BOTSWANA POVERTY ASSESSMENT RENEWING PATHWAYS for Poverty and Inequality Reduction © [2024] International Bank for Reconstruction and Development / The World Bank 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy, completeness, or currency of the data included in this work and does not assume responsibility for any errors, omissions, or discrepancies in the information, or liability with respect to the use of or failure to use the information, methods, processes, or conclusions set forth. 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. Nothing herein shall constitute or be construed or considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. 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. Any queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org Cover design: Florencia Micheltorena BOTSWANA POVERTY ASSESSMENT RENEWING PATHWAYS for Poverty and Inequality Reduction Contents Executive Summary vii 1 Botswana’s macroeconomic context and recent labor market outcomes 1 1.1  Macroeconomic context and growth slowdown 2 1.2  Recent labor market trends 5 2 Progress in poverty reduction has slowed, and inequality remains high 12 2.1  Trends in poverty, inequality, and shared prosperity 13 2.2  Demographic, education, and labor market characteristics of the poor, near poor, and non-poor 19 2.3  Drivers of poverty and inequality 25 3 Non-monetary dimensions of poverty and inequality 29 3.1  Access to electricity and sanitation at the village level is low and highly unequal 30 3.2  Access to basic services increased in 2016 but some services remain low for an upper-middle-income country 32 3.3  Multidimensional poverty 38 3.4  Poorest locations face higher risk from natural hazards 41 4 Social protection in COVID times: Responses and lessons learned 44 4.1  The pre-COVID-19 social protection system 45 4.2  Social protection policy responses to COVID-19 46 4.3  Assessment of responses 48 4.4  COVID-19 and longer-term reforms 50 5 Policy considerations for poverty and inequality reduction 51 Appendixes 57 Appendix 1  Structural Peers 58 Appendix 2  Population Pyramid 2022 59 Appendix 3  Probability of Being Employed 60 Appendix 4  Earnings regression 61 Appendix 5  Population distribution from Population and Housing Census 2001, 2011, 2022 62 Appendix 6  Maps of villages by deciles 63 Appendix 7  Redefining Employment in Botswana 64 Appendix 8  Technical report on Botswana’s household surveys and the creation of a comparable spatially-deflated consumption aggregate 67 Appendix 9  Botswana SWIFT Poverty Projections for 2019-2022 82 List of Boxes Box 1.1  Comparable labor market statistics 8 Box 2.1  Data challenges, comparable poverty estimates, and poverty projections 14 Box 2.2  Poverty Map 2011 20 Box 3.1  Electrification Rate Maps 2012-2019 33 Box 3.2  The Human Opportunity Index 36 Box 3.3  Multidimensional Poverty Measure 40 List of Figures Figure E.1  Poverty reduction has slowed viii Figure E.2  Economic growth has slowed and become more volatile ix Figure E.3  Poverty is high relative to income ix Figure E.4  Rural households fared worse than urban across the consumption distribution x Figure E.5  Botswana faces high levels of multidimensional poverty relative to some peers xiii Figure E.6  Its Human Capital Index is low relative to its GDP per capita, 2020 xiii Figure E.7  Electrification rates have improved more in the northeast and southeast, while gaps between rich and poor villages remain large xiv Figure 1.1  Economic growth has slowed and become more volatile 2 Figure 1.2  Wholesale and retail trade has seen strong growth but remains far behind mining and public administration value added 3 Figure 1.3  Reaching high-income status requires a significant, sustained boost in economic growth 5 Figure 1.4  GDP per capita is higher than among structural peers but below the UMIC average 5 Figure 1.5  Labor Market Trends, 2002-2021Q4 (%) 6 Figure 1.6  Labor Market Trends, 2010-2022Q4 (%) 7 Figure B1.1  Comparable labor force statistics show increasing labor force participation and unemployment rates, while employment rates stagnate 8 Figure 1.7  Lack of transition out of agriculture while employment in services has grown 9 Figure 1.8  Employment in Agriculture, Forestry and Fishing (2016-2022) 10 Figure 1.9  Employment and Value Added (2022) 10 Figure 1.10  Share of workers by type of employment, 2009-2022 (%) 11 Figure 1.11  Employment by type of worker, 2009-2022 11 Figure 1.12  Public-private wage gap, 2016-2022 (ratio) 11 Figure 1.13  Average monthly real wages by sector and area, 2016-2022 (constant 2010 Pula) 11 Figure 2.1  Poverty reduction has slowed 13 Figure 2.2  The link between poverty and GDP growth became weaker 13 Figure 2.3  Botswana is poorer than most of its structural peers 15 Figure 2.4  Poverty is high relative to income level 15 Figure 2.5  Strong declines in urban poverty yet rural poverty increased 16 Figure 2.6  The intensity of poverty improved across strata (Poverty Gap by area, %) 16 Figure 2.7  The poorest of the poor were better off (Poverty Severity by area, %) 16 Figure 2.8  Rural households fared worse than urban households across the income distribution 17 Figure 2.9  Botswana is no longer one of Africa’s top performers on shared prosperity indicator 17 Figure 2.10  Fewer people lived below the $2.15 poverty line in 2016 but more lived below $6.85 17 Figure 2.11  Inequality improved across Botswana, with better outcomes in cities and towns 18 Figure 2.12  Botswana’s inequality is among the highest in the world 18 Figure 2.13  Regional poverty is diverging, increasing in the west and decreasing in the east 19 Figure B2.1  2011 Poverty Map by Village Deciles 21 Figure B2.2  The poorest 50 villages were primarily in the Southern district while the least poor included Gaborone and Francistown and their surroundings, 2011 21 Figure B2.3  The four poorest deciles had the largest rural shares, 2011 22 Figure 2.14  Poverty Rate by Household Size (%) 24 Figure 2.15  Poverty Rate by Education Level of Household Head (%) 24 Figure 2.16  Poverty Rate by Age Group (%) 24 Figure 2.17  Poverty Rate by Labor Force Status of Household Head (%) 24 Figure 2.18  Employment by Sector and Poverty Status (%) 24 Figure 2.19  Decomposition of the change in poverty into growth, distribution, and price effects, National, 2003-2009 and 2009-2016 (percentage points) 25 Figure 2.20  Decomposition of the change in poverty into growth, distribution, and price effects, by area, 2009-2016 (percentage points) 25 Figure 2.21  Shapely decomposition of changes in poverty by income source (2009-2016) 26 Figure 2.22  Decomposition of Inequality 27 Figure 2.23  Decomposition of Inequality by Income Source in Botswana at National level and for Rural Households (%) 27 Figure 2.24  Marginal effect of income source change 27 Figure 2.25  Gini Coefficient by Country and Income Source 28 Figure 3.1  Access to electricity and improved sanitation is highly unequal while access to improved water is more universal (2011 Census) 30 Figure 3.2  Poor villages had much lower access to electricity and improved sanitation than wealthy ones, whereas access to improved water was almost universal 31 Figure 3.3  Poor households face significant differences in access to electricity and sanitation 32 Figure B3.1  Electrification rates are higher and have improved more in the northeast and southeast 33 Figure B3.2  Electrification Rate by 2011 Village Poverty Deciles, 2012-2015-2019 34 Figure 3.4  Expected Years of Schooling 35 Figure 3.5  Human Capital Index and log GDP per capita, 2020 36 Figure 3.6  Human Capital Index, select countries, 2020 36 Figure B3.3  Human Opportunity Index, 2016 37 Figure B3.4  Human Opportunity Index, 2016, by Area 37 Figure B3.5  HOI Decomposition, 2003-2009-2016 38 Figure B3.6  D-index Decomposition, 2003-2009-2016 38 Figure 3.7  Multidimensional Poverty Measure, Deprivation by Dimension, 2009 and 2016 (%) 39 Figure 3.8  Deprivation in Access to Services by Area, 2016 (%) 39 Figure 3.9  Individuals in households deprived in monetary poverty and overall MPM 39 Figure 3.10  Palmer Severity Drought Index (PDSI) and Village 2010 Poverty Rate by Deciles, 2009-2021 41 Figure 3.11  Palmer Severity Drought Index (PDSI) for 2015/16 by Village 2010 Poverty Rate 42 Figure 3.12  Average annualized rainfall shocks by village poverty decile, 2009-2019 43 List of Tables Table 1.1  Labor force growth is outpacing employment growth 6 Table 2.1  The number of poor increased in the west while declining in the Central and South-East districts 20 Table 2.2  Profile of the Poor: Official Poverty Line 23 Table B3.1  Electrification rate for 2012 - 2019 by 2011 Village Poverty Deciles 34 Table B3.2  Multidimensional Poverty Measure Indicators, Weights, and Thresholds 40 Table 4.1  Overview of social cash transfer programs in Botswana 45 Table 4.2  Overview of Botswana’s social protection responses to COVID-19 49 BOTSWANA POVERTY ASSESSMENT v Acknowledgments The Botswana Poverty Assessment was coauthored by Carolina Diaz-Bonilla (Team Lead), Santiago Garriga, and Giselle del Carmen. The extended team included Ugo Gentilini and Victoria Monchuk on social protection (Chapter 4); Ifeanyi Nzegwu Edochie and Daylan Salmeron Gomez on geospatial data and village-level census data; and Danielle Aron and Nobuo Yoshida on poverty projections. The report was finalized under the guidance of Country Director Marie Francoise Marie-Nelly; Acting Country Director and Operations Manager Asmeen Khan; Practice Manager Pierella Paci; and Botswana Resident Representative Liang Wang. The report benefited greatly from collaboration with Statistics Botswana. The team would particularly like to thank Dr. Burton Mguni (then-Statistician General) and Mr. Moffat Malepa (Manager, Labour and Poverty Statistics), who were instrumental in supporting and facilitating the work. The report benefited from feedback provided at different stages by Javier Baez, the Botswana Country Team, and the following peer reviewers: Nga Thi Viet Nguyen (Senior Economist, EECPV), Monica Robayo (Senior Economist, EECPV), and Sonia Araujo (Senior Economist, EAEM2). Logistical assistance was provided by Tsehaynesh H Michael Seltan during the preparation of this report. Florencia Micheltorena created the graphic design. vi BOTSWANA POVERTY ASSESSMENT Abbreviations and Acronyms BCWIS Botswana Core Welfare Indicators Survey BMTHS Botswana Multi-Topic Household Survey BURS Botswana United Revenue Service CEM Country Economic Memorandum CEQ Commitment to Equity COICOP Classification of Individual Consumption According to Purpose EMIS Educational Management Information SYstems FIES Food Insecurity Experience Scale GDP gross domestic product GMD Global Monitoring Database GNI gross national income GRIDMET Gridded Surface Meteorological HCI Human Capital Index HIES Household Income and Expenditure Survey HREA High-Resolution Electricity Access HOI Human Opportunity Index ICT information and communication technology ICLS International Conference of Labour Statisticians IFC International Finance Corporation ILO International Labour Organization ISSF Informal Sector Stimulus Fund LEA Local Enterprise Authority MPM Multidimensional Poverty Measure MSMEs micro-, small, and medium enterprises PDSI Palmer Drought Severity Index PPP purchasing power parity QMTS Quarterly Multi-Topic Household Survey SACU Southern African Customs Union SCD Systematic Country Diagnostic StatsBots Statistics Botswana SWIFT Survey of Well-being via Instant and Frequent Tracking TVET Technical or Vocational Education Training UMIC upper middle income country UNESCO United Nations Educational, Scientific, and Cultural Organization UNICEF United Nations Children’s Fund VIIRS Visible Infrared Imaging Radiometer Suite WHO World Health Organization All dollar amounts are US dollars unless otherwise indicated. BOTSWANA POVERTY ASSESSMENT vii EXECUTIVE SUMMARY Botswana’s strong growth since independence had significantly improved living standards B otswana’s fast diamond-based growth, significant public investment, and political stability catapulted it into a stable upper-middle-income country and improved living standards. A large, sparsely populated, land-locked country in Southern Africa, Botswana has significant mineral wealth and a relatively small and young population (2.3 million).1 The discovery of diamond deposits turned it into one of the world’s fastest-growing economies into the 1990s, with average annual GDP and GDP per capita growth above 10 and 7 percent, respectively. The growth model relies heavily on diamonds (representing around 90 percent of total goods exports, followed by nature-based luxury tourism) and a large public sector, while close to one-fifth of employment is in agriculture (half subsistence farming) and two-thirds in services. Historically, the government invested diamond revenues well to improve infrastructure and human capital, significantly expanding the road network, access to electricity, water, and sanitation, and primary school enrollment. This significantly improved life expectancy, mortality rates, nutrition, and poverty. The country maintained stability thanks to sustainable macroeconomic and fiscal policies and repeatedly ranked among the top African performers across many governance indicators. Despite its income status, the country faces multiple data challenges that limit its ability to make informed and effective policy decisions. The challenges include the availability, quality, and use of data, and a weak statistical infrastructure. Botswana does not conduct frequent income and expenditure surveys to track poverty. In the last twenty years, only three such surveys have been conducted: 2002/03, 2009/10, and 2015/16.2 Efforts were made to create comparable consumption data3 (see Appendix 8); however, income data is less reliable, creating a challenge in understanding income dynamics. Data on informality is also limited despite its importance and the need for further research. Ministries and line agencies are also not capturing information regularly to inform progress and better outcomes and lack integrated data systems. To fill some data gaps, the country began collecting quarterly multi-topic household surveys in 2019 that capture primarily labor force data and some rotating modules. Using the available data, this Poverty Assessment presents poverty estimates up to 2016, preliminary poverty projections based on imputations into the 2019-2022 quarterly data (see Appendix 9), and recent labor market trends (using a longer comparable series, see Appendix 7). It also uses satellite and geospatial data to understand recent trends in electricity access and rainfall shocks (see Chapter 3) after the negative poverty impact of the 2015 drought and electricity and water crises. Looking ahead, a recently completed Population and Housing Census and the upcoming 2024/25 income and expenditure survey will fill other important data gaps. Yet, evidence-based policymaking requires Botswana to act on its commitment to invest in frequent, timely, and relevant data across its statistical system and strengthen its monitoring and evaluation. 1  See Appendix 2 for Botswana’s projected 2022 population pyramid. 2  The surveys are the 2002/03 Household Income and Expenditure Survey (HIES) (June 2002-August 2003; Central Statistics Office 2004), the 2009/10 Botswana Core Welfare Indicators Survey (BCWIS) (April 2009-March 2010; Statistics Botswana 2013), and the 2015/16 Botswana Multi-Topic Household Survey (BMTHS) (November 2015-December 2016; Statistics Botswana 2018). This document refers to these surveys as 2003, 2009, and 2016, corresponding to the year with the most survey-month coverage. 3  Statistics Botswana and the World Bank have updated Botswana’s official poverty measurement methodology using international best practice. See Appendix 8 for a technical overview of the updated measures. viii BOTSWANA POVERTY ASSESSMENT Progress in poverty reduction and shared prosperity has slowed, and inequality remains high Poverty reduction and employment growth slowed in 2016. In contrast to the rapid decrease in the share of the population living below the official poverty line in 2003-09 (from 31 to 19.3 percent), poverty declined more slowly to 16.1 percent between 2009-16 (Figure E.1). In 2016 over 330,000 Batswana were living in poverty. Despite better GDP per capita growth between 2009-16 than 2003-09, the ability of economic growth to lead to poverty reduction declined in 2016 and requires further investigation.4 Although inequality declined (see below), this was due more to real consumption declines among FIGURE E.1 wealthier households and weaker pro-poor growth FIGURE E.1  Poverty reduction has slowed than in the past. Unlike in the past, GDP growth (based on low-employment diamond wealth and a large 35 31 public sector) did not translate to higher average (real) 30 household per capita consumption between 2009-16. Instead, the major regional drought in 2015 affected 25 19 the 2016 harvest, while electricity and water shortages 20 16.1 18.0 15.6 further limited private sector growth, resulting in lower 15 average household consumption than in 2009, higher 10 14.0 unemployment rates, employment declines in rural 5 areas, and increased rural poverty. Both employment 0 and labor force growth slowed between 2009 and 2003 2009 2016 2017 2018 2019Q3 2019Q4 2020Q1 2020Q4 2021Q4 2022Q4 2016, with labor force growth outpacing job creation (2.0 versus 1.8 percent, respectively; see Chapter 1). Official Full Model (Quarterly) Job creation in 2009-16 was driven by public sector Limited Model (Quarterly) Elasticity (Annual) expansion from local government, parastatals, and the Source: 2002/03 BHIES, 2009/10 BCWIS, 2015/16 BMTHS Ipelegeng public works program. The growth model (Stats Botswana 2013, 2018; also see Appendix 8); 2017- 2022 projections based on Quarterly Multi-Topic Household was less able to create sufficient jobs, with larger Surveys and national accounts GDP data. Note: For projection impacts on the youth. methodologies see Appendix 9. Poverty reduction likely continued to slow in recent years. Poverty projections up to 2022 using different methodologies suggest poverty reduction has slowed even further due to the weaker labor market (Figure E.1, Appendix 9). More generally, private sector participation in non-mineral exports and transformative sectors has been limited, restricting diversification and job creation over the longer period. In addition, since 2015, the country has faced floods in 2017, another drought in 2019, the 2020 COVID-19 shock, and a surge in prices in 2022, in addition to sluggish productivity growth and an economic decline in South Africa, Botswana’s main trade partner. Botswana has been increasingly facing less robust and more volatile economic growth (Figure E.2) and higher unemployment due to increased competition from synthetic diamonds, higher local production costs, lack of diversification, and low job creation. By 2016, Botswana was no longer one of Africa’s top performers on the shared prosperity indicator, while poverty remained high given its income level. Consumption per capita among the poorest 40 percent of the population only grew by an annualized 1.2 percent in 2009–16, down from 5 percent in 2003–09, while it declined among the top 60 percent of the population.5 The latter negative consumption growth increased poverty from 60.4 to 63.5 percent (in 2016) using the global poverty line for upper-middle-income countries ($6.85 per day, 2017 PPP). Households above the official poverty line but below the $6.85 poverty line represent more than 4  The elasticity of GDP per capita to poverty fell from -3.8 in 2003-2009 to -0.6 in 2009-2016, a decline of 84 percent. Alternatively, using GDP data post the 2009 recession, the elasticity fell from -2.1 in 2003-2010 to -0.86 in 2010-2016, a decline of 60 percent. 5  Shared prosperity, which measures the extent to which economic growth is inclusive, is expressed as the annualized growth rate in the average consumption per capita of the poorest 40 percent of the population. Real consumption per capita overall for Botswana declined by an annualized -1.8 percent in 2009-2016. BOTSWANA POVERTY ASSESSMENT ix FIGURE E2 FIGURE E.2  Economic growth has slowed and become more volatile 30 20 10 0 -10 -6.7 -10.4 -15.9 -20 1971 1974 1977 1980 1983 1986 1989 1992 1995 1998 2001 2004 2007 2010 2013 2016 2019 2022 GDP growth (annual %) GDP per capita growth (annual %) 2015 PA New PA Source: World Development Indicators (database), databank.worldbank.org/source/world-development-indicators, version 7/25/2023 half the population and faced welfare declines, on E3  Poverty is high relative to income FIGUREE.3 FIGURE average, in 2016. In addition, more than 40 percent of employed Batswana in 2016 were considered poor $2.15 poverty and GDP per capita (2017 PPP) when using the higher global poverty line and require 90 Extreme poverty rate (percentage) a different set of policies for poverty alleviation than 80 people in extreme poverty. Lastly, despite a decline 70 in poverty under the $2.15 PPP International Poverty 60 Line from 17.7 to 15.4 percent, the (extreme) poverty 50 rate remains four times higher than GDP per capita 40 would predict (Figure E.3). 30 20 BWA Botswana remains among the world’s top 10 10 0 most unequal countries, dampening its growth 6.5 7.5 8.5 9.5 10.5 potential. Consumption inequality, measured by the LN (GDP per capita, PPP, US$)* Gini Coefficient, decreased from 60.5 in 2009 to 54.9 Source: World Development Indicators and pip.worldbank.org percent in 2016 (using official figures). This primarily reflected pro-poor growth in urban areas, especially cities and towns, where the Gini declined from 59.9 to 48.7 percent, with Francistown reaching 43.4 percent. Inequality remains high in rural areas, declining less than four percentage points to 53.3 percent in 2016, with urban villages at 51.1 percent. However, the decline in rural inequality also reflects consumption declines among wealthier rural households. Moreover, among its structural peers6, only South Africa and Namibia register worse inequality rates, and Botswana remains exceptionally unequal by international standards. International experience suggests that Botswana would find it hard to reach an inclusive growth path and high-income status without addressing these levels of inequality and poverty. Poverty is increasingly concentrated in rural areas while improving in urban areas, increasing the urban-rural gap Between 2009 and 2016, both the rural poverty rate and the share of the poor in rural areas increased, while urban poverty declined, widening the urban-rural gap. In this period, poverty in cities and towns and in urban villages declined to 3.3 and 13.7 percent, respectively. However, poverty in rural villages increased from 6  For Botswana, eight countries were identified as structural peers: Gabon, Georgia, Lebanon, Mongolia, Namibia, the Republic of Congo, South Africa, and Tunisia. See Appendix 1. x BOTSWANA POVERTY ASSESSMENT FIGURE E.4 FIGURE E4  Rural households fared worse than urban across the consumption distribution a. Rural growth incidence curve (%) a. Urban growth incidence curve (%) 10 10 Annualized Growth Rate (%) Annualized Growth Rate (%) 8 8 6 6 4 4 2 2 0 0 -2 -2 -4 -4 -6 -6 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Percentile Percentile 2009-2003 2016-2009 2009-2003 2016-2009 Source: 2002/03 Household Income and Expenditure Survey (HIES), 2009/10 Botswana Core Welfare Indicators Survey (BCWIS), and 2015/16 Botswana Multi-Topic Household Survey (BMTHS). The variable in the horizontal axis divides individuals into ventiles, from lowest to highest consumption per capita. 24.4 to 26.8 percent, leaving rural poverty rates two and a half times higher than in urban areas. Although the share of all Batswana living in rural areas declined from 43 to 34.9 percent, the share of poor Batswana living in rural areas increased from 56.2 to 58.2 percent. While the population of Botswana (per the household survey) increased by 10.6 percent between 2009-16, the working-age population by 9.6 percent, and the under-15 population by 12.7 percent, the corresponding rural populations instead declined by 10.2, 10.4, and 9.9 percent, respectively, suggesting entire families migrated to urban areas. The economic shocks, however, resulted in more substantial rural employment declines (14 percent) than its population. Poverty increased in the remote North-West, Ghanzi, and Kgalagadi districts while declining in and around cities to the south- and northeast (Chapter 2). In terms of consumption per capita growth, rural households across the distribution fared much worse than between 2003-09 and worse than urban households (Figure E.4). The working poor were primarily employed in services (public administration and as household workers) while the share employed in agriculture declined by more than half in 2016. Approximately half of rural public administration workers were low-paying Ipelegeng workers, suggesting the working poor pushed out of agriculture and other private sectors by the 2015 shocks sought income via the Ipelegeng program (or as household domestic workers). Droughts continue to negatively impact poor rural households. They are the second most common natural disaster to impact Botswana (Chapter 3). In 2016, almost 28 percent of Botswana’s rural labor force worked in the agricultural sector, a sector whose productivity is sensitive to changes in climatic conditions, especially due to Botswana’s challenges with water scarcity. In 2015 and 2019, Botswana suffered severe droughts, with another, but less severe, drought impacting villages in 2020.7 Although drought hit most villages in 2015, it produces more damage among poorer villages where subsistence farmers entirely depend on agriculture. The 2019 drought was similar to 2015, suggesting that subsistence farmers may have again faced declines in welfare in 2019. Although there is no household survey with consumption data available to measure poverty in that year directly, the poverty projections undertaken via survey-to-survey imputations suggest an increase in poverty in 2019. However, concerns with droughts go beyond the vulnerability of subsistence farmers; Botswana’s tourism depends on water-based wildlife and is considered the most important service sector export. 7  See the strong negative values of the Palmer Drought Severity Index (PDSI) in chapter 3. The PDSI is created by the Gridded Surface Meteorological (GRIDMET) at 4km resolution to show driest (-5) to wettest areas (+5). BOTSWANA POVERTY ASSESSMENT xi Labor income and education remain key for reducing poverty and inequality Labor income was the main driver of poverty reduction in urban areas between 2009 and 2016, while a decline in the share of employed adults contributed to higher rural poverty. A decomposition of changes in poverty by components of household income shows that 80 percent of the reduction in urban poverty came from improvements in labor income, with non-labor income and a higher share of adults in the household (a lower dependency ratio) also contributing. In rural areas, the decline in the share of employed adults in households outweighed the beneficial impacts of factors such as lower inequality, higher labor and nonlabor incomes, and demographic changes (a lower dependency ratio). Social protection programs in Botswana have a significant impact. Without Botswana’s social protection transfers (from 29 programs across nine ministries), the poverty rate of 16 percent in 2016 would have been almost 24 percent, and the poverty gap would have been 9.5 instead of 4.6 percent. Removing just the thirteen social assistance programs would increase the poverty headcount by nearly a third (to 22.8 percent). The primary school feeding program, which reaches approximately 269,000 children, and the old-age pension, which reaches more than 126,000 people (about 5.5 percent of the population in fiscal year 2020), have the largest impact on poverty but low levels of expenditure (Chapter 4).8 Social protection can also contribute to human capital outcomes, resilience, and shock response. However, during the 2020 crisis, the government did not supplement existing social grants or introduce new emergency social grant programs but instead oversaw an extensive once-off emergency feeding program. It was also quick to reallocate funds and introduce wage subsidies for formally employed workers (Chapter 4). Reducing inequality is also key and requires strengthening labor market skills, reducing public-private wage differentials, and improving educational attainment.9 Labor market factors10 contributed the most (35.6 percent) to inequality in 2016, followed by demographics (27.9 percent), education (25.9 percent), and location (10.6 percent). In contrast, differences in educational attainment were the most important drivers of overall inequality for other countries in the Southern Africa Customs Union. Among the labor market factors, occupation type (professionals and senior managers, which suggest differences in skills or abilities) explains 29 percent of total inequality relative to only 6 percent for labor force participation. Post-secondary education, with its high earnings, explains 24 percent of total inequality, followed by differences in age (17.4 percent) and location (10.6 percent). Unlike in other SACU countries, location increased in importance as a source of inequality in 2016, explained primarily by a divergence in inequality across regions in Botswana. Among income sources, wage income inequality was the main contributor to inequality in 2016, suggesting declines in inequality may be partially explained by smaller wage gaps.11 Wage income accounts for 85 percent of inequality (both at the national level and for rural areas), a rate higher than the 72.3 percent average for SACU members. A marginal change in wage income inequality is estimated to change the Gini Coefficient by 5.2 percent at the national level and 8.3 percent in rural areas. Economic growth does not support sufficient job creation The current growth model does not support intensive and high-productivity job creation, while labor force growth continues to outpace employment growth. The government’s aim of poverty elimination and high-income-country status by 203612 requires a significant and sustained boost in economic growth and employment and, therefore, substantial reforms. Botswana ranks in the bottom 30 percent of countries 8  For more detail on Botswana’s social protection programs, see World Bank (2022c). 9  The results from a decomposition of inequality by spatial, demographic, education, and labor market dimensions, as well as separately by income sources, are used to shed light on the drivers of inequality in Southern Africa (Sulla et al., 2022). 10  Labor market factors include labor force status (whether people work or not), industry of employment, and occupation type. 11  Business income, social protection transfers, and remittances make up the rest of the income sources. 12  Government of Botswana. 2016. Vision 2036: Achieving Prosperity for All. xii BOTSWANA POVERTY ASSESSMENT worldwide in terms of employment ratios (at 131 out of 187 countries).13 Labor demand bottlenecks limit job creation, and labor supply constraints result in skills mismatches. Recent labor market data for 2016-21 shows that employment continues to grow slower than the labor force (3.3 versus 4.5 percent). As of 2022, services accounted for two-thirds of jobs14, primarily public administration and wholesale and retail trade. The public sector15 remains the country’s largest employer, with 27 percent of total employment using international estimates (compared to a global average of 17 percent) and 47 percent of formal employment (compared to 38 percent globally).16 Structural unemployment continues to be high, with increasing levels in recent years, while the labor market continues to reflect substantial barriers to the inclusion of women. National unemployment is higher than in all other upper-middle-income countries except South Africa; it increased from 17.6 percent in 2016 to 22.7 percent by 2022 year-end (using comparable unemployment measures).17 Significant rural- urban migration has deepened the problem in urban areas. Unemployment is highest in urban villages, among women, and among the 15-24 age group.18 As of 2022, labor force participation rates were higher for men (69 percent) than women (59 percent). Moreover, women in the labor force earn less than men19, account for over a third of the agricultural workforce, and own more informal agricultural businesses than men, underlining their less secure economic standing. Even though Botswana has the second-highest rate of female entrepreneurship globally, women continue to face significant challenges in the business environment, including a lack of access to finance, assets, skills, education, training, and networks.20 Moreover, as the primary caregivers in their families, women bear the responsibility of caring for sick relatives, often at the expense of pursuing economic, educational, or training opportunities. For example, women faced more substantial negative employment impacts in 2020. Multidimensional measures of poverty suggest faster improvements but large gaps remain The Multidimensional Poverty Measure declined more strongly than poverty, yet further investments, particularly among rural households, are required to reach the country’s structural peers. The World Bank’s MPM for Botswana dropped nearly 11 percentage points, from 31.8 percent in 2009 to 21.1 percent in 2016. Rural areas experienced more than three times the deprivation rates in access to electricity (64.8 percent) compared to urban areas (19.6 percent). Similarly, nearly two-thirds of rural households did not have access to improved sanitation in contrast to 45 percent of urban households. Cross-country comparisons for Botswana show similar rates of multidimensional poverty relative to Namibia and South Africa but large gaps with other structural as well as aspirational peers (Figure E.5).21 13  International Labour Organization. 2023. Employment to population ratio, 15+, total (%). https://ilostat.ilo.org/data/. 14  Services account for 526,000 out of 788,000 total jobs; where total jobs includes subsistence farmers. 15  The term “public sector” includes central and local government (18.2 percent in 2022), parastatals (2.4 percent), and Ipelegeng workers (6.6 percent). It is larger than “public administration” as it also includes public wage workers in education, health, and other sectors. Ipele- geng is a workfare program for adults 18 years or older and is limited to only one-month employment. 16  Worldwide Bureaucracy Indicators dataset (version 3.0). August 2022. https://www.worldbank.org/en/data/interactive/2019/05/21/worldwide-bureaucracy-indicators-dashboard#2 17  See Box 1.1 and Appendix 7. The official unemployment rate for Botswana in the fourth quarter of 2022 is 25.4 percent, applying the standard of the Nineteenth International Conference of Labour Statisticians (ICLS). This standard introduced a new classification for “forms of work”: own-use production work (e.g., subsistence farming) is now a separate indicator and no longer counts towards employ- ment or labor force participation. This change means that labor force statistics are no longer comparable with those of earlier years. To maintain comparability with official unemployment rates from 2002 to 2016, the previous ICLS standards were applied here, giving an unemployment rate of 22.7 percent in 2022. 18  Among Botswana’s three strata, unemployment was 26.3 percent in urban villages, 21.8 in rural villages, and 15.8 percent in cities and towns. It was 25 percent for women versus 20.5 percent for men in 2022, and 43.6 percent among the 15-24 age group versus 20.9 and 10.9 percent for the 25-54 and 55-64 age groups, respectively. 19  Botswana’s gender pay gap remains statistically significant after controlling for differences in human capital accumulation, age, geo- graphic location, and sector of work. Over the last decade, on average, male earnings have been 39 percent higher than female earnings. On the other hand, Inequality in Southern Africa (World Bank, 2022) shows that when occupation is also taken into account, the earnings gap in 2016 is estimated to be around 24 percent. 20  Rudhumbu, du Plessis, and Maphosa (2020); Mastercard (2022); and forthcoming World Bank Gender Assessment (2024). 21  The MPM takes into account different dimensions of poverty and provides a way for policymakers to monitor improvements in this broader concept of welfare. BOTSWANA POVERTY ASSESSMENT xiii FIGURE E5 FIGURE E.5  Botswana faces high levels of multidimensional poverty relative to some peers 45 42 40 Monetary Poverty MPM 35 30 28 25 22 21 20 15 10 2 8 5 35 5 16 15 21 2 2 0 Chile 1 Costa Rica 1 Estonia 1 Lebanon 1 Seychelles 1 Malaysia 0 Mauritius 0 Botswana Congo, Rep South Africa Namibia Georgia Gabon Mongolia Tunisia Source: World Bank Global Monitoring Database (GMD). Multidimensional Poverty Measure, April 2023. https://www.worldbank.org/ en/topic/poverty/brief/multidimensional-poverty-measure FIGURE E6 Despite advancements in education and health, FIGURE E.6  Its Human Capital Index is low Botswana’s Human Development results are relative to its GDP per capita, 2020 disappointing for a country with its income and characteristics. The Human Capital Index (HCI)22 0.9 slightly increased from 0.37 in 2010 to 0.41 in 2020, Human Capital Index, 2020 0.8 placing it just above the average HCI score for Sub- 0.7 Mongolia Saharan Africa (0.40) and considerably lower than 0.6 the average for Upper-Middle-Income Countries Georgia 0.5 Lebanon Tunisia Gabon (0.56) (Fig E.6). This indicates that a Botswana child 0.4 Namibia South Africa is only 41 percent as productive as an adult as he Botswana could have been had he received complete education 0.3 and full health.23 Moreover, the Human Opportunity 0.2 Index (HOI) also shows that children in Botswana 6.5 7 7.5 8 8.5 9 9.5 10 10.5 11 11.5 12 face stark differences in life prospects depending Natural logarithm GDP per capita, 2020 (PPP, US$) on their circumstances at birth and during their Source: Human Capital Project (2020). Note: The Human Capital Index is designed to capture the amount of human capital a child early years (Chapter 3). Access to basic goods and born today could expect to attain by age 18. The HCI is higher on services is highly relevant to children’s development, average in rich countries than poor countries and ranges from around 0.3 to around 0.9. The units of the HCI have the same yet access remains largely unequal between urban interpretation as the components measured in terms of relative and rural areas and far from universal in some productivity. cases. Household per capita income, dwelling location, and parental education are the most important factors determining whether a child has access to essential childhood opportunities. Between 2012 and 2019, electrification was gradual and remained unequal across villages, with more robust progress concentrated around Gaborone and Francistown (Figure E.7).24 The electricity shortage in 2015 seems to have negatively impacted the poorest 40 percent of villages, which had estimated electrification rate declines in that year. Poverty rates among rural households also increased in 2016. In 2019, despite growth in the estimated electrification rate across all village poverty deciles, the gap between the poorest and wealthiest villages in terms of electrification remained large. 22  The Human Capital Index (HCI) measures the expected productivity as a future worker of a child born today. It is a function of educa- tion and health, underscoring their importance for the productivity of people. It ranges between 0 and 1, where 1 indicates the benchmark of complete education and full health. 23  World Bank. World Development Indicators. 24  To estimate electrification rates across Botswana beyond the latest available census (2011), high-res Visible Infrared Imaging Radiom- eter Suite (VIIRS) satellite data from the High-Resolution Electricity Access (HREA) project was used. xiv BOTSWANA POVERTY ASSESSMENT FIGURE E.7  Electrification rates have improved more in the northeast and southeast, while gaps betweenE7 FIGURE rich and poor villages remain large Electrified & Non-Electrified village in 2019 Electrification rate by 2011 Village Poverty Deciles 80 Electrification Rate (%) 60 poverty 35 30 40 25 20 15 20 10 0 1 2 3 4 5 6 7 8 9 10 Village Poverty Deciles 2012 2015 2019 Source: World Bank calculations using Min and O’Keeffe (2021) dataset for electricity access and 2009/10 BCWIS for poverty by district. Note: White circles represent electrified villages; yellow circles represent unelectrified villages. Renewing pathways for poverty and inequality reduction Renewed policy efforts will be needed for Botswana to reach its goals of poverty eradication and high- income status. The suggested policy considerations are summarized in four areas. First, accelerating inclusive economic growth based on dynamic private-sector-led job creation is crucial to increasing the income of people in poverty. Second, further investments in quality human capital among the poor are essential to improve welfare and boost workforce productivity. Third, additional investments in infrastructure and shock-responsive systems, especially in rural areas, are required to better connect and protect the most vulnerable population. Fourth, strengthening data is critical for evidence-based policy design that provides better outcomes for all. Accelerating inclusive economic growth and private-sector-led job creation For stronger poverty and inequality reduction, Botswana needs to transition towards a more diversified and inclusive growth model that supports private sector job creation and is more resilient to shocks. The country’s fiscal vulnerabilities, weak economic diversification, and high inequality, plus the declining ability of economic growth to reduce poverty, require a competitive, export-oriented private sector that maximizes economic inclusion. The Country Economic Memorandum (World Bank, 2024a) proposes three guiding criteria to help select priority sectors more systematically. The sectors should be: 1) labor–intensive and tradable to create productive jobs for more people and foster innovation; 2) globally competitive given Botswana’s small and undiversified economy; and 3) strategic, such as Botswana’s potential in renewable energy, base minerals, and eco-tourism. Promoting competition and expanding formal and informal employment opportunities will also require reducing the excessive public sector footprint that creates barriers to entry, discourages diversification, and causes significant economic inefficiencies. It will require facilitating external trade in goods and services and implementing adequate foreign investment rules to improve competitiveness and generate good jobs. As structural transformation takes time, it will be important to boost the productivity of farm and non- farm household enterprises and create jobs for the large unskilled population. Policies that could boost farm productivity for the 10 percent of farm self-employed include investments in agricultural research and extension, irrigation (given the high vulnerability to droughts), and rural infrastructure, and efforts to bring poor farmers into the value chain more effectively. Livestock-related services need to reach the small-scale BOTSWANA POVERTY ASSESSMENT xv sector and poor subsistence farmers so that they can also benefit from Botswana’s niche value chains in beef and livestock.25 For non-farm household enterprises, policies to boost productivity include increasing access to credit (and building skills; see next section). The IFC Country Private Sector Diagnostic estimates the financing gap of small enterprises to be 19 percent of GDP. The limited ability to access available credit limits the ability of the sector to create employment. Improving human capital particularly for the poorest Improving the welfare of the poor requires substantial improvements in human capital to increase the productivity of the labor force. The country needs to redouble efforts toward more efficient investments in health and education to enhance the productivity of the current and next generation, particularly the poor.26 The quality of education services is an important constraint, requiring policies addressing the coordination of service delivery amongst several Ministries, the shortage of classrooms and learning materials, giving teachers adequate training and coaching in the classroom, and fine-tuning education assessment systems to track progress and facilitate corrective action.27 The allocation of resources across education and training subsectors needs to be more pro-poor since the poor learn less and are more likely to drop out of education and training. As the data shows, Botswana’s high degree of childhood inequality of opportunity constrains the upward mobility of the poor.28 For health challenges among the poor, policies should focus on strengthening the quality of health service delivery, including improving clinical guidelines and protocols, enhancing staff core competencies, and improving the availability of essential medicines.29 Strengthening skills training and qualifications and better coordinating and monitoring employment programs are also crucial for boosting the productivity of poor and vulnerable workers. The many existing technical programs and trainings for skills development are fragmented and uncoordinated and lack systematic evaluations of interventions. To better support youth transitions into productive (formal or self-) employment also requires stronger coordination of programs. This coordination requires significant collaboration across ministries and better partnerships with the private sector. In addition, the focus needs to be on skills with a high growth potential, such as digital and green skills. Even in urban areas, a lack of job creation and skills mismatches hamper the ability of disadvantaged households to generate income, and high wage inequality fuels overall inequality. Investing in infrastructure and shock-responsive systems to connect and protect Increasing access to quality infrastructure and basic services increases the productive capacity of the most vulnerable population. Large gaps remain in access to electricity and sanitation between rural and urban areas and between poor and non-poor households. Where access is high, such as for water, the quality of services is the important constraint. Promoting the inclusion of the rural population requires making connectivity (electricity, digital) affordable and reliable. To reach its goal of universal access to electricity, Botswana will need to mainstream off-grid solutions to deal with the deep spatial inequalities. Workers need support to develop new skills to reap the benefits of digitization. Investments to improve water service quality and increase access to sanitation are also priorities in rural areas, but also in urban areas as urbanization 25  For a discussion of sustainable livestock value chains, see Syed and others (2022). World Bank (2023) also highlights that Botswana’s livestock smallholders, lacking the scale to access large markets, could benefit from collaboration with Namibia on veterinary services, traceability, transboundary animal disease surveillance and control, and research and development. 26  Botswana’s Human Capital Index is low relative to its GDP and despite high public expenditure in health and education. Nevertheless, a fiscal policy analysis via the Commitment to Equity (CEQ ) methodology showed education and health transfers contributing strongly to inequality reduction in 2009/10 (World Bank, 2022; Lusting and Higgins, 2016). 27  See the Botswana Systematic Country Diagnostic Update (World Bank, 2023) 28  See Human Opportunity Index (chapter 3) while Sulla et al. (2021) shows that inherited circumstances account for 20 percent of Bo- tswana’s inequality. In addition, the forthcoming Africa Poverty and Inequality Report highlights intergenerational mobility in education for African countries is well below that of the developing world on average. 29  A forthcoming Public Expenditure Review in Health will dive more deeply into policies to strengthen health outcomes. xvi BOTSWANA POVERTY ASSESSMENT continues. Per the Updated Systematic Country Diagnostic (SCD Update): “The interlinked challenges of water scarcity, limited wastewater treatment, the under-provision of sanitation services, and malnutrition continue to undermine people’s health, limit the development of their skills, and hinder their productive participation in the labor force” (World Bank, 2023). Investments in the capacity of social protection systems to respond to shocks while improving the targeting of safety net programs protect the poorest. Although Botswana responded better on some dimensions to the COVID-19 pandemic than many other African countries, the pandemic exposed significant gaps in its social protection system’s shock responsiveness and resilience. Botswana needs to develop productive and shock- responsive social protection systems to mitigate shocks (including droughts and pandemics) and disaster vulnerability. Despite Botswana’s social protection programs’ significant impact on reducing poverty in 2016, they were insufficient to assist many poor rural households. Uninsured risks, such as from the 2015 drought, can trap families in a cycle of poverty. Mitigating fragility helps build household resilience to avoid families falling back into poverty. The social protection delivery system also needs to identify poor and vulnerable households. Since the pandemic, Botswana has embarked on reforms to strengthen the administration of social assistance, including developing a unified social registry. Still, reforms will need to be broadened and deepened. A fully operational social registry would significantly improve the targeting of human capital investments. In addition, the government also took important steps towards developing and piloting a proxy means test for enhanced determination of eligibility for poverty-focused programs. The government should continue its plan to roll out this tool to target social assistance beneficiaries – starting with the Destitute Persons Programs. Strengthening data for evidence-based policy design Improvements in statistical data collection and use, data infrastructure, and monitoring and evaluation systems are important for transparency, accountability, and robust evidence-based policy design. Tracking progress on poverty reduction requires frequently collecting reliable quality data on human capital, livelihoods, and welfare. The last available income and expenditure survey was collected in 2015/16. Fortunately, a new survey is expected to go into the field this year (2024/25), but a commitment to reduce these data gaps is needed. More broadly, investing in data is required to make informed public policy decisions and to track progress (or lack thereof), such as for malnutrition trends. The development and implementation of Botswana’s National Development Plans require timely, reliable, high-quality data across sectors, the modernization of the National Statistical System, the integration of data systems, and strengthened monitoring and evaluation systems. BOTSWANA POVERTY ASSESSMENT 1 CHAPTER 1 BOTSWANA’S MACROECONOMIC CONTEXT AND RECENT LABOR MARKET OUTCOMES 2 BOTSWANA POVERTY ASSESSMENT 1.1  MACROECONOMIC CONTEXT AND GROWTH SLOWDOWN B otswana’s history of strong growth and low corruption has helped catapult it from among the world’s poorest countries to a stable upper-middle income country. Botswana was one of the world’s fastest-growing economies from independence (1966) to the late 1990s, with average annual GDP and GDP per capita growth above 10 and 7 percent, respectively. It is a large, sparsely populated, land- locked country in Southern Africa with significant mineral (diamond) wealth and a relatively small population of around 2.3 million people (2022).30 The fast economic growth was due to the discovery of one of the world’s largest diamond deposits, turning Botswana into one of the largest diamond exporters. The country also maintained stability thanks to sustainable macroeconomic and fiscal policies and repeatedly ranked among the top African performers across many governance indicators. Strong economic growth and significant fiscal revenues are reliant on diamonds. Agriculture remains a low-productivity sector largely dominated by subsistence farming. Some manufacturing and construction activities are growing, but they primarily serve the domestic market and account for around 15 percent of GDP. Services are growing the most, but public administration has the largest share. Overall, the mining sector remains the most important for growth, with diamonds representing around 90 percent of total goods exports31 and accounting for over a third of overall revenues. However, the extractive industry employs few people; thus, its growth does not directly reach many households. The significant revenue inflows led the authorities to establish the Pula Fund in 1994 to preserve part of the income for future generations. The government has invested diamond revenues to improve infrastructure, health, and education. Since independence, significant expansions have occurred in the road network, in access to electricity, water, and sanitation, and in primary school enrollment, with substantial improvements in life expectancy, mortality rates, and nutrition.32 Economic growth has slowed since 2010 and become more volatile. During 2002-2009, the period of poverty reduction covered by the 2015 Poverty Assessment, GDP and GDP per capita grew at an annual average rate of 2.7 and 0.7 percent, respectively. This represented strong GDP and GDP per capita growth of 5 and 3 percent, respectively, before the 2009 Global Financial Crisis. Although not as high as in earlier decades, this strong FIGURE 1.1 FIGURE 1.1  Economic growth has slowed and become more volatile 30 20 10 0 -10 -6.7 -10.4 -15.9 -20 1971 1974 1977 1980 1983 1986 1989 1992 1995 1998 2001 2004 2007 2010 2013 2016 2019 2022 GDP growth (annual %) GDP per capita growth (annual %) 2015 PA New PA Source: World Development Indicators (database), databank.worldbank.org/source/world-development-indicators, version 7/25/2023 30  The preliminary results of the 2022 Population and Housing Census shows an estimated population of 2,346,179. 31  Followed by nature-based luxury tourism, which attracted more than 330,000 international visitors in 2019 (21.3 percent of the total). 32  For example, the road network grew from under 10 kilometers at independence to 32,564 kilometers in 2021; coverage rates in water and electricity reached 92 percent and 72 percent in 2020, respectively, from about 68 percent and under 10 percent in the 1980s; and universal enrollment in primary education was achieved around 1997. BOTSWANA POVERTY ASSESSMENT 3 FIGURE 1.2  Wholesale and retail trade has seen strong growth but remains far behind mining and FIGURE 1.2 public administration value added Value Added by Type of Economic Activity 2006-2022 (constant 2016 prices, P million) 60,000 Agr Min Industry Services 50,000 40,000 30,000 20,000 10,000 0 Agriculture Mining Manufacturing Water/Electricity Construction Wholesale/Retail Diamond Traders Transport/Storage Accom/Food Info/Comm Tech Finance/Insurance Real Estate Professional Admin & Support Public Admin Education Health & Social Other Services Source: World Bank calculations using Statistics Botswana (2023b). growth still raised overall incomes and delivered good economic and welfare outcomes. Since then, the country has faced recessions in 2012, 201533, and 2020 (COVID-19 shock), highlighting that the reliance on minerals and the public sector makes the economy vulnerable to external shocks. GDP and GDP per capita between 2009-15 grew at an annual average of 2.1 and 0.1 percent, respectively, and between 2015-21 at a still-low annual average of 2.4 and 0.4 percent, respectively (Figure 1.1). GDP growth rebounded strongly after each recession in the last 15 years, yet overall average growth has remained relatively low and more volatile. The 2015 recession was caused by both external and domestic factors and highlighted the weakness of the diamond sector-led development model. A decline in the global demand for diamonds began in late 2014, linked to the slowdown in China, that, like the 2008/09 global financial crisis, led to a decline in the real prices of rough diamond exports and lower production. Mining as a share of value added declined sharply from 39 to 26 percent during 2009, and despite a rapid GDP recovery in 2010, the share has remained lower than pre-2009. The 2015 crisis led to a further decline in the share of mining GDP. Other mineral exports also collapsed owing to reduced international prices and the closure of the largest copper-nickel mine in 2016. Although wholesale/ retail, construction, and manufacturing were all negatively impacted in 2015, manufacturing (which has the fifth highest value added) has not fully recovered to 2014 levels (Figure 1.2). The challenges of 2015 were exacerbated by a major regional drought plus electricity and water shortage crises that limited private sector growth. The electricity crisis stemmed from problems commissioning a major power plant (Morupule B) plus service provision inefficiencies, resulting in Botswana having to import 39 percent of its electricity needs. Furthermore, the termination of a long-term power purchase agreement with a major supplier forced the country to import electricity at higher rates and without a secure supply. This contributed to shortfalls, significant fiscal transfers to the energy company, and escalating tariffs. The water supply was affected by a regional drought and aggravated by significant evaporation levels. Additionally, Botswana’s scattered population and the spatial disconnect between water sources and population centers contributed to increased water distribution costs. 33  The last available income and expenditure household survey for this report’s poverty analysis was collected in 2015/16. 4 BOTSWANA POVERTY ASSESSMENT The COVID-19 pandemic and global lockdowns resulted in the 2020 recession, but Botswana’s pace of growth had already slowed due to increased competition from synthetic diamonds, higher local production costs, and low job creation. Weakening global demand for diamonds and another drought in 2019 had already led to a growth slowdown from 4.2 percent in 2018 to 3.0 percent in 2019, even before the pandemic hit. The global restrictions on economic activity from the 2020 COVID-19 pandemic further weakened the external demand for diamonds, while travel restrictions affected the tourism sector. These impacts, plus the domestic lockdown, led to a strong economic contraction of -8.7 percent in 2020 (-10.4 percent for GDP per capita). The shocks that hit the economy between 2015 and 2020 highlighted the weak economic diversification. The service industry grew as mining declined during the 2015 and 2020 shocks, mainly due to the proliferation of small, low-productivity firms primarily serving the small domestic market. Consequently, the shift towards services has not led to significant economic growth or employment opportunities. Instead, most emerging economic activity has been in non-tradable goods and services, with limited growth in tradable industries beyond mining. Sectors such as wholesale and retail trade, hospitality, and food service have been more dynamic, but part of their growth is closely tied to downstream diamond industries34 as well as nature-based tourism. While there has been some diversification, there has been little improvement in productivity and a persistent long-standing role of the public sector35 (Figure 1.2). Despite the country’s efforts to invest in infrastructure, health, and education, private sector participation in nonmineral exports and transformative sectors has been limited, restricting economic growth, diversification, and job creation. The non-mining private sector has yet to attain the level of expertise, training, and efficiency that characterizes the diamond industry. Even though Botswana has adopted multiple policy instruments to encourage private sector investment (including prudent macroeconomic management and efforts to create a well-functioning market environment), the large participation of the government in the economy has stifled private sector investment, leading to minimal economic contributions from non-mining sectors and entrenched inefficiencies in the economy. In addition, efforts to lure foreign direct investment are hindered by high production costs, a scarcity of skilled labor, contradictory policies (for example, between openness versus protectionism), and the country’s landlocked position. The government’s aim of high-income-country status by 2036 requires a significant and sustained boost in economic growth, but Botswana still performs better than most of its structural peers. Botswana is considered an upper middle-income country based on its Gross National Income per capita of $6,940.36 Assuming the threshold for high-income country status remains the same, Botswana’s GNI per capita would have to grow 4.7 percent per year over the next 14 years for the country to reach high-income status by 2036 (Figure 1.3).37 In comparison, before the COVID-19 pandemic, Botswana’s GNI per capita grew at 3.1 percent on average between 2011 and 2019, or at 2.3 percent for the entire 2011-21 decade. GDP growth in recent years has also been slow for many countries considered Botswana’s structural peers (Figure 1.4; Appendix 1). Botswana’s strong growth throughout the 1990s and early 2000s resulted in one of the highest incomes among its structural peers. Over the past decade, Botswana has surpassed the GDP per capita levels of South Africa, a regional peer and Botswana’s main economic partner. However, its GDP per capita of $7,000 is below the average for upper middle-income countries ($10,055 constant US$ 2015) in 2021. 34  The relocation of the Diamond Trading Company International from the United Kingdom to Botswana in 2012 had a considerable impact on the wholesale sub-sector. 35  The public sector includes central and local government, parastatals, and the Ipelegeng workfare program, therefore it is larger than just the size of the “public administration” economic activity. 36  For fiscal year 2023, upper middle-income economies are defined as those with a GNI per capita between $4,256 and $13,205, calcu- lated using the World Bank Atlas method; high-income economies are those with a GNI per capita of $13,205 or more. 37  The GNI would have to grow at a yearly 6.1 percent over 14 years for Botswana to reach HIC status by 2036. The new 2022 Population Census shows the population growing at an annualized 1.4 percent (rather than 1.9 percent in the previous census). BOTSWANA POVERTY ASSESSMENT 5 FIGURE 1.3  Reaching high-income status FIGURE 1.4  GDP per capita is higher than requires a significant, sustained boost in among structural peers but below the UMIC FIGURE 1.3 economic growth FIGURE 1.4 average GNI per capita, Atlas method (current US$) GDP per capita (constant 2015 US$) 14,000 14,000 High income threshold 12,000 US$ 13,205 12,000 GNI per capita would have to increase by 10,000 at least 4.7% per year in real terms for 10,000 Botswana to reach HIC status by 2036. 8,000 8,000 6,000 6,000 4,000 Upper middle- 4,000 income threshold, US$ 4,256 2,000 2,000 0 0 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 2020 2022 2024 2026 2028 2030 2032 2034 2036 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 2020 2022 Congo, Rep. Gabon Georgia Botswana Scenario (5.0% p.a.) Namibia Tunisia South Africa Scenario (2010-2019 average) Lebanon Mongolia Botswana Source: World Bank calculations using World Development Source: World Development Indicators, version 7/25/2023. Indicators. Botswana faces a fast-changing local and global context characterized by growing challenges and risks, but also fresh opportunities. The country faces increasing international uncertainty, continued volatile commodity prices, high inflation and interest rates, revamping global value chains, geopolitical transformation, and protracted conflicts. South Africa has faced slow growth for the last 15 years, negatively impacting Botswana’s economy, while droughts and floods continue to be a challenge. However, the Updated Botswana Systematic Country Diagnostic (World Bank 2023) highlights how several global trends, while adding to the challenges, offer novel opportunities. For example, the SCD Update highlights that Botswana could build on its potential comparative advantages in the areas of renewable energy, exports of battery minerals, livestock value chains, and ecotourism. 1.2  RECENT LABOR MARKET TRENDS Since 2009, job creation has not matched the growing labor force. Between 2003-2009, a period of strong poverty reduction, employment growth outpaced labor force growth by 1.2 percentage points (Table 1.1). Even though both employment and labor force growth slowed between 2009 and 2016, job creation was not at par with the growing labor force (1.8 percent vs. 2.0 percent, respectively). Recent labor market data for 2016-21 shows that employment continues to grow at a slower rate than the labor force (3.3 versus 4.5 percent). Figure 1.5 shows that the labor force participation rate (for the population 15 years or more) has been generally increasing since 2009, whereas the employment rate passed 50 percent in 2016 but has declined a percentage point to 49.2 percent in 2021 and 2022. The country’s current growth model does not support intensive, high-productivity job creation. Botswana ranks in the bottom 30 percent of countries worldwide in terms of employment ratios (at 131 out of 187 countries).38 Labor demand bottlenecks limit job creation, and labor supply constraints result in skills mismatches. A context of low competition, high input costs, an uncertain regulatory environment, and skills mismatches poses structural constraints to private sector job creation. The economy heavily depends on 38  International Labour Organization. “ILO Modelled Estimates and Projections database (ILOEST)” ILOSTAT. Accessed October 10, 2023. https://ilostat.ilo.org/data/. Employment to population ratio, 15+, total (%). 6 BOTSWANA POVERTY ASSESSMENT TABLE 1.1  Labor force growth is outpacing employment growth Annualized growth in employment and the labor force (%) Labor Force Employed 2002-2009 2.7 3.9 2009-2016 2.0 1.8 2016-2021 4.5 3.3 Source: World Bank calculations using 2002/03 BHIES, 2009/10 BCWIS, 2015/16 BMTHS, and 2021-Q4 QMTS (18th ICLS standard). diamonds and the public sector, with much of the private sector playing a peripheral role. Job creation up to 2016 was driven by public sector expansion. Although private sector job creation has grown stronger in more recent years, these tend to be jobs in low-productivity, non-tradable services. The public sector remains the country’s largest employer, with 27 percent of total employment using international estimates (compared to a global average of 17 percent) and 47 percent of formal employment (compared to 38 percent globally).39 Botswana continues to experience high structural unemployment, with increasing levels in recent years. National unemployment is higher than in all other upper-middle-income countries, except for South Africa; it increased from 17.6 percent in 2015 to 22.7 percent by 2022 year-end (using comparable unemployment measures, see Box 1.1 and Appendix 7). Even though unemployment was traditionally a rural phenomenon, significant rural-urban migration has shifted part of the problem into urban areas. Unemployment is highest in urban villages, among women (25 percent versus 20.5 percent for males in 2022), and among the 15-24 age group (43.6 percent versus 20.9 and 10.9 percent for the 25-54 and 55-64 age groups, respectively). Between 2009 and 2016, urban villages saw an increase in employment and a drop in unemployment. On the other hand, households in rural villages experienced declining labor force participation and employment rates and increasing unemployment in 2016 amid a drought and electricity crisis. The most recent six-year period shows a decline in the employment rate for both urban and rural villages and a surge in unemployment, reaching 26.3 and 21.8 percent, respectively, in 2022 (Figure 1.6). Cities and towns also experienced an increase in the unemployment rate, from 13.3 in 2016 to 15.8 in 2022, as more people entered the labor force while employment rates remained flat. After controlling for differences in education, age, and gender, the probability of being employed is higher in cities and towns relative to urban and rural villages. In 2022, workers in urban villages were 25 percent less likely to be employed than workers in cities and towns, higher than the 23 percent estimated in 2016 (Appendix 2). Workers in rural villages were 10 percent less likely to be employed than workers in cities and towns, holding all else constant. FIGURE 1.5 FIGURE 1.5  Labor Market Trends, 2002-2021Q4 (%) 80 63.0 64.4 63.5 60.0 61.3 58.6 63.7 Labor Force 60 Participation Rate 50.5 50.3 50.7 49.2 (%, 15 yrs +) 48.6 45.7 49.2 Employment Rate 40 (%, 15 yrs +) Unemployment 20 22.7 Rate (%) 23.9 22.5 20.1 21.4 17.6 17.1 0 2003 2009 2016 2019- 2020- 2021- 2022- Q4 Q4 Q4 Q4 Source: World Bank calculations using the 2002/03 BHIES, 2009/10 BCWIS, 2015/16 BMTHS, and QMTS (various years). 39  Worldwide Bureaucracy Indicators dataset (version 3.0). August 2022. https://www.worldbank.org/en/data/interactive/2019/05/21/worldwide-bureaucracy-indicators-dashboard#2 BOTSWANA POVERTY ASSESSMENT 7 The labor market continues to reflect substantial barriers to the inclusion of women in Botswana. As of 2022, labor force participation rates were higher for men (69 percent) compared to women (59 percent), while unemployment was 4.5 percentage points higher among women (as mentioned earlier). After controlling for differences in education, age, and geographic location, men were 32 percent more likely to be employed than women in 2022 (Appendix 2). Moreover, women in the labor force earn less than men, account for over a third of the agricultural workforce, and own more informal agricultural businesses than men, underlining their less secure economic standing. Men are 40 percent more likely to be wage workers, while women are 58 percent more likely to be unpaid family workers. In rural areas, nearly 1 out of 3 men are wage employees relative to only 16 percent of women. While men control the high-value agriculture livestock sector, women in the sector own more chickens. Despite having more arable land, women are unable to exploit it due to limited access to the required inputs to develop the land. Even though Botswana has the second-highest rate of female entrepreneurship globally, women continue to face significant challenges in the business environment, including a lack of access to finance, assets, skills, education, training, and networks.40 Moreover, as the primary caregivers in their families, women bear the responsibility of caring for sick relatives, often at the expense of pursuing economic, educational, or training opportunities. The COVID-19 pandemic had a worse employment impact in urban villages and among women and youth. Botswana’s employment survey for workers employed before the pandemic estimates that over 67 thousand people lost jobs or businesses and 58 percent of these losses were female jobs. Moreover, job/business loss was more prevalent among the 25-29 age group and in urban villages (53 percent relative to 30 percent in rural villages). By the end of 2020, only 4 percent of workers who lost a job were able to find work; of these, 70 percent were male. Approximately 19 thousand people found jobs during the COVID-19 outbreak, of which nearly 77 percent were women and 72 percent were in education or public administration. Nonetheless, more than a third of employees were hired in the temporary workfare Ipelegeng program that offers low-paying jobs. The program is run by local governments and is meant as an intervention to temporarily combat poverty and unemployment. 1.6 FIGURE1.6 FIGURE   Labor Market Trends, 2010-2022Q4 (%) a. Cities and Towns b. Urban villages c. Rural villages 80 70.2 71.8 70.0 80 80 64.0 59.8 61.6 62.0 62.2 59.2 59.7 54.4 56.9 60 60 60 60.8 47.1 48.0 45.7 57.6 59.0 56.1 42.2 49.9 47.8 49.1 48.7 40 40 40 22.1 26.3 19.7 22.4 21.2 21.8 13.3 15.8 16.0 17.7 20 12.3 20 20 15.7 0 0 0 2010 2016 2019-Q4 2020-Q4 2021-Q4 2022-Q4 2009 2016 2019-Q4 2020-Q4 2021-Q4 2022-Q4 2010 2016 2019-Q4 2020-Q4 2021-Q4 2022-Q4 Labor Force Participation Employment Rate Unemployment Source: World Bank calculations, using 2009/10 BCWIS and 2015/16 BMTHS (Statistics Botswana 2013, 2018), and QMTS (Statistics Botswana 2019a, 2019b; 2020a, 2020b; 2021; 2022). Note: QMTS employment figures are adjusted to include subsistence farmers (18th ICLS standard). 40  Rudhumbu, du Plessis, and Maphosa (2020); Mastercard (2022); and forthcoming World Bank Gender Assessment (2024). 8 BOTSWANA POVERTY ASSESSMENT BOX 1.1  Comparable labor market statistics Botswana’s employment and unemployment statistics and their trend over time vary depending on which international standard is applied. Since the launch of the first Quarterly Multi-Topic Survey (QMTS) in quarter 3 of 2019, Statistics Botswana has been applying the new international definition of employment in its official labor market statistics, causing a break with previous official statistics. The comparison of employment and unemployment statistics over time requires the application of the same definition across the entire time period. The labor market statistics published by Statistics Botswana in 2002/03, 2009/10, and 2015/16, applied the standard of the Eighteenth International Conference of Labour Statisticians (ICLS), while the statistics published in the QMTS surveys in 2019, 2020, 2021, and 2022 all applied the Nineteenth ICLS standard. The new standard introduced a new classification for “forms of work”: own-use production work (e.g., subsistence farming) is now a separate indicator and no longer counts towards employment or labor force participation. In other words, subsistence farmers that do not sell the majority of their production are no longer considered employed. The current statistics are not inaccurate per se, they are in line with new international definitions from ILO, but they are not comparable to the earlier data. To maintain comparability with official unemployment rates from 2002 to 2016, the previous 18th ICLS standards were applied throughout the poverty assessment. A concrete result of this is that the current official unemployment rate for Botswana in 2022-Q4 (considering the labor force as 15 years or above) is 25.4 percent, based on implementing the 19th ICLS standards, while it is 22.7 percent using the 18th ICLS standards (see Figure B1.1). The comparable data shows that the labor force participation rate has been increasing since 2009, while the employment rate is relatively stagnant. Moreover, unemployment has been continuously increasing since 2010, including between 2021 and 2022. Box B1.1 FIGURE B1.1  Comparable labor force statistics show increasing labor force participation and unemployment rates, while employment rates stagnate 70 64.4 63.0 63.5 60.0 61.3 58.6 63.7 60 59.7 59.0 50.5 50.3 50.7 48.6 49.2 50 45.7 49.2 44.5 45.9 40 30 22.2 25.4 22.7 20 23.9 22.5 20.1 21.4 17.1 17.6 10 2003 2009 2016 2019- 2020- 2021- 2022- Q4 Q4 Q4 Q4 Labor Force Participation Rate (%, 15 yrs +) Employment Rate (%, 15 yrs +) Unemployment Rate (%) Official LFP (19th ICLS) Official Employment (19th ICLS) Official Unemployment (19th ICLS) See Appendix 7 for more information. BOTSWANA POVERTY ASSESSMENT 9 Employment in 2022 remains concentrated in services, primarily public administration and wholesale and retail trade, while agricultural employment has increased since 2016. Employment growth between 2009 and 2016 was almost entirely due to service sector growth, primarily public sector expansion. More recently, between 2016 and 2022, 56 percent of employment growth came from the agricultural sector, 25 percent from industry, and 20 percent from the services sector, while employment declines in mining contributed a negative 1 percent. As of 2022, services accounted for two-thirds of jobs (526,000 out of 788,000 total jobs, where total jobs include subsistence farmers). The 2015 drought may have forced nearly 80 thousand workers out of agriculture and into services between 2009 and 2016 (Figure 1.7a). Some sought income via the workfare Ipelegeng program in rural areas while others sought income as domestic personnel in private households, both low productivity employment in the service sector. Since then, the share of workers employed in agriculture increased from 11.7 percent in 2016 to 17.6 percent in 2022 (Figure 1.7b). This was driven primarily by increases in non-subsistence farming, possibly linked to policies restricting food imports, and resulted in the share of subsistence farmers in agricultural employment declining from 85 percent to 51 percent between 2016 and 2022 (Figure 1.8). Nonetheless, agriculture’s labor productivity (proxied by value added per worker) remains low. As of 2022, not including mining, agricultural labor productivity was 12 times lower than in industry and nine times lower than in the service sector. The mining sector has the highest levels of labor productivity because it generates substantial value but employs a relatively small workforce (Figure 1.9). FIGURE 1.7 FIGURE 1.7   Lack of transition out of agriculture while employment in services has grown a. Employment by sector over time (‘000) b. Share of employment by sector (%) 900 100% 800 20.0 24.3 25.2 24.3 26.2 26.9 Number employed ('000) 32.4 700 206 201 211 80% 212 8.4 5.2 8.0 8.9 9.5 7.3 600 66 74 14.6 7.7 221 76 16.4 146 58 60% 18.0 16.1 14.1 16.6 500 147 133 31 53 114 131 13.4 10.2 19.5 400 92 99 40% 9.1 7.3 17.0 17.9 16.7 15.8 39 133 139 148 134 125 300 67 61 7.8 14.3 6.1 6.3 44 53 49 10.6 2.6 7.3 6.4 62 98 59 50 5.4 8.3 7.5 200 42 47 60 67 59 20% 3.0 7.9 7.2 37 65 6.9 1.4 1.5 1.8 49 47 26.3 1.1 100 20.8 2.2 158 127 147 142 139 15.5 17.8 17.6 17.6 96 80 11.7 0 0% 2003 2009 2016 2019q4 2020q4 2021q4 2022q4 2003 2009 2016 2019q4 2020q4 2021q4 2022q4 Agriculture. Fishing Mining Manufacturing & Utilities Construction Wholesale & Retail Public Administration Education Rest of Services Source: World Bank calculations, using 2009 BCWIS and 2016 BMTHS (Statistics Botswana 2013, 2018), and QMTS (Statistics Botswana 2019a, 2019b; 2020a, 2020b; 2021; 2022). Note: QMTS employment figures are adjusted to include subsistence farmers (Eighteenth ICLS standard). Evidence also suggests a recent movement from farm self-employment and public wage work into private wage work. Among employed people in 2022, 16.4 percent were classified as nonfarm self-employed, 9.8 percent as farm self-employed, 46.8 percent as private wage workers (including domestic personnel), 18.2 percent as wage workers in central or local government, 2.4 percent as parastatal workers, and 6.6 percent as Ipelegeng public workers (Figure 1.10).41 Public wage workers include those who work in the public administration and defense sector, as well as government workers in other sectors, such as education, health, etc., while also including parastatal and Ipelegeng workers. Between 2009 and 2016, most of the increase in employment came 41  Private wage work includes nongovernmental organizations (0.6 percent). Ipelegeng is a workfare program for adults 18 years or older and is limited to only one-month employment. The public sector includes central and local government, parastatals, and Ipelegeng workers. 10 BOTSWANA POVERTY ASSESSMENT FIGURE1.8 FIGURE 1,8 Employment FIGURE 1,9 FIGURE 1.9  Employment and Value Added (2022) in Agriculture, Forestry and Fishing (2016-2022) 160,000 160,000 45,000 Value Added at constant 2016 prices (Pula million) 2022 Value Added Employment 140,000 140,000 40,000 120,000 35,000 120,000 Employment 68,401 30,000 100,000 100,000 25,000 80,000 11,851 80,000 20,000 60,000 60,000 15,000 40,000 40,000 10,000 70,273 69,406 20,000 20,000 5,000 0 0 0 Agriculture, Forestry & Fishing Public Administration & Defence Wholesale & Retail Education Manufacturing Construction Other Services Administrative & Support Activities Accomodation & Food Services Human Health & Social Work Transport & Storage Professional, Scientific & Technical Activities Mining & Quarrying Finance, Insurance & Pension Funding Water & Electricity Information & Communication Technology Real Estate Activities 2016 2022 Subsistence Not subsistence Source: World Bank calculation Source: World Bank calculations using 2022 QMTS and national accounts (Statistics using 2016 BMTHS and 2022 QMTS. Botswana, 2023b). from growth in the public sector (Figure 1.11). This was due to employment growth from local government, parastatals, and Ipelegeng. On the other hand, between 2016 and 2022, most of the employment growth was private sector growth. The public sector declined marginally in this later period, among central and local government and parastatal employment, while Ipelegeng increased. Farm self-employment declined while nonfarm self-employment grew. Although public sector employment decreased between 2016 and 2022, the number of people in the Ipelegeng workfare program increased, and the wage gap between the public and private sectors has grown. As part of public sector reforms, in 2021, the authorities announced they would eliminate half of the public sector open positions and review the size of the civil service. More than 18 thousand workers left central and local government or parastatals between 2016 and 2022. However, the Ipelegeng workfare program increased by more than 7000 people. As a result, employment in Ipelegeng has become a larger share of overall public sector employment, rising 4.5 percentage points from 19.7 in 2016 to 24.2 in 2022. As of 2022, public sector wages were 1.7 times higher than in the private sector, although declining from a 2.1 rate in 2021 (Figure 1.12 and Figure 1.13). At 2.6, the gap between public and private wages is particularly high in rural areas and growing over time, despite low-paid Ipelegeng workers constituting half of rural public jobs. Disparities in earnings arise from differences in demographics, educational attainment, and employment industry. Higher levels of schooling and living in urban areas (urban villages and cities and towns) are associated with increases in earnings (monthly real wages; Appendix 3). Relative to primary education or less, in 2022, secondary and tertiary schooling show 56 and 181 percent higher earnings, respectively. After controlling for differences in education, age, and employment sector, rural earnings were 40 percent lower than in cities and towns. Botswana’s gender pay gap remains statistically significant after controlling for differences in human capital accumulation, age, geographic location, and sector of work. Over the last decade, on average, male earnings have been 39 percent higher than female earnings (although the earnings gap drops to 24 percent when occupation is also considered).42 Sector of employment is another strong predictor of earnings. Except for 42  When occupation is also considered, this earnings gap is estimated to be around 24 percent in 2016. See Inequality in Southern Africa, World Bank 2022. BOTSWANA POVERTY ASSESSMENT 11 FIGURE 1.10 FIGURE 1.11 FIGURE 1.10  Share of workers by type of FIGURE 1.11  Employment by type of worker, employment, 2009-2022 (%) 2009-2022 100% 400,000 15.8 16.6 15.6 15.6 15.4 16.4 350,000 80% 9.6 9.9 12.8 12.2 9.8 16.0 300,000 60% 250,000 41.4 44.4 200,000 44.1 44.5 46.8 43.4 40% 150,000 100,000 20% 32.3 30.1 24.8 27.5 27.9 27.1 50,000 0% 0 2009 2016 2019Q4 2020Q4 2021Q4 2022Q4 2009 2016 2019Q4 2020Q4 2021Q4 2022Q4 Public wage worker Private wage worker Public wage worker Private wage worker Farm self-employed Non-Farm self-employed Farm self-employed Non-Farm self-employed Source: World Bank calculations using BCWIS, BMTHS, and Source: World Bank calculations using BCWIS, BMTHS, and 2019-2022 QMTS. 2019-2022 QMTS. FIGURE 1.13 FIGURE 1.14 FIGURE 1.12  Public-private wage gap, 2016- FIGURE 1.13  Average monthly real wages by 2022 (ratio) sector and area, 2016-2022 (constant 2010 Pula) 3.0 7,000 Ratio public to private real wages 2.6 6,000 2.3 2.3 (constant 2010 Pula) Monthly real wages 2.1 2.1 2.0 5,000 2.0 1.7 1.7 1.7 1.5 1.6 4,000 1.3 3,000 1.0 2,000 1,000 0.0 0 2016 2019 2021 2022 2016 2019 2020 2021 2022 National Urban Rural Urban Public Wage Rural Public Wage Urban Private Wage Rural Private Wage Source: World Bank calculations using BMTHS and 2019-2022 Source: World Bank calculations using BMTHS and 2019-2022 QMTS. QMTS. public administration, in 2016, earnings in all sectors were statistically significantly higher than in agriculture, with mining and education showing the highest returns, at 95 percent and 57 percent, respectively. While mining and education continue to show statistically significant returns, these dropped nearly 17 percentage points for both sectors between 2016 and 2021. Higher earnings from public administration work are not statistically significant, suggesting that other factors, like education, may contribute more to the income disparities between public administration and agriculture. Additionally, workers in the Ipelegeng program are classified as public administration workers, resulting in a mix of workers and wage rates in the sector. 12 BOTSWANA POVERTY ASSESSMENT CHAPTER 2 PROGRESS IN POVERTY REDUCTION HAS SLOWED, AND INEQUALITY REMAINS HIGH BOTSWANA POVERTY ASSESSMENT 13 2.1  TRENDS IN POVERTY, INEQUALITY, AND SHARED PROSPERITY P overty reduction slowed in 2016 and likely continued to slow in recent years. The share of the population living below the official poverty line declined strongly from 30.6 to 19.3 percent between 2002/03 and 2009/10 while declining more slowly to 16.1 percent in 2015/16 (hereafter referred to as 2003, 2009, and 2016; Figure 2.1).43, 44 This deceleration since the last Poverty Assessment aligns with the economic slowdown since 2010 and the more volatile economic conditions. In 2016, after the recession of 2015 and during a drought, over 330,000 Batswana were living in poverty. Despite better GDP per capita growth between 2009-16 than 2003-09, the ability of economic growth to lead to poverty reduction declined by 84 percent in the last available survey (as the elasticity of GDP per capita to poverty fell from -3.8 in 2003-09 to -0.6 in 2009-16), requiring further investigation (Figure 2.2).45 Unlike in the past, GDP growth (based on diamond wealth and a large public sector) did not translate to higher average household per capita consumption between 2009-16. Instead, the drought affected the 2016 harvest, while the electricity and water shortages further limited private sector growth, resulting in lower average household consumption than in 2009. Moreover, the country has since faced floods in 2017, another drought in 2019, the 2020 COVID-19 shock, and a surge in prices in 2022, in addition to sluggish productivity growth and an economic decline in South Africa, Botswana’s main trade partner. Poverty projections up to 2022 using different methodologies suggest poverty reduction has slowed even further as the labor market has weakened (see Box 2.1; Appendix 9). FIGURE 2.1 FIGURE 2.2 FIGURE 2.1  Poverty reduction has slowed FIGURE 2.2  The link between poverty and GDP growth became weaker 35 0% 0.0 31 Elasticity of Poverty to GDP Per 30 -10% -1.0 Change in Poverty 25 Capita Growth (annualized) 19 18.0 20 16.1 15.6 -20% -2.0 15 14.0 10 -30% -3.0 5 0 -40% -4.0 2003-2009 2009-2016 2003 2009 2016 2017 2018 2019Q3 2019Q4 2020Q1 2020Q4 2021Q4 2022Q4 Official Full Model (Quarterly) Poverty Change Elasticity Limited Model (Quarterly) Elasticity (Annual) Source: 2002/03 BHIES, 2009/10 BCWIS, 2015/16 BMTHS Source: World Bank calculations using 2009/10 BCWIS, 2015/16 (Stats Botswana 2013, 2018; also see Appendix 8); 2017-2022 BMTHS, and World Development Indicators for GDP data. projections based on Quarterly Multi-Topic Household Surveys and national accounts GDP data. Note: Projection methodologies include “limited” and “full” quarterly models using survey-to- survey imputations (SWIFT models) and annualized elasticity projections. See Appendix 9 for details. 43  Statistics Botswana and the World Bank have updated Botswana’s official poverty measurement methodology using international best practice. The previous methodology would suggest a poverty rate of 16.3 percent instead of 16.1 percent for 2015/16. See Annex 8 for a technical overview of the updated measures. 44  The household surveys correspond to the 2002/03 Household Income and Expenditure Survey (HIES) conducted from June 2002 to August 2003, the 2009/10 Botswana Core Welfare Indicators Survey (BCWIS) conducted from April 2009 to March 2010, and the 2015/16 Botswana Multi-Topic Household Survey (BMTHS) conducted from November 2015 to December 2016. This document will refer to these surveys as 2003, 2009, and 2016, corresponding to the year with the most survey month coverage. 45  Alternatively, using GDP data post the 2009 recession, the elasticity fell 60 percent from -2.1 in 2003-10 to -0.86 in 2010-16. 14 BOTSWANA POVERTY ASSESSMENT BOX 2.1  Data challenges, comparable poverty estimates, and poverty projections Data challenges Botswana faces multiple data challenges that limit its ability to inform and implement effective policymaking. The challenges include availability, quality, and use of data, and a weak statistical infrastructure. The country does not conduct frequent income and expenditure surveys to track poverty. Ministries and line departments are also not capturing information regularly to inform progress and better outcomes and lack integrated data systems. Evidence-based policymaking requires Botswana to act on its commitment to invest in frequent, timely, and relevant data across its statistical system and strengthen its monitoring and evaluation system. The poverty statistics and projections presented in this Poverty Assessment result from several years of collaboration between Statistics Botswana (StatsBots) and the World Bank Group (WBG). The country has undertaken five household income and expenditure surveys (HIES). These include the 1985/86 BHIES, the 1993/94 BHIES, the 2002/03 BHIES, the 2009/10 Botswana Core Welfare Indicators Survey (BCWIS), and the 2015/16 Botswana Multi-Topic Household Survey (BMTHS). The next HIES for poverty measurement is planned for 2024/25. Comparable poverty estimates between 2009/10 and 2015/16 In 2019, the WBG and StatsBots began a collaboration to update and strengthen the country’s official poverty measurement methodology and poverty rate for 2015/16. Appendix 8 provides a technical report that covers the comparability of the 2009/10 BCWIS and 2015/16 BMTHS; the construction of a comparable nominal consumption aggregate between those two surveys; the construction of a spatially deflated (“real”) consumption aggregate; the construction of a comparable harmonized poverty line for poverty trend analysis; and the resulting updated poverty and inequality measures. The latter poverty and inequality measures are used throughout the Poverty Assessment. SWIFT Poverty Projections for 2019, 2020, 2021, and 2022. More recently, to provide timely statistics for planning and decision-making, Statistics Botswana began to field a lighter, quarterly, multi-topic household survey whose primary module would be on labor force statistics while other modules would rotate. This Quarterly Multi-Topic Survey (QMTS) would not collect income and expenditure data for poverty measurement. Nevertheless, it presents essential information on employment, wages, and other labor force statistics, as well as including (in different rounds) data on household demographics, dwelling characteristics, education variables, and select ICT data, among others. The QMTS has been undertaken in quarters 3 and 4 of 2019, quarters 1 and 4 of 2020, quarter 4 of 2021, and quarter 4 of 2022. In 2021, the WBG and StatsBots began a collaboration to fill the poverty data gaps using a methodology called Survey of Well-being via Instant and Frequent Tracking (SWIFT). The SWIFT program was created in 2014 to produce cost-effective and timely poverty statistics. It uses multiple imputation, survey-to- survey, and machine learning techniques to train poverty projection models and produce poverty rate estimates. The WBG and StatsBots began to use SWIFT to impute poverty into the QMTS surveys for 2019 through 2022, using the last poverty survey (2015/16 BMTHS) as the training data to create the poverty models. Appendix 9 provides a brief technical overview of the poverty projections. Botswana’s poverty rate remains high relative to structural peers and upper-middle-income countries, placing it within the world’s top 25 percent poorest countries. Poverty declined under the international BOTSWANA POVERTY ASSESSMENT 15 FIGURE 2.3 FIGURE 2.3  Botswana is poorer than most of its structural peers a. Most recent $2.15 poverty estimate by country (%) 90 80 70 Poverty rate (%) 60 Congo, Rep. 50 South Africa 40 Botswana Namibia 30 20 Georgia Mongolia Lebanon Gabon Tunisia 10 0 Madaga scar Moza mbique South Sudan Zambia Ha iti Nig er Turkmenistan Zimbabw e Kenya Congo, Rep. Angola Cha d Ethiopia Solomon Isla nds Guinea-Bissau Ghana Timor-Leste Mali Yemen, Rep. Djibo uti Gambia, The Sao Tome and Principe Botsw ana Guinea Benin India Senega l Guatemala Lao PDR Tajikistan Geor gia St. Lucia Pakistan Nicar agua Brazil Ecuador No rth Macedo nia Gabon Trinidad and Toba go Montenegro Ro ma nia Na uru Eg ypt, Arab Rep. Paraguay Panama Mexico Suriname Sri Lanka Marshall Islands Armenia Japan Bulg aria Argentina Spain West Bank and Gaza Austria Algeria Ko so vo Latvia Grenada Estonia Tunisia Lithuania Canada Israel Portugal China Ur ugua y Mauritius Slovak Republic Bosnia a nd Herzegovina Ireland France Jorda n Belgium Finland Tonga Thailand Cyprus Belarus Azerba ijan Malaysia Lebanon Maldives Taiwan, C hina Source: Poverty and Inequality Platform, pip.worldbank.org. poverty line ($2.15 per day, 2017 PPP) from 17.7 to FIGURE 2.4  Poverty is high relative to income 15.4 percent between 2009 and 2016. However, it FIGURE 2.4 level increased by 3.1 percentage points when measured b. $2.15 poverty and GDP per capita (2017 PPP) under the upper-middle-income-country global poverty line ($6.85 per day, 2017 PPP), reaching 90 Extreme poverty rate (percentage) 63.5 percent in 2016. Under this higher global line, 80 Botswana ranked second poorest among its structural 70 peers46 and as the poorest among upper-middle- 60 income countries. Botswana’s poverty rate under the 50 $2.15 PPP line exceeds those of its structural peers 40 Georgia, Gabon, Mongolia, Tunisia, and Lebanon, 30 and approaches rates in South Africa and Namibia; 20 BWA 10 only the Republic of Congo is significantly poorer 0 (Figure 2.3). Botswana’s estimated poverty rate for 6.5 7.5 8.5 9.5 10.5 2019 in terms of the international poverty line ($2.15 LN (GDP per capita, PPP, US$)* PPP) is more than four times higher than its GDP per Source: Poverty and Inequality Platform, pip.worldbank.org; and capita would predict (Figure 2.4). World Development Indicators, version 7/25/2023. Between 2009 and 2016, rural poverty increased and urban poverty decreased, widening the urban-rural gap as rural labor outcomes weakened. Data from 2016 indicated the poverty rate in rural areas (26.8 percent) was two and a half times higher than in urban areas (10 percent) and 66 percent higher than the national average. Even as national poverty declined between 2009 and 2016, it increased by 2.5 percentage points in rural areas. Thus, the overall decline in poverty in Botswana was driven primarily by welfare improvements among urban households. Poverty decreased in cities and towns from 8 to 3.3 percent and in urban villages from 19.9 to 13.7 percent between 2009 and 2016 (Figure 2.5). Even though the depth and severity of poverty have declined since 2003, levels in rural areas remain higher, with a slower decline in recent years. The poorest of the poor remain primarily in rural areas (Figure 2.6 and Figure 2.7). 46  See Appendix 1 for the method for choosing Botswana’s structural peers. For international poverty rates, see pip.worldbank.org. 16 BOTSWANA POVERTY ASSESSMENT FIGURE 2.5 FIGURE 2.5  Strong declines in urban poverty yet rural poverty increased 50 45.2 45.2 40 30.6 30 26.8 26.8 24.4 24.9 24.4 19.3 19.9 18 20 16.1 15 13.7 10 10.7 8.0 10 3.3 0 Rural Urban Cities &Towns Urban Villages Rural Villages National Area Strata 2003 2009 2016 Source: 2002/03 Household Income and Expenditure Survey (HIES), 2009/10 Botswana Core Welfare Indicators Survey (BCWIS), and 2015/16 Botswana Multi-Topic Household Survey (BMTHS) FIGURE 2.6 FIGURE 2.7 FIGURE 2.6  The intensity of poverty improved FIGURE 2.7  The poorest of the poor were better across strata (Poverty Gap by area, %) off (Poverty Severity by area, %) 20 18.3 12 9.8 10 15 11.6 8 6.0 10 8.6 8.2 8.0 6 6.2 4.0 4.0 6.0 3.4 4.6 4 2.9 2.6 5 3.3 3.4 1.9 2.5 1.5 1.3 1.3 2 1.0 0.7 0 0 National Cities & Urban Rural National Cities & Urban Rural Towns Villages Villages Towns Villages Villages 2003 2009 2016 2003 2009 2016 Source: 2002/03 Household Income and Expenditure Survey Source: 2002/03 Household Income and Expenditure Survey (HIES), 2009/10 Botswana Core Welfare Indicators Survey (HIES), 2009/10 Botswana Core Welfare Indicators Survey (BCWIS), and 2015/16 Botswana Multi-Topic Household Survey (BCWIS), and 2015/16 Botswana Multi-Topic Household Survey (BMTHS). (BMTHS). Consumption per capita growth was also weaker in rural areas. Whereas rural areas experienced strong consumption growth across all consumption deciles between 2003 and 2009, and faster than in urban areas, the 2015 recession and drought disproportionately affected households just above the poverty line that relied on subsistence farming and small holding of livestock. In 2016, consumption growth dropped for all but the poorest 15 percent of the population in rural areas (Figure 2.8a). In contrast, urban areas saw higher consumption per capita for all but the top 20 percent of the urban population (Figure 2.8b). In 2016, Botswana was no longer among Africa’s top performers in terms of shared prosperity. Between 2009 and 2016, shared prosperity, expressed as the annualized growth rate in consumption per capita of Botswana’s poorest 40 percent (Bottom 40), reached only 1.2 percent, down from the 5 percent annualized growth estimated between 2003 and 2009. Nonetheless, consumption growth was negative for the rest of the population in that period, resulting in an overall -1.8 percent annualized consumption per capita decline BOTSWANA POVERTY ASSESSMENT 17 nationwide.47 Using international estimates (the Global Database for Shared Prosperity), Botswana’s shared prosperity indicator is no longer among the highest in Africa (Figure 2.9). The decline in consumption per capita among the middle and higher parts of the consumption distribution explains the increase in the poverty rate under the higher $6.85 per day global poverty line. Figure 2.10 shows the distribution of per capita consumption in 2009 (dotted line) and 2016 (the solid line with the colored bands for different consumption lines). While fewer people are living on less than $2.15 per day, there is a higher density of people below the $3.65 and $6.85 poverty lines. FIGURE 2.8  Rural households fared worse than urban households across the income distribution FIGURE 2.8 a. Rural growth incidence curve (%) a. Urban growth incidence curve (%) 10 10 Annualized Growth Rate (%) Annualized Growth Rate (%) 8 8 6 6 4 4 2 2 0 0 -2 -2 -4 -4 -6 -6 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 Percentile Percentile 2009-2003 2016-2009 2009-2003 2016-2009 Source: World Bank calculations using 2002/03 Household Income and Expenditure Survey, 2009/10 Botswana Core Welfare Indicators Survey, and 2015/16 Botswana Multi-Topic Household Survey. On the horizontal axis, the population is divided into ventiles. FIGURE 2.9 FIGURE 2.10 FIGURE 2.9  Botswana is no longer one of Africa’s FIGURE 2.10  Fewer people lived below the $2.15 top performers on shared prosperity indicator poverty line in 2016 but more lived below $6.85 6 0.50 4 Botswana 0.40 2 0.30 Density 0 0.20 -2 0.10 -4 -6 0.00 $2.15 $3.65 $6.85 $14 $80 BFA 2009-2014 NER 2007-2014 NAM 2009-2015 UGA 2009-2012 SWZ 2009-2016 GMB 2010-2015 RWA 2010-2013 SYC 2013-2018 MRT 2008-2014 MWI 2010-2016 MUS 2012-2017 SLE 2011-2018 TGO 2011-2015 MOZ 2008-2014 CMR 2007-2014 ETH 2010-2015 CIV 2008-2015 BWA 2009-2015 RWA 2013-2016 UGA 2016-2019 NER 2011-2014 TZA 2011-2018 GHA 2012-2016 ZMB 2010-2015 ZAF 2010-2014 UGA 2012-2016 MWI 2016-2019 MDG 2010-2012 ZWE 2011-2017 BEN 2011-2015 Per capita daily consumption (ppp2017) 2009 2015 Source: World Bank Global Database of Shared Prosperity, April Source: World Bank calculations using 2009/10 Botswana Core 2023 vintage; data spells circa 2010-2015. Welfare Indicators Survey and 2015/16 Botswana Multi-Topic Household Survey. 47  Using the official welfare aggregate. These rates are slightly different from the World Bank’s Global Database of Shared Prosperity (GDSP) that shows 0.42 and -3.3 percent estimates for the Bottom 40 and overall, respectively. 18 BOTSWANA POVERTY ASSESSMENT Despite the decline in inequality, Botswana’s society remains among the world’s top 10 most unequal countries. As measured by the Gini Coefficient, consumption inequality decreased from 60.5 in 2009 to 54.9 percent in 2016 (using official figures). This primarily reflected pro-poor growth in urban areas, especially cities and towns, where the Gini declined from 59.9 to 48.7 percent, with Francistown reaching 43.4 percent (Figure 2.11, a and b). The Gini coefficient for rural villages declined less than four percentage points; thus, inequality remained high at 53.3 percent, with urban villages at 51.1 percent. The North-West region has shown the strongest improvements in reducing inequality, surpassing the South-East, South-West, and North-East in 2016. Nonetheless, as discussed above, part of the drop in overall inequality was explained by declines in consumption levels among the highest deciles more so than by consumption improvements among the poor. In addition, among its structural peers, only South Africa and Namibia register worse inequality rates, and Botswana remains among the top 10 most unequal countries in the world (Figure 2.12). International experience suggests that Botswana would find it hard to reach an inclusive growth path and high-income status without addressing these levels of inequality and poverty. 2.11 FIGURE2.11 FIGURE   Inequality improved across Botswana, with better outcomes in cities and towns a. Gini coefficient, by strata b. Gini coefficient, by regions 70 75 2002 2009 2015 64.7 70 65 61.3 65 71,5 60.5 60.9 59.9 60 58.4 58.3 60 62,5 62,1 62,1 56.8 60,5 60,3 59,3 58,8 58,9 54.9 55 58,3 58,3 58,2 57,0 55 53.3 55,4 54,6 51.1 50 52,2 52,1 50 48.7 45 46,7 46,5 46,3 40 43,4 45 35 40 30 National Cities & Urban Rural Gaborone Francistown Other towns South East North East North West South West Gaborone Francistown Other towns South East North East North West South West Gaborone Francistown Other towns South East North East North West South West Towns Villages Villages 2003 2009 2016 Source: 2002/03 Household Income and Expenditure Survey (HIES), 2009/10 Botswana Core Welfare Indicators Survey (BCWIS), and 2015/16 Botswana Multi-Topic Household Survey (BMTHS). FIGURE 2.12 FIGURE 2.12  Botswana’s inequality is among the highest in the world Gini Coefficient, most recent estimate by country (%) S. Africa 0.7 Namibia Botswana Congo, Rep. 0.6 0.5 Gabon Mongolia Georgia Lebanon Tunisia 0.4 0.3 0.2 0.1 0 South Africa Colombia Botswana Brazil Angola Mozambique Congo, Rep. Guatemala Costa Rica Ecuador Guyana Lesotho Congo, Dem. Rep. South Sudan Rwanda Mexico Chile Madagascar Cameroon Djibouti Bolivia Argentina Philippines Uruguay Peru Jamaica United States Suriname Bulgaria El Salvador Samoa Malawi Gabon Togo Burundi Chad Solomon Islands Mauritius Lithuania Viet Nam Sierra Leone Marshall Islands Liberia Ethiopia Italy Portugal Australia Latvia Georgia Spain Switzerland West Bank and Gaza North Macedonia Bangladesh Bosnia and Herzegovina Niger Nepal Luxembourg Nauru Seychelles Egypt, Arab Rep. Lebanon Germany France Korea, Rep. Cyprus Fiji Myanmar Sweden Pakistan Albania Kazakhstan Kosovo Kyrgyz Republic Poland Denmark Kiribati Finland Tonga Belgium Czechia United Arab Emirates Moldova Belarus Slovak Republic Source: Poverty and Inequality Platform, pip.worldbank.org. See Appendix 1 for information on structural peers. BOTSWANA POVERTY ASSESSMENT 19 2.2  DEMOGRAPHIC, EDUCATION, AND LABOR MARKET CHARACTERISTICS OF THE POOR, NEAR POOR, AND NON- POOR The poor are increasingly more concentrated in rural areas despite a surge in rural-to-urban migration. In 2016, over a third of Batswana (34.9 percent) lived in rural areas, a drop from 43 percent five years earlier. Nonetheless, between 2009 and 2016, the share of the poor living in rural areas increased from 56.2 to 58.2 percent. Remote areas like the North-West and Ghanzi districts, already the poorest districts (although not housing the poorest villages in 2011, see Box 2.2), experienced increased levels of poverty along with population growth, resulting in an increase of the share of the poor from 14 to almost 24 percent in 2016 (Figure 2.13 and Table 2.1).48 Although the poverty rate only increased by 1.5 percentage points in the Kweneng district, its population share increased from 16.3 to 16.9 percent, resulting in a strong increase in the share of the poor population from 17 to 22.6 percent. On the other hand, poverty fell in and around cities to the south- and north- east. The Kgatleng, Central, and South-East districts saw some of the most substantial declines in poverty rates between 2009 and 2016. The Central district’s population share declined from 33 to 31 percent, but its share of the poor population declined more strongly from 39 to 30 percent in the period. The South-East, where the capital of Gaborone is located, had a decline in the share of the poor population from 7.5 to 3 percent while being home to 17 percent of the population of Botswana. The regional convergence in poverty seen in 2009 has dissipated as welfare disparities across Botswana have grown. FIGURE 2.13 FIGURE 2.13   Regional poverty is diverging, increasing in the west and decreasing in the east a. Poverty rate by region 2009 (official line, %) b. Poverty rate by region 2016 (official line, %) 33–39 33–39 28–33 28–33 North-West North-West 23–28 23–28 Sowa 18–23 Sowa 18–23 North-East 13–18 North-East 13–18 Francistown Francistown 8–13 8–13 Central 3–8 Central 3–8 Ghanzi Ghanzi 0–3 0–3 Kweneng Kweneng Kgalagadi Kgatleng Kgalagadi Kgatleng Jwaneng Gaborone Jwaneng Gaborone Southen Lobatse Southen Lobatse South-East South-East Source: World Bank calculations using the 2009/10 Botswana Core Welfare Indicators Survey (BCWIS) and 2015/16 Botswana Multi- Topic Household Survey (BMTHS). 48  The Kgalagadi district also had a strong increase in the poverty rate, but the decline in the population between 2009 and 2016 (in an already sparsely populated area with desert conditions) resulted in around the same number of poor people. This suggests that the non- poor left the district, possibly affected by drought conditions and in search of better opportunities. 20 BOTSWANA POVERTY ASSESSMENT TABLE 2.1  The number of poor increased in the west while declining in the Central and South-East districts Poverty Rate by Region (national poverty line, %) 2009 2016 Region Poverty Rate Number of Poor Poverty Rate Number of Poor Southern 19.6 35,031 17.4 35,832 South East 8.9 27,035 2.9 10,225 Kweneng 20.1 61,559 21.6 75,570 Kgatleng 19.5 16,221 11.2 10,557 Central 22.9 141,604 15.4 100,327 North East 12.7 20,288 8.3 11,561 North West 30.3 39,997 31.7 61,385 Ghanzi 26.0 10,509 38.5 17,356 Kgalagadi 19.7 10,232 26.1 10,843 Source: World Bank calculations using 2009/10 BCWIS and 2015/16 BMTHS. BOX 2.2  Poverty Map 2011 The poorest ten percent of villages in 2011 were primarily rural and located in southeastern Botswana, but many could also be found spread throughout the country. As shown earlier, nearly one-fifth of Batswana (or more than 362,000 people) lived below the official poverty line in 2009.49 This corresponded to a quarter of the rural population, 20 percent of the population in urban villages, and 8 percent in cities and towns. The North-West and Ghanzi districts had the highest poverty rates, at 30.3 and 26 percent, respectively, accounting for more than 50,000 poor people (14 percent of the poor population).50 However, the 2011 poverty map51 of Botswana shows that most of the poorest villages were found in the Southern district (Figure B2.1 and Figure B2.2). While the Southern district had an overall poverty rate of 19.6 percent in 2009, representing 35,000 people (or less than 10 percent of all the poor), the 20 poorest villages (18 of which were in the Southern district) had poverty rates between 62 and 77 percent.52 Botswana’s 500 “villages” (some of which represent cities or towns) were ranked from poorest to richest and grouped into ten deciles of 50 villages each, with decile 1 representing the poorest 50 villages and decile 10 the wealthiest 50 villages. Villages from the poorest decile (red dots) are spread throughout the map and are the ones that require the most urgent attention. The poorest decile included predominantly rural villages (85 percent), while the rural share of the following three poorest deciles varied from 76 to 92 percent (Figure B2.3). 49  The only available poverty map is based on the 2009/10 survey and 2011 Census, therefore this section focuses on 2009 poverty to provide the spatial context and later incorporates more recent data. 50  As shown earlier, by 2016 the poverty rates in the North-West and Ghanzi districts had increased to 31.7 and 38.5, respec- tively, and together accounted for almost 79,000 poor people or 23.6 percent of all the poor. 51  A poverty map allows for an estimation of poverty rates at much more disaggregated geographical levels than the household surveys alone can provide. Poverty maps combine a household survey and a population census (usually collected within a few years of each other) to provide a highly detailed spatial distribution of poverty within a country. The population census provides complete coverage of a country’s population and includes some data on infrastructure and socio-economic characteristics, while the survey data provides the detailed consumption data needed for poverty analysis but lacks the highly detailed level of spatial disaggregation. The accuracy and precision of a poverty map rely on the degree of comparability between the variables in the census and the survey, and the power of these comparable variables to predict consumption in the survey data. 52  These 20 villages have relatively small populations and therefore they represent only 3.5 percent of the poorest population. BOTSWANA POVERTY ASSESSMENT 21 BOX 2.2 (cont.) FIGURE B2.1 The wealthiest ten percent of villages were 96 FIGURE B2.1  2011 Poverty Map by Village Deciles percent urban and almost all were in or near the main cities of Gaborone and Francistown. These villages (dark green dots) were primarily in the Southern, South-East, and Kgatleng decile districts (all near the capital of Gaborone) as 1 well as in the North-East district (home to the 2 second largest city of Francistown, Figures B2.1 3 4 and B2.2). The South-East district, home to 5 Gaborone, had a low poverty rate of 8 percent 6 in 2009 but represented around 27,000 poor 7 8 people given its large population size. The North- 9 East district had the second lowest poverty 10 rate in 2009 at 12.7 percent with 20,000 poor people. By 2016, poverty in these two districts had declined to 2.9 percent in the South-East and 8.3 percent in the North-East. Source: World Bank calculations based on the 2011 Botswana poverty map, World Bank (2015). https://www.statsbots.org. bw/sites/default/files/publications/POVERTY%20Mapping%20 2010_May%2028%202015.pdf. Note: decile 1 = 50 poorest villages; decile 10 = 50 richest villages. FIGURE B2.2  The poorest 50 villages were primarily in the Southern district while the FIGURE least B2.2 poor included Gaborone and Francistown and their surroundings, 2011 a. Decile 1 b. Decile 10 Source: World Bank calculations based on the 2011 Botswana poverty map, World Bank (2015). Note: decile 1 = 50 poorest villages; decile 10 = 50 richest villages. 22 BOTSWANA POVERTY ASSESSMENT BOX 2.2 (cont.) FIGURE B2.3 The urban population has grown steadily, FIGURE B2.3  The four poorest deciles had the primarily in urban localities outside of cities, largest rural shares, 2011 and is projected to be around 70 percent of the total population in 2022. Preliminary 100 90 estimates from the 2022 Population Census 80 show a population of 246,325 for Gaborone Percent rural (%) 70 and 102,444 for Francistown.53 Gaborone is the 60 country’s economic hub, home to government 92.2 85.01 82.89 50 76.34 offices, ministries, institutions, diamond cutting 40 59.67 56.29 56.5 and polishing operations, and diamond sorting 30 34.86 33.68 and trading centers. The mining industry also 20 6.08 dominates Francistown. Botswana’s relatively 10 small cities lack economic diversification and 0 1 2 3 4 5 6 7 8 9 10 density, making it difficult to achieve economies Decile of scale. Unemployment levels in 2022 are also Source: World Bank calculations based on the 2010 Botswana higher than in 2009 or 2016. Despite population poverty map, World Bank (2015). Note: decile 1 = 50 poorest villages; decile 10 = 50 richest villages. growth in cities and towns between 2011 and 2022, these areas represent a smaller share of the total population, declining from 21.7 percent in 2011 to 19.2 percent in 2022. The proportion of Batswana living in Gaborone declined by a percentage point from 11.4 to 10.4 percent over the same period. Instead, neighboring localities (including Kweneng East and the census districts of Kgatleng and South-East near Gaborone) gained a share of the total population as these areas provide alternative accommodation and business opportunities. Poverty remains highest among children, large families, female-headed households, and people with lower levels of education. Between 2009 and 2016, the share of poor people in female-headed households declined from 58.1 to 56.8 percent. In contrast, the share with secondary or higher education increased from 14.6 to 20.4 percent as education improved for the poor and non-poor (Table 2.2). The poverty rate increased with household size, from 8 percent for households of four people to 34 percent for households with seven or more people (Figure 2.14). However, these poverty rates are all lower than in 2009. Households whose head had low levels of education faced poverty rates of 21 percent on average, down from 24 percent in 2009. In contrast, those with secondary or tertiary education had much lower poverty rates (13 and 2 percent, respectively, Figure 2.15). Nonetheless, nearly 4 out of 5 poor individuals continue to have low levels of schooling (Table 2.2). At 12 percent, working-age individuals continued to have the lowest poverty rate, in contrast to nearly 1 in 4 children ages 0-5 living in poverty, albeit down from 27 percent in 2009 (Figure 2.16 and Figure 2.17). The working poor were primarily employed in services while the share in agriculture declined by more than half in 2016. Since 2009, the share of the working poor in agriculture decreased from 44 to 18 percent (Figure 2.18). In contrast, 72 percent of the working poor were employed in the service sector, primarily in public administration (28 percent) and other services (29 percent). The latter, in large part, include domestic workers. Overall, employment in public administration grew by 34.7 percent between 2009 and 2016, with 16 percent of this increase coming from the working poor. Approximately half of rural public administration includes 53  Updated (but still preliminary) Census 2022 estimates as of December 2022. See statsbots.org.bw December 2022. BOTSWANA POVERTY ASSESSMENT 23 low-paying Ipelegeng workfare program workers, relative to only 20 percent of urban workers.54 This suggests that the working poor pushed out of agriculture by the 2015 shocks sought income via the workfare Ipelegeng program in rural areas or as domestic personnel in private households. As shown in Chapter 1, however, the decline in agricultural employment seems to have been temporary. TABLE 2.2  Profile of the Poor: Official Poverty Line 2002/03 2009/10 2015/16 Region Non- Non- Non- Poor Total Poor Total Poor Total poor poor poor Demographics Age (household head) 25.5 22.9 24.7 27.6 22.3 26.6 27.6 22.5 26.7 Share living in households 53% 46% 51% 52% 42% 50% 49% 43% 48% headed by a male Household size 5.6 8.1 6.4 4.9 8.5 5.6 5.0 9.0 5.6 Share Living in Cities and 30% 6% 23% 24% 7% 21% 24% 4% 21% Towns Share Living in Urban Villages 36% 28% 33% 36% 37% 36% 45% 37% 44% Share Living in Rural Villages 34% 66% 44% 40% 56% 43% 30% 58% 35% Education (household head) Low education (primary or 66% 94% 75% 57% 85% 63% 47% 78% 52% less) Mid education (secondary) 34% 6% 25% 27% 13% 24% 31% 17% 29% High education (+ than 16% 2% 13% 21% 3% 18% secondary) Sector (household head) Agriculture and Fishing 16% 42% 21% 24% 44% 26% 11% 18% 12% Mining 3% 1% 3% 3% 1% 3% 2% 0% 2% Manufacture and Utilities 11% 10% 11% 8% 5% 8% 7% 5% 7% Construction 10% 7% 9% 7% 6% 7% 5% 4% 5% Wholesale and Retail 14% 11% 13% 10% 10% 10% 15% 11% 14% Public Administration 15% 11% 15% 17% 13% 16% 19% 28% 19% Education 9% 4% 8% 6% 1% 5% 8% 4% 8% Other services 21% 15% 20% 25% 20% 24% 33% 29% 32% Type of employment (household head) Wage employee 40% 15% 34% 36% 17% 33% 40% 20% 37% Farm self-employed 5% 8% 6% 7% 10% 8% 5% 5% 5% Non-farm self-employed 7% 6% 7% 8% 6% 8% 9% 4% 8% Access to services Water 96% 95% 96% 97% 95% 96% Sewage 38% 14% 34% 51% 31% 48% Electricity 35% 3% 25% 49% 24% 44% 71% 30% 64% Welfare Daily consumption per capita 13.2 1.4 9.6 13.3 1.5 11.0 11.3 1.6 9.7 (PPP 2017) Source: World Bank calculations using 2002/03 Household Income and Expenditure Survey (HIES), 2009/10 Botswana Core Welfare Indicators Survey (BCWIS), and 2015/16 Botswana Multi-Topic Household Survey (BMTHS). 54  Nearly 26,000 out of an estimated 54,000 employed in rural public administration. 24 BOTSWANA POVERTY ASSESSMENT FIGURE 2.14 FIGURE 2.15 FIGURE 2.14  Poverty Rate by Household Size (%) FIGURE 2.15  Poverty Rate by Education Level of Household Head (%) 50% 44.9 50% 45% 38.6 36.8 40% 40% 33.8 31.9 35% 30% 30% 24.3 24.7 21.2 25% 20.1 19.2 19.7 17.1 15.5 15.9 20% 20% 14.6 13.0 12.4 15% 10.1 9.3 7.8 7.3 10% 10% 4.5 3.9 4.1 2.6 2.4 2.2 5% 1.1 0.0 0% 0% 2003 2009 2016 2003 2009 2016 HhSize== 1 HhSize== 2 HhSize== 3 HhSize== 4 Education-Low Education-Mid Education-High HhSize== 5 HhSize== 6 HhSize>= 7 Source: World Bank calculations using 2002/03 BHIES, 2009/10 Source: World Bank calculations using 2002/03 BHIES, 2009/10 BCWIS, and 2015/16 BMTHS. BCWIS, and 2015/16 BMTHS. FIGURE 2.16 FIGURE 2.17 FIGURE 2.16  Poverty Rate by Age Group (%) FIGURE 2.17  Poverty Rate by Labor Force Status of Household Head (%) 50% 50% 42.1 37.5 37.6 35.2 40% 34.4 40% 31.6 30.9 28.4 26.8 26.7 27.2 30% 24.7 30% 23.7 22.9 21.9 20.8 21.6 21.1 19.2 19.0 18.4 17.9 17.0 15.9 16.2 15.6 20% 20% 14.5 14.2 13.3 11.6 11.6 10.6 7.3 10% 10% 0% 0% 2003 2009 2016 2003 2009 2016 [0-5] [6-14] [15-19] [20-24] [25-54] [55-64] [65+] Employed Unemployed Inactive Less than 15 years Source: World Bank calculations using 2002/03 BHIES, 2009/10 Source: World Bank calculations using 2002/03 BHIES, 2009/10 BCWIS, and 2015/16 BMTHS. BCWIS, and 2015/16 BMTHS. 2.18 FIGURE2.18 FIGURE   Employment by Sector and Poverty Status (%) Poor Non-Poor 100% 100% 15 20 27 25 4 29 33 80% 1 80% 11 12 13 6 4 11 8 60% 10 60% 20 17 7 28 6 19 10 5 18 10 40% 40% 7 11 12 15 8 4 14 20% 42 44 5 20% 5 7 18 24 21 11 0% 0% 2003 2009 2016 2003 2009 2016 Agriculture Mining Manuf-Utilities Agriculture Mining Manuf-Utilities Construction Wholesale-Retail PublicAdmin Construction Wholesale-Retail PublicAdmin Education RestServices Education RestServices Source: World Bank calculations using 2002/03 Household Income and Expenditure Survey (HIES), 2009/10 Botswana Core Welfare Indicators Survey (BCWIS), and 2015/16 Botswana Multi-Topic Household Survey (BMTHS). BOTSWANA POVERTY ASSESSMENT 25 2.3  DRIVERS OF POVERTY AND INEQUALITY Over the last ten years, a more equal distribution of consumption contributed strongly to poverty reduction, while low or negative consumption growth in recent years pushed poverty higher. Between 2003 and 2009, consumption growth explained over a third of poverty reduction while a better distribution of consumption contributed two-thirds. Conversely, negative consumption growth in 2009-16 increased poverty by 3.5 percentage points (p.p.), while a more equal distribution of consumption more than offset this effect, resulting in lower overall poverty (Figure 2.19). In rural areas, the negative growth of mean per capita consumption had a larger negative impact on poverty (+8.3 p.p) than the positive impact of an improved distribution of consumption (-6.8 p.p), resulting in an overall increase in rural poverty between 2009 and 2016 (Figure 2.20). In urban areas, these two effects also worked in opposite directions, but the improved distribution of consumption was larger and resulted in a decline in urban poverty. In both areas and time periods, price effects had minimal impact on the poverty results. Labor income was the main driver of poverty reduction in urban areas between 2009 and 2016, while the significant decline in the share of employed adults in rural households increased rural poverty. In urban areas, 80 percent of the poverty reduction came from improvements in labor income, with non-labor income and a higher share of adults in the household (that is, a lower dependency ratio) also contributing (Figure 2.21). This aligns with the labor market analysis presented in Chapter 1; urban areas register better labor market outcomes than rural areas. In rural areas, the significant decline in the share of employed adults in rural households pushed poverty higher. The change in the consumption-to-income ratio was also correlated with higher rural poverty as poor rural households faced lower consumption levels and an increased proportion worried about not having enough food. Although income from employment is central to reducing poverty and inequality, social protection programs in Botswana have a significant impact. Without Botswana’s social protection transfers (from 29 programs across nine ministries), the poverty rate of 16 percent in 2016 would have been almost 24 percent, and the poverty gap would have been 9.5 instead of 4.6 percent. Removing just the thirteen social assistance programs would increase the poverty headcount by nearly a third (to 22.8 percent). The primary school feeding program, which reaches approximately 269,000 children, and the old-age pension, which reaches more than FIGURE 2.19  Decomposition of the change in FIGURE 2.20  Decomposition of the change 2.19growth, distribution, and price poverty into FIGURE in poverty FIGURE into growth, distribution, and price 2.20 effects, National, 2003-2009 and 2009-2016 effects, by area, 2009-2016 (percentage points) (percentage points) 5.0 3.5 10.0 8.3 0.5 0.1 0.0 5.0 3.5 3.3 1.6 -3.2 0.07 0.08 0.08 -5.0 -4.1 0.0 -6.8 -7.6 -3.2 -10.0 -5.0 -4.5 -11.2 -6.8 -6.8 -7.8 -15.0 -10.0 2003-2009 2009-2016 National Urban Rural Total change Growth Effect Total change Growth Effect Distribution Effect Price Effect Distribution Effect Price Effect Source: World Bank staff calculations. Source: World Bank staff calculations. 26 BOTSWANA POVERTY ASSESSMENT FIGURE 2.21 FIGURE 2.21  Shapely decomposition of changes in poverty by income source (2009-2016) 14 9 Percentage points 4 Non Labor Labor Share of employed adults -1 Share of adults Consumption-to-income ratio Total -6 Urban Rural Source: World Bank staff calculations. 126,000 people (about 5.5 percent of the population in fiscal year 2020), have the largest impact on poverty, but some of the lowest levels of expenditure (see Chapter 5).55 Labor market factors, particularly skills differences, were the primary contributors to inequality in 2016, whereas demographics and education had been the largest contributors in 2009. Labor market factors contributed the most (35.6 percent) to inequality in 2016, up from 20.3 percent in 2009, followed by demographics (27.9 percent), education (25.9 percent), and location (10.6 percent, Figure 2.22).56 In contrast, for other countries in the Southern Africa Customs Union (SACU), differences in educational attainment among adult household members were the most important driver of overall inequality. Differences in occupation type (professionals and senior managers, which suggest differences in skills or abilities) explain 29 percent of total inequality relative to only 6 percent for labor force participation. Post-secondary education, with its high earnings, explains 24 percent of total inequality,57 followed by differences in age (17.4 percent) and location (10.6 percent). The age factor contributed less to inequality, suggesting the “demographic dividend” (more working-age household members and fewer dependents) became more even across households since 2009. In addition, an increase in the share of tertiary education among adults in Botswana may have helped reduce the contribution of higher education to inequality relative to the contribution of the labor market. Lastly, unlike in other SACU countries, location increased in importance as a source of inequality in 2016, explained primarily by a divergence in inequality across regions in Botswana. Wage income is the primary driver of inequality, more so than in the rest of SACU, suggesting declines in inequality may be partially explained by smaller wage gaps. Wage income accounts for 85 percent of inequality58 (both at the national level and for rural areas) (Figure 2.23), a rate higher than the 72.3 percent average for SACU members. A marginal change in wage income is estimated to change the Gini Coefficient by 5.2 percent at the national level and 8.3 percent in rural areas (Figure 2.24). Social protection transfers 55  For more detail on Botswana’s social protection programs, see World Bank (2022c). 56  See Sulla et al. (2022) (Inequality in Southern Africa: An Assessment of the Southern Africa Customs Union). This decomposition of inequality is based on a technique proposed by Fields (2003), which adopts a regression-based approach to estimate standard income- or consumption-generating equations. The main drivers of inequality can be identified from the contributions of explanatory variables (such as education, labor market factors, and demographics) to the distributional changes in welfare aggregates captured by the size of the es- timated coefficients (Heshmati 2004). The estimated coefficient of each variable in the regression captures its estimated share in overall inequality. Labor market factors include labor force status (“participation”; whether people work or not), industry of employment, and occupation type (proxying for skills or abilities). 57  Individuals with university-level education, who in the majority are English-speaking non-nationals (34 percent of non-nationals have a university degree as compared to 18 percent of nationals), show much higher levels of consumption expenditure. The costs of sending children to school, both direct and indirect, are higher (in proportion to their average consumption) for rural and less affluent households and become prohibitive at higher levels of education. (Sulla et al. [2022]) 58  See Sulla et al. (2022). The decomposition of inequality by income sources follows Lerman and Yitzhaki (1985) and Stark and others (1986). A module developed by López-Feldman (2008) implements this approach in Stata. BOTSWANA POVERTY ASSESSMENT 27 FIGURE 2.22 FIGURE 2.22  Decomposition of Inequality 45 Gender 39.1 graphics Demo- 40 Hh Size 35.1 35.6 35 Age 30 27.9 Secondary Education 25.9 Primary 25 20.3 Post-Secondary 20 Industry market 15 Labor Participation 10.6 10 Skills 5.5 Urban 5 Loca- tion Region 0 Demographics Education Labor market Location -10 0 10 20 30 40 50 2010 2015 2010 2015 Source: Inequality in Southern Africa: An Assessment of the Southern Africa Customs Union (World Bank 2022). FIGURE 2.23 FIGURE 2.23  Decomposition of Inequality by FIGURE2.24 FIGURE 2.24  Marginal effect of income source Income Source in Botswana at National level change and for Rural Households (%) 100 85.1 84.6 Wage 5.2 Income 8.3 80 Business -0.6 60 Income -1.4 Social -1.3 40 Protection -1.5 Transfers 20 8.7 7.3 -3.3 4.7 4.7 Remittances -5.4 1.5 3.4 0 Wage Business Social Remittances -10 -5 0 5 10 Income Income Protection Transfers National Share Rural Share National Elast Rural Elast Source: World Bank calculations based on Sulla et al. (2022). Source: World Bank calculations based on Sulla et al. (2022). Numbers vary slightly due to the use of a methodology that Numbers vary slightly due to the use of a methodology that considers survey weights. considers survey weights. are a larger share of total inequality than remittances (4.7 percent versus 1.5 percent). Yet, on the margin, their equalizing effect on incomes is small and smaller than that of remittances (3.3 versus 1.3 percent).59 Nonetheless, the impact of social protection transfers and remittances is small relative to the disparities caused by differences in wage income. The marginal elasticity for rural households is 3.6 times larger for remittances than for social protection transfers despite smaller shares of total inequality. These results are consistent with the significant contributions of labor market outcomes to inequality. These results suggest that beyond continued improvements in educational attainment and dependency ratios, reducing inequality will require policies that reduce differences in wage incomes, for example, by strengthening skills and abilities and by reducing wage differences between the public and private sectors. 59  These estimates are slightly different than those reported in Sulla et al. (2022) (the regional report “Inequality in Southern Africa”) because a different methodology was used that takes into account survey weights. 28 BOTSWANA POVERTY ASSESSMENT FIGURE 2.25 FIGURE 2.25  Gini Coefficient by Country and Income Source 0.80 0.70 0.632 0.60 0.599 0.50 0.530 0.40 0.30 0.20 Romania (2016) Croatia (2014) Mauritius (2017) Uruguay (2009) Panama (2016) Montenegro (2015) Belarus (2015) Armenia (2017) Jordan (2017) Russia (2014) Albania (2015) Venezuela (2013) Indonesia (2017) Iran (2011) Turkey (2016) Ecuador (2011) Peru (2011) Guatemala (2014) Argentina (2017) Georgia (2013)* Costa Rica (2010) Dominican Republic (2013) Paraguay (2014) Mexico (2014) Colombia (2017) China (2014)* Brazil (2009) Namibia (2016) Botswana (2010) South Africa (2015) Kyrgyz Republic (2016) Myanmar (2017) Ukraine (2016) Egypt (2015) Moldova (2017) Sri Lanka (2009) India (2011) Tanzania (2011) El Salvador (2017) Ivory Coast (2015) Mongolia (2016)* Ghana (2012) Tunisia (2010) Comoros (2014) Bolivia (2015) Nicaragua (2009) Kenya (2015) eSwatini (2017) Lesotho (2017) Honduras (2011) Zambia (2015) Guinea (2012)* Niger (2014)* Tajikistan (2015) Ethiopia (2016) Gambia (2016) Burkina Faso (2014) Togo (2015) Uganda (2016) Mali (2014)* Market Income Disposable Income Consumable Income Final Income Source: Adapted from World Bank 2022a. Poverty and Shared Prosperity Report, 2022. Fiscal policy has also contributed to reducing inequality. The Commitment to Equity (CEQ )60 methodology traces how inequality evolves as different transfers and taxes are added or subtracted from welfare aggregates. Botswana’s most recent CEQ analysis uses the 2009/10 BCWIS.61 Results show a 10-percentage point decrease in the Gini Coefficient after accounting for the fiscal policy impact, with education and health transfers contributing two-thirds of this reduction (Figure 2.25). Botswana’s direct taxes and transfers contributed three percentage points to reducing inequality in 2009, reflecting a stronger impact on inequality than in two-thirds of the 60 countries where the CEQ framework has been implemented. Nevertheless, Botswana still has one of the highest levels of inequality (after South Africa). 60  The CEQ methodology is a diagnostic tool that helps identify how fiscal policy affects equity. Three different income stages are estimated: market income (before any fiscal policy), disposable income (after direct taxes and transfers), and consumable income (after indirect taxes and subsidies). Changes in the Gini Coefficient between the different concepts quantifies the distributional impact of taxes and transfers. Lastly, final income also incorporates health and education expenditure (Lustig and Higgins, 2016). 61  International Monetary Fund 2018. https://commitmentoequity.org/wp-content/uploads/2020/05/cr18268.pdf. The 2009/10 sur- vey used for the CEQ analysis in Botswana corresponds to a year of fiscal stimulus which could amplify the fiscal incidence on welfare aggregates. If there have been significant changes in the structure of household income and expenditures or the structure of taxation and social spending since 2010, these findings would not reflect accurately how fiscal policy impacts Botswana today. BOTSWANA POVERTY ASSESSMENT 29 CHAPTER 3 NON-MONETARY DIMENSIONS OF POVERTY AND INEQUALITY 30 BOTSWANA POVERTY ASSESSMENT T he chapter presents non-monetary measures of welfare to capture deprivations beyond consumption and income, including education and basic infrastructure such as electricity, water, and sanitation. It also analyzes the impact of drought and rainfall shocks on poor villages. The first section combines the 2011 poverty map and census to examine spatially disaggregated village-level outcomes. The results show that access to electricity and sanitation was low and highly unequal in 2011, whereas access to water was almost universal. The following section focuses on urban-rural trends: poor households had important improvements in access to sanitation and little improvement in access to electricity in 2016, while levels overall remain low, especially in rural areas. Satellite data then shows that despite policy goals of universal access, electricity remained unequal and city-centric in 2019. The Human Opportunity Index, which focuses on opportunities among children, highlights the need to target quality infrastructure programs in rural areas. Health and education outcomes show improvements but face quality challenges, and the Human Capital Index remains considerably lower than the average for countries with similar income. The Multidimensional Poverty Measure then combines data on monetary poverty, basic infrastructure, and education to monitor a broader measure of poverty and shows bigger improvements than monetary poverty alone. However, large gaps remain relative to peers. The last section’s drought and rainfall spatial data shows stronger negative impacts on the poorest locations, particularly during a year of known increases in rural poverty (2015/16) but also during 2019, in line with projected poverty increases. 3.1  ACCESS TO ELECTRICITY AND SANITATION AT THE VILLAGE LEVEL IS LOW AND HIGHLY UNEQUAL Access to electricity and improved sanitation was low and highly unequal across households in 2011, while access to improved water was more universal. The 2011 Population and Housing Census showed large differences in access to electricity and sanitation across the country, with access highest among households living in or near the largest cities, such as Gaborone and Francistown in the east (Figure 3.1.a and Figure 3.1.c). A few western and northern villages also showed relatively high access to services (Figure 3.1.a). Many villages had limited access to electricity, with fewer villages having higher access rates (Figure 3.1.b). In contrast, for sanitation, a comparable number of villages had varying degrees of access (Figure 3.1.d). The 2011 Census also showed access to improved water was near universal (95.7 percent), with high levels of access for villages across the country (Figure 3.1.e and 3.1.f). In terms of poverty, households in poorer villages experienced worse access rates to electricity and sanitation than those in wealthier villages but similar rates of access to improved water (Figure 3.2). FIGURE 3.1 FIGURE 3.1  Access to electricity and improved sanitation is highly unequal while access to improved water is more universal (2011 Census) a. Access to electricity (%), 2011 b. Number of villages at each percentage of access 70 60 57 60 50 Village Count 40 37 34 33 30 29 24 23 21 22 acce_elec 20 18 18 17 0 to 20 15 14 12 20 to 40 9 9 9 10 8 40 to 60 6 60 to 80 4 4 5 1 1 0 1 1 1 80 to 100 0 0 25 50 75 100 Percentage access to electricity c. Access to improved sanitation (%), 2011 d. Number of villages at each percentage of access 34 33 Village 30 29 24 23 21 22 acce_elec 20 18 18 BOTSWANA POVERTY ASSESSMENT 17 31 0 to 20 15 14 12 20 to 40 9 9 9 10 8 40 to 60 6 60 to 80 4 4 5 1 1 0 1 1 1 80 to 100 0 0 25 50 75 100 FIGURE 3.1 (cont.) Percentage access to electricity c. Access to improved sanitation (%), 2011 d. Number of villages at each percentage of access 30 27 26 25 25 24 24 24 23 22 20 20 19 19 19 19 Village Count 18 18 17 16 16 15 15 15 14 14 13 12 improved 10 10 9 sanitation 0 to 20 6 20 to 40 5 5 40 to 60 60 to 80 2 2 80 to 100 0 0 25 50 75 100 Percentage access to improved sanitation e. Access to improved water (%), 2011 f. Number of villages at each percentage of access 300 251 250 200 Village Count 150 126 improved 100 water 0 to 20 20 to 40 50 38 40 to 60 60 to 80 19 24 10 80 to 100 2 0 1 1 0 1 0 1 0 0 0 0 0 0 2 2 1 0 0 0 2 3 6 3 0 0 25 50 75 100 Percentage access to improved sanitation Source: World Bank calculations based on the 2011 Botswana poverty map, Statistics Botswana (2015) and the 2011 Population and Housing Census. Note: decile 1 = 50 poorest villages; decile 10 = 50 richest villages. FIGURE 3.2 FIGURE 3.2  Poor villages had much lower access to electricity and improved sanitation than wealthy ones, whereas access to improved water was almost universal Electricity Improved Sanitation Improved Water 100 Average access to service 90 80 70 60 50 40 30 20 10 0 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Deciles Source: World Bank calculations using the 2011 Population and Housing Census and the 2011 Poverty Map. 32 BOTSWANA POVERTY ASSESSMENT 3.2  ACCESS TO BASIC SERVICES INCREASED IN 2016 BUT SOME SERVICES REMAIN LOW FOR AN UPPER-MIDDLE-INCOME COUNTRY Despite efforts to achieve universal access by 2040, at 65 percent, access to electricity in 2016 is low relative to Botswana’s per capita income. The 0.005 US$/kWh tariff on electricity customers to connect low-income households to the grid has not been enough to fund universal grid access, and off-grid solutions have yet to be promoted in Botswana. In addition, the COVID-19 pandemic in 2020 and 2021 interrupted and slowed the long-term trend of increasing access by approximately 1.5 percentage points per year. In rural areas, access to electricity was only 35.2 percent in 2016, leading many rural households to rely on wood fuel for cooking, heating, and lighting, resulting in hazardous indoor air pollution. At 80.4 percent, urban areas had much higher access to electricity in 2016. Moreover, despite improvement over time, only 30 percent of poor households had access to electricity in 2016, compared to 71 percent of non-poor households (Table 2.2 and Figure 3.3). In addition, improvements in access to electricity in 2009 -16 favored non-poor households. Low levels of access to electricity in rural areas and among poor households result in less productivity and lower levels of development. Botswana has near universal access to safe drinking water, yet supply is limited to certain hours and days in many parts of the country. As of 2016, 96.3 percent of households had access to improved drinking water (Figure 3.3) and 9 out of 10 households received water from a piped network. Even though access rates align with structural peers, water supply is limited to eight hours per day, three days a week in many parts of the country, forcing households with a piped connection to use alternative sources such as mobile water tanks. In addition, groundwater is overexploited, and saline intrusion is a recurrent problem across Botswana. Critical industries like agriculture, mining, and tourism depend on an adequate water supply.62 Moreover, the 2015 drought highlighted the country’s vulnerability to climate change, as cereal crop production declined by 70 percent and livestock mortality increased by 20 percent. Botswana is forecast to become highly water-stressed by 2040.63 If no action is taken, the annual water supply-demand gap will increase from an estimated 20Mm3 in 2020 to 114Mm3 by 2035. FIGURE B3.3 FIGURE 3.3  Poor households face significant differences in access to electricity and sanitation a. Access to electricity b. Access to safe drinking water c. Access to improved sanitation 96.6 96.3 94.8 96.4 96.1 95.1 70.5 80 100 100 64.2 70 90 90 80 80 49.0 60 70 70 44.1 51.2 50 47.9 60 60 38.3 30.1 40 50 50 33.6 30.8 23.7 30 40 40 30 30 20 14.1 20 20 10 10 10 0 0 0 Total Non- Poor Total Non- Poor Total Non- Poor Total Non- Poor Total Non- Poor Total Non- Poor poor poor poor poor poor poor 2009 2016 2009 2016 2009 2016 Source: World Bank calculations using 2009/10 BCWIS and 2015/16 BMTHS. 62  The livestock subsector accounts for 66 percent of agricultural water use. Mining’s water consumption share is third after agriculture and domestic consumption. 63  The Water Resources Institute indicates that water stress levels for Botswana, Namibia, South Africa, and Lesotho will reach levels between 40-80 percent by 2040. Available at: https://www.wri.org/insights/ranking-worlds-most-water-stressed-countries-2040 BOTSWANA POVERTY ASSESSMENT 33 Access to improved sanitation has expanded but remains quite low, especially among poor households. Access to improved sanitation increased from 34 to 48 percent between 2009 and 2016. Poor households benefitted the most, as access more than doubled for this group, from 14 to 31 percent during the same period (Table 2.2 and Figure 3.3). Despite these advances, large gaps remain between urban and rural areas and between poor and non-poor households, while access levels are below the average in Upper-Middle-Income Countries. Low access to improved water and improved sanitation adversely affects the human capital formation of poor and rural households. BOX 3.1  Electrification Rate Maps 2012-2019 Satellite data is used to generate estimates of electrification at the highest resolution. The rates are estimated for 2012-2019 and validated against population estimates based on computer vision techniques identifying human settlements. This chapter utilizes the estimated predicted likelihood of electrification HREA data to generate the electrification maps for Botswana in 2012, 2015, and 2019. Since 2012, progress in electrification has been gradual and remains unequal amongst Botswana’s villages, with stronger growth in the northeast and southeast. To estimate electrification rates across Botswana beyond the latest available census (2011), satellite data from the High-Resolution Electricity Access (HREA) project was used. The analysis maps the HREA likelihood, calibrated to replicate the share of households with access to electricity in the latest available household survey (2015/16 BMTHS). Electrification rates are estimated for 2019, a year without available survey or census data, using HREA calibrated data and validated against population estimates based on computer vision techniques that identify human settlements. The results show a gradual increase in the predicted likelihood of electrification across villages between 2012 and 2019, with a more significant concentration around Gaborone and Francistown (Figure B3.1; map shows higher concentration of white circles around the main cities). FIGURE B3.1  Electrification rates are higher and have improved more in the northeast and FIGURE B3.1 southeast Electrified & Non-Electrified village in 2012 Electrified & Non-Electrified village in 2019 poverty poverty 35 35 30 30 25 25 20 20 15 15 10 10 Source: World Bank calculations using Min and O’Keeffe (2021) dataset for electricity access and 2009/10 BCWIS for poverty by district. Note: White circles represent electrified villages; yellow circles represent unelectrified villages. 34 BOTSWANA POVERTY ASSESSMENT BOX 3.1 (cont.) The electricity shortage of 2015 negatively impacted the poorest 40 percent of villages, which had estimated electrification rate declines in that year. While all poverty deciles have seen growth in estimated electrification rates between 2012 and 2019, there is significant inequality in power access growth across groups and intertemporally. Holding villages constant at their 2010 poverty level (or decile rank), Table B3.1 shows that between 2012 and 2015, the poorest 40 percent of villages saw declines in the proportion of their electrified villages. As discussed in Chapter 1, electricity shortages in 2015 led to the importation of nearly 40 percent of the country’s electricity demand at premium tariffs and without a guaranteed supply. This result suggests that the electricity shortages experienced in Botswana in 2015 had a stronger and more direct negative impact on the poorest households. It is important to note that the rising incidence of privately owned generators in less poor neighborhoods will not be captured by the satellite data employed. Nonetheless, the conclusion is still valid. Despite growth in the estimated electrification rate across all deciles in 2019, the gap between the poorest and wealthiest villages in terms of electrification remains large. Between 2015 and 2019, all ten deciles saw a rise in the electrification rate (Table B3.1 and Figure B3.2). Nevertheless, the gap in the electrification rate between the poorest and wealthiest villages remains extremely large. The richest ten percent of villages had, on average, an electrification rate of 70 percent in 2019, while the poorest ten percent had a rate of less than 12 percent. In addition, the increase in the electrification rate between 2012 and 2019 was the smallest among the poorest ten percent of villages, while the rate was 12 or 17 percentage points higher for some of the wealthier villages. FIGURE B3.2 TABLE B3.1  Electrification rate for 2012 - FIGURE B3.2  Electrification Rate by 2011 2019 by 2011 Village Poverty Deciles Village Poverty Deciles, 2012-2015-2019 80 Decile 2012 2015 2019 1 9.8 4.2 11.8 Electrification Rate (%) 60 2 6.1 6.3 14.3 3 9.4 8.0 20.8 40 4 12.5 8.9 20.8 5 10.0 14.0 16.0 20 6 26.5 27.7 38.8 7 23.1 29.2 40.4 0 8 40.0 43.8 52.0 1 2 3 4 5 6 7 8 9 10 Village Poverty Deciles 9 39.6 41.7 50.0 2012 2015 2019 Source: World Bank calculations using Min and O’Keeffe (2021) dataset for electricity access and 2010 Poverty Map to calculate the village poverty deciles. Note: Decile 1 = poorest 50 villages; decile 10 = wealthiest 50 villages. While Botswana has made some headway regarding health outcomes, significant challenges remain. By providing antenatal care and antiretroviral treatment to 9 out of every ten pregnant women, Botswana became the first high-burden country to bring mother-to-child transmission of HIV below 5 percent. The HIV case rate has dropped below 500 per 100,000 live births64, accompanied by an increase in life expectancy from 60.2 years in 2010 to 69.3 years in 2018. While maternal mortality declined from 188.9 per 100,000 live births in 2011 to 133.7 in 201865, it remains considerably higher than the average in other Upper-Middle-Income countries (57 64  See this report https://www.unaids.org/en/resources/presscentre/pressreleaseandstatementarchive/2021/december/emtct_bo- tswana 65  Statistics Botswana (2020) Key Statistics @ http://www.statsbots.org.bw/ BOTSWANA POVERTY ASSESSMENT 35 maternal deaths per 100,000 births) and not commensurate to the level of health sector investment.66 Similarly, Botswana continues to experience rates of malnutrition like other countries in Southern Africa that have much lower income levels. Infant and under-five mortality also declined between 2015 and 2021, from 36.6 to 28.3 deaths per 1,000 live births and 45.4 to 34.9 deaths per 1,000 live births, respectively.67 Nonetheless, infant mortality is slightly higher than 28.4 deaths registered globally and three times higher relative to Upper-Middle- Income countries.68 In contrast, the neonatal mortality rate and levels of stunting among children under five remained relatively unchanged. The neonatal mortality rate decreased from 23.3 to 21.9 per 1,000 live births between 2015 and 2020, while levels of stunting stagnated at 22 percent in the past two decades69. Botswana continues to register one of the highest incidences of tuberculosis globally and is facing a considerable burden from non-communicable diseases. Estimated death rates from cancers and cardiovascular disease have remained higher than 100 and 250 per 100,000 population, respectively. Considerable progress has been made in increasing school enrollment at younger ages; however, the quality remains low. Education Management Information Systems (EMIS) show the share of children aged 5 enrolled in pre-primary schools increased from 20 percent in 2013 to 43 percent in 2018. Household survey data from 2015/16 indicates that less than a third of children between the ages of three and five were enrolled in community-based preschool centers by private providers. Nonetheless, at the end of junior secondary, dropout rates remain high, the senior secondary gross enrollment rate is only 62 percent, there is a shortage of classrooms and learning materials, and teachers receive very limited in-service training at the secondary level. Moreover, Botswana’s expected years of schooling, adjusting for the quality of education, are low for its income level and below its structural peers (Figure 3.4).70 Public spending on education is high, representing over a fifth of the government budget (22.3 percent) and 8.4 percent of GDP in 2020 (among the highest in the world). A large share is earmarked for tertiary education, primarily benefiting students from the top two income quintiles. Technical or vocational (TVET) per-student spending is lower than secondary education per- student spending. Thus, school dropouts go to underfunded, bad-quality TVET institutions, if at all. Only 20 percent of poor households have household heads with secondary or higher education compared to 52 percent of nonpoor households (Table 2.2). As discussed in Chapter 2, poorer households have lower levels of human capital accumulation, FIGURE 3.4 limiting the productive capacity of the workforce. FIGURE 3.4  Expected Years of Schooling 16 14 Years of Schooling 12 10 8 6 4 2 0 a a ia lia p. on a a ia ca e e e m m m an ric si gi ib on Re go Ri n ni or co co co m w Af ba ed Tu on a Ge o, ts in in in Na st Le ac h ng M Bo w e gh ut Co M dl Co Lo So Hi id h rt m No er w Lo Source: Human Capital Project (2020). 66  Total current expenditure on health as a percentage of GDP was 6 percent in 2019 with total spending increasing in real terms by 8.6 percent between 2013/2014 and 2018/2019. World Development Indicators 67  World Bank. World Development Indicators. 68  Ibid. 69  World Bank from UNICEF/WHO/World Bank Joint Child Malnutrition Estimates - Country Level Models, April 2021, 70  Note: Expected years of school is calculated as the sum of age-specific enrollment rates between ages 4 and 17. Age-specific enroll- ment rates are approximated using school enrollment rates at different levels: pre-primary enrollment rates approximate the age-specific enrolment rates for 4- and 5-year-olds; the primary rate approximates for 6-11 year-old; the lower-secondary rate approximates for 12-14 year-olds; and the upper-secondary approximates for 15-17 year-olds. 36 BOTSWANA POVERTY ASSESSMENT FIGURE 3.5 FIGURE 3.6 FIGURE 3.5  Human Capital Index and FIGURE 3.6  Human Capital Index, select log GDP per capita, 2020 countries, 2020 0.9 0.8 Human Capital Index, 2020 0.8 0.7 0.7 0.6 Mongolia 0.6 0.5 Georgia Lebanon 0.4 0.5 Tunisia Gabon 0.3 0.4 Namibia South Africa Botswana 0.2 0.3 0.1 0.2 0 6.5 7.0 7.5 8.0 8.5 9.0 9.5 10.0 10.5 11.0 11.5 12.0 in e ng olia st nia M ibia Af a Tu on a e me e rt Ge sia ac ia Na ica ba . Le ep gh om ut an m ic M org n ni Co do w Lo a R r R id inco co Co ng m So tsw Hi inc o, e o Natural logarithm GDP per capita, 2020 (PPP, US$) h Bo w dl h m No er Lo Source: Human Capital Project (2020). Note: The Human Capital Index is designed to capture the amount of human capital a child born today could expect to attain by age 18. The HCI is higher on average in rich countries than poor countries and ranges from around 0.3 to around 0.9. The units of the HCI have the same interpretation as the components measured in terms of relative productivity. Despite recent advancements, Botswana’s Human Development outcomes remain poor for a country with similar income and characteristics. The Human Capital Index (HCI)71 slightly increased from 0.37 in 2010 to 0.41 in 2020, placing it just above the average HCI score for Sub-Saharan Africa (0.40) and considerably lower than the average of Upper-Middle-Income Countries (0.56) (Figure 3.5 and Figure 3.6). This indicates that a Botswana child is only 41 percent as productive as an adult as he could have been had he received complete education and full health.72 Moreover, children in Botswana face stark differences in life prospects depending on their circumstances at birth and during their early years, as shown via the results of the Human Opportunity Index (see Box 3.2). 73 BOX 3.2  The Human Opportunity Index The Human Opportunity Index (HOI) measures how individual circumstances such as place of residence, gender, and education of the household head can affect a child’s access to basic goods and services such as education, water, electricity, and sanitation.73 The index considers two concepts: the average coverage of a basic good or service and the inequality in the distribution of the good or service across circumstances. The HOI is based on children aged 14 and under to remove the effects of individual effort and choices and focus on opportunities essential to early development (Barros et al. 2009). Promoting more egalitarian access to basic goods and services of quality early in life will likely reduce inequality of outcomes in adulthood and increase economic efficiency. The HOI focuses on the extensive margin - whether or not there is access to a service. Nonetheless, the quality of an ‘opportunity’ could be heterogeneous across or within regions. 71  The Human Capital Index (HCI) measures the expected productivity as a future worker of a child born today. It is a function of educa- tion and health, underscoring their importance for the productivity of people. It ranges between 0 and 1, where 1 indicates the benchmark of complete education and full health. 72  World Bank. World Development Indicators. 73  The index is expressed on a scale of 1 to 100, with higher figures reflecting good levels of equity and lower figures reflecting poor and/ or inequitable access. See Paes de Barros et al. (2008) and Molinas et al. (2012). BOTSWANA POVERTY ASSESSMENT 37 BOX 3.2 (cont.) Access to basic goods and services is highly relevant to the development of children, yet remains largely unequal between urban and rural areas and far from universal in some cases. Poor households often lack access to adequate water and sanitation, electricity, and education, limiting their ability to participate in and contribute to growth. Despite progress in providing access to electricity, as of 2016, the HOI value for improved sanitation (40.6) remains low. This is of particular concern given that water and sanitation influence health and other important childhood opportunities, such as not missing school days due to preventable illnesses. Nonetheless, Botswana has near-universal coverage and HOI values for school enrollment, with slight increases between 2009 and 2016 (Figure B3.3). In addition, there are substantial differences between urban and rural areas regarding access to basic services. For all the opportunities analyzed, urban coverage is more than double that of rural areas, with access to improved sanitation showing the largest gap (Figure B3.4). An increase in available opportunities primarily drives improvements in the HOI. Changes in the HOI can be decomposed into three components: (i) the composition effect, which refers to changes in the composition of the group analyzed (for example, an increase in parent’s education or rural-to-urban migration); (ii) the scale effect, which represents a change in the available number of opportunities (for example, more electricity connection spots/towers); and (iii) the equalization effect, which measures the shift towards a more equal distribution of the opportunity under consideration. In 2003-16, the scale effect explained most HOI changes, particularly for electricity (Figure B3.5). Greater electricity coverage allowed more households to connect to the grid. In contrast, changes in sewage and water in 2009-16 were primarily driven by the composition effect. Household per capita income, dwelling location, and parental education are the most important factors determining whether a child has access to essential childhood opportunities. The D-index provides insight into how an opportunity would have to be reallocated across children to ensure no association between access to services and their circumstances at birth. Approximately 62 percent of the differences in access to safe water and adequate sanitation relate to parent’s education or household per capita income. Likewise, opportunities to live in a home with access to water appear heavily correlated with place of residence (Figure B3.6). Rural-urban disparities in basic infrastructure, such as access to safe running water and improved sanitation, highlight a need to target quality infrastructure programs in rural areas. FIGURE B3.3 FIGURE B3.3  Human Opportunity Index, FIGURE B3.4  Human Opportunity Index, 2016 2016, by Area 100 100 80 80 60 60 40 40 20 20 0 0 2003 2009 2016 2003 2009 2016 2003 2009 2016 2003 2009 2016 2003 2009 2016 2003 2009 2016 2003 2009 2016 2003 2009 2016 2003 2009 2016 2003 2009 2016 2003 2009 2016 2003 2009 2016 electricity sanitation water school electricity sanitation water school electricity sanitation water school enroll enroll Rural Urban HOI Coverage HOI Coverage Source: World Bank calculations using HIES (2002/03), BCWIS (2009/10), and BMTHS (2015/16). 38 BOTSWANA POVERTY ASSESSMENT BOX 3.2 (cont.) FIGURE FIGURE B3.5  HOI Decomposition, B3.3 FIGURE B3.6  D-index Decomposition, 2003-2009-2016 2003-2009-2016 2003-2009 2009-2016 Electricity Sanitation Water 70 0.6 60 0.5 50 0.4 40 0.3 30 0.2 20 10 0.1 0 0 Electricity Sanitation Water Electricity Sanitation Water 2003 2009 2016 2003 2009 2016 2003 2009 2016 Equalization Scale Composition Location (urban) Income per capita Education head Number kids Gender head (male) Two parents in hhd Source: World Bank calculations using 2002/03 HIES, 2009/10 BCWIS, and 2015/16 BMTHS. Note: The circumstance groups used in this analysis are urban or rural residence, family per capita income, education of the household head, number of children 0-14 years old, gender of the household head, gender of the child, and presence of both parents in the household. When all possible groups have the same access, the D-Index is zero and the HOI is equal to the coverage rate. In the other extreme, if one group has full access while another has no access, the D-index equals one and the HOI is zero. 3.3  MULTIDIMENSIONAL POVERTY Multidimensional poverty in Botswana shows stronger improvements than monetary poverty. The country’s Multidimensional Poverty Measure (MPM) dropped nearly 11 percentage points, from 31.8 to 21.1 percent in 2009-16 (Figure 3.7). The World Bank’s MPM measures the share of households deprived in monetary poverty, education, and basic infrastructure services (Box 3.3). At 52 percent, access to sanitation remains the highest deprivation rate among basic infrastructure services, followed by electricity (35.5 percent) and access to water (3.7 percent). Even though electricity continues to have the second-highest deprivation rate, access to this service improved significantly since 2009, resulting in a 37 percent drop in the deprivation rate in 2016. Although starting from much lower deprivation rates, access to water and indicators for education also showed slight improvements in this period. Deprivation rates are highest in rural areas, with significant gaps between urban and rural areas. In 2016, at 64.8 percent, rural areas experienced more than three times the deprivation rates in access to electricity compared to urban areas (19.6 percent). Similarly, nearly two-thirds of rural households did not have access to improved sanitation compared to 45 percent of urban households. Disparities also persist in education variables; 16 percent of rural households have at least one adult with less than complete primary education, relative to less than 4 percent in urban households (Figure 3.8). In addition, only 5 percent of urban households have at least one school-aged child up to grade 8 not enrolled in school compared to 8 percent among rural households. BOTSWANA POVERTY ASSESSMENT 39 FIGURE 3.7 FIGURE 3.8 FIGURE 3.7  Multidimensional Poverty Measure, FIGURE 3.8  Deprivation in Access to Services by Deprivation by Dimension, 2009 and 2016 (%) Area, 2016 (%) 80 80 64.8 65 60 56 52 60 60 45 35.5 31.8 40 40 19.6 21.1 16.2 17.7 14.1 10.3 15 20 20 8.2 8.1 7.9 6.3 3.8 5.3 4.3 3.7 1.4 0 0 MPM Electricity Sanitation Water Adult Educ Child Educ Poverty Electricity Sanitation Water Adult Educ Child Educ 2009 2016 Rural Urban Source: World Bank calculations. Source: World Bank calculations. FIGURE 3.9 FIGURE 3.9  Individuals in households deprived in monetary poverty and overall MPM 45 42 40 35 30 28 25 22 21 20 15 10 2 8 5 5 35 16 15 21 2 2 0 Chile 1 Costa Rica 1 Estonia 1 Lebanon 1 Seychelles 1 Malaysia 0 Mauritius 0 Botswana Congo, Rep South Africa Namibia Georgia Gabon Mongolia Tunisia Monetary Poverty MPM Source: World Bank Global Monitoring Database (GMD). Multidimensional Poverty Measure, April 2023. https://www.worldbank.org/en/topic/poverty/brief/multidimensional-poverty-measure Cross-country comparisons show similar rates of multidimensional poverty in Namibia and South Africa but large gaps with other structural as well as aspirational peers (Figure 3.9). Botswana and Namibia had similar levels of extreme poverty circa 2015 under the international poverty line of $2.15 per day (2017 PPP). However, Namibia had a higher multidimensional poverty measure (at 28 percent versus 21 percent for Botswana) due to worse electricity, water, and sanitation deprivation rates, while similar education rates. Botswana and South Africa had similar levels of MPM circa 2015 because better results across all non- monetary dimensions compensated for South Africa’s higher monetary poverty rate.74 Comparisons with other structural peers, such as Georgia, Gabon, Mongolia, Lebanon, and Tunisia, or some possible aspirational peers, such as Chile, Estonia, Seychelles, and Mauritius, show large gaps with Botswana’s non-monetary dimensions of poverty. The main drivers of Botswana’s high MPM are the high levels of monetary poverty and the high deprivation rates in access to sanitation and electricity. It is harder to achieve higher living standards when poverty in all its forms is considered, but the MPM provides a way for policymakers to monitor improvements in this broader concept of welfare. 74  Recent concerns regarding access to electricity in South Africa could result in a worse MPM. 40 BOTSWANA POVERTY ASSESSMENT BOX 3.3  Multidimensional Poverty Measure The World Bank’s Multidimensional Poverty Measure (MPM) seeks to understand poverty beyond monetary deprivations by including dimensions such as access to education and basic infrastructure. However, maintaining monetary poverty as one of the dimensions (in the form of the monetary headcount rate at the $2.15 international poverty line) also allows for a broader definition of poverty. Including monetary and nonmonetary dimensions differentiates the WB’s MPM from other prominent global multidimensional measures (such as the Multidimensional Poverty Index developed by the United Nations Development Programme and Oxford University). A country’s MPM is at least as high as or higher than monetary poverty, reflecting the additional role of nonmonetary dimensions to poverty and their importance to general well- being. A focus on non-monetary deprivations highlights to policymakers the importance of improving other aspects of human welfare that may not be well-captured by monetary measures alone. For example, some rural households may be considered nonpoor in income but lack access to water, electricity, or education. Alternatively, households that are both income-poor and deprived in nonmonetary dimensions are much worse off in terms of well-being than households that are income-poor but have access to basic services. Considering multidimensional poverty can provide a means for monitoring improvements in broader welfare. The MPM comprises six indicators mapped into three dimensions of well-being: monetary standard of living, education, and basic infrastructure services (Table B3.2). These six standardized indicators include consumption- or income-based poverty, educational enrollment, educational attainment, and access to drinking water, sanitation, and electricity. Each is defined as a 0-1 variable, where “1” means the individual or household is deprived in that indicator. The MPM summarizes the number of deprivations into a single index, which requires a decision on how to weight each indicator. In the World Bank’s MPM, dimensions and indicators within each dimension are weighted equally. Individuals are considered multidimensionally deprived if they fall short in at least one dimension or a combination of indicators equal in weight to a full dimension. In other words, if a household faces deprivations in indicators whose weight adds up to 1/3 or more, it is considered poor. Since the monetary dimension has only one indicator and there are three equally weighted dimensions, anyone who is income-poor is also poor under the broader multidimensional poverty concept. In addition to selecting the dimensions and the indicators, one must also specify the deprivation parameters or thresholds for each indicator. As an example, the threshold chosen for the educational enrollment indicator is that at least one school-age child up to the age of grade 8 is not enrolled in school. Table B3.2 presents the detailed indicators, weights, and thresholds for all dimensions and indicators that comprise the MPM. TABLE B3.2  Multidimensional Poverty Measure Indicators, Weights, and Thresholds Dimension Parameter Weight Monetary Daily consumption or income is less than US$ 2.15 per person 1/3 At least one school-age child up to the age of 8 grade is not enrolled in school 1/6 Education No adult in the household (age of grade 9 or above) has completed 1/6 Access The household lacks access to limited-standard drinking water 1/9 to basic The household lacks access to limited-standard sanitation 1/9 infrastructure The household lacks access to electricity 1/9 Source: World Bank. 2020b. Poverty and Shared Prosperity 2020: Reversals of Fortune. https://openknowledge.worldbank.org/handle/10986/34496 BOTSWANA POVERTY ASSESSMENT 41 3.4  POOREST LOCATIONS FACE HIGHER RISK FROM NATURAL HAZARDS Droughts are the second most common natural disaster in Botswana, particularly in the southern and eastern regions. Approximately 40 percent of the total population experiences low levels of effective precipitation, and under future projected climate conditions, the share of the population affected by droughts is expected to increase. The losses disproportionately affect the agricultural sector, as 40 percent of the country’s livestock and all crops (except for maize and sorghum) are affected by droughts.75 The poorest villages are the most negatively affected by droughts and flooding. In 2016, almost 28 percent of Botswana’s rural labor force worked in the agricultural sector, a sector whose productivity is sensitive to changes in climatic conditions, especially due to Botswana’s challenges with water scarcity. In 2015 and 2019, Botswana suffered severe droughts, as seen via the strong negative values of the Palmer Drought Severity Index in Figure 3.10, with another but less severe drought impacting villages in 2020.76 On the other extreme, Botswana seems to have faced very wet conditions in 2021, raising concerns about possible urban flooding, such as during the damaging floods of 2017. On the other hand, the PDSI was around zero during 2009/10, coinciding with a year of strong poverty reduction in Botswana. FIGURE 3.10  Palmer Severity Drought Index (PDSI) and Village 2011 Poverty Rate by Deciles, 2009-2021 PDSI 2009/10 (R2 = 0.02) PDSI 2015/16 (R2 = 0.03) PDSI 2019 (R2 < 0.01) 5.0 Palmer Drought Severity Indices 2.5 0.0 −2.5 0.0 0.2 0.40 0.6 0.8 PDSI 2020 (R2 < 0.01) PDSI 2021 (R2 < 0.01) 5.0 Decile: Palmer Drought Severity Indices 1 2 3 2.0 4 5 6 0.0 7 8 9 −2.5 10 0.0 0.2 0.40 0.6 0.8 0.0 0.2 0.40 .6 0.8 Village HeadCount Rates Source: World Bank calculations using the PDSI created by the Gridded Surface Meteorological (gridMET) database. Abatzoglou (2013). Note: The PDSI ranges from –5 (extreme drought) to 5 (extremely wet). 75  “Syed, T.; Goodman, P.; Mataruka, Z.; Evans, D. E.; Pachena, H.; Canales Gomez, A. C. C. 2022. Roadmap for sustainable livestock value chains in Southern Africa. © World Bank.”. https://documents1.worldbank.org/curated/en/099430006082232802/pdf/P174621055c7850b80808d09c38c18d5ed9.pdf 76  The Palmer Severity Drought Index (PDSI) is created by the Gridded Surface Meteorological (gridMET) database (Abatzoglou, 2013) at 4km resolution to show driest (-5) to wettest areas (+5). 42 BOTSWANA POVERTY ASSESSMENT FIGURE 3.11 FIGURE 3.11  Palmer Severity Drought Index (PDSI) for 2015/16 by Village 2011 Poverty Rate 0.4 Decile: R² = 0.04 0.3 1 Annual rainfall shocks (2015/16) 2 0.2 3 0.1 4 5 0 Stronger negative 6 −0.1 rainfall shocks 7 −0.2 8 9 −0.3 10 −0.4 Poorest villages −0.5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Village poverty rates (2009/10) Source: World Bank calculations using the PDSI created by the Gridded Surface Meteorological (gridMET) database. Abatzoglou (2013). Note: The PDSI ranges from –5 (extreme drought) to 5 (extremely wet). Although drought hit most villages in 2015, it produced more damage among poorer villages where subsistence farmers work (Figure 3.11). The 2016 household survey shows that the number of people working in agriculture declined steeply relative to 2009, which, at a minimum, seems correlated with the drought conditions experienced in 2015/16. As shown in Chapter 2, poverty in rural areas increased in 2016 as subsistence and non-subsistence farmers could not make up for lost consumption via other sources of income. This points to the importance of efficient water management and irrigation policy, given the prevalence of subsistence (low-scale) agriculture in poorer areas. The Palmer Severity Drought Index shows similar drought conditions in 2019 as in 2015, which suggests that subsistence farmers may have again faced declines in welfare in 2019. Although there is no household survey with consumption data available to directly measure poverty that year, the poverty projections undertaken via survey-to-survey imputations show an increase in poverty in 2019 (Figure 2.1 and Appendix 9). Concerns with droughts go beyond the vulnerability of subsistence farmers, however; Botswana’s tourism depends on water-based wildlife, and it is considered the most important services sector export. Similar results are confirmed by an indicator that focuses on annualized rainfall shocks. Seasonal rainfall provides an important means of natural irrigation for farmers as it is a determinant of the quality of the subsequent harvest season. However, a significant proportion of the agricultural population (particularly in subsistence agriculture) live around the poverty line and are often one bad harvest away from falling into extreme poverty. This section uses standardized differences between mean annual rainfall at the village level in 2009/10, 2015/16, 2016/17, 2017/18, and 2018/19 to historical average annual rainfall between 1982 and 2019 to see how poor villages (defined based on the 2011 poverty map) might have been affected by the difference in annual expectations of rainfall (rainfall shocks). Similar to Figure 3.10, villages with higher poverty levels correlate with more negative rainfall shocks. In addition, these annual rainfall shocks appear stronger in the poorest deciles in both 2015 and 2019 (Figure 3.12, where decile 1 represents the poorest villages). These results suggest that the poorest villages with the most cropland cover suffer more severely from these downturns. BOTSWANA POVERTY ASSESSMENT 43 FIGURE 3.12 FIGURE 3.12  Average annualized rainfall shocks by village poverty decile, 2009-2019 2009/10 2015/16 2016/17 0.75 Annual Rainfall Shock 0.50 0.25 0.00 -0.25 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Village Poverty Rates 2017/18 2018/19 0.75 Annual Rainfall Shock 0.50 0.25 0.00 -0.25 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Village Poverty Rates Source: World Bank calculations using CHIRPS data (Funk et al., 2015) for rainfall shocks and 2011 Poverty Map for village poverty rates. Note: Village poverty rates are maintained constant at the 2011 Poverty Map rates and villages are grouped into deciles from poorest to richest. 44 BOTSWANA POVERTY ASSESSMENT CHAPTER 4 SOCIAL PROTECTION IN COVID TIMES: RESPONSES AND LESSONS LEARNED BOTSWANA POVERTY ASSESSMENT 45 4.1  THE PRE-COVID-19 SOCIAL PROTECTION SYSTEM 77 Social protection programs have a significant impact on poverty reduction. As discussed in Chapter 2, the 2016 poverty rates could be nearly five percentage points higher without these programs. Botswana’s primary school feeding and Old-Age Pension programs contribute the most to poverty reduction while having low expenditure. Social protection can also contribute to human capital outcomes as well as resilience and shock response. Productive economic inclusion programs could be used to support income generation and resilience. Botswana’s steps towards developing systems such as its social registry or stronger poverty targeting could help shock responsiveness. TABLE 4.1  Overview of social cash transfer programs in Botswana Program Launch Targeting Beneficiaries Benefit value Budget P530/month cash Old Age Universal, age 65 years P736 million 1996 126,424 (increased to P630 in Pension and older in 2019/20 2022 and P730 in 2023) Categorical, age 18 and 1999/ 20,146 in P600-P700 food basket, P246 million OVC Grant under, orphaned or in need 2009 2021/22 other in-kind benefits in 2019/20 of care Categorical, age 18 and P450/month cash Disability P35 million 2015 older, medical declaration 7,805 (increased to P550 in Allowance in 2019 of disability 2022) War Categorical, age 65 and P600/month cash P10 million Veteran’s 1998 older, fought in WWII, and 1,270 (increased to P700 in in 2019/20 Pension surviving dependents 2022) Destitute P500-650 in-kind, 1980/ P332 million Persons’ Means-tested 38,973 P300/month cash, 2002 in 2019/20 Allowance additional services Vulnerable Categorical, vulnerable P714 million Group 1988 children, women, sick and 307,225 Take-home food rations in 2017/18 Feeding destitute, no means test School Categorical, primary feeding 390,294 in 1997 school pupils in Cooked meal daily programme 2018/19 government schools (primary) School Categorical, secondary P308.5 feeding 183,896 in Two cooked meals daily 1997 school pupils in million in programme 2015/16 (+1 if boarding) government schools 2017/18 (secondary) Community P500 food basket/ Categorical, means-tested P15 million Home 1995 1,252 in P1,200 oral tube feeding, with same criteria as DPA in 2018/19 Based Care additional services Age 18 years and older, P567/month (casual Ipelegeng 1978/ 70,000/ P635 million able-bodied, ‘lottery labourers)/ P651/month (workfare) 2008 month in 2019 system’, limited to 1 month (supervisors) cash Source: Gronbach et al. (2023 forthcoming) Note: Beneficiary numbers for 2019/20, unless otherwise specified. The value of the Botswana Pula in relation to the dollar crashed in March/April 2020, from about $0.09 to about $0.08 (i.e., $1 bought about P11 before the COVID-19 lockdowns, then about P12.5). The Pula strengthened against the dollar in late 2020 and 2021 before depreciating in 2022. In October 2022, the Pula was worth US$ 0.074 (i.e., $1 bought about P13). 77  This chapter is based on Lena Gronbach, Jeremy Seekings and Winnie Arthur, ‘Lessons learned on social protection responses to Covid-19 in Botswana’. Forthcoming. CSSR Working Paper from the Centre for Social Science Research at the University of Cape Town. See also Gentilini et al. 2022: Social Protection and Jobs Responses to COVID-19: A Real-Time Review of Country Measures.https://open- knowledge.worldbank.org/entities/publication/fa7a2f3c-efbd-5950-bfac-4b2b4bfc8cad 46 BOTSWANA POVERTY ASSESSMENT The social protection system comprises three pillars: feeding schemes, workfare, and support for specific categories of deserving poor (the elderly, people living with disabilities, and ‘destitutes’). Table 4.1 provides an overview of the country’s social programs before the COVID-19 outbreak.78 The school feeding and the Old-Age Pension programs had the largest number of beneficiaries and the highest expenditure levels among the ten programs analyzed. Primary and secondary education school feeding programs catered to more than 390,000 and 183,000 students, respectively. Likewise, the Old-Age Pension program covered about 5.5 percent of the population (126,424). At 635 million Pula (USD 60 million in 2019), the Ipelegeng workfare program had the third largest budget in 2019. According to the Quarterly Multi-Topic Household of 2019 (quarter 3), approximately 57 percent of households benefitted from at least one of the country’s social protection programs. Moreover, coverage rates were higher in rural areas and among poorer households. The conservatism of Botswana’s system79 is evident in comparison with South Africa. More than half of the population lives in households that benefit from a feeding scheme, workfare, or social grants for the elderly, orphans, and ‘destitutes’. However, benefits are parsimonious or provided in kind, the means test for some cash transfer programs is severe, and none of the programs are statutory. The social protection systems in Botswana and South Africa exhibit similarities in their common emphasis on social assistance rather than social insurance and their shared inclusion of ‘residual’ elements (some programs are means-tested or otherwise targeted to the poor). Both countries retain extensive workfare and feeding schemes alongside cash transfer programs. Even though Botswana today has a higher GDP per capita than South Africa (having first caught up in the late 1990s) and coverage of social protection is extensive in both countries, benefits and total expenditure are much lower in Botswana. South Africa spent about 3.3 percent of GDP on social assistance prior to COVID-19, while Botswana spent less than 1 percent of GDP (excluding substantial expenditure on stipends for university students). Moreover, the Government of Botswana has resisted calls for a Child Support Grant along the lines of the South African program. 4.2  SOCIAL PROTECTION POLICY RESPONSES TO COVID-19 Botswana’s COVID-19 response for poor and vulnerable households was modest. The government did not supplement existing social grants or introduce new emergency social grant programs. Thus, during the 2020 lockdown and economic contraction, Botswana did not expand vertically or horizontally its social protection system even though its workfare program, Ipelegeng, was kept partially active. While the authorities did oversee an extensive once-off emergency feeding program, the benefits were offset at least partly by the suspension of school feeding schemes. Net additional expenditure on social protection in 2020 only amounted to 0.1 percent of GDP. In contrast, the government quickly reallocated funds and introduced wage subsidies for formally employed workers. The wage subsidy was given for three months (April through June 2020) and cost approximately 0.5 percent of GDP. There was reluctance to expand the country’s social protection system – except for food distribution – during the 2020 pandemic. Botswana’s social protection system evolved out of disaster relief during droughts and retained some flexibility to expand; even prior to COVID-19 there were strong calls for its expansion. In addition, Botswana has a well-established delivery system – both in cash and in-kind – to many citizens through various social programs. Moreover, the governing party has long emphasized compassion and concern for the poorest members of society. Even as the economic impact of the pandemic was evident within weeks80 and 78  Table 4.1 excludes tertiary education funding, job creation and agricultural support programmes and community-based early child- hood development schemes. 79  Botswana’s ‘conservative welfare state’ originated in drought relief programs between the 1960s and 1980s before being institutional- ized and expanded after 1990 (World Bank, 2022). 80  As early as 24th April 2020, the Minister of Finance and Economic Development predicted a 13 percent contraction in GDP and a 22 percent decline in government revenues. (Government of Botswana, Economic Briefing by Honorable Minister of Finance and Economic Development, 24 April 2020, cited in the Terms of Reference for the National Social Protection Recovery Plan, p.53.) BOTSWANA POVERTY ASSESSMENT 47 some reforms were proposed in a National Social Protection Recovery Plan as early as July 2020, the COVID-19 crisis was not accompanied by the anticipated social protection response. In contrast, Botswana’s neighbors provided various options for COVID-19 responses, including the broad and expensive package in South Africa, once-off cash distribution in Namibia and Lesotho, and emergency programs targeted at the urban poor in Zambia. In Botswana, as elsewhere, social protection measures were implemented after a lockdown had been imposed. The government’s subsequent social protection reforms were shaped by both the character of the lockdown and the country’s existing social protection system. The lockdown not only suspended school feeding programs but also limited access to markets or farms to purchase food. The food basket program In mid-April 2020, the country organized a massive food relief operation to mitigate the impacts of the COVID-19 lockdown. Household needs were assessed based on the 2002 National Policy on Destitute Persons.81 District Councils mobilized to assist Social Welfare officers in procuring and distributing food supplies, and unemployed social work graduates were hired temporarily to assist with the household assessments. By the end of April, over 300,000 households had been assessed, and 244,000 had been recommended for assistance. The actual distribution of food baskets depended on procuring the required supplies and establishing distribution mechanisms. At the time, less than one in five households (47,000 of 244,000) identified needing food baskets had received them.82 Overall, the assistance brought temporary relief, as nearly 80 percent of more than 500 thousand assessed households (approximately 430,000 households by one report) received a food basket in 2020, representing more than P400 million in additional government expenditure.83 As lockdown restrictions eased in early May, the government ceased plans for a second round of food baskets and completed the only round by the end of May. Despite the delay in distributing food baskets, the speed and reach of food distribution in Botswana were impressive, especially compared to neighboring South Africa.84 Botswana was also much quicker than South Africa to reopen schools and resume school (and clinic) feeding schemes. The monetary benefit of the once-off food basket program per household was comparable to the once-off Emergency Income Grant in Namibia and substantially higher than benefits under emergency cash transfer programs in Zambia and most other African countries. However, it was much lower than the value of vertical and horizontal extensions of social grants in South Africa or the donor-funded emergency program in Kenya. Wage subsidies As in many African countries during the 2020 pandemic, Botswana implemented wage subsidies for formal workers.85 Eligible employers could claim 50 percent of the wages of eligible employees, with a minimum subsidy of P1,000 per month and a maximum of P2,500 per month. Qualifying businesses had to apply on 81  The standard questionnaire recorded the number of people in the household, education level, employment status, among others. 82  Masisi, M. E. K. (2020b). State of the Nation Address by His Excellency Dr. Mokgweetsi E.K. Masisi, President of the Republic of Botswana, to the First meeting of the Second Session of the Twelfth Parliament, 09 November 2020. Gaborone: Government of Botswana. Retrieved from https://vision2036.org.bw/sites/default/files/resources/SONA%202020.pdf (20.04.2022). 83  According to one official report, a total of 537,466 households had been assessed and 429,255 households had received assistance by end of May. https://www.bocra.org.bw/sites/default/files/covid19-docs/NEOC%20BULLETIN%20ISSUE%2076.pdf. However other official reports apparently indicated different numbers. 84  See: Rodolfo Beazley, Jana Bischler and Alexandra Doyle (2021) ‘Towards shock-responsive social protection: Lessons from the COVID-19 response in six countries’, Towards Shock-Responsive Social Protection, Oxford Policy Management, Oxford; and Beazley, M. Marzi, and R. Steller (2021) ‘Drivers of Timely and Large-Scale Cash Responses to COVID-19: what does the data say?’, Social Protection Approaches to COVID-19 Expert Advice Service (SPACE), DAI Global UK Ltd, United Kingdom. 85  Similar wage subsidy schemes for formal sector workers were implemented in Benin, Cape Verde, eSwatini, Gabon, Lesotho, Mauritius, Namibia, Seychelles and South Africa (International Policy Centre for Inclusive Growth, 2021). 48 BOTSWANA POVERTY ASSESSMENT behalf of their employees through the Botswana United Revenue Service (BURS). Nonetheless, the subsidy focused on formal workers, so informal workers were excluded. The subsidy was set to run for three months (April – June 2020) and was later extended for businesses and employees working in the tourism sector. By the end of April 2020, 12,440 companies had applied for the wage subsidy, 12,413 of which had been approved. This covered 165,681 employees in April and amounted to payments totaling P233.7 million.86 Overall, P833 million were paid between April and June 2020.87 Informal Sector Stimulus Fund The lockdown caused considerable financial hardship for households that depended on informal sector livelihoods (such as street trading) that were prohibited at the time. While most of the funding for sectors hit by the lockdown and recession was reserved for formal sector enterprises, informal traders and Small, Medium, and Micro-sized Enterprise (SMME) owners could apply for P1,000 payments through a P100 million Informal Sector Stimulus Fund (ISSF), administered by the Local Enterprise Authority (LEA). The program was launched in May 2020. By June, LEA had received 30,929 applications, almost all approved and paid out. In the face of slow take-up, the program was extended but only reached 47,000 beneficiaries, significantly lower than the wage subsidy scheme for formal workers. Workfare Along with lockdown restrictions, the Ipelegeng workfare program was partially suspended. Workfare had long been a cornerstone of Botswana’s drought relief and social protection strategies. Ipelegeng workers were paid until the end of April, including non-workdays due to the lockdown. A small number of people were employed in special programs. For instance, in the second half of 2020, about 20,000 people were hired to clean schools. Between the end of April and September 2020, the Ipelegeng program operated at a small fraction of its usual level of 70,000 people per month. Nonetheless, by September 2020, the Ipelegeng program resumed fully. The government’s resistance to expand the Ipelegeng workfare program while the economy was in recession in 2020 was likely rooted in its pre-COVID ambivalence over the program. The program had been criticized within and outside the government for its lack of developmental impact: Ipelegeng participants were widely viewed as unproductive and did not appear to acquire any significant skills but became ‘dependent’ on the program. In July 2020, the Ministry of Local Government and Rural Development announced it was reviewing and reforming Ipelegeng ‘with a view to making it more productive and worthy of the budget it attracts on a yearly basis’.88 The cabinet approved the proposed reforms in August 2020. The pilot reform would entail engaging workers for an extended period to work with trained artisans to construct housing for destitutes. 4.3  ASSESSMENT OF RESPONSES In sum, while the scale and cost of COVID-19 emergency response programs were low relative to neighboring countries, measures were implemented faster. Unlike its neighbors, Botswana did not introduce a special social cash transfer for vulnerable individuals or households.89 As previously discussed, the government did roll out, reasonably quickly, a massive food parcel operation, provided modest support for a small number 86  Relief Fund to be audited: Matsheka. Retrieved from https://twitter.com/bwgovernment/status/1253950300018413570 (09.04.2022). 87  https://www.sundaystandard.info/covid-wage-subsidy-was-costly-political-mess/ 88  Ipelegeng Temporarily Suspended due to Covid-19 Compliance - Hon. Molale. Presidential (Covid-19) Task Force Bulletin, Issue 75. Gaborone: The National Emergency Operation Centre. 89  The South African Government paid supplements to the recipients of the 18 million social grants that it paid every month (i.e. it extended its social grant system ‘vertically’) and, after a short delay, introduced a very modest emergency grant (the COVID-19 Social Relief of Distress Grant) that soon reached 5 million more people (i.e. the Government extended the social grant system ‘horizontally’). In Namibia, a once-off cash transfer was quickly implemented. BOTSWANA POVERTY ASSESSMENT 49 of informal businesses, and provided massive wage subsidies to formal workers, benefitting primarily non-poor urban households (Table 4.2). Nonetheless, the existing workfare and feeding schemes were interrupted, and the government did not consider supplements or expansion of its existing, modest social grant programs. The net cost of emergency programs amounted to a little over 0.1 percent of GDP, ten times lower than the assigned budget in South Africa (1 percent of GDP). In Botswana, existing social assistance programs largely continued using their established payment infrastructure, both during and after the initial lockdowns, with only relatively minor modifications. The use of mobile payment technologies for cash transfers in response to COVID-19 has been a defining feature of the pandemic response across the African continent. Similarly, digital and mobile application and registration tools were used for various cash transfers and other social protection responses, in many cases for the first time. The trend was most pronounced in new cash transfers, explicitly launched in response to the pandemic. As Botswana did not implement new emergency programs, the pandemic did not trigger significant reforms in its payment system, and pre-existing cash transfer programs continued to use their pre-COVID TABLE 4.2  Overview of Botswana’s social protection responses to COVID-19 Timeliness Benefit Value Reach Duration Cost Generous Disbursements BWP 833m for low-paid In April, 165,000 3 months approved in (first 3 months) workers; employees (except for late April, + BWP 143m Wage subsidies averaging (potentially tourism sector, presumably (tourism, further BWP 1,500/ more in May/ further 6 effected quickly 6 months) = employee/ June) months) thereafter BWP 976m month BWP 1,500 or Sectoral Most paid in BWP 2,500 per 10,000, less 3 months BWP 20m subsidies June month for 3 than anticipated months Pay-outs started in June, Informal sector BWP 1,000 47,000, less continued Once-off BWP 50m support once-off than anticipated throughout 2020 Some distributed in late April, Once-off; plans BWP 350-431m Baskets worth Widespread, most in May, for further (but savings Food baskets approximately approximately some in June distribution on other BWP 1,000 430,000 (but existing dropped programmes) programmes suspended) [Widespread, Largely approximately Net savings Existing feeding suspended under Not applicable 900,000 Not applicable (estimated at schemes lockdown beneficiaries, BWP 100m) suspended] [About 73,000 Benefits paid workfare in April but Net savings opportunities Workfare programme then Not applicable Not applicable (estimated at per month, largely suspended BWP 100m) suspended for 4 to September months] Cash transfers No reform Not applicable Not applicable Not applicable None Source: Gronbach et al. (2023 forthcoming). Note: Negative responses in italics. 50 BOTSWANA POVERTY ASSESSMENT payment channels. Nonetheless, the pandemic accelerated the use of e-wallets for Ipelegeng and the Local Enterprise Authority informal enterprise subsidy, which was initiated prior to the 2020 crisis and expanded in 2020. However, COVID-19 may have encouraged a shift in thinking toward digital payment channels, as well as shifting from providing benefits in kind to cash-based programs and the consolidation of payment systems. 4.4  COVID-19 AND LONGER-TERM REFORMS Despite its shock-responsive origins in drought relief, Botswana’s social protection system proved surprisingly inflexible in response to the pandemic and ensuing economic hardship in 2020-21. Programs designed originally to respond to drought proved ill-suited to the conditions that resulted from a severe national ‘lockdown’ in April and May 2020. Faced with severe hardship under the COVID-19 lockdown, the state distributed food parcels widely, after an impressively short delay, and offered substantial wage subsidies. While wage subsidies were of little benefit to poor Batswana, the distribution of food parcels was a considerable achievement, helping to mitigate hardship in late April and May. However, the suspension of other feeding schemes and workfare and the reluctance to consider any expansion of cash-based programs meant that the overall response was modest. The absence of any cash-based response also meant that COVID-19 had little effect on consolidating and digitizing payment systems. Evidence suggests the 2020 pandemic had little effect on the Government of Botswana’s approach to social protection. The emergency provided an opportunity for international organizations to renew their advocacy of expansionary reforms – including targeted child grants. Nevertheless, the Government of Botswana showed interest only in those reforms that accorded with its existing approach, which focused on very bespoke programs implemented by social workers at the local level. The Government’s response to COVID-19 was not to expand social protection in new directions but rather to ease the lockdown to return quickly to ‘normality’. The authorities emphasized its existing preference for discretionary social protection rather than to introduce large-scale cash transfer programs. The Government of Botswana’s more cautious response to the shock of COVID-19 may reflect the limits of political pressure for change and the goal of self-reliance. The response mirrored the enduring and widespread commitment within Botswana to the norms and values underpinning Botswana’s social protection system. These rested, above all, on the idea that assistance from the state should be linked, wherever possible, to the goal of people achieving self-reliance through productive work and thereby being able to fulfill their responsibilities to the wider society. Reform proposals that got traction within Botswana were generally those that moved social protection in a more developmental direction. In the future, this may include linking social assistance to productive economic inclusion approaches, investment in human capital, and support for labor market activation. BOTSWANA POVERTY ASSESSMENT 51 CHAPTER 5 POLICY CONSIDERATIONS FOR POVERTY AND INEQUALITY REDUCTION 52 BOTSWANA POVERTY ASSESSMENT Renewed policy efforts will be needed for Botswana to reach its goals of poverty eradication and high- income status. Several policy implications emerge from the analysis presented throughout the report and from sectoral work recently produced by the World Bank and/or recently included in the Botswana SCD Update and Country Economic Memorandum (CEM).90 The suggested policy considerations are summarized around four areas. First, accelerating inclusive economic growth based on dynamic private sector-led job creation is crucial to increasing the income of poor people. Second, further investments in quality human capital among the poor are essential to improve welfare and boost workforce productivity. Third, additional investments in infrastructure and shock-responsive systems, especially in rural areas, are needed to better connect and protect the most vulnerable population. Fourth, strengthening data is vital for better evidence-based policy design that improves outcomes for all, particularly the poor. Accelerating inclusive economic growth with private-sector-led job creation Stronger poverty and inequality reduction requires a more diversified and inclusive growth model that supports private-sector job creation and is more resilient to shocks. The shocks that hit the economy since the last Poverty Assessment, as explained in Chapter 1, highlighted the country’s fiscal vulnerabilities and weak economic diversification. More concerning, Chapter 2 showed that the ability of economic growth to lead to poverty reduction (i.e., the elasticity of poverty to GDP growth) declined in 2016. Per the SCD Update, Botswana needs to “develop a competitive, export-oriented private sector, leveraging regional integration” that maximizes economic inclusion. The CEM then proposes three guiding criteria, based on economic theory and empirical evidence, that could help policymakers more systematically select priority sectors. The sectors should be labor–intensive and tradable as they are most likely to create productive jobs for more people and foster innovation; globally competitive given Botswana’s small and undiversified economy; and strategic towards high expected regional and global demand to increase market size, such as Botswana’s potential in renewable energy, base minerals, and eco-tourism. Finally, promoting competition and expanding formal and informal employment opportunities will require a reduction in the excessive public sector footprint that creates barriers to entry, discourages diversification, and causes large economic inefficiencies. It will require facilitating external trade in goods and services and implementing adequate foreign investment rules to improve competitiveness and generate good jobs. As structural transformation takes time, it is important to boost the productivity of farm and non-farm household enterprises, particularly in rural areas, and create jobs for the large unskilled population. In 2022, 9.8 percent of workers were farm self-employed, and 16.4 percent were non-farm self-employed, hardly unchanged over time. Farm policies that could boost productivity include investments in agricultural research and extension, irrigation (given the high vulnerability to droughts), and rural infrastructure, as well as efforts to bring poor farmers into the value chain more effectively. Livestock-related services (including from regional collaboration) need to reach the small-scale sector and poor subsistence farmers so that they can also benefit from Botswana’s niche value chains in beef and livestock.91 For non-farm household enterprises, policies to boost productivity include increasing access to credit (and building skills; see below). The Country Private Sector Diagnostic (IFC 2022) estimates the financing gap of small enterprises to be 19 percent of GDP. Limited access to credit limits the ability of the sector to create employment. Since 2009, job creation has not matched the growing labor force; instead, chapter 1 shows Botswana ranks in the bottom 30 percent of countries in terms of employment and has the second-highest unemployment rate among upper-middle-income countries. 90  World Bank (2023) and World Bank (2024a, forthcoming). 91  For a discussion of sustainable livestock value chains, see Syed and others (2022). World Bank (2023) also highlights that Botswana’s livestock smallholders, lacking the scale to access large markets, could benefit from collaboration with Namibia on veterinary services, traceability, transboundary animal disease surveillance and control, and research and development. BOTSWANA POVERTY ASSESSMENT 53 Improving human capital particularly for the poorest Improving the welfare of the poor and reinvigorating the growth model require substantial improvements in human capital to increase labor force productivity. As mentioned in Chapter 3, despite Botswana’s considerable progress and despite investing significant resources, its human development outcomes remain well below the levels expected for a country of its income and characteristics. The government needs to redouble efforts toward more efficient investments in health and education to enhance the productivity of the current and next generation, particularly the poor.92 The quality of education services is an important constraint, requiring policies addressing the coordination of service delivery among several Ministries; the shortage of classrooms and learning materials; giving teachers adequate classroom training and coaching; and fine-tuning education assessment systems to track progress and facilitate corrective action (see SCD Update, World Bank 2023). The allocation of resources across education and training subsectors needs to be more pro-poor since the poor learn less and are more likely to drop out of education and training. As the data shows, Botswana’s high degree of childhood inequality of opportunity constrains the upward mobility of the poor.93 Only 20 percent of poor households have more than primary education compared to 52 percent of non-poor households. The country’s low growth and high inequality require that fiscal resources are used efficiently and reallocated to high social value-added uses, such as policies that improve early life outcomes (World Bank 2022a). Significant challenges remain also in health outcomes among the poor, with both maternal and infant mortality rates much higher than the average for upper-middle-income countries. Policies could focus on strengthening the quality of health service delivery, including improving clinical guidelines and protocols, enhancing staff core competencies, and improving the availability of essential medicines (World Bank 2023).94 Strengthening skills training and qualifications, and better coordinating and monitoring employment programs, are also important for boosting the productivity of poor and vulnerable workers. Chapter 2 shows that differences in occupation types in the labor market (reflecting skill and ability differences) were the primary contributors to inequality in 2016, while poverty levels were much higher for Batswana with low education. However, skills development through technical and vocational programs or on-the-job training faces many limitations. The many existing technical programs and trainings for skills development are fragmented and uncoordinated and lack systematic evaluations of interventions. To better support youth transitions into productive employment (formal or self-employment) also requires stronger coordination of programs. This coordination requires significant collaboration across ministries and better partnerships with the private sector. In addition, the focus needs to be on skills with a high growth potential, such as digital and green skills. Even in urban areas, a lack of job creation and skills mismatches hamper the ability of disadvantaged households to generate income, and high wage inequality fuels overall inequality. Investing in infrastructure and shock-responsive systems to connect and protect Increasing access to quality infrastructure and basic services increases the productive capacity of the most vulnerable population. Access to basic services has increased significantly over the last decades, but some services remain below the average for upper-middle-income countries. Large gaps remain in access to electricity and sanitation between rural and urban areas and between poor and non-poor households. The 92  Fiscal policy analysis via the Commitment to Equity (CEQ ) methodology showed education and health transfers contributing strongly to inequality reduction in 2009/10 (World Bank, 2022; Lusting and Higgins, 2016). 93  See Human Opportunity Index (chapter 3) while Sulla et al. (2021) shows that inherited circumstances account for 20 percent of Bo- tswana’s inequality. In addition, the forthcoming Africa Poverty and Inequality Report highlights intergenerational mobility in education for African countries is well below that of the developing world on average. 94  A forthcoming Public Expenditure Review in Health will dive more deeply into policies to strengthen health outcomes. 54 BOTSWANA POVERTY ASSESSMENT Human Opportunity Index shows that large gaps also remain between rural and urban children, with access far from universal. Worse, progress has slowed among poor households relative to nonpoor households. Geospatial data suggests that the electricity shortage of 2015 had particularly negative impacts on the poorest 40 percent of villages. Where access is high, such as for water, the quality of services is the important constraint. Promoting the inclusion of the rural population requires making connectivity (electricity, digital) affordable and reliable. To reach its goal of universal access to electricity, Botswana will need to mainstream off-grid solutions to deal with the deep spatial inequalities. Workers need support to develop new skills to reap the benefits of digitization. Investments to improve water service quality and access to sanitation are also priorities, particularly in rural areas and as urbanization continues. Per the SCD Update: “The interlinked challenges of water scarcity, limited wastewater treatment, the underprovision of sanitation services, and malnutrition continue to undermine people’s health, limit the development of their skills, and hinder their productive participation in the labor force.” Investments in the capacity of social protection systems to respond to shocks while improving the targeting of safety net programs protect the poorest. Although Botswana responded better on some dimensions to the COVID-19 pandemic than many other African countries, the pandemic exposed significant gaps in its social protection system’s shock responsiveness and resilience. Botswana needs to develop productive and shock-responsive social protection systems to mitigate shocks (including droughts and pandemics) and disaster vulnerability. Rural poverty increased in 2016 during a drought and a recession as the country faced water and electricity crises. Despite Botswana’s social protection programs’ significant impact on reducing poverty in 2016, they were insufficient to assist many poor rural households. Uninsured risks can trap families in a poverty cycle. Mitigating fragility helps build household resilience to avoid families falling back into poverty. The social protection delivery system also needs to identify poor and vulnerable households. Since the pandemic, Botswana has embarked on reforms to strengthen the administration of social assistance, including developing a unified social registry, but it will need to broaden and deepen the reforms.95 A fully operational social registry would significantly improve the targeting of human capital investments. In addition, the government also took important steps towards developing and pilot testing a proxy means test for improved eligibility determination for poverty-focused programs. The government should continue its plan to roll out this tool for targeting social assistance beneficiaries, starting with the Destitute Persons Programs. Strengthening data for evidence-based policy design Improvements in statistical data collection and use, data infrastructure, and monitoring and evaluation systems are important for transparency, accountability, and more robust evidence-based policy design. Tracking progress on poverty reduction requires frequently collecting reliable quality data on human capital, livelihoods, and welfare. The last available income and expenditure survey was collected in 2015/16. Fortunately, a new survey is expected to go into the field this year (2024/25), but a commitment to reduce these data gaps is needed. More broadly, investing in data is required to make informed public policy decisions and to track progress (or lack thereof), such as for malnutrition trends. The development and implementation of Botswana’s National Development Plans require timely, reliable, high-quality data across sectors, the modernization of the National Statistical System, the integration of data systems, and strengthened monitoring and evaluation systems. 95  The National Social Protection Framework (NSPF) approved by the Cabinet in 2020 aims to establish a comprehensive and well-co- ordinated social protection system. See Government of Botswana, Economic Briefing by Honorable Minister of Finance and Economic Development, 24 April 2020, cited in the TOR for the NSP Recovery Plan, p.53 BOTSWANA POVERTY ASSESSMENT 55 References Abatzoglou, J. T. 2013. “Development of gridded surface International Monetary Fund (2018). Botswana 2018 meteorological data for ecological applications and Article IV Consultation. IMF Country Report No. modelling.” International Journal of Climatology, 18/268. https://commitmentoequity.org/wp-content/ 33: 121–131. rmets.onlinelibrary.wiley.com/doi/ uploads/2020/05/cr18268.pdf full/10.1002/joc.3413 Lanjouw, Jean Olson & Lanjouw, Peter, 1997. “Poverty Amendola, Nicola and Giovanni Vecchi. 2014. “Durable comparisons with non-compatible data: theory and goods and poverty measurement.” No 7105, Policy illustrations,” Policy Research Working Paper Series Research Working Paper Series, The World Bank. 1709, The World Bank. Beazley, M. Marzi, and R. Steller (2021) ‘Drivers of Timely Lerman, Robert and Shlomo Yitzhaki. 1985. “Income and Large-Scale Cash Responses to COVID-19: what Inequality Effects by Income Source: A New does the data say?’, Social Protection Approaches to Approach and Applications to the United States.” COVID-19 Expert Advice Service (SPACE), DAI Global Review of Economics and Statistics 67(1): 151–56. doi: UK Ltd, United Kingdom. 10.2307/1928447. Beegle, Kathleen & De Weerdt, Joachim & Friedman, Lustig, N., and S. Higgins, 2016. “The CEQ Assessment: Jed & Gibson, John, 2012. “Methods of household Measuring the Impact of Fiscal Policy on Inequality consumption measurement through surveys: and Poverty.” Chapter 1, Commitment to Equity Experimental results from Tanzania,” Journal of Handbook. National Social Protection Framework, Development Economics, Elsevier, vol. 98(1), pages 3-18 Ministry of Local Government and Rural Development. Deaton, Angus, and Salman Zaidi. 2002. “Guidelines for Government of Botswana, June 2017. constructing consumption aggregates for welfare Masisi, M. E. K. (2020b). State of the Nation Address by analysis.” Living Standards Measurement Study His Excellency Dr. Mokgweetsi E.K. Masisi, President Working Paper No. 135. of the Republic of Botswana, to the First meeting of the Funk, C., Peterson, P., Landsfeld, M. et al. 2015. “The Second Session of the Twelfth Parliament, 09 November climate hazards infrared precipitation with stations—a 2020. Gaborone: Government of Botswana. Retrieved new environmental record for monitoring extremes.” from https://vision2036.org.bw/sites/default/files/ Sci Data 2, 150066. https://doi.org/10.1038/ resources/SONA%202020.pdf (20.04.2022). sdata.2015.66 MasterCard. 2022. ”Africa Claims Top Three Spots in Gaddis, Isis, Gbemisola Oseni, Amparo Palacio-Lopez, and Mastercard Index for Highest Concentration of Janneke Pieters. 2020. “Who is employed? Evidence Women Business Owners in the World” MasterCard from Sub-Saharan Africa on Redefining Employment”. Social Newsroom. 13 January 2021. https://www. Policy Research Working Paper 9370. World Bank. mastercard.com/news/eemea/en/newsroom/ August. https://documents1.worldbank.org/curated/ press-releases/press-releases/en/2021/january/ en/468881598538973944/pdf/Who-Is-Employed- africa-claims-top-three-spots-in-mastercard-index-for- Evidence-from-Sub-Saharan-Africa-on-Redefining- highest-concentration-of-women-businesses-owners- Employment.pdf in-the-world/#:~:text=Africa%20claims%20top%20 Gibson, John & Huang, Jikun & Rozelle, Scott, 2001. “Why three%20spots,in%20the%20world%20%7C%20 is income inequality so low in China compared to other Mastercard%20Newsroom countries?: The effect of household survey methods,” Min, Brian and O’Keeffe, Zachary. 2021. High Resolution Economics Letters, Elsevier, vol. 71(3), pages 329-333, Electricity Access Indicators Dataset. Ann Arbor, MI: June. Center for Political Studies, University of Michigan. Government of Botswana. 2016. Vision 2036: Achieving Molinas, J., R. Paes de Barro, J. Saavedra and M. Giugale. Prosperity for All. Gaborone: Government of Botswana. 2012. “Do Our Children Have a Chance?” The 2010 https://www.statsbots.org.bw/sites/default/files/ Human Opportunity Report for Latin America and the documents/Vision%202036.pdf. Caribbean. Washington, DC: World Bank. Gronbach, Lena, Jeremy Seekings, and Winnie Arthur. Paes de Barros, R., Ferreira, F., Molinas Vega, J. and Forthcoming. “Lessons learned on social protection Saavedra Chanduvi, J. 2009. Measuring Inequality of responses to Covid-19 in Botswana”. Centre for Social Opportunities in Latin America and the Caribbean. Science Research Working Paper, University of Cape Washington, DC: World Bank. Town. Ravallion, Martin, 2016. “The Economics of Poverty: IFC (International Finance Corporation). 2022. Country History, Measurement, and Policy,” OUP Catalogue, Private Sector Diagnostic: Creating Markets in Oxford University Press, number 9780190212773 Botswana. A Diamond in the Rough: Toward a New Rodolfo Beazley, Jana Bischler and Alexandra Doyle (2021) Strategy for Diversification and Private Sector Growth. ‘Towards shock-responsive social protection: Lessons Washington, DC: World Bank. https://www.ifc.org/en/ from the COVID-19 response in six countries’, Towards insights-reports/2022/cpsd-botswana. Shock-Responsive Social Protection, Oxford Policy International Labour Organization. 2023. “ILO Modelled Management, Oxford; Estimates and Projections database (ILOEST)” Rudhumbu, N, du Plessis, EC, and Maphosa, C. 2020. ILOSTAT. Accessed October 10, 2023. https://ilostat. ”Challenges and Opportunities for Women ilo.org/data/. Employment to population ratio, 15+, Entrepreneurs in Botswana: Revisiting the Role of total (%). Entrepreneurship Education.” Journal of International 56 BOTSWANA POVERTY ASSESSMENT Education in Business, v13, n2. https://eric. Statistics Botswana. 2023a. Consumer Price Index ed.gov/?id=EJ1266818 December 2022. https://www.statsbots.org.bw/sites/ Smith, Lisa C., Olivier Dupriez, and Nathalie Troubat. default/files/Consumer%20Price%20Index%20 2014. “Assessment of the Reliability and Relevance December%202022.pdf of the Food Data Collected in National Household Statistics Botswana. 2023b. Gross Domestic Product: Consumption and Expenditure Surveys.” IHSN Fourth Quarter of 2022. Gaborone: Statistics Working Paper No. 008. Statistics Botswana. 2013. Botswana. March. https://www.statsbots.org.bw/sites/ Botswana Core Welfare Indicators Survey 2009/10. default/files/publications/Gross%20Domestic%20 Gaborone: Statistics Botswana. https://www.statsbots. Product%20Q4%202022.pdf org.bw/sites/default/files/BCWIS%202009%20 Sulla, Victor; Precious Zikhali, and Pablo F. Cuevas. 10%20MAIN%20REPORT.pdf. 2022. Inequality in Southern Africa: An Assessment Statistics Botswana. 2015. Mapping Poverty in Botswana of the Southern African Customs Union (English). 2010. https://www.statsbots.org.bw/sites/default/ Washington, DC: World Bank. http://documents. files/publications/POVERTY%20Mapping%202010_ worldbank.org/curated/en/099125303072236903/ May%2028%202015.pdf P1649270c02a1f06b0a3ae02e57eadd7a82. Statistics Botswana. 2018. 2015/16 Botswana Multi-Topic Syed, Tahira, Peter Goodman, Zano Mataruka, David Household Survey. Gaborone: Statistics Botswana. Evans; Hope T. Pachena. 2022. Roadmap for https://www.statsbots.org.bw/sites/default/files/ Sustainable Livestock Value Chains in Southern Africa. publications/Botswana%20Multi%20Topic%20 Washington, DC: World Bank. http://documents. Household%20Survey%20REPORT%202015%2016. worldbank.org/curated/en/099430006082232802/ pdf. P174621055c7850b80808d09c38c18d5ed9. Statistics Botswana. 2019a. Quarterly Multi-Topic Survey: UNDRR. 2019. https://www.undrr.org/media/73965 Labour Force Module Report – Quarter 3: 2019. UNICEF (United Nations Children’s Fund), WHO (World Gaborone: Statistics Botswana. https://www.statsbots. Health Organization), and World Bank. 2021. Levels org.bw/sites/default/files/publications/Quarterly%20 and Trends in Child Malnutrition – UNICEF/WHO/ Multi%20Topic%20Survey%20Labour%20Force%20 World Bank Joint Child Malnutrition Estimates: Key Module%20Report-%20Q3%202019.pdf. Findings of the 2021 edition. https://www.who.int/ Statistics Botswana. 2019b. Quarterly Multi-Topic Survey: publications/i/item/9789240025257. Labour Force Module Report – Quarter 4: 2019. World Bank. 2020a. The Human Capital Index 2020 Gaborone: Statistics Botswana. https://www.statsbots. Update: Human Capital in the Time of COVID-19. org.bw/sites/default/files/publications/Multi%20 Washington, DC: World Bank. https://www. Topic%20Survey%20Q4%20Labour%20Force%20 worldbank.org/en/publication/human-capital#Data. Module%20Report_0.pdf. World Bank. 2020b. Poverty and Shared Prosperity 2020: Statistics Botswana. 2020a. Quarterly Multi-Topic Survey: Reversals of Fortune. Washington, DC. https:// Labour Force Module Report – Quarter 1: 2020. openknowledge.worldbank.org/handle/10986/34496 Gaborone: Statistics Botswana. https://www.statsbots. World Bank. 2022a. Poverty and Shared Prosperity 2022: org.bw/sites/default/files/publications/Multi- Correcting Course. Washington, DC: World Bank. Topic%20Survey%20%20Q%201%202020%20 doi:10.1596/978-1-4648-1893-6. https://www. LABOUR%20FORCE%20MODULE%20REPORT.pdf. worldbank.org/en/publication/poverty-and-shared- Statistics Botswana. 2020b. Quarterly Multi-Topic prosperity Survey: Labour Force Module Report – Quarter 4: World Bank. 2022b. Botswana Social Protection 2020. Gaborone: Statistics Botswana. https://www. Programs and Systems Review. Washington, DC: statsbots.org.bw/sites/default/files/publications/ World Bank. https://documents1.worldbank. MULTI-TOPIC%20SURVEY%20QUARTER%204%20 org/curated/en/099645103062256866/pdf/ 2020%20LABOUR%20FORCE%20MODULE%20 P1721750af4ddb0ce098d30b331412b0ab8.pdf. REPORT.pdf. World Bank. 2023. At a Crossroads: Reigniting Efficient Statistics Botswana. 2021. Quarterly Multi-Topic Survey and Inclusive Growth. Botswana Systematic Country Quarter 4, 2021. Gaborone: Statistics Botswana. Diagnostic Update. Washington, DC: World Bank. https://www.statsbots.org.bw/sites/default/ World Bank. 2024a (forthcoming). In Search of New files/publications/Quarterly%20Multi-Topic%20 Drivers of Inclusive Growth. Botswana Country Survey%20Report%20Q4%2C%202021.pdf. Economic Memorandum. Statistics Botswana. 2022a. Population & Housing World Bank. 2024b (forthcoming). Trends and Census 2022: Population of Cities, Towns, Villages & Opportunities to Advance Gender Equality in Associated Localities. Gaborone: Statistics Botswana. Botswana. A Country Gender Assessment. https://www.statsbots.org.bw/sites/default/files/ Worldwide Bureaucracy Indicators dataset (version 3.0). publications/Population%20%26%20Housing%20 August 2022. https://www.worldbank.org/en/data/ Census%202022-%20Population%20of%20 interactive/2019/05/21/worldwide-bureaucracy- Cities%2C%20Towns%2C%20Villages%20%26%20 indicators-dashboard#2 Associated%20Localities.pdf. Yoshida, N., R. Munoz, A. Skinner, C. Kyung-eun Lee, M. Statistics Botswana. 2022b. Quarterly Multi-Topic Survey Brataj, W. Durbin, and D. Sharma. 2015. SWIFT Data Quarter 4, 2022. Gaborone: Statistics Botswana. Collection Guidelines Version 2. Washington, DC: https://www.statsbots.org.bw/sites/default/ World Bank. https://documents1.worldbank.org/ files/publications/Quarterly%20Multi-Topic%20 curated/en/591711545170814297/pdf/97499-WP- Survey%20Quarter%204%2C%202022_0.pdf P149557-OUO-9-Box391480B-ACS.pdf. BOTSWANA POVERTY ASSESSMENT 57 APPENDIX 58 BOTSWANA POVERTY ASSESSMENT APPENDIX 1 STRUCTURAL PEERS Structural peers refer to countries around the globe that share economic characteristics similar to those of a particular country. In the case of Botswana, 20 indicators were selected for analysis from the World Development Indicators database, covering the period between 2012 and 2018. These indicators include GDP and its composition, population characteristics, and labor market indicators. The indicators were standardized, and the differences among countries concerning the best performance in each category were used to identify similar structural countries. For Botswana, eight countries were identified as structural peers, robust to different combinations of the indicators: Gabon, Georgia, Lebanon, Mongolia, Namibia, the Republic of Congo, South Africa, and Tunisia. The 20 indicators chosen from the World Development Indicators database between 2012 and 2018: 1. The labor force participation rate (% of total population ages 15-64) 2. The Gini Index 3. GDP per capita, PPP (current international $) 4. Industry (including construction), value added (% of GDP) 5. Manufacturing, value added (% of GDP) 6. Imports of goods and services (% of GDP) 7. Gross fixed capital formation (% of GDP) 8. Exports of goods and services (% of GDP) 9. General government final consumption expenditure (% of GDP) 10. Vulnerable employment, total (% of total employment) (modeled ILO estimate) 11. Mortality rate, under-5 (per 1,000 live births) 12. Rural population (% of total population) 13. Economically active population in agriculture (number) 14. Employment in agriculture (% of total employment) (modeled ILO estimate) 15. Access to electricity, rural (% of rural population) 16. Arable land (hectares per person) 17. Population, total 18. Total natural resources rents (% of GDP) 19. Survey mean consumption or income per capita, total population (2011 PPP $ per day) 20. Poverty headcount ratio at $1.90 a day (2011 PPP) (% of population) Source: World Development Indicators (database), databank.worldbank.org/source/world-development-indicators, version 2022. BOTSWANA POVERTY ASSESSMENT 59 APPENDIX 2 POPULATION PYRAMID 2022 A population pyramid graphically shows the age-sex structure of a given population. Botswana’s population pyramid for 2022 has a wide base, highlighting that Botswana has a relatively young population. The rapidly tapering top suggests that Botswana has a low life expectancy. The results of the 2022 Population and Housing Census show an estimated population of 2,346,179 and a population growth rate of around 1.4 percent. Life expectancy is around 66 years. FIGURE A2 FIGURE A2.1  Botswana Population Pyramid 2022 Botswana Population 2022 80 and above 70-74 60-64 50-54 40-44 female male 30-34 20-24 10-14 0-4 12% 10% 8% 6% 4% 2% 0% 2% 4% 6% 8% 10% 12% Source: World Bank calculations based on World Development Indicators, version 7/25/2023. The 2022 values are projections from the 2011 Population and Housing Census. 60 BOTSWANA POVERTY ASSESSMENT APPENDIX 3 PROBABILITY OF BEING EMPLOYED TABLE A3.1  Probability of being employed (Probit) 2015-2022 BMTHS 2019Q3 2019Q4 2020Q1 2020Q4 2021Q4 2022Q4 2016 Demographics Age 0.172*** 0.166*** 0.175*** 0.163*** 0.170*** 0.162*** 0.167*** Age squared -0.002*** -0.002*** -0.002*** -0.002*** -0.002*** -0.002*** -0.002*** Gender (=1 if male) 0.347*** 0.345*** 0.341*** 0.340*** 0.406*** 0.375*** 0.316*** Education (Primary or less = omitted category) Middle education 0.350*** 0.388*** 0.236*** 0.231*** 0.403*** 0.375*** 0.265*** High education 0.615*** 0.512*** 0.482*** 0.395*** 0.557*** 0.754*** 0.499*** Strata (Cities&Towns = omitted category) Urban Villages -0.234*** -0.133*** -0.155*** -0.182*** -0.240*** -0.294*** -0.253*** Rural areas -0.162*** -0.111*** -0.038*** -0.021*** -0.117*** -0.146*** -0.105*** Constant -3.780*** -3.693*** -3.801*** -3.565*** -3.814*** -3.711*** -3.633*** Number of 1,360,089 1,572,922 1,618,129 1,637,913 1,631,858 1,637,768 1,611,721 observations Adjusted R2 0.173 0.159 0.174 0.166 0.176 0.178 0.165 Source: World Bank calculations. Note: The asterisk (*) indicates statistical significance at 1% (***), 5% (**), or 10% (*) level. BOTSWANA POVERTY ASSESSMENT 61 APPENDIX 4 EARNINGS REGRESSION TABLE A4.1  Earnings regression (log monthly real wages, base 2010) BMTHS 2019Q3 2019Q4 2020Q1 2020Q4 2021Q4 2022Q4 2015 Demographics Age 0.113*** 0.081*** 0.094*** 0.086*** 0.066*** 0.086*** 0.065*** Age squared -0.001*** -0.001*** -0.001*** -0.001*** -0.000*** -0.001*** -0.001*** Gender (=1 if male) 0.393*** 0.474*** 0.413*** 0.355*** 0.361*** 0.340*** 0.402*** Strata (Cities&Towns = omitted category) Urban Villages -0.169*** -0.147*** -0.229*** -0.161*** -0.267*** -0.117*** -0.202*** Rural areas -0.382*** -0.430*** -0.493*** -0.420*** -0.566*** -0.408*** -0.398*** Education (Primary or less = omitted category) Secondary 0.652*** 0.599*** 0.449*** 0.551*** 0.608*** 0.611*** 0.559*** Tertiary 1.788*** 1.725*** 1.496*** 1.631*** 1.767*** 1.750*** 1.813*** Sector (Agro & Fishing = omitted category) Mining 0.954*** 1.023*** 0.986*** 0.897*** 0.758*** 0.785*** 1.106*** Manufacturing & 0.307*** 0.182* 0.353*** 0.199** 0.074 0.108 0.163* Utilities Construction 0.376*** -0.085 0.133 0.029 -0.074 -0.146 -0.015 Whole Sale & 0.191** 0.222*** 0.171** 0.060 -0.052 -0.076 0.118* Retail Public 0.078 0.117 0.156** 0.024 0.091 0.082 0.066 administration Education 0.567*** 0.610*** 0.618*** 0.540*** 0.329*** 0.398*** 0.430*** Rest Serv 0.248*** 0.252*** 0.316*** 0.193*** 0.073 0.101 0.163** Constant 3.667*** 4.193*** 4.177*** 4.289*** 4.727*** 4.203*** 4.523*** Number of 5,259 2,920 2,745 2,690 2,678 2,601 2,992 observations Adjusted R2 0.501 0.526 0.514 0.534 0.538 0.545 0.530 Source: World Bank calculations based on BMTHS and QMTS surveys (various years). Note: The asterisk (*) indicates statistical significance at 1% (***), 5% (**), or 10% (*) level. 62 BOTSWANA POVERTY ASSESSMENT APPENDIX 5 POPULATION DISTRIBUTION FROM POPULATION AND HOUSING CENSUS 2001, 2011, 2022 TABLE A5.1  Population Distribution District Population Distribution Populalfon Distribution (%) SN Code Census Dlsfrtcts 2001 2011 2022 2001 2011 2022 1 01 Gaborone 186.007 231.592 244,I07 11,1 11,4 10,4 2 02 Francistown 83.023 98.961 102.444 4,9 4,9 4,4 3 03 Lobotse 29.689 29.007 29.457 1,8 1,4 1,3 4 04 Selebi_phikwe 49.849 49.411 41.839 3 2,4 1,8 5 05 Orapa 9.151 9.531 8.614 0,5 0,5 0,4 6 06 Jwaneng 15.179 18.008 18.576 0,9 0,9 0,8 7 07 Sowa Town 2.879 3.598 2.901 0,2 0,2 0,1 8 10 Ngwaketse 113.704 129.247 140.321 6,8 6,4 6 9 11 Barolong 47.477 54.831 58.394 2,8 2,7 2,5 10 12 Ngwaketse West 10.471 13.689 23.253 0,6 0,7 1 11 20 South Ea.st 60.623 85.014 111.474 3,6 4,2 4,8 12 30 Kweneng East 189.773 256.752 330.442 11,3 12,7 14,1 13 31 Kweneng West 40.562 47.797 57.261 2,4 2,4 2,4 14 40 Kgatleng 73.507 91.660 121.411 4,4 4,5 5,2 Central Serowe 15 50 153.035 180.500 201.775 9,1 8,9 8,6 Palapye 16 51 Central Mahalapye 109.811 118.875 130.530 6,5 5,9 5,6 17 52 Central 8obonong 66.964 71.936 76.922 4 3,6 3,3 18 53 Central 8oteti 48.057 57.376 74.099 2,9 2,8 3,2 19 54 Central Tutume 123.514 147.377 164.228 7,3 7,3 7 20 60 North East 49.399 60.264 68,9IO 2,9 3 2,9 21 70 Ngamiland East 75.070 90.334 120.603 4,5 4,5 5,1 22 71 Ngamiland West 49.642 59.421 73.122 3 2,9 3,1 23 72 Chobe 18.258 23.347 28.388 1,1 1,2 1,2 24 73 Delta 2.529 2.849 0,1 0,1 25 80 Ghanzi 33.170 43.095 55.396 2 2,1 2,4 26 81 CKGR 260 488 0 0 27 90 Kgalagadi South 25.938 30.016 35.160 1,5 1,5 1,5 28 91 Kgalagadi North 16.111 20.476 23.215 1,0 1,0 1,0 Total 1.680.863 2.024.904 2.346.179 100 100 100 Statistics Botswana: Pg 8 https://www.statsbots.org.bw/sites/default/files/publications/2022%20Population%20and%20Housing%20 Census%20Preliminary%20Results%20Aug.pdf BOTSWANA POVERTY ASSESSMENT 63 APPENDIX 6 FIGURE Apendix 6 MAPS OF VILLAGES BY DECILES Decile 1 Decile 2 Decile 3 Decile 4 Decile 5 Decile 6 Decile 7 Decile 8 Decile 9 b. Decile 10 Source: World Bank calculations using 2011 Poverty Map. 64 BOTSWANA POVERTY ASSESSMENT APPENDIX 7 REDEFINING EMPLOYMENT IN BOTSWANA Redefining Employment in Botswana: Comparable Labor Market Statistics for 2002-2022 Carolina Diaz-Bonilla Santiago Garriga Introduction Botswana’s employment and unemployment statistics and their trend over time vary depending on which international standard is applied. Since the launch of the first Quarterly Multi-Topic Survey (QMTS) in quarter 3 of 2019, Statistics Botswana has been applying the new international definition of employment in its official labor market statistics, causing a break with previous official statistics. Although one could argue the benefits of either definition, in the end the possibility to compare employment and unemployment statistics over time requires the application of the same definition across the entire time period. This note briefly explains the changes and presents a set of comparable labor market statistics for the full 2002-2022 period in Botswana. Background on the new definition of employment In October 2013, the 19th International Conference of Labour Statisticians (ICLS)1, hosted by the International Labor Organization (ILO), adopted a new set of standards regarding the definition of employment, among other changes.2 The new and more narrow definition of employment would refer to “work performed for pay or profit” while a different headline indicator was created for “own-use production work”. Effectively this meant that the production of goods for own use, such as crop cultivation mainly or only for own/family consumption, would no longer be considered employment and, therefore, would no longer count towards employment or labor force participation. In the case of agricultural activities, this meant that employment would be determined by the self-declared main intended use of the output produced. As explained by Gaddis et al. (2020)3: “The main objective of these changes was to increase visibility of all forms of work, paid and unpaid, by advocating for separate measurement and acknowledging that individuals are often engaged in different types of work. These changes have significant implications for the calculation of labor statistics in developing countries, and especially in Sub-Saharan Africa, where a large share of the population is engaged in the production of goods for family use. They also raise important issues with how to measure this definition in a consistent and robust manner.” Comparable labor market statistics in Botswana for 2002-2022 The labor market statistics published by Statistics Botswana in 2002/03, 2009/10, and 2015/16, applied the 18th ICLS standard, while the statistics published in the QMTS surveys in 2019, 2020, 2021, and 2022 all applied the 19th ICLS standard. The current statistics are not inaccurate per se, they are in line with new  international  definitions from  ILO, but they are not comparable to the earlier data. The simplest explanation is that subsistence farmers are no longer considered employed in the 2019-2022 QMTS, 1  The International Conference of Labour Statisticians (ICLS) is a “vehicle for standard-setting in labour statistics, hosted by the Interna- tional Labour Organization (ILO) every five years”. “Experts in labour statistics from all over the world attend the ICLS. This includes most notably statisticians who work in national statistical offices, ministries of labour and selected representatives of workers’ and employers’ organizations.” The standards adopted are meant to reflect “best practice” but are not prescriptive nor binding. See: https://ilostat.ilo.org/ about/standards/icls/?playlist=4194a13&video=38313ec 2  Available here (link). 3 https://documents1.worldbank.org/curated/en/468881598538973944/pdf/Who-Is-Employed-Evidence-from-Sub-Saharan-Afri- ca-on-Redefining-Employment.pdf. The accompanying blog: https://blogs.worldbank.org/opendata/working-hard-and-not-being-count- ed-evidence-sub-saharan-africa-redefining-employment BOTSWANA POVERTY ASSESSMENT 65 while they were considered employed in 2015/16.4 A concrete result of this is that the current official unemployment rate for Botswana in 2022-Q4 (considering the labor force as 15 years or above) is 25.4 percent, based on implementing the 19th ICLS standards, while it is 22.7 percent using the 18th ICLS standards. The comparable labor market series presented below use the 18th ICLS standards for two important reasons. First, National Accounts data include subsistence farming as part of GDP. Therefore, to maintain comparability with National Accounts, the employment statistics are adjusted to also include subsistence farmers (which are a significant category across many Sub-Saharan African countries). Second, it is not clear whether it would be possible to adjust the oldest surveys to match the new methodology, especially since many of the questions in the recent questionnaire were not collected in the past. Therefore, it is simpler to harmonize the most recent data to match the older definitions. Figure 1 below shows the new (comparable) labor market series that incorporates subsistence farmers. The three solid lines from 2002 and 2022 correspond to the labor force participation rate, the employment rate, and the unemployment rate (all consider a population of fifteen years or more) when subsistence farmers are maintained in the employment statistics. Note that we use the full range of available survey rounds to visualize long-run trends in labor market outcomes. The three dashed line correspond to the official statistics published by Statistics Botswana using the 2015/16 BMTHS and the 2019, 2020, 2021, and 2022-Q4 QMTS. The large drop in labor force participation and the large drop in the employment rate are due to a methodological change in the definition of these statistics and not due to changes in actual economic conditions. The comparable labor market statistics show that the labor force participation rate was 63.7 percent in 2022 and the employment rate 49.2 percent, both higher than the official published statistics. The unemployment rate using the comparable data reached 22.7 percent in 2022, less than the official figure suggests, but nevertheless a worrying continuous increase since 2010. It also suggests that unemployment continued to increase between 2021-Q4 and 2022-Q4, rather than the decline seen in the official statistics. The 19th ICLS also changes the sectoral composition of the labor force towards non-agricultural sectors. As shown below, the 18th ICLS standard suggests that the services sector accounts for 66.7 percent of total employment in 2022 while the 19th ICLS standard would suggest that this is 73.2 percent instead. In other words, the new standard suggests that Agriculture is only 9.5 percent of employment, whereas using the 18th ICLS that is comparable to 2003-2016 shows that Agriculture makes up 17.6 percent of total employment in 2022. A comparison of subsistence and non-subsistence agricultural workers between 2016 and 2022 shows that the number of subsistence farmers may have remained at similar levels in that period. However, non-subsistence agricultural employment increased since 2016 (Figure 4). Applying the 18th ICLS standard would result in 70,273 subsistence farmers being considered “employed” in 2022. The 19th ICLS standard would instead consider them to be working for own production but not employed. In conclusion, the statistics that are officially published for 2019-2022 by Statistics Botswana are correctly following the ILO’s 19th ICLS standard. However, this results in statistics that are not comparable to the data for 2003-2016. Therefore, if the intention is to analyze long-term trends from before 2019, then an adjustment needs to be done to make the data comparable. 4  For instance, there is not a large drop in rural employment between 2015 and 2019 – it is, indeed, a change in methodology. 66 BOTSWANA POVERTY ASSESSMENT Fig 7.1 FIGURE A7.1  Labor force statistics for 2002-2022 using the 18th ICLS standard compared to 2019-2022 official data using the 19th ICLS standard 70 64.4 63.0 63.5 Labor Force Participation 60.0 61.3 63.7 58.6 Rate (%, 15 yrs +) 60 59.7 59.0 Official LFP (19th ICLS) 48.6 50.5 50.3 50.7 49.2 50 45.7 49.2 Employment Rate 44.5 40 45.9 (%, 15 yrs +) Official Employment 30 (19th ICLS) 22.2 25.4 22.7 Unemployment Rate (%) 20 23.9 22.5 20.1 21.4 Official Unemployment 17.1 17.6 (19th ICLS) 10 2003 2009 2016 2019- 2020- 2021- 2022- Q4 Q4 Q4 Q4 Source: World Bank calculations using 2002/03 HIES, 2009/10 BCWIS, 2015/16 BMTHS, and QMTS (various years). FIGURE A7.2 FIGURE A7.3 FIGURE A7.2  Employment Shares by Sector - FIGURE A7.3  Employment by Sector - 2022 2022 using 18th vs 19th ICLS standard using 18th vs 19th ICLS standard Employment Shares by Sector Employment by Sector 100% 900,000 460,145 601,912 683,603 787,997 717,725 11.7 9.5 90% 20.8 17.6 800,000 26.3 80% 12.2 15.2 700,000 138,674 13.9 68,401 70% 79,715 19.8 15.0 600,000 109,443 109,443 60% 158,390 83,637 500,000 50% 400,000 95,854 40% 90,396 73.9 73.2 30% 66.7 300,000 90,886 56.4 56.1 504,907 525,547 525,547 20% 200,000 337,629 259,641 10% 100,000 0% 0 2003: 2009: 2016: 2022: 2022: 2003: 2009: 2016: 2022: 2022: 18th 18th 18th 18th 19th 18th 18th 18th 18th 19th ICLS ICLS ICLS ICLS ICLS ICLS ICLS ICLS ICLS ICLS Agriculture Mining Industry+Const Services Agriculture Mining Industry+Const Services Total Source: World Bank calculations using 2002/03 HIES, 2009/10 Source: World Bank calculations using 2002/03 HIES, 2009/10 BCWIS, 2015/16 BMTHS, and 2022-Q4 QMTS. BCWIS, 2015/16 BMTHS, and 2022-Q4 QMTS. FIGURE A7.3 FIGURE A7.4  Subsistence and non-subsistence employment in 2016 and 2022 (Applying the 18th ICLS standard in both years) 160,000 140,000 120,000 100,000 68,401 80,000 Subsistence 11,851 Not subsistence 60,000 40,000 69,406 70,273 20,000 0 2016 2022 Source: World Bank calculations using 2015/16 BMTHS and 2022-Q4 QMTS. BOTSWANA POVERTY ASSESSMENT 67 APPENDIX 8 TECHNICAL REPORT ON BOTSWANA’S HOUSEHOLD SURVEYS AND THE CREATION OF A COMPARABLE SPATIALLY-DEFLATED CONSUMPTION AGGREGATE5 1. Comparability of the BCWIS 2009/10 and the BMTHS 2015/16 Household Income and Expenditure Surveys (HIES) have a long and well-established tradition in Botswana. The first Botswana HIES (BHIES) was conducted in 1985/1986, which was followed by the 1993/94 BHIES, 2002/03 BHIES, 2009/10 Botswana Core Welfare Indicators Survey (BCWIS), and 2015/16 Botswana Multi- Topic Household Survey (BMTHS). This annex provides a brief account of the methodologies underlying the surveys conducted in 2009/10 and 2015/16. How comparable is the BMTHS 2015/16 with the BCWIS 2009/10? What is the impact of different definitions and methods underlying the surveys on inequality and poverty estimates? A growing literature suggests that changes in survey design need to be considered when analyzing trends in consumption, inequality, or poverty over time (Lanjouw and Lanjouw, 1997; Beegle et al. 2012). The “details” related to: i) the method of data capture (e.g., diary versus recall); ii) the respondents (individuals versus households); iii) the reference periods for which consumption is reported (whether on week, two weeks, one month, all the way to one year); and iv) the degree of commodity detail (e.g., the detail of the Classification of Individual Consumption According to Purpose - COICOP system) greatly matter. Gibson et al. (2001), for instance, have shown the high sensitivity of inequality measures in China when incomes are collected for one month instead of one year. A large number of other examples could be easily cited (Smith, Dupriez, and Troubat, 2014). In 2009/10, the Central Statistics Office, now Statistics Botswana, conducted the Botswana Core Welfare Indicator Survey. The survey was carried out between April 2009 and March 2010 and collected information on household consumption and expenditure, on income, employment, asset ownership, agriculture, health and nutritional status, and education. Preliminary results were published in Statistics Botswana (2013a) and a full report in Statistics Botswana (2013b). In 2015/16, a new survey was launched, and fielded out between November 2015 and November 2016. The new Botswana Multi-Topic Household Survey improved on the 2009/10 BCWIS in several ways, as summarized in Table A8.1. The table summarizes a selection of indicators that play a role in constructing welfare indicators. The information has been organized into two broad categories: (i) the survey design and (ii) the questionnaire. Both BCWIS and BMTHS were conducted nationwide, using administrative district and sub-district boundaries. Only private dwellings were within the scope of the surveys; institutional dwellings (prisons, hospitals, army barracks, hospitals, hotels, and other institutions) and places completely within industrial areas were excluded. However, there are two key differences between the two surveys: (a) the reduction of the period of diary-keeping in each household from one month to 14 days; and (b) the switch from an expenditure- to a consumption-based food diary module. Both changes help align the BMTHS with the best practice of surveys of this kind elsewhere, the general tendency being to reduce the diary-keeping periods due to respondent fatigue. A drawback of these innovations, however, is that they are likely to create comparability issues (Beegle et al. 2012; Caeyers et al 2012). Non-food expenditures were inquired by recall, as in the previous surveys. 5  This appendix is from an unpublished report prepared by Giovanni Vecchi in May 2019. The original report was produced as part of a collaboration on data production and poverty analysis between the World Bank and Statistics Botswana, with a focus on the 2015/16 Botswana Multi-Topic Household Survey. The report benefited from discussions with the Statistics Botswana poverty statistics team: Mr. Moffat Malepa, Mrs. Kutlwano Sebolaaphuti, led by Dr. Burton Mugani (then-Deputy Statistician General for Economic and Social Statistics). 68 BOTSWANA POVERTY ASSESSMENT TABLE A8.1  Comparison of the 2009/10 BCWIS and 2015/16 BMTHS 2009/10 BCWIS 2015/16 BMTHS Survey design 2002/03 Population and Housing 2011 Botswana Population and Sample frame Census Housing Census Sample design Two-stage stratified Two-stage stratified Strata 3 (1) 3(1) PSU 288 out of 4,114 enumeration areas 599 Response rate (%) 98.2 Actual sample size (hh) 7,731 7,060 Actual sample size (ind) 27,211 24,719 Population (est.) 1,874,414 2,073,675 Households (est.) 541,593 589,725 Household size (est.) 3.46 3.52 Reference period April 2009 – March 2010 Nov 2015 – Oct 2016 Rounds 12 12 National, by stratum within each of National, by stratum within each of Representativeness 7 regions 7 regions Questionnaire Data collection PAPI CAFE Food items: diary vs. recall diary diary Information collected on food expenditure consumption and expenditure Reference period (food expenditure) 4 weeks 14 days Non-food expenditures: recall recall diary vs. recall Reference period (non-food 1, 3, and 12 months 1 week, 1 and 12 months expenditure) Notes: (1) three strata: cities/towns, urban villages, and rural areas. “PAPI” is for Paper-Assisted Personal Interviewing, “CAFE” is for Computer-Assisted Field-based data entry. Overall, the 2009/10 BCWIS and the 2015/16 BMTHS are comparable surveys in terms of geographical representation, sample size, and the core modules of the questionnaires. A number of methodological differences might have an impact on the analysis, however. The main concerns are related to: 1. the sample frame (the new 2011 population census has replaced the 2002/03 one used in past), 2. the data entry system (PAPI vs. CAFE), 3. the diary duration (2 weeks vs. 4 weeks), and 4. the change of the information collected in the food diary (based on expenditures in the past, now based on both consumption and expenditures). All this warrants caution in interpreting the results and deserves further research and investigation, a task beyond the scope of this report. BOTSWANA POVERTY ASSESSMENT 69 2. Construction of a comparable nominal consumption aggregate This section focuses on the construction of a nominal consumption aggregate (CA) for 2015/16 that is as closely comparable as possible with the one used in 2009/10. One distinctive feature of the official estimates produced by Statistics Botswana is that two versions of the consumption aggregate are used. A first definition is meant to describe the expenditure pattern of the households. This is what one finds in section 3.0 (Technical Information) of Statistics Botswana (2018), here reproduced in Table A8.2. TABLE A8.2  Average monthly household expenditure by consumption item and amount in Pula - 2009/10 and 2015/16 2009 2015 Type of consumption Cities/ Urban Cities/ Urban Rural National Rural National expenditure/outlay Towns Villages Towns Villages Food 694.78 649.07 352.89 537.61 485.97 501.00 513.98 501.77 Alcohol and Tobacco 268.51 259.10 292.38 275.38 161.42 148.83 113.89 139.89 Clothing and Footwear 343.44 266.11 140.21 233.68 365.14 235.61 150.68 238.23 Housing Costs 739.38 450.11 182.16 413.29 1196.82 736.01 297.06 698.20 H/hold Goods and Services 412.39 257.51 138.35 247.91 339.02 156.99 134.85 194.15 Medical/Health care 27.35 48.10 12.77 28.03 277.51 107.65 66.80 135.38 Transport 1150.40 629.49 305.76 629.27 1346.19 982.34 592.33 937.52 Communication 262.62 178.81 82.89 160.53 416.88 300.03 151.68 277.68 Recreation and Culture 241.13 144.59 67.34 137.37 201.72 111.48 50.86 112.80 Education 88.31 54.99 16.23 47.46 427.17 141.28 60.63 183.84 Restaurants and Hotels 223.49 47.87 17.05 80.39 269.10 140.31 66.30 146.51 Miscellaneous 479.14 255.05 115.91 255.00 586.35 362.31 197.45 360.65 Final Consumption Exp 4930.93 3240.80 1723.94 3045.93 6073.30 3923.83 2396.51 3926.63 Source: Statistics Botswana (2018: 15). https://www.statsbots.org.bw/sites/default/files/publications/BMTHS%20POVERTY%20STATS%20BRIEF%202018.pdf When it comes to measuring poverty and inequality, Statistics Botswana uses a second, more comprehensive definition of the consumption aggregate. Schematically, this second welfare indicator includes – in addition to all expenditure categories listed in Table A8.1 – the following items: • non-food wages in-kind (section 5, questions 508-510), • own production from livestock (section 11, question 1106), • own production from crops (section 11, question 1106), • education aid (section 2, question 233-238), • social protection food aid (section 8, question 830), • government in-kind social protection (section 8, question 831, 836). The data shared by Statistics Botswana allow us to reproduce the official estimates (the first definition mentioned above) in Table A8.2. Table A8.3 shows the results (the upper panel refers to expenditure levels, while the lower panel refers to budget shares). For most consumption categories, we managed to match official estimates pretty closely. Two exceptions are food expenditures and housing costs. 70 BOTSWANA POVERTY ASSESSMENT TABLE A8.3  Average household nominal expenditure (per household per month) by main consumption categories Official estimates New estimates Cities/ Urban Cities/ Urban Rural National Rural National Towns Villages Towns Villages Food 486 501 514 502 942 1,049 1,095 1,039 Alcohol and Tobacco 161 149 114 140 206 184 169 184 Clothing and Footwear 365 236 151 238 372 241 154 243 Housing Costs 1,197 736 297 698 1,365 970 403 872 H/hold Goods and 339 157 135 194 342 164 139 199 Services Medical/Health care 278 108 67 135 283 110 62 136 Transport 1,346 982 592 938 1,295 943 549 894 Communication 417 300 152 278 428 311 148 283 Recreation and 202 111 51 113 211 118 58 120 Culture Education 427 141 61 184 426 140 62 183 Restaurants and 269 140 66 147 285 151 71 156 Hotels Miscellaneous 586 362 197 361 512 330 203 331 Final Consumption 6,074 3,925 2,397 3,927 6,666 4,709 3,114 4,642 Exp. budget shares (%) budget shares (%) Cities/ Urban Cities/ Urban Rural National Rural National Towns Villages Towns Villages Food 8.0 12.8 21.4 12.8 14.1 22.3 35.2 22.4 Alcohol and Tobacco 2.7 3.8 4.8 3.6 3.1 3.9 5.4 4.0 Clothing and Footwear 6.0 6.0 6.3 6.1 5.6 5.1 4.9 5.2 Housing Costs 19.7 18.8 12.4 17.8 20.5 20.6 13.0 18.8 H/hold Goods and 5.6 4.0 5.6 4.9 5.1 3.5 4.5 4.3 Services Medical/Health care 4.6 2.7 2.8 3.4 4.2 2.3 2.0 2.9 Transport 22.2 25.0 24.7 23.9 19.4 20.0 17.6 19.3 Communication 6.9 7.6 6.3 7.1 6.4 6.6 4.7 6.1 Recreation and 3.3 2.8 2.1 2.9 3.2 2.5 1.9 2.6 Culture Education 7.0 3.6 2.5 4.7 6.4 3.0 2.0 4.0 Restaurants and 4.4 3.6 2.8 3.7 4.3 3.2 2.3 3.4 Hotels Miscellaneous 9.7 9.2 8.2 9.2 7.7 7.0 6.5 7.1 Final Consumption 100.0 100.0 100.0 100.0 100.0 100.0 100.0 100.0 Exp. Source: left panel is from Statistics Botswana (2018: 15); right panel is from our estimates. BOTSWANA POVERTY ASSESSMENT 71 Regarding housing, the category includes housing related costs and rent. In particular, in terms of rent, we include the rent paid as well as the approximation of rent for those who occupy a dwelling, reported in section 9 (question 913 and 915). Additionally, from the same section we include housing costs as spelled out in terms of electricity. Finally, from the recall of weekly, monthly, and annual payments (section 12), we include expenditure on all housing relative expenditure categories as defined by the respective COICOP codes. In order to improve on the comparability of the total household expenditure between 2009/10 and 2015/16 each component was recalibrated so as to match the variables provided by the official Statistics Botswana datasets. Table A8.4 shows the estimated values of the components 1) to 6) above. The rest of this section illustrates the details on how each main component was estimated based on the 2015/16 BMTHS. TABLE A8.4  Average monthly household expenditure by consumption item and the adjustment terms of the comparable consumption aggregate Component Value (pula/hh/month) Total consumption aggregate 4,642 Education aid 34 Own production consumption from livestock 27 Own production consumption from crops 20 Government in-kind aid 46 Wages in-kind (non food) 17 Social protection food aid 14 2.1 Food The main component of food expenditures, food consumption expenditure, was calculated from the diary. We use the diary information on daily consumption of food. In particular, information on food eaten in the 2015/16 diary is organized in two parts. In Part A1 households report, on a daily basis, the consumption of food. This includes i) items purchased, ii) received as a gift, and iii) produced at home, but only reports the consumed quantities, and not the corresponding amounts. This makes it necessary to estimate the value of these items in order to construct the food component of the consumption aggregate. Figure A8.1 shows the structure of this first part of the diary in the 2015/16 BTHS questionnaire. For consumption quantities, we use the variable “qty_g” produced by Statistics Botswana. This variable already converted non-standard measurement units (question 102 in Figure 2) into grams. Quantities were priced using the variable “unit_price”, also extracted from the official Statistics Botswana datasets. 72 BOTSWANA POVERTY ASSESSMENT FIGURE A8.1  Food consumption: excerpt from the diary of the 2015/16 BMTHS Source: 2015/16 BMTHS questionnaire. To obtain comparability of consumption aggregates between 2009/10 and 2015/16, we apply a bottom-coding procedure on the 2015/16 MBTHS data and adjust food consumption expenditure accordingly. This same procedure was implemented by the analysts at Statistics Botswana (as a way of dealing with extreme values and/or other data flaws/inconsistencies). In addition to food consumption expenditure, we include meals consumed outside home, which is available from the questionnaire. Part A of the diary reports expenditures on meals and drinks, separately by breakfast, lunch, and dinner (Figure A8.2). The values in the diaries have been included in the food aggregate, as reported by the households, after annualizing the fortnightly expenditures. Moreover, we include an estimate of the value of school meals for children studying in government schools. Based on the prices shared by Statistics Botswana, a meal for each primary school student is estimated at 103.3 Pula. Similarly, a meal for each secondary school student is valued at 182.5 Pula. In addition, we take into account the value of food received as in-kind wage. This is available in section 5 of the questionnaire (“Wage earners: income, deductions, and employee benefits”, section 5, question 506 and 507). The total amount obtained by summing up all the components described in this section is used as the first building block for the construction of the consumption aggregate. In the next section, we discuss the second building block, namely how non-food expenditures were calculated. BOTSWANA POVERTY ASSESSMENT 73 FIGURE A8.2  Meals outside home: excerpt from the diary of the 2015/16 BMTHS Source: 2015/16 BMTHS questionnaire. 2.2  Non-food expenditures Total consumption includes the following non-food expenditures: • Health expenditures, as available from Section 3 Part B (questions 341, 327, 329, 330, 333); • Education expenditures, as available from Section 2 (questions 225, 226, 227, 228, 229, 230, 231). • All non-food expenditures listed in Section 12 (Part A, expenditures in the past week, and Part B, expenditure in the past 12 months) as well as expenditure in Part C (regular monthly and annual payments). • For alcohol and tobacco, in order to impute the relative expenditure, we include information on their use from section 3, part C. We combine the information with tobacco and alcohol prices supplied by Statistics Botswana. We thus use questions 344, 351, 352, 353, 354, section 3, part C, and by reconstructing the relative standard units of measurement, we combine the quantities with the prices and construct the relative monthly expenditure. Finally, we use the weekly recall expenditure on alcohol and tobacco, and in cases where the weekly recall expenditure estimate outweighs the previously obtained aggregate, we augment it with the residual part. 2.3  Durable goods and housing As explained in Deaton and Zaidi (2002), both durable goods and housing require special treatment, which we briefly discuss in Box A8.2. Regarding durable goods we did not estimate the consumption flow that comes from owning or having access to durable goods, for consistency with the official definition used in 2009. The assumption is that it was used the ‘acquisition approach’ (see, for instance, Amendola and Vecchi 2014), and this is what we have implemented for the comparable CA. 74 BOTSWANA POVERTY ASSESSMENT Regarding housing, we included actual rent, as well as the imputed rent, which we estimate through a standard hedonic regression model, for households who own their dwellings. Additionally, from the same section we include housing costs as spelled out in terms of electricity. Finally, from the recall of weekly, monthly and annual payments (section 12), we include expenditure on all housing relative expenditure categories as defined by the respective COICOP codes. BOX A8.1  The consumption aggregate in Botswana and in theory The welfare indicator described in section 2 can be assessed against the theoretical recommendations provided by Deaton and Zaidi (2002). In the first place, Deaton and Zaidi argue that “Consumption is a theoretically more satisfactory measure of well-being” (p. 23): the Botswana consumption aggregate is therefore in line with the general theoretical framework used by welfare analysts. Regarding food expenditures, Deaton and Zaidi argue that the consumption aggregate should include “not just (i) food purchased in the market place, including meals purchased away from home for consumption at or away from home, but also (ii) food that is home-produced, (iii) food items received as gifts or remittances from other households, as well as (iv) food received from employers as payment in-kind for services rendered, In some cases where food can be and is stored over long periods of time, and where the questionnaire permits it, “food consumed” can be distinguished from “food purchased”. In principle, it is the value of the former that should go into the consumption aggregate.” (p. 27). This is in line with the choices documented in section 3.1. Regarding non-food expenditures, the consumption aggregate used by Statistics Botswana includes health expenditures, whereas Deaton and Zaidi argue that they should be excluded, or, more precisely, that they “should only be included if they have high income elasticity in relation to their transitory variance or measurement error” (p. 39). Durable goods are not dealt with consistently with Deaton and Zaidi, whose recommendation is to “calculate an annual rental equivalent using an appropriate real rate of interest and median depreciation values for each item calculated across all households owning that item.” (p. 39). Finally, housing. Here the consumption aggregate of Statistics Botswana does not include imputed rent for owners, while Deaton and Zaidi argue that “if the dwelling is owned by the household or received free of charge, an estimate of the annual rental equivalent must be included in the consumption aggregate” (p. 39). We found no empirical reason that would advise against this choice, that is rental equivalents do not seem inaccurate. 2.4 Aggregation The nominal consumption aggregate is obtained by summing up the four building blocks constructed in sections 2.1-2.3. We calculate the nominal per capita expenditure (PCE) by dividing the nominal consumption aggregate by the total number of household members. The nominal PCE is expressed in Pula per month. We check the cumulative distribution functions (CDFs) of the nominal PCE. The CDF F(x), for any given expenditure level x, gives the proportion of people who have expenditures below that level. If the expenditure level is taken to be the poverty line z, then F(z) would give the proportion of people who have expenditures below z, i.e. the proportion of poor people (the incidence of poverty). A second reason that makes CDFs an attractive tool is their use in dominance analysis. Suppose there are two different income distributions, A and BOTSWANA POVERTY ASSESSMENT 75 B, for instance two different years, and suppose that FA(x) < FB(x) for all values of x; if this is the case, then we say that the distribution A first-order-stochastically dominates (FOD) the distribution B. This implies that the distribution A’s CDF lies everywhere below the distribution B’s CDF, and therefore the incidence of poverty is always higher in B than in A, no matter where the poverty line is drawn (Atkinson 1987; Ravallion 1994). Figure A8.3 shows the cumulative distribution functions (CDF) of nominal PCE, separately by stratum. The curves depict a ranking consistent with the descriptive statistics published in Table 3: the CDF for cities and towns first-order stochastically dominates the CDF for urban areas, which in turn first-order stochastically dominates the CDF for rural areas. This implies that 1) the incidence of poverty will always be higher in urban areas than in in cities and town, and 2) the incidence of poverty will always be higher in rural than in urban areas. FIGURE A8.3  Cumulative Distribution Functions of Nominal PCE, by stratum 1 0.8 Population (%) 0.6 Botswana 0.4 Cities/towns Urban Rural 0.2 0 0 2,000 4,000 6,000 8,000 10,000 12,000 14,000 16,000 PCE (pula/person/month) Source: our estimates. 3. From nominal to real consumption aggregate: spatial price variation Geographical differences in price levels are of major concern for welfare comparisons. A higher level of the cost of living clearly decreases the real purchasing power of a given level of expenditure, thereby decreasing household welfare. The nominal consumption aggregate constructed in section 2 needs to be adjusted for spatial price differences. In this section we estimate a survey-based household-level spatial price index. The official estimates of poverty used by Statistics Botswana already account for spatial differences in the cost of living by using the Poverty Datum Lines (PDLs). The PDLs are household-specific poverty lines which embed an adjustment for differences in the purchasing power across Botswana. The fact that a spatial deflator is implicitly built-in the official PDL (World Bank 2015: 174) can be easily seen by noting that, for instance, the ratio between a regional poverty line and the national level poverty line can be interpreted as an implicit spatial “true” cost-of-living index (Deaton and Muellbauer, 1980). This is because the regional level poverty lines, keeping constant the standard of living, take into account the differences in the households’ consumption pattern and the local market prices structure. A simple estimate of the spatial cost of living index for the region r is given by: 76 BOTSWANA POVERTY ASSESSMENT where is the poverty line for region r and year t and is the average national poverty line. Table A8.5 (col. 3 and 4) shows the estimate for the regional cost of living differences obtained applying the equation above. TABLE A8.5  Implicit spatial price indices, 2009/10-2015/16 PLrt1 PLrt2 SPI SPI (pula/hh/month) (pula/hh/month) (BWA=100) (BWA=100) 2009 2016 2009 2016 Gaborone 1,068 987 81 78 Francistown 1,133 1,363 86 108 Other Cities & Towns 1,094 1,157 83 92 Rural South-East 1,296 1,153 98 91 Rural North-East 1,428 1,116 108 88 Rural North-West 1,479 1,531 112 121 Rural South-West 1,388 1,534 105 121 Cities/Towns 1,091 1,160 82 84 Urban villages 1,532 1,216 116 118 Rural areas 1,269 1,352 96 98 Botswana 1,318 1,246 100 100 How to calculate scalar regional poverty lines consistent with the official estimates? The only way is through the following method, where we use the quantile function calculated at the official regional poverty rates: where is the inverse of the cumulative density function of the total households’ expenditure and is the official regional poverty headcount for the region r. Equation (2) was calculated separately and independently for 2009/10 and 2015/16, each set of estimates being based on the official household-level poverty lines (both at current prices), kindly shared by Statistics Botswana. The spatial price deflators in Table 6 (last column) are the ones implicitly used by Statistics Botswana to obtain official poverty measures in 2009/10 and 2015/16. The availability of spatial deflators allows us to adjust the nominal PCE calculated in section 2.4, and calculate the real consumption aggregate, that is a measure that accounts for the differences in the cost of living across the country, and is therefore comparable to official poverty estimates. The latter, produced using household- level PDLs, are implicitly adjusting for spatial differences in prices. Figure A8.4 shows that the adjustment for spatial differences in prices is not negligible but does not change the shape of the CDFs significantly. Figure A8.5 shows that the ranking in Figure A8.3 (cumulative distribution functions of nominal PCE) is preserved when using real PCE. It implies that based on real PCE: 1) the incidence of poverty will always be higher in urban areas than in in cities and town, and 2) the incidence of poverty will always be higher in rural than in urban areas. BOTSWANA POVERTY ASSESSMENT 77 FIGURE A8.4  Cumulative Distribution Functions of Nominal vs Real PCE 1 0.8 Population (%) 0.6 Real 0.4 Nominal 0.2 0 0 2,000 4,000 6,000 8,000 10,000 12,000 14,000 16,000 PCE (pula/person/month) Source: our estimates. FIGURE A8.5  Cumulative Distribution Functions of Real PCE, by stratum 1 0.8 Population (%) 0.6 Botswana 0.4 Cities/towns Urban Rural 0.2 0 0 2,000 4,000 6,000 8,000 10,000 12,000 14,000 PCE (spi-adjusted pula/person/month) Source: our estimates. 4. A harmonized poverty line After constructing a consumption aggregate for 2015/16 comparable to that used in 2009/10, this subsection defines a poverty line consistent with the one used in the past. The difficulty here is with the fact that official poverty lines in Botswana are calculated at the household level. We find 3,803 unique poverty lines for the 2009/10 survey and 3,403 poverty lines for 2015/16. The official poverty line is referred to as the “poverty datum line” (PDL). For household h in region r it is defined as follows: 11 7 5 (A1) 78 BOTSWANA POVERTY ASSESSMENT The first component is the minimum food expenditure, calculated as the product of , a vector of food commodities that identifies a “basic need” food basket, (in our notation p identifies age-gender groups), and , the corresponding minimum prices vector. The term denotes the number of members in household h belonging to the age-gender class p. The second and third components (clothing and personal items) are defined in a similar way. The last two components of the PDL are the cost of household’s goods in region r (Hr ), and the cost of housing for household type u in region r ( ). The strategy used in this report consists in anchoring poverty measurement to the standard of living identified by the 2009/10 PDLs. Analytically, this amounts to identifying a (scalar) poverty line consistent with the official poverty rate in 2009/10, and then adjusting its purchasing value for inflation.6 By so doing, we obtain a new poverty line, expressed in 2015/16 Pula, that we can use in combination with the 2015/16 consumption aggregate constructed in section 3. We shall refer to this line as to the harmonized poverty line. A scalar poverty line for 2009/10 is obtained by inverting the 2009/10 quantile function at the percentile corresponding to the official poverty rate (0.193, according to Table 1 in World Bank, 2018: 4). Thus: If we use 233.5 Pula/month/person on the 2009/10 data, we obtain an estimated headcount poverty rate equal to 19.3 percent, that is, we obtain the official estimate of the incidence of poverty in 2009/10. The task to update this poverty line to 2015/16 is conceptually simple. In order to anchor the new poverty line to the same (minimum) standard of living set for 2009/10, one needs to adjust the monetary value of the latter (that is, the 2009/10 poverty line) for changes in purchasing power (that is, for inflation). This method – the old line updated for inflation over the period using the best available price index – is to be preferred to the alternative of re-estimating a new line. The reason is that “repeating the calculations used in the base period [2009/10, in our context] may well introduce some differences in the real value of the poverty line associated with shifts in the Engel curve, such as shifts due to changes in relative prices or tastes” (Ravallion, 2016: 203). In practice, we need a (monthly) time series for the consumer price index (CPI) between 2009/10 and 2015/16. This is what we need to adjust for inflation the 2009/10 poverty line. Statistics Botswana provided us with official estimates of the CPI. Given the importance of the CPI in the rest of the analysis, it is worth exploring its dynamics. Figure A8.6 shows the monthly CPI between 2009 and 2016, comprising all survey months, separately by strata. Overall, the dynamics of inflation is similar in all strata. According to the official CPI, the cumulative inflation adjustment factor between April 2009 and October 2016 equals 1.433. The corresponding factors for cities/towns, urban and rural areas equal 1.442, 1.426, and 1.425, respectively. In our context, we decided to take advantage of the new information available from the 2015/16 to rebase the official CPI using the new budget shares from the 2015/16 survey, as a way of better capturing the changes in the consumption pattern between 2009/10 and 2015/16. In Table A8.6 we compare the budget shares underlying the official CPI (column 2) with the budget shares calculated on the 2015/16 BMTHS (column 3), separately by major categories of goods and services. 6  As argued in World Bank (2015: Annex B, p. 179), the only way to replicate the official estimates using aggregated poverty lines is that of imposing ex-ante, i.e. by construction, the consistency between official poverty estimates and the aggregated poverty lines. BOTSWANA POVERTY ASSESSMENT 79 FIGURE A8.6  Official CPI, by stratum 100 90 Botswana 80 Cities/towns Urban Rural 70 2009m1 2010m7 2012m1 2013m7 2015m1 2016m7 date Source: our elaboration on official estimates. TABLE A8.6  Budget shares for the consumer price indices Categories Official CPI Based on 2015/16 BMTHS Food 0.15269 0.22375 Alcohol and Tobacco 0.07820 0.03970 Clothing and Footwear 0.06637 0.05240 Housing Costs 0.04559 0.18786 H/hold Goods and Services 0.01348 0.04297 Medical/Health care 0.00796 0.02931 Transport 0.07041 0.19259 Communication 0.25230 0.06107 Recreation and Culture 0.07242 0.02590 Education 0.03901 0.03951 Restaurants and Hotels 0.02283 0.03364 Miscellaneous 0.17872 0.07131 Total 1.00000 1.00000 According to the official CPI, the cumulative inflation adjustment factor between April 2009 and October 2016 equals 1.433. By applying the adjustment for inflation to the poverty line: we obtain 334.6 (=233.51.433). The poverty line used in the rest of this report is therefore 334.6 Pula/person/month, which is our best estimate of the amount required in 2015/16 for achieving the same standard of living set by the official PDL for 2009/10. This way of calculating the poverty line ensures the consistency of poverty comparisons (Ravallion, 2016), and entirely relies on Statistics Botswana official methodology. 80 BOTSWANA POVERTY ASSESSMENT 5. Poverty estimates for 2015/16 The main results are reported in Table A8.7, based on the real consumption aggregate discussed in section 3 and the harmonized poverty line discussed in section 4. The first column reports the point estimates for each poverty measure; the second column is the estimated standard error, after accounting for the complex survey design; the last two columns report the lower and upper bound of the estimates using a 95 percent confidence interval. A clear picture is confirmed: poverty level, poverty gap, and poverty severity are consistently highest in rural areas. TABLE A8.7  Poverty estimates in 2015/16, by stratum Incidence of poverty (headcount poverty index) Stratum Estimate Std. error lower b. upper b. Cities/Towns 3.3 0.7 2.0 4.7 Urban Villages 13.7 1.4 10.9 16.4 Rural areas 26.8 2.1 22.7 30.9 Botswana 16.1 1.0 14.1 18.0 Depth of poverty (poverty gap index) Estimate Std. error lower b. upper b. Cities/Towns 1.3 0.3 0.7 1.9 Urban Villages 3.4 0.5 2.5 4.4 Rural areas 8.0 0.9 6.3 9.7 Botswana 4.6 0.4 3.8 5.3 Severity of poverty (poverty gap squared index) Estimate Std. error lower b. upper b. Cities/Towns 0.7 0.2 0.3 1.1 Urban Villages 1.3 0.2 0.8 1.8 Rural areas 3.4 0.5 2.5 4.3 Botswana 1.9 0.2 1.5 2.3 6. Inequality estimates Table A8.8 shows selected estimates of inequality based on the 2015/16 BMTHS using the expenditure-based consumption aggregate both in real terms (that is adjusted for spatial variation of prices) and in nominal terms. Overall, the estimates are robust and spatial deflation has a modest impact on the estimates. A formal t-test was carried out to test whether the difference between the Gini indices in 2009/10 and 2015/16 equals 0, and the null hypothesis was not rejected. Statistically, the spatial deflation has no significant impact on the Gini index. Similar results apply to other inequality measures. BOTSWANA POVERTY ASSESSMENT 81 TABLE A8.8  Inequality estimates for 2015/16 nominal std. 95% confidence std. 95% confidence real CA CA error interval error interval Gini 53.3 1.1 51.2 55.4 54.9 1.1 52.8 57.1 Cities/Towns 47.1 1.1 44.9 49.2 48.7 1.2 46.4 51.1 Urban Villages 50.6 1.8 47.1 54.0 51.1 1.8 47.5 54.6 Rural areas 52.6 3.1 46.5 58.7 53.3 3.2 47.0 59.7 Generalized Entropy Indices MLD 49.7 2.2 45.3 54.1 53.2 2.4 48.5 58.0 Theil 59.9 6.1 47.9 72.0 63.6 6.2 51.4 75.8 Atkinson index e=1 39.2 1.4 36.5 41.9 41.3 1.4 38.5 44.1 e=2 58.0 1.3 55.5 60.5 60.5 1.3 57.8 63.1 e=3 68.6 1.2 66.2 71.1 71.0 1.2 68.5 73.4 Percentile and share ratios p90/p10 10.6 0.5 9.7 11.5 11.5 0.6 10.3 12.7 p90/p50 4.0 0.1 3.7 4.2 4.1 0.2 3.8 4.4 s80/s20 8.3 0.4 7.5 9.1 9.1 0.5 8.1 10.0 Source: our estimates. 82 BOTSWANA POVERTY ASSESSMENT APPENDIX 9 BOTSWANA SWIFT POVERTY PROJECTIONS FOR 2019-20227 Like most countries in Sub-Saharan Africa, Botswana does not conduct frequent surveys to track poverty trends. In fact, in the last 20 years, only three such surveys were conducted: the 2002/03 Household Income and Expenditure Survey (HIES), the 2009/10 Botswana Core Welfare Indicators Survey (BCWIS), and the 2015/16 Botswana Multi-Topic Household Survey (BMTHS).8 However, in 2019 Botswana began to field the Quarterly Multi-Topic Survey (QMTS), a labor force survey with additional modules, which had been collected six times by 2022.9 On its own, the QMTS cannot provide information on monetary poverty and inequality. To help fill the poverty data gaps, the World Bank and Statistics Botswana applied survey-to-survey imputations to the QMTS, using a methodology called Survey of Well-being via Instant and Frequent Tracking (SWIFT).10 The SWIFT methodology combines machine learning and multiple imputations techniques to project household expenditure and impute poverty statistics in surveys without such data. The BMTHS 2015/16, a year-long survey with data for all four quarters, was used in its entirety as the training data for the SWIFT models. The variable set used for the modeling was limited to what could be harmonized between the BMTHS 2015/16 and the six QMTS datasets, such as household demographics, dwelling characteristics, sources of household income, education, and labor market information. Overall, four types of models or methodologies were used to project poverty up to 2022: 1. Limited model (quarterly). The limited model applied the SWIFT survey-to-survey imputation methodology, using the base year BMTHS 2015/16 survey, to all six available QMTS datasets. The name of this model reflects the limited set of variables common to all quarters of the QMTS and the 2015/16 BMTHS, such as demographics, education, and labor market variables. 2. Full model (quarterly). The QMTS for the third and fourth quarters of 2019, the first quarter of 2020, and the fourth quarter of 2022 collected two additional modules with questions on household dwelling conditions and sources of household income, which could also be harmonized to the BMTHS 2015/16 survey. The name of the model reflects this larger set of variables, which strengthened the projection model. However, it could only be applied to four quarters. 3. Full+FIES model (quarterly). The QMTS for the fourth quarter of 2022 collected the same information as the 2019 surveys but added a set of food security questions to estimate Food Insecurity Experience Scale (FIES) indicators. These variables represent quickly changing poverty correlates that usually strengthen poverty projection models. The results, however, were not very different. 4. Elasticity (annual). A simple, non-SWIFT methodology was also used to estimate annual poverty projections for 2017–22 using a growth-poverty elasticity. The elasticity estimate is based on the official poverty estimates and GDP growth data for 2009 and 2016 and is applied to later years using actual or projected GDP data, assuming a passthrough rate of 0.87. 7  The results presented in this appendix are based on a collaboration between the World Bank and Statistics Botswana. The World Bank team included Carolina Diaz-Bonilla and Danielle Aron, with inputs from Nobuo Yoshida. For details on the SWIFT guidelines see: Yoshida, N., R. Munoz, A. Skinner, C. Kyung-eun Lee, M. Brataj, and D. Sharma. 2015. SWIFT Data Collection Guidelines version 2. The World Bank. 8  The dates of these surveys were as follows: 2002/03 HIES: June 2002–August 2003 (Central Statistics Office 2004), 2009/10 BCWIS: April 2009–March 2010 (Statistics Botswana 2013), and the 2015/16 BMTHS: November 2015–December 2016 (Statistics Botswana 2018). As noted, for ease of reference, the rest of document uses the year in which a survey covered the most months as its date; hence, the 2003 HIES, the 2009 BCWIS, and the 2016 BMTHS. 9  The six QMTS dates were Quarters 3 and 4 of 2019, Quarters 1 and 4 of 2020, Quarter 4 of 2021, and Quarter 4 of 2022 (Statistics Bo- tswana 2019a, 2019b; 2020a, 2020b; 2021; 2022c). 10  For background on SWIFT, see Yoshida and others (2015). BOTSWANA POVERTY ASSESSMENT 83 All the models show a similar slowdown in poverty reduction since 2016. 1. What is SWIFT? Survey of Well-being via Instant and Frequent Tracking The SWIFT program was created in 2014 to produce poverty statistics in a cost-effective and timely manner. The SWIFT methodology uses multiple imputation and machine learning techniques to train poverty projection models and produce poverty rate estimates. 2. How does SWIFT work? A model to estimate household expenditure/income is created from dataset A, which contains both expenditure/ income data AND poverty correlate variables. Dataset A is referred to as the baseline or training dataset. The model is applied to dataset B, referred to as the imputation dataset, which contains ONLY poverty correlate data, to estimate household expenditure/income. Estimated household expenditure/income is compared to the poverty line to get estimated poverty rate. 3. SWIFT MODELING & VARIABLE SELECTION SWIFT models assume a linear relationship between household expenditure/income and poverty correlates, with an error term: where ( ) are drawn from OLS regression coefficients and error terms A Cross-Validation exercise is run ensure that the SWIFT model can perform well outside the training dataset, preventing the “over-fitting” problem. A stepwise OLS regression is applied to all available SWIFT variables and produces a model containing only significant variables. Multiple Imputation is used to test how well the model can project household expenditure/income within the training dataset. 4. Variables available for selection for Botswana SWIFT model The variable set used for the modeling was limited to what could be harmonized between BMTHS 2015/16 and QMTS, taking into consideration question phrasing, respondent answer options, and recall periods. This included household demographics, dwelling characteristics, sources of household income, and education and employment information. However, the harmonized variable set was further limited when considering QMTS 2020 Q4 and QMTS 2021 Q4. These two-quarter surveys did not collect information on household dwelling conditions or sources of household income. Additionally, food security information comparable to BMTHS 2015/16 was collected in 2022 Q4. 5. Botswana’s SWIFT model and results Due to the available data and the differences in poverty levels in different areas of Botswana, nine different SWIFT models were created for this analysis. Three models were created for each region: cities and towns, urban villages, and rural villages. Full, limited, and full + FIESmodels were created for each region. The full model utilizes the larger variable set that was harmonized between BMTHS 2015/16 and QMTS, including dwelling conditions and sources of household income. The limited model utilizes the smaller harmonized variable set, excluding dwelling conditions and sources of household income, so that poverty projections could be produced for QMTS 2020 Q4 and QMTS 2021 Q4. And finally, the full + FIES model utilizes the full variable set with the addition of a food security variable; the full + FIES model is only used to produce estimates for 2022 Q4. For quarterly monitoring of poverty, it would be more appropriate to incorporate fast-changing poverty correlates, such as food and non-food consumption or food security indicators, into the models. However, because QMTS is focused primarily on labor conditions, there are no available consumption variables to 84 BOTSWANA POVERTY ASSESSMENT harmonize. For food security, the reference period on the FIES questions varies between BMTHS 2015/16 and QMTS 2019 Q3 – 2022 Q4, making the data not comparable. However, after feedback and cooperation with Statistics Botswana, a comparable reference period on the food security questions was added to QMTS 2022 Q4, allowing for creating new models to include this data. For QMTS 2019 Q3 through 2020 Q1, where both the full and limited models could be applied, the imputed poverty rates are very similar, differing by less than two percentage points and most differing by less than one percentage point. The full + FIES models also show near identical results to the regular full model in QMTS 2022 Q4. This supports the accuracy of both the limited and full models, but it should still be noted that the full + FIES model still has a significantly lower proportion of fast-changing poverty correlates (the food security indicator) compared to slow-changing poverty correlates. Results show that poverty at the national level has remained relatively stagnant, with a slight increase in 2021 Q4 and then a decrease in 2022 Q4 (Figure A9.1). The trend in urban areas closely follows that of the national level. In cities and towns, the poverty rate has remained the same over the entire period, while in rural areas, we see fluctuation. Rural areas saw a decline in poverty in 2020 Q1, an increase through 2021 Q4, and another decrease in 2022 Q4. FIGURE A9.1 FIGURE A9.1  Poverty Projections for Botswana 30% 25% 26.90% 27.08% 27.38% 26.81% 24.96% 25.44% 24.34% 20% 18.04% 17.05% 17.01% 16.65% 16.05% 15.84% 14.47% 15% 15.63% 13.80% 13.79% 14.30% 13.66% 13.32% 10% 11.41% 5.15% 4.68% 5.25% 5.07% 4.70% 4.40% 5% 3.21% 0% BMTS 2015/16 QMTS 2019 Q3 QMTS 2019 Q4 QMTS 2020 Q1 QMTS 2020 Q4 QMTS 2021 Q4 QMTS 2022 Q4 NAT FULL NAT LIM NAT FIES CT FULL CT LIM CT FIES UV FULL UV LIM UV FIES RV FULL RV LIM RV FIES Source: Prepared by authors using BMTHS 2015/16 and QMTS 2019-Q3, 2019-Q4, 2020-Q1, 2020-Q4, 2021-Q4, 2022-Q4. Note: The data labels in the figure are only for the limited models. This work is one of the first attempts in Africa to monitor official, comparable poverty statistics quarterly. Furthermore, the SWIFT team at the World Bank has worked closely with Statistics Botswana during this project, paving the way for future efforts to improve the models and projections. As a result of this collaboration, Statistics Botswana has already made some amendments to the QMTS (starting for 2022 Q4) that will allow for comparability on some indicators of household food security. Including these types of fast-changing indicators will allow the models to be better suited to capturing more sudden changes in economic conditions. Further efforts like this will be important steps towards routinely producing reliable poverty estimates for Botswana.