender and Youth Employment in Sub-Saharan Africa: A Review of Constraints and Effective Interventions.” World Bank Policy Research Working Paper 8245. Chaudhuri, S., J. Jalan, and A. Suryahadi. 2002. “Assessing Household Vulnerability to Poverty from Cross-sectional Data: A Methodology and Estimates from Indonesia.” Columbia University Department of Economics Discussion Paper Series 0102-52. Chaudhury, Nazmul, Jeffrey Hammer, Kremer Michael, Karthik Muralidharan, and F. Halsey Rogers. 2006. “Missing in Action: Teacher and Health Worker Absence in Developing Countries.” Journal of Economic Perspectives 20 (1): 91–116. Chayanov, Alexander V. 1926. The Theory of Peasant Economy. University of Wisconsin Press. Chen, Daniel L., Paul Glewwe, Michael Kremer, and Sylvie Moulin. 2001. Interim Report on a Teacher Attendance Incentive Program in Kenya. Cambridge: National Bureau of Economic Research. Cherondo, S. 2013. “How Much is Free? The Riddle of Kenya’s Health Care System.” Daily Nation, June 9. Christiaensen, L., Punam Chuhan-Pole, and Aly Sanoh. 2013. Africa’s Growth, Poverty and Inequality Nexus-Fostering Shared Prosperity. Washington, DC: World Bank. Christiaensen, L., and K. Subbarao. 2005. “Towards and Understanding of Household Vulnerability in Rural Kenya.” Journal of African Economies 14 (4): 520–558. Cowell, Frank, and Emmanuel Flachaire. 2002. “Sensitivity of Inequality Measures to Extreme Values.” SSRN Scholarly Paper ID 1094842. Social Science Research Network, Rochester, NY. Dang, H.-A. H., P. F. Lanjouw, and R. Swinkels. 2017. “Who Remained in Poverty, Who Moved Up, and Who Fell Down?” In Poverty Reduction in the Course of African Development, edited by M. Nissanke and M. Ndulo. Oxford: Oxford University Press. Daniels, Benjamin, Amy Dolinger, Guadalupe Bedoya, Khama Rogo, Ana Goicoechea, Jorge Coarasa, Francis Wafula, Njeri Mwaura, Redemptar Kimeu, and Jishnu Das. 2017. “Use of Standardised Patients to Assess Quality of Healthcare in Nairobi, Kenya: A Pilot, Cross-sectional Study with International Comparisons.” BMJ Global Health 2 (2): e000333. 254 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead References Das, Jishnu, and Jeffrey Hammer. 2014. “Quality of Primary Care in Low-Income Countries: Facts and Economics.” Annual Review of Economics 6 (1): 525–553. Das, Jishnu, Jeffrey Hammer, and Kenneth Leonard. 2008. “The Quality of Medical Advice in low-income countries.” Journal of Economic Perspectives 22 (2): 93–114. Demobynes, G., and S. K. Trommlerová. 2016. “What Has Driven the Decline of Infant Mortality in Kenya.” Economics & Human Biology 21: 17–32. Datt, G., and H. Hoogeveen. 2003. “El Niño or El Peso? Crisis, poverty and income distribution in the Philippines.” World Development 31 (7): 1103–1124. Datt, G., and Martin Ravallion. 1992. “Growth and Redistribution Components of Changes in Poverty Measures: A Decomposition with Applications to Brazil and India in the 1980s.” Journal of Development Economics 38 (2): 275–95. Davis, K., et al. 2012. “Impact of Farmer Field Schools on Agricultural Productivity and Poverty in East Africa.” World Development 40 (2): 402–413. De Laat, J., Michael Kremer, and C. Vermeersch. 2008. “Local Participaton and Teacher Incentives: Evidence from a Randomized Experiment.” Harvard University Working Paper. Deere, Carmen D., and Cheryl R. Doss. 2006. “The Gender Asset Gap: What Do We Know and Why Does It Matter.” Feminist Economics 12 (1–2): 1–50. Demirguc-Kunt, Asli, Leora Klapper, Dorothe Singer, and Peter Van Oudheusden. 2015. “The Global Findex Database 23015: Measuring Financial Inclusion Around the World.” World Bank Policy Research Working Paper 7255. Dercon, S. 2004. “Risk, Insurance and Poverty: A Review.” In Insurance Against Poverty. Oxford: Oxford University Press. Dercon, S., J. Hoddinott, and T. Woldehanna. 2005. “Shocks and Consumption in 15 Ethiopian Villages, 1999–2004.” Journal of African Economies 14 (4): 559. DHS (Demographic and Health Survey) StatCompiler. 2018. https://www.statcompiler.com/en/ Dizon, Felipe, Erick Gong, and Kelly Jones. 2017. “The Effect of Promoting Savings on Informal Risk-Sharing: Experimental Evidence from Vulnerable Women in Kenya.” Djuikom, Marie Albertine, and Dominique van de Walle. 2018. “Marital Shocks and Women’s Welfare in Africa.” World Bank Policy Research Working Paper 8306. Doss, Cheryl, Caitlin Kieran, and Talip Kilic. 2017. “Measuring Ownership, Control and Use of Assets.” World Bank Policy Research Working Paper 8146. Duflo, Esther, Pascaline Dupas, and Michael Kremer. 2015a. “Education, HIV, and Early Fertility: Experimental Evidence from Kenya.” American Economic Review 105 (9): 2757–97. ———. 2015b. “School Governance, Teacher Incentives, and Pupil-teacher Ratios: Experimental Evidence from Kenyan Primary Schools.” Journal of Public Economics 123: 92–110. ———. 2013. “Savings Constraints and Microenterprise Development: Evidence from a Field Experiment in Kenya.” American Economic Journal: Applied Economics 5 (1): 163–92. Evans, D, M. Kremer, and M. Ngatia. 2008. The Impact of Distributing School Uniforms on Children’s Education in Kenya. Washington, DC: World Bank. Fischer, Elisabeth, and Matin Qaim. 2012. “Linking Smallholders to Markets: Determinants and Impacts of Farmer Collective Action in Kenya.” World Development 40 (6): 1255–268. Fiszbein, A., and N. Schady. 2009. Conditional Cash Transfers: Reducing Present and Future Poverty. World Bank Policy Research Report. Washington, DC: World Bank. Friedman, W., and A. Keats. 2017. “What Can We Learn from Babies Born During Health Worker Strikes?” Unpublished manuscript. Friedman, Willa, Michael Kremer, Edward Miguel, and Rebecca Thornton. 2016. “Education as Liberation?” Economica 83 (329): 1–30. Gaddis, Isis, Rahul Lahoti, and Wenjie Li. 2018. “Gender Gaps in Property Ownership in Sub-Saharan Africa.” World Bank Policy Research Working Paper WPS 8573. Gettleman, Jeffrey. 2007. “Disputed Vote Plunges Kenya in Bloodshed.” New York Times December 31. Glewwe, Paul, Nauman Ilias, and Michael Kremer. 2010. “Teacher Incentives.” American Economic Journal: Applied Economics 2: 205–227. Glewwe, Paul, and Hanan G. Jacoby. 1995. “An Economic Analysis of Delayed Primary Enrollment in a Low Income Country: The Role of Early Childhood Nutrition.” Review of Economics and Statistics 77 (1): 156–169. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 255 References Glewwe, Paul, Michael Kremer, Sylvie Moulin, and Eric Zitzewitz. 2004. “Retrospective vs. Prospective Analyses of School Inputs: The Case of Flip Charts in Kenya.” Journal of Development Economics 74: 251–268. Goldstein, Markus, and Christopher Udry. 2008. “The Profits of Power: Land Rights and Agricultural Investment in Ghana.” Journal of Political Economy 116 (6): 981–1022. Government of Kenya. 2007. Kenya Vision 2030. The Popular Version. Nairobi: Government of the Republic of Kenya. ———. 2010. Constitution of Kenya. Nairobi: Government of Kenya. ———. 2012. Kenya Social Protection Review. Nairobi: Ministry of State for Planning, National Development and Vision 2030. Grepin, Karen, James Habyarimana, and William Jack. 2017. “Cash on Delivery: Results of a Randomized Experiment to Promote Maternal Health Care in Kenya.” GUIDE Working Paper Series 3. Günther, I., and K. Harttgen. 2009. “Estimating Households Vulnerability to Idiosyncratic and Covariate Shocks: A Novel Method Applied in Madagascar.” World Development 37 (7): 1222–1234. Hallward-Driemeier, Mary. 2013. Enterprising Women. Expanding Economic Opportunities in Africa. Africa Development Forum. Washington, DC: Agence Française de Développement and World Bank. Halvorsen, Robert, and Raymond Palmquist. 1980. “The Interpretation of Dummy Variables in Semilogarithmic Equations.” American Economic Review 70 (3): 474–475. Handa, S., C. Halpern, A. Pettifor, and H. Thirumurthy. 2014. “The Government of Kenya’s Cash Transfer Program Reduces the Risk of Sexual Debut among Young People Age 15–25.” PLoS ONE 9 (1). Hanushek, Eric A., and Ludger Woessmann. 2015. The Knowledge Capital of Nations. Cambridge: MIT Press. Harrington, Andrew, and Tanja Chopra. 2010. Arguing Traditions: Denying Kenya’s Women Access to Land Rights. Justice for the Poor Research Report 2/2010. Washington, DC: World Bank. Healthy Nation. 2017. “What National Insurer Can Do For You Now.” Daily Nation, April 18 . Hicks, Joan H., Michael Kremer, Isaac Mbiti, and Edward Miguel. 2016. Evaluating the Impact of Vocational Education Vouchers on Out-Of-School Youth in Kenya. 3ie Impact Evaluation Report 37. New Delhi, India: International Initiative for Impact Evaluation (3ie). Human Rights Watch. 2008. “Ballots to Bullets, Organized Political Violence and Kenya’s Crisis of Governance.” IFC (International Finance Corporation). 2017. Tackling Childcare: The Business Case for Employer-Supported Childcare. Washington, DC: World Bank. ———. 2018. Driving Toward Equality: Women, Ride-Hailing and the Sharing Economy. Washington, DC: World Bank. Jack, William, and James Habyarimana. 2018. “High Hopes: Experimental Evidence on Saving and the Transition to High School in Kenya.” GUIDE Working Paper Series 4. Jack, William, Adam Ray, and Tavneet Suri. 2013. “Transaction Networks: Evidence from Mobile Money in Kenya.” American Economic Review 103 (3): 356–61. Jack, William, and Tavneet Suri. 2014. “Risk Sharing and Transactions Costs: Evidence from Kenya’s Mobile Money Revolution.” American Economic Review 104 (1): 183–223. Jacoby, H., and E. Skoufias. 1997. “Risk, Financial Markets, and Human Capital in a Developing Country.” Review of Economic Studies 64 (3): 311–335. Jakiela, Pamela, and Owen Ozier. 2016. “Does Africa Need a Rotten Kin Theorem? Experimental Evidence from Village Economies.” Review of Economic Studies 83 (1): 231–68. Jann, Ben. 2008. “The Blinder-Oaxaca Decomposition for Linear Regression Models.” Stata Journal 8 (4): 453–79. Kassie, Menale, et al. 2011. “Agricultural Technology, Crop Income, and Poverty Alleviation in Uganda.” World Development 39 (10): 1784–1795. Kenya CT-OVC Evaluation Team. 2012. “The Impact of the Kenya Cash Transfer Program for Orphans and Vulnerable Children on Household Spending.” Journal of Development Effectiveness 4 (1): 9–37. KNBS (Kenya National Bureau of Statistics). 2005 through 2015 editions. Economic Survey. Nairobi: Kenya National Bureau of Statistics. ———. 2012. “2009 Kenya Population and Housing Census: Analytical Report on Mortality.” KNBS and Ministry of State for Planning, National Development and Vision 2030, Nairobi. ———. 2016. Kenya Economic Survey. Nairobi: Kenya National Bureau of Statistics. Kenya National Bureau of Statistics et al. 2010. Kenya Demographic and Health Survey 2008-09. Nairobi: Kenya National Bureau of Statistics. 256 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead References ———. et al. 2014. Kenya Demographic and Health Survey 2014. Nairobi: Kenya National Bureau of Statistics. Kevane, Michael. 2004. Women and Development in Africa: How Gender Works. Boulder, CO: Lynne Rienner Publishers, Inc. Kikulwe, Enoch M., Elisabeth Fischer, and Matin Qaim. 2014. “Mobile Money, Smallholder Farmers, and Household Welfare in Kenya.” PLoS ONE 9 (10). Kimenyi, Mwangi S., and Josephine Kibe. 2014. Africa’s Powerhouse. Washington, DC: The Brookings Institution. Kiplang’at, J. 2016. “Kenya: How TSC Will Appraise Teachers’ Performance.” Daily Nation. Klasen, S., and J. Pieters. 2015. “What Explains the Stagnation of Female Labor Force Participation in Urban India?” World Bank Economic Review 29 (3): 449–78. Klasen, S., and F. Povel. 2013. “Defining and Measuring Vulnerability: State of the Art and New Proposals.” In Vulnerability to Poverty, edited by S. Klasen and H. Waibel. London: Palgrave Macmillan. Klasen, S., and H. Waibel. 2015. “Vulnerability to Poverty in South-east Asia: Drivers, Measurement, Responses, and Policy Issues.” World Development 71: 1–3. Klugman, Jeni, Lucia Hanmer, Sarah Twigg, Tazeen Hasan, Jennifer McCleary-Sills, and Julieth Santamaria. 2014. Voice and Agency: Empowering Women and Girls for Shared Prosperity. Washington, DC: World Bank. Korinek, Anton, Johan A. Mistiaen, and Martin Ravallion. 2006. “Survey Nonresponse and the Distribution of Income.” Journal of Economic Inequality 4 (1): 33–55. Kremer, M., E. Miguel, S. Mullainathan, C. Null, and A. P. Zwane. 2011. Social Engineering: Evidence from a Suite of Take-up Experiments in Kenya. Atlanta: Emory University. Kremer, M., E. Miguel, and R. Thornton. 2009. “Incentives to Learn.” Review of Economics and Statistics 91 (1): 437–56. Kruk, Margaret E., Emily Goldmann, and Sandro Galea. 2009. “Borrowing and Selling to Pay for Health Care in Low-and Middle-income Countries.” Health Affairs 28 (4): 1056–1066. Loayza, Norman V., and Claudio Raddatz. 2010. “The Composition of Growth Matters for Poverty Alleviation.” Journal of Development Economics 93 (1): 137–151. Lucas, Adrienne M., and Isaac M. Mbiti. 2012a. “Does Free Primary Education Narrow Gender Differences in Schooling? Evidence from Kenya.” Journal of African Economies 21 (5): 691–722. ———. 2012b. “Access, Sorting, and Achievement: The Short-Run Effects of Free Primary Education in Kenya.” American Economic Journal: Applied Economics. 4 (4): 226–253. ———. 2014. “Effects of School Quality on Student Achievement: Discontinuity Evidence from Kenya.” American Economic Journal: Applied Economics 6 (3): 234–263. Lundgren, Rebecka, Miranda Beckman, Surendra Prasad Chaurasiya, Bhawna Subhedi, and Brad Kerner. 2013. “Whose Turn to Do the Dishes? Transforming Gender Attitudes and Behaviours Among Very Young Adolescents in Nepal.” Gender and Development 21: 127–45. Martin, G. H., and O. Pimhidzai. 2013. Service Delivery Indicators: Education and Health. Washington, DC: World Bank. Matata, L. 2015. “New Secondary School Fee Structure Gazetted, Day Scholars to Pay Sh9,374 and Sh53,554 for National.” The Star, March 11 . Mbiti, Isaac, and David Weil. 2011. “Mobile Banking: The Impact of M-PESA in Kenya.” National Bureau of Economic Research Working Paper Series 17129. McConnell, Margaret, Allison Ettenger, Claire Watt Rothschild, Faith Muigai, and Jessica Cohen. 2016. “Can a Community Health Worker Administered Postnatal Checklist Increase Health-seeking Behaviors and Knowledge?: Evidence from a Randomized Trial with a Private Maternity Facility in Kiambu County, Kenya.” BMC Pregnancy and Childbirth 16: 136. McKenzie, David, and Susana Puerto. 2017. “Growing Markets through Business Training for Female Entrepreneurs. A Market-Level Randomized Experiment in Kenya.” World Bank Policy Research Working Paper 7993. McKenzie, David, and Christopher Woodruff. 2014. “What Are We Learning from Business Training Evaluations Around the Developing World?” World Bank Research Observer 29 (1): 48–82. Medeiros, M., Juliana de Castro, and Luisa de Azevedo. 2016. “Correcting the Underestimation of Top Incomes: Combining Data from Income Tax Reports and the Brazilian 2010 Census.” Social Indicators Research 135 (1): 233–244. Melesse, Mequaint B., Adane Dabissa, and Erwin Bulte. 2018. “Joint Land Certficiation Programmes and Women’s Empowerment: Evidence from Ethiopia.” Journal of Development Studies 54 (10): 1756–74. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 257 References Meredith, Jennifer, Jonathan Robinson, Sarah Walker, and Bruce Wydick. 2013. “Keeping the Doctor Away: Experimental Evidence on Investment in Preventative Health Products.” Journal of Development Economics 105: 196–210. Merotto, Dino, Fayavar Hayati, David Stephan, and William Bataille. 2015. “Dismal Science, Accounting and Newton’s Second Law. Identifying Force and Rigidity in Public Expenditure Analysis.” World Bank Policy Research Working Paper 7431. Merttens, F., A. Hurrell, M. Marzi, R. Attah, M. Farhat, A. Kardan, and I. MacAuslan. 2013. Kenya Hunger Safety Net Programme Monitoring and Evaluation Component: Impact Evaluation Final Report: 2009 to 2012. Oxford: Oxford Policy Management. Ministry of Health. 2013. The State of the Health Referral System in Kenya: Results from a Baseline Study on the Functionality of the Health Referral System in Eight Counties. Nairobi: Ministry of Health. ———. 2017. Kenya National Health Accounts 2015/16. Nairobi: Ministry of Health. Minten, Bart, and Christopher B. Barrett. 2008. “Agricultural Technology, Productivity, and Poverty in Madagascar.” World Development. 36 (5): 797–822. Morawczynski, Olga, and Mark Pickens. 2009. “Poor People Using Mobile Financial Services: Observations on Customer Usage and Impact from M-PESA.” CGAP Brief. World Bank, Washington, DC. Morduch, J. 1999. “Between the State and the Market: Can Informal Insurance Patch the Safety Net?” World Bank Research Observer 14 (2): 187–207. Moulin, Sylvie, Michael Kremer, and Paul Glewwe. 2009. “Many Children Left Behind? Textbooks and Test Scores in Kenya.” American Economic Journal: Applied Economics 1 (1): 112–135. Munoz Boudet, Ana Maria, Paola Buitrago, Benedicte Leroy de la Briere, David Newhouse, Eliana Rubiano Matulevich, Kinnon Scott, Pablo Suarez-Becerra. 2018. “Gender Differences in Poverty and Household Composition through the Life-cycle.” World Bank Policy Research Working Paper 8306. Muraguri, Nicholas. 2015. “Key Findings of the Demographic Health 2014 Survey.” PowerPoint presentation. Kenya Ministry of Health, Nairobi. National Social Protection Secretariat. 2018. Retrieved January 19, 2018. www.socialprotection.go.ke. Ngaware, Moses M., Moses Oketch, and Maurice Mutisya. 2014. “Does Teaching Style Explain Differences in Learner Achievement in Low and High Performing Schools in Kenya.” International Journal of Educational Development 36: 3–12. NHIF (National Hospital Insurance Fund). 2015. “Benefit Package: Explanation of the Benefit Package for the National Scheme.” Nairobi: National Hospital Insurance Fund. Nyabola, Nanjala. 2017. “Why did Kenya’s Supreme Court annul the elections?” Al Jazeera, September 2. O’Sullivan, Michael. 2017. “Gender and Property Rights in Sub-Saharan Africa. A Review of Constraints and Effective Interventions.” World Bank Policy Research Working Paper 8250. O’Sullivan, Michael, Arathi Rao, Raka Banerjee, Kajal Gulati, and Margaux Vinez. 2014. Levelling the Field: Improving Opportunities for Women Farmers in Africa. Washington, DC: World Bank and ONE Campaign. Okeke, Edward N., and A. V. Chari. 2016. “Can Institutional Deliveries Reduce Newborn Mortality? Evidence from Rwanda.” RAND Working Paper Series WR-1072. Olewe, Dickens. 2017. “What Next in Kenya Election Crisis?” BBC, October 11. Ozier, Owen. 2016. “The Impact of Secondary Schooling in Kenya: A Regression Discontinuity Analysis.” Policy Research Working Paper WPS 7384. World Bank Group, Washington, DC. Plummer, David. 2010. “Is Learning Becoming Taboo for Caribbean Boys?” In Challenging HIV & AIDS: A New Role for Caribbean Education, edited by Michael Morrissey, Myrna Bernard, and Donald Bundy. Paris: United Nations Educational, Scientific and Cultural Organization (UNESCO) and Ian Randle Publishers, 174–183. Plyler, Megan G., Sherri Haas, and Geetha Nagarajan. 2010. “Community-Level Economic Effects of M-PESA in Kenya: Initial Findings.” Financial Services Assessment Project. University of Maryland, College Park, Maryland. Pontefract, Caroline, and Frank Hardman. 2005. “The Discourse of Classroom Interaction in Kenyan Primary Schools.” Comparative Education 41 (1): 87–106. Priebe, Jan, and Johannes Gräb. 2009. “Low Malnutrition but High Mortality: Explaining the Paradox of the Lake Victoria Region.” Proceedings of the German Development Economics Conference. Rao, Elizaphan J.O., and Matin Qaim. 2011. “Supermarkets, Farm Household Income, and Poverty: Insights from Kenya.” World Development 39 (5): 784–96. 258 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead References Ravallion, Martin. 1994. “Measuring Social Welfare with and Without Poverty Lines.” American Economic Review 84 (2): 359–364. Republic of Kenya. 2015. “Law of Succession Act.” Revised Edition of 2015. Reuters. 2017. “Kenyan Opposition Leader to Challenge Election Result in Court.” The Guardian, August 16. Rodriguez-Oreggia, E., A. de la Fuente, R. de la Torre, and H. Moreno. 2013. “Natural Disasters, Human Development and Poverty at the Municipal Level in Mexico.” Journal of Development Studies 49 (3): 442–455. Romero, Mauricio, Justin Sandefur, and Wayne Aaron Sandholtz. 2017. “Outsourcing Service Delivery in a Fragile State: Experimental Evidence from Liberia.” Center for Global Development Working Paper No 462. Salon, Deborah, and Sumila Gulyani. 2010. “Mobility, Poverty and Gender: Travel ‘Choices’ of Informal settlement Residents in Nairobi, Kenya.” Transport Reviews 30 (5): 641–57. Sandefur, J. 2018. “Internationally Comparable Mathermatics Scores for Fourteen African Countries.” Econonomics of Education Review 62: 267–286. ———. 2017. “The Cost of Convenience? Transaction Costs, Bargaining Power, and Savings Account Use in Kenya.” Journal of Human Resources 52 (4): 919–45. Slavchevska, Vanya, Ana Paula de la O Campos, Chiara Brunelli, and Cheryl Doss. 2017. “Beyond Ownership: Women’s and Men’s Land Rights in Sub-Saharan Africa.” Paper presented at the 2017 World Bank Conference on Land and Poverty. Rome: Food and Agriculture Organization of the United Nations. Smoke, Paul. 2012. “Design and Management of Decentralized and Intergovernmental Revenues.” In Decentralization and the Changing Role of Central Finance Agencies: Forum Report and Expert Papers. Manilla: Asia–Pacific and Latin American Interregional Forum on Managing for Results. Sundet, Geir Scanteam, and Eli Moen. 2009. Political Economy Analysis of Kenya. Norwegian Agency for Development Cooperation Report 19/2009. Suri, Tavneet. 2017. “Mobile Money.” Annual Review of Economics 9 (1): 497–520. Suri, Tavneet, and William Jack. 2016. “The Long-run Poverty and Gender Impacts of Mobile Money.” Science 354 (6317): 1288–92. Teachers Service Commission. 2007. Revised Scheme of Service for Non-Graduate Teachers. Thome, K., M. Filipski, J. Kagin, J. E. Taylor, and B. Davis. 2013. “Agricultural Spillover Effects of Cash Transfers: What Does LEWIE Have to Say?” American Journal of Agricultural Economics 95 (5): 1338–1344. Tiwari, S., S. Daidone, M. A. Ruvalcaba, E. Prifti, S. Handa, B. Davis, O. Niang, L. Pellerano, P. Q. van Ufford, and D. Seidenfeld. 2016. “Impact of Cash Transfer Programs on Food Security and Nutrition in Sub-Saharan Africa: A Cross- country Analysis.” Global Food Security 11: 72–83. USAID (U.S. Agency for International Development). 2016. Measurement and Research Support to Education Strategy Goal I: Boys’ Underachievement in Education: A Review of the Literature with a Focus on Reading in the Early Years. Research Triangle Park, NC: RTI International. Uwezo. 2016. Are Our Children Learning? Uwezo Kenya Sixth Learning Assessment Report. Nairobi: Twaweza East Africa. Wafula, Francis, Amy Dolinger, Benjamin Daniels, Njeri Mwaura, Guadalupe Bedoya, Khama Rogo, Ana Goicoechea, Jishnu Das, and Bernard Olayo. 2017. “Examining the Quality of Medicines at Kenyan Healthcare Facilities: A Validation of an Alternative Post-market Surveillance Model that Uses Standardized Patients.” Drugs-real World Outcomes 4 (1): 53–63. Wafula, Francis, et al. 2017. “Examining the Quality of Medicines at Kenyan Healthcare Facilities: A Validation of an Alterantive Post-Market Surveillance Model That Uses Standardized Patients.” Drugs Real World Outcomes 4 (1): 53-63. Wakaba, Mabel, Patrick Mbindyo, Jacob Ochieng, Rose Kiriinya, Jim Todd, Agnes Waudo, Abdisalan Noor, Chris Rakuom, Martha Rogers, and Mike English. 2014. “The Public Sector Nursing Workforce in Kenya: A County-level Analysis.” Human Resources for Health 12 (1): 6. Wamalwa, Fredrick M., and Justine Burns. 2017. “Private Schools and Student Learning Achievements in Kenya.” ERSA Working Paper 689. Wanzala, O. 2016. “Teachers Attaining Higher Qualifcation Won’t Be Automatically Promoted.” Daily Nation,October 30. Ward, P., A. Hurrell, A. Visram, N. Riemenschneider, L. Pellerano, C. O’Brien, I. MacAuslan, and J. Willis. 2010. Cash Transfer Programme for Orphans and Vulnerable Children (CT-OVC), Kenya: Operational and Impact Evaluation, 2007– 2009. Oxford: Oxford Policy Management. Wattanga, Humphrey. 2015. “Perspectives on Impact Bonds: Working around Legal Barriers to Impact Bonds in Kenya to Facilitate Non-state Investment and Results-based Financing of Non-state ECD Providers.” Blog post, December 21. KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead 259 References WHO (World Health Organization). 2018. “Kenya Country Profile – Malaria.” WHO, UNICEF, UNFPA, World Bank Group, and the United Nations Population Division. 2015. Trends in Maternal Mortality: 1990 to 2015. Geneva: World Health Organization. Wineman, A., N. Mason, J. Ochieng, and L. Kirimi. 2016. “Let it Rain: Weather Extremes and Household Welfare in Rural Kenya.” Michigan State University Food Security Collaborative Working Papers 245109. World Bank. 2007. Malawi: Poverty and Vulnerability Assessment. Washington, DC: World Bank. ———. 2008. Kenya Poverty and Inequality Assessment: Volume I, Synthesis Report. Poverty Reduction and Economic Management Unit. Washington, DC: World Bank. ———. 2011. World Development Report 2012: Gender Equality and Development. Washington, DC: World Bank. ———. 2012. Devolution Without Disruption: Pathways to a Successful New Kenya. Washington, DC: World Bank. ———. 2013a. Republic of Kenya: Kenya Gender Policy Note. Report No: ACS5140. Agriculture and Rural Development Unit (AFTA2). Washington, DC: World Bank. ———. 2013b. A Background Brief on Kenya’s Devolution. Washington, DC: World Bank. ———. 2013c. World Development Report 2014: Risk and Opportunity - Managing Risk for Development. Washington, DC: World Bank. ———. 2014a. Decision Time: Spend More or Spend Smart? Kenya Public Expenditure Review, Volume I. Nairobi: World Bank. ———. 2014b. Laying the Foundation for a Robust Health Care System in Kenya, Kenya Public Expenditure Review, Volume II. Nairobi: World Bank. ———. 2014c. SABER Country Report. Washington, DC: World Bank. ———. 2015a. World Bank Group Gender Strategy (FY16–23): Gender Equality, Poverty Reduction and Inclusive Growth (English). Washington, DC: World Bank Group. ———. 2015b. Women, Business and the Law 2016: Getting to Equal. Washington, DC: World Bank. ———. 2015c. The State of Social Safety Nets 2015. Washington, DC: World Bank. ———. 2017a. Kenya Tax Policy Studies: Value Added Tax and Corporate Income Tax. Washington, DC: World Bank. ———. 2017b. World Development Indicators. ———. 2018a. Sustainable Financing of Vertical Programs. Washington, DC: World Bank. ———. 2018b. Kenya Economic Update: Policy Options to Advance the Big 4. Washington, DC: World Bank Group. ———. 2018c. World Development Indicators. Yonga, Paul O., Henry M. Muchiri, and Andrew W. Onyino. 2012. “Perceptions of Medical Brain Drain among Young Doctors in Kenya: A Cross-sectional Survey.” The Lancet 380: S22. Zumbyte, Leva. 2018. Kenya DHS 2014 Analysis of Gender Based Violence. Washington, DC: World Bank. 260 KENYA POVERTY AND GENDER ASSESSMENT 2015/16 · Reflecting on a Decade of Progress and the Road Ahead World Bank Group Delta Center Menengai Road, Upper Hill P. O. Box 30577 – 00100 Nairobi, Kenya Telephone: +254 20 2936000 Fax: +254 20 2936382 http://www.worldbank.org/en/country/kenya