Poverty and Vulnerability in the Ethiopian Lowlands: Building a More Resilient Future The World Bank Group and the UK’s Department for International Development 2019 POVERTY AND VULNERABILITY IN THE ETHIOPIAN LOWLANDS BUILDING A MORE RESILIENT FUTURE The World Bank Group and the UK’s Department for International Development 2019 Poverty and Vulnerability in the Ethiopian Lowlands Contents Abbreviations and Acronyms xiii Acknowledgments xvii Overview xviii Introduction xviii Pastoral Livelihoods in the Lowlands xix Changes in Monetary and Nonmonetary Poverty xx Vulnerability xxv Sources of Vulnerability xxvi Changes in Pastoral Livelihoods over the Long Term xxxiii Promoting Resilience in the Lowlands: The Components of a Broader Development Policy Package xxxiv PART 1: LIVELIHOODS, POVERTY, AND VULNERABILITY IN THE LOWLANDS 1 Chapter 1: Pastoral Livelihoods 2 Summary 2 Introduction 2 Pastoralism and Agropastoralism 2 Gender Roles in Livestock, Crop Production, and Marketing 4 Sources of Income 4 (1) Dependence on Livestock 5 (2) Variable, Semiarid and Drought-Prone Environment 10 (3) Remote Geographic Location 14 (4) Political and Economic Marginalization from the Ethiopian State 14 Chapter 2: Poverty and Vulnerability 15 Summary 15 Introduction 15 Expenditure-Based Poverty 16 Trends in Nonmonetary Poverty 18 Humanitarian Food Aid (HFA) and Poverty in the Lowlands 19 Poverty by Main Source of Household Income (livelihood) 22 Structural Factors behind the Decline in Expenditure-Based Poverty 23 Vulnerability to Poverty in the Lowlands 26 Comparing Vulnerability and Its Sources among Ethiopia’s Agroecological Zones 28 iii Comparing Vulnerability and Its Sources between Urban and Rural Areas of the Drought-Prone Lowlands and Pastoral Areas 29 Comparing Vulnerability and Its Sources among Different Livelihoods in the Rural Drought Prone Lowlands and Pastoral Areas 30 Conclusions 30 Chapter 3: Low Human Capital as a Source of Vulnerability 32 Summary 32 Introduction 32 Status of Human Development in Lowland Areas of Ethiopia 33 Education 34 Health 38 Health-Related Quality Indicators 42 Sociocultural and Economic Constraints in Health and Education 43 Education and Health Services in the Face of Shocks 43 HD Delivery Systems’ Analysis 44 Education System Analysis 49 Health System Analysis 51 Conclusions 52 Chapter 4: Climate Change and Environment as Sources of Vulnerability 54 Summary 54 Introduction 54 Rainfall 55 Temperature 56 Growing Season 58 Climate Change 60 Droughts 61 Vegetation 62 Water Availability 64 Chapter 5: Changing Land Use and Land Cover 69 Summary 69 Introduction 69 Drivers of Land Use Changes 73 Chapter 6: Regional Governance 85 Summary 85 Introduction 85 Regional Governance 86 Politics of Regional Governance 86 Lowland Contexts 86 Conclusions 90 PART 2: PROMOTING RESILIENCE AND DEVELOPMENT IN THE LOWLANDS 91 Chapter 7: The Role of Public Finances 92 Summary 92 Introduction 92 Sources of Funds 92 Expenditures 98 Conclusions 108 iv  Contents Chapter 8: Humanitarian Interventions and Safety Nets Working in Tandem 109 Summary 109 Introduction 109 Humanitarian Intervention 110 Impact and Performance of the PSNP 115 Impact and Performance of the HFA 117 Necessary and Essential Measures to Transform the Current Approach 117 Additional Measures to Respond to the Causes of Poverty and Vulnerability 119 Chapter 9: Economic Development 121 Summary 121 Agriculture Overview 121 Pastoralism and Livestock 123 Camels 124 Dairy 124 Agroforestry 124 Spices 125 Extractive Sector 125 Tourism 126 Key Drivers of Growth 127 Governance and Stability 127 Infrastructure 127 Market Development 128 Access to Finance 128 Urbanization 129 Growth Corridors 130 Cross-Border Trade 132 Industrial Parks (IP) 132 Enabling Social Inclusion 133 Conclusions 134 References 136 Annex 1: Environment and Land Use Tables 140 Annex 2: Methodology for Measuring Household Vulnerability 152 List of Tables Table E.1: Mean livestock holdings, by livestock holding quintile, region, and year in tropical livestock units xxxiii Table E.2: Mean share of total household income (%) (ILRI survey data) xxxiv Table 1.1: Rural population by livelihood zone (2007 Census) 3 Table 1.2: Livelihoods of households (%) 5 Table 1.3: Mean share of total household income (%) (ILRI survey data) 5 Contents  v Table 1.4: Market engagement by wealth group, northern Kenya and Southern Ethiopia, 2000–2002 6 Table 1.5: Contribution of livestock to annual household income in pastoralist areas, Shinile zone 7 Table 1.6: Access to livestock asset in TLU by region 7 Table 1.7: Trends in household livestock ownership in Borana and Karrayu 7 Table 1.8: Mean livestock holdings, by livestock holding quintile, region, and year in tropical livestock units 8 Table 1.9: Changes in wealth ranking of Somali pastoral households according to livestock holding over a 60-year period (1944–2004) 8 Table 1.10: Changes in household livestock composition in TLU by livelihood type for Somali region (%), 2006 to 2017 12 Table 2.1: Proportion of drought affected woredas by region, March 2016 20 Table 2.2: Changes in terms of trade between livestock and cereals comparing January 2017 with January 2014, by woreda hot spot area, percent change 26 Table 3.1: Ever attended school (schooling experience) 35 Table 3.2: Average class size and availability of textbooks 37 Table 3.3: Average distance from school in km 38 Table 3.4: Place where medical assistance was received, by ecological zone and wealth strata 41 Table 3.5: Sector specific expenditures assignments/service delivery functions at different tiers of government, district level decentralization program 44 Table 3.6: Sectoral allocation of the MDG transfers in EFY12 by region 46 Table 3.7: Percentage of woreda level operational budget share in 2009 and 2010 EC from the total operational budget of the region, February 2019 47 Table 5.1: Lowland grazing and forage (1986 and 2016) 70 Table 5.2: Lowland cropland (1986, 2000, and 2016) 72 Table 5.3: Population of selected urban centers in Somali region, Ethiopia (based on CSA census, and 2015 projected) 74 Table 5.4: Lowland settlement (1986 and 2016) 75 Table 5.5: Large farms and concessions by region 78 Table 5.6: Prosopis expansion in Afar 2000–2016/17 80 Table 5.7: Total number of IDPs by cause and region 84 Table 5.8: Percent of IDPs by cause and region 84 Table 7.1: Per capita transfers to lowland (pastoral) zones in Oromia and SNNP in ETB, FY11–16 95 Table 7.2: Regional share of federal block grant allocation 96 Table 7.3: Regional share of MDG/SDG Transfers 96 Table 7.4: Sectoral allocation of the MDG Transfers in FY12 by region 97 Table 7.5: Rate of disbursement for the MDG/SDG Transfers 98 vi  Contents Table 7.6: Share of capital spending to total regional expenditures, in percent 100 Table 7.7: Sectoral disaggregation of regional expenditures during FY12–18, percent of total expenditures 100 Table 7.8: Budget outturn—recurrent 101 Table 7.9: Budget outturn—capital 101 Table 9.1: Structure of agriculture—GDP and employment 2011 121 Table 9.2: Major exploration projects in lowland areas of Ethiopia 126 Table A.1: Area by region, lowland/highland and lowland AEZ (km2) 140 Table A.2: Average annual rainfall by region, lowland/highland, and lowland AEZ (mm/year) 141 Table A.3: Rainfall trend by region, lowland/highland, and lowland AEZ (mm/year) 142 Table A.4: Key short-term agro-meteorological trends by seed location 143 Table A.5: Average NDVI by region, lowland AEZ, and lowland/highland (index) 144 Table A.6: NDVI trend by region, lowland/highland, and lowland AEZ (%) 145 Table A.7: Average EVI by region, lowland AEZ, and lowland/highland (index) 146 Table A.8: EVI trend by region, lowland/highland, and lowland AEZ (%) 147 Table A.9: Land use and land cover (LULC) classes and definitions 148 Table A.10: Summary of LULC by highland/lowland (1986 and 2016) 148 Table A.11: Lowland LULC (1986, 2000 and 2016) 148 Table A.12: Lowland forest (1986 and 2016) 149 Table A.13: Lowland cropland (1986, 2000 and 2016) 149 Table A.14: Lowland grazing and forage (1986 and 2016) 149 Table A.15: Lowland woodland (1986 and 2016) 149 Table A.16: Lowland shrub and bush (1986 and 2016) 150 Table A.17: Lowland grassland (1986 and 2016) 150 Table A.18: Lowland bare land (1986 and 2016) 150 Table A.19: Lowland swamp and marsh (1986 and 2016) 150 Table A.20: Lowland waterbody (1986 and 2016) 150 Table A.21: Lowland settlement (1986 and 2016) 150 Table A.22: EHAIA land concession details 151 Table A.23: Large farms and concessions by region 151 Table A.24: Prosopis expansion in Afar 2000–2016/17 151 List of Figures Figure E.1: Overall decline in headcount poverty by region 2000–2016 (HCE surveys) xxi Figure E.2: Headcount poverty rate by agroecological zone: 2011 vs. 2016 (HCES) xxii Figure E.3: Multidimensional poverty index by region: 2011 vs. 2016 xxii Contents  vii Figure E.4: The prevalence of stunting among children 0–5 years of age by region: 2011 vs. 2016 xxiii Figure E.5: Trend in the poverty headcount rate without aid/donations: 2011 vs. 2016 xxiii Figure E.6: Changes in the percentage of population in different livelihoods—drought-prone lowlands and pastoral areas in 2011 and 2016 xxiv Figure E.7: Poverty rate (headcount) by livelihood—drought-prone lowlands and pastoral areas in 2011 and 2016 xxiv Figure E.8: Poverty gap and severity of poverty by household livelihood—drought-prone lowlands and pastoral areas in 2011 and 2016 xxv Figure E.9: The prevalence of poverty and vulnerability—lowlands vs. other agroecological zones in 2016 xxvi Figure E.10: Sources of vulnerability in 2016—poverty induced vs. risk induced xxvii Figure E.11: Distribution of livestock wealth in TLU by household quartile and by region (based on ILRI survey data 2016) xxxiv Figure E.12: Expenditures on rural development and water by region, FY12–18 xxxvi Figure E.13: Expenditures on education and health by region, FY12–18 xxxvi Figure E.14: Humanitarian requirements, 2008–18 xxxvii Figure E.15: PSNP caseloads in Afar and Somali regional states: 2008–2017 xxxvii Figure 1.1: Distribution of livestock wealth by household quartile and by region (based on ILRI survey data 2015–17) 10 Figure 1.2: Average TLU per household by region (Ethiopia) 13 Figure 2.1: Overall decline in headcount poverty by region, 2000–2016 (HCE surveys) 17 Figure 2.2: Headcount poverty rate by agroecological zone: 2011 vs. 2016 (HCES) 17 Figure 2.3: Multidimensional poverty index by region: 2011 vs. 2016 18 Figure 2.4: The prevalence of stunting among children 0–5 years of age by region: 2011 vs. 2016 19 Figure 2.5: Poverty headcount with and without aid/donations in 2016 20 Figure 2.6: Poverty rate (headcount) by region—with and without aid/donations in 2016 21 Figure 2.7: Trend in the poverty headcount rate without aid/donations: 2011 vs. 2016 21 Figure 2.8: Changes in the percentage of population in different livelihoods—DPL&P areas in 2011 and 2016 22 Figure 2.9: Poverty rate (headcount) by household livelihood— DPL&P areas in 2011 and 2016 23 viii  Contents Figure 2.10: Poverty gap and severity of poverty by household livelihood—DPL&P areas in 2011 and 2016 23 Figure 2.11: Changes in expenditure-based and asset-based poverty—DPL&P areas in 2011 and 2016 24 Figure 2.12: Changes in expenditure-based and asset-based poverty by livelihood—DPL&P areas in 2011 and 2016 25 Figure 2.13: Vulnerability to poverty characterized by the mean and variance of welfare 27 Figure 2.14: The prevalence of poverty and vulnerability—lowlands vs. other agroecological zones in 2016 28 Figure 2.15: Sources of vulnerability in 2016—poverty induced vs. risk induced 29 Figure 2.16: Sources of vulnerability in 2016: idiosyncratic vs. covariate shocks 29 Figure 2.17: The prevalence of poverty and vulnerability in the DPL&P areas in 2016—urban vs. rural 30 Figure 2.18: Sources of vulnerability in urban and rural areas of the lowlands in 2016—poverty induced vs. risk induced 30 Figure 2.19: Sources of vulnerability in urban and rural areas of the lowlands in 2016—idiosyncratic vs. covariate 30 Figure 2.20: The prevalence of poverty and vulnerability in the rural lowlands in 2016 31 Figure 2.21: Sources of vulnerability by livelihood in the rural lowlands in 2016—poverty induced vs. risk induced 31 Figure 3.1: School attendance by region, gender, and highland/lowland 35 Figure 3.2: Schooling experience and literacy by gender and ecological zones 36 Figure 3.3: Problems faced in school by ecological zones 37 Figure 3.4: Households who sought health services by gender and type of residence (highland and lowland) 38 Figure 3.5: Type of problems faced at health facilities by location of residence (highland/lowland) 39 Figure 3.6: Birth delivery at a facility by location of residence 39 Figure 3.7: Mortality rate and fertility rate by location of residence 40 Figure 3.8: Incidence of health problems by ecological zones and wealth quintiles 40 Figure 3.9: Wasting prevalence among children under five 42 Figure 3.10: Distance to health facility in km by ecological zones 42 Figure 3.11: Sectoral disaggregation of regional expenditures during FY12–18, percent of total expenditures 46 Figure 3.12: Per capita out-of-pocket expenditures and facility by-pass rate by region 48 Figure 3.13: Percent of civil servants with a university degree, by region 48 Figure 4.1: Lowland (< 1,500 m elevation) mean annual temperature and trend 59 Contents  ix Figure 4.2: Prolonged recent droughts recorded by pastoralists 63 Figure 5.1: Lowland climate migrants, 2015–2050 (shaded areas represent the 95 percent confidence intervals) 77 Figure 7.1: Percent of own revenue to total budget, FY12–18 93 Figure 7.2: Per capita own revenue collection in ETB, FY12–18 93 Figure 7.3: Federal block grant allocation to regional governments 94 Figure 7.4: Per capita block grant transfers in ETB, FY12–18 95 Figure 7.5: MDG Transfers allocation and disbursements, million ETB 97 Figure 7.6: Trends in regional expenditures 98 Figure 7.7: Per capita regional expenditures in ETB, FY12–18 99 Figure 7.8: Composition of regional expenditures, FY12–FY18 99 Figure 7.9: Regional expenditures in Afar, in million ETB 102 Figure 7.10: Sectoral breakdown of regional spending, FY08–17 102 Figure 7.11: Functional and sectoral disaggregation of regional spending in Afar, FY08–17 103 Figure 7.12: Regional expenditures in Somali region, FY 08–17 104 Figure 7.13: Regional expenditures in Benishangul-Gumuz region, FY08-17 105 Figure 7.14: Regional expenditures in Gambella region, FY08–17 106 Figure 7.15: Block grant allocation to pastoral and non-pastoral zones in SNNP, FY 12–17 106 Figure 7.16: Public expenditures in pastoral zones of SNNP, FY15–17 107 Figure 7.17: Block grant allocation to pastoral and non-pastoral zones in Oromia region, FY11–17 108 Figure 8.1: Humanitarian requirements, 2008–18 110 Figure 8.2: Humanitarian requirement by type of need, 2008–18 110 Figure 8.3: Number of people requiring food aid, 2008–18 111 Figure 8.4: Regional distribution of people requiring food aid (2008–18), as a proportion of total population 111 Figure 8.5: Food aid contributions, 2009–17 112 Figure 8.6: PSNP caseloads in Afar and Somali regional states (2008–2017) 114 Figure 8.7: HFA needs in Afar and Somali regional states (2008–2017) 114 Figure 8.8: Total needs in Afar and Somali regions (2008–2017) 114 Figure 9.1: Growth in crop yields and population across highland and lowland regions 122 Figure 9.2: Contribution of extractive sector exports 2013–35 (ETB M) scenarios 126 Figure 9.3: Sectors and drivers for economic growth in the lowlands 127 Figure 9.4: Access to a bank account 2016 129 Figure 9.5: Human capital and empowerment 133 x  Contents List of Maps Map E.1: Administrative regions with highland/lowland relief xix Map E.2: Regional water supply coverage xxviii Map E.3: Trend in annual rainfall 1982–2017 (mm/year) xxx Map E.4: Drought hot spots 12 xxx Map E.5: Internal displacement in the lowlands as result of conflict and extreme climate events (drought and floods) 2016–2018 (WB based on data from IOM) xxxi Map 2.1: Ethiopia’s four agroecological zones 16 Map 3.1: Agroecological zones map of Ethiopia 34 Map 4.1: Administrative regions with highland/lowland relief 55 Map 4.2: Ethiopia’s agroecological zones 56 Map 4.3: Trend in annual rainfall 1982–2017 (mm/year) 57 Map 4.4: Trend in annual rainfall 1982–1999 (mm/year) 57 Map 4.5: Trend in annual rainfall 2000–2017 (mm/year) 58 Map 4.6: Mean annual temperature (1986–2017) 59 Map 4.7: Distribution of seed locations for in-depth agroecologic trend analysis 60 Map 4.8: Seasonal rainfall in Ethiopia’s eastern and southern pastoral areas 61 Map 4.9: Drought hot spots 62 Map 4.10: Trend in NDVI 2000–2017 (%) 63 Map 4.11: Trend in NDVI 2000–2009 (%) 64 Map 4.12: Trend in NDVI 2010–2017 (%) 65 Map 4.13: Major river basins 66 Map 4.14: Regional water supply coverage distribution 67 Map 5.1: Distribution of lowland grazing and forage land (1986 to 2016) 70 Map 5.2: Distribution of lowland forest (1986 to 2016) 71 Map 5.3: Expansion of cropland 72 Map 5.4: Night-light data on urban settlement changes (ha) 2000–2017 74 Map 5.5: Population density in the lowlands in 2007 (CSA Census) 75 Map 5.6: Population growth 2007–17 lowland woredas 76 Map 5.7: Hot spots of climate in-migration and out-migration for 2050 77 Map 5.8: Land concessions and large farms 2016–2017 79 Map 5.9: Afar—large farm detail 2016–2017 79 Map 5.10: Afar—Prosopis juliflora extent (left) and intensity (right) in 2016/17 80 Map 5.11: Prosopis juliflora coverage 2016/17 81 Map 5.12: Frequency of conflict events per woreda, 2015–2018 82 Map 5.13: Number of deaths from conflict events per woreda, 2015–2018 83 Contents  xi Map 5.14: Internal displacement in the lowlands as result of conflict and extreme climate events (drought and floods) 2016–2018 (WB based on data from IOM) 84 Map 8.1: Frequency of hot spot woredas 2008–2017 111 Map 9.1: Models of agricultural production 122 Map 9.2: Spatial distribution of livestock 122 Map 9.3: Transport corridors for the Ethiopian lowlands in the IGAD region 131 xii  Contents Poverty and Vulnerability in the Ethiopian Lowlands Abbreviations and Acronyms ABE Alternative Basic Education ACLED Armed Conflict Location and Event Data project AE Adult equivalent AEZ Agro-ecological zones AfDB African Development Bank AFP Chatham House Africa Programme Briefing Papers ANC Ante-natal care ANDP Afar National Democratic Party ANOVA Analysis of Variance test ARRA Agency for Refugee and Returnee Affairs ASAL Arid and Semi Arid Lands AU African Union AWD Acute Waterborne Disease BG Benishangul-Gumuz BORESHA Building Opportunities for Resilience in the Horn of Africa Programme CAPE  Centre for Aid and Public Expenditure—Overseas Development Institute CARE Cooperative for Assistance and Relief Everywhere CBHI Community-based health insurance CGAP Consultative Group to Assist the Poor CHIRPS Climate Hazards Group InfraRed Precipitation with Stations data CMI Christiam Michelsen Institute COMESA Common Market for Eastern and Southern Countries CSA Central Statistical Agency DFID Department for International Development DHS Demographic and Health Survey DPL&P Drought-prone lowlands and pastoral DRS Developing Regional States EB Ethiopian Birr EC Ethiopian Calendar ECHO European Civil Protection and Humanitarian Aid Operations xiii EDC Economic development corridor EEU Economic Empowerment Unit EFY Ethiopian Fiscal Year EHAIA Ethiopian Horticultural and Investment Agency ELAP Ethiopia Land Administration Program EMIS Education Management Information System ENDF Ethiopian National Defense Forces ENSO El Niño–Southern Oscillation EPRDF Ethiopian Peoples’ Revolutionary Democratic Front ESDS Education Service Delivery Strategy EVI Enhanced Vegetation Index FBO Faith Based Organization FDI Foreign Direct Investment FMD Foot and Mouth Disease FMOH Federal Ministry of Health GCM Global Climate Model GDP Gross Domestic Product GEID General Education Inspection Directorate, Minsitry of Education GEQIP General Education Quality Improvement Project GFDRR Global Facility for Disaster Reduction and Recovery GoE Government of Ethiopia GPLM Gambella People’s Liberation Movement GTP II Growth and Transformation Plan II HAZ Climate Hazard Group HC Health centers HCE Household Consumption-Expenditure HCES Household Consumption-Expenditure Survey HCI Human Capital Index HD Human Development HEA Household economy analysis HEP Health Extension Program HEW Health Extension Worker HFA Humanitarian Food Assistance HICES Household Income, Consumptions and Expenditure Surveys HIV/AIDS Human immunodeficiency virus infection and acquired immune deficiency syndrome HLF High-level Forum HP Health posts HRD Humanitarian Requirement Document HRP Humanitarian Response Plan HSNP Household Safety Net Program HSTP Health Sector Transformation Plan ICP Incident Command Post ICRAF World Agroforestry Center ICT Information and Comunications Technology IDP Internally Displaced Person xiv   Abbreviations and Acronyms IDS Institute of Development Studies IEC Information, Education and Communications IGAD Intergovernmental authority on development IGC International Growth Centre IGFT Intergovernmental Fiscal Transfer ILRI International Livestock Research Institute IMF International Monetary Fund IMR Infant Mortality Rate IP Industrial Parks IPCC Intergovernmental panel on Climate Change IUCN International Union for Conservation of Nature LAPSSET Lamu Port and Lamu—Southern Sudan—Ethiopia Corridor LIFT Land Investment for Transformation LULC Land Use and Land Cover MDG Millenium Development Goals MFI Micro Finance Institutions MHNT Mobile Health and Nutrition Teams MNCH Maternal, Newborn and Child Health MNT Mobile Nutrition Teams MODIS Moderate Resolution Imaging Spectroradiometer MoE Ministry of Education MoFEC Ministry of Finance and Economic Cooperation MoH Ministry of Health MoWIE Ministry of Water, Irrigation and Electricity MPI Multi-Dimensional Poverty Index NDVI Normalized Difference Vegetation Index NEAEA National Education Assessment and Examinations Agency NER Net Enrollment Rate NGO Nongovernmental Organization NWoW New Way of Working OLF Oromo Liberation Front ONLF Ogaden National Liberation Front OOPS Out-of-pocket spending OPD Outpatient Services OPHI Oxford Poverty & Human Development Initiative OPM Oxford Policy Management OSSREA  Organisation for Social Science Research in Eastern and Southern Africa OWNP One WASH National Program PBF Performance Based Financing PCDP Pastoral Community Development Project PFM Public Financial Management PFSA Pharmaceuticals Fund and Supply Agency PHCU Primary Health Care Unit PNC Post-natal Care PRIME Pastoralist Areas Resilience Improvement and Market Expansion Acronyms  xv Abbreviations and  PSA Pharmaceuticals Supply Agency PSN Productive Safety Net PSNP Productive Safety Net Program PTA Parent Teacher Association RBH Regional Bureau of Health RDA Regional Development Association RIF Reproductive Maternal and Neonatal Health Innovation Fund RMNCH Reproductive Maternal and Neonatal Health and Child services RVF Rift Valley Fever SAM Severe acute malnutrition SBCC Social and behavioral change communication SDG Sustainable Development Goals SDI Spatial development initiative SLE Center for Rural Development SNNP Southern Nations, Nationalities and Peoples SNNPR Southern Nations, Nationalities and Peoples Region SOS SOS Sahel (an NGO) SPLA Sudan People’s Liberation Army SR Sub-Recipient SRS Somali Regional State SST Sea surface temperature SWReGAP Southwest Regional Gap Analysis Project TA Technical assistance TBA Traditional birth attendants TFR Total fertility rate THDRS Transform Health in Developing Regional States TLU Tropical Livestock Unit TVET Technical and Vocational Education Training U5MR Under-five Mortality Rate UN United Nations UNDP United Nations Development Program UNHCR United Nations High Commission for Refugees UNICEF United Nations Children’s Fund US United States USAID United States Agency for International Development USD United States Dollar VSLA Village Savings and Loans Associations WaSH Water, Sanitation and Hygiene WB World Bank WIDE Well-being and Ill-being Dynamics in Ethiopia WISP World Initiative for Sustainable Pastoralism WLRC Water and Land Resource Center, Addis Ababa University WMS Welfare Monitoring Survey YL Young Lives xvi   Abbreviations and Acronyms Poverty and Vulnerability in the Ethiopian Lowlands Acknowledgments This report was led jointly by Richard Hogg (Pro- Written comments at the concept note stage gram Leader, World Bank) and Emmanuel Skou- of this report were provided by: Professor Peter fias (Lead Economist, Poverty & Equity GP, World Little (Emory University), Simon Levine (Overseas Bank). The following individuals and institutions Development Institute), Margaret Grosh (Senior contributed substantively to this report (by chap- Adviser, World Bank), Dominick de Waal (Senior ter): Chapter 1: Dejene Negassa Debsu, and Asfaw Economist, World Bank), Ruth Hill (Lead Economist, Negassa. Chapter 2: Emmanuel Skoufias, Katja World Bank), Ladisy Chengula (Lead Agriculture Vinha, Berhe Mekonnen Beyene, Tom Bundervoet, Economist, World Bank), Robin Mearns (Practice Adrian Cullis, and Asfaw Negassa. Chapter 3: Manager, World Bank), and Malcolm Smart (Senior Victoria Monchuk, Berhe Mekonnen Beyene, Tom Economic Adviser, DFID). The final report benefit- Bundervoet, Anne Bakilana, Ayele Abebe, Belay ted from written comments provided by: Ruth Hill, Addise, Eshete Yilma Tefera, Fiseha Terefe, Fitsum Robin Mearns, Malcolm Smart, Fiona Flint (Senior Zewdu Mulugeta, Hiroshi Saeku, Mary Breen, Paul Scientist, ILRI), Daniel Clarke (Director, Center for Jacob Robyn, Roman Tesfaye, Roman Tesfaye Geb- Disaster Protection), and the UN team in Ethiopia. remedhin, and Ziyn Engdasew Woldab. Chapters 4 The team is grateful to the peer reviewers for their and 5 are based on contributions by Daniel C. Mon- constructive comments. chuk, Habab Taifour, Rebecca Lacroix, Benjamin The team is also thankful for the suggestions Petrini, and Manex Bule Yonis, aWhere, the Water provided throughout the process of generating and Land Resources Center of Addis Ababa Uni- this report by the following colleagues: Carolyn versity, and a background report prepared for the Turk (Country Director, World Bank), Nicole Klin- World Bank by the Center for International Earth gen (Country Program Coordinator, World Bank), Science Information Network (CIESIN), Columbia Pierella Paci (Practice Manager, World Bank), Chris- University, and the CUNY Institute for Demographic tian Rogg (Head, DFID Ethiopia), Sam Yates (Team Research (CIDR). Chapter 6: Alexander Meckelburg. Leader Livelihoods and Humanitarian team, DFID), Chapter 7: Samuel Mulugeta. Chapter 8: Matthew Anne Bakilana (Program Leader, World Bank), Hobson, Sarah Coll-Black, Simon Narbeth, Lucian Richard Spencer (Program Leader, World Bank), Pop, Judith Sandford, and Wout Soer. Chapter 9 Nataliya Mylenko (Program Leader, World Bank), is based on a report on the “Sources of Growth and Vikas Choudhary (Senior Agriculture Specialist, in Lowland Peripheries” prepared for DFID by IPE World Bank). Tripleline. xvii Poverty and Vulnerability in the Ethiopian Lowlands Overview Introduction To identify a policy agenda fostering resilience This study responds to a request in March 2018 by in the lowlands, the report adopts an analytical the Ministry of Finance and Economic Cooperation framework with three main features. First, the (MoFEC), to the World Bank and DfID to carry out livelihood system of the population in the lowlands a study of the lowlands with a view to strengthen- is examined in detail for the purpose of providing a ing the resilience of lowland populations to external better understanding of the patterns of behavior shocks. It builds on previous work undertaken by observed and the factors that are associated with the United Nations (UN) in examining practical ways the prevalence of poverty at a given point in time to end humanitarian need in the lowlands (see HLF and changes in poverty over time. Second, a dis- Concept Note, 2018),1 but goes beyond this original tinction is made between the poverty status of a study to describe the interplay of factors contrib- household at a given point in time and the vulnera- uting to the poverty and vulnerability2 of Ethiopia’s bility of a household to poverty, which is about the lowland populations and to make recommendations likelihood of a household being poor in the future. to improve the effectiveness of current humanitar- This distinction between poverty and vulnerabil- ian and development responses. ity is particularly useful for the development of a This overview synthesizes the nine self- safety net system that reduces the need for recur- standing chapters of the report that examine rent humanitarian appeal by supporting not only different dimensions of poverty, vulnerability, the chronically poor but also being able to expand and resilience to shocks in the Ethiopian lowlands. to include new beneficiaries during (and contact People’s resilience to shocks is generally affected after) droughts, and long-term development inter- by three important factors: the level of their expo- ventions. Third, emphasis is placed on understand- sure, their sensitivity to the shock, and their cop- ing the different sources of household vulnerability ing capacity (De Haan, 2016; Cervigni and Morris, in the lowlands, how these sources of vulnerability 2016). Building resilience is about ensuring that have evolved over time, and how the capacity of households can respond to shocks without losing households to cope with changes has evolved. An access to sustainable livelihoods and depleting their understanding of the causes of vulnerability to assets to the point of no recovery. poverty is necessary for the design of policies that increases the overall resilience of household welfare to shocks. The frequent exposure of households 1 Omzigt, Dirk-Jan, A Study on Lowland Resilience in Ethiopia—Building to droughts and other shocks in a highly variable a More Developmental Approach to Responding to Shocks, May 2018, UNDP. and marginal environment has led to the develop- 2 Vulnerability is predictive or an ex ante measure of welfare since it ment of adaptive strategies for coping with shocks, refers to the prospective immediate impact of a shock (or the likelihood including herd diversification, human and livestock of being poor in the event of a future shock). Poverty is an ex post mea- mobility, and common access to grazing resources. sure of welfare and an important factor that increases a household’s or community’s susceptibility to the impact of shocks and the ability Changes in the availability of water, the amount of to recover from them. In contrast, resilience is about “the speed of rainfall and its distribution across space and over recovery from shocks” or the longer time path of well-being in the face time, the frequency and duration of droughts, the of shocks, and especially the likelihood that any adverse outcomes of either risk avoidance or a realized shock do not persist for an extended stock of human capital assets and livestock owned, period. declines in the availability of land for grazing, and xviii Map E.1: Administrative regions with highland/lowland relief political and economic marginalization, all com- in a highly variable and marginal environment. bined or in isolation, tend to strain the ability of Altogether, the lowlands (below 1,500 meters) hosueholds to cope with shocks. This, in turn, constitute over 60 percent of the country and results in increased household vulnerability to pov- are home to some 10–12 percent of the country’s erty in the future, as well as their speed and ability population. Administratively they cover parts of to recover from the shocks. Tigray and Amhara, the four “emerging” regions Based on the preceding analytic framework, of Benishangul-Gumuz (BG) and Gambella in the a selective policy agenda for the promotion of west, and Somali and Afar in the east and south- resilience in the lowlands is identified consisting east, as well as parts of Oromia and SNNP regions of three pillars (elaborated in more detail in the last in the south (see Map E.1). The backbone of the section of this overview): economies of these lowland areas is extensive live- stock herding by pastoralists and agropastoralists, • investing in human capital development, through who constitute some 56 percent and 44 percent, improved delivery of human development respectively, of the total pastoralist population of services; some 6.6 million people in the country.3 The main • developing a safety net system that can deal difference between these livelihood categories is in with short-term crises and longer term develop- their degree of dependence on livestock. According mental needs; and to one definition, pastoralists derive 50 percent or • improving the efficiency of the livestock econ- more of household gross revenue from livestock or omy while encouraging economic diversification livestock-related activities, while agropastoralists and improving infrastructure and connectivity. (those who engage in both pastoralism and agri- culture) derive more than 50 percent of household Pastoral Livelihoods in the Lowlands Ethiopia’s lowlands are home to some of the coun- try’s most vulnerable populations who depend 3 Atlas of Ethiopian Livelihoods (2010), based on data from the 2007 on extensive livestock herding and/or farming national census. Overview  xix gross revenue from farming and 10–50 percent demand for livestock and livestock products and from livestock (Ellis and Swift, 1988). improved infrastructure and connectivity in many Although pastoral livelihoods in the lowlands areas that have allowed for easier access to mar- have fairly diverse forms, their basic economics kets and more effective and timely humanitarian consist of the simultaneous need to manage risks aid delivery. These changes have had both positive to financial assets (livestock) and food security and negative effects on the resilience of lowland in the context of shocks. Major shocks in these populations. For those who still have access to areas commonly refer to extreme weather events, livestock, rising livestock prices have compen- such as droughts and floods, sharp fluctuations in sated for smaller household livestock herds. But it the terms of trade, human and animal diseases, also means that livestock owners are increasingly and conflicts between groups. These shocks are dependent on the market and vulnerable to fluc- compounded by vulnerability driven by low levels of tuations in future terms of trade for their prod- human capital, exposure to environmental factors, ucts. Generally, while there are fewer poor people weak governance, and remoteness. Generally, the in the lowlands than two decades ago, there are poorer you are in terms of both consumption and many more that are vulnerable to falling below multidimensional poverty, the more vulnerable you the poverty line in the event of external shocks. are to these shocks, as even minor fluctuations Indeed, lowland populations are the most vulnera- may affect your ability to cope and recover after ble to covariate shocks of any agroecological zone the shock. in the country. While there is a strong positive Households engaged in pastoralism consume story of change—that is too often overlooked by meat and milk that are sourced directly from their commentators—­ there are also clearly winners and livestock and cereals that are sourced mainly losers in the changes that have been taking place. through crop production and/or livestock sales. Considering the high prevalence of vulnerability Prolonged drought shocks can quickly decimate to poverty, the study focuses on the challenges livestock herds, cause crop failures, and increase faced to inform the broader development policy food insecurity.4 In the event of a drought, vulnera- package and promote resilience in the lowlands. bility to poverty and food insecurity increases due The frequent exposure of households to droughts to livestock mortality, declines in the milk yields of and other shocks in a highly variable and marginal animals, and increased constraints in the capacity environment has led to the development of adap- of pastoralists to sell livestock and buy cereals. As tive strategies for coping with shocks, including the drought progresses, the value of livestock may herd diversification, human and livestock mobility, also fall and the price of cereals increase, meaning and common access to grazing resources. The pre- less favorable terms of trade between livestock vailing view among observers and policy makers is and cereals (the kgs of staple grain that could be that vulnerability to shocks is increasing over the purchased with cash earned from the sale of an long term in the lowlands. animal). Herd re-constitution after a drought is a long and slow process because of the mortality rate Changes in Monetary and Nonmonetary in the female reproductive stock. In fact, reductions Poverty in herd sizes below a certain threshold and compo- sition make it very unlikely that a pastoralist can Pastoralists have done relatively well in recent years ever recover his losses. despite the 2015 drought. Poverty rates among Despite significant falls in headcount poverty pastoralists declined, while poverty rates among rates in the past two decades in lowland regions, full-time farmers increased. At the same time, the largely on the back of increasing national and severity of poverty has increased in many groups. global livestock prices, many people continue to be trapped in chronic poverty or are vulnerable Poverty headcount rates in each of the eight to sliding back into poverty due to insufficient regions of Ethiopia declined significantly between income, assets (including livestock), or human 2000 and 2016 (Figure E.1).5 In 2000 there was capital. Some populations in the lowlands have been able to take advantage of increased global 5 An estimate of expenditure-based or “monetary” poverty is derived by comparing total (food and nonfood) household expenditures per adult 4Food security consists of four broad dimensions: availability, access, equivalent against the nominal (or inflation-adjusted) poverty line per adult utilization, and stability (over time) (Barrett, 2009). equivalent (AE) (3,781 Birr per AE in 2011, and 7,184 Birr per AE in 2016). xx  Overview Figure E.1: Overall decline in headcount poverty by region 2000–2016 (HCE surveys) 68 63 58 53 48 43 38 33 28 23 18 2000 2005 2011 2016 Tigray Afar Amhara Oromia Somali Benishangul-Gumuz SNNP Gambella Source: World Bank staff estimates based on the respective HCES and WMS. a considerable diversity in the prevalence of pov- rate in the DPL&P was also accompanied by a small erty across regions, ranging between 38 percent in decline in the poverty gap (distance from the pov- the Somali region and just under 63 percent in the erty line) and in the severity (depth) of poverty. Tigray region. By 2016 the poverty rates in all eight The decline in the monetary measures of pov- regions ranged between 20 and 28 percent, exhibit- erty is accompanied by declines in the value of ing a rapid convergence to a lower headcount pov- the Multidimensional Poverty Index (MPI) in the erty rate. With the unique exception of the Somali two main regions of the drought-prone and pas- region where poverty seems to have increased toral areas, Afar and Somali, albeit from a high temporarily between 2000 and 2005, poverty in all value of deprivation in basic services in 2011 (see other regions has either declined rapidly or did not Figure E.3). The Multidimensional Poverty Index increase from survey to survey. constructed by Oxford Poverty & Human Develop- A closer inspection of the poverty headcount ment Initiative (OPHI) focuses on the combination rate across agroecological zones between 2011 of deprivations that simultaneously afflict a house- and 2016 reveals that decline in poverty in the hold and consists of weighted indicators on child drought-prone lowlands and pastoral (DPL&P) mortality, child nutrition (measured by child weight areas is the largest among all agroecological for age), child school enrollment, completed years zones (Figure E.2).6 The DPL&P areas had the of schooling by adult household members, and higher poverty rate of 37 percent in 2011, but by standard of living (e.g., access to electricity, drink- 2016, the poverty rate was 27 percent, a decline ing water, sanitation, type of floor, cooking fuel, and by 10 percentage points (pp), considerably greater asset ownership (excluding livestock). than the decline by 7 pp in the poverty rate at the A similar pattern is observed with respect to national level. The decline in the headcount poverty the change in the prevalence of stunting among children between zero and five years of age (Fig- ure E.4). There is an increasing consensus that child 6 The country is divided into four agroecological zones: the drought- height for age z scores (HAZ) provide a better mea- prone and pastoral lowlands (DPL&P), the moisture reliable lowlands, the drought-prone highlands, and the moisture reliable highlands fol- sure of the nutritional status of children and of their lowing EDRI, 2009 (see Map 1.1 in Chapter 1). potential welfare as adults. In a pattern consistent Overview  xxi Figure E.2: Headcount poverty rate by agroecological zone: 2011 vs. 2016 (HCES) 0.40 0.37 0.35 0.30 0.30 0.30 0.27 0.28 0.25 0.25 0.25 0.24 0.23 0.21 0.20 0.15 0.10 0.05 0 Drought-prone Humid moisture Drought-prone Moisture National lowlands and reliable highlands reliable pastoral areas lowlands highlands 2011 2016 Source: Bank staff estimates based on the 2011 and 2016 HCES and WMS. Note: All comparisons between 2011 and 2016 are based on the zones in Afar and Somali regions that were surveyed in both 2011 and 2016 (these are zones 1, 3, and 5 in Afar and 1, 2, and 9 in Somali). In 2016 some additional zones were included in the HCES/WMS survey in Afar and Somali to increase the representativity of the survey in pastoral areas. Figure E.3: Multidimensional poverty index by region: 2011 vs. 2016 0.70 0.60 0.50 0.40 0.30 0.20 0.10 0 ay ar ra ia i uz NP lla ri a a al ab w ra om Af ha m be um gr Da SN Ha Ab So Ti Am m Or l-G re Ga s di Di gu Ad an sh ni Be 2011 2016 Source: Oxford Poverty & Human Development Initiative (OPHI) accessible at: https://ophi.org.uk/multidimensional-poverty-index/global-mpi-2018/ with that observed for expenditure-based poverty food aid and lack of safety nets7 in Gambella and and the MPI, Afar and Somali regions also appear Benishangul-­ Gumuz had much less of a cushion- to have experienced a decline in the prevalence of ing effect on poverty rates (than in other lowland stunting between 2011 and 2016. areas). Expenditure-based or monetary poverty rates The declines in the prevalence of poverty in the declined in the (DPL&P) areas of Ethiopia, despite drought-prone lowlands and pastoral areas have the 2015–16 El Niño triggered droughts that been accompanied by an increase in the reliance affected large parts of Afar and northern Somali on crop production as the main source of income region (see Figure E.5). In the humid moisture of households between 2011 and 2016 (Figure E.6). reliable lowlands, however, the headcount poverty The share of the population (urban and rural rate seems to increase between 2011 and 2016, areas combined) reporting livestock as their main when food aid is excluded, while it remains con- stant when food aid is included (see Figures E.5 and E.2). The main reasons for this appear to be that 7 PSNP is absent in Gambella and Benishangul-Gumuz. xxii  Overview Figure E.4: The prevalence of stunting among children 0–5 years of age by region: 2011 vs. 2016 60 50 40 30 20 10 0 ay r ra ia i uz NP la ri a a al a ab w ra om l Af ha m be um gr Da SN Ha Ab So Ti Am m Or l -G re Ga is Di d gu Ad an sh ni Be 2011 2016 Source: Ethiopia Demographic and Health Surveys 2011 and 2016. Figure E.5: Trend in the poverty headcount rate without aid/donations: 2011 vs. 2016 0.40 0.37 0.35 0.33 0.30 0.29 0.29 0.29 0.28 0.30 0.25 0.25 0.20 0.15 0.10 0.05 0 Drought-prone Humid moisture Drought-prone Moisture lowlands and reliable highland reliable pastoralists lowland highland 2011 2016 Source: Bank staff estimates based on the 2011 and 2016 HCES and WMS. Notes: Poverty rates in the absence of food aid are estimated by subtracting from total household consumption all food aid and donations received either from humanitarian organizations and/or from Ethiopia’s social safety net system (see Chapter 2 for more details). All comparisons between 2011 and 2016 are based on the zones in Afar and Somali regions that were surveyed in both 2011 and 2016 (these are zones 1, 3, and 5 in Afar and 1, 2, and 9 in Somali). In 2016 some additional zones were included in the HCES/WMS survey in Afar and Somali to increase the representativity of the survey in pastoral areas. livelihood declined only slightly, from 16.3 percent and other shocks (see Figure E.7). In contrast, the in 2011 to 15 percent in 2016, whereas the share of poverty rate among farmers and agropastoralists the population in crop and livestock almost halved, has not changed significantly between 2011 and from 18. 5 percent in 2011 to 9.5 percent in 2016. 2016. However, despite the reduction in their pov- Most of the population shifting away from only erty rates, many pastoralists hover just above the livestock or crop and livestock seems to be moving poverty line and could become poor if they experi- into full-time crop production, with the fraction of enced serious shocks. Figure E.7 reveals that there the population in crop production increasing from has been a substantial reduction in the headcount 35 percent in 2011 to 41.6 percent in 2016, and the poverty rate among households with livestock as rest moving into other livelihoods/occupations. their main livelihood, with the proportion of the Contrary to the prevailing view, consumption-­ poor population among those largely dependent based or monetary poverty among pastoralists on livestock decreasing from 58 percent in 2011 to has declined significantly despite the drought 29 percent in 2016 (a reduction of 29 percentage Overview  xxiii Figure E.6: Changes in the percentage of population in different livelihoods—drought-prone lowlands and pastoral areas in 2011 and 2016 0.45 0.42 0.40 0.35 0.34 0.35 0.30 0.30 0.25 0.20 0.19 0.16 0.15 0.15 0.10 0.10 0.05 0 Livestock Crop and Crop All other livestock production occupations 2011 2016 Source: Bank staff estimates based on the 2011 and 2016 HCES and WMS (urban and rural areas combined). Note: All comparisons between 2011 and 2016 are based on the zones in Afar and Somali regions that were surveyed in both 2011 and 2016 (these are zones 1, 3, and 5 in Afar and 1, 2, and 9 in Somali). In 2016 some additional zones were included in the HCES/WMS survey in Afar and Somali to increase the representativity of the survey in pastoral areas. Figure E.7: Poverty rate (headcount) by livelihood—drought-prone lowlands and pastoral areas in 2011 and 2016 0.70 0.60 0.58 0.50 0.40 0.37 0.32 0.34 0.30 0.29 0.26 0.26 0.20 0.15 0.10 0 Livestock Crop and Crop production All other livestock occupations 2011 2016 Source: Bank staff estimates based on the 2011 and 2016 HCES and WMS. Note: All comparisons between 2011 and 2016 are based on the zones in Afar and Somali regions that were surveyed in both 2011 and 2016 (these are zones 1, 3, and 5 in Afar and 1, 2, and 9 in Somali). In 2016 some additional zones were included in the HCES/WMS survey in Afar and Somali to increase the representativity of the survey in pastoral areas. points). In contrast, the headcount poverty rate in whether the standard of living among poor house- all other livelihoods appears to have declined more holds in different livelihood categories improved or moderately (by 11 percentage points in all other deteriorated. Figure E.8 presents the poverty gap and occupations and only by 3 percentage points in the severity of poverty measures, with the former crop production). measuring the average distance of household expen- However, even though the headcount poverty ditures from the poverty line among the poor, and the rate may have declined, overall poverty has “deep- latter placing a higher weight on the poverty gap of ened” among poor households engaged in agropas- those further away from the poverty line. Figure E.8 toralism (crop and livestock production) or in crop reveals that the depth of poverty among those production alone. The reductions in the headcount engaged in crop and livestock, as well as in crop pro- poverty rate by livelihood are not informative as to duction only, increased between 2011 and 2016. xxiv  Overview Figure E.8: Poverty gap and severity of poverty by household Vulnerability livelihood—drought-prone lowlands and pastoral areas in 2011 and 2016 Despite significant falls in headcount poverty rates Poverty Gap in the past two decades, too many people continue 0.25 to be trapped in chronic poverty or are vulnerable 0.20 0.19 to sliding back into poverty. 0.15 0.11 0.12 0.10 0.08 0.09 0.09 Efforts at building a more resilient future for 0.07 0.04 the population living in the lowlands require a 0.05 better understanding of the interplay of factors 0 contributing to the poverty and vulnerability of Livestock Crop and Crop All other livestock production occupations Ethiopia’s lowland populations. The familiar head- count poverty rate is not necessarily a very useful 2011 2016 guide to the future poverty status of a household. Severity of Poverty Households in the lowlands are frequently hit by a variety of shocks resulting in high income and con- 0.10 0.08 sumption volatility. In a context of high volatility, 0.08 a household’s currently observed poverty status 0.06 0.06 0.05 0.05 is, therefore, unlikely to be a very good guide to a 0.04 0.04 household’s vulnerability to poverty, or its “poverty 0.03 0.03 0.02 risk.” Poverty is a backward looking or ex post mea- 0.02 sure of welfare, whereas vulnerability is predictive 0 Livestock Crop and Crop All other and indicates what might happen to a household livestock production occupations or community exposed to a particular hazard. For example, a household may currently be nonpoor, 2011 2016 meaning that its current level of welfare measured Source: Bank staff estimates based on the 2011 and 2016 HCES and WMS. by consumption is above the poverty line, but in Note: All comparisons between 2011 and 2016 are based on the zones in the event of a future shock, the same household Afar and Somali regions that were surveyed in both 2011 and 2016 (these may end up falling below the poverty line. Such a are zones 1, 3, and 5 in Afar and 1, 2, and 9 in Somali). In 2016 some household may be said to be “vulnerable to poverty” additional zones were included in the HCES/WMS survey in Afar and Somali to increase the representativity of the survey in pastoral areas. even though currently it is not poor. In general, the extent of vulnerability to poverty depends on the risk management strategies of households and At first sight, the short-run trends in monetary communities, the abilities of households to cope poverty appear to be contradictory to the prevail- after the incidence of the shock (e.g., assets owned, ing views about poverty in the lowlands. The expla- herd size and composition of herds, social capital), nation for these short-run changes in poverty lies in and access to safety nets (e.g., PSNP).8 the fact that the terms of trade prevailing between In spite of the decline in the headcount poverty livestock and staple grain prices were favorable to rate (Figure E.2), the drought-prone lowlands and pastoralism. As elaborated in more detail in Chap- pastoral areas have the highest vulnerability to ter 2, the welfare of households engaged in pastoral- ism in the lowlands has increased due to increasing livestock prices relative to those of cereals. However, 8 Vulnerability to poverty is related to the concept of “resil- these favorable terms of trade between livestock ience,” which concerns the longer time path of well-being in the face of shocks, and especially the likelihood that any adverse and staple grains have not benefited all households outcomes of either risk avoidance or a realized shock do not equally. Farmers and agropastoralists, who generally persist for an extended period. For example, a nonpoor house- have considerably less livestock (or no livestock at all) hold may be vulnerable to becoming poor due to a job loss and yet be quite resilient if the prospects for finding follow-on compared to pastoralists and are more dependent on employment offering similar compensation are high and/or sales of grains, have been less able to benefit. Despite formal or informal safety net programs provide adequate sup- the improved livestock prices, many agropastoralists port promptly. In statistical terms, a nonpoor household with high conditional variance of income might be both vulnerable are hovering just above the poverty line and, given the (to becoming poor) and resilient (because the poverty is suffi- much smaller herd sizes they have today, could easily ciently short in duration, intensity, and/or likelihood (Barrett slip below the poverty line when faced with shocks. and Constas, 2014). Overview  xxv Figure E.9: The prevalence of poverty and vulnerability—lowlands vs. other agroecological zones in 2016 0.70 0.66 0.60 0.50 0.48 0.46 0.43 0.40 0.38 0.29 0.30 0.30 0.25 0.26 0.22 0.20 0.10 0 Drought-prone Humid moisture Drought-prone Moisture reliable National lowland and reliable lowland highland highland pastoral Poverty rate Vulnerability rate Source: Bank staff estimates based on the 2016 HCES and WMS. poverty rates among all agroecological zones rather than the potential implications of these in Ethiopia (see Figure E.9). According to esti- investments on welfare. mates detailed in Chapter 2 of the report, about Vulnerability to poverty from low human cap- two-thirds or 66 per cent of the population in ital or high consumption volatility is higher than the drought-prone lowlands and pastoral areas in any other part of the country (see Figure E.10). are either poor or at risk of becoming poor in the Figure E.10 reveals that in the drought-prone low- event of a shock in the future. In the drought-prone lands and pastoral areas, both poverty-induced and lowlands and pastoral areas, vulnerability due to risk-induced vulnerability are higher than in other idiosyncratic shocks is relatively more important agroecological zones. than vulnerability due to covariate shocks, such In the lowlands, vulnerability associated with as drought (ratio equal to 0.86). Nevertheless, of low human capital and assets (poverty-induced all agroecological zones in Ethiopia covariate vul- vulnerability) is relatively more important than nerability, such as that associated with droughts, vulnerability associated with climatic and envi- is highest in the drought-prone and pastoral ronmental shocks (risk-induced vulnerability).9 lowlands. This is in sharp contrast to the other agroecological zones where the relatively more important source Sources of Vulnerability of vulnerability to poverty is high variability in con- An understanding of the causes of vulnerability sumption. This attests to the unique nature of the to poverty is necessary for the design of policies lowlands in comparison to other agroecological that increase the overall resilience of household zones and informs the design of policies that can welfare to shocks. If vulnerability to poverty is increase the overall resilience of household welfare associated mainly with low assets and low human to shocks. capital endowments (i.e., it is poverty-induced), Thus, alleviating vulnerability rather than then cash transfer programs or programs enhanc- poverty, per se, should be the primary aim of ing the delivery of basic services facilitating invest- policy in the lowlands. The current focus of the ments in physical and human capital are likely to safety net system in Ethiopia is primarily on the be the most appropriate. In contrast, if vulnerability chronically poor households. Yet, many households to poverty is associated mainly with variability in the lowest wealth strata across lowland areas in consumption volatility caused by climatic and are vulnerable or a “shock away from poverty” and environmental factors (i.e., it is risk-induced), then an insurance type of program may be needed to increase resilience. The insurance provided would 9 In the lowlands, the ratio of the vulnerability rate that is poverty allow households to make production and invest- induced to the vulnerability rate that is risk induced is 1.2 (=0.37/0.30), ment choices based on their expected returns, the highest among all other agroecological zones. xxvi  Overview Figure E.10: Sources of vulnerability in 2016—poverty induced vs. risk induced 0.40 0.37 0.35 0.30 0.30 0.25 0.25 0.26 0.26 0.25 0.22 0.20 0.19 0.17 0.15 0.14 0.10 0.05 0 Drought-prone Humid moisture Drought-prone Moisture National lowlands and reliable highland reliable pastoral lowland highland Poverty induced vulnerability Risk induced vulnerability Source: Bank staff estimates based on the 2016 HCES and WMS. require support during droughts, as well as alterna- Water availability and access tive income earning opportunities. Considering that Water availability is a critical component of low- vulnerability to poverty in the lowlands is primarily land livelihood systems. However, there is a wide “poverty induced” (i.e., due to low assets and human variety of conditions in the lowlands related to capital endowments), cash transfer programs and/ groundwater presence and availability. Large areas or programs strengthening the delivery of basic have brackish or salty groundwater, with high min- services and facilitating investments in human and eral content that makes it unsuitable for drinking physical capital are the most appropriate to the water and most other purposes. As a result, water context of the lowlands. resources in pastoralist areas, particularly in the two regional states of Afar and Somali, are usu- Vulnerability to poverty in the lowlands is primarily due ally scarce and costly to access for the rural water to low human capital and assets (including livestock). supply. Surface water is largely confined to the per- nennial and seasonal rivers that run through the The lowlands, especially the drought-prone lowlands, whereas groundwater availability var- lowlands, lag the highlands across a range of ies across the lowlands. Afar state is underlain by human development indicators (see Chapter 3 for several deep and prolific aquifers of regional impor- more details). Across a range of human develop- tance. The high salinity of the waters in some areas, ment outcomes, from school attendance to infant however, limits use of the water for irrigation and and child mortality rates, and severe acute malnu- drinking purposes. In Somali region, by contrast, trition, the lowlands do worse than the highlands. water tables in some areas run extremely deep and Accessibility and quality of services are also worse. aquifers are low yielding. Notable areas for better The poor levels of education in the lowlands are groundwater potential include: (i) the Shinile area also a key constraint to making progress on gen- in northern Somali region, (ii) the Jerer valley, and der equality. Negative social norms remain a major (iii) the shallow groundwaters in the Fafan valley in barrier to women and girls for accessing basic ser- the northern part of the regional state. Elsewhere vices and full participation in market opportunities. in the region, groundwater of limited extent can be encountered along the banks of the major rivers. However, a deeper understanding of the long-term trends Shallow groundwater also occurs in the Ogaden of the climatic and enviromental factors associated with zone. vulnerability is also essential for informing the broader Access to improved water supply and sanita- development policy package for increasing resilience in the lowlands. tion facilities in pastoralist areas of Afar, Somali, and pockets of pastoralist areas in Oromia and Overview  xxvii Map E.2: Regional water supply coverage SNNPR lag highland regions and the national coverage (53%). The same survey also suggests average. Access levels for water and sanitation that there is very low sanitation coverage in Afar range from 39–61 percent, and sanitation coverage (23%) and Somali (5%) regions. ranges from 6–21 percent, respectively, compared The lower rainfall in pastoralist and agropas- to 62–95 percent and 41–76 percent, respec- toralist areas, combined with the the scarcity of tively, in highland areas (OWNP Phase II program, potential groundwater water sources, increase November 2018). According to the Ethiopian WaSH the costs of accessing water. While there are Poverty Diagnostic study, agrarian highland wore- shallow aquifers in the alluvial beds of many rivers, das were significantly more likely to have access they are not perennial and may last a few weeks to improved water than agropastoralist woredas, to a few months after the rainy season. There are by about 10 percentage points. Map E.2 highlights also areas that have sedimentary or volcanic for- regional coverage rates by woreda. mations, with deep to very deep aquifers. However, According to a survey conducted in lowland some of the deep aquifers could have a meager woredas of Somali, Afar, and South Omo zones yield or may have poor water quality, often with of SNNP regions (USAID, May 2016), 48 percent high mineral content and high temperatures, which do not have access to improved drinking water may make the water unsuitable for drinking, small- sources, of which 49 percent were in Afar, 81 per- scale irrigation, and even for animal watering. cent in Somali, and 15 percent in SNNP. The same The scattered settlement patterns of pasto- survey highlighted that of the 51 percent that have ralist and agropastoralist areas have made water access to water in Afar, 86 percent of households service schemes expensive to maintain. Given the use water from drilled boreholes. Baseline data transient nature of some communities, the oper- collected in 366 woredas in 2018 also indicated ation and maintenance of sophisticated pumping that 80 percent of the woredas in SNNP, Somali, systems have been inadequate. Systems that and Afar have below the national average access require extensive pumping of deep groundwater xxviii  Overview sources, for example, require diesel to operate Rainfall and droughts pumps and replacement of spare parts that are Rainfall is highly variable, both in distribution often imported (electromechanical equipment, steel and amount. This variability is particularly marked casing, etc.). Consequently, water investments in in the arid and semiarid, drought-prone lowland sparsely populated areas have fallen into disrepair areas which cover most of Afar and Somali regions due to lack of accountability, community engage- and parts of Oromia and SNNP. Contrary to the ment, and capacity for repairs. conventional narrative that rainfall has been declin- The hydrology and geology in lowland Ethio- ing in past years, lowland rainfall appears to have pia are complex, and little is known about the increased slightly during the years 1982–2017, with quality and quantity of water resources, which the western lowlands showing the largest increases makes the cost of reconnaissance before drilling (see Map E.3). In some areas, including the south- expensive. Master plan studies of the different eastern Somali region, rainfall has declined slightly, river basins focus only on significant tributaries while in other areas, such as the western lowlands and detail the surface water resource potential for of Benishangul-Gumuz and Gambella and southern hydropower and irrigation use. Information avail- lowlands in SNNPR, rainfall has been increasing. able about groundwater resources in the lowlands However, despite a long-term slight increase in is fragmented. The hydro-geological data from rainfall in parts of the lowlands, rainfall is highly drilled boreholes in the lowland areas have not been variable, both in distribution and amount. The high systematically collected and analyzed to assess levels of interannual rainfall variability in the low- groundwater potential to orient design and man- lands are too significant for people to take advan- agement of water service delivery. tage of long-term rainfall trends, and continue The cost of drilling and groundwater exploita- to pose challenges and constrain livelihoods and tion in many of the pastoralist regions is increas- opportunities for economic growth. Climate change ingly expensive, with a technological shift toward projections also point to more frequent extreme deep boreholes and limited drilling capacities in weather events in the future. lowland regions. Although the knowledge base to Historically, Ethiopia’s drought hot spots have orient investment in groundwater is fragmented, occurred in the northeast, southeast, and south- the reliance on groundwater is increasing. To ern lowlands (see Map E.4). Eighty-seven percent reverse the high rate of drilling failure or the lower of the identified 450 drought-prone woredas are than average water yield or chances of boreholes in lowland regions—169 of these woredas are also “drying up,” the solution to date has been to drill flood prone. The most recent prolonged droughts in deeper. Borehole drilling in lowlands is expensive the northern pastoral areas of Afar and the north- and can be as high as US$291,000 per borehole, ern Somali region were in 2015–2016, and in the against a national median cost of US$128,000. southern pastoral areas of southern Somali, south- Review of actual drilling contracts in Oromia, ern Oromia, and south-west SNNP, in 2016–2017. SNNPR, Somali, and Afar regions have revealed As result of these frequent droughts, the that the average cost of drilling a borehole ranges drought-prone lowlands have been consistently from ETB 4.69 million in Somali to ETB 0.44 million targeted for humanitarian aid. On average the in Oromia. Besides the high cost of drilling, there is humanitarian appeal in Ethiopia has been equivalent also a higher rate of non-functionality—according to 1.4 percent of GDP over the last six years; spend- to the Ministry of Water, Irrigation and Electricity ing on safety nets (rural and urban PSNP) is less (MoWIE), the national average is 11 percent, while in than 0.5 percent of GDP.10 The cost of the humani- some lowland regions this rate is three times higher. tarian appeal is on average equivalent to 8.9 percent As a result of these factors, water develop- of government revenue.11 This cost has been rising ment in the lowlands needs to be undertaken with over time. Food accounts for 76 percent of these careful consideration of its wider environmental impact. Investing large sums of money into wide- spread bore hole drilling may not be the best way to use scarce budget resources and may do more 10 Endale, et al., Financing Social Protection in Ethiopia: A long-term per- spective, OECD Development Policy Papers, February 2019, No. 15, p. 32. long-term environmental harm than good. 11 Omzigt, 2018. Overview  xxix Map E.3: Trend in annual rainfall 1982–2017 (mm/year) Map E.4: Drought hot spots12 12 Based on UN-OCHA map. Hot spots are woredas in need of emergency support. xxx  Overview Map E.5: Internal displacement in the lowlands as result of conflict and extreme climate events (drought and floods) 2016–2018 (WB based on data from IOM) appeal amounts, on average. Nutrition, health, and drought and conflicts has been to create a human- water when combined, represent a further 10 per- itarian crisis in some areas with large numbers of cent of the need.13 In Somali and Afar regions, the IDPs crowded into camps and dependent on food percentage of the population reached by human- relief. With close to half of IDPs under 15, pressure itarian food assistance or PSNP has exceeded is also increasing on health and education ser- 50 percent in 2016 and 2017. Benishangul-Gumuz vices in receiving areas. In other areas, social ser- and Gambella also receive repeat food aid but to vices have been disrupted as teachers and health much small portions of their population. staff abandoned schools. While the steep rise in the number of IDPs is recent, most are unlikely to Conflict return home soon. A majority of IDPs (85 percent) In recent years, interethnic conflict, in addition indicate preferring to integrate locally. to drought, has seen the displacement of many Ethiopia has also experienced multiple refugee thousands of pastoralists and agropastoralists crises (some more protracted, some more recent) in the lowlands, particularly along the Somali- that have put a strain on the coping capacity of Oromo regional border (see Map E.5). Of the 1.5 national and local authorities in the lowlands. As million IDPs in the lowlands, half have been dis- of August 2018,14 Ethiopia is host to 905,831 refu- placed by conflict, and 4 out of 10 by drought. gees and 1,964 asylum seekers. The main popula- Internal displacement in the lowlands primarily tion groups are comprised of 422,240 refugees and affects the Somali region (with close to 7 out asylum seekers from South Sudan; 257,283 from 10 IDPs in the lowlands) and Oromia (with a quar- Somalia; 173,879 from Eritrea; and 44,620 from ter of IDPs in the lowlands). The overall effect of Sudan; as well as, 1,891 from Yemen. 13Attributing to largest national donor funded programs, and consid- ering that health and nutrition are now a combined investment within the health sector. 14 From https://data2.unhcr.org/en/country/eth Overview  xxxi Most refugees are in Gambella (401,594), fol- of total lowland area, covering a strip running lowed by Somali (256,200), Benishangul-­ Gumuz north–south through the region. This also cov- (62,461), and Tigray regions (43,740), with the ers large areas of BG (245,000 ha) and SNNPR majority of the 26 refugee camps situated in (134,000 ha) representing 5.5 percent and 2.2 the lowlands. Refugees make up a significant percent of their respective lowland areas. In the proportion of the population in some regions and eastern lowlands, large commercial farms are outnumber the local population in Gambella. While primarily located along rivers and rely on expen- the presence of refugees brings benefits to the sive irrigation; in Somali, the majority are located local population in terms of access to services and along the Shebelle river and along the Awash in stimulating demand for goods and services, the river in Afar. implications of long-term refugee hosting are being • The introduction of permanent and semipermanent felt at the local host-community level, where their water points has led to a decline in access to and protracted presence has wrought environmental quality of grazing areas. The construction of bore- degradation and uneven service delivery between holes, large surface ponds, and birkedas (large refugee and host populations, and has the potential cement lined cisterns) in wet season grazing to cause local tension. A more sustainable man- areas in the Haud in the Somali region and parts agement of refugees that gradually provides ser- of Borana has led to settlement, agriculture, vices through national systems has the potential and range enclosure in these once seasonally to enhance service delivery for refugees and host exploited grazing areas. This, in turn, has led to a communities. decline in grazing quality and availability in these areas. Land fragmentation and loss of access to common • Bush encroachment by invasive species, such as grazing areas Prosopis juliflora, is an increasing problem in At the same time, there has been significant land parts of the lowlands, particularly in Afar, parts fragmentation in the lowlands and a decline in of Somali, and SNNPR. The presence of P. juliflora access to common grazing areas, which may be in Afar has increased by 42 percent between undermining pastoralist coping strategies. The 2000 and 2016 and now covers nearly 1.2 mil- drivers behind many of these observed land use lion ha, while the portion characterized by heavy changes are complex and interrelated, and include infestation (where 80 percent or more of the land the following: is covered) is over 300,000 ha, an increase of over 600 percent from 2000 levels. Many pas- • An increase in human population and settlement toralists regard bush invasion as a more serious has led in many areas to increased cropping and threat to their way of life than drought. It also land enclosures. According to data collected for appears that increasing resource competition this study, there has been almost a 1,000 per- between groups has resulted in increasing com- cent increase in the area of settlements in the petition and conflict over access to grazing and lowlands of SNNP and Afar between 1986 and water, and in the moist western lowlands, access 2016, and overall across the lowlands an increase to forest resources. of 179 percent. In terms of cropped area, the increase in parts of the lowlands has been almost Political and economic marginalization, as dramatic with an over 400 percent increase in and weak governance Afar, nearly 300 percent in Gambella, and overall The vulnerability from droughts, conflicts, and a doubling in cropped area in the lowlands, par- diminished access to common grazing areas is ticularly in better watered areas along the high- compounded by vulnerability driven by political land-lowland interface. and economic marginalization, and weak gover- • The excision of key dry season grazing areas for nance. Since 1995 the government has sought to commercial agricultural plantations along some address this marginalization through designating of the main rivers that run through the low- the main lowland regions as ‘emerging’ states lands has resulted in loss of riverine forest with additional financial benefits. Despite this and dry season grazing. This has particularly recognition and the benefits it has brought, weak affected Afar and parts of Gambella and SNNP. governance and accountability has continued to Gambella (438,000 ha) has the highest share limit the development of lowland areas. The low- of concessions representing nearly 15 percent lands are less able to raise their own revenue, which xxxii  Overview means that they are more dependent on the federal different ethnic groups, and between highland- government, and their financial needs are signifi- ers and lowlanders. Opening the space to greater cant considering the lagging human development accountability and dialogue, combined with a indicators, low population density, and recurrent deeper understanding of the vulnerabilities of its humanitarian shocks. Compared to the highlands, inhabitants, will be key to addressing the needs of the lowlands do well in terms of per capita fis- lowland populations. cal transfers. However, the rationale for different spending choices by lowland governments is not Changes in Pastoral Livelihoods always clear. over the Long Term Ethiopia’s model of local governance is inter- ventionist; it has been designed and implemented The impacts of many of the climatic and environmental from the center to effect change in the regions changes are manifested as a long-term decline in (Vaughan, 2017). In many areas this change has household herd sizes, a shift in composition of herds been rapid and profound,15 with interventions (keeping more browsers than grazers), a shift toward designed to transform socioeconomic, education, farming, and increased wealth inequality. health, and cultural norms. But, in the lowland periphery, the impact of this system has been In Afar and Somali regions, between 2010 mixed.16 Top-down decision making has led to chal- and 2016, herd sizes declined by more than lenges around accountability in many regions. Eth- three-quarters from 78 tropical livestock units nic parties affiliated with the federal government (TLU)17 to less than 19 TLU among the wealthiest formally govern the developing regional states, quintile in Afar, and among Somali by more than but these parties are not part of the ruling federal half from 37 to just over 17 TLU (see Table E.1). In coalition. In some regions, disagreements between the poorest category the decline was even more regional officials and federal officials have devel- dramatic, falling from 4.8 to 0.2 TLU and 1.1 to oped over the level and nature of federal interven- 0.2 TLU among Afar and Somali respectively, tion in their respective regions. State penetration effectively forcing this group to exit pastoralism in the lowlands is perceived by many lowlanders to altogether. be driven by the priorities and needs of the highland The growing inequality in livestock ownership majority, and development strategies such as vil- among households is confirmed by data collected lagization, land appropriation, and turning pasture by the International Livestock Research Institute or cropping land into commercial farms, have not (ILRI) in 2016. Based on livestock ownership as a always produced their intended results, and have critical dimension of wealth, Figure E.11 shows the often come at a high cost. In many areas these share of TLUs per household by wealth quartile approaches have further marginalized lowland across the four regions. The large share of livestock peoples, while also increasing tensions between owned by the wealthiest quartile is consistent Table E.1: Mean livestock holdings, by livestock holding quintile, region, and year in tropical livestock units Livestock AFAR SOMALI holding quintile 2010 2012 2014 2016 2010 2012 2014 2016 Poorest  4.8  1.5  0.8  0.2  1.1  0.5  0.6  0.2 2nd 13.7  4.9  2.6  1.4  4.2  3.0  3.6  1.5 Middle 25.5  8.8  5.1  3.3  7.7  7.1  7.2  3.5 4th 42.4 18.0 10.1  6.6 14.6 14.6 12.7  6.1 Wealthiest 78.5 61.7 35.6 18.9 36.7 38.7 27.5 17.1 Source: IFPRI/EDRI, 2018 Targeting Social Transfers in Pastoral Societies-Household Surveys for PSNP. 15 Pankhurst, (Ed), Change and Transformation in Twenty Rural Com- munities in Ethiopia, 2017. 16 Vaughan, Revolutionary democracy as an uncompromising political 17Tropical Livestock Units are livestock numbers converted to a common strategy: Revolutionary democratic state-building: party, state and unit. Conversion factors are: camels = 1.4, cattle = 1, sheep = 0.1, people in the EPRDF’s Ethiopia, 2012. goats = 0.1.  Overview  xxxiii Figure E.11: Distribution of livestock wealth in TLU by household quartile and by region (based on ILRI survey data 2016) .682 SNNP 11.3 23.4 64.6 .57 2.28 Somali 10.8 22.2 64.7 .558 3.94 Gini Afar 14.1 25.4 56.5 .47 Oromia 4.52 14.9 25.9 54.6 .448 2.82 Total 12.7 23.7 60.8 .517 0 20 40 60 80 100 Outcome share (percent) 0–25 25–50 50–75 75–100 Table E.2: Mean share of total household income (%) (ILRI survey data) Afar Oromia Somali SNNP Livestock income 67 63 52 43 Livestock product income 1 1.7 2.6 7.7 Income from income generating activities 3 7 31 14 Crop income 4.8 1 2.7 28 PSNP income 14 21 3 0.0 Wage income 9.5 5.7 8.6 7.4 Household with share of livestock income >50% (%) 76.0 71.0 55.0 52.0 across all four regions and is confirmed by the find- growing importance of the government’s Produc- ings of other case studies. tive Safety Nets Program or PSNP, as an important source of income in Afar (14 percent) and Oromia (21 percent), but less so in Somali and in SNNP Despite the smaller household livestock herds, livestock continues to be an important source of income for many, regions, although it is increasing in importance in thanks to the increased price of livestock. However, with the Somali region. Wage employment is of modest reduced livestock assets, susceptibility to shocks has importance in all areas but income generation is also increased. particularly important among Somalis providing 31 percent of household income in the region (see Despite growing inequality and falling house- Table E.2 for breakdown). hold herd sizes, livestock continue to dominate sources of income in all the drought-prone regions Promoting Resilience in the Lowlands: and, according to ILRI data, represent 67 percent of The Components of a Broader Development mean household income in Afar, 63 percent in Oro- Policy Package mia, 52 percent in Somali, and 43 percent in SNNP To address the needs of lowland populations, gov- regions (see Table E.2).18 The ILRI data confirm the ernment needs to address the primary sources of fragility and vulnerability in the region. PSNP and 18 While important, the study also showed that the value of livestock recurrent humanitarian interventions may alleviate sales was more significant for wealthier households, with an average of the symptoms but cannot address the primary 70 percent of all household income coming from livestock. Income from livestock was much lower for the poorest quartile with only 17.5 percent causes of this vulnerability identified above. The and 28.8 percent of sales in SNNP and Afar regions, respectively. analysis in this report identifies a selective policy xxxiv  Overview agenda for the promotion of resilience in the low- would be in improving the human capital of its lands with the following three pillars: population. The delivery of human development services in • investing in human capital development through the lowlands is limited by a confluence of factors, improved delivery of human development services; including low population densities, inefficien- • developing a safety net system that can deal cies with the Intergovernmental Fiscal Transfer with short-term crises and longer term develop- System, lack of operating and capital budgets, mental needs; and low prioritization by some regional governments, • improving the efficiency of the livestock econ- inadequate number of and insufficiently skilled omy while encouraging economic diversification. professionals, and the lack of incentives and accountability structures to deliver human devel- Each one of these pillars is elaborated in more opment outcomes. In the decentralized system, detail below. the mandate to deliver basic services rests largely Human capital development with woredas. Due to the challenging circumstances in the lowlands, including climatic conditions, low Building resilience in the lowlands requires a large population density, and settlement patterns, sec- investment in human capital through effective toral ministries have developed specific strategies and accountable basic services. The analysis in and approaches for delivery of human development this report suggests that overall, there needs to services in the lowland areas—especially for reach- be a much stronger focus on improving the human ing pastoral populations. However, these strategies capital in the drought-prone lowlands. Without a remain largely untested. Compared to the large focus on improving the human capital, there will human capital needs, regions and woredas tend be little chance to break the cycle of vulnerability. to allocate too little of their budgets for health With respect to investments in human capital, and education sectors, with minimal resources important policy considerations emerging from this available for capital and operational costs. Afar study include: and Somali regions have predominantly focused • the allocation of more on-budget resources for on investments in water and rural development human development in the lowlands; due to increased drought to the relative neglect of • the introduction of more results-based education and health (see Figures E.12 and E.13). In approaches; constrast, Gambella and Benishangul-Gumuz have • the introduction of Specific Purpose Grants, tar- preferred to allocate expenditures to education and geted to regions/woredas with low human devel- health (see Figure E.13). opment outcomes; and These differences in regional allocations • the alignment of civil servant incentives and pay extend to capital allocations from the MDG/SDG to results; and grants regional governments receive to boost • the recentralization of some local and regional the capital expenditures. For example, Afar and functions to higher levels and review mecha- Somali regions allocated their entire MDG/SDG nisms of local planning enforcement and sectoral budget to the water sector in EFY 2012, meaning budget allocation in view of national targets. that health and education did not benefit from these resources. In contrast, Amhara region, in Limited human capital is also a binding con- addition to the already high 43 percent of its straint to growth and gender equality in the regional expenditures going to human develop- lowlands. The lowlands are an area of economic ment, is spending 22 percent of its MDG/SDG on disadvantage which will require additional invest- human development infrastructure. Benishangul-­ ment (public and private) and government support Gumuz and Gambella are spending a commendable to fully participate in growth. Key areas include 23 percent and 38 percent, respectively, on the support to extensive and mobile pastoralism, sup- human development sectors. porting connectivity and investing in human capital Given the large human capital deficits in Afar without which many people will be unable to partic- and Somali regions, there appears to be a critical ipate in the benefits of growth. The main barrier to knowledge gap regarding the relative benefit-­ growth in the lowlands and greater gender equality to-cost ratio associated with the allocation of is the limited human capital of lowland populations. resources to the water sector versus the educa- The biggest return on investment in the lowlands tion and health sectors in their regions. Overview  xxxv Figure E.12: Expenditures on rural development and water by region, FY12–18 18 17 16 16 15 14 14 14 12 12 11 11 10 8 8 8 8 7 7 6 6 6 4 4 3 3 4 2 2 2 0 0 r i uz la ay ra ia NP ri a a al a ad w ra om l Af ha m be um gr Da SN Ha Ab So Ti Am m Or -G re Ga is l Di gu d Ad an sh ni Be Agriculture and rural development Water Source: MoFEC (see also Table 7.7). Figure E.13: Expenditures on education and health by region, FY12–18 35 32 30 28 25 26 25 24 21 20 20 17 16 15 14 14 13 12 12 13 11 11 11 10 8 8 9 7 5 0 ar i uz lla ay ra ia NP ri a a al ad w ra om Af ha m be um gr Da SN Ha Ab So Ti Am m Or l-G re Ga s di Di gu Ad an sh ni Be Education Health Source: MoFEC (see also Table 7.7). Weak human development services in the low- make response more difficult, the impact of shocks lands are also a significant impediment to effec- harder to identify, and the overall effort more tive humanitarian response; where basic services expensive. are stronger, the humanitarian response is more effective. A key to an effective and early response Safety net system for coordinating humanitarian is stronger basic services in the lowlands, to both and development efforts reduce the need for large-scale response and to Building resilience in the lowlands requires the ensure that responses are more effective. This can ability to respond effectively, both to transitory be most effectively accomplished when humanitar- needs arising from rapid onset emergencies, such ian and development actors work through a single as conflict, and chronic needs arising from slow and/or coordinated systems. Currently, humanitar- onset environmental changes, such as drought ian responses are fragmented and expensive. and extreme weather events. The instruments The quality of basic services is critical to to do both should complement rather than under- humanitarian assessment and response. Where mine each other so that the humanitarian response investments in basic services have been made, should seamlessly transition to a safety net sys- humanitarian response is stronger and more cost tem that supports chronic and predictable needs. effective. Weak services and low human capital Building resilience is ultimately about building xxxvi  Overview Figure E.14: Humanitarian requirements, 2008–18 (a) Number of people requiring food aid, 2008–18 (b) Regional distribution of people requiring food aid (2008–18), as a proportion of total population 12 12 25 10 10 20 8 8 15 6 6 10 4 4 5 2 2 0 0 0 ay Am r ra ia i NP l -G l a uz a ri al a w ra om el Af ha m um gr 08 09 10 11 12 13 14 15 16 17 18 Da SN an mb Ha So Ti 20 20 20 20 20 20 Or 20 20 20 20 20 re Ga Di gu In millions (left) sh ni In % of total population (right) Be Source: HRD, various issues and CSA. sustainable systems that can both tackle the varied across the regional states. The lowland short-term crisis but also the long-term needs of regions, including Somali, Gambella, and Afar, had the population to protect themselves against future a relatively large proportion of their population shocks and smooth out the high degree of variabil- requiring food assistance over the past 10 years. ity and risk in lowland environments. To date the The number of households enrolled in the two have not always worked in close tandem as lowland PSNP is significant. Between 2008 and humanitarian interventions have been relied on to 2017, the number of PSNP beneficiaries in Afar too great extent to tackle predictable and chronic region remained fairly constant, fluctuating slightly need, while the latter should be built into govern- between 470,00 and 560,000. In Somali region the ment safety net and development interventions. number of PSNP beneficiaries increased steadily Bringing these two together will be key in building over the same period, responding to the periodic lowland resilience. reviews of needs assessments, growing from over The number of people requiring food assistance 700,000 in 2015 to well over 1.6 million in 2016 over the past decade averaged about 4.8 million and 2017. Together, the number of households that people a year or 6 percent of the entire popula- are addressed through either the PSNP or human- tion, ranging between 2.5 and 10 million people itarian assistance in the lowlands is increasing. (Figure E.14). The number of people impacted Figure E.15 shows the combined annual PSNP and Figure E.15: PSNP caseloads in Afar and Somali regional states: 2008–2017 2,500,000 2,000,000 1,500,000 1,000,000 500,000 0 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Afar Somali Overview  xxxvii humanitarian caseloads in Somali and Afar regions • Ensure there are appropriate, scalable instru- since 2008. The numbers have increased over ments that can deliver support to those who time, with 2016 and 2017 representing the high- are poor and those who are vulnerable to pov- est figures. The overall increase is reflective of the erty. With the numbers of vulnerable house- national trend, which is an increase in the number holds approximately two times the number of of people in need of social transfers (cash or food). poor households in lowland areas, it follows In view of the increasing dependence of lowland that appropriate instruments able to rapidly populations on the PSNP and HFA, as well as the respond to their needs are required (for example, overall costs of the current approach, it makes a transfer program with flexible delivery mech- sense to bring the overall relief effort together anisms that can reach those who may not be into one unified system of response. This will allow poor today, but who are likely to become poor the concentration of development resources on if impacted by a ’shock’). Being vulnerable to supporting the livestock economy of the lowlands poverty and risk does not mean that households to get more value added out of the livestock it pro- are ‘in need’ all the time—rather, that vulnerable duces while encouraging alternative income earning households may need rapid support as livelihood opportunities or migration to urban areas or out- conditions change. When assistance is needed, side the region for those without livestock. a temporary and rapid scale-up to a larger case- A common policy framework for initiatives to load must happen before the ‘alarm’ sounds for address poverty and vulnerability in the lowlands the drought—multiple research shows that early needs to be developed. The key elements of this interventions are more cost effective20—and framework are set out below. with a ‘no regrets’ approach. This means that there should be pre-positioned funds and/or an • Ensure the program design and financing annual allocation to enable an annual response responds to the relative size of poor and vul- to transitory needs among the lowland regions. nerable caseloads in the lowlands, which are • Introduce one financial framework with appro- different from the highlands. In practice, the priate financial instruments for a scalable PSNP has a fixed national ratio between a ‘core’ system. The multiple systems and actors in the caseload of chronically food insecure bene- lowlands currently leverage multiple sources of ficiaries and a ‘transitory’ caseload, which it financing through government systems, as well can modestly respond to as vulnerability con- as parallel to government. Adopting one frame- ditions change and available financing allows. work will lead to humanitarian financing flowing The humanitarian appeals augment the PSNP’s through PSNP systems and will clarify the limits response to the ‘transitory’ caseload.19 The ratio to what government will accept financial respon- between a ‘core’ caseload of PSNP beneficia- sibility for, when responding to vulnerability. ries and a ‘transitory’ caseload in the lowlands This will then enable a realistic assessment of should better reflect local conditions (it currently the related costs in addressing responsibilities, reflects highland dynamics) and will need to leading in turn to a more effective deployment change on an annual (possibly seasonal) basis, of resources and ensuring complementar- as rainfall and vegetation changes. According to ity between government and nongovernment poverty data, a ‘core’ regular caseload in the low- resources. lands may need to be smaller than some regions’ current caseload (for example in Somali region) Improving the efficiency of the livestock economy while and the annual allocation to a ‘transitory’ case- encouraging economic diversification load larger than currently programmed. Within Within the lowlands, the pastoral livestock econ- regional boundaries, beneficiaries are likely to omy is a critical resource that needs investment differ in proportions, according to the incidence and support. Rural to rural migration has had weak of poverty in pastoral, agropastoral, and farming poverty reduction impacts. The livestock sector areas. is the most important source of livelihoods and 19While the PSNP can support a transitory caseload in all woredas 20 See Cabot-Venton et al. (2012). The Economics of Early Response within a PSNP region, in practice the HFA is directed to woredas with- and Disaster Resilience: Lessons from Kenya and Ethiopia, DFID; out a PSNP core caseload and to woredas where the “need” surpasses USAID, Economics of Resilience to Drought in Ethiopia, Kenya and the ability of the PSNP to respond. Somalia, (2018), among others. xxxviii  Overview growth for the lowlands. For pastoralists, support federal government to attract appropriate private to livestock is critical, as this sector is under threat investments that will help create some of these from the increasing frequency of droughts and opportunities while ensuring effective environmen- floods, and bush encroachment and loss of access tal and social safeguards are maintained. Barriers to grazing, which is forcing a reduction in overall to economic migration need to be removed so the livestock herd sizes and a shift in composition of poorer members of the community can look for jobs livestock. Given global and national demand for elsewhere, both within and outside Ethiopia. meat, livestock are an important economic asset Public investment will have to focus on improv- for pastoralists but require adequate support in ing connectivity and market infrastructure to terms of veterinary services and access to markets, enable greater links between lowland economic including export. For most pastoralists, domestic activity, urban demand, industrial activity, and markets may provide greater scope for income trade corridors. There is no clear case to build more increases because of higher prices compared with industrial parks in lowland areas. export prices. With respect to supporting and pro- The manufacturing sector based on industrial moting pastoral livelihoods in the lowlands, import- park development is not expected to have a major ant policy considerations emerging from this study impact in the lowlands apart from the compar- include: atively small agro-industrial sector. The study underlines that investors will not invest in areas • the adoption of measures supporting pastoralist without connectivity, labor force and local capacity, mobility as an important adaptation to rainfall appropriate resources, or adequate water supplies. variability and uneven forage availability; Industrial parks could be a source of employment • the legal recognition and support by the gov- for migrating rural lowlanders to the emerging ernment of pastoralist land tenure regimes to manufacturing hubs who could then support their ensure equitable access to grazing; communities with remittance flows. In areas of the • the improvement of extention services; and lowlands there are perceptions that new jobs, such • access to grazing and water; and veterinary, as in industrial parks, are for highlanders; lower abattoir, and cold chain services to better meet levels of human capital in lowland areas are likely domestic and external demand for livestock and to limit labor force participation locally and hinder livestock products. economic migration. There may be other opportunities to attract Given the nature of lowland pastoral livelihood completely new investments and new activities to systems, the provision of permanent water in lowland areas. This should be done based on dis- lowland areas needs to be carefully considered cussions with investors and making evidence-based in relation to other natural resources, including public investment decisions. Investment in the its long-term impact on grazing availability for capacity of regional governments will be critical, livestock. Too much water can lead to permanent as investors, all things being equal, will prefer areas settlement and overgrazing. Environmental change with better security and a high capacity of regional induced by water availability may have permanent administrations. and negative land use consequences. The princi- In general, the benefits of large-scale private pal challenge in these areas is to understand how and/or public investment in commercial agri- water development can contribute toward provid- culture in the lowlands needs to be assessed ing adequate water for human consumption and carefully against costs such as potential long- settlement, including exploiting untapped irrigation term damage to the environment and other local potential, while not upsetting traditional adaptive livelihoods. Harnessing the irrigation potential of strategies to environmental and climate variabil- the main rivers that run through the lowlands is ity on which large numbers still depend for their attractive but comes at a high cost both in terms livelihoods. of water management and the long distances to Economic diversification and alternative markets involved. Many agricultural concessions employment opportunities should also be encour- given to private/state investors remain largely dor- aged, particularly in or near urban areas and in mant. The long-term viability of new sectors, such areas of higher agricultural potential. This will as agroforestry in the west and spices in the south, empower regional governments to work with the Overview  xxxix is dependent on private investment and govern- Greater regional integration through devel- ment support to enable market infrastructure, and oping existing trade corridors to Djibouti and market capability to enable access to highland and Berbera and the opening of new growth corridors international markets. Experience of the success to Assab/Massawa should benefit the lowlands. of floriculture in the highlands should be heeded The Djibouti corridor has the highest potential as to enable the development of smallholder and out it reaches the greatest number of lowland res- grower market systems. idents, including the emerging urban cluster of The extractive sector will become a major Dire Dawa-Harar-Jijiga, and handles most of the source of growth, but this will mainly be through trade. While growth corridors offer opportunities national foreign exchange earnings and royal- for employment and increased competitive advan- ties, with limited employment opportunities and tage, there will also be some negative impacts direct benefits to the lowlands. Revenues from to be addressed (e.g., shifting from road to rail extractives should be shared equitably with regions impacts the market for ancillary services in the and local communities that bear the brunt of any lowland areas). Cross-border trade needs to be possible negative consequences. Balancing national facilitated and formalized to ensure a more level and local objectives by ensuring that new extractive playing field among traders and livestock produc- industries do not undermine the development of the ers. Cross-­border trade, most of which is largely tradeable manufacturing sector (e.g., Dutch dis- informal, plays a crucial economic, political, and ease) and that host communities feel the benefits social role for pastoralists in the Horn of Africa. of the sector (e.g., through better services), will be For example, the value of the livestock trade on critical. Low levels of human capital will be a con- the Somaliland border is estimated to be between straint to access new jobs from new investments. US$200–300 million, four to five times the official ­ Increasing urbanization will lead to the growth of recorded level. Because of the informal nature of the service sector in the lowlands, and with lower the trade, the benefits are captured by powerful levels of poverty and vulnerability in urban areas, elites. Greater support to formalizing trade and rural to urban migration provides a better opportu- improving veterinary inspection services would sup- nity than rural to rural migration for reducing pov- port pastoralists in a more inclusive way. erty and supporting growth. However, investment in cities will require strong regional governments. xl  Overview Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Poverty and Vulnerability in the Ethiopian Lowlands Chapter 1: Pastoral Livelihoods Summary census, pastoralists constitute 56 percent and This chapter provides a description of the key features agropastoralists 44 percent of the total pastoralist of pastoral livelihoods in the drought-prone lowlands. population of some 6.6 million people in the coun- The chapter documents a long-term decline in house- try. The main difference between these livelihood hold herd sizes, a shift in the composition of herds, and a categories is in their degree of dependence on growing economic diversification, including an increase livestock. According to one definition, pastoralists in cropping, and an increase in wealth inequality. Live- derive 50 percent or more of household gross rev- stock, however, continue to be an important source of enue from livestock or livestock-related activities, income for many households. This has been buoyed by while agropastoralists (those who engage in both the high prices that livestock command in the market. pastoralism and agriculture) derive more than However, as livestock assets have reduced susceptibility 50 percent of household gross revenue from farm- to poverty, the result of shocks has also increased. ing and 10–50 percent from livestock (Ellis and Swift, 1988). However, most of Ethiopia’s pastoral Introduction societies pursue multi-resource economies in which the balance between pastoral and non-­ pastoral The nature of agriculture-based livelihoods varies activities is constantly shifting in response to across the country based on ecological conditions. changing circumstances. The highlands and lowlands represent two con- Pastoralism is a set of specialized economic trasting ecological and economic spaces. The latter activities and techniques revolving around the herd- historically lag the development of the highlands ing and care of livestock, which may be used alone and are beset by problems of poorly developed or in combination with other economic activities at infrastructure and communications, low population different times and places. Pastoralism and pas- densities, highly variable and uncertain rainfall, and toralists are defined in the literature in different low agricultural productivity and human develop- ways. Pastoralism is “a mode of production which ment outcomes. While there are pockets of higher depends on natural forage. In the arid regions this potential agricultural productivity in the moist requires constant or periodic movement in search western lowlands and along the main rivers, such of pasture, a factor that differentiates this form of as Awash, Omo, and Shabelle, are suitable for irri- livestock production from those practiced by farm- gation agriculture, and the backbone of the econ- ers and ranchers” (Markakis, 1993, 1). Pastoralism omies of these lowland areas is extensive livestock is also simply defined as a way of life, emphasiz- herding by pastoralists and agropastoralists. ing the complex social, cultural, and institutional practices and value system surrounding livestock Pastoralism and Agropastoralism keeping. According to the Atlas of Ethiopian Livelihoods Historically, the kinds of pastoralism practiced (2010), based on data from the 2007 national and degree of dependence on livestock alone or in 2 combination with other activities has changed over human population pressure (49%); expansion time. Groups which are typically labelled pastoral- of settlements (19%); and a decline in livestock ist, such as Somali, may have subgroups which, numbers per household owing to drought (12%). because of the favorable conditions in which they Consequently, pastoralists started to culti- live, are able to specialize in both cropping and vate small plots of land to grow cereal crops in pastoralism at the same time. This is the case, situations which were marginal for cropping. for instance, for many Somali clans that live in About 94% of the respondents had increased the Jijiga and Teferi Ber areas of northern Somali the area of land devoted to cultivation at the region. Currently, in the Haud area of Somali region expense of grazing area” (pp. 203–4). with the new technology of concrete lined cis- terns or birkas, many pastoralists can stay in this The distribution of farmers, pastoralists, and formerly wet season grazing area all year round, agropastoralists across Ethiopia’s main regions are leading to cultivation and development of private set out in Table 1.1. While the population numbers in enclosures in what were once wet season grazing each category should be taken with a grain of salt, areas. Opportunistic rain-fed cropping appears to as accurately distinguishing between the categories be becoming much more extensive in many for- is extremely difficult, they give a sense of orders of merly pastoral areas, both driven by population magnitude. Interestingly, as Ethiopia’s largest pas- pressure in the highlands and poorer pastoralists toralist region, Somali region is also home to many taking up cropping. full-time farmers (nearly 600,000). Abate et al. (2010) describe how the shift from Pastoral groups subsist off their animals both pastoralism to agropastoralism occurred among directly through milk consumption and to a lesser pastoralists in the Rayitu district in Bale, Oromia extent by eating their meat, and indirectly by region, as follows: exchanging them or their products for other com- modities, such as grain they obtain in the market. “Historically (30 years earlier), livestock pro- It is a widespread misconception that pastoralists duction was practiced by 94% of families and largely survive on milk and meat. Most pastoralists the inhabitants were totally pastoralists. Live- have too few animals to survive on a purely pas- stock and livestock products played a major toral diet and depend on the market to buy grain. role and the rangelands were used mainly for One of the significant changes taking place today grazing. Currently, only 36% are purely live- as the numbers of livestock per capita decline (see stock producers, with 63% combining livestock more on this below) is the increasing dependence of and crop production (agro-pastoralists). Most poorer pastoralists on alternative income sources, of the families (73%), who commenced growing and for ‘middle’ households’ increasing dependence crops, did so during the period 1974–1991, with on the market for sales of livestock, especially 13% doing so before that time. Reasons for small stock, to make up for losses in milk yields adopting an agro-pastoral lifestyle included: (HEA Change Analysis in Somali and Afar Regions, the need to diversify household income (81%); 2017/18) Table 1.1: Rural population by livelihood zone (2007 Census) Region Cropping Agropastoral Pastoral Total Tigray 3,471,733 0 0 3,471,733 Amhara 15,003,844 0 0 15,003,844 Benishangul-Gumuz 572,882 0 0 572,882 SNNPR 13,021,540 374,491 0 13,396,031 Oromia 22,011,592 1,289,805 484,806 23,786,203 Gambella 110,113 118,925 0 229,038 Somali 584,929 998,792 2,177,992 3,761,713 Afar 0 148,897 1,051,432 1,200,329 National 54,776,634 2,930,909 3,714,230 61,421,773 Source: Atlas of Ethiopian Livelihoods (2010). Chapter 1: Pastoral Livelihoods   3 Gender Roles in Livestock, Crop Production, that are smuggled across the borders from Djibouti and Marketing and Somaliland is important. In Gambella, Afar, Women are actively engaged in livestock manage- and SNNPR, women produce handicrafts, such as ment and production in all regions. This includes wooden vessels, utensils, baskets, mats, artificial taking care of weak animals and providing tradi- fans, and brooms, as alternative sources of income tional veterinary services, herding small ruminants, (Adugna and Sileshi, 2013). Gambella women also fodder production (in SNNPR and Borena), milking sell local wine and alcohol as off-farm activities. In goats and cows, sewing and smoking of milk con- Borana, women who live near forested areas use tainers in Borana, and preparing dung for manure. forest products to complement their income. For Women frequently engage in the production and example, they produce a scented wood product marketing of small livestock, as well as in the that is used as a perfume, they collect gum Arabic processing and marketing of livestock products from trees, and sell incense. Sales of firewood and in all of the regions (Owitti, 2015; Nigussie, et al., charcoal are one of the main sources of income for 2014). This includes goats, sheep and chicken, milk both men and women in Somali and Afar. and milk products, and petty trade. Women are also dominant in the production and marketing Sources of Income of chicken in Gambella and East Hararghe (Ojulu, According to data collected by the International 2015). Although chicken production was not a tradi- Livestock Research Institute (ILRI),21 Somali and tional feature of Borana’s livelihood, it has recently Afar regions are overwhelmingly pastoral, with become one of the main income sources for Borana 71 percent of households in Somali region and women (Desta, et al., 2011). 62 percent in Afar reporting that they depend Men are generally dominant in the production largely or wholly on livestock for their income (see and marketing of large livestock, such as camels, Table 1.2). In Oromia and SNNP the proportions cattle, and donkeys. In parts of Southern Omo and are much less, with only 30 percent of house- in the Borana lowlands, cattle have a central posi- holds in SNNP and 20 percent in Oromia saying tion in the economy, and men are usually occupied they depend largely on livestock. In these last two with the cattle, whereas women are responsible regions some 45 percent of pastoralists report they for the production and marketing of small stock. also grow crops compared to only 14 percent in Afar In Somali, men are involved in milking camels, and 23 percent in Somali region. whereas this is mainly women’s work in Borana Despite the number of households that report and Afar. The herding of cattle and camels is the they also crop in Oromia and SNNP, cropping as responsibility of men and boys, while goats and a source of income appears to be only important sheep are boys’ responsibility. The slaughtering in SNNP, where it represents 28 percent of mean of cattle and camel is done by men, while women household income (see Table 1.3).22 Livestock con- slaughter goats in the Somali region. tinue to dominate sources of income in all the Both men and women engage in crop production, drought-prone regions and, according to the ILRI though to what extent differs between regions. Women mainly engage in sowing, weeding, trans- porting, storing, and harvesting of subsistence 21 The data were collected through a series of household surveys car- crops such as cereals and pulses, whereas men ried out at different times between 2015–2017, based on a stratified are actively involved in land preparation, plough- (by livelihood zone and regional state) approach using two stages of ing, weeding, harvesting, and threshing in crop sampling (by woreda and kebele) to draw a sample of 2,667 house- production. holds in Afar, Somali, and the lowlands of Oromia, SNNPR. The initial first survey covered 1,295 households and was analyzed as part of The main off-farm activity for women especially the baseline for the World Bank financed Regional Pastoral Resilience in Afar, Somali, and Gambella is petty trade, which Project—see Table 10. Subsequent surveys covered the remaining includes managing tea stalls, and selling coffee/tea households. The moist western lowlands of Gambella and Benishangul-­ Gumuz were not part of the survey. and biscuits. Due to their lack of access to finance, 22 While income from rain-fed agriculture is notoriously variable, it women favor these businesses because it requires might have been expected that income from irrigated cropping on the low capital, they are familiar with it, and the risks Awash River would have been higher, in particular, as it is technically to start up are low. For women in Somali and Afar, possible to produce two crops per year. It might be that the timing of the ILRI data collection coincided with the El Niño drought which, the contraband trade in secondhand and new cloth- among other impacts, resulted in particularly hydrological drought in ing, food items (e.g., sugar), and electronic items the Awash Basin. 4   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Table 1.2: Livelihoods of households (%) Region Proportion of households by livelihood type (A) (B) (C) (D) (E) A ‘pure’ pastoral A pastoral An agropastoral An agropastoral An agropastoral household with household household where household where household where no permanent where members members move members move members own residence, 100% move mobile mobile livestock mobile livestock livestock but do not income from livestock regularly regularly to agreed regularly to agreed move them from livestock and no to agreed grazing grazing areas grazing areas their private land crop growing and areas within the within the woreda, within the woreda, (i.e., livestock are no other income- woreda, with with milking/weak with milking/ kept in enclosures) generation activities milking/ weak animals kept around weak animals with crop growing animals kept around the settlement (i.e., kept around the and with other the settlement (i.e., a ‘satellite’ system) settlement (i.e., a income-generation a ‘satellite’ system), with crop growing ‘satellite’ system) activities with no crop and no other with crop growing growing and income-generation and with other no other activities income-generation income-generation activities activities Afar 51.0 11.2 10.5 3.1 24.1 Oromia (Borana)  5.6 13.8 43.9 1.6 35.2 Somali region 36.1 34.8 18.1 4.6  6.5 SNNPR (S. Omo &  0.0 29.4 40.9 4.8 25.0 Bench Maji) Total 23.6 22.2 28.3 3.4 22.5 Source: Gebremedhin et al., 2017 (based on ILRI survey data of sample of 1,295 households across the four regions collected in 2015–16 for the WB financed Regional Pastoral Resilience Project). Table 1.3: Mean share of total household income (%) (ILRI The ILRI data confirm that the government’s survey data) safety net program for the poor, PSNP, is an Afar Oromia Somali SNNP important source of income in Afar (14 percent) and Oromia (21 percent), but less so in Somali and Livestock income 67 63 52 43 in SNNP regions, although it is increasing in impor- Livestock product income 1 1.7 2.6 7.7 tance in Somali region. Wage employment is of Income from income 3 7 31 14 modest importance in all areas, but income gen- generating activities eration is particularly important among Somalis, Crop income 4.8 1 2.7 28 providing 31 percent of household income in the PSNP income 14 21 3 0.0 region but much less in other regions (see Table 1.3 Wage income 9.5 5.7 8.6 7.4 for breakdown). Household with share of 76.0 71.0 55.0 52.0 Overall, the ILRI data confirm, despite significant livestock income >50% (%) falls in household herd sizes, the continuing impor- tance of livestock in lowland economies. To understand how recent changes have affected data, represent 67 percent of mean household pastoralists, it is important to understand four income in Afar, 63 percent in Oromia, 52 percent critical characteristics of pastoral societies and in Somali, and 43 percent in SNNP regions (see economies. Table 1.3).23 (1) Dependence on Livestock 23 While important, the study also showed that the value of livestock This dependence on livestock has several sales was more significant for wealthier households, with an average of consequences. 70 percent of all household income coming from livestock. Income from livestock was much lower for the poorest quartile with only 17.5 percent It is a characteristic of pastoral capital that it and 28.8 percent of sales in SNNP and Afar regions, respectively. can reproduce itself without intervention of the Chapter 1: Pastoral Livelihoods   5 market. In pastoralism, capital gains are a normal for goats, 84 percent for cattle, and 71 percent part of the process of herd growth. Unless herd for camels between 2003 and 2015 alone (USAID, owners have viable alternative forms of investment, 2017). The higher livestock prices and lower staple the tendency is to reinvest in herd growth. A con- food prices helped offset herd losses and boost sequence of this is that without intervening factors terms of trade. such as drought and disease and/or high market However, poor pastoralists with only a few ani- offtake rates, livestock populations may eventually mals cannot take full advantage of the improving exceed the carrying capacity of the land. It also terms of trade between livestock and grain. Their leads to the typical saw tooth graph of herd growth market participation both as buyers and sellers of peaks and troughs coinciding with good and bad is extremely limited (see Table 1.4). For them, their rainfall years that is commonly recorded in pastoral market engagement tends to be limited to emer- areas. gencies, such as droughts, when they are forced Because pastoralism is geared toward herd to sell livestock. At other times they depend more reproduction, there is an inevitable surplus of ani- on other sources of income to buy food and other mals, such as most of the males which can be sold essential goods. On the other hand, rich pastoral- or eaten without affecting the reproductive capac- ists can take advantage of the fluctuating live- ity of the herd. This feature encourages the sale stock prices by selling when prices are good and and/or exchange of surplus stock/livestock com- buying when prices drop due to high supply during mercialization if the price is right. or immediately after droughts (Tari and Pattison, Herd reconstitution after drought is a long and 2014). See Table 1.4 for market engagement (sales slow process because many of the female reproduc- of livestock) by wealth categories in northern tive stock will have died. If a farmer loses his crop Kenya and southern Ethiopia between 2000 and one season he can plant again the next, but a pas- 2002. toralist will take much longer to recover his herds and flocks. Indeed, below a certain threshold, it is Table 1.4: Market engagement by wealth group, northern unlikely he can ever recover his losses. Therefore, it Kenya and Southern Ethiopia, 2000–2002 pays to have large herds and flocks in good times in preparation for the inevitable bad times that may Mean livestock holdings Number of quarters when per household market activity recorded follow. Pastoralists are vulnerable to fluctuations in the 10.8 TLU 0 terms of trade when milk yields decline in the dry 16.0 TLU 1 season and at times of drought when they need 45.9 TLU 7 to buy grains, but their livestock hold little value. 48.3 TLU 8 At these times they risk losing their reproductive Source: Aklilu and Catley, 2009. stock. In recent years, given increasing national and international demand for livestock, livestock prices have seen an upward trend in relation to Reliance on diverse income sources varies by grain prices, which has helped many pastoralists wealth group. While most wealthy pastoralists earn compensate for the long-term decline in herd sizes their income largely from the sale of livestock, this (HEA, Somali and Afar Change Reports, 2017/18, is not the case for the poorest category of pastoral- and Chapter 2). ists and agropastoralists. For example, among the In the past, the unfavorable terms of trade Borana/Guji pastoral and agropastoral groups, the between livestock and grain disproportionately very poor earned their cash income from livestock disadvantaged pastoralists and agropastoralists. (58%), labor (12%), firewood (8%), and safety nets Davies and Bennett (2007) observed that terms (22%), unlike for middle and better-off groups who of trade fluctuated between a high of 183 kg of generated almost all cash income from the sale of maize per goat to a low of 93 kg per goat (aver- livestock (Aklilu and Catley, 2009). In Shinille zone aging around 133 kg per goat) in 2001 in Afar. The of the Somali region, similar patterns exist in terms following year it dropped to 14 kg of maize per goat. of the contribution of livestock to annual household In the past decade this has significantly improved income (Table 1.5). in favor of livestock producers, at least in ‘normal’ As Table 1.5 shows, households in the better-off years. A recent USAID report from the Afar region and medium category derive almost all of their shows that terms of trade improved 96 percent annual income from livestock-related sources, such 6   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Table 1.5: Contribution of livestock to annual household income Afar, Oromia (Borana zone), Somali, and SNNP in in pastoralist areas, Shinile zone 2016/17. These numbers represent a significant Total annual income (EB) from decline over the years. livestock/total income (%) A study by Berhanu (2017) in Borana zone based Wealth group 1998–1999 2004–2005 on a panel subsample of pastoral households tracked from 2003–2013 shows the scale of the Better-off 6,050/7,000 (86%) 9,300/9,300 (100%) drop in livestock per household over time. According Medium 3,978/4,350 (91%) 5,800/6,200 (94%) to this study, more than 90 percent of previously Poor 1,980/2,200 (90%) 2,200/3,950 (56%) asset-poor households remained in poverty, with Source: Catley and Iyasu, 2010. a further deterioration in their situation, while a significant percentage of nonpoor stock owners slipped into poverty. Other studies among Borana as livestock sales, renting of pack camels, and sale and Karrayu pastoralists also indicate a decline in of ghee. Poor households fill the deficits in annual household livestock assets over time (see Table 1.7). income through gifts, relief food, wild foods, and The study by Elias (2014) shows that both other income generating activities. among the Borana and Karrayu pastoralists, live- stock substantially declined over a thirty-year Decline in household herd size period, making households more vulnerable to food There has been a long-term decline in household insecurity and fluctuating terms of trade. Similar herd size. Table 1.6 shows the number of live- herd size declines among all wealth groups have stock of different types owned by households in been reported for the Afar and Somali in a series Table 1.6: Access to livestock asset in TLU by region Access to livestock asset Afar Oromia Somali SNNP All Households owning livestock (%) 91.3 93.6 90.6 82.7 90.5 Cattle 39.8 86.8 43.8 74.4 59.4 Sheep 54.5 19.8 55.6 31.5 41.5 Goats 86.4 68.8 76.2 52.9 73.5 Camels 34.0  7.4 32.0  0.0 20.8 Equines 30.0 43.9 56.6 11.8 38.9 Chicken  0.4 36.9  4.7 59.3 20.5 Livestock owned per household (TLU)  7.3  4.5  6.3  5.0  5.9 Livestock owned per capita (TLU)  2.3  0.9  1.4  1.0  1.5 Source: ILRI survey data, 2016/17. Table 1.7: Trends in household livestock ownership in Borana and Karrayu Location/wealth Mean livestock numbers in the past (30 years ago) Mean livestock numbers at present group Cattle (no.) Camel (no.) Cattle (no.) Camel (no.) Borana: Rich  94  12 74 12 Medium  39   9 23  6 Poor  22  10  6  3 Mean  30  11 12  5 Karrayu: Rich 176 111 36 30 Medium 108  79 19 23 Poor 109  35  6  5 Mean 133  79 12 16 Source: Elias (2014). Chapter 1: Pastoral Livelihoods   7 Table 1.8: Mean livestock holdings, by livestock holding quintile, region, and year in tropical livestock units Livestock Afar Somali holding quintile 2010 2012 2014 2016 2010 2012 2014 2016 Poorest  4.8  1.5  0.8  0.2  1.1  0.5  0.6  0.2 2nd 13.7  4.9  2.6  1.4  4.2  3.0  3.6  1.5 Middle 25.5  8.8  5.1  3.3  7.7  7.1  7.2  3.5 4th 42.4 18.0 10.1  6.6 14.6 14.6 12.7  6.1 Wealthiest 78.5 61.7 35.6 18.9 36.7 38.7 27.5 17.1 Source: IFPRI/EDRI, 2018 Targeting Social Transfers in Pastoral Societies-Household Surveys for PSNP. Table 1.9: Changes in wealth ranking of Somali pastoral households according to livestock holding over a 60-year period (1944–2004) No. Wealth ranks overtime Cattle Sheep Goats Camels Donkey Total % 30-year period before 1974 1 Wealthy households 400 200 250 50 20 920 56.6 2 Medium households 200 100 150 20 10 480 29.5 3 Below medium households 80 50 80 10 5 225 13.9 30-year period after 1974 1 Wealthy households 100 350 500 120 10 1,070 63.3 2 Medium households 50 150 300 60 5 565 33.6 3 Poor households 3 10 22 1 2 38 2.2 4 Very poor households 0 5 12 0 1 18 1.1 Changes over 60 years Decrease/increase –527 +165 +354 +101 –17 +66 — Percent change +/– –77.5 +47.1 +73.7 +126.2 –48.6 +4.0 — Source: Gezahegn, 2006. of baseline surveys carried out over the period (Mussa, 2004). It appears that this has been a 2010 to 2016 for PSNP (see Table 1.8). According to trend since the 1960s, especially in Borana. Holt- these surveys, between 2010 and 2016 herd sizes land (2011, 6) reports that Borana had 7.5 TLU24/ declined by more than three-quarters from 78 TLU capita in 1960, and this had declined to 4.5 TLU in to less than 19 TLU among the wealthiest quintile 1990 and more recently to less than 3 TLU per cap- in Afar, and among Somali by more than half from ita. According to Desta and Coppock (2002) median 37 to just over 17 TLU. In the poorest category the household herd size was 58 cattle in 1980/81 but decline was even more dramatic, falling from 4.8 to only 29 in 1996/97, having dipped as low as 22 cat- 0.2 TLU and 1.1 to 0.2 TLU among Afar and Somali tle per household in the drought year of 1992/93. respectively, effectively forcing this group to exit pastoralism altogether. Wealth differentiation Devereux (2006, 46) reported from his study of In Somali Region, Gezahegn (2006) assessed the Somali region that almost all his respondents wealth ranks and livestock species kept by house- “claim that the numbers of animals they own today holds over 60 years in Shinille zone and showed is fewer than in the past.” A recent study comparing changes in both indicators (Table 1.9). He divided household economy changes between 2003/06 the period into two by taking 1974 as a reference and 2013/15 in Somali region similarly reports a year and compared changes over 30 years prior to decline in household herds by 25 percent between and after 1974. The result showed drastic changes the two reference years (HEA, 2018). A study con- ducted in Borana area also shows that the average 24 Tropical livestock units are a measure of livestock holdings in which livestock holding per household decreased by 37% different livestock types are converted to a common unit: cattle = 1, over 17 years prior to 2001, mainly due to drought sheep and goats = 0.1, and camels = 1.4. 8   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands in livestock ownership and the type of species kept in the former and decline in the latter. Following due to increased drought frequency. According to this shift and progressive loss of livestock assets, Gezahegn, the frequent droughts caused shrink- many households adopted agriculture as a coping age of the grasslands and the expansion of woody mechanism. plants, which favored camels and small ruminants. Similarly, based on his study of pastoralists The average species composition of livestock hold- in Northern Kenya, Fratkin (2001, 9) reported ing per household changed, with camels increasing that “Increased commoditization of the livestock by 126.2%, goats by 73.7%, and sheep by 47.1%, economy has led to a growing polarization of pas- whereas the number of cattle and donkeys declined toralists into haves and have-nots, particularly by 77.5% and 48.6%, respectively (Gezahegn, 2006, in . . . areas that are close to urban markets.” The 43). Over the period, while the species composi- social, political, and economic challenges that are tion of household herds changed, there was an responsible for greater livelihood diversification are increasing shift of livestock holdings in favor of the also responsible for increased social and economic wealthy, with the poor getting distinctly poorer in stratification, urban migration, and diminished terms of their livestock asset holdings. nutrition for women and children (Fratkin, 2001). This growing gap between the few emerging The wealthy often diversify to expand their eco- wealthy (and commercial) herders and the grow- nomic opportunities while the poor diversify for sur- ing number of stockless and poor pastoralists and vival. According to Aklilu and Catley (2009) wealth agropastoralists has been reported on by a vari- disparities have increased markedly in Somali ety of commentators (Little et al., 2015; Fratkin, region. They estimate that over a 10-year period 2001; Fratkin and Mearns, 2003; Aklilu and Catley, between 1996 and 2005, there was an increase 2009). Little et. al. (2015, 391) found greater wealth of 2.5 percent in the number of wealthy house- differentiation among the pastoral groups in the holds, 0.8 percent in the number of middle wealth Horn of Africa and showed that the wealthiest households, and 4.1 percent in the number of poor 10 percent of herd owners control about 46 percent households. “If this trend continues within 20 years of aggregate TLUs, while the poorest 20 percent the number of poor households will double, whereas control less than 3 percent. Holtland (2011), who the number of wealthy households will increase by observed an increasing gap between the wealth about 50%” (Aklilu and Catley, 2009, 22) groups in Borana, also stated: This growing inequality between households is confirmed by data from the ILRI surveys referred to “. . . in 2007 only 7% owned 35% of all cattle above.25 Based on livestock ownership as a critical and the poorest 70% owned only 30%. Thirty dimension of wealth, Figure 1.1 shows the share of years before this was respectively 21% and TLU’s per household by wealth quartile across the 50%. For goats the poorest 70% still owns 50% four regions. The large share of livestock owned by of the herd (down from 66%). Today probably the wealthiest quartile is consistent across all four only 25–33% of all households have a herd regions and is confirmed by the findings of other that can sustain their families” (p. 6). case studies. Tufts University in their research on pastoralism The study by Tache and Oba (2010) in the Borana in the Horn of Africa, including case studies in Ethi- area also shows that there were considerable shifts opia, noted: in individual wealth ranks over time, while the desti- tute category remained unchanged. Using the tra- “In 2009, research . . . explained the increas- ditional Borana categories of wealth, the authors ing domestic and international livestock classified households into wealthy (duuresa ciccitaa), trade from some pastoralist areas, and yet rich (dureessa), self-reliant (nama ufirraa bulu), tran- at the same time, the increasing destitution sitional (harka qalleessa—‘has weaker capacity’), and levels of humanitarian assistance. The poor (deega), very poor (deega bombii), and destitute findings showed that although most pasto- (qollee guutuu hiikanaa), each category respectively ralists were engaged in livestock markets, owning 179, 58, 30, 11, 7, 5, and less than 2 TLU, respectively. The study indicates that the greatest shifts occurred in the numbers of the very poor and 25 The ILRI survey was carried out after the 2015–16 El Niño drought in the very wealthy categories after 1992 due to fre- the northern pastoral areas, and it coincided with the 2016–17 Indian quent droughts, leading to a progressive increase Ocean Dipole drought in the southern pastoral areas. This may well have depressed the size of herds reported in the four regions. Chapter 1: Pastoral Livelihoods   9 Figure 1.1: Distribution of livestock wealth by household quartile and by region (based on ILRI survey data 2015–17) .682 SNNP 11.3 23.4 64.6 .57 2.28 Somali 10.8 22.2 64.7 .558 3.94 Gini Afar 14.1 25.4 56.5 .47 Oromia 4.52 14.9 25.9 54.6 .448 2.82 Total 12.7 23.7 60.8 .517 0 20 40 60 80 100 Outcome share (percent) 0–25 25–50 50–75 75–100 commercialization was associated mainly are few and far between. Pastoral areas, while they with wealthier pastoralists and herd owners. may produce crops in good years, are generally Over time, and as human populations grew, marginal to intensive crop production. Extensive commercialization was also linked to a grad- livestock production is often the only way to make ual shift of livestock from poorer to wealthier use of the fluctuating forage resources of these producers, and related trends such as increas- environments. ing privatization of rangelands, and declining The variability and low productivity of pastoral social capital related to livestock transactions environments has imposed certain constraints and support. These changes were evident in on livestock production and the nature of human parts of Ethiopia, Kenya, and Somalia despite adaptive strategies in these areas. diverse central political ideologies and policies Livestock are bred for their resilience to drought over several decades, and within a context and disease rather than their productivity. Growth of multiple conflicts and frequent droughts. and fertility rates are generally poor, and in addi- This research became known as the “Moving tion, herd composition is adapted to the nature and Up, Moving Out” analysis, and notably, the availability of forage. Over the last few decades findings were consistent with other studies as bushy species have invaded many rangeland on poverty and livestock development, the areas, there has been a general shift from grazers economics of pastoralism, and frameworks (cattle and sheep) to browsers (camels and goats), for understanding future trends. Although which are better able to exploit these invasive spe- not widely recognized at the time, Moving cies. Many Boran cattle pastoralists in the Moyale Up, Moving Out also concluded that for many area of the southern rangelands, for example, have poorer households, future livelihoods were not adopted camels as an adaptive response, both to only a matter of moving out of pastoralism, increased drought and to bush encroachment in but also of moving out of pastoralist areas” the area. (Catley, 2017, 1). Research in Borana zone in the 1980–90 found that when cattle populations exceed a threshold of 30 cattle per square km, increased mortality (2) Variable, Semiarid and Drought-Prone becomes more likely in years when annual rainfall Environment is less than 400 mm/year.26 In the droughts of These environments are characterized by the 1983–85, 1989–90, and 1991–93, cattle mortality ­ extreme variability and unreliability of rainfall, both between different years and between different 26 Coppock, L., Gebru, G., Desta, S., Mesele, S., and Tezerra, S. (2008). places in the same year, by the scarcity and sea- Are Cattle Die-Offs Predictable on the Borana Plateau. Environment and Society Faculty Publications. Paper 212. Utah State University. Global sonal variability of vegetation and water, and by Livestock Collaborative Research Support Program. https://digital vulnerability to drought. Permanent water points commons.usu.edu/envs_facpub/212 10   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands rates of 40 percent were recorded. Pastoralists in This is because poorer households are less mobile— two kebeles in Dillo woreda estimated that cattle they are dependent on alternative sources of food losses in the 2016–17 drought rose to as much as and income that are typically available only in or 70 percent, while agropastoralists in the neigh- near to small urban centers—and therefore their boring Dirre woreda estimated cattle losses in the livestock have access only to forage resources 2010–11 drought to exceed 85 percent.27 While pas- around the towns that are grazed throughout toralists’ reflections should not be confused with an the year. Wealthier households have the labor accurate representation of losses, they are proba- and capacity to move their herds to less severely bly valid illustrations of the scale of losses. Hence, drought affected areas, including neighboring the higher the reported losses, the more serious the regions and even countries.32 Wealthier pastoralists drought. The reported losses therefore suggest that also maintain stronger social networks that help the ‘die-offs’ of the 1980s and 1990s have contin- buffer them against drought impacts33 through the ued to the present. This is confirmed by zonal offi- provision of additional labor, livestock exchanges, cials with reports of the loss of 200,000 cattle in and facilitation of higher levels of mobility through the 2010–11 drought and 350,000 in the 2015–16 herd splitting. Other coping strategies include: drought. The pre-drought value of the cattle lost in diversification of livestock species—more than the 2015–16 drought is estimated to be more than 50 percent of households keep at least three types US$65 million.28 High cattle mortality rates can be of livestock; investing in herd growth; the purchase expected to continue in Borana as large areas of of livestock feed; destocking of surplus animals; savannah grassland continue to be lost to enclo- and accessing local social safety nets.34 sure and invasive thorny, inimical scrub to cattle Despite high levels of livestock mortality in many pastoralism. pastoral areas due to drought, significant loss of Pastoral household health and well-being is human lives during these periods has been avoided extremely vulnerable to prolonged drought. Drought due to significant improvements in the delivery of causes consumption losses and loss of assets in humanitarian assistance and/or scaled up safety pastoral areas. A study in the drought-prone low- net programs like the Productive Safety Net Pro- land areas of Ethiopia confirms that a moderate gram (PSNP). According to ILRI data35 more than drought may reduce consumption by 8 percent,29 65 percent of pastoral households reported receiv- rising to 20 percent in a prolonged drought.30 While ing food aid during recent droughts. In Somali and droughts in pastoral areas have similar impacts on Afar regions the share of PSNP and humanitarian household consumption, these impacts are typically beneficiaries to the total regional population has different from consumption changes in drought- consistently been near to or greater than 30 per- affected agrarian communities. cent since 2009, and in 2016 and 2017 reached Pastoral households suffer different levels of loss above 50 percent. High levels of external assistance according to wealth, with higher losses typically appear to be the new norm for many communities recorded by poorer, more sedentary households.31 in the lowlands. Because of the variability of rainfall and there- fore grazing, a degree of mobility is enforced on 27 Information collected using the Participatory Rural Appraisal propor- tional piling method. 28 Tufts University (2017). The 2017 East Africa Drought: Drought and 32 Pastoralists interviewed in Arbale kebele, Dillo woreda, Borena zone resilience perspectives from southern Oromia. Field Notes. Agricul- reported that in the 2016-17 drought they trekked their livestock as far ture Knowledge Learning Documentation and Policy (AKLDP) Project as Samburu District in northern Kenya. Ethiopia. 33 Little, P., McPeak, J., Barret, C., and Kristjanson, P. (2005). The Mul- 29 World Bank (2015). tiple Dimensions of Poverty in Pastoral Areas of East Africa. Paper pre- 30 Porter, C. (2012). Shocks, Consumption and Income Diversification in sented at ‘Pastoralism and Poverty Reduction in East Africa: A Policy Rural Ethiopia. Journal of Development Studies, Taylor and Francis Jour- Research Conference’, June 27–28, 2006 in Nairobi, Kenya. World Bank nals. Vol. 48 (9). Cited in World Bank (2015). and World Bank’s Arid Lands Resource Management Project (ALRMP), 31 A study in pastoral areas of northern Kenya classified pastoral USAID’s Global Livestock Collaborative Research Support Program households as: wealthy with 4.5+ tropical livestock unit (TLU) per (GL-CRSP) and Strategies and Analysis for Growth and Access Pro- household member; medium with 2–4.49 TLU category; poor with gram (SAGA), and the International Livestock Research Institute (ILRI). fewer than 2 TLU; and very poor with under 1 TLU. Little, P., McPeak, J., 34 Catley, A., Lind, J. and Scoones, I. (2016). The Futures of Pastoralism and Kristjanson, P. (2008). Challenging Orthodoxies: Understanding in The Horn of Africa: pathways of growth and change. Rev. Sci. Tech. Poverty in Pastoral Areas of East Africa. Development and Change Off. Int. Epiz., 2016, 35 (2). http://doc.oie.int:8080/seam/resource/ 39(4): 587–611 (2008). Institute of Social Studies 2008. Published by directMedia/_jU2DnHa7TNgYWICszidt0yVl-KUH6GJ;jsessionid= Blackwell Publishing, 9600 Garsington Road, Oxford OX4 2DQ, UK and 82b105cb5607b2ab935984979dea?binaryFileId=13875&cid=68 350 Main St., Malden, MA 02148, USA. 35 ILRI Livestock Surveys 2016–2017 (WB/ILRI analysis for this study). Chapter 1: Pastoral Livelihoods   11 pastoralists and their livestock. Human and live- Devereux (2006) reported that Somali pasto- stock mobility is a response to seasonal variations ralist and agropastoralist households on average in forage and water availability. The degree of this owned 9.7 camels, 7.7 cattle, 1.7 goats, 1.5 sheep, mobility and the extent to which it affects people, and 0.8 donkeys. In TLU, each species respec- as opposed to their livestock, varies considerably. tively constituted 46 percent, 34 percent, 8 per- A common pattern among Boran pastoralists cent, 7 percent, and 3 percent of the household is for livestock to be herded in mobile livestock herd. The corresponding share of livestock species camps by the young men, while the women, chil- recorded by Gebremedhin et al., in 2017 is 27 per- dren, and older men stay in more settled villages. cent for camels, 17 percent for cattle, 31 percent for Among Somali camel pastoralists households goats, and 24 percent for sheep for the region (see tend to be more mobile and shift more regularly. Table 1.10). Access to dry season grazing areas of higher rain- In other pastoral areas similar trends toward fall potential or along riverbanks is vital to most increasing numbers of small stock are reported. forms of pastoralism. As parts of rangeland areas For example, in 1988 Cossins and Upton reported are privatized and enclosed, either because of the that small stock comprised about 7.4 percent of the development of commercial private or govern- TLUs on the Borana plateau. More than 25 years ment agricultural plantations in areas of higher later, McPeak et al. (2015) reported a 76 percent potential, or because of the actions of wealthier cattle, 16 percent small stock, and 8 percent camel pastoralists themselves to enclose grazing areas, composition for the Borana, which is a significant the overall adaptive strategies of pastoralists are shift toward small stock and camels. Getachew affected. (2001) argued that the shift in livestock compo- Herd diversification is common, with many sition in Afar from grazers to browsers occurred wealthier pastoralists herding a variety of differ- in response to the loss of riverine land to irrigated ent stock in different areas, partly to spread risk. agriculture. A recent report from Afar compares This is labor intensive and often requires employ- two baseline surveys over a decade apart and indi- ing herders, which only the wealthiest can afford. cates that household herd composition has changed Not only do different animals have different niche in some livelihood zones because of a changing specializations but they have different vulnera- resource base and reduced rainfall. “The typical herd bilities to drought. As pastoralist herds become now contains more browsers (goats and camels) and depleted, the tendency is to specialize in small fewer grazers (sheep and cattle)” (USAID, 2017, 56). stock pastoralism rather than the herding of larger Research carried out in northern Kenya sug- animals (cattle and/or camels). The risk of losing gests that below a threshold of 4.5 TLU per capita, small stock is much less than large stock because pastoral households without alternative income their unit value is less to start with, and they can sources will struggle to provide adequate food and reproduce at a much faster rate than large stock, income for family members from their herds and so they provide the fastest re-entry back into flocks. Households of fewer than 2 TLU are typi- pastoralism. cally caught in a cycle of poverty from which they Table 1.10: Changes in household livestock composition in TLU by livelihood type for Somali region (%), 2006 to 2017 Household livestock composition in TLU (%) Period Livelihood type Camels Cattle Goats Sheep Donkeys All Pastoralist 57 19 11 9 3 100 2006 Agropastoralist 35 50 5 5 4 100 Average 46 34.5 8 7 3.5 100 Pastoralist 28 15 32 25 — 100 2017 Agropastoralist 26 20 30 24 — 100 Average 27 17.5 31 24.5 100 % change over Pastoralist –29 –4 +21 +16 — 10 years (+/–) Agropastoralist –9 –30 +25 +19 — Source: Adapted from Devereux (2006); Gebremedhin et al. (2017). 12   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands are unable to exit, even in good rainfall years.36 In Recurrent shocks may not only delay full recov- the ILRI survey referred to above, the average TLU ery but may also prevent it altogether, as con- per household in 2016/17 ranged from 7 TLU in firmed by ex-pastoralists in Yabello. While some Afar to 4 TLU in Oromia (see Figure 1.2). As would had once been quite wealthy, they became pro- be expected, richer households own more than the gressively impoverished as the result of recurrent mean, but what is surprising is the percentage in drought and were forced to abandon pastoralism the poorest quartile—43 percent (Afar), 75 per- and migrate to Yabello. Many were destitute. They cent (Oromia), 69 percent (Somali) and 46 percent also noted that once a household has moved to a (SNNP)—who owned no animals at all and are town, it is extremely difficult if not impossible to either wholly dependent on their relatives who con- accrue adequate livestock to return to pastoralism. tinue to own livestock or have moved out of pasto- The longer they stay in the towns, the smaller their ralism altogether. social networks.40 Communal ownership of the range is an adap- Figure 1.2: Average TLU per household by region (Ethiopia) tation to variability. Without access to a variety 8 of grazing areas, no pastoralist can be assured 7 that his livestock will find grazing when needed. As 6 Ethiopia’s rangelands have become increasingly TLU per household 6 5 fragmented with important parts excised for com- 4 4 mercial plantations or enclosed as a result of crop expansion and growth of permanent settlements, 2 common access to grazing is becoming increasingly constrained. Newly created political boundaries 0 between Ethiopia’s regions since 1995 have also Afar SNNP Somali Oromia imposed new restrictions on access to traditional Source: ILRI (2017). pastures and well complexes for some groups. Protecting traditional pastoralist common Herd rebuilding after drought takes time. For access regimes is vital to pastoralist livelihoods households that have lost in excess of 40 percent remaining viable, given increased rangeland frag- of their livestock, this recovery will take years: for mentation experienced by many pastoralist groups sheep and goats the recovery may take 4–7 years; today. Economic planning, in particular of land use, cattle 7–9 years; and camels up to 13 years.37 needs to take into account the potential impact During such herd rebuilding phases, pastoralists on pastoralist livelihoods. The scope for livestock minimize consumption demands and avoid live- intensification as population density increases is stock sales as much as possible, as such losses limited. Excess population cannot be absorbed by simply further extend recovery periods.38 It is existing production technologies. The inevitable however necessary to cover basic consumption consequence of a growing population trying to demands and meet human and animal health costs eke out an existence on a diminishing area of land and school fees, and hence all households are peri- (with the same level of technology) is likely to lead odically forced to sell animals. Poorer households to increasing competition between groups, lower typically select immature animals for which they do economic returns, and increasing poverty for those not receive the full benefits.39 Full recovery is also with few livestock assets and no obvious alterna- inevitably dependent on subsequent weather- tive economic opportunities. To date, pastoralists related and other shocks. have benefited from increasing livestock prices and improving terms of trade at the same time as herd sizes have declined. But in the future more and 36 Ibid. more people will need to look outside pastoralism 37 Toulmin, C. (1994). Tracking through Drought: Options for Destock- and pastoral areas for their livelihoods. This will ing and Restocking, in Scoones, I., ed., Living with Uncertainty: New Directions in Pastoral Development in Africa, London: Intermediate require much more emphasis than before on human Technology. capital development and out-migration to areas of 38 McPeak, J., and Little, P. (2017). Applying the concept of resilience to higher employment. pastoralist household data. Pastoralism Research, Policy and Practice, https://doi.org/10.1186/s13570-017-0082-4 39 Wealthy pastoralists therefore typically benefit from investment in livestock marketing in pastoral areas. 40 Discussions with ex-pastoralists in Yabello, August 2018. Chapter 1: Pastoral Livelihoods   13 (3) Remote Geographic Location been viewed by highlanders as an outmoded form A third important characteristic of pastoral areas is of production, and government pastoral policy has their remoteness from centers of population, many been largely driven by the desire to settle pastoral- basic services, and large urban markets. This has ists and turn them into farmers or commercial live- affected the costs of goods and services, which stock producers. There has been little appreciation tend to be higher than in more densely populated of the finely balanced adaptation of pastoralists areas. Relief interventions are also costlier. Because to a difficult environment and the validity of their of their remoteness and distance, infrastructure overarching objective to maximize the numbers of and communications are generally poorly devel- their livestock to sustain as many people as possi- oped, and services limited. While this is changing in ble in a highly variable and marginal environment. some lowland areas as result of new infrastructure The early livestock development projects in the investments, for example, the new all paved road 1960s and 1970s were based on the ideas that pas- from Jijiga to Gode has resulted in a considerable toralists must keep less livestock, exchange more increase in commercial traffic between the two, in livestock, and collectively or individually regulate many lowland areas the orientation of pastoralists access to pasture and/or change their traditional is still outwards across national borders rather than land tenure system. Since then, frequent emer- inward toward the Ethiopian highlands. This has gency relief interventions in pastoral areas after important implications for trade and markets, as drought have compounded the widespread view large numbers of livestock are traded across the that pastoralism is in a crisis. border to Kenya and Somalia/Somaliland without As this study shows this view is not supported passing through official channels. This has led to by the evidence. There are a variety of reasons for frequent prohibitions on such trade by the Ethio- the vulnerability of lowland populations, and pasto- pian authorities. ralists are not the poorest of the poor but do suffer from severe human capital deficits and may be (4) Political and Economic Marginalization particularly vulnerable to shocks. Traditional pasto- ralism in Ethiopia is changing toward an increasing from the Ethiopian State integration of livestock herding and cropping, and a A fourth characteristic of pastoral areas and greater degree of commercialization driven by high lowlands in general is their political marginaliza- national and global demand for livestock. However, tion. These areas, historically, have formed the this brings with it its own attendant risks as pas- borderlands and frontiers of the Ethiopian state. toralists become less mobile and pastoralism itself Many pastoral groups were only incorporated into becomes an increasing preserve of the wealthy, Ethiopia in the late 19th century during the reign as poorer pastoralists are pushed out to become of Emperor Menelik. The relationship between pas- full-time croppers and/or to pursue other eco- toralists and the Ethiopian state has therefore been nomic opportunities from petty trade to charcoal ambivalent at best, and at worst openly hostile production. (Markakis, 1993). In general, mobile pastoralism has 14   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Poverty and Vulnerability in the Ethiopian Lowlands Chapter 2: Poverty and Vulnerability Summary Introduction This chapter examines changes in monetary and This chapter analyzes the trends in expenditure-­ nonmonetary poverty in the lowlands between 2011 based poverty in the Ethiopian lowlands, with the and 2016. The analysis indicates a reduction in both objective of having a better understanding of the dimensions of poverty from an initially high level. Those factors behind the changes over time in the poverty engaged largely in livestock production (pastoralists) rate in the lowlands. Considering that monetary seem to have experienced large decreases in the pov- poverty is only one of many dimensions of welfare, erty headcount as well as declines in the depth of pov- an effort is made to cross-validate the observed erty. On the other hand, agropastoralists seem to have trends in expenditure-based poverty with trends borne the brunt of the drought and have increasingly in other non-monetary dimensions of poverty. transitioned, presumably due to emergency livestock Attention is also paid to the potential effects of the sales or deaths, to being full-time farmers (cropping large inflow of humanitarian food aid distributed only). These new farmers have not done well as evi- in response to the 2015–16 drought that affected denced by the increase in the depth of poverty among many areas in the lowlands, on the observed trends agropastoralists and crop producers. Increases in the of poverty. price of livestock accompanied by decreases in the In a context such as that of the lowlands, char- price of cereals combined to reduce monetary poverty, acterized by repeated droughts and a high degree even in the case of smaller herd sizes and increasing of volatility in consumption, a household’s currently asset-based poverty. Our analysis also reveals that observed poverty status is unlikely to be a very in the lowlands: (i) the proportion of the vulnerable good guide for a household’s vulnerability to pov- population is more than two times the proportion of erty, or its “poverty risk.” For this reason, a measure the poor population in the lowlands; (ii) vulnerability of vulnerability to poverty is also estimated for due to aggregate shocks such as droughts is relatively comparison with the standard headcount poverty more important than vulnerability due to idiosyncratic rate, and a concerted effort is made at understand- shocks; and (iii) poverty-induced vulnerability is rela- ing the sources of vulnerability in the lowlands. tively more important than risk-induced vulnerability, The basic sources of data used in this chapter which is in sharp contrast to the other zones where are the Ethiopian government’s Central Statisti- the relatively more important sources of vulnerability cal Agency’s, Household Consumption Expendi- to poverty is high consumption volatility. This points ture Surveys (HCES), and the Welfare Monitoring to the unique nature of the drought-prone lowlands Surveys (WMS) in 2011 and 2016. It is important in comparison to the other agroecological zones in to bear in mind that there are some serious data the country and the need for policies and programs limitations associated with the HCES and WMS tailored to these special features. surveys, especially when it comes to assessing the 15 Map 2.1: Ethiopia’s four agroecological zones level and the trend in the welfare of pastoralists household expenditures per adult equivalent and agropastoralists in Ethiopia. Prior to 2016, against the nominal (or inflation-adjusted) poverty pastoral areas were underrepresented in the HCES line per adult equivalent (AE) (3,781 Birr per AE in since largely pastoralist zones were not included 2011, and 7,184 Birr per AE in 2016). in the surveys.41 In 2016 some additional zones Figure 2.1 reveals that poverty headcount rates were included in the HCES/WMS survey in Afar and in each of the eight regions of Ethiopia has declined Somali regions to increase the representativity of significantly over the last 16 years (between 2000 the surveys in pastoral areas.42 To ensure compara- and 2016). In 2000 there was a considerable diver- bility, all comparisons between 2011 and 2016 are sity in the prevalence of poverty across regions based on the zones in Afar and Somali regions that ranging between 38 percent in the Somali region were surveyed in both 2011 and 2016 (these are and just under 63 percent in the Tigray region. By zones 1, 3, and 5 in Afar and 1, 2, and 9 in Somali). 2016 the poverty rates in all eight regions ranged In order to allow comparisons between agro- between 20 and 28 percent, exhibiting a rapid ecological zones, the country is divided into four convergence to a lower headcount poverty rate. agroecological zones: the drought-prone and pas- With the unique exception of the Somali region toral lowlands, the moisture reliable lowlands, the where poverty seems to have increased tempo- drought-prone highlands, and the moisture reliable rarily between 2000 and 2005, poverty in all highlands following EDRI, 2009 (see Map 2.1). other regions has either declined rapidly or did not increase from survey to survey. A likely explanation Expenditure-Based Poverty for the increase in poverty in the Somali region may An estimate of expenditure-based poverty is be due to the regional ban by Saudi Arabia and derived by comparing total (food and nonfood) the Gulf states on livestock imports from Soma- lia, Ethiopia, Eritrea, Kenya, Sudan and Djibouti because of the outbreak of Rift Valley Fever (RVF). 41 The nomadic lifestyle of pastoral populations makes it difficult to The import ban was imposed in 2000 and ulti- identify the exact location of pastoral households in advance and thus mately lifted in 2003. very difficult to send enumerator teams to carry out the survey. 42 In 2016 the sampling of pastoral households was based on sending A comparison of the change in the poverty enumerators to boreholes and birkedas where pastoralists are known to headcount rate across agroecological zones visit during the seasonal migration of their herds in search of pasture. between 2011 and 2016 reveals that the decline 16   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Figure 2.1: Overall decline in headcount poverty by region, 2000–2016 (HCE surveys) 68 63 58 53 48 43 38 33 28 23 18 2000 2005 2011 2016 Tigray Afar Amhara Oromia Somali Benishangul-Gumuz SNNP Gambella Source: Bank Staff Estimates based on the respective HCES and WMS. Figure 2.2: Headcount poverty rate by agroecological zone: 2011 vs. 2016 (HCES) 0.40 0.37 0.35 0.30 0.30 0.30 0.27 0.28 0.25 0.25 0.25 0.24 0.23 0.21 0.20 0.15 0.10 0.05 0 Drought-prone Humid moisture Drought-prone Moisture National lowlands and reliable highlands reliable pastoral areas lowlands highlands 2011 2016 Source: Bank staff estimates based on the 2011 and 2016 HCES and WMS. Note: All comparisons between 2011 and 2016 are based on the zones in Afar and Somali regions that were surveyed in both 2011 and 2016 (these are zones 1, 3, and 5 in Afar and 1, 2, and 9 in Somali). In 2016 some additional zones were included in the HCES/WMS survey in Afar and Somali to increase the representativity of the survey in pastoral areas. in poverty in the drought-prone lowlands and greater than the decline by 7 pp in the poverty pastoral (DPL&P) areas is the largest among all rate at the national level. The decline in the head- agroecological zones (Figure 2.2) The DPL&P areas count poverty rate in the DPL&P is also accom- had the higher poverty rate of 37 percent in 2011, panied by a small decline in the poverty gap but by 2016, the poverty rate was 27 percent, a (distance from the poverty line) and in the severity decline by 10 percentage points (pp), considerably (depth) of poverty. Chapter 2: Poverty and Vulnerability   17 Trends in Nonmonetary Poverty • Cooking fuel (wood/charcoal/dung are poor) Expenditure-based poverty is only a measure • Assets (poor if do not own more than one of of poverty in monetary terms. Yet, poverty has these: radio, tv, telephone, bike, motorbike) many different dimensions, most of which are In line with the changes in the monetary mea- nonmonetary. The Multidimensional Poverty Index sures of poverty, the values of the MPI by region in constructed by Oxford Poverty & Human Devel- 2011 and 2016 suggest that the two main regions opment Initiative (OPHI) focuses on the combina- of the drought-prone and pastoral areas of Afar tion of deprivations that simultaneously afflict a and Somali have also experienced a decline in the household (Alkire and Foster, 2011). A household is value of the MPI, albeit from a high value of depri- identified as multidimensionally poor if, and only vation in basic services in 2011 (see Figure 2.3). if, it is deprived in some combination of indica- The value of the MPI appears to decline also in the tors whose weighted sum exceeds 30 percent of regions of Benishangul-Gumuz and Gambella, the deprivations. The dimensions and indicators are lowland areas which comprise the humid moisture presented below reliable lowlands, whereas Figure 2.2 suggests that the headcount poverty rate based on consump- 1. Health (each indicator weighted equally at 1/6) tion in the lowlands areas of these regions did not • Child mortality (if any child has died in the change. The reasons for this are unclear but are family) likely to be related to a deterioration in local con- • Nutrition (if any adult or child in the family ditions in some zones due to land alienation and is malnourished—low weight for age) localized conflict, and the absence of coverage 2. Education (each indicator weighted equally by any safety net program like PSNP, which has at 1/6) become an important source of income to many in • Years of schooling (if no household member the drought-prone lowlands. has completed five years of schooling) To complete the picture, Figure 2.4 presents the • Child enrollment (if any school-aged child is prevalence of stunting among children between out of school in years one to eight). zero and five years of age. There is an increasing 3. Standard of living (each of the six indicators consensus that child height for age scores (HAZ) weighted equally at 1/18) provides a better measure of the nutritional status • Electricity (no electricity is poor) of children and of their potential welfare as adults. • Drinking water (MDG definitions) In a pattern consistent with that observed for • Sanitation (MDG definitions, including that expenditure-based poverty and the MPI, Afar and toilet is not shared) Somali regions also appear to have experienced a • Flooring (dirt/sand/dung are poor) decline in the prevalence of stunting. Figure 2.3: Multidimensional poverty index by region: 2011 vs. 2016 0.70 0.60 0.50 0.40 0.30 0.20 0.10 0 ay ar ra ia i uz NP lla ri a a al ab w ra om Af ha m be um gr Da SN Ha Ab So Ti Am m Or -G re Ga s di ul Di g Ad an sh ni Be 2011 2016 Source: Oxford Poverty & Human Development Initiative (OPHI) accessible at: https://ophi.org.uk/multidimensional-poverty-index/global-mpi-2018/ 18   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Figure 2.4: The prevalence of stunting among children 0–5 years of age by region: 2011 vs. 2016 60 50 40 30 20 10 0 ay r ra ia i uz NP la ri a a al a ab w ra om l Af ha m be um gr Da SN Ha Ab So Ti Am m Or l -G re Ga is Di d gu Ad an sh ni Be 2011 2016 Source: Ethiopia Demographic and Health Surveys 2011 and 2016. Humanitarian Food Aid (HFA) social safety net system (PSNP), the counterfac- and Poverty in the Lowlands tual household expenditures in the absence of any The decline in the expenditure-based poverty rate food aid received are estimated by deducting the in the drought-prone lowlands and pastoral areas amount of food aid and donations received from of Ethiopia appears to have taken place despite the total household consumption.44 2015–16 El Niño—triggered droughts that affected Figure 2.5 compares the poverty rates inclusive large areas of the country, especially in the low- of food aid received against the estimated poverty lands. The Belg rains of 2015 failed in large parts of rate that would have prevailed in the absence of the country and this was followed by further inade- any food aid, by agroecological zone in Ethiopia quate rainfall during the Meher season, which is the in 2016. Clearly, food aid made a difference. This main growing season (FEWS NET, 2016).43 This rain- suggests that food aid in general was directed fall failure led to a fall in crop output and to a loss toward the poorer and vulnerable segments of the of livestock in parts of the country. Consequently, population, thus preventing the poverty rate from a significant share of the rural population living in increasing. In the drought-prone lowlands and pas- these drought-affected areas suffered hardship, toral areas, for example, the poverty rate of 28 per- with an estimated 10.2 million people requiring cent with aid in 2016 is lower than the 30 percent emergency food assistance in 2016, on top of the poverty rate that would have prevailed without 7.9 million already covered by the Productive Safety food aid. Figure 2.5 also reveals that there is con- Net Program (HRD, 2016) (see Box 2.1). siderable heterogeneity in the extent to which the Considering these events between the two sur- distribution of food aid was successful at reducing vey rounds of the HCES, it is important to know the the poverty rate. extent to which the large inflow of food aid contrib- The preceding pattern is reinforced when actual uted to the apparent reduction in poverty between poverty rates are compared against counterfactual 2011 and 2016 in the lowlands. To estimate the poverty rates by region in 2016 (see Figure 2.6). poverty rate that might have prevailed in the absence of food aid provided either through human- 44 The value of food consumed from in-kind donations from itarian organizations and/or through Ethiopia’s Government/NGOs (code 62), cash received from the sale of in-kind donations from Government/NGOs (code 63), and donations in cash from Government/NGOs (code 64) are all deducted from the total value of food consumed by the household. The construction of the 43 The Meher is the main cropping season in Ethiopia and depends on counterfactual expenditure in this manner assumes that there are no the major rains during May through September. The Belg season, which behavioral responses to household expenditures on food and nonfood is based on the shorter rains that start in March, is important only in items because of the loss of food aid, and thus the interpretation of the some parts of the country. results must bear this caveat in mind. Chapter 2: Poverty and Vulnerability   19 Box 2.1: Ethiopia’s hot spot woreda classification in the 2015–16 drought The Ethiopian government and its partners have classified that were agreed upon through expert consultations (EWRD, the woredas (districts) in the country into hot spot catego- 2014). Operationally, this classification triggers a prioritized ries based on the impact of the drought, whereby woredas response, most notably in supplementary feeding (HRD, severely affected are categorized as hot spot 1 and those 2016) According to this categorization, 27, 18, and 9 percent with decreasing severity of drought as hot spots 2 and 3. The of the 743 woredas in the country were categorized as hot hot spot woreda classification is derived using six multisector spots 1, 2, and 3, respectively, in March 2016. In the region of indicators—food availability; water, sanitation, and hygiene; Afar, in particular, 93.5 percent of the woredas were classified access to markets; health and nutrition; education; and into the hot spot 1 category, whereas in the Somali region, other factors (increased migration, significant disruption to 43.6 and 45.5 percent of the woredas were classified as hot normal livelihoods, etc.)—at zonal, regional, and federal levels spots 1 and 2, respectively (see Table 2.1). Table 2.1: Proportion of drought affected woredas by region, March 2016 Number of Percent of woredas categorized as Region woredas Hot spot 1 Hot spot 2 Hot spot 3 Uncategorized All regions 743 26.5 17.8 9.3 46.4 Tigray 47 40.4 6.4 17.0 36.2 Afar 31 93.5 — — 6.5 Amhara 139 31.7 19.4 4.3 44.6 Oromia 279 21.9 11.8 7.2 59.1 Somali 55 43.6 45.5 — 10.9 Benishangul-Gumuz 20 — — 40.0 60.0 SNNP 147 11.6 27.2 15.0 46.3 Gambella 13 15.4 30.8 30.8 23.1 Harari 1 — — 100.0 — Addis Ababa 10 — — — 100.0 Dire Dawa 1 100.0 — — — Source: Bachewe et al., 2017. Note: Woredas in hot spot 1 are the most severely affected by the drought. Figure 2.5: Poverty headcount with and without aid/donations in 2016 0.35 0.33 0.30 0.29 0.30 0.28 0.28 0.27 0.26 0.25 0.22 0.20 0.15 0.10 0.05 0 Drought-prone Humid moisture Drought-prone Moisture lowlands and reliable highland reliable pastoral areas lowland highland With aid/donations Without aid/donations Source: Bank staff estimates based on the 2016 HCES and WMS, and inclusive of the additional zones in Afar and Somali aimed at increasing the representativity of the survey in the pastoral areas. 20   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Figure 2.6: Poverty rate (headcount) by region—with and without aid/donations in 2016 0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0 ay r ra ia i uz NP la ri a a al a ad w ra om l Af ha m be um gr Da SN Ha Ab So Ti Am m Or -G re Ga is l Di gu d Ad an sh ni Be Head count with Head count without Source: Bank staff estimates based on the 2016 HCES and WMS, and inclusive of the additional zones in Afar and Somali aimed at increasing the representativity of the survey in the pastoral areas. Food aid, through the PSN program or through the poverty headcount with aid is 17.5 percent, NGOs appears to have made a significant differ- and the poverty rate without is estimated to be ence in Afar, which was severely affected by the 19.2 percent. drought, where the poverty headcount with aid is Although food aid appears to have had an 27.3 percent and the poverty rate without food impact on the poverty rate prevailing in 2016, aid is estimated to be 36.9 percent. In contrast, it remains to be determined whether food aid is food aid appears to have made very little differ- also responsible for the observed decline in over- ence in the Somali region, where over 43 percent all poverty in the lowlands. Figure 2.7, comparing of the woredas were classified as hot spot 1, and the counterfactual poverty rates in both 2011 and some 45 percent of the woredas were classified in 2016 across agroecological zones of Ethiopia, as hot spot 2 (see Box 2.1). In the Somali region, reveals that the downward trend in overall poverty Figure 2.7: Trend in the poverty headcount rate without aid/donations: 2011 vs. 2016 0.40 0.37 0.35 0.33 0.30 0.29 0.29 0.29 0.28 0.30 0.25 0.25 0.20 0.15 0.10 0.05 0 Drought-prone Humid moisture Drought-prone Moisture lowlands and reliable highland reliable pastoralists lowland highland 2011 2016 Source: Bank staff estimates based on the 2011 and 2016 HCES and WMS. Note: All comparisons between 2011 and 2016 are based on the zones in Afar and Somali regions that were surveyed in both 2011 and 2016 (these are zones 1, 3, and 5 in Afar and 1, 2, and 9 in Somali). In 2016 some additional zones were included in the HCES/WMS survey in Afar and Somali to increase the representativity of the survey in pastoral areas. Chapter 2: Poverty and Vulnerability   21 in the (DPL&P) areas persists, even without the crop and livestock, seem to be moving into full-time contribution of humanitarian food aid. In contrast, crop production, with the fraction of the population in the humid moisture reliable lowlands, the head in crop production increasing from 35 percent in count poverty rate increases slightly between 2011 to 41.6 percent in 2016, and the rest moving 2011 and 2016, when food aid is excluded, while it into other livelihoods/occupations. remains constant when food aid is included (com- Figure 2.9 presents the prevalence of poverty pare Figure 2.2 and Figure 2.7). associated with different livelihoods in the (DPL&P) areas in 2011 and 2016. Figure 2.9 reveals that there has been a substantial reduction in the head- Poverty by Main Source of Household count poverty rate among households with live- Income (livelihood) stock as their main livelihood, with the proportion The next step in the analysis is to investigate more of the poor population among those largely depen- closely the prevalence and trends in poverty by dent on livestock decreasing from 58 percent in livelihood in the drought-prone lowlands and pas- 2011 to 29 percent in 2016 (a reduction by 29 per- toral areas. Based on the information collected by centage points). In contrast, the headcount poverty HCES on the main sources of income, households rate in all other livelihoods appears to decline more are classified into the following livelihood types: moderately (from 11 percentage points in all other livestock, crop and livestock (at comparable level), occupations to only 3 percentage points in crop crop production (temporary and permanent crops), production). and all other (which includes, salary casual labor, The reductions in the headcount poverty rate manufacturing, trade, wholesale, retail, services, by livelihood are not informative as to whether etc.). Figure 2.8 reveals that there have been sub- the standard of living among poor households in stantial changes in the main source of income different livelihood categories improved or deteri- of households (urban and rural areas combined) orated. Figure 2.10 presents the poverty gap and between 2011 and 2016. The share of the popula- the severity of poverty measures, with the former tion reporting livestock as their main livelihood has measuring the average distance of household declined only slightly, from 16.3 percent in 2011 to expenditures from the poverty line among the poor, 15 percent in 2016, whereas the share of the popu- and the latter placing a higher weight on the expen- lation in crop and livestock has almost halved, from ditures of those further away from the poverty 18.5 percent in 2011 to 9.5 percent in 2016. Most line. Figure 2.10 reveals that the depth of poverty of the population shifting away from livestock, and among those engaged in crop and livestock, as well Figure 2.8: Changes in the percentage of population in different livelihoods—DPL&P areas in 2011 and 2016 0.45 0.42 0.40 0.35 0.34 0.35 0.30 0.30 0.25 0.20 0.19 0.16 0.15 0.15 0.10 0.10 0.05 0 Livestock Crop and Crop All other livestock production occupations 2011 2016 Source: Bank staff estimates based on the 2011 and 2016 HCES and WMS (urban and rural areas combined). Note: All comparisons between 2011 and 2016 are based on the zones in Afar and Somali regions that were surveyed in both 2011 and 2016 (these are zones 1, 3, and 5 in Afar and 1, 2, and 9 in Somali). In 2016 some additional zones were included in the HCES/WMS survey in Afar and Somali to increase the representativity of the survey in pastoral areas. 22   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Figure 2.9: Poverty rate (headcount) by household livelihood—DPL&P areas in 2011 and 2016 0.70 0.60 0.58 0.50 0.40 0.37 0.32 0.34 0.30 0.29 0.26 0.26 0.20 0.15 0.10 0 Livestock Crop and Crop production All other livestock occupations 2011 2016 Source: Bank staff estimates based on the 2011 and 2016 HCES and WMS. Note: All comparisons between 2011 and 2016 are based on the zones in Afar and Somali regions that were surveyed in both 2011 and 2016 (these are zones 1, 3, and 5 in Afar and 1, 2, and 9 in Somali). In 2016 some additional zones were included in the HCES/WMS survey in Afar and Somali to increase the representativity of the survey in pastoral areas. Figure 2.10: Poverty gap and severity of poverty by household livelihood—DPL&P areas in 2011 and 2016 Poverty Gap Severity of Poverty 0.25 0.10 0.19 0.08 0.20 0.08 0.06 0.15 0.06 0.05 0.05 0.11 0.12 0.04 0.10 0.08 0.09 0.09 0.04 0.07 0.03 0.03 0.04 0.02 0.05 0.02 0 0 Livestock Crop and Crop All other Livestock Crop and Crop All other livestock production occupations livestock production occupations 2011 2016 2011 2016 Source: Bank staff estimates based on the 2011 and 2016 HCES and WMS. Note: All comparisons between 2011 and 2016 are based on the zones in Afar and Somali regions that were surveyed in both 2011 and 2016 (these are zones 1, 3, and 5 in Afar and 1, 2, and 9 in Somali). In 2016 some additional zones were included in the HCES/WMS survey in Afar and Somali to increase the representativity of the survey in pastoral areas. as in crop production only, has increased between expenditures (and the consequent reduction pov- 2011 and 2016. Thus, even though the headcount erty); or (ii) an improvement in the terms of trade poverty rate may have declined, overall poverty between livestock and staple grains. There is little has “deepened” among poor households engaged in evidence of the former and much stronger evidence crop and livestock production or in crop production of the latter. alone. There is scattered but generally consistent evi- dence in the pastoralist literature that household herd sizes or per capita livestock holdings have Structural Factors behind the Decline been declining in the last 20–30 years in Ethio- in Expenditure-Based Poverty pia (see Chapter 1 of this report). In addition, the The decline in expenditure-based poverty in the 2015–16 drought contributed significantly to a fur- ­ drought-prone lowlands and pastoral areas requires ther decline rather than to an increase in livestock a better understanding of the structural factors herd sizes. Yet, concrete evidence on the extent to behind it. Two potential explanations are possible: which household livestock assets have declined, (i) an accumulation of new livestock assets that especially between 2011 and 2016 in the drought- could sustain the increased household consumption prone lowlands and pastoral areas, is lacking. Chapter 2: Poverty and Vulnerability   23 Figure 2.11: Changes in expenditure-based and asset-based poverty—DPL&P areas in 2011 and 2016 Expenditure-based Poverty Rate Asset-based Poverty Rate (Head Count) (Head Count) 0.40 60 0.36 53.2 0.35 50 0.30 0.28 42.6 40 0.25 0.20 30 0.15 20 0.10 10 0.05 0 0 2011 2016 2011 2016 Source: Bank staff estimates based on the 2011 and 2016 HCES and WMS. Note: All comparisons between 2011 and 2016 are based on the zones in Afar and Somali regions that were surveyed in both 2011 and 2016 (these are zones 1, 3, and 5 in Afar and 1, 2, and 9 in Somali). In 2016 some additional zones were included in the HCES/WMS survey in Afar and Somali to increase the representativity of the survey in pastoral areas. Following Carter and Barrett (2006, 2013), the declines across all livelihoods, asset-based poverty asset-based poverty line is constructed based on or the fraction of asset poor households increases the value of livestock assets held by households within each livelihood category (see Figure 2.12). around the expenditure-based poverty line in 2016 The preceding estimates confirm that the main in the drought-prone lowlands.45 The same asset- explanation for the decline in expenditure-based based poverty line is used in 2011. A household in poverty between 2011 and 2016 is likely to be in each survey round is classified as “asset-poor” if the the increased returns to livestock owned. There is value of livestock owned (based on 2016 median strong evidence regarding the positive trends in prices for different types of livestock) is less than or the livestock terms of trade over the last 15 years equal to the asset-based poverty line. in the lowlands of Ethiopia. For example, a recent Figure 2.11 provides estimates of the change in USAID report on the changes in livelihoods between expenditure-based poverty and asset-based pov- 2003–2006 and 2013–2015 in the Afar region erty (headcount) in the drought-prone lowlands. As confirms that “in relation to staple grains, livestock Figure 2.11 highlights, the fraction of poor house- are worth significantly more in the new baseline holds based on expenditures declines between 2011 (2013–2015) than they were in the old baseline and 2016, while at the same time the fraction of (2003–2006)” (HEA Change Analysis, Afar Regional households that are asset poor (or with livestock Overview, USAID, 2017). Specifically, in Afar it is assets lower than the asset-based poverty line) estimated that the terms of trade (the kgs of sta- increases from 42.6 percent in 2011 to 53.2 percent ple grain that could be purchased with cash earned in 2016. In fact, it is estimated that the proportion from the sale of an animal) increased on average of nonpoor households based on expenditures, who by 84 percent for cattle, 96 percent for goats, and are asset poor increased from 45 percent in 2011 to 71 percent for camels. Moreover, livestock prices 57 percent in 2016. (adjusted for inflation) have increased significantly The same pattern prevails when expenditure-­ over the reference years allowing households to based poverty and asset-based poverty are com- partially compensate for losses of livestock. In pared across different livelihood categories in the Afar, for example, the reported increase in prices drought-prone lowlands and pastoral areas in 2011 for camels are between 40–45 percent, for goats and 2016. Although expenditure-based poverty between 60–65 percent, and for cattle between 50–55 percent. 45 This is based on the livestock holdings of households +/–3 percent of The same pattern, in general, seems to hold in the expenditure-based poverty line. the Somali region, where the terms of trade (the 24   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Figure 2.12: Changes in expenditure-based and asset-based poverty by livelihood—DPL&P areas in 2011 and 2016 Poverty Rate (Head Count) 0.70 0.60 0.58 0.50 0.40 0.37 0.32 0.34 0.30 0.29 0.26 0.26 0.20 0.15 0.10 0 Livestock Crop and Crop production All other livestock occupations 2011 2016 Asset-based Poverty Rate (Head Count): Urban and Rural Areas 100 90 87.8 81.0 80 70 60 54.7 50 40.6 40 37.6 30 28.8 21.9 20 13.7 10 0 Livestock Crop and Crop production All other livestock occupations 2011 2016 Source: Bank staff estimates based on the 2011 and 2016 HCES and WMS. Note: All comparisons between 2011 and 2016 are based on the zones in Afar and Somali regions that were surveyed in both 2011 and 2016 (these are zones 1, 3, and 5 in Afar and 1, 2, and 9 in Somali). In 2016 some additional zones were included in the HCES/WMS survey in Afar and Somali to increase the representativity of the survey in pastoral areas. kgs of staple grain that could be purchased with prices of major cereals were lower in January 2017 cash earned from the sale of an animal) increased compared to three years earlier, especially for between baselines (reference years) on average by maize, sorghum, and wheat. These crops are the 15–20 percent for camels, 10–15 percent for goats, major source of calories in areas that were most hit and 25 percent for sheep (HEA Change Analysis, by drought, such as Afar and the northern areas of Somali Region, World Bank, 2019). Cattle, on the the Somali region. The decline in the cost of cereals other hand, in the Somali region were worth less by in the food basket in January 2017 compared to 10–15 percent. three years earlier was estimated at 13.3 percent The recent study by Bachewe et al. (2017) pro- at the national level, and the decline in cereal costs vides additional evidence of increases in the terms was highest in areas most affected by the drought, of trade between livestock and cereals, even in the possibly indicating the effect of major cereal context of the 2015–16 drought. Their analysis of imports and food aid directed to these areas. prices over the period between January 2014 and When crop and livestock prices are consid- January 2017 confirms that the inflation-adjusted ered jointly, their analysis also reveals that Chapter 2: Poverty and Vulnerability   25 Table 2.2: Changes in terms of trade between livestock and cereals comparing January 2017 with January 2014, by woreda hot spot area, percent change Items compared Overall Hot spot 1 woredas Hot spot 2 woredas Hot spot 3 woredas Uncategorized woredas Cattle vs. cereals 9.8 0.9 12.8 11.9 13.3 Cows vs. cereals 12.1 1.3 19.6 5.0 17.4 Oxen vs. cereals 8.3 0.7 8.5 17.0 10.8 Sheep and goats vs. cereals 17.8 7.5 16.5 36.6 20.1 Sheep vs. cereals 15.2 2.4 15.0 34.7 18.6 Goats vs. cereals 20.4 12.8 18.0 38.5 21.8 Sheep and goats vs. teff –1.6 –9.3 0.1 9.4 0.3 Sheep and goats vs. wheat –18.0 24.4 27.6 24.2 15.9 Sheep and goats vs. maize 95.5 22.4 25.8 42.7 38.1 Sheep and goats vs. sorghum 36.7 19.8 23.1 72.3 43.6 Source: Bachewe et al. (2017) based on computations using CSA producer price data. Note: Woredas in hot spot 1 are the most severely affected by the drought (see Box 2.1). livestock-cereal terms of trade improved between current consumption of a household may be above the period of January 2014 to January 2017. the poverty line and thus classified as nonpoor, but, Table 2.2, copied from Bachewe et al. (2017), pro- in the future after a shock, the same household vides the percentage changes in the terms of trade may end up dropping below the poverty line. Such between various types of livestock and cereals in a household may be said to be “vulnerable to pov- the county overall as well as by woreda hot spot erty” even though currently it is not poor. In general, area (for additional details see Box 2.1) the extent of vulnerability to poverty depends on the risk management strategies of households and Vulnerability to Poverty in the Lowlands communities, the abilities of households to cope The standard poverty measures employed in the after the incidence of the shock (e.g., assets owned, previous section, such as the poverty headcount herd size, social capital), and access to safety nets rate or the poverty gap (see, e.g., Foster et al., (e.g., PSNP).47 1984), are only able to assess the current poverty In contrast to the recent reduction in poverty status of a household but are not necessarily a very among pastoralists in the lowlands, the prevailing useful guide to its future poverty status. House- view is that vulnerability to poverty from both idio- holds in the lowlands are frequently hit by a variety syncratic and covariate shocks48 is increasing over of shocks resulting in high income volatility. While the long term in the lowlands. The increase in vul- households in the risky environment of the lowlands nerability is attributed to a variety of factors that characterized by periodic droughts have developed various ex ante and ex post risk-coping strategies 47 Vulnerability to poverty is related to the concept of “resilience” which to reduce income fluctuations or to insure con- concerns the longer time path of well-being in the face of shocks, sumption against these fluctuations, the variance and especially the likelihood that any adverse outcomes of either risk of their consumption over time is likely to remain avoidance or a realized shock do not persist for an extended period. generally high.46 In a context where there is high For example, a nonpoor household may be vulnerable to becoming poor due to a job loss yet be quite resilient if the prospects for finding volatility in consumption, a household’s currently follow-on employment offering similar compensation are high and/ observed poverty status is, therefore, unlikely to or formal or informal safety net programs reliably provide adequate be a very good guide to a household’s vulnerability support promptly. In statistical terms, a nonpoor household with high conditional variance of income might be both vulnerable (to becoming to poverty, or its “poverty risk.” For example, the poor) and resilient (because the poverty is sufficiently short in duration, intensity, and/or likelihood (Barrett and Constas, 2014). 48 In this report, idiosyncratic shocks refer to household specific shocks 46 As emphasized by the extensive literature on consumption smooth- (e.g., injury, birth, death, or job loss of a household member) that are ing, household consumption may be insulated from idiosyncratic only weakly correlated across households within a community. Covari- shocks through community-based risk sharing arrangements, but is ate shocks refer to shocks that are correlated across households within likely to be affected by covariate or aggregate shocks that are more communities but only weakly correlated across communities (e.g., difficult to insure (Townsend, 1994). droughts, floods, earthquakes, or epidemics). 26   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Figure 2.13: Vulnerability to poverty characterized by the mean and variance of welfare A B C D Households E F G H I J Poverty line Mean and variance of welfare are discussed in this report, including the loss of places it below the poverty line, depicted by the ver- grazing areas for pastoralists, increased population, tical line in Figure 2.13, whereas household B is on expansion of cropland and agricultural concessions, average a non-poor household with a higher level of and conflicts (see Chapters 3 and 4). These fac- mean consumption located above the poverty line. tors, combined with the increasing frequency and In Figure 2.13, households A, D, G, and I are on aver- incidence of droughts and extreme weather events age poor households and their vulnerability is “pov- associated with climate change, lead to an overall erty induced,” meaning that is determined primarily decline in herd sizes per household, making house- by low endowments of livestock assets and human holds more vulnerable to poverty and more food capital, which are the primary determinants of their insecure due to fluctuating terms of trade. low mean value of their welfare. Households B, C, The “vulnerability to poverty” measure employed E, F, H, and J are, by contrast, on average nonpoor in this chapter rests on the characterization of households as their mean welfare is above the welfare (income, consumption) by its mean value poverty line. However, some of these households and its variance. Figure 2.13 summarizes these are vulnerable to poverty while others are not. two dimensions of welfare, mean and variance, Households B, E, and H, for example, have variabil- for 10 hypothetical households (households A ity in their consumption but the variance line never through J). The mean consumption expenditure of crosses the poverty line. In contrast, households C, a household (or the average value of consumption F, and J may end up below the poverty line under expenditures, for example, associated with many some circumstances, as depicted by the fact that different shocks or states of the world over time) the variance of their consumption around the mean is depicted by the orange square. Different shocks crosses the poverty line. For households C, F, and J, at different points in time lead to variation around vulnerability to poverty is “risk induced.”49 the mean value of expenditure, and the variance of welfare is depicted by the width of the horizon- tal lines to the right and to the left of the mean value of expenditure. Thus, households differ with 49 Figure 2.13 is also useful as an example of the difference between respect to the mean level of their expenditure and poverty and vulnerability headcount measures. The poverty headcount is based on the fraction of households whose consumption happens to the variance of their expenditure around that mean, be below the poverty line at a particular point in time. Depending on with some households having a low (or high) mean the timing of measurement and the type of shocks experienced (as well level of consumption and a low (or high) variance of as the capacity of a household to mitigate the impacts of such shocks consumption. ex ante or cope with these shocks ex post) the poverty head count may range anywhere between 20 percent (only households A and I poor) and Household A, for example, is on average a poor 70 percent (households A, C, D, F, G, I, and J). In contrast, the vulnera- household with a low mean consumption, which bility head count rate is simply 70 percent. Chapter 2: Poverty and Vulnerability   27 Figure 2.14: The prevalence of poverty and vulnerability—lowlands vs. other agroecological zones in 2016 0.70 0.66 0.60 0.50 0.48 0.46 0.43 0.40 0.38 0.29 0.30 0.30 0.25 0.26 0.22 0.20 0.10 0 Drought-prone Humid moisture Drought-prone Moisture reliable National lowland and reliable lowland highland highland pastoral Poverty rate Vulnerability rate Source: Bank staff estimates based on the 2016 HCES and WMS. An understanding of the causes of vulnerabil- rate in the drought prone lowlands and pastoral ity to poverty in the lowlands is necessary for the areas, estimated at 66 percent of the population, is design of policies that increase the overall resilience the highest among all agroecological zones in Ethio- of household welfare to shocks. To the extent vul- pia. Thus, 66 percent of the population is either poor nerability to poverty in the lowlands is “poverty or at risk of becoming poor in the event of a shock in induced” (i.e., low assets and human capital endow- the future. Moreover, in the lowlands the proportion ments), then cash transfer programs or programs of the vulnerable population is more than two times enhancing the delivery of basic services facilitating the proportion of the poor population (ratio of vul- investments in physical and human capital are nerability rate to poverty rate is 2.26). likely to be the most appropriate to the context of Figures 2.15 and 2.16 shed light on the sources the lowlands. In contrast, if vulnerability is primarily of vulnerability by agroecological zone in Ethiopia “risk induced” (i.e., high uninsured income fluctua- in 2016. As discussed above, vulnerability to pov- tions), then an insurance type of program may be erty can be either poverty-induced or risk induced. needed to increase resilience in the lowlands. Figure 2.15 reveals that in the drought-prone low- In the remainder of this chapter, a quantitative lands and pastoral areas, poverty-induced and measure of vulnerability to poverty for households risk-induced vulnerability are both higher than in in the lowlands is derived based on an ex ante esti- other agroecological zones. As a means of deter- mate of the mean and variance of consumption. mining which of these two sources is relatively more The methodology employed allows the distinction important, the ratio of poverty-induced to risk- between “poverty-induced” (or chronic poverty) and induced vulnerability reveals that poverty-induced “risk-induced” vulnerability, as well as the decom- vulnerability is relatively more important than position of the variance of consumption into dif- risk-induced vulnerability (ratio equal to 1.25). This ferent sources, such as idiosyncratic and covariate is in sharp contrast to the other zones where the shocks (a more detailed discussion of the method relatively more important source of vulnerability to used is contained in Annex 2).50 poverty is high consumption volatility. This attests to the unique nature of the lowlands in compari- Comparing Vulnerability and Its Sources son to the other agroecological zones and points among Ethiopia’s Agroecological Zones in favor of policies and programs tailored to the Figure 2.14 presents the prevalence of poverty and special features of the drought-prone lowlands and vulnerability by agroecological zone in Ethiopia in pastoral areas. 2016. The vulnerability to the poverty headcount It is also possible to decompose the variance of consumption into different sources, such as idio- syncratic and covariate shocks (a more detailed 50 See Hill and Poerter (2017) for a related approach to measuring vul- discussion of the method is contained in Annex 2). nerability using explcit measures of shocks reported by households. Figure 2.16 reveals that the vulnerability due to 28   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Figure 2.15: Sources of vulnerability in 2016—poverty induced vs. risk induced 0.40 0.37 0.35 0.30 0.30 0.25 0.25 0.26 0.26 0.25 0.22 0.20 0.19 0.17 0.15 0.14 0.10 0.05 0 Drought-prone Humid moisture Drought-prone Moisture National lowlands and reliable highland reliable pastoral lowland highland Poverty-induced vulnerability Risk-induced vulnerability Source: Bank staff estimates based on the 2016 HCES and WMS. Figure 2.16: Sources of vulnerability in 2016: idiosyncratic vs. covariate shocks 0.70 0.61 0.60 0.52 0.50 0.43 0.40 0.40 0.36 0.34 0.33 0.32 0.29 0.30 0.25 0.20 0.10 0 Drought-prone Humid moisture Drought-prone Moisture reliable National lowland and reliable lowland highland highland pastoral Idiosyncratic vulnerability Covariate vulnerability Source: Bank staff estimates based on the 2016 HCES and WMS. either idiosyncratic shocks or covariate shocks is rural areas. The prevalence of both poverty and the highest in the (DPL&P) areas. In addition, the vulnerability is higher in the rural areas than in ratio of covariate to idiosyncratic vulnerability urban areas of the lowlands (Figure 2.17), a finding reveals that in the (DPL&P) areas, vulnerability due consistent with the prevailing view that infrastruc- to covariate shocks, such as droughts, is relatively ture (e.g., roads) and easier access to markets are more important than vulnerability due to idiosyn- associated with reduced poverty and vulnerability cratic shocks (ratio equal to 0.86). to poverty. Moreover, the fraction of the vulnerable population in the rural areas is twice the size of the Comparing Vulnerability and Its Sources poor population. In contrast, in the urban areas, between Urban and Rural Areas of the the difference in the size of the poor and vulnerable Drought-Prone Lowlands and Pastoral Areas populations is considerably smaller (ratio of the vul- nerability rate to the poverty rate is 1.24). A closer comparison of the vulnerability and its A closer look at the sources of vulnerability in sources between urban and rural areas of the urban and rural areas of the lowlands reveals that drought-prone lowlands and pastoral areas in 2016 in rural areas poverty-induced and risk-induced reveals substantial differences between urban and Chapter 2: Poverty and Vulnerability   29 Figure 2.17: The prevalence of poverty and vulnerability in the Figure 2.19: Sources of vulnerability in urban and rural areas of DPL&P areas in 2016—urban vs. rural the lowlands in 2016—idiosyncratic vs. covariate 0.70 0.60 0.56 0.60 0.59 0.50 0.49 0.52 0.44 0.50 0.40 0.38 0.40 0.30 0.27 0.29 0.18 0.30 0.20 0.12 0.20 0.19 0.10 0.15 0.10 0 Total pop Urban Rural 0 in DPL&P Total pop Urban Rural in DPL&P Idiosyncratic vulnerability Covariate vulnerability Poverty rate Vulnerability rate Source: Bank staff estimates based on the 2016 HCES and WMS. Source: Bank staff estimates based on the 2016 HCES and WMS. Figure 2.18: Sources of vulnerability in urban and rural areas of engaged in livestock as well as in crop and livestock the lowlands in 2016—poverty induced vs. risk induced (pastoralists and agro-pastoralists) livelihoods, the 0.35 0.33 vulnerability rate in 2016 is more than twice as high 0.30 0.29 0.27 as the poverty rate (the ratio of the vulnerability 0.25 0.24 rate to the poverty rate is 2.34 for livestock and 2.39 for crop and livestock; for the rest of the liveli- 0.20 hoods this ratio is less than 2). 0.15 The analysis of the sources of vulnerability 0.10 0.09 reveals that the high variance of consumption (or 0.10 0.05 risk-induced vulnerability) is almost as important as low household assets and human capital endow- 0 Total pop Urban Rural ments (or poverty-induced vulnerability) especially in DPL&P for the population in pastoral, agropastoral, and cropping activities (see Figure 2.21). Nevertheless, it Poverty-induced vulnerability Risk-induced vulnerability should be noted that among those engaged in crop production (crops only, and crops and livestock) risk-induced vulnerability tends to be relatively vulnerability are both significantly higher than in more important than poverty-induced vulnerabil- urban areas, and that low household assets and ity, which is a likely reflection of the inherent risks human capital endowments (or poverty-induced associated with crop production in the drought- vulnerability) are relatively more important than prone lowlands of Ethiopia. consumption volatility (risk-induced vulnerability). See Figure 2.18. Conclusions Also, in rural areas of the drought-prone low- This chapter examined changes in monetary and lands, covariate shocks are found to have a rel- nonmonetary poverty in the lowlands between 2011 atively higher contribution to vulnerability than and 2016. The analysis revealed a reduction in both idiosyncratic shocks (see Figure 2.19). dimensions of poverty from an initially high level. Those engaged largely in livestock production (pas- Comparing Vulnerability and Its Sources toralists) seem to have experienced large decreases among Different Livelihoods in the Rural in the poverty head count as well as declines in the Drought-Prone Lowlands and Pastoral Areas depth of poverty. On the other hand, agropastoral- To complete the analysis, Figure 2.20 presents the ists seem to have transitioned, presumably due to prevalence of poverty and vulnerability for the pop- emergency livestock sales or deaths, to being full- ulation in the rural (DPL&P) areas grouped by dif- time farmers (cropping only). These new farmers ferent livelihoods. Figure 2.20 reveals that for those have not done well, as evidenced by an increase in 30   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Figure 2.20: The prevalence of poverty and vulnerability in the rural lowlands in 2016 0.70 0.63 0.62 0.60 0.58 0.57 0.50 0.46 0.40 0.34 0.30 0.29 0.27 0.26 0.26 0.20 0.10 0 Rural pop in Livestock Crops and Crops only All other DPL&P only livestock occupations Poverty rate Vulnerability rate Source: Bank staff estimates based on the 2016 HCES and WMS. Figure 2.21: Sources of vulnerability by livelihood in the rural lowlands in 2016—poverty induced vs. risk induced 0.40 0.35 0.33 0.32 0.30 0.29 0.30 0.30 0.30 0.28 0.27 0.27 0.25 0.20 0.19 0.15 0.10 0.05 0 Rural pop in Livestock Crops and Crops only All other DPL&P only livestock occupations Poverty-induced vulnerability Risk-induced vulnerability Source: Bank staff estimates based on the 2016 HCES and WMS. the depth of poverty among agropastoralists and poverty per se. The regular occurrence of droughts crop producers. Increases in the price of livestock in the lowlands and the transitory food insecurity accompanied by decreases in the price of cereals associated with such events need to become inte- combined to reduce monetary poverty even in the gral components of the risk management strat- case of smaller herd sizes and increasing asset- egy of the Ethiopian government for increasing based poverty. resilience. The analysis also provided firm quantitative The next three chapters drill down on the deter- evidence that in the lowlands aggregate shocks minants of poverty-induced and risk-induced vul- such as droughts combined with low human and nerability. Chapter 3 documents the low level of livestock capital are major sources of vulnerability human capital assets of households in the lowlands. to poverty for households. This points to the need Chapter 4 examines the changes affecting the cli- for policies and programs in the lowlands tailored mate and environment in the lowlands, and Chap- to alleviating vulnerability to poverty and not just ter 5 examines changes in land use and land cover. Chapter 2: Poverty and Vulnerability   31 Poverty and Vulnerability in the Ethiopian Lowlands Chapter 3: Low Human Capital as a Source of Vulnerability Summary The chapter recommends to (i) allocate more Across a range of human development (HD) outcomes, on-budget resources for human development in the from school attendance to infant and child mortality lowlands; (ii) introduce results-based approaches; rates, and severe acute malnutrition, the lowlands do (iii) introduce Specific Purpose Grants, targeted to worse than the highlands. Accessibility and quality of regions/woredas with low human development out- services are also worse. As a result, sectoral ministries comes; (iv) align civil servant incentives and pay to have developed specific strategies and approaches for results; and (v) recentralize some local and regional delivery of human development services—especially functions to higher levels and review mechanisms of to reach mobile pastoral populations. However, these local planning enforcement and sectoral budget alloca- strategies remain largely untested. tion in view of national targets. The main constraints to basic service delivery in the lowlands revolves around the nature of the Intergov- Introduction ernmental Fiscal Transfer System; low capacity and Ethiopia has achieved significant advances in availability of skilled professionals stemming from low human development over the last several years. education, especially of girls; and lack of incentives It has reduced child mortality, and increased net and accountability systems to ensure results. Com- enrollment rates (NERs) and the share of the rural pared to the large human capital needs, regions and population with access to clean water. Between woredas tend to allocate too little of their budgets for 1990 and 2017, Ethiopian life expectancy at birth health and education sectors, with minimal resources increased by 18.8 years, mean years of school- available for capital and operational costs. Afar and ing increased by 1.2 years, and expected years of Somali regions have predominantly focused on invest- schooling increased by 5.4 years. Between 2000 ments in water and rural development to the relative and 2017, Ethiopia’s Human Development Index neglect of education and health, while Gambella and value also increased from 0.283 to 0.463, an Benishangul-­ Gumuz have preferred to allocate expen- increase of 63.5 percent (UNDP, 2018). ditures to rural roads and education and health. These achievements were made possible through Overall, there needs to be a much stronger focus on the government’s continued commitment to financ- improving the human capital in these areas, especially ing basic services primarily through the Intergov- in the drought-prone lowlands. Without this there will ernmental Fiscal Transfers (IGFT), known as block be little chance to break the cycle of vulnerability. grants, to regions and woredas (see Chapter 7). This 32 is complemented by the government’s additional (with most of them in Oromia, Somali, and SNNP MDG/SDG Fund for capital expenditures in pro-poor regions), 4.4 million children and pregnant lactat- sectors (education, health, agriculture, water, and ing women are targeted to receive treatment for rural roads) and through the sectoral investments acute malnourishment, and a further 4.8 million of the government and sector specific programs people are targeted to receive emergency health financed by development partners. A large portion interventions. Approximately 3.1 million school of the general government’s expenditures (about children in 212 woredas are affected by drought 63 pecent in 2017) continues to be allocated to and conflict-induced emergencies. Among these the pro-poor sectors, with block grant subsidies are 444,839 displaced school-age children resid- to regions showing a much higher growth rate in ing in temporary settlements, of which 125,035 2016/17 than in preceding fiscal years (see Chap- (28%) have no access to any educational services. ter 5 for more details). In addition, the mortality rate in children under five In the 2018 World Bank Human Capital Index is likely to increase given the vicious cycle of poor (HCI)51 Ethiopia ranked 135 out of 157 countries. nutrition/malnutrition, increasing their susceptibil- With an HCI score of 0.38, a child born in Ethiopia ity to infection and illnesses. Newly displaced peo- today will be 38 percent as productive when she ple in Oromia and Somali regions also increase the grows up as she could be if she enjoyed complete burden on existing health facilities in host areas. education and full health. That means that her In Ethiopia’s lowlands, livelihood patterns are human capital endowment is around 2.5 times changing in response to climate factors and gov- below its potential. In addition, the consumption ernment policies. Frequent droughts, coupled with growth of the bottom 40 percent of the population the loss of rangelands, have led to an increase in has not kept pace with the top 60 percent, and the the shift from traditional, mobile livestock herding consumption of the bottom wealth quintile has con- pastoralism to a mix of livestock herding and crop tracted since 2005 (HCES 2016). So, while growth production. The government’s “voluntary based has been a main driver of poverty reduction, the resettlement” program has also resulted in large benefits have not been felt equally. Similarly, the sections of the lowland population in some areas poorest have not been able to benefit equally from moving to more urban environments and abandon- services. On average, their human capital outcomes ing their traditional livelihoods. (e.g., on literacy, school attendance, and skilled birth deliveries) are well below the wealthiest quin- Status of Human Development in Lowland tiles. Finally, falling and converging poverty rates Areas of Ethiopia across regions hide disparities between regions and This section examines the human development among woredas within regions. HCES 2016 data (mainly health, nutrition, and education) situation show that regional disparities are growing. in the lowlands (below 1,500 m), which sheds light In addition, Ethiopia is impacted by frequent on the effectiveness of these sectors in support- droughts that have the potential to undermine ing human capital development. The quantitative the development gains the country has made. The data on which the analysis is based are the Welfare El Niño–induced drought of 2016/17 resulted in Monitoring Survey (WMS), Household Consumption almost 20 percent of the rural population needing and Expenditure Survey (HCES), and the Demo- support to meet their basic food needs. The 2019 graphic and Health Survey (DHS), all of which were Humanitarian Response Plan (HRP) estimates conducted in 2016. Additionally, it explores other that in addition to PSNP beneficiaries, 8.13 million reports of studies conducted in the lowlands with people will need humanitarian food assistance a focus on human development including sociocul- tural norms, traditions and practices in the lowland areas which affect health seeking behavior, nutri- 51The HCI measures the amount of human capital that a child born today can expect to attain by age 18. It conveys the productivity of tion practices, and attitudes and behaviors related the next generation of workers compared to a benchmark of complete to schooling. education and full health. It is constructed for 157 countries. It is made To examine variations within the lowlands, addi- up of five indicators: the probability of survival to age five, a child’s tional classifications of drought-prone, moisture expected years of schooling, harmonized test scores as a measure of quality of learning, adult survival rate (fraction of 15-year olds that will reliable, and pastoral lowlands are made whenever survive to age 60), and the proportion of children who are not stunted. possible. Ecologically, the lowlands are divided in Chapter 3: Low Human Capital as a Source of Vulnerability   33 Map 3.1: Agroecological zones map of Ethiopia two—drought prone and moisture reliable. The within the lowlands, followed by some recommen- residents in moisture reliable lowlands are mainly dations to address these challenges. settled and rely on sedentary agriculture for their livelihood. Drought-prone lowlands on the other Education hand have both settled (agropastoralists) and more Education differences between highlands and lowlands mobile pastoralists. For this analysis, drought- Schooling experiences (having ever been in school) prone lowlands refer to the sedentary agropastoral- and literacy rates are significantly lower in the ists, while pastoralist refer to areas dominated by lowlands than in the highlands (Table 3.1). Nation- pastoralists (see Map 3.1).52 Disaggregation was also made by sex, age, ally, 63 percent have been to school at some point and expenditure/wealth quantiles. We have used in their life, while this number rises to 73 percent expenditure data from HCES together with WMS to for current school-age children (4- to 18-year- classify households based on their socioeconomic olds) (WMS). Overall, males are more likely to have status. There are five such quintiles with the lowest attended school than females (69 percent versus being referred to as the poorest, followed by poor, 56 percent). Among adults (19–65-year-olds), the middle, wealthy, and wealthiest quintiles. gender gap widens (67 percent for males versus The chapter is organized into two main sections: 40 percent for females). However, for school-age education and health/nutrition. Under each of the children, both males and females are equally repre- sections we start with a comparison of the low- sented in terms of having attended school (73 per- lands versus highlands (above 1,500 m) in terms of cent). However, fewer children have been to school their human development characteristics. This is in the lowlands—only 68 percent—compared to the then followed by comparisons of variation within highlands with 74 percent of children having been the lowlands based on agroecological zones, wealth, to school. and gender. The section concludes with an analysis Nationally, 64 percent of school-age children of the cross-cutting sociocultural challenges that were attending school in 2016 with the proportion are impeding the development of human capital being almost equal for boys and girls. However, the proportion of school-age children in school is much higher in the highlands (67%) than the low- 52 The only difference between Map 3.1 and Map 2.1 is that Map 3.1 dis- lands (59%). Interestingly, the proportion of lowland tinguishes the pastoralist areas from the drought-prone lowlands. school children attending school is similar across 34   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Table 3.1: Ever attended school (schooling experience)53 Frequency Percent Frequency Percent School age 33,138 73.1 School age  Male 17,034 72.7  Highland 18,669 74.9***  Female 16,104 73.4  Lowland 5,287 67.9 Adult 32,353 53.3 Adult  Male 19,677 67.2***  Highland 19,967 55.3***  Female 12,676 40.4  Lowland 3,705 41.6 ***Differences are statistically significant at the 1 percent level of significance using Pearson’s chi-squared test. Figure 3.1: School attendance by region, gender, and highland/lowland 100 100 80 80 60 60 40 40 20 20 0 0 ay ar ra ia i uz NP lla a ay ar ra ia i uz NP lla a al al ab ab om om Af ha Af ha m m be be um um gr gr SN SN Ab Ab So So Ti Ti Am Am m m Or Or l-G l-G Ga Ga s s di di gu gu Ad Ad an an sh sh ni ni Be Be Male Female Highland Lowland Highland Lowland all regions, with the striking exceptions being (wealth) groups do not have a clear pattern, except Benishangul-Gumuz and Gambella, where male that the top quintile having higher education attendance in both regions is much higher than the outcomes. lowland average (Figure 3.1). Furthermore, female Pastoralists in the lowlands have the lowest pro- attendance is exceptionally high in the lowlands portion of school-age children who have ever been of Gambella but very low in Benishangul-Gumuz, to school.54 Only 61 percent of school-age children where it falls even below the level for the Afar from pastoral areas have ever been to school region. It should be noted that the predominantly compared to 72 percent for children from drought- highland regions allocate the largest share of their prone and 81 percent from moisture reliable low- regional budgets to the education sector, while this lands. In terms of wealth groups, it is the wealthiest is not the case for the lowlands.53 (quintile 5) who have the highest proportion of schooling experience for school-age children. As can Education differences within lowlands be expected, the proportion of students attend- The moisture reliable lowlands (mainly Gambella ing secondary school increases with the wealth and Benishangul-Gumuz regions) perform much quintile. In the poorest quintiles about 92 percent better across a range of indicators from “having of those attending school are at the primary level ever been to school,” “literacy,” “current school with only 3.6 percent at secondary. On the other attendance,” and being “at a closer distance to hand, among the richest quintile, 78 percent of stu- school,” than the drought-prone and pastoral dents were in primary and the number in secondary lowlands. Differences within the socioeconomic reached 16 percent. 53Tables and figures are WB staff calculations using WMS/HCES 54 One-way ANOVA result being significant at 1 percent level of unless stated otherwise. significance. Chapter 3: Low Human Capital as a Source of Vulnerability   35 Figure 3.2: Schooling experience and literacy by gender and ecological zones 100 100 82 79 80 75 80 70 63 60 58 60 47 46 48 41 43 43 38 40 40 40 33 20 20 0 0 Drought- Moisture Pastoralist Lowland Male Female prone reliable Moisture reliable Drought-prone Pastoralist Male Female Pastoralist areas are also among the worst per- particularly for drought prone, pastoralist, and the formers in terms of the ability to read and write, 4th quintile (21 percent, 23 percent and 26 percent, with a 38 percent literacy rate followed by drought- respectively). Distance to school is also another prone (41%) and moisture reliable lowlands topping factor that came up frequently among drought the list but still at only 47 percent (see Figure 3.2). prone (16%), and pastoralist (15%). For girls, pastoralist areas have the lowest literacy The most common reasons for being absent from rate at 33 percent while for boys, it is the drought- school (defined as being absent for more than a prone lowlands (40%). Literacy rates increase from week in the month preceding the WMS survey) in the poorest (36%) to the wealthiest quintile (60%). the lowlands are having to work to help the family, school being temporarily not functioning, and illness. Reasons for education outcomes in the lowlands Nonfunctioning schools (temporarily) were the main Among the main reasons for not attending for- reason for not attending school in the pastoral areas mal education, ‘family not letting’ tops the list for (86%). Poor health as a reason for being absent from drought-prone and pastoralist lowlands, while ‘child school was exceptionally high for drought-prone is too young’ is the primary reason for moisture lowlands at 23 percent. Of course, the timing of the reliable lowlands. These reasons are demand side WMS study may have influenced these data. factors and might have been influenced by the fact that the response options had only one category Education quality issues in the lowlands related to the supply side, namely ‘no school around The Young Lives (YL)55 survey shows that pastoral the household’. Family not letting the child go to communities in the lowlands have been relatively school emerged as a top reason for drought-prone disadvantaged in terms of school facilities, such as lowlands (38%), pastoralists (49%), 3rd, 4th, and laboratories, libraries, and computers for students. 5th expenditure quintiles (42 percent, 53 percent, In terms of textbook availability, all students from and 41 percent, respectively). This could be related Addis Ababa sites have one math and English text- to the need for children to work or help with chores book each, while possession of an English textbook related to animals or household activities. It could falls to a low of 38 percent in Afar and 44 percent also be related to sociocultural norms related to in Somali (Table 3.2). This means that more than the usefulness of schooling and early marriage and half of students in these lowland communities do childbearing for girls. not have their own textbooks. Reasons for dropping out of school once they Shortage of books was the most severe problem have started are a mix of demand side (having among the range of problems identified by respon- to work) as well as supply side (school being too dents in WMS, particularly for the pastoralist group far and lack of education materials). The most common reason was having to work, which is the 55The results cited here are from the school surveys of Young Lives highest among the richest quintiles at 45 per- conducted at different highland and lowland sites. These were not cent followed by lack of educational materials, meant to be nationally representative. 36   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Table 3.2: Average class size and availability of textbooks Mean class size Proportion of students who Proportion of students who Location (number of students) have math textbook (%) have English textbook (%) Addis Ababa sites 49 100 100 Amhara sites 44  86  93 Oromia sites 52  74  74 SNNP sites 51  77  68 Tigray sites 50  78  68 Somali sites 62  52  44 Afar sites 46  54  38 All sites 51  75  70 Source: Young Lives School survey, 2016. Figure 3.3: Problems faced in school by ecological zones 70 60 50 41 42 43 40 28 28 30 25 25 26 26 26 23 23 20 17 10 10 10 8 6 2 0 Shortage of Poor Teacher Crowded Poor Insecurity books teaching shortage class facility Drought-prone Moisture reliable Pastoralist (Figure 3.3). Other problems included poor teaching, In summary, the overall picture emerging is shortage of teachers, overcrowding of classrooms, that education outcomes and quality indicators poor facilities, and insecurity. Among these prob- in the lowlands are much lower than in the high- lems, poor facilities were the second most reported lands. Furthermore, within the lowlands, moisture problem after shortage of books. Poor teaching and reliable lowlands (Benishangul-Gumuz and Gam- shortage of teachers were also reported frequently bella) fare better than drought-prone and pas- in both the pastoral and drought-prone lowlands, toralist lowlands, with the pastoralist lowlands but the moisture reliable lowlands appeared much having the worst education outcomes and quality better off for these two indicators. indicators. While there may be many factors con- Regarding school accessibility (Table 3.3), chil- tributing to these results, including demand-side dren in the lowlands on average must walk for up sociocultural factors (which are discussed in more to 3.7 km to access Grades 5–8, while their peers detail later), it should be noted that this trend is from the highlands only travel 2.7 km to attend the mirrored in the investment in education by the same level of schooling. Moisture reliable lowlands regions. The highland regions (Tigray, Amhara, have the lowest average distance to school, 2.5 km Oromia, and SNNPR) generally place higher pri- (Grades 5–8), while pastoralists have the longest at ority on education than the lowlands in terms of nearly 5.0 km. The difference is particularly large budget allocation, and within the lowlands, for secondary schools in the lowlands where chil- Benishangul-Gumuz and Gambella spend a dren travel up to 14.0 km to attend Grades 9–10. greater proportion of their budgets on education The equivalent distances for children in the high- than Afar and Somali regions. lands are 9.9 km (WMS, 2016; HCES, 2016). Chapter 3: Low Human Capital as a Source of Vulnerability   37 Table 3.3: Average distance from school in km Group Primary (G1–4) Primary (G5–8) Secondary (G9–10) Area  Highland 1.99 (2.0956)*** 2.73 (2.9364)*** 9.87 (9.6361)***  Lowland 1.90 (2.7066) 3.66 (5.5555) 13.95 (15.9300) Ecological zone   Drought prone 1.69 (2.0194)*** 3.18 (3.372)*** 14.70 (23.9776)***   Moisture reliable 1.41 (1.8183) 2.5 (3.7433) 15.17 (14.2456)  Pastoralist 2.3 (3.4310) 4.88 (7.2292) 14.87 (15.6799) Socioeconomic status  Poorest 1.94 (2.3120)*** 3.52 (4.8446)*** 16.46 (17.8573)***  2 1.78 (2.6172) 3.23 (4.8644) 14.77 (15.7314)  3 1.91 (2.9099) 4.29 (6.7503) 12.95 (14.0905)  4 2.08 (3.0700) 4.07 (5.9719) 12.91 (14.9461)  Wealthiest 1.69 (2.6187) 2.55 (4.18458) 10.16 (15.5881) Significant at **5% and ***1% level of significance (one sample t-Test for gender and one-way ANOVA for the rest), SE in parenthesis. Health highlands and lowlands (Figure 3.4). Households Health differences between highlands and lowlands in the highlands and lowlands demonstrate sim- ilar trends in the type of medical facilities used. The highlands fare much better than the lowlands However, in the lowlands 11% of households use across a range of key health service indicators, government health posts, which offer a less com- including availability of key health professionals, prehensive package of health services than health availability of essential drugs at health facilities centers or hospitals, and which there are more of and attendance at delivery by skilled birth atten- than in the highlands. This is largely due to the dants at health facility level. Under five mortality inaccessibility of health centers and hospitals in the rates and total fertility rates are also significantly lowlands. lower in the highlands. The two most frequently diagnosed illnesses, When comparing health seeking behavior within malaria and diarrhea, are much more common in the last two months prior to the WMS survey, lowland areas than in the highlands. Both lowlands nationally 15% of households received medical and highlands exhibit almost the same kinds and assistance after facing a health problem, with a levels of challenges regarding availability of quality higher incidence of women than men in both the Figure 3.4: Households who sought health services by gender and type of residence (highland and lowland) 100 80 60 40 20 0 ay ar a ia i uz NP lla a al r ab om Af ha m be um gr SN Ab So Ti Am m Or l-G Ga s di gu Ad an sh ni Be Male HL Female HL Male LL Female LL 38   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Figure 3.5: Type of problems faced at health facilities by location of residence (highland/lowland) 50 45 40 35 30 25 20 15 10 5 0 em e ls e ne ry t aff en t im iv na to ici st ns bl m io ra ed o ng pe of ip ss pr bo m qu ex ti s fe la n es ai of le o t io o w of pr en To ica y ta ng lit y iv th ed lit ni bi at Lo al Sa bi la m er he la ai op of ai av of av co ge un ge Un ta of / ta ge or ck or ta Sh La Sh or Sh Highland Lowland health services (Figure 3.5). Lack of medicines and Under five mortality and total fertility rates in cost of service are the most reported problems in the lowlands were much higher than in the high- the lowlands followed by lack of laboratories and lands according to the 2016 DHS (Figure 3.7). Under long waiting times to receive service. five mortality rates in the lowlands accounted for While the national rate for institutional deliveries around 76 deaths per 1,000 live births compared has increased to 26 percent in 2016 from 5 per- to the highlands, which were around 65 deaths per cent in 2000, almost 80 percent of households 1,000 live births. However, infant mortality rates in the lowlands reported that their child was born were similar in both highland and lowland areas at home, which is much higher than the rate in (48 deaths per 1,000 live births). the highland areas (67%). This may be due to the The total fertility rate is notably higher for the relatively better availability of key health profes- lowlands at 5.9 compared to the highlands at 4.4. sionals attending deliveries in the health centers in When we further disaggregate it by urban and rural the highlands compared to the lowlands, shorter areas, the total fertility rate is much higher for the distances to facilities, or sociocultural factors that rural areas of the lowland regions. On average, rural may deter women in the lowlands from having women will give birth to nearly three times more institutional deliveries. Addis Ababa and Tigray children during their reproductive years than urban report the highest facility births, followed by Gam- women. Importantly, total fertility has also been bella in the lowlands (Figure 3.6). increasing in Afar and Somali since 2000. Figure 3.6: Birth delivery at a facility by location of residence 100 80 60 40 20 0 ay ar ra ia i uz NP lla a al ab om Af ha m be um gr SN Ab So Ti Am m Or l-G Ga s di gu Ad an sh ni Be Highland Lowland Chapter 3: Low Human Capital as a Source of Vulnerability   39 Figure 3.7: Mortality rate and fertility rate by location of residence Infant and Under-5 Mortality Rates Highland 65.4 48.2 Lowland 75.5 47.6 National 67.0 48.1 0 20 40 60 80 U5mr Imr 8 6 4 2 0 –2 –4 Total fertility rate (2016) Change in TFR (2000–2016) Somali Afar Oromia Tigray Ethiopia Benishangul-Gumuz SNNP Harari Amhara Gambella Dire Dawa Addis Ababa Source: WB staff calculation based on DHS, 2016. Health and nutrition results within the lowlands of health problems (18%). Incidence of health prob- The moisture reliable lowlands of Benishangul-­ lems by wealth does not seem to have a clear Gumuz and Gambella report the most health prob- trend for drought-prone lowlands and pastoralists lems (23%) followed by the drought prone lowlands while it seems to be increasing from poorest to (19%) and pastoral lowlands (8%) (Figure 3.8). Look- richest expenditure quintiles for moisture reliable ing at the same indicator by expenditure quintiles, lowlands. Higher incidence of reporting illnesses the richest quintile reported the highest incidence among wealthier households may be due to greater Figure 3.8: Incidence of health problems by ecological zones and wealth quintiles 30 25 26 25 23 24 22 22 21 20 20 20 15 14 9 10 11 10 8 6 5 0 Poorest 2 3 4 Richest Drought-prone lowland Moisture reliable lowland Pastoralist 40   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Table 3.4: Place where medical assistance was received, by ecological zone and wealth strata Ecological zones Expenditure quintiles Drought Moisture Type of health facility Overall prone reliable Pastoral Poorest Poor Middle Rich Richest Health center 46% 37% 47% 44% 48% 48% 48% 41% 35% Private clinic 16% 30% 11% 10% 13% 14% 19% 17% 28% Government hospital 14%  8%  8% 28% 12% 10% 12% 22% 21% Health post 12% 13% 18% 10% 12% 18% 10% 10%  5% Total number of observations 4,108 630 2,248 842 779 786 705 877 961 knowledge of signs of illness in households of better As can be expected from the high rates of home economic and educational status. Of those who delivery, most of the births (about 78 percent) were faced a health problem, around 80 percent or more assisted with untrained traditional birth attendants consulted a medical institution or traditional56 (TBA) and family members. This rate is the highest healer. The highest proportion of health service among pastoralists (87%) followed by drought- seekers live in the drought-prone lowlands (94%) prone lowlands (66%) and moisture reliable lowlands followed by moisture reliable lowlands (83%) and (56%). Similarly, wealth dictates the type of birth pastoralist (81%). assistance used where 81 percent of births among Government health centers (HC) are the most the poorest quintile use TBA and family members common places visited for medical care regardless but this number drops to 50 percent among the of ecological zones or expenditure quintiles. HCs top quintile. In fact, 47 percent of deliveries among are followed by private clinics for drought-prone the top quintile were assisted by trained medical lowlands, government health posts (HP) for mois- personnel. ture reliable lowlands, and government hospitals The lowlands and emerging regions have for pastoralists (Table 3.4). The reasons given for extremely high rates of acute malnutrition (wast- not seeking health care included both demand and ing) among children under five compared to the supply side constraints. Even though the number of highlands, with an alarming upward trend in prev- observations was limited (587), the most frequently alence. This is of serious concern because wast- reported reasons were ‘no need for consulting’ and ing is a strong predictor of under-five mortality; ‘financial limitations’ among the different groups. and is usually the result of acute significant food Other reasons that followed these two include ‘not shortage and/or disease (and persistent food inse- believing in medicine’ (quintiles 2 and 4 as well as curity in the lowlands is threatening to worsen moisture reliable lowlands groups), ‘service being the nutrition situation). Wasting prevalence above too far away’ (quintile 3 and moisture reliable 10 percent is the threshold for determining an lowlands), and ‘lack of confidence in the service’ emergency nutrition situation, and five out of the (drought-prone lowlands). six lowland/emerging regions have wasting rates Overall, around 80 percent of the children in above 10 percent (the national average is 9.9 per- the lowlands under five, according to WMS, were cent): Somali (23%), Afar (18%), Gambella (14%), born at home. The home delivery rate is highest Benishangul-Gumuz (12%), and Oromia (11%). As among pastoralists at 89 percent, then for drought shown in Figure 3.9, wasting rates have increased prone (73%) and lowest for moisture reliable low- between 201157 and 2016 in Somali, Gambella, lands (64%). With increasing wealth, deliveries Benishangul-Gumuz, and Oromia (DHS). Reports seem to shift away from home delivery and toward suggest that climate-­ induced IDPs have higher institutional delivery. Home deliveries drop from levels of malnutrition than host communities, 85 percent for the poorest to 53 percent for the which contributes to this lower access to services. wealthiest quintile. This is also the case for conflict IDPs; 42% of IDPs 56Going to a traditional healer as a primary choice for health problems account for less than 2 percent among those who reported facing a 57 Wasting rates from the 2011 DHS survey: Somali (22%); Gambella health problem. (12.5%), Benishangul (9.9%), and Oromia (9.7%). Chapter 3: Low Human Capital as a Source of Vulnerability   41 Figure 3.9: Wasting prevalence among children under five 30 25 20 Percent 15 10 5 0 r i uz la ay ra ia NP ri a a l al ta a ab w ra om l Af ha m be um gr To Da SN Ha Ab So Ti Am m Or -G re Ga is l Di d gu Ad an sh ni Be 2000 2005 2011 2016 identified in need of support by the 2019 humani- Figure 3.10: Distance to health facility in km by ecological tarian appeal are defined as in acute need. zones Afar region has the lowest level of early initia- 80 tion of breastfeeding (42%) and the highest level 60 of pre-lacteal feeding (41%). A mixed methods survey by UNICEF (2017) conducted in selected 40 zones of Afar, Amhara, Benishangul-Gumuz, and 20 Tigray found a much lower rate of early initiation of breastfeeding in Afar compared to other regions. It 0 reported that “only 27 percent of surveyed partici- HP Clinic HC Hospital pants from zones in Afar with a child 0–23 months Drought-prone lowland initiated breastfeeding within one hour of birth, as Moisture reliable lowland compared with zones in other regions (57 to 86 per- Pastoralist cent).” There is significant regional variation in the proportion of children 6–23 months who receive a minimum acceptable diet, with the lowest levels the farthest away from health facilities. In the case (2–3%) in Afar, Somali, and Amhara. of hospitals, all lowland households were on aver- age around 70 km away. This must be seen in rela- Health-Related Quality Indicators tion to the service standards and catchment areas that are expected of health facilities. Among the lowland regions, Afar has the highest In summary, like the education sector, the mean availability of basic amenities58 accounting overall picture emerging is that health outcomes for 52 percent. In comparison, Addis Ababa was the in the lowlands are poorer than in the highland region with highest mean availability of basic ame- areas. Within the lowlands there is not such a clear nities among all regions with 79 percent. People in distinction between ecological zones in terms of the drought-prone lowlands live closest to health overall health outcomes, though the pastoralist facilities (except for hospitals) followed by the mois- areas fare significantly worse in terms of facility ture reliable lowlands (Figure 3.10). Pastoralists live deliveries. Wealth appears to be a more dominant factor than agroecological zone in terms of some 58 Service readiness for basic amenities was assessed based on sanita- health outcomes, such as facility deliveries and tion facilities, communication equipment, consultation room, improved water source, power source, emergency transportation, and computer breastfeeding (except for pastoralist areas). Again, with Internet access. it should be noted that most of the four highland 42   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands regions (apart from Tigray) allocate a greater per- also affects school enrollment (Bacha et al., 2014). centage of their budgets to health than the low- Where school timetables have not adjusted to child lands, with Somali region only allocating 8% of its labor demands, dropout levels are higher. In Borena budget to the sector. This is even more significant zone, the Alternative Basic Education (ABE) system when factors such as low population density, inac- offers evening classes for children, which reduces cessibility, harsh environment, and human capacity dropout rates (Dyer, 2016), whereas this flexibility is are considered. These will be discussed in more rarely practiced in Afar, Somali, and South Omo. detail later in the chapter. Education and Health Services Sociocultural and Economic Constraints in the Face of Shocks in Health and Education Lowland areas are prone to both natural and man- In analyzing the causes and contributing factors to made hazards that have led to high mortality poor health outcomes in the lowland areas, gender rates, displacement, and loss of property. Among and other sociocultural determinants play major these, flood, drought, conflicts, and political insta- roles in the lowland health care system. People in bility have all taken their toll on the health care the lowlands adhere strongly to traditional cul- system and disrupted access to livelihoods and tural values and beliefs and have limited access to food security, resulting in escalating rates of child information and services, although there is some under­ nutrition. According to the 2019 Humanitar- evidence of gradual changes in some traditional ian Needs Report, there are 5.91 million people in practices. Deep rooted traditional practices such need of nutrition assistance and 4.4. million who as female genital mutilation, early marriage, and have been targeted for humanitarian support. a variety of traditional ceremonies and rituals are Somali region over the last two years has repre- critical bottlenecks influencing the demand side of sented at least 25 percent of the national detected the health care system. The use of traditional heal- Severe Acute Malnutrition (SAM) cases, compared ers and religious leaders to cure a variety of medical to its population size, which is estimated to be conditions is also widespread. 6 percent of the Ethiopian population. Afar is also Apart from most maternal and child health ser- overrepresented with 6 percent of the SAM cases vices being free of charge, there still exists very few nationally but with only 2 percent of the national financial protection mechanisms to protect house- population. holds from out-of-pocket expenses; consequently, Conflicts and political instability in the region fees are an important barrier to service utilization are further worsening the health care system. A in lowland areas. In the lowlands, health care costs significant proportion of “highlanders” (health pro- of households increase because of high transporta- fessionals that come from other areas) abandoned tion costs due to long distances to health facilities. health facilities in lowland areas because of the Financial barriers to care are the main reason given recent inter-ethnic conflicts in the area. During for not seeking health care in the lowlands. With the field visits for this study the regional health increasing wealth, for example, deliveries shift away bureau estimated that about 50 percent of from home delivery and toward institutional deliv- middle-level health professionals (nurses, health ery. Community-based health insurance schemes officers, midwives, etc.) had left their jobs from (CBHI) are implemented in very few geographic health centers and hospitals in Somali region. areas in the lowlands and are completely absent in As noted from the field visit in South Omo zone, Somali, Afar, and South Omo. more than 20 health posts are ruined/destroyed The nomadic nature of pastoralism and the because of conflict which left several kebeles with- need for child labor to herd cattle results in higher out health posts. primary school dropout levels at certain times of In Somali and Oromia regions conflict has left the year among Borana in Oromia and in Afar and hundreds of thousands displaced, often in areas Somali regions (Dyer, 2016). Child labor require- already experiencing ongoing drought. Nearly all ments also deter families from sending children to districts along the regional borders are affected. boarding schools. Retention rates are much higher For IDPs, accessing services in their area of dis- for agropastoralists and urban residents. Child placement (or while on the move) including human- labor demands are also factors in Gambella region itarian assistance, social protection, health, and where subsistence farming is practiced, which education is challenging, which makes them highly Chapter 3: Low Human Capital as a Source of Vulnerability   43 vulnerable. Displaced persons who reside in host HD Delivery Systems’ Analysis communities report having difficulty accessing Decentralized service delivery systems—relevance services and may not be recognized officially as for the lowland context displaced. For those IDPs living in relief centers or In addressing the challenge of low density and informal settlements, there are further concerns. mobile populations in the lowlands, the government These include lack of space and privacy, lack of initiated “voluntary based resettlement and villa- shelter, lack of latrines, exchanging sex for food gization” programs in the four lowland emerging or money, and gender-based violence. The risk of regions. The purpose of villagization is to provide outbreaks of preventable disease is extremely high, access and utilization of basic socioeconomic infra- and due to poor living conditions could lead to high structure and services. The status of villagization levels of morbidity. and settlement programs varies across the regions As of December 2018, Ethiopia was host to more but is more advanced in Benishangul-Gumuz and than 900,000 refugees, the majority who live in Gambella than in Afar and Somali. In general, camps in the lowlands. In 2016, the government settlement makes it easier to deliver services at articulated a strategic approach to the plight scale and could be one factor contributing to bet- of refugees in a series of nine pledges aimed at ter service outcomes in Benishangul-Gumuz and improving rights and expanding services to benefit Gambella than in other lowland areas. In Gambella, both refugees and host communities. Currently, for example, primary school attendance, especially services in camps are provided through the Agency for girls, is very high in villagization sites and areas for Refugee and Returnee Affairs (ARRA) with the with large numbers of highlanders but continues to support of UNHCR and the international commu- lag elsewhere. Yet these programs are controver- nity. In practice there is often overlap with host sial and need to be carefully assessed in terms of community members accessing some services in their impact on the environment and local livelihood camps and vice versa. Building on the pledges, the strategies. draft National Comprehensive Refugee Response Decentralized service delivery is organized fol- Strategy outlines a vision of gradual integration for lowing sector specific expenditure assignments refugees who have been in Ethiopia for more than (Table 3.5). The two main elements of decentral- 20 years within host communities and gradually ization (particularly in a devolved system) are moving away from the current in-camp assistance expenditure assignment and revenue assignment to refugees and toward an integration of service by level of government. In the health and education provision to refugees into national systems. Table 3.5: Sector specific expenditures assignments/service delivery functions at different tiers of government, district level decentralization program Kebele Woreda Region Federal Education Primary school Primary and 1st cycle 2nd cycle secondary school, Tertiary/higher education, admin, PTA, secondary school, primary up to public colleges, national policies and matching finance school teachers, salaries, management institutes, standards primary textbook distribution, vocational and technical primary boarding school, school training, training centers, management and cost sharing, regional policies and adult education standards Health Primary health Health centers up to 25,000, Up to public hospital care, Specialized hospital, care admin, HIV/ health posts up to 5,000, referral hospital, HIV/AIDA medical training institutes, AIDS prevention, health extension, clinic admin, immunization and malaria educational materials, health posts up HIV/AIDS and malaria control control, allocation of regional large procurement, to 5,000, health and prevention, allocation of funds, regional policies and national policies and extension woreda funds standards standards, resource mobilization Source: Ministry of Capacity Building, 2011. 44   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands sectors most of the basic service delivery functions due to low population density, limited urbanization, are assigned at the woreda and kebele levels, with inaccessibility, harsh environmental conditions, and regional and federal levels responsible for higher lower capacity to manage and deliver services. The level education and health care (upper secondary current formula used by the federal government and tertiary education and referral and specialized to allocate block grants to the regions does not hospitals), as well as regional and federal policy adequately compensate for the increased costs of and standards setting and monitoring. Hence, the service delivery in the lowlands. While the federal mandate and responsibility of basic health and government gives some consideration to the unique education provision rest strongly with woredas. This circumstances of the lowlands in determining the is consistent with the high degree of woreda-level budget allocation formula for block grants, it is not autonomy in the allocation of budgets across consistent with the real costs of delivering services sectors. in the lowlands. Generally, the lowlands receive However, in Oromia and Benishangul-Gumuz approximately 1 percent more in block grants than regions, while woredas are responsible for pro- their population size warrants. Afar, for example, viding services at health post, health center, and constitutes 2 percent of Ethiopia’s population and recently primary hospital levels, they do not have has received around 3 percent of the block grants the mandate to construct the facilities except for over the last decade. Similarly, the pastoralist pop- health posts. Construction of health centers is ulation in SNNPR comprises about 14 percent of still the mandate of the regional government in the regional population and has received approx- both regions, and this was reported to be asso- imately 15 percent of the region’s block grants ciated with low capacity at woreda levels and to over the last five years. In Oromia, pastoral and ensure value for money through bulk contracting agropastoral zones constitute about 30 percent of of construction work at regional levels. Delegated the population and received around the same per- financing is mostly controlled at the regional level, centage in block grants over the past seven years. and district administrations are underfunded and Benishangul-­ Gumuz and Gambella received higher under-capacitated. grants in per capita terms. However, the changes As a result, the influence of federal sectoral min- to the block grant formula in 2018 have resulted in istries on how basic services are provided and orga- an increase for Somali region of about 2 percent, no nized, and the matching of resources for adequate change for Afar, and a decrease for Benishangul-­ provision of those services (compared to set service Gumuz and Gambella. Nevertheless, the slightly standards) varies. In the health sector, because higher than average block grants to the regions the Federal Ministry of Health (FMOH) manages with lowland areas still fall well short of what is the SDG Performance Fund, primarily for essential required compared to the significantly higher ser- commodities and drugs, it retains some control of vice delivery costs and the level of needs of these how these resources are used to deliver improved areas to boost human development outcomes. health results. The federal sector level is also At the regional level, at the same time, the ratio- responsible for the procurement of drugs through nale for the sectoral allocations made by regional the specialized agency Pharmaceuticals Supply governments are often unclear. Significant differ- Agency (PSA) which procures and distributes all ences emerge between the priorities of the various major drugs to regions and woredas. For education regions. In Afar and Somali regions, the economic however, the influence of the federal Ministry of sectors (water and agriculture) are prioritized over Education (MoE) over the management and oper- the human development sector, with the combined ation of primary and secondary education in the health and education sectors only receiving 27 per- regions is extremely limited. cent and 22 percent respectively, of the regional budgets over the last five years (Figure 3.11). This Resources for health and education service delivery is in stark contrast with Amhara allocating 43 per- are very limited and not allocated based on severity cent to these sectors. Benishangul-Gumuz and of needs or performance Gambella allocate significantly higher proportions The cost of human development (health and edu- on education, which may partially account for the cation) service delivery in the lowlands is much better education outcomes in moisture reliable higher in per capita terms than in other regions lowlands. All four lowland regions allocate relatively Chapter 3: Low Human Capital as a Source of Vulnerability   45 Figure 3.11: Sectoral disaggregation of regional expenditures during FY12–18, percent of total expenditures 100 80 60 40 20 0 r i uz lla ay ra ia NP ri a a al a ab w ra om Af ha m be um gr Da SN Ha Ab So Ti Am m Or l -G re Ga is Di d gu Ad an sh ni Be Education Health Other higher proportions of their budgets to administra- (Table 3.6). In contrast, Amhara region, in addi- tion and general services than highland regions. The tion to the already high 43 percent of its regional implication of this is evident in the health sector expenditures going to human development, spend where regions with the best service coverage indi- 22 percent of its MDG/SDG on human development cators (i.e., availability of health facilities) spend infrastructure. Benishangul-Gumuz and Gambella the most on the sector (Addis Ababa, Dire Dawa, spend a commendable 23 percent and 38 percent, Harari, and Gambella), while the region with the respectively, on the HD sectors. worst in the country at only 14 percent (Somali), Operational budgets are inadequate and excep- spends the second least amount per capita after tionally limited at the woreda levels in the lowlands Benishangul-Gumuz. (see Table 3.7). While Somali region fares relatively These differences in regional allocations extend better than the other lowland regions, the oper- to capital allocations from the MDG/SDG grants ational budgets for both health and education in regional governments receive to boost the cap- Gambella and education in Benishangul-Gumuz are ital expenditures. For example, Afar and Somali particularly low, and it is difficult to perceive how regions allocated their entire MDG/SDG budget to the systems can function within such constraints. the water sector in EFY 2012, meaning that health Furthermore, the percentage of operational bud- and education did not benefit from these resources gets going to the woredas as opposed to the Table 3.6: Sectoral allocation of the MDG transfers in EFY12 by region Agriculture Water Rural roads Education Health Tigray 3.9% 63.5% 28.4% 4.0% 0.2% Afar — 100.0% — — — Amhara 5.7% 14.1% 57.8% 10.3% 12.2% Oromia 2.8% 8.9% 68.0% 11.3% 9.0% Somali — 100.0% — — — Benishangul-Gumuz 14.3% 10.5% 51.0% 17.9% 6.2% SNNPR 18.4% 7.0% 57.0% 8.0% 9.5% Gambella 11.2% 29.3% 21.1% 19.9% 18.5% Harari 1.9% 23.9% 41.6% 27.2% 5.4% Dire Dawa 11.8% 32.1% 12.8% 22.3% 20.9% Total 6.8% 25.0% 50.8% 9.1% 8.4% Source: MoF. 46   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Table 3.7: Percentage of woreda level operational budget share in 2009 and 2010 EC from the total operational budget of the region, February 2019 Woreda level Woreda share from oper. expn. in million birr (office, total operational budget health center/posts, non-salary) of the region in % Region Description FY2009 EC FY2010 EC FY2009 EC FY2010 EC Somali Education  67.87  53.44 17.8% 32.1% Health  75.42  84.85 26.9% 44.7% Gambella Education   2.09   3.42  5.1%  9.8% Health   2.11   2.95  7.4%  7.9% Afar Education  41.79  41.49 56.8% 51.2% Health  55.13  42.06 54.9% 42.9% Benishangul-Gumuz Education  17.04  20.46 33.4% 30.0% Health  46.27  58.67 47.8% 52.4% Tigray Education 115.05  98.37 30.4% 24.9% Health 92.59  87.37 28.6% 25.9% Amhara Education 956.60 960.06 79.1% 89.8% Health 727.89 806.23 68.4% 78.5% SNNPR Education 729.54 929.84 68.3% 72.6% Health 921.89 1021.89 92.3% 92.6% Source: MoF. Note: EC = Ethiopian calendar. regional level is much lower for the lowlands than constructed residential homes for health extension the highlands, with Gambella and Somali regions workers alongside health posts. In the health sec- doing particularly poor. This is inconsistent with the tor, NGOs fund essential drugs and compensate, fact that major expenditure assignments for basic to a certain extent, for unavailability of adequate health and education should be at the woreda level. operational budgets for the duration of the NGO It also means that woreda health and education support. However, this by no means makes up for staff are unable to adequately monitor and support the shortfall in overall operational budgets from the work of remote health and school facilities due the government. While out-of-pocket expenditures to lack of per diems and transport allowances. at facilities are relatively low in lowland regions, Off-budget resources supplement the low gov- patients often chose to bypass the nearest facil- ernment HD expenditures but do not address the ity to go to other facilities with better quality care extent of the challenges. Given the limited budgets and hence incur additional costs related to travel that are available for capital expenditures or oper- (Figure 3.12). Overall, poor coordination between ational costs in the lowlands, off-budget resources off-budget resources and on-budget government from community driven, externally funded projects resources lead to inefficiencies and sustainability supplement the limited human development in the problems in health care funding. lowlands. For instance, the World Bank financed Pastoral Community Development Project (PCDP) Local level capacity and accountability processes are not aligned with delivering results in Afar and Somali regions has supported the con- struction of health facilities, particularly health Leadership and management capacity in the health posts, furnishing and equipping health facilities as and education sectors lack effective accountability well as training health workers in targeted woredas. and incentive mechanisms. Health and education PCDP supported the construction of health posts managers at all levels of the service delivery sys- at the sub-kebele level (village level) due to the tems in the lowland areas lack capacity to ade- large size of kebeles in lowland areas and seasonal quately understand and implement health and mobility of many communities. The project also education policies, guidelines, standards, protocols, Chapter 3: Low Human Capital as a Source of Vulnerability   47 Figure 3.12: Per capita out-of-pocket expenditures and facility by-pass rate by region Per capita OOP expenditure (USD) By-passed nearest health facility for OPD services (%) 25 Tigray 20 Amhara 15 Dire Dawa Addis Ababa 10 SNNP 5 Oromia Harari 0 Afar l -G l l a So uz i ar Or ri Ad S a A P a Am wa Ti a Na ay l al na i Di bab r di NN ra om Af ha Somali m gu be um gr Da tio Ha an am re s Benishangul-Gumuz G Gambella sh ni Be 0 20 40 60 80 and reforms. The political commitment from Figure 3.13: Percent of civil servants with a university degree, regional government and woreda councils toward by region the provision of quality education and health ser- National average vices is also lacking. Too often poor execution of plans and programs fails to result in any perceived Somali consequences, and there is a lack of an appraisal system that rewards high performance. The use of Gambella feedback mechanisms to improve human develop- Afar ment service performance is weak and often deci- sion making follows a top-down approach. While 0 20 40 60 80 there are incentives from the Ministry of Health Source: Ethiopia Civil Servant Survey (2016). (MoH) to support emerging regions and lowland areas, they are not necessarily addressing key chal- lenges related to performance, motivation, skills, not speak the local language, are filling many civil capacity, and retention of the health workforce. service positions in the lowlands. This poses chal- lenges for health and education service delivery in Human capacity in the lowlands is very low terms of communication, and understanding and and inadequate to provide adequate human appreciation of sociocultural factors, which can development services result in poor quality and low uptake of services. There is an overall shortage of trained health The pool of suitably qualified women to take up and education professionals in the lowlands (Fig- government jobs in health and education is partic- ure 3.13). The government struggles to find suitably ularly scarce. One reason for this is the shortage qualified personnel to fill vacancies. Only 70 percent of girls completing high school in Afar and Somali of civil servants in the lowlands have university regions. The lack of female teachers is a deterrent degrees (Afar being the lowest with 57 percent), to families to send their daughters to school, which which is in stark comparison to 83 percent in the reinforces the problem of an inadequate supply of highlands. Civil servants in Somali region are the women in the workforce. Similarly, for the health youngest in Ethiopia (29 years), 4 years younger sector, the lack of local female health extension than the national average of 33. Correspondingly, workers is a deterrent to women to seek health ser- their years of service (and therefore experience) are vices, particularly in relation to reproductive health. shorter (eight years), with some four years less than Thus, without targeted interventions, this situation the national average. Because of the lack of locally will not be rectified and any progress in human available expertise, highlanders, who have distinct development will be difficult to achieve. cultural differences, possible biases and negative Turnover of staff, particularly at the woreda level, stereotypes toward lowland populations, and do is very high in the lowlands. At times this is due 48   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands to voluntary departures by the staff and at other Education System Analysis times due to widespread reshuffling of staff. The Education strategy not adapted to lowland context top five drivers of voluntary turnover in the edu- The provision of education services for lowland cation sector, as stated by Directors and Heads of communities in Ethiopia was given special atten- Organization, are: (i) the lack of opportunities for tion in the 1994 National Education and Training promotion (33%), (ii) the workload is too much (19%), policy, which clearly spelled out the government’s (iii) work climate (17%), (iv) poor pay and benefits intention to offer special support to lowland com- (15%), and (v) problems with one’s manager (8%). munities that had been deprived of educational The top five drivers of voluntary turnover in the opportunities in the past. The policy provided for health sector, as stated by Directors and Heads of a new structure, decentralized management, and Organization, are: (i) the lack of opportunities for local languages as mediums of instruction. The promotion (38%), (ii) problems with one’s supervisor lowland education service delivery was guided by (18%), (iii) work climate (13%), (iv) poor pay and bene- a Pastoralist Education Strategy, developed in fits (10%), and (v) workload (10%). Interestingly, poor 2008, which proposed institutionalizing different pay and benefits are placed lower on the list than educational modalities (Alternative Basic Education one may expect. [ABE], mobile schools, para-boarding schools, hos- Health workers often do not remain in their posts tels, distance education, adult education, educa- for more than a year in the lowlands. In December tion radio programs, and formal primary schools), 2018, a large-scale reassignment of government improving the quality and relevance of primary personnel took place in Afar and Somali regions, education, and addressing cultural, environmental, with new staff being appointed without any hand- and economic constraints of lowland communi- over mechanisms in place to ensure orientation and ties. Despite the recommendations of a variety of onboarding of the new appointees. In general, there education strategies and modalities, most of them is a lack of appropriate deployment, retention, and have never been implemented due to technical, motivation mechanisms, and few monetary and managerial, and financial limitations. nonmonetary incentives available, to encourage The new Pastoral Education Strategy developed staff to stay for longer. in 2017 builds on the previous strategy. In addition, The Management Information System for education the MoE has developed an Education in Emergency and health care in the lowland areas has very limited Operational Strategy—2017, to create a nexus capacity to generate, analyze, disseminate, and use between humanitarian and development initiatives information and assist in the mobilization of resources in emer- In the heath sector, data are often inaccurate, gencies. The strategy sets out defined roles and late, incomplete, and inconsistent. Third-party responsibilities of partners, harmonizes emergency verification is almost nonexistent without checks response, applies well-coordinated monitoring and and balances for false reporting. Inadequate sys- evaluation approaches, and seeks to minimize and/ tems for quality data collection and analysis for or avoid duplication of efforts (MoE, 2017). preprimary and primary education are also evi- These policies and strategies provided impetus dent. Education data are often not available or and opportunities to promote educational develop- unreliable and often appear to be inconsistent. ment in the lowland areas. However, there are policy Moreover, there are no joint databases and reports and implementation gaps in the education service from EMIS, NEAEA, and GEID within the Ministry delivery system in the lowlands. The national edu- of Education that would provide insights into the cation and training policies are developed based on education system of the country in general and a “one-size-fits-all” approach that does not explic- that of the lowlands, in particular. The utilization itly address the different livelihoods of lowland of modern ICT systems to assess the education communities. Though it suggests “special support,” system performance and development in lowland this support is not backed up by appropriate edu- areas is only at the rudimentary stage. Some of cation service delivery modalities which can fit the the factors that contribute to the underdeveloped distinctive ways of life of pastoralists. National EMIS in the lowlands include lack of ICT struc- strategies, such as timely delivery of school grants, tures, remoteness of the schools and health facil- are often not fully reflected in the budget (block ities, and deliberate incorrect reporting to conceal grant) at regional and woreda levels, and the Minis- low performance. try of Education has little leverage to overcome the Chapter 3: Low Human Capital as a Source of Vulnerability   49 financial and human resource shortages and weak there are still many children in mobile pastoralist political commitments of the decentralized educa- families who do not have access to schooling. As tion service delivery system. In various policy and an Afar Elder said, “We will learn if you bring us a strategy discussions, it was suggested that a sep- school which has a foot and can walk with us.” arate education policy and road map for the low- The use of mother tongue as medium of instruc- lands needs to be developed instead of having one tion could be promoted further in the lowlands. policy for both highlands and lowlands. However, Ethiopia has progressed well in promoting mother this has never materialized. tongues as a medium of instruction, compared to many African countries. In fact, most students in Even though various education delivery modalities grades 1–4 receive their education in their mother were recommended in the pastoralist strategies, tongue. However, in SNNPR where 45 languages few of them are being implemented are used, there is room for further expansion of While formal education is relatively accessible education in local languages. The major constraint at primary levels in the more moisture reliable to such expansion is the limited number of local lowlands and agropastoral or drought-prone low- teachers. lands of south Omo and Oromia, it has inherent Other modalities recommended in the pastoral- limitations for more mobile pastoralist children in ist education strategy for delivering education in Afar and Somali regions. Expanding static formal remote areas include mobile schools, para-boarding schooling and inflexible academic calendars for schools, hostels, distance education, adult educa- mobile and semi-mobile lowland communities is tion, and education radio programs. However, they unlikely to deliver the intended results. Pastoralists have not yet been implemented as planned. More- have the lowest proportion of school-age children over, preprimary education, specifically 0-class who have ever been to school. education is not fully functional and implemented Over the last few years, the most common in the lowlands as per the required national stan- means of addressing the gap in educational access dards due to resource shortages in terms of both in pastoral areas has been ABE (see Box 3.1). How- finance and personnel, lack of trained teachers, ever, it has not been implemented according to the poor facilities and materials, classrooms, poor standards and guidelines set by the MoE. The major incentives, and lack of a proper functioning system strength of ABE is that it is a form of education to run the program.59 In general, 0-classes are not delivery that does not undermine the pattern of accessible for most young children because the labor organization and mobility on which pastoral- distance from home to schools is sometimes up to ist livelihoods depend. Despite the potential of the 5–8 km in pastoralist and drought-prone areas. fixed ABE modality to extend educational access, There is no strategy and plan to ensure edu- cation services can continue during periods of emergency. The education service delivery system, Box 3.1: Alternative basic education particularly in the drought-prone and pastoralist areas, is unable to adapt and flex in response to The Alternative Basic Education (ABE) program is an orga- nized basic education program designed to be flexible and emergencies such as drought and conflict. Both responsive in character, delivered outside formal school- ABE and formal primary schools in rural settings ing, and offered in small cost-effective buildings. It is an in conflict-prone areas are generally closed until alternative to formal education and has a four-year cycle. peace is restored. For instance, currently many It is equivalent to the first cycle formal primary education schools in Oromia and Somali are closed due to (Grades 1–4). Under this program, education is provided for conflict. Many children miss school every year due out-of-school children. ABE is meant to improve access to quality basic edu- to an absence of proactive planning and budget- cation for children excluded from or beyond the reach of ing, extended periods between assessments and formal education, which especially concerns children in responses, and the absence of an organizational pastoral and agropastoral areas (MoE, 2010) and “disad- structure within the education system to deal with vantaged and previously under-served ethnic groups” (MoE, emergencies. As the lowland areas are vulnerable 2015). It targets children at the age of being enrolled in the primary cycle of primary school (7 to 10 years old) but also to natural disasters, the absence of an emergency overage children with ranges from 11 to 14 years old (MoE, 2011). 590-class is a one-year pre-primary program, delivered by primary Source: MoE. schools, organized for children before they enter Grade 1. 50   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands response system will affect students’ participation expanded to pastoralist and lowland areas without and quality of education. Moreover, the Education sufficient adaptation to the change of context. Service Delivery Strategy (ESDS) is not integrated There are few lowland tailored health service deliv- with the early warning and disaster risk manage- ery modalities for drought-prone pastoralist popu- ment system and early preparedness and response lations, and usually they are introduced years after planning. being introduced in highland areas. In recognition of the needs of the lowlands, the Federal Ministry of Health is providing health sys- Health System Analysis tem strengthening to support the four emerging The health transformation agenda is relevant, regions and least performing zones in agrarian but implementation remains poor regions. The Ministry of Health has developed a Currently, the health care delivery system is guided five-year plan to address geographic inequity, pro- by a 20-year rolling Health Sector Transformation viding budget subsidies and deployment of addi- Plan (HSTP), which is part of the second Growth tional health professionals to the lowland regions. and Transformation plan of the country. HSTP-I, However, in practice the support is irregular and the five-year health sector strategic plan (2015– insufficient. 2020), is mainly focused on four transformation The FMOH has also embarked on optimization of objectives alongside other programs. These are: the health extension program and has developed a health extension strategy that seeks to better fit • Woreda transformation with the aim of woreda the lifestyle of lowland populations. The program health programs delivering improved health envisages (i) improving the leadership, governance, outcomes. The woreda transformation aims at and performance management capacity of health narrowing the gap between the high and low per- sector managers, (ii) community engagement, forming woredas. It has three components: high (iii) revisiting the service delivery modality, (iv) pro- performing primary health care units (PHCUs, viding special support, (v) enhancing integration health posts, health centers, and primary hos- with key sectors, and (vi) close follow-up, moni- pitals), expansion of model kebeles (healthy toring, and evaluation. While the Health Extension community behavior), and financial protection Program (HEP) implementation has not yet begun, through expansion of community-based health there is a guideline for its implementation in the insurance (CBHI). pastoralist areas. The guideline outlines roles and • Transformation in equity and quality of health responsibilities of all actors but does not include care, which has elements to ensure equity of key aspects such as implementation arrangements, access, service utilization, and quality of care. financing, procurement, and monitoring. • Information revolution which aims to reform the There are some promising emerging practices in methods and practice of collecting, analyzing, alternative health care delivery in the lowlands. Two presenting, and disseminating information. It is such examples are discussed in Box 3.2. The fea- a radical shift from traditional ways of data utili- sibility and sustainability of potentially scaling up zation to a systematic information management these experiences, however, would require careful system. review, as currently they are at a relatively small • Caring, respectful, and compassionate health scale and financed off-budget. workforce—this is a multi-pronged approach and There are several recommendations to be consid- calls for a mechanism to promote the values, ered to strengthen the delivery of health services in hopes, and aspirations within health profession- the lowlands. These include: als that brought them into health care in the first place. • Tailor the design of national service delivery platforms (Primary Health Care Unit, Health While all these transformation goals are relevant Extension Program, etc.) to the specific context to the lowland areas, their implementation remains of lowlands and pastoralist areas; poor due to an absence of effective implementa- • Establish a strong context-specific compre- tion arrangements, lack of political prioritization, hensive early warning system. Strengthening and financial and institutional capacity. Conven- emergency preparedness and response activities tional health service delivery modalities are often through an innovative approach such as risk- designed and piloted in highland areas and then informed programming and creating a lowland Chapter 3: Low Human Capital as a Source of Vulnerability   51 Box 3.2: Promising practices in health care delivery in the lowlands The DFID-funded Reproductive Maternal and Neona- and implementing Phase I of the PBF Show Case Project in tal Health Innovation Fund (RIF). The RIF project aims to Borana, Ethiopia, from May 2015 to June 2018 in nine health improve the health outcomes of women, newborns, and facilities (Yabello hospital and eight health centers) in the young people in the pastoralist regions of Ethiopia using four woredas of the Borana zone. The main objectives of the innovative interventions to increase the demand for ser- PBF Showcase Project were improving the quantity (coverage), vices. To implement RIF interventions, FMOH has contracted quality, and health information system by introducing 16 Sub-Recipients (SRs) over two funding rounds. SRs include performance-based financing complementing the existing gov- RBHs, RDAs, NGOs, FBOs, and professional associations. In ernment input-based financing system. The recently conducted the Somali region, SRs have worked collaboratively under the external evaluation has found that the six elements/principles leadership of the ESRHB to promote and improve RMNCH upon which the approach of PBF is based were successfully services through co-opting TBAs and mother’s support applied. The separation of roles/responsibilities, formalizing the groups; using solar powered mobile health centers for hard- same with contract agreements (accountability) and linking to-reach communities; and through engaging key influencers payments with verified results, vetting performance reports and gatekeepers in the use of mini-media for IEC. Final eval- through an independent verifying body, and increasing auton- uation of the RIF project shows that “there is strong evidence omy (decision space) to the health facility on the use of the that interventions improved service utilization and mutually subsidy earned through their performance improved the func- reinforcing to improve the uptake of RMNH services.” tionality of the health system in the pilot area. The quantity of CordAid has experimented with Performance Based health services has increased—in some health centers increas- Financing (PBF) in the lowlands. Cordaid has been financing ing to seven-fold over the three years of support. health emergency fund will contribute to create Conclusions more resilient systems; Across a range of HD outcomes, from school atten- • A model for financial protection that fits with dance to infant and child mortality rates to severe the lifestyle of lowland populations needs to acute malnutrition, the lowlands do worse than the be designed. Community-Based Health Insur- highlands. Accessibility and quality of services are ance (CBHI) as currently designed may not be also worse. Due to the challenging circumstances appropriate; in the lowlands, including climatic conditions, low • The coordination of Performance Based Financ- population density, and settlement patterns, sec- ing (PBF) with CBHI or other financial protection toral ministries have developed specific strategies mechanisms could be a practical, relevant, and and approaches for delivery of human development financially sustainable approach to create resil- services in the lowland areas—especially for reach- ient systems that can withstand shocks. Experi- ing pastoral populations. However, these strategies ence from Borana shows promising results in the remain largely untested. lowlands; The main constraints to basic service delivery in • A specific human resource development and the lowlands revolve around limited budgets and management program should be developed. A competing regional priorities; low capacity and well-designed and systematic incentive scheme availability of skilled professionals stemming from that includes financial and nonfinancial pack- low education, especially of girls; and lack of incen- ages to attract qualified front line health workers tives and accountability systems to ensure results. is essential; Compared to the large human capital deficit in the • Establish effective engagement of local organi- lowlands, regions and woredas tend to allocate too zations and nongovernment actors to promote little of their budgets to their health and education culturally appropriate social and behavioral sectors, with minimal resources available for capi- change communication (SBCC); and tal and operational costs. Afar and Somali regions • Improve coordination of NGOs supporting have predominantly focused on investments emergency responses in the health sector and in water and rural development to the relative strengthen integration of “emergency” health neglect of education and health, while Gambella services into the national health system and rou- tine service delivery. 52   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands and Benishangul-Gumuz have preferred to allocate approaches; (iii) introduce Specific Purpose Grants, expenditures to rural roads, education, and health. targeted to regions/woredas with low human devel- Overall, there needs to be a much stronger focus opment outcomes; (iv) align civil servant incen- on improving the human capital in these areas, tives and pay to results; and (v) recentralize some especially in the drought-prone lowlands. Without local and regional functions to higher levels and this there will be little chance to break the cycle of review mechanisms of local planning enforcement vulnerability. This chapter recommends to (i) allo- and sectoral budget allocation in view of national cate more on-budget resources for human devel- targets. opment in the lowlands; (ii) introduce results-based Chapter 3: Low Human Capital as a Source of Vulnerability   53 Poverty and Vulnerability in the Ethiopian Lowlands Chapter 4: Climate Change and Environment as Sources of Vulnerability Summary southeast and south and wetter lowlands to the This chapter sets out the geographical and climate west along the South Sudan and Sudan borders context of the lowlands and explores some of the (see Map 4.1). Administratively, they cover parts of longer term climate trends affecting them. The moist Tigre and Amhara, the four “emerging” regions of western lowlands of Gambella and Benishangul-Gumuz Benishangul-Gumuz (BG) and Gambella in the west, are significantly wetter than the drought-prone low- and Somali and Afar in the east and southeast, as lands of Afar, Somali, and parts of Oromia and SNNP well as parts of Oromia and SNNP regions in the regions. Despite these trends, which suggest a long- south (see Map 4.1). term slight increase in rainfall, the most significant Hurni (1998) divides the Ethiopian lowlands into features of these environments are their enormous four agroecological zones (AEZ) based on elevation rainfall variability, lack of surface and groundwater, and rainfall.60 With an elevation threshold at 500 m and ability to regenerate after drought. Water devel- and annual precipitation at 900 mm/year, the four opment is critical but extremely expensive and, unless AEZs are: (i) dry Berha (<500 m, <900 mm/year); integrated into range management plans, may cause long-term environmental damage. 60 The associated shapefile with AEZ extents was further refined by Hurni (2008) and provided by the Ethiopia Water and Land Resources Introduction Center (WLRC) and available as part of EthioGIS-II database (Krauer Ethiopia is a country of huge geographic diversity et al., 2015). Without loss of generalization, minor adjustments to geometry were made, such as smoothing boundaries and a merging with a central highland massif, where altitudes rise of the ‘wet Kolla’ AEZ as part of the ‘moist Kolla’ AEZ, and is reflected to above 4,000 meters, surrounded by an apron of in the current definition specifying precipitation is >900 with no upper semiarid lowlands below 1,500 m in the northeast, limit (the updated classification by Hurni (2008) specified precipitation range >900 and <1,400 mm/year). 54 Map 4.1: Administrative regions with highland/lowland relief (ii) dry Kolla (500–1,500 m, <900 mm/year); the arid and semiarid, drought-prone lowland areas (iii) moist Berha (<500 m, >900 mm/year); and which cover most of Afar and Somali regions and (iv) moist Kolla (500–1,500 m, >900 mm/year). The parts of Oromia and SNNP. dry lowlands—dry Berha and dry Kolla—cover most Contrary to the conventional narrative that of the lowlands, completely characterizing those rainfall has been declining in recent years, lowland areas in Afar and Somali as well as large areas rainfall appears to have increased slightly during of Oromia, Tigray, and Benishangul-Gumuz (see the years 1982–2017, with the western lowlands Map 4.2).61 Moist Kolla is predominantly found along showing the largest increases (see Map 4.3). A the western periphery of the highlands and pockets summary of annual mean precipitation for high- in Oromia to the east which also border the high- land and lowland as well as lowland AEZs for lands. Moist Berha (<500 m) is limited to the west- three time periods is given in Table A.2 in Annex 1. ern areas of Gambella bordering South Sudan. Nationally, the average rainfall over this period was roughly 805 mm/year with the lowlands receiving Rainfall only half the rainfall received in the highlands. The Rainfall is highly variable both in distribution and data also reveal considerable variation between amount. This variability is particularly marked in lowland regions. For example, despite average annual rainfall of 627 mm/year in the lowlands from ­1982–2017, Benishangul-­ Gumuz (BG) received 61 1,220 mm per year on average, which was nearly Regional and subregional boundaries are based on shapefiles from the following site: https://datacatalog.worldbank.org/dataset/ four times that of Somali region (331 mm/year). ethiopia-woreda-districts Within the lowlands the different agroecological Chapter 4: Climate Change and Environment as Sources of Vulnerability    55 Map 4.2: Ethiopia’s agroecological zones zones further highlight spatial differences. For Gambella 3.0 mm/year), SNNPR (5.4 mm/year), example, in dry Berha (i.e., the lower lowlands) and Tigray (2.5 mm/year) showing respectable areas, the 36-year average is 285 mm but varies long-term increases (see Table A.3, Annex 1). The considerably by region (i.e., 742 mm/year in SNNP lowlands of Somali and Oromia in contrast indicate vs. 228 mm/year), underscoring the considerable a (small) decreasing trend at –0.6 and –0.3 mm/yr. diversity across ‘lowland’ areas. Focusing on the most recent period (2000–2017), Comparing rainfall from the earlier period we again find that for a majority of the lowlands ­ (1982–1999) with the later period (2000–2017) there is an increasing precipitation trend averaging (see Maps 4.4 and 4.5 respectively) nationally 2.5 mm/year (Map 4.5 and lower panel of Table A.3, there has been a slight increase (801 mm/year vs. Annex 1) with the lowlands in BG, Gambella, and 809 mm/year), a trend that is reflected in both the SNNPP showing fairly large increasing trends (8.2, lowland (624 mm vs. 630 mm) and highland areas 4.6, and 7.1 mm/year respectively). By AEZ, large (1,142 mm vs. 1,154 mm) as a whole. However, the increases during this period were found in the moist data do not support wetter conditions in all regions. Berha (3.7 mm/year) and moist Kolla (6.2 mm/year). In Somali region, for example, annual mean pre- Results for lowland areas in Afar and Tigray indicate cipitation since 2000 is less than the 1982–1999 negligible change (0.1 and –0.1 mm/year respec- period for both dry Berha and dry Kolla zones. tively). Within Afar, the increasing trend in the dry Nationally, the trend has seen an increase of Berha AEZ (0.9 mm/year) is offset by the declining a little over 1 mm/year between 1982–2017 (see trend in the dry Kolla zone (–0.6 mm/year). Table A.3, Annex 1) although there is considerable spatial variation (green represents increasing while red indicates a decreasing trend in Map 4.3). In the Temperature lowlands, the long-term trend is one of increasing Regarding temperature, a clearer picture of precipitation (0.9 mm/year), with lowland areas recent climate trends emerges. Oxfam reported of Amhara (3.0 mm/year), BG (4.2 mm/year), that “between 1960 and 2006, the mean annual 56   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Map 4.3: Trend in annual rainfall 1982–2017 (mm/year) Map 4.4: Trend in annual rainfall 1982–1999 (mm/year) Chapter 4: Climate Change and Environment as Sources of Vulnerability    57 Map 4.5: Trend in annual rainfall 2000–2017 (mm/year) temperature increased by 1.3°C, at an average rate in severe drought stress days, and other relevant of 0.28°C per decade” (Regassa et al., 2010, 5). agro-meteorological variables. However, this anal- Temperature increases reduce soil moisture and put ysis also highlights the considerable interannual stress on plants, animals, and humans, leading to variability, making it difficult to capitalize on trend a variety of negative outcomes, including reduced improvements in weather conditions. crop yields and plant and human disease. To complement the long-term trend analy- The spatial distribution of mean annual tem- sis, an in-depth assessment of recent trends in peratures (1986–2016) is shown in Map 4.6. For growing seasons for select lowland locations was lowland areas (<1,500 m) annual averages are plot- conducted.62 This contributes to the more gen- ted in Figure 4.1, which tends to show a generally eral understanding of how the changing climate is increasing temperature trend. Specifying a simple affecting agricultural and pastoral systems across linear regression where temperature is a function the lowlands by assessing rainfall and tempera- of time indicates that temperatures have been ture trends during the primary growing season, increasing by approximately 0.03 degrees/year, a including the frequency and timing of rains and the finding which is aligned with other analyses show- incidence of drought-stress days. This analysis uses ing increases of 0.28 degrees per decade (Regassa a large dataset63 on temperatures, precipitation, et al., 2010). relative humidity, and wind speed, a daily ‘virtual’ ag-meteorology station worth of data for over Growing Season For some lowland areas (i.e., Gambella and SNNPR), 62This work was carried out by aWhere. an in-depth analysis of a sample of locations indi- 63This foundation asset is the innovation of aWhere, Inc., a USA-based cates an improvement in growing season condi- agricultural intelligence Benefit Corporation. More information at: tions as implied by an increase in rainfall, a decline http://blog.awhere.com/foundation-big-data-asset 58   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Map 4.6: Mean annual temperature (1986–2017) Figure 4.1: Lowland (< 1,500 m elevation) mean annual temperature and trend 26.6 26.4 26.2 Temperature (˚C) 26.0 25.8 25.6 y = 0.0291x–32.429 25.4 R² = 0.6254 25.2 25.0 24.8 1980 1985 1990 1995 2000 2005 2010 2015 2020 13,000 locations distributed across Ethiopia. Given Trend estimates for annual total rainfall, maxi- the data-intensive nature of this exercise, analyses mum temperature, and number of drought stress were limited to just 12 locations (see Map 4.7). days for each location are given in Table A.4 in While the lowlands of Ethiopia are heterogeneous Annex 1). Since 2008, rainfall in the Afar and in terms of the pattern of rainy seasons through- east-central clusters in Somali region have seen out the year, this study focused on the “best” rainy increasing rainfall trends, as have the sample season for each location studied. For each location, locations in SNNPR and Gambella. In contrast, the ‘optimal’ 90-day growth period was determined declining rainfall is found for the locations in Oro- based on agroclimatic characteristics and applying mia (–7.4 mm/year), BG (–3.1 mm/year), Amhara the same criteria across sample areas to ensure (–15 mm/year), and Tigray (–13.5 mm/year). A trend consistency and comparability. in high stress drought days is also on the decline for Chapter 4: Climate Change and Environment as Sources of Vulnerability    59 Map 4.7: Distribution of seed locations for in-depth agroecologic trend analysis most areas. Only in the far southeast sample areas In general, an analysis shows that while there of Somali #5 and #6 are there trends toward more are statistically significant trends in several agro- of these extremely drought stress periods during meteorological variables in several lowland sam- the 90-day season. For most of Ethiopia, there is ple areas, given high year-to-year variability it will a weak trend showing a reduction in extremely dry be challenging for farmers and/or pastoralists to conditions. The interannual variability, however, is respond to any potential weather improvements. high, reflecting again the variability of dry periods Without an additional analysis it is unclear how during the growing season across the lowlands. meaningful favorable trends in weather condi- The second trimester, days 31–60, are critical tions are when faced with such high degrees of for plant growth. A slight reduction in the number variability. of extreme drought days continues to be observed except for the southeastern areas (#6, #7, and #8), Climate Change where a weak trend toward more drought stress Studies projecting climate change suggest uncer- days continues. Growing season days 31–60 show a tainty regarding the exact impact of future rising mixed signal for the lowland sample areas in terms temperatures on precipitation. Coastal East Africa, of moist days. Variability in the number of very including Ethiopia, is semiarid due to the cool moist days dominates the signal in the west and waters offshore Climate models tend to underes- northwest sample areas (#9, #10, #11, and #12), timate east-west sea surface temperature (SST) while the weak trend shows some spatial consis- gradients in the Indian Ocean and to weaken them tency in the southeast (a few more moisture laden further under greenhouse gas forcing (Yang et al., days) as well as the southwest (#8 and #9). 2015). Hence, they predict coastal East Africa to 60   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Map 4.8: Seasonal rainfall in Ethiopia’s eastern and southern pastoral areas get wetter as surface water temperatures are pro- Indian Ocean Dipole64 episodes result in drought jected to rise in the western Indian Ocean. However, events in the southern pastoral areas. The different in fact coastal East Africa has become drier over pastoral zones and associated seasonal rains are the past century (Williams and Funk, 2011). This presented in Map 4.8. While not always the case, may partially be due to natural variability. This below average rainfall and drought in one pastoral uncertainty carries over to Global Climate Model area is often mirrored by above average rainfall in (GCM) projections. Funk et al. (2012) project less the other. rainfall through 2039, though this is largely an Historically, Ethiopia’s drought hot spots have extrapolation of observed trends. The IPCC Fifth occurred in the northeast, southeast, and southern Assessment Report reflects greater uncertainty: lowlands (see Map 4.9). Eighty-seven percent of the “GCM projections over Ethiopia indicate a wide identified 450 drought-prone woredas are in low- range of rainfall spatial pattern changes and in land regions—169 of these woredas are also flood some regions GCMs do not agree on the direction of prone. Of the 450 identified woreda, 190 are priority precipitation change, for example, in the upper Blue 1 hot spots, and 169 are vulnerable to both floods Nile basin in the late 21st century” (Niang et al., and droughts. 2014, 1210). All droughts are characterized by phases: early onset, onset, peak intensity, withdrawal—following Droughts the onset of rains—and recovery. The early onset While there are uncertainties about future climate of a pastoral drought is characterized by a reduc- change trends—either wetter or drier—what is clear tion in the availability of surface water and drier is the high degree of variation from one year to the and poorer quality forage for livestock, resulting in next and an increased likelihood of extreme weather declines in milk production. In the event of a second conditions in the drought-prone lowlands. Serious poor/failed seasonal rain, the onset phase is char- droughts in the Horn of Africa are associated with acterized by further and more dramatic declines the El Niño–Southern Oscillation (ENSO) phenom- ena: El Niño episodes result in drought events in 64Indian Ocean Dipole episodes are La Niña episodes that are centered Ethiopia’s northern pastoral areas; and La Niña and on the Indian Ocean as opposed to the Pacific Ocean. Chapter 4: Climate Change and Environment as Sources of Vulnerability    61 Map 4.9: Drought hot spots in the availability of water and forage resources. and quality of forage in normal years. In this study Livestock also expend increasing amounts of time a pastoral drought is defined as two consecutive and energy trekking between water and forage poor/failed seasonal rains over 12 months. A pro- resources and hence lose weight, and milk produc- longed pastoral drought is defined as three or more tion declines sharply. In the event of a third poor/ consecutive poor/failed seasonal rains or 18 months failed seasonal rain, the distances trekked and of poor/failed rains. In the last 10 years, Ethio- the quality of forage become such that significant pia’s northern and southern pastoral areas have livestock body weight is lost, and, without rain experienced at least one prolonged drought, while and access to improved forage and water, animals some have experienced two such droughts (see slowly starve. Figure 4.2). There are four types of drought: meteorolog- ical,65 hydrological,66 agricultural, and pastoral. Vegetation Agricultural drought in Ethiopia is defined as a Inhabiting some of the harshest ecologies in Ethio- 30 percent reduction in crop yield,67 while pasto- pia, livestock convert sparse vegetation—seasonal ral drought has been defined as ‘a lack of forage grasses and the leaves of shrubs and trees—into availability as result of a particular sequence of milk and meat that can be utilized by human pop- meteorological drought’.68 The lack of forage might ulations for food, and hence enable pastoralists to however equally describe an extended dry season, occupy areas of Ethiopia that would otherwise be as there are significant variations in the availability uninhabitable. The quality and extent of grazing and browsing available are therefore critical to pas- 65 Meteorological drought happens when dry weather patterns domi- toral livelihoods, but assessing trends in lowland nate an area in Ethiopia when in normal years seasonal rainfall would be expected. vegetation, given the spatial and temporal vari- 66 Hydrological drought occurs when low water supply becomes evi- ability of rainfall in the lowlands, is extremely diffi- dent, especially in streams, reservoirs, and groundwater levels, usually cult. Improvements in vegetation cover in one area after many months of meteorological drought. are frequently offset by deterioration in another. 67 World Bank (2015). 68 FAO (2011). Drought. FAO Land and Water. http://www.fao.org/ Grasslands also show enormous ability to regen- docrep/017/aq191e/aq191e.pdf erate after heavy rainfalls. A variety of vegetation 62   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Figure 4.2: Prolonged recent droughts recorded by pastoralists Northern pastoral areas: Afar and Northern Somali Southern pastoral areas: Southern Somali, Southern regions Oromia, and southwest SNNP regions –1984–1985 –1984–1985 –1988–1989 –1994 –2000–2001 –2010–2011 –2015–2016 –2016–2017 Map 4.10: Trend in NDVI 2000–2017 (%) indices can be employed to provide empirical mea- in bush encroachment and invasive inimical species sures of vegetation. In this study, trends in vege- to livestock. Nevertheless, despite these limitations, tation are assessed using two indices: Normalized they give some indication of the direction of change Difference Vegetation Index (NDVI) and Enhanced and trend information in vegetation cover in the Vegetation Index (EVI). Each of these indices have lowlands. been used extensively in the literature for assess- For the period 2000–2017, vegetation in the ing vegetation cover and direction of change. But lowlands as measured by NDVI shows a small neither are very good at assessing the quality of increase in contrast to the highlands where there vegetation available for livestock grazing. Increases has been a decrease (see Map 4.10, and Tables A.5 in greening, for example, can also reflect increases and A.6 in Annex 1). During this extended period, Chapter 4: Climate Change and Environment as Sources of Vulnerability    63 Map 4.11: Trend in NDVI 2000–2009 (%) vegetation has increased across much of Somali, external interventions. This management may be Oromia, and Gambella. In contrast, lowland areas in influenced by a variety of factors, some deliberate Afar, Amhara, and Tigray exhibit a downward trend such as bush and grass burning, range enclosures, (SNNP largely unchanged). Looking at the two sub- shifts in the composition of herds from grazers to periods, there are some interesting differences that browsers, and introduction of permanent water emerge. For example, between 2000–2009 the points, to others which may be unintended, such lowlands in Afar appear to have experienced wide- as the consequences of over- and understocking spread declines (Map 4.11) compared to the later caused by population increase, conflict, and forced 2010–2017 period characterized by widespread displacement. increases in the NDVI trend (Map 4.12). In fact, in this later period, all lowlands in all regions except for SNNPR showed growth in NDVI (0.33%) and EVI Water Availability (0.23%) (see Tables A.5, A.6, A.7, and A.8 in Annex 1). Water availability is a critical component of low- By AEZ, the moist Berha and moist Kolla zones land livelihood systems. However, there is a wide reported large improvements during this period variety of conditions in the lowlands related to across all regions. In the dry lowlands, the story groundwater presence and availability. Large areas was a little more mixed with the dry Kolla showing have brackish or salty groundwater with high min- an overall increase (both according to NDVI and EVI) eral content that makes it unsuitable for drinking but with regional variations. water and most other purposes. As a result, water Vegetation, however, is not only responsive to resources in pastoralist areas, particularly in the the amount and distribution of rainfall and tem- two regional states of Afar and Somali, are usually perature, but also to human intervention. In pas- scarce and costly to access for rural water supplies. toral ecosystems the rangelands are a product Surface water is largely confined to the perennial of human management of the range as well as and seasonal rivers that run through the lowlands. 64   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Map 4.12: Trend in NDVI 2010–2017 (%) The northern and central highlands drain westward groundwater potential include: (i) the Shinile area into Ethiopia’s most extensive river system, the in northern Somali region, (ii) the Jerer valley, and Abbay, or the Blue Nile, or into the Tekeze River, (iii) the shallow groundwaters in the Fafan valley in a tributary of the main Nile, and the Baro river, a the northern part of the regional state. Elsewhere branch of the White Nile. The eastern highlands in the region, groundwater of limited extent can be drain into the Awash, Wabi-Shebele, and Genale- encountered along the banks of the major rivers. Dawa rivers. The Awash River never reaches the Shallow groundwater also occurs in the Ogaden sea but is ultimately absorbed into a succession zone. of lakes and marshes near the Djibouti border. The Access to improved water supply and sanitation Wabi-Shebele and Genale-Dawa rivers flow across facilities in pastoralist areas of Afar, Somali, and the eastern lowlands. In the south, the Omo River pockets of pastoralist areas in Oromia and SNNPR drains into Lake Turkana, and several streams flow lag highland regions and the national average. into the other Rift Valley lakes. In the southeast, the Access levels for water and sanitation range from mountains of Arsi, Bale, and Sidamo drain toward 39–61 percent, and sanitation coverage range Somalia and the Indian Ocean, but only the Genale from 6–21 percent, respectively, compared to or Juba rivers permanently flows into the sea. See 62–95 percent and 41–76 percent respectively in Map 4.13 for major river basins in the country. highland areas (OWNP Phase II program, November Groundwater availability varies across the low- 2018). According to the Ethiopian WaSH Poverty lands. Afar regional state is underlain by several Diagnostic study, agrarian highland woredas were deep and prolific aquifers of regional importance. significantly more likely to have access to improved The high salinity of the waters in some areas, water than agropastoralist woredas—by about however, limits use of the water for irrigation and 10 percentage points. Map 4.14 highlights regional drinking purposes. In Somali region, by contrast, coverage rates by woreda. water tables in some areas run extremely deep and According to a survey conducted in lowland aquifers are low yielding. Notable areas for better woredas of Somali, Afar, and South Omo zones of Chapter 4: Climate Change and Environment as Sources of Vulnerability    65 Map 4.13: Major river basins SNNP regions (USAID, May 2016) 48 percent do not the alluvial beds of many rivers, they are not peren- have access to improved drinking water sources of nial and may last a few weeks to a few months which 49 percent were in Afar, 81 percent in Somali, after the rainy season. There are also areas that and 15 percent in SNNP. The same survey high- have sedimentary or volcanic formations, with deep lighted that of the 51 percent that have access to to very deep aquifers. However, some of the deep water in Afar, 86 percent of households use water aquifers could have a meager yield or may have from drilled boreholes. Baseline data collected in poor water quality, often with high mineral content 366 woredas in 2018 also indicated that 80 per- and high temperature, which may make the water cent of the woredas in SNNP, Somali, and Afar have unsuitable for drinking, small-scale irrigation, and below the national average access coverage (53%). even for animal watering. The same survey also suggests that there is very Given the nature of lowland pastoral livelihood low sanitation coverage in Afar (23%) and Somali systems, provision of permanent water in lowland (5%) regions. areas needs to be carefully considered in relation Several factors have contributed to the chal- to other natural resources, including its long-term lenges and disparities in water service delivery impact on grazing availability for livestock. Too between highlands and lowland areas. These much water can lead to permanent settlement include: and overgrazing. Environmental change induced Pastoralist and agropastoralist areas receive less by water availability may have permanent and rainfall than other regions, potential water sources negative land use consequences. As such, water are usually scarce, and groundwater sources are coverage needs in the lowlands and highland costly to access. While there are shallow aquifers in communities may not be comparable, as water 66   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Map 4.14: Regional water supply coverage distribution development in the lowlands needs to be part of an of accountability, community engagement, and integrated natural resource and livelihoods plan. capacity for repairs. The principal challenge in these areas is to The hydrology and geology in lowland Ethiopia understand how water development can contribute are complex, and little is known about the quality to provide adequate water for human consump- and quantity of water resources, which makes the tion and settlement, including exploiting untapped cost of reconnaissance before drilling expensive. irrigation potential, while not upsetting traditional Master plans studies of the different river basins adaptive strategies to environmental and climate focus only on significant tributaries and detail variability on which large numbers still depend for the surface water resources potential for hydro- their livelihoods. power and irrigation use. Information available The scattered settlement patterns of pasto- about groundwater resources in the lowlands are ralist and agropastoralist areas have made water fragmented. The hydrogeological data from drilled service schemes expensive to maintain. Given the boreholes in the lowland areas have not been transient nature of some communities, operation systematically collected and analyzed to assess and maintenance of sophisticated pumping sys- groundwater potential to orient design and man- tems have been inadequate. Systems that require agement of water service delivery. extensive pumping of deep groundwater sources, Cost of drilling and groundwater exploitation for example, require diesel to operate pumps and in many of the pastoralist regions is increasingly replacement of spare parts that are often imported expensive, with a technological shift toward deep (electro-mechanical equipment, steel casing, etc.). boreholes and limited drilling capacities in lowland Consequently, water investments in sparsely pop- regions. Although the knowledge base to orient ulated areas have fallen into disrepair due to lack investment in groundwater is fragmented, the Chapter 4: Climate Change and Environment as Sources of Vulnerability    67 reliance on groundwater is increasing. To reverse of Water, Irrigation and Electricity (MoWIE) the the high rate of drilling failure or lower than aver- national average is 11 percent, while in some low- age water yield or chances of boreholes “drying up,” land regions this rate is three times higher. the solution to date has been to drill deeper. Bore- As a result of these factors water development hole drilling in the lowlands is expensive and can in the lowlands needs to be undertaken with careful be as high as US$291,000 per borehole, against consideration of its wider environmental impact. a national median cost of US$128,000. Review of Investing large sums of money into widespread actual drilling contracts in Oromia, SNNPR, Somali, borehole drilling may not be the best way to use and Afar regions have revealed that average cost scarce budget resources and may do more long- of drilling a borehole ranges from ETB 4.69 million term environmental harm than good. in Somali to ETB 0.44 million in Oromia. Besides In the next chapter we examine changing land the high cost of drilling, there is also a higher rate use and land cover trends in the lowlands, and of nonfunctionality—according to the Ministry some of the principal drivers behind them. 68   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Poverty and Vulnerability in the Ethiopian Lowlands Chapter 5: Changing Land Use and Land Cover Summary to generate land cover maps temporally compa- rable, classification is based primarily on Landsat This chapter examines long-term trends in land use imagery. and land cover in the lowlands. These trends indicate an In considering the extent of change in grazing expansion of cropland, and a decline in grazing avail- and forage areas, lowlands grasslands are one of ability due to a variety of factors, including significant the most important landscapes in terms of the bush encroachment in some areas, land enclosures, and functions they provide for pastoralists. However, population increases causing increased pressure on, and from a land use and pastoral livelihoods point of competition and conflict over natural resources. Land view, grasslands represent only one portion of the fragmentation and loss of access to common grazing potential land available since grazing and forage areas are undermining established coping strategies. typically include areas such as shrub and bush and even woodlands (i.e., non-forest, woody vegetation). Introduction Additionally, grasslands are also one of the most Distinguishing long-term environmental changes difficult classes to accurately map and detect from short-term fluctuations due to rainfall vari- change in, especially in the lowlands, where signals ability is extremely difficult in many of these low- can be easily confused and distinction between land environments, but there are some long-term bush and shrub and other non-forest woodlands trends which can be identified with reasonable difficult to make given the available imagery. Rec- confidence. To better understand the changing ognizing that suitability and palatability of veg- landscape across the Ethiopian lowlands, land use etation in shrub and woody landscapes varies by and land classification maps were generated for livestock species, grasslands are combined in this 1986, 2000, and 2016 to look at trends in land use analysis with non-forest woody vegetation—shrub and land cover.69 Given the spatial extent and need and bush and woodland Land Use and Land Cover (LULCs)—to represent potential grazing and forage 69 The main source of data are LULC maps generated by the Water and land more broadly. Land Resource Center (WLRC), Addis Ababa University, for this study. Potential grazing and forage land, as repre- The methods described reflect the approach adopted by WLRC’s team of GIS and remote-sensing experts. For 2016, two additional sources of sented by combining grassland with shrub and data were used to supplement the LULC map produced by WLRC. First, brush and woodland LULCs, covers most of the settlement areas were revised using data from the European Space lowlands (Map 5.1). Since 1986 there has been a Agency’s 2016 Africa land cover map (ESA, 2018) to more accurately decline of 3 percent across the lowlands with nearly capture villages and the extent of urban areas. Second, data generated on the location and extent of large commercial farms as part of this all major regions experiencing declines (Table 5.1 lowland project were used to update cropland areas. and Table A.9 in Annex 1). The exception is SNNPR 69 Map 5.1: Distribution of lowland grazing and forage land (1986 to 2016) Table 5.1: Lowland grazing and forage (1986 and 2016) that cropland expansion is taking over some areas that had previously been used for grazing. Area (1,000 ha) Change (%) This slight long-term decline in cover may con- Region 1986 2016 1986–2016 ceal a much sharper decline in access to prime Afar 3,006.6 2,786.8 –7% grazing areas. Differentiating access from cover Amhara 3,204.6 2,826.6 –12% cannot be done by Landsat imagery interpretation Benishangul-Gumuz 3,844.9 3,545.5 –8% but requires a more detailed knowledge of changes Dire Dawa 49.3 47.2 –4% in grazing patterns by pastoralists on the ground Gambella 2,185.2 2,139.0 –2% and the extent to which land fragmentation, in the form of land enclosures, have reduced access to Hareri 3.6 3.9 7% certain areas for a variety of reasons (see below). Oromia 12,053.4 11,631.5 –4% Between 1986 and 2016 several trends in land SNNPR 3,939.6 4,136.5 5% use change were reflected in the highlands. For Somali 27,015.8 26,872.4 –1% example, lowland and highland regions both experi- Tigray   1,928.2 1,676.1   –13% enced reductions in forest and woodland areas, and Total   57,231.3 55,665.4   –3% increasees in cropland and settlements. However, the relative changes and magnitudes differed con- Source: WB calculations based on data from WLRC, ESA, and DevSeed. siderably, for example, the area of cropland which more than doubled in the lowlands (2.3 M ha to where there was a 3 percent increase—although in 4.7 M ha) compared to only a 30 percent increase SNNPR much of the increase appears to correspond in the highlands. In 1986 the forested area in the to decreases in forest cover (black shaded areas lowlands was slightly more than 5 M ha, while in in Map 5.2). In addition, there is a tendency for the the highlands forest covered approximately 5.9 M declines to appear along the highland/lowland tran- ha. The lowlands between 1986 to 2016 have expe- sition, which corresponds to increases in cropland rienced notably large declines in forested area of (pink shaded areas in Map 5.3) supporting the claim approximately 1.7 M ha (–34%), a reduction roughly 70   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Map 5.2: Distribution of lowland forest (1986 to 2016) four times that in the highlands where a decline of 2016. In the last three decades, total lowland crop- 440,000 ha or 7 percent over this 30-year period land more than doubled (2.3 M ha to 4.7 M ha in took place. Including area classifications for 2000 2016) with significant increases found in all lowland as well, these trends appear to be largely balanced regions. As a percentage, the cropped area in Afar over time with a few exceptions (i.e., bare land increased by a factor of four (431%) and in Gambella showing a decline in 1986–2000 but then increas- by a factor of three (292%) with large area increases ing in 2000–2016). Regional summaries and net in Oromia (790,000 ha), SNNPR (440,000 ha), change from 1986 to 2016 in lowland areas are pre- BG (313,000 ha), Somali (263,000 ha), and Amhara sented for each LULC in Annex 1, Tables A.9–A.19. (253,000 ha) (Map 5.3 and Table 5.2). These trends While forest area declined in all lowland regions, are generally in agreement with recent studies cov- the largest declines were in Oromia, SNNPR, and ering a similar time frame (Degife et al., 2018), but Somali, where the combined reduction totaled more this contrasts with some recent research such as than 1.6 M ha, with the area in Tigray nearly halv- Mekuyie et al. (2018) who looked at land use change ing from 59,000 to 32,000 ha (Map 5.2). Spatially, in southern Afar for a similar period (1985–2015) most of the decline appears to be reductions along and found a declining trend for cultivated land.70 river banks, although there are discernible clusters Spatially, much of the lowland cropland expansion such as in SNNPR which could be due to large-scale appears to have taken place along the escarpment, clearing for agricultural plantations and in different transitioning from highland to lowland (Map 5.3). patches in Somali region, one notable area being This expansion along the highland/lowland along the Kenya border near to Moyale (Map 5.2). There has been a steady expansion of cropland in lowland areas both in higher potential areas within 70While not directly comparable due to different areas involved, the the lowlands but also along the lowland-highland conflicting findings for Afar, such as an increase in area cultivated and bare land, are likely due to our combining data sources (in the case of interface, as farmers have pushed down from the cultivated area), reference years (i.e., 2015 vs. 2016) or different thresh- highlands in search of land to crop, with an esti- olds and classification criteria (i.e., designation of bare vs. sparsely mated doubling in cropland between 1986 and vegetated land captured by woody vegetation or grassland). Chapter 5: Changing Land Use and Land Cover   71 Map 5.3: Expansion of cropland Table 5.2: Lowland cropland (1986, 2000, and 2016) Area (1,000 ha) Change (%) Region 1986 2000 2016 1986–2000 2000–2016 1986–2016 Afar 27.7 45.7 147.1 65% 222% 431% Amhara 463.1 615.7 716.2 33% 16% 55% Benishangul-Gumuz 205.9 370.2 519.3 80% 40% 152% Dire Dawa 10.8 9.9 11.6 –8% 17% 7% Gambella 27.3 39.0 106.8 43% 174% 292% Hareri 6.2 6.0 7.0 –3% 16% 12% Oromia 588.9 947.1 1,378.5 61% 46% 134% SNNPR 414.5 446.7 853.8 8% 91% 106% Somali 147.1 250.7 410.7 70% 64% 179% Tigray   342.7 500.8 540.4   46% 8% 58% Total   2,234.1 3,231.9 4,691.3   45% 45% 110% Source: WB calculations based on data from WLRC, ESA, and DevSeed. 72   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands transition is likely being driven by changing con- of the population on a little over 40 percent of the ditions in the highlands. During the past several land area reflects the close relationship between decades, an increasing highland population coupled physiography, climate, economy, and population with limited potential land suitable for expanding distribution in the country. The average population cultivation has led to a spillover into the lowlands density in the highlands is almost eight times that (Hurni et al., 2005; Kassawmar et al., 2018; Zeleke of the lowlands, but the population in the lowlands and Hurni, 2001). is growing at a faster pace, with Somali and Afar The satellite data in Map 5.3 show the expan- regions experiencing the highest fertility rates sion of cropland (in pink) over a thirty-year period in the country at 7.2 and 5.5, respectively (DHS, between 1986 and 2016. This trend is confirmed 2016). by data from recent HCE surveys in 2011 and 2016 High population growth poses a future risk to showing an increasing number of full-time farmers Ethiopia’s development gains. The country’s popu- in the drought-prone lowlands (see Chapter 1). lation more than doubled between 1990 and 2016, Associated with this expansion in cropped area rising from 48 million to more than 100 million has been the expansion of land enclosures, both to (World Bank, 2017). With a current annual popula- protect cropland from animals but also for more tion growth rate of around 2.5 percent Ethiopia’s intensive forage production. These changes are population is expected to reach 160 million by 2035 likely to increase the risks in the future of inten- and nearly 205 million by 2050 (UN Population sified resource competition between pastoralists Division World Population Prospects, 2019). Afar looking for grazing for their livestock and farmers and Somali regions have seen their populations protecting their crops. increase by nearly a quarter in the last decade. The high proportion of people under 18 creates future Drivers of Land Use Changes challenges for labor markets and economic stabil- The drivers behind many of these observed land use ity, particularly if continued population growth can- changes are complex and interrelated and include: not be absorbed in nonagricultural sectors. Given (i) an increase in human population and settlement existing pressure on land resources in the lowlands, leading in many areas to increased cropping and and limited alternatives to extensive livestock land enclosures; (ii) excision of key dry season graz- herding, it is unlikely that future productivity gains ing areas for commercial agricultural plantations in lowland economies will be able to absorb such along some of the main rivers that run through a rapidly growing population. Population growth the lowlands, resulting in loss of riverine forest is likely to exert enormous pressure on traditional and grazing; (iii) introduction of permanent and lowland livelihood systems and the environment. semipermanent water points, such as boreholes Population density increased in the majority of and birkedas (large cement lined cisterns) in wet lowland woredas from 2000 to 2017, but decreased season grazing areas, e.g., in the Haud in Somali in 49 woredas, including 15 in northern Afar. Several region, leading to permanent settlement, agricul- woredas in the lowlands of SNNPR have seen an ture, and range enclosure in these once seasonally increase of more than 100,000 people in the last exploited grazing areas; (iv) bush encroachment by 17 years. Night-light analysis shows that Afar, Dire invasive species, such as Prosopis juliflora, reducing Dawa, and SNNPR have recorded the most signifi- grassland availability; and (v) increasing resource cant changes, as well as the area around Jijiga, the competition between groups leading to intercom- regional capital of Somali region (see Map 5.4). munal violence and the creation of de facto grazing What is notable from the data that have been exclusion areas, for example, in parts of Borana collected is how population density is increasing in zone along the Somali-Oromia regional border, and many parts of the lowlands, especially in Ogaden in Afar between Afar and Somali Issa pastoralists. and northeast Afar, recording some of the highest population growth rates between 2007 and 2017 (see Map 5.6). Population increase and urbanization Combined with the population increase is a The impact of these two on the lowlands is increas- growing trend toward increased settlement and ing. The highlands, which cover only some 40 per- urbanization in parts of the lowlands. Between cent of the land area of the country, are home to 1986 and 2016 the lowlands saw a 179 percent over 80 percent of the population who are depen- increase in settlement areas (see Table 5.4). Within dent on rain-fed agriculture. The high concentration pastoral regions themselves there has been a Chapter 5: Changing Land Use and Land Cover   73 Map 5.4: Night-light data on urban settlement changes (ha) 2000–2017 steady growth in the numbers and size of small Table 5.3: Population of selected urban centers in Somali towns. This has been encouraged more recently by region, Ethiopia (based on CSA census, and 2015 projected) humanitarian interventions in settling populations, % increase many of whom have been displaced by drought 1984 1994 2007 2015 1994–2015 and conflict, through provision of permanent water Jijiga 23,187 56,821 125,876 159,300 180 (boreholes) and relief supplies. This has eased the Gode NA 45,755 43,234 54,700 19 burden of service provision but also created many Degehbur NA 28,708 30,027 38,000 32 nascent new permanent settlements in once remote pastoral areas, such as in Warder in Somali Kebridehar NA 24,263 29,241 37,000 52 region. These small settlements open opportu- nities for development of alternative livelihoods. As a result these nodes of potential growth have predict that levels of urbanization (within Somali attracted both the poor and destitute as well as the Region) and migration (to destinations outside wealthy looking for opportunities to diversify their the region) will rise, while the proportion of Ethi- incomes. In Somali region, from 1984–2015, small opian Somalis who remain engaged in livestock-­ and medium-sized towns in the region grew at dif- dependent livelihoods will decline” (p. 30). ferent rates; some like Jijiga grew much faster than In Afar urbanization has been more pronounced— others, as they attracted people from rural areas as some parts of northern Afar have lost population (see Table 5.3). In 1997 the region had 30 towns other parts have gained. During 1997–2006 the with a population of at least 5,000 and 7 with population of towns with populations greater than more than 20,000. In 2007 the urban population in 1,000 grew at least 32 percent over this period. In the region was 623,000 or 14 percent of the total. 2007 the urban population in Afar as a percentage Over the last decade this increased considerably. In of the total was only slightly less than in Somali 2006 Devereux predicted: “Together with anecdotal region at 13 percent (CSA, 2008), but judging from evidence . . . that the aspirations of Somali youth night-light data and changes in the area covered might lie increasingly outside pastoralism, we might by settlements, Afar has seen a large expansion in 74   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Table 5.4: Lowland settlement (1986 and 2016) settlement area (see Table 5.4). Some of this expan- sion has been fueled by highlanders moving down Area (ha) Change (%) the escarpment looking for work on agricultural Region 1986 2016 1986–2016 plantations. The devastating effects of drought in Afar 235.0 2,482.0 956% the region also stimulates migration to towns. Each Amhara 2,029.0 5,490.0 171% successive drought damages livestock holdings and Benishangul-Gumuz 739.0 1,614.0 118% forces more stockless and near-stockless herders Dire Dawa 717.0 2,769.0 286% to towns. Many who have urbanized have done Gambella 937.0 1,348.0 44% relatively well; poverty has declined faster in urban areas, vulnerability is lower, and access to services Hareri 32.0 20.0 –38% is better. Oromia 3,415.0 5,000.0 46% Changes in population density in the lowlands SNNPR 466.0 4,765.0 923% since 2007 are shown in Maps 5.5 and 5.6. Somali 2,449.0 5,583.0 128% Tigray   374.0 2,671.0   614% Climate-induced migration Total   11,393.0 31,742.0   179% A finding of this study is that future climate change is emerging as a potent factor in the mobility of Source: WB calculations based on data from WLRC, ESA, and DevSeed. the population. Lowland climate migration levels Map 5.5: Population density in the lowlands in 2007 (CSA Census) Chapter 5: Changing Land Use and Land Cover   75 Map 5.6: Population growth 2007–17 lowland woredas are high and make up a higher proportion of total Map 5.7 shows the hot spots of climate change climate migration than in the highlands (Map 5.7). in-migration (in red) and out-migration (in blue). According to various scenarios (Figure 5.1) the low- The results suggest that the great majority of the land areas will see the following levels of climate movement in Ethiopia will be within the highlands— in-migration by 2050:71 that is, out of the northern highlands and into the southern highlands (to the south and east of the • Pessimistic/reference scenario (SSP4/RCP8.5): Great Rift Valley). For the lowlands, according to 5.9 m migrants the models run, the largest areas of in-migration • Climate-friendly scenario (SSP4/RCP2.6): 4.5 m are in the Somali Region, particularly around Denan migrants and Kebri Dehar, and farther to the northeast near • Inclusive development scenario (SSP2/RCP8.5): Aware. The water model inputs suggest modest 4.2 m migrants increases in water availability in this region of zero • Optimistic scenario (SSP2/RCP2.6): 3.6 m to two times the baseline, which is very low. The migrants ecosystem model inputs suggest that these regions will also see increases in net primary productivity. Even under the most optimistic climate scenario, there will be a significant in-migration to lowland areas. Excision of critical dry season grazing areas Pastoralists have been losing parts of their range- 71Based on a report prepared for the World Bank, 2019, Ethiopia Cli- lands to other users for years as population mate Change-Induced Migration Modeling, by Center for International pressure in the highlands pushes people down Earth Science Information Network (CIESIN), Columbia University, and the escarpment looking for land to crop, and gov- the CUNY Institute for Demographic Research (CIDR) with contribu- ernments look for making use of higher potential tions of Jacob Schewe and Fang Zhao of the Potsdam Institute for Cli- mate Impact Research (PIK), who provided ISI-MIP sectoral projections areas in the lowlands for commercial agricultural for crop production, water availability, and flood risk to 2050. plantations. The overall area of land lost to grazing 76   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Map 5.7: Hot spots of climate in-migration and out-migration for 2050 Figure 5.1: Lowland climate migrants, 2015–2050 (shaded areas represent the 95 percent confidence intervals) Pessimistic (reference) (RCP8.5/SSP4) More inclusive development (RCP8.5/SSP2) 10.0 10.0 7.5 7.5 5.0 5.0 Climate migrants (millions) 2.5 2.5 0 0 More climate-friendly (RCP2.6/SSP4) Optimistic (RCP2.6/SSP2) 10.0 10.0 7.5 7.5 5.0 5.0 2.5 2.5 0 0 2020 2030 2040 2050 2020 2030 2040 2050 Year Chapter 5: Changing Land Use and Land Cover   77 may be small, although in some regions still sig- Table 5.5: Large farms and concessions by region nificant, such as in Gambella where it is estimated Region   Area (ha)   % of lowland area some 15 percent of the land area has been taken up Afar 84,889 0.90% by agricultural concessions, many of which remain underutilized or unutilized, but the impact on the Amhara 16,998 0.42% wider rangeland ecosystem may be significant, Benishangul-Gumuz 254,448 5.50% given that many of these are the best watered Gambella 437,932 14.81% parts of the rangelands. As pointed out by Hatfield Oromia 24,595 0.17% and Davies (2006, 24) “Pastoralism is a system, in SNNPR 134,369 2.20% which small pockets of rich resources render vast Somali 12,542 0.04% tracts of low-grade rangeland productive.” Establishment of commercial farms in the Awash Tigray   44,060   1.80% Valley began in the 1960s, displacing pastoralists Total   1,009,833   1.36% from their traditional dry season grazing lands. Source: WB calculations using information on land concessions from Between 1950 and 1970 alone, 52,000 hectares Ethiopian Horticulture and Agriculture Investment Agency (EHAIA) plus of land were taken out of the prime grazing areas DevSeed. Area reflect areas identified as land concessions (from EHAIA data), DevSeed, or where these datasets overlap. and converted to agricultural lands by private and government firms (Harbeson, 1978, 481). After the military government came to power in 1974, it nationalized all commercial farms and managed Awash River in Afar. Details on commercial farming them as state farms. The size of the state farms along the Awash River show much of this is located considerably expanded under the military govern- within 5–10 km of the Awash River (Map 5.9). ment, increasing to 69,000 hectares in the late Developing the irrigation potential of the low- 1980s (Said, 1994). lands may be a national public good, but the Finding complete and up-to-date data on land tradeoffs in terms of increased vulnerability among concessions is difficult. Partial data on land con- lowland populations to shocks and longer term cessions were provided by the Ethiopian Horticul- impacts on the environment and rangeland pro- tural and Investment Agency (EHAIA) covering BG, ductivity need to be carefully assessed. Many con- Gambella, and SNNPR. According to EHAIA, these cessions remain dormant, while others have proven data represent most land concessions granted extremely costly to develop and the benefits exag- from 2015–2017. The data do not, however, appear gerated (Heady, Tafesse and You, 2013). to include earlier land concessions. For example, a previous analysis indicates that for the years Introduction of permanent and semipermanent 2004–2009 there were over 400 large land acquisi- water points tions covering nearly 1.2 M ha in Ethiopia (Deininger The potential negative effects of introducing per- et al., 2011). To supplement this analysis, spatial manent water points in wet season grazing areas data on large ‘commercial’ farms in Afar were gen- are well researched (Sandford, 1983). Lack of well- erated that involved processing of high-resolution planned water interventions in rangeland areas satellite imagery supplemented with information have caused extensive land degradation in parts of from open sources (DevSEED, 2018). When combin- Borana zone between Yavello and Arero and in parts ing these sources of information, total land conces- of Somali region. While it is easy to characterize sions appear to be over 1 M ha (Table 5.5). Gambella rangeland areas as lacking water, it is not water (438,000 ha) has the highest share of concessions per se but water in relation to grazing which is key. representing nearly 15 percent of total lowland area Introducing large dams or birkedas or boreholes into (Map 5.8), covering a strip running north-south normal wet season areas may open these areas through the region (Map 5.8). This also covers large to permanent occupation and overgrazing. This areas of BG (245,000 ha) and SNNPR (134,000 ha) has already happened in the Haud in the Ogaden representing 5.5 percent and 2.2 percent of their area of Somali region, where the extensive develop- respective lowland areas. In the eastern lowlands, ment of birkedas (large cement lined cisterns) over large commercial farms are primarily located along many decades, has led to permanent settlement rivers and rely on irrigation. In Somali, the majority and cropping in areas marginal to agriculture, and is located along the Shebelle River and along the the same may be currently happening in areas of 78   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Map 5.8: Land concessions and large farms 2016–2017 Map 5.9: Afar—large farm detail 2016–2017 Chapter 5: Changing Land Use and Land Cover   79 Map 5.10: Afar—Prosopis juliflora extent (left) and intensity (right) in 2016/17 Source: WLRC. Table 5.6: Prosopis expansion in Afar 2000–2016/17 % change 2000 2016/17 2000–2016/17 Prosopis High Prosopis High Prosopis High present infestation present infestation present infestation Area (ha) 828,411 41,529 1,173,300 306,409 42% 638% % of lowland area   8.8% 0.4%   12.4% 3.2%       Source: WLRC. humanitarian relief intervention where boreholes The expansion of Prosopis juliflora in Afar, for are drilled to support the settlement of IDPs. instance, has increased at a rapid rate. Presence of Prosopis juliflora has increased by 42 percent Bush encroachment between 2000 and 2016 and now covers nearly While the data show a positive greening in parts 1.2 M ha, while the portion of this characterized by of the lowland, particularly for the 2010–17 period, heavy infestation (where 80 percent or more of the this may be due to increased bush encroachment land is covered) is over 300,000 ha, an increase of as result of grassland degradation, overgrazing, over 600 percent from 2000 levels (see Map 5.10 and lack of rangeland management techniques and Table 5.6). designed to manage bush, such as annual “burn- Prosopis juliflora was first introduced into Ethi- offs,” which were banned under previous govern- opia in the late 1970s in Dire Dawa and was later ments. This appears to be the case in many parts planted over large areas including under the Food of Afar, Somali region, and Borana zone in Oromia. for Work Program from 1986 to 1988 for soil and 80   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Map 5.11: Prosopis juliflora coverage 2016/17 Source: WLRC. water conservation. One of the main invasive spe- with Prosopis juliflora present was nearly 1.2 M cies in Ethiopia, Prosopis juliflora has been declared ha, increasing by an additional 345,000 ha, or a a noxious weed and has been aggressively invading 42 percent increase, since 2000. pastoral areas, including in the Middle and Upper Nationwide, Prosopis poses a threat far and Awash Valley, and Eastern Harerge (Mwangi and beyond just Afar (see Map 5.11). There are signifi- Swallow, 2005). cant infestations of the bush in parts of SNNP and To better understand the potential threat posed in Somali region as well. to agricultural and pastoral systems by Prosopis Prosopis is not the only bush invading the range- juliflora, we look at the incidence and expansion of lands. Coppock reports extensive bush encroach- this invasive species in Afar.72 The results of this ment by Acacia drypanalobium in the southern analysis are summarized in Table 5.6 with the spa- rangelands in Borana zone (Coppock, 1994). As the tial extent and intensities shown in Map 5.10. The bush takes hold it is increasingly difficult to control, results in Map 5.10 show the coverage of Prosopis resulting in loss of access to large areas of grazing juliflora based on whether its presence is detected land. Pastoralists cope either by abandoning the (even at a small amount) and whether the grid cell area or by switching to herding camels and goats is nearly covered (i.e., high infestation with 80 per- that can cope much better with the bush. cent of area covered). In 2000 Prosopis juliflora was present in nearly 830,000 ha (8.8 percent of Conflict and displacement the lowland area). These results are higher than, Competition over water and grazing resources but consistent with earlier reports that as much between pastoral groups and between pastoral- as 700,000 ha of land in Afar was invaded in the ists and farmers in the lowlands is long standing. early 2000s (Tegegn, 2008). By 2016/17 the area This competition has been exacerbated in recent years by reduction in grazing areas as a result 72This analysis was undertaken by the Water and Land Resource Cen- of population increase, land excision, localized ter (WLRC), Addis Ababa University. enclosures, expansion of cropland, and permanent Chapter 5: Changing Land Use and Land Cover   81 settlements. The 1995 Constitutions’ recognition of northern Ethiopia (Kelemework, 2013). The major regional borders based on ethnicity has also hard- reasons for the recurrent conflicts between the Afar ened what were once more fluid natural resource and Issa Somali is competition over scarce pastoral boundaries between different groups. As a result resources that has been exacerbated by what Afar some groups have found they have lost access to considered the unfair absorption of some of their traditional grazing and wells across regional-ethnic traditional grazing areas into the Somali region. borders. Conflict is both a consequence and cause This conflict affected the customary system of of land pressure, as in some areas it has created rangeland management along the border between widespread human and livestock displacement, these two groups (Oba, 2009, 41). resulting in the abandonment of some areas, and Human Rights Report (2015, 2) indicated that increased pressure on other areas. “the drivers of the conflict in the South Omo zone Land use change in the Afar basin since the include economic marginalization of local commu- 1960s has contributed to deteriorating Afar rela- nities, loss of traditional grazing lands to large- tions with their neighbors. The conversion into scale, government-financed sugar plantations, and cropland of prime dry season grazing areas has restrictions on hunting.” restricted pastoralists’ customary access to riv- In the Borana pastoral area, even though low- erine resources. Therefore, “Many Afar clans have scale conflict has existed for many years between been forced to move their livestock further into Boran and Somali over rangeland use, it has inten- the highlands where they have come into conflict sified in recent years following the rezoning of the with crop cultivators” (Rahmato, 2007, 5). The Afar border between Oromia and Somali region after the most commonly come into conflict with the Issa 2003 referendum, which transferred a part of the Somali to the east, Karayyu Oromos to the south, Borana traditional grazing areas and deep wells to and the Wajirat highlanders in the Tigray Region of the south of Wachile to the Somali Region. These Map 5.12: Frequency of conflict events per woreda, 2015–2018 82   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands long-standing land claims culminated in the recent The overall effect of these conflicts has been to large-scale conflict in 2017/18 between the Borana create a humanitarian crisis in some areas with and Somali in the Moyale area. large numbers of IDPs crowded into camps and In 2015/16 ethnic violence escalated in many dependent on external food relief. In other areas, parts of Ethiopia, together with violent protests social services have been disrupted as teachers and against the EPRDF government (ICG, 2009). Much health staff have abandoned schools and clinics. of this violence was allegedly stoked by political Table 5.7 shows the number of IDPs by cause of actors but fed off of underlying and long-standing displacement. Conflict is the overwhelming reason grievances. Tensions between Somalis and Oromos given for displacement, with the majority of IDPs along their joint border exploded into open con- in Oromia (85 percent), Somali region (65 percent), flict in 2016–17. Maps 5.12, 5.13, and 5.14 show the and Gambella (85 percent) citing conflict as the incidence, fatalities, and displacement caused by main reason for displacement (see Table 5.8 for per- recent conflict outbreaks between 2015–2018 in centages by cause) Ethiopia according to data collected by ACLED and While the government is keen for IDPs to return UN-OCHA. The scale of the interethnic conflict has to their areas of origin, many are reluctant to do been large, resulting in the displacement of over a so as long as the underlying causes of the conflict million Ethiopians. While not confined to lowland remain. These conflicts may create long-term envi- areas, many of the conflict incidents are in lowland ronmental and economic damage in border areas, areas along the border between Somali and Oromia creating “no go” areas between ethnic groups and regions. sharpening already acute interethnic differences. Map 5.13: Number of deaths from conflict events per woreda, 2015–2018 Chapter 5: Changing Land Use and Land Cover   83 Map 5.14: Internal displacement in the lowlands as result of conflict and extreme climate events (drought and floods) 2016–2018 (WB based on data from IOM) Table 5.7: Total number of IDPs by cause and region Tigray Afar Oromia Somali Gambella Dire Dawa Total IDPs by cause Conflict 7,062 3,417 325,533 658,582 20,943 11,245 1,026,782 Climate 42,115 32,681 347,694 3,746 426,236 Other 5,087 22,591 27,678 Total IDPs by region 7,062 50,619 380,805 1,006,276 24,689 11,245 1,480,696 Table 5.8: Percent of IDPs by cause and region   Tigray Afar Oromia Somali Gambella Dire Dawa Conflict 100 6.75 85.49 65.45 84.83 100 Climate 83.2 8.58 34.55 15.17 Other 10.05 5.93   100 100 100 100 100 100 84   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Poverty and Vulnerability in the Ethiopian Lowlands Chapter 6: Regional Governance Summary Markakis73 termed the phrase ‘lowland periphery’ This chapter highlights some key political economic to describe the southernmost regions of Oromia risks in the lowlands. Vulnerability in the lowlands is and SNNPR, along with the so-called ‘Developing not only a function of low human capital and a highly Regional States (DRS)’ of Gambella, Benishangul-­ variable climate but also of the long history of political Gumuz, Afar, and Somali regions—effectively and economic marginalization of the area. Historically, highlighting their difference from the highland core the lowlands have existed on the margins of the Ethi- (Tigray, Amhara) and the highland periphery (Oro- opian state. The 1995 Constitution sought to redress mia, northern SNNPR). The lowland periphery refers the marginalization of these areas by conferring spe- to land and people spread across a large proportion cial status on the regions of Afar, Somali, Benishangul- of Ethiopia’s borderlands. Largely pastoralists and Gumuz, and Gambella. Despite this recognition, and the agropastoralists, to the east and north (Somali benefits it has brought, weak governance and account- Regional State and Afar) they are overwhelmingly ability has continued to limit the development of the Muslim and less diverse in terms of language and area. Opening the space to greater accountability and ethnicity than in the south and west (Borana in dialogue, combined with a deeper understanding of the Oromia, South Omo, Bench Maji in SNNPR, Gam- vulnerabilities of its inhabitants, will be key to address- bella, and Beneshangul-Gumuz). These lowland ing this marginalization. areas were formally incorporated into the Ethiopian state in the late nineteenth century after the con- Introduction quests of Emperor Menelik. During the Derg regime all land was national- Ethiopia’s lowland regions were formally incorporated ized. People could hold usufruct but not ownership into the Ethiopian state in the late nineteenth century. rights. Under the 1995 Ethiopian constitution, Three regimes—imperial, socialist, and federal—have land administration was devolved to the regional imposed their own cultural, political, and economic governments.74 The constitution introduced the governance arrangements on the lowlands (Marka- Ethiopian Peoples’ Revolutionary Democratic Front kis, 2013). Although federalism has led to greater (EPRDF) coalition of ethnically defined parties run- government access for groups previously excluded ning ethnically demarcated regional governments (Vaughan and Tromvoll, 2003), decision making has and operating in their own languages. Ethiopian remained largely centralized. The lowlands are heavily federalism encouraged political parties to organize dependent on the federal government for their pub- along ethnic lines and championed an ethnicized lic finances, including humanitarian assistance. In more ethnically diverse western Ethiopia, the regional states of Gambella and ­ Benishangul-Gumuz continue to be defined by contestation over ethnic ‘ownership’ 73 Markakis, Ethiopia: The Last Two Frontiers, 2013. of land and resources. 74 Ibid. 85 federal state.75 It has also allowed the federal gov- accountability in many regions. Ethnic parties affil- ernment to follow a path of political expediency iated to the federal government formally govern over time (Van Der Beken, 2017).76 the developing regional states, but these parties are not part of the ruling federal coalition. In some Regional Governance regions disagreements between regional officials Administrative decentralization has taken place and federal officials have developed over the level across the country with the architecture of the and nature of federal intervention in their respec- state now reaching the remotest areas. Each region tive regions. has its own elected regional government charged with overseeing a structure of zones, districts Lowland Contexts (woredas), and wards (kebeles). The process of devo- Lowland regions, including southern areas of lution took place in two phases. Initially in 1995 the SNNPR and Oromia, share several similarities as bulk of public service provisions were devolved to well as important differences. These similarities the subregional units. In 2002 substantial authority include: was devolved to the woreda administrations in the form of subsidiarity. The constitutional principle • Federal patronage defines federal to regional gov- of subsidiarity informs the division of responsibil- ernment relations. Although regional autonomy in ity between different levels of government; each Ethiopia is grounded in the federal constitution, regional unit provides public services at its level as it is also subject to the federal economic, social, well as implementing social and economic policies and development framework. In addition, the and law and order. The federal government main- limitations on political space, which have led tains all powers that were not delegated to the to a dominant one-party system at the center, regional level, as well as shared responsibilities. restrict opposition, debate, and participation at Public expenditure in the regions is based on fed- regional levels. Regional governments in the low- eral transfers combined with local revenue collec- land periphery are affiliated with the ruling party tion (see Chapter 7). The decentralization of public and subject to its centralized policy making. sector governance in Ethiopia has been extensive, • Customary governance. Various forms of custom­ and regional governments have the discretion to pri- ary governance systems endure across the oritize where to spend their funding allocations. Yet, lowlands of Ethiopia. In pastoralist areas these administrative and budgetary decentralization has customary systems operate alongside govern- been insufficient to substantially make up for the ment, and claims from people—for relief, justice, low capacity of the public sector in lowland regions. and services—are often made and heard within both systems. Practically, this requires engage- Politics of Regional Governance ment with two forms of governance for many of the people of the lowlands. With time and as Ethiopia’s model of local governance is interven- the reach of government has increased, hybrid tionist; it has been designed and implemented from forms have developed that unite systems and the center to effect change in the regions (Vaughan, bring together clan and state leadership, for 2017). In many areas this change has been rapid example, the Borana and Guji elect tribal leaders and profound,77 with interventions designed to to the kebele council or as cabinet members to transform socioeconomic, education, health, and intentionally bring together the clan and state.79 cultural norms. But, in the lowland periphery—the In some cases, hybridization has been used to impact of this system has been mixed.78 Top-down undermine the informal structures of governance decision-making has led to challenges around in favor of strengthening the state. • Long-standing, unresolved conflict dynamics. All the developing states have a legacy of separatist 75 Habte, Ethnic Federalism in Ethiopia: Background, Present Condi- movements that fought the state for the libera- tions and Future Prospects, 2003. 76 Van Der Beken, The Challenge of Reform within Ethiopia’s Constitu- tion of material and human resources. Added to tional Order, 2018. these (now mainly defunct, but not extinguished) 77 Pankhurst, (Ed), Change and Transformation in Twenty Rural Com- liberation struggles are multiple conflicts. Federal munities in Ethiopia, 2017. 78 Vaughan, Revolutionary democracy as an uncompromising political strategy: Revolutionary democratic state-building: party, state and people in the EPRDF’s Ethiopia, 2012. 79 Brocklesby et al., 2010, IDS working papers. 86   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands forces regularly intervene to manage these, but • Limited formal economic opportunities. Despite few have been conclusively resolved. At the root large investments in rural roads by regional gov- of many of these conflicts are political power ernments, many lowland areas remain difficult and representation, resource scarcity, and to reach, especially to the southeast. The very competition over land and water, as well as the remoteness of these areas creates difficulties preservation of ethnic or tribal identity. Conflicts for pastoralists who choose to or have no option operate between groups, within groups, between but to transition away from pastoralism. The perceived outsiders and indigenes, between high- absence of nearby urban centers severely limits landers and lowlanders, and between the state economic opportunities and has encouraged and the people (Feyissa, 2010). Conflicts have informal migration out of Ethiopia. Even in areas also been carried across the long and porous reasonably close to urban settings, research borders. Weapons are common. Herders have has shown a marked difference between jobs traditionally carried small arms, but proximity that young people are prepared to do in another to violent conflicts in the region has also made country compared with those they are willing to weapons readily available. It is not unusual for take close by in their own communities. In Sitti a localized incident to rapidly escalate into vio- zone, of Somali region, peer pressure appears to lence. And, mechanisms for either restorative or prevent young people from taking menial jobs retributive justice are often not available or are under the potential scrutiny of their peers and inadequate. contributes to migration to Djibouti to under- • Large and growing concentration of refugees. take similar work illegally (Gray, 2016).82 Illicit Benishangul-Gumuz, Gambella, and the Somali trade in contraband over the borders, including in region—three out of the four developing regions— Chat, livestock, and electronics, has also grown, are major refugee hosting areas. Given the prox- adding to the informal economy. The result has imity of international borders, and the fact that been economic disempowerment from the for- ethnic groups may straddle both sides of the mal Ethiopian economy and reliance on informal border, the distinction between host80 and ref- economic opportunities. Lowlanders, especially ugees is complicated. The definitions ‘host’ and those from subsistence backgrounds, have found ‘refugee’ are often fluid and can merge, with fam- it difficult to access capital and in general have ily groups adopting one or other identity based not been able to compete with highlanders in the on calculations of self-interest.81 The implication Ethiopian economy (see Feyissa, 2013). is that a formal encampment policy is often not enforced, with people moving in and out of camps While lowland pastoralist societies share similar and engaging in informal economic activity. This traits, there are also distinct variations across the situation has heightened tensions in some areas region. The pastoralist belt stretching through Afar, between groups along multiple lines (inter-clan, Somali region, and southern Oromia have more in intra-clan, highland-lowland, cross-border, and common given shared livelihoods and the relative state-clan), including driving greater resource lack of ethnic and religious diversity in these areas. competition in some areas. The influx of refu- Southern SNNPR, Gambella, and Benishangul-­ gees has also led to an increase in international Gumuz are considerably more ethnically diverse funding, which has contributed to growth in some and more varied in terms of livelihoods, customs local economies. Gambella town, for instance, has and traditions. As a result, their forms of gover- gone through a relative ‘boom’ period. nance vary, as do their customs, traditions, and • Limited implementation capacity. All lowland levels of integration into the dominant highland regions and zones suffer from limited capacity to Ethiopian political and economic system. implement projects. In Somali and Afar regions the federal government provides direct imple- Somali Regional State mentation support. This lack of capacity is a Somali Regional State (SRS) covers a large part major constraint to raising revenues and to the of Ethiopia’s low-lying eastern plateau, internally delivery of social services. bordering Afar and Oromia and sharing interna- tional borders with Djibouti, Somalia, and Kenya. 80 UNHCR defines ‘host’ as communities in a 20 km radius of refugee camps. 82Save the Children International Briefing Note on Migration for 81 DFID assessment of conflict risks in Gambella, 2017. Employment. David Gray, 2016. Chapter 6: Regional Governance   87 Its population is overwhelmingly Somali speaking, region. In 2016 the regional government increased a language shared with the peoples of its ­ adjacent the number of woredas from 68 to 102 and the international borderlands with who ­ Ethiopian number of zones from 9 to 11. Funding disbursed to Somalis share close clan ties. A 2007 census esti- these woredas was used to reward clans and allies mated the Somali regional population at over that had stayed loyal to the president.83 4.3 million; in 2018 it was projected to be approxi- Presidential power largely dictated how Ethio- mately 5.5 million, making Somalis the third largest pian Somalis in the region participated in the sys- ethnic group in Ethiopia. The population constitutes tem of governance. Daily processes of bargaining about 6 percent of the total Ethiopian population. to get things done, access to services or welfare, Despite historically low levels of development, or engagement in the regional economy were all over the last five years, the Somali region has seen determined by the degree of influence or marginal- considerable improvements. The pace of increase ity of a clan group to the president. Under Abdi Iley in regional expenditure over the past ten years there were a set of winners and losers, with some has been about 33 percent annually. International pastoralist groups consolidating and increasing development funding has also increased in the last their wealth and influence. Others, who were either few years. Yet, a succession of droughts has led to marginal or made marginal due to the president’s an increase in humanitarian need. There is a large network saw their vulnerability increase. Political-­ internally displaced population. Dependence on economic factors were important in directing government welfare has increased. Ethnic conflict humanitarian relief, both between different areas of has flared, and the regional president was recently the region and in the overall volume of assistance removed by the federal government. the region received, which in 2016 and 2017 reached The political settlement in this region has been over 50 percent of the population (see Chapter 8). closely tied to the national political context. Somali region was unique in the lowlands in forging a rel- Afar national regional state atively autonomous relationship with the state Afar covers the northeastern part of Ethiopia. The that transcended the more normal expectations northern part of the region is largely desert scrub- of federal-­regional relations. Here, the political land with shallow, salty lakes and a long chain of settlement was based on a combination of heavily volcanoes. The Awash River valley forms the south- securitized governance, supported by the Ethio- ern part. The region is characterized by an arid pian National Defense Forces (ENDF), and a highly and semiarid climate with low and erratic rainfall personalized, president-led style of regional gov- and has frequently been affected by drought. The ernance. Abdi “Iley” Mohamoud Omar, deposed as majority of Afar people live in the region, but there president in 2018, was extremely effective at seiz- are significant minorities in neighboring Eritrea and ing opportunities to consolidate his powers in the Djibouti. Afar is the principle language of the region region, first as head of regional security and then as and Islam (Sunni) the principle religion. the regional president. He presided over a regional The region has 32 woredas that receive block executive that was unusually resilient and able to grants to provide services to their respective com- wield an unprecedented level of autonomy from munities. The population of Afar constitutes about the center. This resilience was partially due to the 2 percent of the country’s population; Afar’s share president’s ties with the ENDF, earned through the of block grant transfers from the federal govern- impact of the long-standing campaign against the ment has remained around 3 percent over the past Ogaden National Liberation Front (ONLF). The liber- decade. As in Somali region implementation capac- ation struggle led the federal government to replace ity in the region is very weak, necessitating reliance federal security units in Somali region with the Liyu on federal institutions. Economic sectors (including (special) police in 2007. Recruited from local Somali agriculture and water) and social sectors received communities, the Liyu police maintained close ties a large proportion of total regional expenditures, with the ENDF despite nominally falling under the though in comparison with other regions invest- authority of the regional president. ment in economic sectors is higher than in social Having gained control of the insurgency in the services. In keeping with the federalist system, Afar region, Abdi Iley was able to change the usual patron-client relationship between the regional and the federal government. He also demonstrated 83 Conciliation Resources Political Unsettlement and the Somali equally adept handling of the clan politics within his Regional State of Ethiopia, 2017. 88   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands is formally governed by the Afar National Demo­ leaving little opportunities for the indigenous cratic Party (ANDP), an affiliate of the EPRDF. population. Like in Benishangul-Gumuz the wider Customary governance systems also coexist with economic developments seem to favor the high- formal governance in the region. Decision-making land population; contractors are mostly recruited power regarding land and other natural resources from the non-titular population, and the secondary rests with the clan council, which consists of clan industry, restaurants and hotels, are also staffed leaders, elders, the feima,84 and local wise men. by nonindigenous groups. Resource users are divided into two categories, Patterns of segregation in terms of occupation primary right holders, those who have the right to and living spaces can be observed between the two use the resources on the rangelands but also to groups in the emerging regions. The titular groups exclude others and transfer rights to their heirs, and are often mostly active in the political sphere in secondary right holders. This second tier consists of the administration, but otherwise pursue tradi- groups of neighboring pastoralists whose demands tional rural activities. Investment and businesses for pastoral resources go beyond their own endow- are largely run by the non-titular groups. Shops, ments, particularly during drought years.85 The businesses, and major market activities are often interests of pastoralists in Afar and the federal in the hands of nonindigenous groups. Since the government have often been in opposition. Major banking sector is also often staffed with foreigners flash points have been the use of land and provision (read non-titular) who keep alliance with their fellow of services, particularly in the riverine areas and migrants, the indigenous often find accumulation of in connection with the development of the sugar credit and capital to start their businesses difficult. industry. Competition is uneven as highlanders have better Along the regional border with Somali region access to credit and are often attached to wealth- there have been frequent clashes between Afar and ier communities in the highlands. Issa Somali pastoralists. Many Afar lay historical With a tradition of north-south migration and claim to grazing lands which are now incorporated large-scale resettlement of highland populations, within Somali regional state. the regions have become the home of migrant workers, investors, and other business people. The Gambella and Benishangul-Gumuz regional states changes to livelihoods as a result of villagization, The Benishangul-Gumuz and Gambella regions have and the removal of forest land for hunting and a history of marginality that has been reinforced by honey farming, have affected local livelihoods. recent politics. For many titular groups in the west- Casual labor on investment farms has attracted ern lowlands, the perception that the centralized more settlers and migrants to the regions, and not state expropriates indigenous land in a quest for necessarily benefited the local population. Low lev- economic growth is fairly widespread today. els of human capital among local populations have Due to their location at the border with the also contributed to limited access to the benefits Sudan and South Sudan, the two regions have long of growth and increased risk of marginalization in been exposed to refugees fleeing the fighting across political processes. In addition, unlike in the drought- the border. Gambella, for example, has provided prone lowlands, there is no safety net program in shelter for Sudanese refugees since the 1960s. At the form of PSNP to support the chronically poor. present, the number of Sudanese refugees almost The current situation in Benishangul-Gumuz equals the total population of Gambella. The pres- and Gambella is volatile, with tensions prevailing ence of such a large number of refugees, many of between the Nuer and Anywaa. In summary, the whom are Nuer, is a source of growing concern to western lowlands are prone to multiple factors that Anywaa who fear being outnumbered in their own impede their economic development. At the core is territory (Ohta and Gebre Yintiso 2005). Jobs are a contestation between local, regional, and federal largely being taken by experts from the highland, actors over the abundant natural resources of the area, and a fear among indigenous groups of their further marginalization. 84 Feima is a rule-enforcing authority in Afar traditional administration. It consists of a principal leader (feima-abba), a deputy leader (erenna- Southern SNNPR and Oromia abba), and ordinary members. Both Oromia and SNNPR have southern areas 85 Cloudburst Group, USAID, 2017. Land use rights, land governance institutions and tenure security indicators in pastoral community: evi- that share similarities with the developing regions. dence from a baseline study in the Afar region, Ethiopia. In both regions these borderlands are home to Chapter 6: Regional Governance   89 nomadic herders who belong to an array of dif- pastoral zones in Oromia over the past seven years ferent ethnic groups. Like the developing regions, was around 30 percent of the total block grants they are defined by their proximity to international allocation. borders, as well as their relationship to those peo- Much of the recent tension between regional ples who live across the border. In Oromia the main states in the lowlands has occurred between Somali pastoralist area is Borena and Guji zones. In SNNPR and Oromia. In Borana zone, Boran pastoralists lay a variety of peoples inhabit the southern Omo val- historical claim to wells and grazing within Somali ley and Bench Maji areas with distinct customs and region. This has frequently resulted in clashes traditions. Given the level of ethnic fragmentation between Somali and Boran pastoralists in the area. along the Omo with a variety of small agropasto- ralist groups, these groups have lacked political Conclusions voice when it has come to protecting their natural Ethiopia’s federal system has enabled the federal resources from state encroachment. As result the government to move forward from a legacy of war, land on which they live has often been categorized famine, and poverty. Interventionist policies have in as unproductive or underutilized by government many cases brought rapid, far reaching, and pos- officials seeking to make way for large sugar plan- itive social and economic changes to Ethiopians. tations or other large-scale investments in the Yet, uniting the country under common economic Lower Omo River valley. policies and a centrist, one-party system has come SNNP region is composed of 14 zonal adminis- at a cost. The difficulties that have developed trations and about 135 woredas including 5 special in recent times at the national political level are woredas, which are similar to autonomous areas. reflected at regional levels. Ethnic federalism has The zonal administrations in SNNP have greater brought greater autonomy and self-governance, administrative authority compared to similar and ultimately has delivered a higher level of ser- administrations in other regions, which tend to vices to all regions. But state penetration in the carry out a coordination role. Pastoralists are lowlands has remained top-down and is perceived grouped into three areas: South Omo, Bench Maji, by many lowlanders to be driven by the priorities and Kaffa, which account for about 14 percent of and needs of the highland majority. the regional population. Like the practice in the What is perhaps more telling is the fragility other regions, the SNNP regional government allo- of the gains that have been made to date. There cates budgetary resources to the zonal administra- appear to be huge pent-up demands for approved tions and the woredas under these zones in the form accountability and voice among lowland commu- of block grants. Over the past five years, allocation nities who feel that their economic interests have to the pastoral zones in SNNP accounted for about often been overlooked. Ultimately, governance in 15 percent of the total block grants allocated by the the lowland regions has not always protected the region to zonal administrations and woredas. The interests of lowland peoples. State development share of administrative and general services expen- strategies, including villagization and land appro- ditures in the pastoral zones was higher than the priation which turns pasture or cropping land into regional average. However, allocation to agriculture commercial farms, have not always produced their and water resource development followed a largely intended results, and have often come at a high similar trend with the regional average. cost. In many areas these approaches have further Oromia region is the largest regional state in marginalized lowland peoples, while also increasing Ethiopia, both in terms of land and population. It tensions between different ethnic groups, as well as is divided into 20 administrative zones, 2 of which between highlanders and lowlanders. have been added over the last two years. Of the As a follow-up to the diagnostic analysis carried 20 zones, 5 are predominantly pastoral and/or out so far, the next part of the report focuses on agropastoral: Borena, Bale, Guji, East Hararghe, the role of public policy through the intergovern- and West Hararghe. The combined population in mental fiscal transfer system, and the coordination these zones accounts for about 30 percent of the of the safety net system working with humanitar- total population in the region. Similarly, the propor- ian interventions, in supporting the promotion of tion of the block grants resources allocated to the resilience and development in the lowlands. 90   Part 1: Livelihoods, Poverty, and Vulnerability in the Lowlands Part 2: Promoting Resilience and Development in the Lowlands Poverty and Vulnerability in the Ethiopian Lowlands Chapter 7: The Role of Public Finances Summary fiscal transfers from the federal government, while This chapter explores the role of intergovernmental the second one is own revenue collection by the fiscal transfers in supporting the lowlands. The lowland regions. regions have less capacity to raise their own revenues Own revenue collection and depend more on block grants from the center than the other regions. In terms of intergovernmental fiscal The revenue assignments between the federal gov- transfers from the federal government, the lowlands ernment and the regional governments in Ethiopia’s have done well compared to highland regions. But sig- federal structure are stipulated in Ethiopia’s Con- nificant regional differences exist in terms of expendi- stitution. The most lucrative sources of revenue, ture priorities. In general, the Somali and Afar regions including customs taxes and duties, domestic indi- have spent per capita much less on human devel- rect taxes, and business profit taxes on medium- opment (education and health) than Gambella and and large-scale businesses, are assigned to the Benishangul-Gumuz. None of the SDG capital grants in federal government. On the other hand, regional Somali and Afar regions, for example, have been spent governments can collect taxes and fees from on education and health. employment income, agriculture and land use, and In the future, consideration should be given to more profits of small businesses. The Constitution also closely aligning the provision of block grants to need stipulates revenue sharing mechanisms for some and the costs of service delivery, rather than popula- of the revenue sources between the federal govern- tion size, and earmarking funds for human develop- ment and the regional governments. ment as a key national priority. As the federal government controls the bulk of the revenue sources, the regional governments Introduction largely rely on transfers from the federal govern- ment to carry out their expenditure assignments. In terms of intergovernmental fiscal transfers The extent to which own revenue collection contrib- from the federal government, they have done well utes to total budget shows significant variations and compare favorably with highland regions. This across the regions. Between FY12–18 the average section analyzes trends in public expenditures in contribution of own revenue collection to total the lowland regions. The analysis looks at both the budget in all regions (excluding Addis Ababa) was sources and uses of budgetary resources in lowland around 25 percent, implying that the regions relied regions and zones in comparison with the highlands. on the federal government to cover about 75 per- cent of their budgets (see Figure 7.1). The Dire Dawa Sources of Funds City Administration had the largest share of own For the regional governments in Ethiopia’s federal revenue to total budget during the period, while the system, there are essentially two major sources of other predominantly urban regional state, Harari, budgetary resources. The first important source is also had one of the largest shares at 30 percent. 92 Figure 7.1: Percent of own revenue to total budget, FY12–18 50 45 43 40 39 35 30 30 24 25 24 25 25 22 20 17 18 15 15 10 5 0 ay r ra ia i uz NP la ri a ns al a w ra om l Af ha m be um gr io Da SN Ha eg So Ti Am m Or -G re lr Ga l Di Al gu an sh ni Be Source: MoFEC. This shows the significant association between the Revenue collection in per capita terms varied level of urbanization and the amount of revenue significantly between the regions (Figure 7.2). Addis that can be mobilized by the regions. Tigray is an Ababa is an outlier with a per capita own revenue exception in this regard as the regional government collection which is higher than the regional aver- managed to mobilize revenue which was much age by more than ten-fold. The two predominantly larger than the regional average. urban regions (Dire Dawa and Harari) also had per In the lowland regions of Afar, Somali, capita revenue collections that are significantly Benishangul-­ Gumuz, and Gambella, the con- higher than the other regions. The larger regions tribution of own revenue to total regional bud- including Oromia, Amhara, and SNNP (which are get during the period was much lower than the considered to have much better public administra- regional average. Afar had the least contribution tion and service delivery capacity) had per capita from own revenue sources of all the regions, while revenue collections that were significantly lower Benishangul-­ Gumuz had a share close to the than the regional average. The dominance of small- regional average and higher than some of the larger scale agricultural activities (which have relatively regions (e.g., SNNP). low revenue raising potential) could be one possible Figure 7.2: Per capita own revenue collection in ETB, FY12–18 6,500 6,419 6,000 5,500 5,000 4,500 4,000 3,500 3,000 2,500 2,000 1,558 1,530 1,500 1,000 681 643 543 528 500 271 287 258 261 260 0 y ar ra ia i uz NP lla ri a a ns al ra ab w ra om Af ha m be um io Da SN g Ha Ab eg So Ti Am m Or -G re lr Ga s di ul Di Al g Ad an sh ni Be Source: MoFEC and CSA. Chapter 7: The Role of Public Finances   93 Figure 7.3: Federal block grant allocation to regional governments 140 50 45 120 40 100 35 30 Billion ETB 80 25 60 20 40 15 10 20 5 0 0 FY12 FY13 FY14 FY15 FY16 FY17 FY18 In billion ETB As % of federal government budget Source: MoFEC. explanation for this very low level of revenue col- regional differences in the capacity to collect reve- lection in these larger regions. Again, Tigray is an nue, were the major flaws of this approach. In later exception to this as well, as it had a per capita reve- formulae, the focus shifted more toward attempt- nue collection that was significantly higher than the ing to fill the fiscal gaps of regions, which were regional average. Afar and Somali had a per capita estimated using representative tax and expenditure revenue collection significantly below the regional data. At least three iterations of the formula have average, while in Gambella and Benishangul-Gumuz so far been implemented using this approach. The it was slightly higher than the regional average. major drawback encountered is the challenge in the availability and quality of data to reliably measure Fiscal transfers from federal government expenditure needs and revenue raising capacities of a) Block grant transfers regions. Regardless of the approaches used, how- Under Ethiopia’s federal system, allocation of grants ever, the final output of all the formulae has been by the federal government to regional governments to decide on the share of each region from the total is a constitutional requirement. Accordingly, the fed- pool86 allocated to be transferred to the regions. eral government has been transferring resources to Block grant allocation from the federal govern- regional governments since 1995, based on annually ment to regional governments has shown sizable approved allocations. The transfers are provided to increases in nominal terms in recent years. During regional governments as “block grants,” implying that the FY 2013–17 period, the federal government the regions have discretion on the sectoral allocation increased its block grant allocation to regions at of the resource. The distribution of the grants among an average rate of about 26 percent annually. The the regions is governed by a block grant allocation share of block grant transfers from total federal formula. The House of Peoples Representatives government expenditures picked up significantly approves the formula behind the block grant allo- over the past two years, increasing from about cations. Once the formula is approved, it remains in 40 percent to almost 46 percent (Figure 7.3). effect for about five fiscal years before being revised. Block grant transfers to regions have also shown The predominant objective of earlier transfer sizable increase in per capita terms, increasing formulae prior to the mid-2000s was equaliza- annually by about 24 percent over that past five tion of fiscal needs. These formulae attempted to fiscal years. Of the lowland regions, Gambella compute weighted indices of expenditure needs for received the highest block grant allocation in per the regions, which would then serve as a basis for capita terms, followed by Benishangul-Gumuz, the distribution of grants among the regions. The subjectivity in the selection of indicators/variables 86 The determination of this pool is entirely at the discretion of the of fiscal needs and the weights assigned to the federal government and is approved on an annual basis by the Council of Minsters as part of the Medium-term Macroeconomic and Fiscal variables, as well as the fact that it did not consider Framework, which is a rolling five-year framework. 94   Part 2: Promoting Resilience and Development in the Lowlands Figure 7.4: Per capita block grant transfers in ETB, FY12–18 3,500 3,051 3,000 2,725 2,500 2,000 1,846 1,562 1,500 1,396 1,187 1,052 1,000 866 831 882 748 500 0 ay r ra ia i uz NP la ri a ns al a w ra om l Af ha m be um gr io Da SN Ha eg So Ti Am m Or -G re lr Ga l Di Al gu an sh ni Be Source: MoFEC and CSA. Afar, and Somali (Figure 7.4). In general, per capita Table 7.1: Per capita transfers to lowland (pastoral) zones in block grant allocations to the lowland regions were Oromia and SNNP in ETB, FY11–16 above the regional average. Per capita block grant Oromia allocations to the larger regions, including Oromia, Pastoral zones   Amhara, and SNNP, were less than the regional Bale zone 789.5 average, with the exception of Tigray which was higher than the average. Guji zone 587.6 Looking at per capita transfers to pastoral zones Borena zone 630.3 in regions that are partially lowland (i.e., Oromia East Hararge zone 544.4 and SNNP), average per capita transfers to the West Hararge zone 585.4 pastoral zones in Oromia are higher than the other Pastoral zones average 627.4 zones, as well as the regional average, they are Non-pastoral zones 608.7 while lower in the case of SNNP (Table 7.1). Regional average 614.2 Over the past 10 years, the share of regions from total block grant allocation showed no significant SNNP  variations despite the application of at least three Pastoral zones   different allocation formulae (Table 7.2). How- South Omo zone 742.4 ever, the current formula, which was introduced in Kefa zone 614.5 FY18, has resulted in some changes to the regional Bench Maji zone 571.3 shares. In the lowland regions, Somali experienced a Pastoral zones average 642.7 significant increase in its share (about 2 percentage Non-pastoral zones 689.7 points), Benishangul-Gumuz and Gambella experi- enced declines, and the share of Afar remained the Regional average 681.4 same. Of the regions that are partially lowlands, Source: MoFEC. Oromia’s share got a relatively significant boost, while that of SNNP was largely unchanged. early 1990s. The stated aim of the MDG Transfers b) MDG/SDG Transfers is to help regions achieve the MDGs by support- In FY12, the federal government introduced a new ing them in areas where they lag. Accordingly, the fiscal transfer instrument for regional govern- sectoral allocation of the MDG Fund was decided ments, which it called Support to the Millennium through a consultative process involving all regions. Development Goals; this was in addition to the When the MDGs were replaced by the Sustainable Federal Block Grant which has been in place since Development Goals (SDGs) in 2015, the transfers the beginning of the decentralization process in the were renamed as Support to the SDGs. Chapter 7: The Role of Public Finances   95 Table 7.2: Regional share of federal block grant allocation FY07–FY09 FY10–FY12 FY13–FY17 FY18 Tigray 6.7% 6.9% 7.1% 6.0% Afar 3.3% 3.3% 3.1% 3.0% Amhara 24.3% 23.3% 23.4% 21.6% Oromia 33.7% 32.6% 32.5% 34.5% Somali 6.6% 8.4% 8.0% 10.0% Benishangul-Gumuz 2.0% 2.0% 2.1% 1.8% SNNP 19.5% 20.0% 20.0% 20.1% Gambella 1.4% 1.6% 1.5% 1.3% Harari 1.0% 0.9% 1.0% 0.7% Dire Dawa 1.4% 1.0% 1.2% 0.9% Total 100.0% 100.0% 100.0% 100.0% Source: MoFEC. The MDG/SDG Transfers are used only for cap- shares while Benishangul-Gumuz and Gambella ital projects related to the five pro-poor sectors saw their shares reduced. The shares of SNNP and (i.e., agriculture and rural development, water, Afar were largely unchanged. rural roads, education, and health). The resources The sectoral allocation of the MDG/SDG Trans- transferred through this channel are managed fers are not uniform in all regions. In the larger at the regional level, with concerned regional regions including Oromia, Amhara, and SNNP, the bureaus responsible for implementing the projects/­ rural roads sector was allocated the bulk of the investments, though woredas also play a crucial role transfers, while in the pastoral regions of Afar and (especially in the monitoring and evaluation pro- Somali, the entire MDG Transfers were allocated to cess). The MDG Transfers are performance-based the water sector (Table 7.4). Of the smaller lowland grants whereby disbursements are made based on regions, Benishangul-Gumuz region also allocated progress assessments on the utilization of previ- the bulk of the transfers to rural roads, while the ously transferred funds. The allocation is largely allocation in Gambella was more evenly distributed based on the same formula used to allocate the between the five sectors. Federal Block Grants (Table 7.3). Oromia and Somali The allocation by the federal government to regions received a relatively large boost in their the MDG/SDG Transfers, which reached nearly Table 7.3: Regional share of MDG/SDG Transfers FY12 FY13 FY14 FY15 FY16 FY17 FY18 Tigray 7.0% 7.2% 7.2% 7.2% 7.2% 7.2% 6.0% Afar 3.3% 3.2% 3.2% 3.1% 3.1% 3.1% 3.0% Amhara 23.3% 23.2% 23.2% 23.2% 23.2% 23.2% 21.6% Oromia 32.5% 32.5% 32.5% 32.5% 32.5% 32.5% 34.5% Somali 8.4% 8.1% 8.1% 8.1% 8.1% 8.1% 10.0% Benishangul-Gumuz 2.0% 2.1% 2.1% 2.2% 2.2% 2.2% 1.8% SNNP 19.9% 20.1% 20.1% 20.1% 20.1% 20.1% 20.1% Gambella 1.6% 1.5% 1.5% 1.5% 1.5% 1.5% 1.3% Harari 0.9% 1.0% 1.0% 1.0% 1.0% 1.0% 0.8% Dire Dawa 1.0% 1.2% 1.2% 1.2% 1.2% 1.2% 0.9% Total 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% Source: MoFEC. 96   Part 2: Promoting Resilience and Development in the Lowlands Table 7.4: Sectoral allocation of the MDG Transfers in FY12 by region Agriculture Water Rural roads Education Health Tigray 3.9% 63.5% 28.4% 4.0% 0.2% Afar — 100.0% — — — Amhara 5.7% 14.1% 57.8% 10.3% 12.2% Oromia 2.8% 8.9% 68.0% 11.3% 9.0% Somali — 100.0% — — — Benishangul-Gumuz 14.3% 10.5% 51.0% 17.9% 6.2% SNNP 18.4% 7.0% 57.0% 8.0% 9.5% Gambella 11.2% 29.3% 21.1% 19.9% 18.5% Harari 1.9% 23.9% 41.6% 27.2% 5.4% Dire Dawa 11.8% 32.1% 12.8% 22.3% 20.9% Total 6.8% 25.0% 50.8% 9.1% 8.4% Source: MoFEC. Figure 7.5: MDG Transfers allocation and disbursements, million ETB 21,000 19,000 17,000 15,000 13,000 11,000 9,000 7,000 5,000 FY12 FY13 FY14 FY15 FY16 FY17 FY18 Allocated Disbursed Source: MoFEC. 20 billion birr in the initial years, has been reduced rates in the early years but significantly improved significantly in recent years (Figure 7.5). However, in the later years. In the last couple of years, how- the disbursement rate of the transfers to the ever, there was a relatively large gap between the regions, which was relatively low during the first allocation and disbursement of the MDG/SDG couple of years after the introduction of the trans- Transfers, especially in the larger regions. This gap fers, has improved over the years, with entire allo- was less due to a deterioration in the utilization cations disbursed in all regions in subsequent fiscal capacity of the regions than to the diversion of years (Table 7.5). The disbursement rate during the some of the funds to other priority areas, including initial years varied among the regions. The lowland the building of regional agroindustry parks, particu- regions generally had below average utilization larly in the larger regions. Chapter 7: The Role of Public Finances   97 Table 7.5: Rate of disbursement for the MDG/SDG Transfers   FY12 FY13 FY14 FY15 FY16 FY17 FY18 Tigray 94.3% 86.6% 100.0% 99.6% 99.4% 60.0% 80.0% Afar 80.6% 85.0% 100.0% 100.0% 99.9% 87.7% 100.0% Amhara 89.1% 87.1% 100.0% 92.9% 100.0% 60.0% 80.0% Oromia 93.4% 87.1% 97.9% 93.1% 97.6% 60.0% 80.0% Somali 58.4% 88.5% 100.0% 100.0% 100.0% 87.3% 100.0% Benishangul-Gumuz 95.3% 80.9% 96.8% 81.0% 97.7% 60.0% 100.0% SNNP 76.3% 87.2% 95.8% 93.1% 100.0% 60.0% 80.0% Gambella 79.3% 52.2% 98.3% 99.9% 100.0% 60.0% 100.0% Harari 68.0% 88.3% 99.8% 100.0% 99.7% 60.0% 100.0% Dire Dawa 79.5% 85.6% 94.3% 67.3% 76.6% 60.0% 100.0% All regions 85.1% 86.5% 98.3% 93.9% 98.8% 63.1% 83.6% Source: MoFEC. Expenditures Following the upward trends in the regions’ own Overview of regional expenditures revenue collection and fiscal transfers from the federal government, regional spending has been Regional governments combine the transfers they increasing in recent years (Figure 7.6). In nominal receive from the federal government with their own terms, total expenditures by the regions increased revenue collection and allocate to expenditures by about 22 percent annually over FY13–18. When that fall under their constitutional mandate. At the adjusting for inflation, the increase was much more regional level, the sectoral allocation of the regional modest at around 11 percent over the period. There budget is approved by the Regional Council. The was also a strong increase of about 19 percent in Regional Council also approves the block grant regional per capita expenditures. allocation to the woreda administrations within the Per capita expenditures across the regions during region. Woreda Councils in each woreda then decide the FY12–18 period show a largely similar pattern on the sectoral allocation of the block grants they as per capita revenue collection. Addis Ababa is an receive from the regional level. The regional expen- outlier with per capita expenditures of more than ditures analysis presented in this section includes four times the regional average (Figure 7.7). The spending at both the regional as well as woreda predominantly urban regions of Harari and Dire levels. Dawa also had per capita expenditures that were Figure 7.6: Trends in regional expenditures 200 2,000 180 1,800 160 1,600 140 1,400 120 1,200 Billion ETB ETB 100 1,000 80 800 60 600 40 400 20 200 0 0 FY12 FY13 FY14 FY15 FY16 FY17 FY18 Nominal Real Per capita (RHA) Source: MoFEC and CSA. 98   Part 2: Promoting Resilience and Development in the Lowlands Figure 7.7: Per capita regional expenditures in ETB, FY12–18 7,000 5,869 6,000 5,000 3,877 4,000 3,157 3,167 3,000 2,030 2,000 1,707 1,546 1,382 1,309 1,111 953 1,055 1,000 0 ay r ra ia i uz NP la ri a a ns al a ab w ra om l Af ha m be um gr io Da SN Ha Ab eg So Ti Am m Or -G re lr Ga is l Di Al d gu Ad an sh ni Be Source: MoFEC and CSA. significantly higher than the regional average. All Recurrent spending absorbed about 60 percent the lowland regions had per capita expenditures of total regional expenditures during the FY12–18 above the regional average, but to varying degrees. period. This share has, however, been increasing Gambella had the highest per capita expenditures, in the last two years as the amount of MDG/SDG which were more than double the regional average, Transfers from the federal government was reduced followed by Benishangul-Gumuz. The predomi- (see Figure 7.8). The predominantly urban regions, nantly pastoral lowland regions of Afar and Somali Addis Ababa, Harari, and Dire Dawa, allocate a had significantly lower per capita expenditures than relatively larger portion of their budgets to capital Gambella and Benishangul-Gumuz despite being spending compared to the other regions (Table 7.6). above the regional average. Somali had the least In the lowland regions, Somali region had the high- per capita expenditures, which was very close to est allocation to capital expenditures while Gam- the regional average. The predominantly highland bella had the lowest, which is also the least of all regions, except for Tigray, had per capita expendi- the regions. The predominantly pastoral regions of tures well below the regional average. Somali and Afar had a better allocation to capital Figure 7.8: Composition of regional expenditures, FY12–FY18 100 90 80 70 60 50 40 30 20 10 0 FY12 FY13 FY14 FY15 FY16 FY17 FY18 Recurrent Capital Source: MoFEC. Chapter 7: The Role of Public Finances   99 Table 7.6: Share of capital spending to total regional expenditures, in percent   FY12 FY13 FY14 FY15 FY16 FY17 FY18 FY12–18 Tigray 44.5 47.1 47.3 40.5 40.1 34.6 30.5 40.7 Afar 50.9 55.0 49.6 46.6 43.6 40.8 19.2 43.7 Amhara 35.0 43.8 41.2 34.2 31.3 27.0 23.5 33.7 Oromia 33.5 42.6 40.0 34.0 30.9 26.3 18.6 32.3 Somali 48.6 56.1 46.2 54.5 57.2 58.2 58.4 54.2 Benishangul-Gumuz 28.7 45.6 35.1 36.2 32.1 28.6 20.7 32.4 SNNP 40.7 41.9 41.9 37.0 32.4 27.4 26.5 35.4 Gambella 39.9 23.3 27.2 19.5 28.0 14.1 13.3 23.6 Harari 53.8 57.1 61.1 55.8 52.7 57.1 45.4 54.7 Addis Ababa 62.0 62.3 64.1 59.2 52.0 52.8 48.4 57.3 Dire Dawa 47.6 44.8 45.2 44.2 46.3 47.3 37.9 44.8 Total 41.5 47.4 46.2 41.1 37.6 34.3 29.8 39.7 Source: MoFEC. expenditures compared to the moisture reliable regions. In the four predominantly highland regions lowland regions of Benishangul-Gumuz and Gam- (i.e., Oromia, Amhara, SNNP, and Tigray), the edu- bella. From the predominantly highland regions, cation sector was allocated the largest share of Tigray had the largest allocation to capital expendi- regional expenditures followed by administration tures while Oromia had the lowest. and general services. Other important sectors Spending on education absorbed the largest include health, roads, agriculture, and water. Mean- chunk of regional expenditures, taking up almost while, the lowland regions allocated the largest 25 percent of total regional spending during share of their budgets to expenditures on admin- FY12–18 (Table 7.7). Administration and general ser- ­ istration and general services. In Afar and Somali, vices, roads, and health were also among the sec- agriculture was the second most important sector, tors with relatively large allocations from regional while in Gambella and Benishangul-Gumuz it was budgets. The sectoral disaggregation of regional education. Spending on the water sector absorbed expenditures exhibits some variations across the the third largest allocation in Somali region while in Table 7.7: Sectoral disaggregation of regional expenditures during FY12–18, percent of total expenditures Administration Agriculture & general & rural General   services development Water Roads Education Health development Others Tigray 17.1 7.9 11.2 2.5 25.0 8.3 6.6 21.3 Afar 24.2 17.0 13.9 2.2 16.0 10.8 4.8 11.1 Amhara 16.8 7.5 7.0 9.0 31.6 11.5 6.4 10.2 Oromia 19.4 7.9 6.9 15.0 27.6 12.3 2.1 8.8 Somali 26.5 15.5 14.8 3.7 13.9 8.0 8.9 8.7 Benishangul-Gumuz 25.2 11.9 3.3 7.8 23.7 13.5 2.7 11.8 SNNP 18.6 10.5 3.3 2.7 26.4 11.0 8.4 19.0 Gambella 31.0 13.7 4.1 3.6 21.0 10.6 6.9 9.1 Harari 16.0 2.4 5.6 15.9 16.6 9.0 8.3 26.3 Addis Ababa 12.4 0.4 6.2 23.4 13.1 6.7 10.3 27.5 Dire Dawa 17.5 3.5 2.4 1.8 20.2 13.0 22.2 19.5 All regions 18.2 7.8 7.1 10.9 24.2 10.4 6.5 15.1 Source: MoFEC. 100   Part 2: Promoting Resilience and Development in the Lowlands Afar it was the fourth largest following education. highest utilization of capital budget, while Amhara Health and agriculture also absorbed a relatively had the lowest utilization rate. On the other hand, significant portion of regional expenditures in Gam- Somali had the highest capital budget utilization bella and Benishangul-Gumuz. rate of all the lowland regions, while Gambella had Regional expenditure outturn against budget the lowest. Afar and Benishangul-Gumuz had a over the past five years was much better for recur- largely similar capital budget utilization rate. rent spending than for capital expenditures. As there were supplementary budgets for FY15 and Expenditures in the lowland regions FY17, the budget outturn for recurrent expenditures was above 100 percent for almost all the regions Afar region during these two years. There was relatively lower Afar region has over 30 woredas that receive block recurrent budget utilization in FY18 in some of the grant resources from the region to provide ser- regions, including Harari, Gambella, Addis Ababa, vices to their communities. The population of Afar and Somali (see Tables 7.8 and 7.9). Of the predom- accounts for about 2 percent of Ethiopia’s popula- inantly highland regions, Tigray and SNNP had the tion. Afar’s share of block grant transfers from the federal government has remained around 3 percent over the past decade. Over the same period, total Table 7.8: Budget outturn—recurrent expenditures by the region (including woredas) has   FY14 FY15 FY17 FY18 FY14–18 increased on average by about 35 percent annu- Tigray 97.0 118.4 105.6 100.0 105.2 ally. The share of capital spending to total regional Afar 94.0 123.1 115.7 96.3 107.3 spending has increased significantly in recent years (Figure 7.9). While this might indicate improved Amhara 102.1 136.9 114.6 98.5 113.0 capacity of the region to implement capital proj- Oromia 102.6 114.1 106.3 99.1 105.5 ects, such capacity in the region is still relatively Somali 90.9 109.2 87.0 84.1 92.8 weak, as Afar is one of the two regions (Somali Benishangul-Gumuz 97.4 114.5 113.0 96.2 105.3 being the other one) that rely on federal-level insti- SNNP 94.6 108.0 106.0 96.1 101.2 tutions to implement capital projects undertaken Gambella 92.0 105.5 124.1 75.1 99.2 using the MDG/SDG Transfers. Harari 102.2 111.3 105.9 62.1 95.4 Economic sectors (including agriculture and Addis Ababa 80.2 87.0 100.4 81.8 87.3 water) took up a large chunk of total regional expen- ditures in Afar over the past decade (Figure 7.10). Dire Dawa 96.9 109.9 107.7 91.1 101.4 Social sectors (which include education and health) All regions 96.2 112.8 106.7 94.6 102.6 also absorbed a large share of regional expenditures. Source: MoFEC. In comparison with other regions, the allocation to the economic sectors has been relatively high, while Table 7.9: Budget outturn—capital allocation to the social sectors was on the lower side. Administration expenses absorbed a slightly   FY14 FY15 FY17 FY18 FY14–18 higher share of regional spending in Afar than in Tigray 99.0 101.0 96.5 102.7 99.8 other regions, presumably due to the relatively less Afar 91.3 112.1 97.6 36.2 84.3 developed infrastructure to start with. Amhara 86.4 73.9 77.0 80.9 79.5 The region’s spending on agriculture and pas- toral development increased by about 60 percent Oromia 88.8 87.7 106.1 63.3 86.4 annually over the past decade. Accordingly, its Somali 80.4 130.7 98.8 84.3 98.5 share of total regional expenditures doubled in Benishangul-Gumuz 88.0 103.5 79.4 68.5 84.8 recent years from 10 percent, 10 years ago. This SNNP 106.2 108.0 83.9 85.2 95.8 was mainly reflected through increased capital Gambella 74.8 51.0 31.8 143.6 75.3 expenditures in the agriculture sector, which drove Harari 100.7 82.9 70.8 67.0 80.4 its share of the region’s total capital spending from Addis Ababa 95.0 73.1 69.2 74.2 77.9 6 percent, 10 years ago to above 30 percent last Dire Dawa 85.1 79.2 78.8 58.1 75.3 year. The increased spending on agriculture in Afar was mainly driven by the provision of additional All regions 92.5 86.8 83.7 75.9 84.7 resources from the federal government through the Source: MoFEC. MDG/SDG Transfers over the past seven years. Chapter 7: The Role of Public Finances   101 Figure 7.9: Regional expenditures in Afar, in million ETB 4,500 4,000 3,500 3,000 2,500 2,000 1,500 1,000 500 0 FY08 FY09 FY10 FY11 FY12 FY13 FY14 FY15 FY16 FY17 Recurrent Capital Total spending Source: MoFEC. Figure 7.10: Sectoral breakdown of regional spending, FY08–17 Afar All Regions Others, Others, 5.5% 6.5% Administration and Administration and general services, general services, 25.1% 20.8% Social Social sectors, sectors, 30.5% 41.8% Economic Economic sectors, sectors, 30.9% 38.9% Source: MoFEC. Spending on water resource development has 10 years ago to 8 percent in recent years. The pro- also been increasing rapidly (by about 50% annu- liferation of woredas in the region is likely to have ally) over the past 10 years. This was also mainly exerted increased pressure on the allocation of driven by higher investment expenditures in the resources to the woredas. The population size in water sector. On average, about 13 percent of Somali region constitutes about 6 percent of the regional expenditures over the past 10 years was total population of the country. allocated for water sector development (Figure 7.11). The pace of increase in regional expenditures in Somali region over the past 10 years has been Somali region much more modest than in Afar (about 33 percent Somali region, which had about 50 woredas five annually). Like Afar region, regional spending on years ago, has increased the number of its woredas infrastructure investments was very small a decade to more than 70 over the last couple of years. The ago, especially considering the huge infrastructure share of the region from block grant transfers from gap in the region, and only picked up in recent years the federal level increased from about 6 percent, mainly due to the MDG/SDG Transfers from the 102   Part 2: Promoting Resilience and Development in the Lowlands Figure 7.11: Functional and sectoral disaggregation of regional spending in Afar, FY08–17 Recurrent Capital Others, 7% Agriculture, 22% Health, Others, 14% 35% Administration and general services, Education, 44% 22% Water, 24% Health, Water, 8% Agriculture, Education, 3% 10% 11% federal government. Because of capacity limita- by about 30 percent annually over the past decade. tions, infrastructure projects built using the MDG/ Allocation to capital expenditures was virtually SDG Transfers are implemented by federal-level nonexistent prior to the introduction of the MDG/ institutions. The sectoral allocation of regional SDG Transfers from the federal government. The spending in Somali region also seems similar to Afar share of capital budget to total regional spending, in that the economic sectors took a larger share which was about 15 percent prior to FY12, jumped than social services (Figure 7.12) to nearly 35 percent after the MDG/SDG Transfers Until recently, the region’s spending on agricul- were introduced. ture and pastoral development was minimal, with Social services, including education and the sector absorbing just over 10 percent of total health, have been important priority sectors regional spending. This has improved in recent in Benishangul-­Gumuz, with the social sectors years with the share reaching more than 15 per- absorbing more than 40 percent of the total cent. While this compares favorably with the total regional budget over the past 10 years (Figure 7.13). regional average (which on average is a little under Economic sectors, which include agriculture, water, 10 percent), it seems to be on the low side in com- trade, and roads, in turn accounted for about parison with the significant investment needs of the 30 percent of the regional budget. The sectoral dis- sector in the region. Investment on water resource aggregation of public expenditures in Benishangul-­ development also improved in the region in recent Gumuz reflects different priorities from Afar and years, with the region allocating up to 20 percent of Somali but is generally closer to the overall regional its capital budget to the sector compared to under average. It would appear that the increased invest- a 10 percent allocation a decade ago. ment correlates with higher levels of human capital, and in particular better educational enrollment Benishangul-Gumuz rates. Benishangul-Gumuz is a relatively small region with Allocations to agriculture and water in only 22 woredas. The population of the region con- Benishangul-­ Gumuz have been well below those stitutes only about 1 percent of the country’s pop- found in Afar and Somali regions over the past ulation. The share of the region from block grant 10 years. The capital investment allocated to the transfers was 2 percent for the past decade but two sectors was just about 18 percent over the declined to just under 2 percent in the latest for- past 10 years compared to 44 percent in Afar and mula applied at the beginning of FY17. 42 percent in Somali. Education and health sectors Public spending by the regional government in took up a significant portion of the regional budget, Benishangul-Gumuz region increased on average both for recurrent as well as capital expenditures. Chapter 7: The Role of Public Finances   103 Figure 7.12: Regional expenditures in Somali region, FY 08–17 Total Expenditures, in Million ETB Sectoral Classification, Total Expenditures 12,000 Others, Administration and 10,000 9% general services, 29% 8,000 6,000 4,000 Social sectors, 2,000 27% 0 08 9 0 1 2 3 4 5 6 7 /1 /1 /1 /1 /1 /1 /0 /1 /1 10 / 16 11 13 14 12 15 09 08 07 20 20 20 20 20 20 20 20 20 20 Recurrent Capital Total expenditures Economic sectors, 35% Sectoral Classification, Recurrent Expenditures Sectoral Classification, Capital Expenditures Others, Agriculture, 4% 18% Health, 11% Others, 36% Education, 18% Water, Administration and 24% Water, general services, 1% 58% Health, 10% Education, Agriculture, 12% 8% Source: MoFEC. Gambella allocated for capital spending was as low as 6 per- Gambella region has only 14 woredas that are cent a decade ago but has improved in recent constituted under three zones. Its population size years. The reduction in the volume of the MDG/SDG roughly accounts for about 0.5 percent of Ethiopia’s Transfers over the past few years, however, has population. The region’s share in block grant trans- again led to a reduction in the allocation to capital fers from the federal government has remained expenditures by the region. around 1.5 percent over the past decade. Similar to Benishangul-Gumuz, social services Public spending in Gambella by the regional including education and health took up a large government has shown only a modest increase share of public spending by the region, while allo- over the past 10 years. There has been very lim- cation to economic sectors was largely lower than ited spending on capital investment by the region, in the other lowland regions. Education and health although the MDG/SDG Transfers helped to partly sectors jointly absorbed nearly 40 percent of both redress this. The proportion of the regional budget recurrent and capital budgets of the regional gov- ernment over the past 10 years. 104   Part 2: Promoting Resilience and Development in the Lowlands Figure 7.13: Regional expenditures in Benishangul-Gumuz region, FY08-17 Total Expenditures, in Million ETB Sectoral Classification, Total Expenditures 3,500 Others, 3,000 2% Administration and general services, 2,500 28% 2,000 1,500 1,000 Social 500 sectors, 41% 0 08 9 0 1 2 3 4 5 6 7 /1 /1 /1 /1 /1 /1 /0 /1 /1 10 / 16 11 13 14 12 15 09 08 07 20 20 20 20 20 20 20 20 20 20 Economic Recurrent Capital Total expenditures sectors, 29% Sectoral Classification, Recurrent Expenditures Sectoral Classification, Capital Expenditures Agriculture, Others, Others, 10% 6% 17% Water, Health, Administration and 8% 14% general services, 38% Health, Urban 8% development 9% Education, Roads, 25% Education, 23% 28% Agriculture, Water, 12% 2% Source: MoFEC. SNNP grants. Over the past five years, allocation to the SNNP region is composed of 14 zonal administra- pastoral zones in SNNP accounted for about 15 per- tions and about 135 woredas, including 5 special cent of the total block grant allocated by the region woredas. The zonal administration in SNNP has to zonal administrations and woredas (Figure 7.15). greater administrative autonomy compared to The sectoral disaggregation of the budget in similar administrative units in other regions. Of the pastoral zones is largely similar to the overall the 14 zones in the region, the pastoral zones are regional average in SNNP. Over the past five years, South Omo, Bench Maji, and Kaffa which account education was allocated a significant portion of the for about 14 percent of the total population in the budget, both at the regional level as well as in the region. Similar to the practice in the other regions, pastoral zones (Figure 7.16). The share of admin- the SNNP regional government allocates budget- istrative and general services expenditures in the ary resources to the zonal administrations and pastoral zones was higher than the regional aver- the woredas under these zones in the form of block age. However, allocation to agriculture and water Chapter 7: The Role of Public Finances   105 Figure 7.14: Regional expenditures in Gambella region, FY08–17 Total Expenditures, in Million ETB Sectoral Classification, Total Expenditures 2,000 Others, 7% Administration and general services, 1,500 32% 1,000 500 Social sectors, 38% 0 08 9 0 1 2 3 4 5 6 7 /1 /1 /1 /1 /1 /1 /0 /1 /1 10 / 16 11 13 14 12 15 09 08 07 20 20 20 20 20 20 20 20 20 Economic 20 sectors, Recurrent Capital Total expenditures 23% Sectoral Classification, Recurrent Expenditures Sectoral Classification, Capital Expenditures Health, Health, 15% 11% Agriculture, 29% Education, 24% Administration and general services, 44% Education, 29% Water, 16% Roads, 16% Water, 2% Agriculture, 14% Source: MoFEC. Figure 7.15: Block grant allocation to pastoral and non-pastoral zones in SNNP, FY 12–17 100 90 80 70 60 50 40 30 20 10 0 2011/12 2012/13 2013/14 2014/15 2015/16 2016/17 Pastoral zones Non-pastoral zones Source: MoFEC. 106   Part 2: Promoting Resilience and Development in the Lowlands Figure 7.16: Public expenditures in pastoral zones of SNNP, FY15–17 SNNP Region South Omo Zone Administration and Others, general services, 6% Administration and 19% Others, general services, 19% 34% Health, 14% Agriculture, 10% Health, 11% Water, 4% Trade and Education, industry, 29% Urban 5% Agriculture, Education, development, Water, 12% 28% 4% Trade and 1% industry, 4% Bench Maji Zone Kaffa Other, Other, 8% 5% Health, Health, 11% 12% Administration and general services, 28% Administration and general services, 36% Education, 37% Agriculture, Education, 11% 28% Agriculture, Water, Water, 11% Trade and 2% Trade and 1% industry industry, 5% 5% Source: SNNP BOFED. resource development followed a largely similar including Borena, Bale, Guji, East Hararghe, and trend as the regional average. West Hararghe zones. The combined population in these pastoral zones accounts for about 30 percent Oromia of the total population in the region. Similarly, the Oromia region, the largest regional state in Ethiopia proportion of the block grants resources allocated both in terms of land and population, is divided into to the pastoral zones in Oromia over the past seven 20 administrative zones (two of which were newly years hovered around 30 percent of the total block added over the last two years). Of the 20 zones, grants allocation (Figure 7.17). there are 5 which are predominantly pastoral, Chapter 7: The Role of Public Finances   107 Figure 7.17: Block grant allocation to pastoral and non-pastoral zones in Oromia region, FY11–17 100 90 80 70 60 50 40 30 20 10 0 2010/11 2011/12 2012/13 2013/14 2014/15 2015/16 2016/17 Pastoral zones Non-pastoral zones Source: MoFEC. Conclusions regions, for example, have predominantly focused Despite the use of different allocation mechanisms on investments in water and agriculture to the over the past two decades, fiscal transfers from relative neglect of education and health, while in the federal government to regional governments Gambella and Benishangul-Gumuz the regional has shown very little change in terms of the share authorities have preferred to allocate expenditures that regions receive from the total pool. Allocations to rural roads and the social sectors, particularly to each region have closely followed population education and health. The pastoral areas of SNNP size. Given the scale of the need in many parts of also allocated a large proportion of their budgets the lowlands and the much higher costs of deliv- toward education and health. Given the large ery in these areas, there may be an argument for human capital deficits in Afar and Somali regions, even higher allocations than those received (see there may be strong case for the regional govern- Chapter 9). ments to allocate more resources to the education Until recently, the lowland regions, including and health sectors in their regions. In general, given Afar, Somali, Gambella, and Benishangul-Gumuz, the very tight fiscal space available to the federal allocated only a very small part of their budget to government, more generous allocations to lagging capital expenditures, preferring to spend money on lowland regions may be politically and financially recurrent expenditures (i.e., salaries and operating difficult. But there may be a strong argument for expenditures). It was only after the introduction of such additional allocations to boost human capital the MDG/SDG Transfers that some relatively mean- development. ingful levels of allocation to capital investments The next chapter examines the role of humani- started to be made. Different regional authorities tarian and safety net responses to emergency and have made different choices about the sectoral chronic needs in the lowlands. allocation of capital investments. Afar and Somali 108   Part 2: Promoting Resilience and Development in the Lowlands Poverty and Vulnerability in the Ethiopian Lowlands Chapter 8: Humanitarian Interventions87 and Safety Nets Working in Tandem Summary environments. To date, the two have not always This chapter examines how government and the inter- worked in close tandem, as short-term humanitarian national community have responded to emergency interventions have been relied on to too great of an and chronic food needs in the lowlands. Building the extent to tackle predictable and chronic needs. Long- resilience of lowland populations to shocks requires term needs should be built into government safety net the ability to respond effectively to both rapid onset and development systems. Bringing these two together emergencies, such as conflict, as well as slow onset will be key in building lowland resilience. environmental and other changes, such as drought and extreme weather events, which are affecting a Introduction household’s ability to cope and recover from shocks. To date the government has been using two main The instruments to do both should complement rather financing instruments to respond to emergency than undermine each other so emergency response and chronic food needs. The first mechanism is should seamlessly transition to a safety net system to the Humanitarian Requirement Document (HRD) support chronic and predictable need.87 in which the government appeals for humanitarian Building resilience is ultimately about building sus- assistance from international donors to respond to tainable systems that can both tackle the short-term emergency needs that arise mainly from drought crisis but also the long-term needs of the population to and other weather-related events, such as flood- protect themselves against future shocks and smooth ing.88 The second mechanism is the Productive out the high degree of variability and risk in lowland 88On average food assistance has accounted for 76 percent of the 87Humanitarian Food Assistance refers to food or cash distributed to appeal. In 2019 the complex needs of 3 million IDPs and returnees have beneficiaries to meet their emergency food needs. changed the profile of the HRP, with 50 percent allocated for food. 109 Figure 8.1: Humanitarian requirements, 2008–18 Figure 8.2: Humanitarian requirement by type of need, 2.0 12% 2008–18 Education, 10% Agriculture, 3% 4% Others, 1.5 2% 8% Water and sanitation, 1.0 6% 5% Health and 4% nutrition, 0.5 10% 2% 0.0 0% 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 In billion US$ (left) 2018 Food, 76% Source: HRD, various issues, and MoFEC. Source: HRD, various issues, and MoFEC. Safety Net Program (PSNP), which is a government-­ led, mainly donor-financed program, that supports been equivalent to 1.4 percent of GDP over the last people that are chronically food insecure. six years or on average equivalent to 8.9 percent of government revenue,90 whereas spending on safety Humanitarian Intervention nets (rural and urban PSNP) has cost the equivalent The HRD is a biannual appeals process correspond- of less than 0.5 percent of GDP.91 More than 75 per- ing to the two main harvest seasons, meher and belg. cent of the total amount appealed through the HRD HRD meher appeal numbers are typically released in process constituted food aid (Figure 8.2). From the January or February, reflecting the number of people nonfood components, the main needs came from in need of food assistance as a result of the perfor- health and nutrition, WaSH, and education. In 2019 mance of the meher harvest. Updated appeal num- the complex needs of IDPs has changed the profile bers, corresponding to the belg harvest, are typically of the response, with only 50 percent of the 2019 released in July or August, reflecting the number of HRD appeal being for food. people in need of assistance either because of the The number of people requiring food assistance performance of the belg harvest, or because of the over the past decade averaged about 4.8 million extent of need from the failure of the meher harvest. people a year, ranging between 2.5 and 10.0 million The start time of the assessments is frequently people (Figure 8.3). On average, the number of ben- delayed (by two or three weeks) in the lowlands until eficiaries constituted about 6 percent of the entire the end of the lowland rains (gu/ganna and deyr/ population. The number of people impacted varied hageya) to ensure their impact can be included. across the regional states. The lowland regions, Humanitarian requirements requested through including Somali, Gambella, and Afar, had a rela- the HRD process over the past 10 years on average tively large proportion of their population requiring amounted to nearly US$700 million a year. The food assistance over the past 10 years (Figure 8.4). amount of appeal saw a significant increase in the Based on the hot spot classifications (hot spots past three years as emergency needs increased are woredas in need of emergency support), the significantly, mainly due to the El Niño–induced drought-prone lowlands have been consistently tar- drought in 2015/16 and La Niña in 2016/17. The geted for humanitarian assistance in Somali, Afar, humanitarian requirements were $1.6 billion and and the Borana zones of Oromia region (Map 8.1). $1.4 billion for 2016 and 2017 respectively (Fig- ure 8.1).89 On average the humanitarian appeal has 90Ibid. 89Omzigt, Dirk-Jan, A Study on Lowland Resilience in Ethiopia—­ 91Endale, et al. Financing Social Protection in Ethiopia: A long-term Building a More Developmental Approach to Responding to Shocks, perspective, OECD Development Policy Papers, February 2019, May 2018, UNDP. No. 15, p. 32. 110   Part 2: Promoting Resilience and Development in the Lowlands Figure 8.3: Number of people requiring food aid, 2008–18 Figure 8.4: Regional distribution of people requiring food aid 12 12 (2008–18), as a proportion of total population 25 10 10 20 8 8 15 6 6 10 4 4 5 2 2 0 ay Am r ra ia i NP l -G a uz a ri al a ll w ra om Af ha m be um gr Da SN Ha So Ti m 0 0 Or re Ga Di gu 08 09 10 11 12 13 14 15 16 17 18 an 20 20 20 20 20 20 20 20 20 20 20 sh ni Be In millions (left) In % of total population (right) Source: HRD, various issues, and CSA. Source: HRD, various issues, and CSA. Map 8.1: Frequency of hot spot woredas 2008–2017 In response to the appeals through the HRD ETB (~US$175 million). Average annual spending process, resources worth a total of about US$4 on emergencies by the regions between 2013/14 billion are estimated to have been mobilized to and 2016/17 (Ethiopian fiscal years) by Oromia was cover food aid only over the past decade. The U.S. highest at 120 million ETB (~US$4.2 million), fol- government has been the major donor, covering lowed by Somali and Amhara.92 nearly half of the total resources mobilized. Other The critical question is whether responding to major donors included the United Kingdom and EU weather-related events in this way is the most (Figure 8.5). The Government of Ethiopia covered a little more than 10 percent of the total resources. Regional spending on droughts and other emer- 92Endale, et al. Financing Social Protection in Ethiopia: A long-term gencies is not included in the figures; this rose perspective, OECD Development Policy Papers, February 2019, significantly in 2015/16 reaching nearly 5 billion No. 15, p. 31. Chapter 8: Humanitarian Interventions87 and Safety Nets Working in Tandem   111 Figure 8.5: Food aid contributions, 2009–17 vulnerability in the lowlands through the provision GoE, of social transfers. Due to data constraints, the 12% discussion focuses primarily on Afar and Somali Others, Regions but includes SNNP and Oromia regional 22% data where available. Unless otherwise stated, data sources are from the PSNP’s independent Impact and Performance Evaluations completed every two years (2008, 2010, 2012, 2014, 2016). EU+ECHO, We make a series of policy recommendations 8% regarding the design of an appropriate safety net system in the lowlands that supports resilience U.S., building. The discussion raises fundamental strate- U.K., 13% 45% gic and design recommendations which need policy direction before any detailed program design work can begin. These recommendations relate inter alia Source: HRD, various issues. to policy issues of allocative efficiency, horizontal equity, governance and accountability, financial planning, and capacity. Social assistance should continue to be a effective use of donor resources.93 Drought is a pre- first-order priority for building resilience in the dictable occurrence in the lowlands and responses lowlands—although alone it is insufficient. The to it should be built into government systems; swift first critical step toward building resilience in the action through government systems can reduce lowlands is to redesign the government-led safety both human suffering and the macroeconomic net system, ensuring it is appropriate for lowland effects of humanitarian crisis.94 Going forward contexts and can flexibly respond to local condi- there is a need to gradually shift to more predict- tions (drought, for example) through one financial able and relatively less costly ways of responding to framework and one implementation rule book for droughts, while preserving emergency humanitar- multiple implementing partners. By recommitting ian interventions for relatively less predictable rapid to this system as a foundational platform, social onset disasters, such as displacement because of assistance programs (or safety nets) can then conflict. In this regard, there may be a variety of be complemented by livelihood and other more ways of smoothing out the consumption peaks ­ development-based interventions to build the and troughs of lowland environments, from the use resilience of poor and vulnerable households in the of contingency budgets and risk transfer (insur- lowlands. ance) instruments, to an adapted social protection There are multiple systems and service providers system like PSNP, that can scale up assistance to that support poor and vulnerable households in the those that need it when required. lowlands. The two largest vehicles for responding to needs in the lowlands are the Productive Safety Reimagining PSNP for the lowlands Net Program (PSNP) and Humanitarian Food This section focuses on the critical policy elements Assistance (HFA). Box 8.1 has more details relat- needed to create one government-led system that ing to these systems. Importantly, both systems effectively and efficiently addresses poverty and are implemented through government agencies and multiple nongovernmental organizations. In 93 A large proportion of humanitarian aid beneficiaries receive support addition, there are several government and non- consistently; more than half of HFA clients received HFA assistance for government livelihood support programs,95 aimed three consecutive years. This figure is over 80 percent in Somali and at strengthening households’ livelihoods through over 50 percent in Afar. Hirvonnen et al., Humanitarian Food Assis- tance 2019: Process Evaluation Report, DRAFT January 2019. 94 IMF Article 4 discussion on the 2015/16 drought noted, ‘Ethiopia’s macroeconomic outturn . . . [had] . . . been adversely affected by a severe drought and the weak global environment. As a result, output 95Such as the Government’s Pastoral Community Development Pro- growth is estimated to have slowed down in 2015/16 to 6.5 percent. gram (PCDP) and bilateral development partner funded programs The slowdown was mitigated by effective and timely policy responses such as USAID’s Pastoral Areas Resilience Improvement and Markets to the drought, and buoyant industrial and services sectors. Expansion (PRIME) program. 112   Part 2: Promoting Resilience and Development in the Lowlands Box 8.1: The Productive Safety Net Program (PSNP) and Humanitarian Food Assistance (HFA) in the lowland areas The Productive Safety Net Program (PSNP) in lowland disciplines, assessing progress toward development goals areas. The design of the national PSNP is based on livelihoods over time. and food security analysis in the highland areas of Ethiopia. Humanitarian Food Assistance (HFA) in the lowland The PSNP aims to address the symptoms and causes of areas. HFA is a complex network of agencies providing food food insecurity by providing predictable transfers in cash or or cash to meet emergency food needs using a common food each month to poor households during the six-month analytical framework on livelihoods and food security. HFA lean season, in exchange for a labor contribution from the attempts to address the symptoms of food insecurity by household to build community assets. A smaller number of providing food (and increasingly cash) transfers to transitory households benefit from ‘direct support’, which is offered for food insecure households to meet immediate food needs. 12 months to those with no able-bodied adults. The focus While there are data on the outputs of agencies, there are of the community asset building is on restoring natural limited data to measure impact on the lives of food assis- resources within watersheds. The PSNP has benefited from tance recipients. multiple impact evaluations and assessments in multiple community resilience initiatives, social service pro- the impacts of HFA across the country is limited, vision, and market access. in the lowlands the challenge is exacerbated by HFA and the vertical or horizontal scale-up the harsh operating environments, poor report- of PSNP use food need assessments as the pri- ing lines, and a lack of demand for evidence (from mary means of determining coverage, meaning donors and government), as well as access issues. that beneficiaries of HFA and the transitory (or Data on the HFA tend to be expressed in terms of extended) PSNP beneficiaries are not (necessar- the number of people in need of support of food ily) the poorest. National food needs assessments assistance or the amounts of support provided. are completed every six months to determine There is little regular or comparable reporting on future numbers of people ‘in need’. While technical the timeliness, number, and size of monthly trans- in nature, the assessments are also informed by fers received by households across operators and political considerations. The assessment results over time on the targeting accuracy of HFA,96 on are used by the government as the basis for issuing accountability to beneficiaries, or on a host of emergency appeals. In recent years the response other aspects of program delivery. There is no reg- was implemented through both HFA and PSNP ular information on or assessment of how the HFA scale-ups. Together, the HFA and the PSNP foot- and PSNP together support poor and vulnerable prints mirror the size of historic food assistance households. While there is reasonable information need assessments, rather than poverty data. As a from the HFA donors and operators on program result, HFA and PSNP caseloads reflect vulnerability inputs and outputs, there is no empirical evidence to drought and self-reported consumption—and on impact although there are impacts that can be while they could be used as a proxy for poverty, inferred—for example, it can be argued that as a they may not reflect poverty incidence or trends result of both HFA and PSNP, there were no famines more broadly. declared in 2011, 2012, 2015, or 2016. Data on how The PSNP aims to address both symptoms and HFA contributes to building resilience are scant. causes of food insecurity with regular assessments The number of households enrolled in the low- of impact and performance. Data are monitored land PSNP is significant and they receive support regularly, and data on inputs, outputs, and out- through accountable government delivery systems. comes are independently evaluated with increas- Between 2008 and 2017, the number of the PSNP ing efforts since 2010 to ensure that the lowland beneficiaries in Afar Regional State remained fairly regions are held accountable in the same way as constant, fluctuating slightly between 470,00 and the highland regions. The PSNP’s evidence-based 560,000. In Somali Regional State the number approach allows the identification of constraints and opportunities to address poverty and vulner- 96 Hirvonnen et al., Humanitarian Food Assistance 2019: Process Eval- ability. The HFA aims to address the symptoms of uation Report, DRAFT January 2019 is a first step in assessing target- food insecurity (vulnerability), although empirical ing of humanitarian assistance and in particular shows the importance evidence of the impact is scarce. While evidence on of PSNP and HFA working together as a combined system. Chapter 8: Humanitarian Interventions87 and Safety Nets Working in Tandem   113 Figure 8.6: PSNP caseloads in Afar and Somali regional states Figure 8.7: HFA needs in Afar and Somali regional states (2008–2017) (2008–2017) 2,500,000 2,500,000 2,000,000 2,000,000 1,500,000 1,500,000 1,000,000 1,000,000 500,000 500,000 0 0 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Jun-08 2009 2010 2011 2012 2013 2014 2015 2016 2017 Afar Somali Afar HFA Somali HFA of PSNP beneficiaries increased steadily over the Figure 8.8: Total needs in Afar and Somali regions (2008–2017) same period, responding to the periodic reviews of 5,000,000 needs assessments. Figure 8.6 clarifies the size of 4,000,000 the PSNP caseloads in the lowland regions. Imple- 3,000,000 mentation of the PSNP across the lowlands is man- aged through a common set of procedures, rules, 2,000,000 and accountability structures, with clear reporting 1,000,000 lines. The PSNP is ‘on budget’, has regular reporting, 0 and a variety of feedback loops to beneficiaries, the 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 various government structures, and donors. The number of households identified as ‘in need’ Total PSNP Total HFA of HFA is large, yet they are supported through weak delivery systems. The number of households ‘in need’ of emergency food support through the HFA have fluctuated significantly over the last increased over time, with 2016 and 2017 repre- decade in both Afar (as low as 86,000 people) senting the highest figures, following a recalcu- and Somali regions (as high as 1.6 million people), lation of those in need of PSNP transfers and the as shown in Figure 8.7. While the delivery of HFA major droughts demanding additional support in Afar is mostly through a government system, through HFA. The overall increase is reflective of the in Somali Region the delivery is through interna- national trend of an increase in the number of peo- tional organizations and NGOs, with government ple in need of social transfers (cash or food). agencies only nominally involved due to limited Responding to poverty and vulnerability inter capacity to effectively manage the operation. In alia through a well-designed safety net remains both regions, the HFA is, however, off budget, with an effective approach in the lowlands of Ethio- no reporting using standard government systems, pia. As detailed in the following section, there is and with significant weaknesses in the reporting an increasing body of evidence to justify invest- and accountability systems. Further, HFA continues ments in safety nets as a policy contribution for to be unpredictable (it is almost exclusively funded addressing poverty and vulnerability, and for build- through international donors), late, and expensive. ing resilient households. The PSNPs documented Together, the number of households that are effectiveness in the highlands, as well as programs addressed through either the PSNP or HFA in the in other lowland areas of East Africa (e.g., Kenya’s lowlands is increasing. Figure 8.8 shows the com- Hunger Safety Net Program), show the significant bined annual PSNP and HFA caseloads in Somali effects that safety net approaches can have on and Afar regions since 2008. The numbers have long-term development outcomes and building 114   Part 2: Promoting Resilience and Development in the Lowlands resilience.97 Safety net approaches are also typi- each year, as rainfall fluctuated and cumulatively cally more cost-effective than alternatives such as affected livelihoods. Inflexible financing instru- HFA approaches—by as much as 30 percent more ments and rigid administrative ‘quotas’ between efficient—saving between $2.3 and $3.3 for every ­ the numbers of chronic food insecure households $1 spent on safety nets.98 In the case of Ethiopia, and transient food insecure households meant that such a cost-effective approach would be built on the PSNP could not provide temporary support to the significant investments already made in sys- all those in need. A tightly controlled financial enve- tems and capacity. lope, with a lack of flexible financial instruments and capacity constraints, resulted in a continued Impact and Performance of the PSNP reliance on the HFA to address transient needs. The While the PSNP in the highlands made important PSNP’s contingency budgets held at the federal and gains in addressing chronic poverty, progress in the woreda levels were inadequate in size and inade- lowlands was slower. Over the course of a decade, quately used to accommodate all transitory needs. the PSNP in highland areas was able to increase The effective targeting of beneficiaries in the food security by 1.48 month for public works par- lowlands was compromised with a lack of adher- ticipants, and by 1.93 months for nonpublic works ence to related program rules. The effectiveness participants, increase nonfood expenditures by of poverty targeting in the highlands increased 43.7 birr per month, increase the number of food steadily over time with additional support, capac- groups consumed by participants by –0.5, limit ity, and training so that by 2012 targeting was con- households’ distress sales, and protect them sidered good by international standards. Despite against the impact of drought. In the lowlands, the similar, significant investments in targeting in the impact of the PSNP was more muted, achieving a lowlands, challenges continued throughout. There reduction of households’ food gap by a maximum was significant ‘rotation’ of beneficiaries within of 0.5 months, a (maximum) 21 percent increase the PSNP, with many registered beneficiaries only in income levels, and the stabilization of household receiving four months of transfers. This may have assets. Progress in the highlands PSNP was in part been an administrative solution to the rigid quotas a result of an appropriately designed program of PSNP beneficiary numbers. Some larger house- that responded to local poverty and vulnerability holds only received as little as 27 percent of their dynamics. Different climate, production systems, entitlement, which may have been a way of provid- social services, and degree of market integration ing transfer sizes tailored to individual households. mean that the ratio of poverty to vulnerability in There is also evidence that some PSNP house- the highlands is half the ratio experienced in the holds circumvented the rigid quotas of program lowlands. Regrettably however, the lowland PSNP beneficiaries in place at a local level by sharing design simply replicated the highland PSNP design, their transfers with non-PSNP households—some despite these differences. A reluctance to ensure 38 percent of PSNP households confirmed this was the lowlands program responded to local dynam- a regular practice. Local notions of ‘fairness’ in the ics was inter alia born out of concerns regarding lowland areas are that everyone should benefit regional allocations and equity. regardless of wealth and that targeting a minority The design of the PSNP in the lowlands meant of community members (albeit poor ones) can that it could not respond adequately to transitory engender tensions.99 Concerns around elite capture food insecurity, or vulnerability, in particular. Pas- of targeting and enrollment are legitimate, with evi- toral households’ needs fluctuate in line with the dence from Somali region indicating that having a ‘sawtooth curve’ of economic ups and downs that household head in an official position increases the reflect expansion and contraction of herd sizes likelihood that the household (and kin) are selected and milk production. Accordingly, the numbers of into the PSNP by 22 percent. These examples show vulnerable households ‘in need’ varied considerably that program rules regarding the poverty-targeting approach of the program were not always strictly adhered to—and go some way toward explaining 97 PSNP achieves both social and productive goals by raising income in beneficiary households while stimulating local and national production (IFPRI, 2016) and improves health and school attendance (Berhane et al., 2015) (Shigute et al., 2017), achieves higher rates of insurance 99 Lind, J., et al., Targeting social transfers in pastoralist societies: Ethi- uptake, and improves cognitive skills in children (Berhane et al., 2016). opia’s productive safety net program revisited, ESSP Working Paper, 98 Cabot Venton (2018). Case studies: Ethiopia, Kenya, and Somalia. IFPRI, 2018. Chapter 8: Humanitarian Interventions87 and Safety Nets Working in Tandem   115 reduced impacts of the program compared to its community members to appeal their exclusion highland implementation. from the program, allow local administrations to The delivery of PSNP transfers to households resource solutions for exclusions, and require var- remained erratic, unpredictable, and irregular. From ious levels of government to monitor and report the highlands, there is strong evidence that the on resources and processes. While these systems extent to which transfers are predictable, regular, could be replicated by the HFA, they were not, and timely affects the intended PSNP outcome and there was little reporting or monitoring from of household food security. Beneficiaries are able HFA to government and weak auditing of HFA to plan household consumption and investments commodities. The lack of a harmonized account- better, and they are seen as more creditworthy. ability mechanism stretches (already weak) local In the lowlands, little progress was made against implementation capacity, could lead to perverse program targets on timeliness and predictability of incentives for regions and households to prefer food transfers (the vast majority of transfers in the one program over the other, and confuses district/ lowlands): food transfers to beneficiaries were up to kebele staff responsible for delivery of transfers. 60 days late, with a two (sometimes three) months’ Investments through the safety net have had payment often being paid together, typically with positive spillover effects for HFA in accountability only one day’s notice, requiring an overnight stay and targeting. In woredas that have both PSNP outside of the local community, and only 60 per- and HFA, targeting is progressive, and communi- cent of PSNP households indicated they could plan ties that have HFA but no PSNP have significantly ahead based on the predictability of the PSNP poorer targeting.100 Kebele Appeals Committees transfers with the amounts received often less are ‘relatively rare’ in areas that receive HFA but than expected. The shift to cash and e-payments do not receive PSNP, but common where PSNP through the PSNP is currently seeking to address is present.101 Fifty-five percent of people in non- some of these challenges and, based on a range of PSNP woredas think that local leaders who help international experiences, will also increase trans- with program implementation should benefit from parency through a robust audit trail of receipts. HFA, compared to just 16 percent of people in PSNP Capacity constraints in the lowlands affected woredas.102 The better performances of HFA in the progress of the PSNP toward its objectives. PSNP woredas is likely due to system strengthen- Capacity constraints across the lowlands have ing investments under the program, which includes materially affected PSNP performance with, for community awareness of the program rules. example, only 50 percent of key financial PSNP The use of a highland design for public works staff in place in Somali Region and only 66 percent hampers the PSNP’S ability to contribute to of staff having received training or having access to addressing the causes of chronic food insecurity vehicles to perform their work tasks. Importantly, in the lowlands. The PSNP, informed by analysis accountability mechanisms need to be further and practice in highland areas, uses a commu- strengthened when compared with performance nity participatory watershed planning system to in the highlands. For example, by 2016 no kebeles identify public works which contribute to tackling had made program identity cards available to new root causes of food insecurity. While a planning PSNP beneficiaries in Somali region, the use of approach using the watershed model in the high- Kebele Appeals Committees was limited, and only lands has proven to be highly successful, the wide 25 percent of households understood the roles of geographic spaces, low gradients, sparse popula- key PSNP structures or were aware of key decision- tions, and labor-based natural resource conserva- making meetings. tion in the lowlands mean that a watershed model The PSNP has clear accountability mechanisms, cannot be expected to contribute strongly to local which are not mirrored by the HFA. While the PSNP livelihoods. Despite several attempts to moderate and HFA are delivered to the same (and neighbor- the public work planning methodology, practice ing) communities using the same frontline staff on the ground still fails to take into account more at kebele and district levels, the same targeting criteria, and delivering the same type of transfers, 100 Hirvonnen et al., Humanitarian Food Assistance 2019: Process Eval- they have different accountability mechanisms. uation Report, DRAFT, January 2019. The PSNP has several administrative, financial, 101 Ibid. and quasi-judicial checks and balances that allow 102 Ibid. 116   Part 2: Promoting Resilience and Development in the Lowlands considered approaches, such as rangeland man- seven rounds of transfers to be delivered. When agement, drought cycle management, or a focus on these ‘breaks’ occur, households do not receive social infrastructure. their expected transfer and are not retrospectively compensated. A country evaluation of operations in Impact and Performance of the HFA Somali noted the importance of actors mitigating Impact data on the HFA is scarce. There is no con- governance and accountability risks in the delivery sistent process to review the targeting or outcomes of food and nutrition assistance in Ethiopia, with of HFA. Though there is limited detailed empirical particular focus on Somali region.104 Persistently evidence on impact of the HFA, there are impacts high levels of food aid accompanied by persistently that can be inferred. For example, the droughts high levels of malnutrition in Somali region merits of 2011, 2012, 2015, or 2016 can be compared in further research. severity and numbers affected to the droughts of 2002, where multiple deaths occurred and famine Necessary and Essential Measures was declared. As a result of both HFA and PSNP, to Transform the Current Approach there no deaths and no famines were declared in There are three policy changes that are necessary 2011, 2012, 2015, or 2016. There are no data how- to transition the current PSNP in the lowlands ever on the impact of HFA on human capital (e.g., into a more effective system. While the PSNP has nutrition) or household economy (e.g., productive proven to be a powerful instrument for tackling assets) or wider development outcomes. As set poverty and vulnerability, further improvements in out in Chapter 4, a recent analysis of Household design and implementation are needed to deliver on Income, Consumption and Expenditure Surveys this promise in the lowlands: (HICES) data shows that humanitarian aid has been poverty reducing but significantly less so in • redesign one government-led safety net, appro- Somali region103 than Afar. There are no data on priate for lowland contexts, whether HFA prevented the sale of distress sales • prioritize the timely delivery of social transfers to during droughts. Data on whether HFA contributes intended households, and to building resilience more generally are equally • strengthen governance and accountability and limited. address capacity bottlenecks. The number of HFA transfers delivered are often not the same as the number of planned transfers—­ A. Redesign the government-led safety net system, and transfers that are not delivered are ‘lost’ to appropriate for lowland contexts the beneficiary. There is little regular or compara- Develop one common policy framework for initia- ble reporting on the timeliness, number, and size tives to address poverty and vulnerability in the of monthly HFA transfers received by households lowlands. There is currently an unenforced agree- over time; on the targeting accuracy of HFA; nor ment of ‘continuity’ between the HFA and PSNP. A on accountability to beneficiaries. From data new policy framework for the lowlands should be that are available, while HFA implementers may established that provides overall direction for how plan to deliver a fixed number of monthly trans- the PSNP and HFA should be brought under one fers to households (so-called ‘rounds’ of transfer), system, outlines operational efficiencies to make scarce and delayed resources result in ‘breaks’ in use of scarce implementation capacity, clarifies one the delivery. Over the last decade, for example, institutional structure required to deliver support, despite the HFA needs assessments indicating that and ensures that the PSNP and HFA are delivered 12 months of food assistance may be necessary with the same levels of accountability and trans- in some zones, it has been unusual for more than parency. The key elements of this framework are set out below. Ensure that the program design and financing 103 The less significant poverty reducing impact in Somali region could responds to the relative size of poor and vulnerable have been because the drought in the HICES reference year 2016 was caseloads in the lowlands, which are different from less severe in Somali (noting that malnutrition indicators remained the highlands. In practice, the PSNP has a fixed at emergency levels in seven zones) but also may indicate targeting challenges or elite capture, or that resources are not reaching those targeted to receive them or are not reaching them in the appropriate quantity. 104 WFP Country Evaluation, 2018. Chapter 8: Humanitarian Interventions87 and Safety Nets Working in Tandem   117 national ratio between a ‘core’ caseload of chron- instrument requires credible data against which ically food insecure beneficiaries and a ‘transitory’ decisions can be made to scale up in certain areas caseload, which it can modestly respond to as vul- to certain households when certain predefined con- nerability conditions change and available financing ditions have been met. The current system does not allows. The HFA augments the PSNP’s response to allow for this: there are limited credible data inputs; the ‘transitory’ caseload.105 The ratio between a triggers for action are not well-established; and ‘core’ caseload of PSNP beneficiaries and a ‘transi- systems are not well-embedded within government tory’ caseload in the lowlands should better reflect or linked to national, regional, zonal, and woreda local conditions (it currently reflects highland level capacities. The current system can also be dynamics) and will need to change on an annual influenced by political factors. An agreement on the (possibly seasonal) basis, as rainfall and vegetation use of objective ‘thresholds’ to “trigger” a response changes. According to poverty data, a ‘core’ regular of differing magnitudes implemented under the caseload in the lowlands may need to be smaller government-led system is required, supported by than some regions’ current caseload (i.e., in Somali the necessary investments that allow for the tech- region) and the annual allocation to a ‘transitory’ nical measurement of the thresholds. This should caseload larger than currently programmed. Within be considered central to an ex ante approach that regional boundaries, beneficiaries are likely to differ addresses poverty and vulnerability. in proportions, according to the incidence of pov- Introduce one financial framework with appro- erty in pastoral, agropastoral, and farming areas. priate financial instruments for a scalable system. Ensure there are appropriate, scalable instru- The multiple systems and actors in the lowlands ments that can deliver support to those who are currently leverage multiple sources of financing poor and those who are vulnerable to poverty. With through government systems, as well as parallel the numbers of vulnerable households approxi- to government. Adopting one framework will lead mately two times the number of poor households to humanitarian financing flowing through PSNP in lowland areas, it follows that appropriate instru- systems and will clarify the limits for what gov- ments able to rapidly respond to their needs are ernment will accept financial responsibility for required (i.e., a transfer program with flexible deliv- when responding to vulnerability. This will then ery mechanisms that can reach those who may enable a realistic assessment of the related costs not be poor today but who are likely to become in addressing responsibilities, leading in turn to a poor if impacted by a ’shock’). Being vulnerable to more effective deployment of resources, and ensur- poverty and risk does not mean that households ing complementarity between government and are ‘in need’ all the time—rather, that vulnerable nongovernment resources. households may need rapid support as livelihood Ensure one rule book for all implementers in the conditions change. When assistance is needed, a lowlands. While all PSNP actors work to a common temporary and rapid scale-up to a larger case- ‘rule book’ of processes and procedures, others— load must happen before the ‘alarm’ sounds for the including the HFA—do not. Standardizing informa- drought—multiple research shows that early inter- tion sets and ways of working across actors are ventions are more cost effective106—and with a ‘no needed to strengthen transparency and account- regrets’ approach. This means that there should be ability and make better use of implementation prepositioned funds, and/or an annual allocation, capacity. Adopting a common administration sys- to enable an annual response to transitory needs tem for all implementers would accelerate the shift among the lowland regions. from fragmented projects to common systems. Make critical and essential investments in the Developing an operational system that can respond Government’s Early Warning System. A scalable to poverty and vulnerability requires flexible admin- istration and strong capacity at local levels. 105 While the PSNP can support a transitory caseload in all woredas B. Prioritize the delivery of social transfers within a PSNP region, in practice the HFA is directed to woredas with- out a PSNP core caseload and to woredas where the “need” surpasses to intended households the ability of the PSNP to respond. Prioritize the timeliness of transfers for all imple- 106 See Cabot-Venton et al. (2012) The Economics of Early Response menters. Despite progress across the highlands, and Disaster Resilience: Lessons from Kenya and Ethiopia, DFID; USAID, Economics of Resilience to Drought in Ethiopia, Kenya and PSNP transfers in the lowlands remain erratic, late, Somalia, (2018), among others. unpredictable, and irregular. While comparable data 118   Part 2: Promoting Resilience and Development in the Lowlands on the HFA are not available, monitoring of ‘rounds’ improves.109 Importantly, transparency can neu- of food assistance show that these are often deliv- tralize the ‘capture’ of resources at the local level. ered late, with some rounds ‘skipped’ due to short- PSNP and HFA actors in the lowlands should adopt falls in resources.107 Remembering that ‘a safety the same oversight, reporting, monitoring, and net delayed is a safety net denied’, the PSNP’s auditing standards. The execution of accountabil- primacy of transfers principle needs to ‘have teeth.’ ity functions should be managed as part of the There needs to be an inviolable rule regarding the system-­ based approach; in particular, food man- delivery of transfers, with local incentives for strong agement requires stronger accountability measures performance and accountability measures to that mirror those in place for cash transfers. encourage staff to meet targets. Increase human and physical capacity to Regular transfers must be delivered to the poor. improve performance. The numbers of local gov- The national program is a poverty-targeted pro- ernment staff in the lowlands are currently inade- gram, but the record of effective targeting in the quate and, more than in highland areas, key staff lowlands is particularly concerning. Despite multi- are often not in place, are heavily and frequently ple technocratic attempts to resolve this for over a rotated, and essential training is erratic. More than decade, there remains friction with local notions of in highland areas, to deliver support and monitor ‘fairness’ and ‘equity’, with communities indicating effectively, local authorities require reliable and that everyone should benefit regardless of wealth, frequent access to quality transport services, which and that targeting a minority of community mem- are currently missing. Addressing these issues bers can engender tensions.108 Senior leadership through policy change will be critical to the success discussions with customary institutions and other of a government-led system. structures are needed to ensure that the safety net system can reach the poor in line with agreed rules. Additional Measures to Respond to the While universal approaches are aspirational (and Causes of Poverty and Vulnerability an achievable policy goal in the long term), they are While social assistance is necessary for poor and currently unaffordable, and in the meantime pov- vulnerable households, it cannot respond to the erty targeting should be strictly enforced. range of causes of poverty and vulnerability alone. Economic diversification and investments in human C. Strengthen governance and accountability capital are needed to provide sustainable options and address capacity bottlenecks for populations in the lowlands to strengthen their Introduce one institutional structure for oversight resilience. These require dedicated strategies, with of initiatives to address poverty and vulnerability safety net professionals and technicians from other in the lowlands. Introducing one clear structure to relevant sectors (livestock, agriculture, labor, etc.) govern implementation of the system minimizes engaged in identifying how they can tailor support the ability of stakeholders to avoid accountability to safety net beneficiaries and work together to at local levels relating to performance, particularly facilitate households’ exit from poverty. by “projectizing” responses. The structures need to To complement safety net initiatives, poor be easy so they can follow a trail of accountability and vulnerable households should be exposed to from community to kebele to district to zone to a range of diversification strategies, including: region to federal levels and be representative of key livestock-based activities, non–livestock-based stakeholders, including customary institutions. activities, and facilitation to allow the uptake of Strengthen local accountability through employment opportunities (including migration). increased transparency. When local decision mak- Poor and vulnerable households require additional ers and officials in the lowlands are transparent, support to diversify activities away from traditional the governance and performance of the PSNP livelihoods. The government is well positioned to 107 As the HFA is not an entitlement, if the plan is to provide monthly support from January to June, for example, and the allocation for Jan- 109 For example, the association between being a kebele official and uary is severely delayed, it is common for this ‘round’ to be renamed being a PSNP beneficiary in Somali is higher when the beneficiary list support for February and for only five months of support to be pro- is not publicly posted; in kebeles where the list is posted, widows are vided rather than six as originally intended. 30 percentage points more likely to be selected; and where lists are 108 Lind et al. Targeting Social Transfers in Pastoralist Societies—­ publicly posted, rich households are 11.3 percentage points less likely to Ethiopia’s Productive Safety Net Programme Revisited, 2018, IFPRI. be selected for the PSNP than are destitute households. Chapter 8: Humanitarian Interventions87 and Safety Nets Working in Tandem   119 lead the development of relevant strategies that barriers to remunerative diversification being skills, would facilitate households’ diversification (even contacts, and capital. As part of the government-­ exit) from unviable, unsustainable livelihoods. Global led system, human capital (education and skill experience indicates that these strategies need development, particularly for women) should be to be tailored to poor and vulnerable households’ considered as critical building blocks toward diver- capacities and capabilities (they should not be sification away from unsustainable livelihoods. A generic strategies). Options exist for households strategy for strengthening human capital around with livestock to diversify into added value prod- the life cycle should be prioritized once the supply ucts (dairy production, hides and skins, etc.), in of relevant social services has evolved adequately addition to strengthening existing primary liveli- (see Chapter 3). hood activities, noting that local markets are often too thin to ensure viability without integration into “[E]ducation investments should be the central larger value chains. While returns to rain-fed crop- pillar of diversification and transformation ping have continued to be low and uncertain, the strategies in ASAL regions, with irrigation numbers of pastoral households now turning to schemes and other attempts to promote cropping has increased, partly from a lack of alter- diversification into sedentary farming a dis- natives. The supply and quality of skills and training tant second. The latter have some potential however remain low. Facilitation of employment to promote economic growth and create jobs, opportunities and migration should also be con- but that potential is limited by basic agroeco- sidered for poor and vulnerable households. While logical and agronomic factors, and by insti- entrepreneurial activities may be appropriate for tutional factors that appear to exclude ASAL some poor and vulnerable households, a significant populations from substantially benefitting proportion will not wish to risk their resources on from these schemes. Education, in contrast, entrepreneurial activities and prefer to gain food is a direct investment in the very young pop- security through employment. In most developed ulations found in ASAL areas, and one likely countries at least 80 percent of all people are of to generate assets and income sources (such this mindset, and poor and vulnerable households as remittances) that are far less vulnerable to are unlikely to be different. Government could covariate shocks, and even more mobile than therefore consider best practices in the areas of pastoralist livestock. Advances in female edu- employment facilitation as part of the system to cation will also reduce population growth (with build resilience. This may also be associated with a lag) and improve health and nutrition out- internal and international labor migration, which comes for children. . . . Hence if the right deliv- should be further considered as part of the govern- ery modalities can be developed, we expect ment system. that education investments will be a high There is a special need to address education return investment, albeit one that will require and skills development in lowland regions to better a decade or so to truly bear fruit.” (Headey support individuals to identify and secure adequate et al. 2014) incomes through wage-based employment. While wage-based diversification is increasing, 89 percent In the next chapter we examine opportunities of the households who left the pastoral system in for long-term sustainable development in the low- 2013 were unskilled and 82 percent were illiterate. lands through increased opportunities for economic Urban employment is generally informal and low- diversification. skilled, and dominated by women, with the entry 120   Part 2: Promoting Resilience and Development in the Lowlands Poverty and Vulnerability in the Ethiopian Lowlands Chapter 9: Economic Development Summary Agriculture Overview Economic development in the lowlands lags the high- Agriculture plays a key role in the economy of Ethi- lands. To fully participate in inclusive national and opia and, including the related value and supply regional growth the lowlands will need considerable chains, accounts for over 84 percent of employ- investment. At the forefront of any strategy is the ment and 53 percent of GDP.110 (See Table 9.1.) In need to take a long-term perspective that addresses the lowlands, agriculture is characterized by an service provision, connectivity, and human capital. At increasing scarcity of grazing land and driven by the same time support and investment are needed to the expansion of land used for irrigated commercial ensure that the economic opportunities available can agriculture, as well as the settlement of rangeland be maximized. Livestock will remain the main source of communities. livelihood for the pastoralist community, and there are potentially good returns from both the domestic and Table 9.1: Structure of agriculture—GDP and employment 2011 export livestock markets. But this will require sup- port to all aspects of market infrastructure, including GDP (%) Employment (%) veterinary services, cold chain logistics, border posts, Agriculture 42.1 79.1 and roads. The extractive sector will feature in many Agroprocessing 2.0 0.7 of the lowland regions over the next 10 years, and the Input production 1.3 0.4 sector could comprise up to 10 percent of Ethiopia’s Trade & transport 7.5 4.2 GDP within 20 years. But this will mainly benefit the TOTAL Agri food system 52.9 84.4 country through national foreign exchange earnings and royalties, with limited employment opportunity Source: IFPRI. and direct benefit to the lowlands. Increasing urbaniza- tion will lead to the growth in the service sector in the While the lowland areas are overwhelmingly lowlands, and with lower levels of poverty and vulnera- dependent on primary production, the balance bility in urban areas, rural to urban migration provides between livestock and cropping varies across a better opportunity than rural to rural migration for regions. Livestock is a key source of income in the poverty reduction and supporting growth. lowlands for most rural households (see Chapter 1 The most important action for the government will of this report). be to strengthen the overall human capital of the low- Maps 9.1 and 9.2 demonstrate the importance land population, with priorities given to education and and distribution of different types of livestock for access to improved services. Such support to human the lowlands—particularly for the arid drought- capital would increase the ability of lowlanders to take prone lowlands. new jobs and is part of a long-term transformational approach to address entrenched gender barriers and social norms. 110 IFPRI (2018) Diagnostic Resilience Building for Ethiopia. 121 Map 9.1: Models of agricultural production111 Map 9.2: Spatial distribution of livestock Figure 9.1: Growth in crop yields and population across highland and lowland regions Cereals Pulses Yield growth > population growth Yield growth > population growth 12 12 SNNP (Bench Maji) 10 SNNP (Bench Maji) 10 8 Amhara SNNP (South Omo) 6 Benishangul-Gumuz 8 Annual yield growth Annual yield growth 4 Tigray Gambella 2 6 Oromia (Borena) 0 Oromia (Guji) 4 Benishangul-Gumuz Amhara –2 Somali Oromia (Guji) Afar SNNP (South Omo) –4 2 –6 Oromia (Borena) –8 Yield growth > population growth 0 Yield growth > population growth 1.5 2.0 2.5 3.0 3.5 4.0 1.5 2.0 2.5 3.0 3.5 4.0 Annual population growth Annual population growth Source: CSA Agricultural Sample Surveys, 2004–2010 Ethiopian Calendar, CSA Population Projection of Ethiopia for all regions at woreda level from 2014–2017. A notable feature of lowland agriculture is the Figure 9.1 demonstrates this: for cereals, most of slowing growth in crop yields, most likely due to the the lowland regions and zones are below the yield/ adverse climatic conditions and the lasting impact population growth parity line, meaning that in the of extreme weather events on soil fertility and no-change scenario, the food security situation is water retention. Despite the fact that agricultural likely to gradually worsen over time. For pulses the productivity in the lowlands continues to increase outlook is better but only for those areas where over time, the rate of the increase is insufficient to production is possible—production volumes in Afar, compensate for population growth. the Somali region, and Gambella are negligible. In terms of livestock productivity, the analysis in the Livestock Master Plan sets out a continued 111 MOARD. (2010) Ministry of Agriculture and Rural Development, growth in feed requirements in the lowlands, which United States Agency for International Development; Government of Ethiopia. An atlas of Ethiopian livelihoods. is forecast to rise to a total of 56 M tons of dry 122   Part 2: Promoting Resilience and Development in the Lowlands matter by 2030. The existing pastureland will not approximately 90 percent of Ethiopia’s meat and be able to support this level of feed demand under animal exports, though the trade is largely informal. any climatic conditions.112 Prices have remained low due to quality concerns, and government attempts to impose a live animal Pastoralism and Livestock export price floor have been circumvented by a Despite the adverse climatic conditions for inten- booming informal and/or illegal cross-border trade, sive fodder production, extensive livestock herding estimated to be almost twice the value of formal is still a high-potential sector for lowlands pasto- exports (up to US$400 million per year). Taking ralists. Ethiopia is the 5th largest livestock pro- advantage of export opportunities requires meeting ducer in the world and thus has potential to claim challenging export quality requirements. Processed a share of the market opportunity created by an exports are minimal (~1 percent of total export rev- estimated 7 percent projected deficit in global beef enue), with only 14 export-oriented abattoirs across supply by 2024. Domestic demand for beef is also the whole country. expected to grow strongly in line with urbanization Considering the mismatch between domestic and increasing incomes. Already, the price differ- demand and supply (current higher domestic prices ential between domestic and export market prices for cattle) and challenges in meeting export quality, (c. ETB 150/­ kg and ETB 74/kg, respectively) is an the incentive for many lowlanders is to produce for indication of domestic supply shortages. The Ethi- the domestic market that offers a greater oppor- opian Government’s Livestock Master Plan (2015) tunity. Nevertheless, the export opportunity should estimates a deficit of 1.33 million tons of meat by not be ignored. Substantial government investment 2028 in the absence of productivity improvements. in livestock infrastructure will help increase produc- Taking advantage of these opportunities will tivity and earning potential. Only 45 percent115 of require substantial investment to improve carcass the country is served by public and private animal quality and feed efficiency, and to strengthen health delivery systems, with pastoralists relying market linkages. Out of a total national cattle on vaccination and treatment at reconditioning/ population of 60 million head, approximately feedlots which serve as pre-quarantine stations 30–40 percent are owned by lowland pastoralists for exports. Investment in disease management to and agropastoralists practicing extensive, low input contain the prevalence of transboundary diseases production systems. Smallholders and pastoralists will help reduce Ethiopia’s vulnerability to trade (with <200 cows) are the dominant livestock pro- bans, predominantly from Middle East countries. ducers, contributing to almost the entirety of the For lowland producers to benefit, substantial 0.39 million metric tons of beef.113 Chapter 1 in this government investment is required in livestock study highlights the increasing concentration of infrastructure, including investment in disease livestock ownership in the population. management to contain the prevalence of trans- Lowland pastoralist cattle producers capture boundary diseases, including but not limited to a relatively low share of the overall value of their Foot-And-Mouth and Rift Valley Fever. Ethiopia produce. They supply most of their live animals is vulnerable to trade bans, predominantly from to small traders earning margins ranging from MENA countries, and sanitary and phytosanitary ETB 7,300 to ETB 13,600 per head of cattle. Only (SPS) and the World Organization for Animal Health 35–40 percent of the terminal value of the beef requirements are becoming more stringent. sold in consumer markets is captured by the pro- Only 45 percent of the country is served by ducers (compared to about 50 percent in Kenya). public and private animal health delivery sys- There is scope for increased income for producers, tems, with pastoralists in particular relying on but this will require improvements to market link- vaccination and treatment at reconditioning/ ages and improved transportation links.114 feedlots which serve as pre-quarantine stations Export earnings from live cattle are signifi- for exports.116 There is an overreliance on Djibouti cant in the lowlands, which are responsible for as the intermediary marketing channel for the Middle East. However, Djibouti does not recog- nize Ethiopia’s quarantine procedures and thus 112 Livestock Master Plan 2015, p. 70. 113 Agricultural Transformation Agency (2018) Beef and Dairy Cluster Strategies. 115 Livestock Master Plan (2015) Roadmaps for Growth 114 USAID (2009) Agricultural Value Chain Financing in Kenya—­ and Transformation. Assessment of Potential Opportunities for Growth. 116 Ibid. Chapter 9: Economic Development   123 requires an additional 21-day quarantine prior to preference for low-input pastureland grazing. re-export, adding time and cost.117 China, Hong Additional challenges include the inherent low Kong, Vietnam, and Malaysia have been identified productivity of local breeds, with little progress in as additional strategic priorities by the Ethiopian improving the genetic stock, and poor access to government, based on existing trade relations and extension and animal health services. market size, although improvement in disease Support to all aspects of the value chain is management will be equally crucial.118 needed, including breeding, veterinary services, and access to markets. The western and ­ eastern Camels lowlands are particularly disadvantaged, and Camels will become an increasingly important approximately 70–90 percent of dairy production source of meat, milk, and transport given the cam- in Ethiopia is marketed through informal channels. el’s versatility and resilience to drought. In terms Consequently, Ethiopia is a net dairy importer, with of marketing potential, it has been estimated119 imports reaching US$14.6 million in 2015 to supply that the northern Ethiopia camel trade alone has the wealthier, quality-orientated urban markets. a sales volume of between US$18.5 million and Within the formal value chain, producers capture $24.5 million per year, surpassing the sales volume approximately 50–60 percent122 of the final retail of the reported formal meat exports. The contri- price. The informal chain may allow for a higher butions of camel to the national livestock GDP in value capture due to the lower costs of unlicensed terms of meat production is estimated to be ETB marketing intermediaries, although the absence 3.6 billion and is projected to grow. Camel milk pro- of rules and regulations creates its own problems duction, prevalent in the eastern lowlands, presents regarding the reliability of marketing opportuni- an underexplored opportunity. Even in conditions ties. Given the perishable nature of the product of fodder scarcity, camels can produce enough (c. 20–35 percent wastage rate),123 increased avail- nutrient-­ rich milk to sustain pastoralist households ability of cold storage or cold chain transportation for several months. is needed so that processing facilities do not neces- sarily need to be near urban centers. New private investment and/or public support is Dairy needed. Due to the climatic advantage and prox- Dairy production is a significant and growing mar- imity to urban markets, the most likely contenders ket in the lowlands, however its growth is hampered to capture this opportunity are the large-scale for- by market inefficiencies. The average consumption malized dairy processing operations located in the of milk in Ethiopia is 19 liters per person per year, highlands. compared to the African average of 40 liters. Dairy demand is strongly income and urban related: the Agroforestry top 10 percent of earners in the Addis Ababa market consume 38 percent of the milk, thus with increas- BG124 is a leading region in the country in terms ing urbanization, demand is expected to increase. of the area of land under private investment.125 In Over 60 percent of Ethiopia’s annual production addition to commonly grown cash crops such as of 3.1 million metric tons of milk are produced in the sesame, maize, and oil seeds, which represent most lowland areas.120 Milk yields are low at 1.5–1.9 liters of the commercial farming activity, bamboo pre- of milk per day due to a range of challenges, most sents a high potential commercial opportunity. notably the availability of quality feed121 and the Investor interest has been registered for pro- cessed bamboo products, such as pulp, paper, fibers, plywood, and wood pellets, driven by: 117 USAID Feed the Future (2017) Value Chain Analysis: Live Animals (i) plentiful availability of natural resource with over and Meat. 750,000 ha of bamboo in the region; (ii) physical 118 Agricultural Transformation Agency (2017) Ethiopian National FMD Control Strategy and Program Design—Final Design Document. 119 Aklilu Y. (2009). Livestock and Animal Product Trade and Their Impor- 122 Agricultural Transformation Agency (2018) Beef and Dairy Cluster tance for Pastoral Economic Growth in Ethiopia. Strategies. 120 Agricultural Transformation Agency (2018) Beef and Dairy Cluster 123 SNV Ethiopia and Target Business Consultants Plc (2017), Inventory Strategies. of Dairy Policy, Ethiopia. 121 Livestock Master Plan, Roadmaps for Growth and Transformation 124 See Regional Report Annex on BG. (2015); USAID AGP-Livestock Market Development Project (2013), 125 The Federal Ministry of Agriculture (MOARD) (2013). Agricultural Value Chain Analysis for Ethiopia: Meat and Live Animals, Hides, Skins Investment Land Handed to Investors, Excel File (Data Set); MoA: Addis and Leather, Dairy. Ababa, Ethiopia. 124   Part 2: Promoting Resilience and Development in the Lowlands attributes of lowland bamboo that make crush- finance by smallholders by advancing inputs and/ ing it both technically simpler and commercially or credit. viable; and (iii) potentially attractive markets with Investment in market infrastructure is required rising prices for pulp, e.g., due to the closure of in all aspects of the marketing chain but especially paper mills in China following tighter environmental post-harvest handling, as 25 percent of output is regulations. lost in the “first mile,”129 and quality is undermined The benefits to the local population can include as a result. the sale of bamboo to investors/processing fac- tories. According to initial estimates, 40–50 per- Extractive Sector cent of bamboo would have to be sourced among The extractive sector will become a significant smallholders outside of the investor-owned plan- feature of the Ethiopian economy within the next tations,126 and there would be employment for 20 years, and most of the resources130 are in the harvesting and post-harvest treatment as well. lowlands. The sector’s contribution will increase to The cost calculations for transporting the product 10 percent of GDP from its current level of 5.6 per- to the nearest port create a strong argument for cent. Given the capital intensity of the sector, the investing in bamboo pulp processing facilities close benefits to the lowland regions will probably mainly to the source of the raw material to reduce trans- come from higher overall government revenues and port costs. This would also create more jobs and the increased scope for regional fiscal transfers capture more value for the country. rather than direct employment. Good governance is a critical precondition to maximize the economic Spices and social benefits of extractives. Unquestionably, Spices are a historically significant commodity this will bring demands for the sharing of benefits in Ethiopia, with the lowlands playing an import- to the region in terms of employment, services, and ant role in their cultivation (smallholders make up livelihood opportunities. 98 percent of total production). The domestic mar- Most of the resources are in the lowlands, includ- ket is reliably strong since spices play an important ing gold (Guji and West Wellega zones in Oromia), role in Ethiopian cooking and over 90 percent of tantalum (Guji zone, Oromia), gemstones (Oromia spices are consumed domestically. and Amhara), oil, natural gas (Somali), potash There is considerable potential for the lowlands (Afar), and salt (Somali and Afar) as set out in (especially in SNNP, Oromia, and Afar) to play an Table 9.2. A key constraint is the ability of firms to increasing role in the US$16 billion world spice mar- prospect for minerals and the willingness of firms ket. For example, SNNP is the leading region for to invest. ginger (~80 percent of national production) and tur- Given the 15–20 years that will be needed for the meric (~75 percent of national production), both of mineral sector in Ethiopia to generate critical mass, which are exportable commodities, as well as black the economic potential from the sector is clearly cardamom.127 Ethiopian turmeric is of particularly difficult to predict. However, based on the Strategic high quality. Assessment of the Ethiopian Mining Sector in 2014, Supply chains tend to be long (with up to six three economic scenarios (conservative, probable, intermediaries between the farmer and the end and possible) have been developed to present the consumer), and the correspondingly high trans- potential based on exploitation of gold, potash, tan- action costs serve to push up prices. Emerging talum, and copper reserves. In a probable scenario, exporting companies offer the potential of outgrow the mineral sector could generate export revenues models, such as those operated by export com- of up to US$1.5 billion by 2024 from the sale of panies.128 The advantage of an outgrow model is gold, copper, tantalum, and potash. that the processor/exporter provides the key role of Figure 9.2 shows the potential revenues from aggregation and can address the lack of access to the projects listed above. The sharp rise in revenue in 2023 is mainly linked to start-up production of 126 Interview with McKinsey, DFID consultancy under EIAF on Industrial parks, January 2019. 129Ibid. 127 Herms S. (2015) Investment Opportunities in the Ethiopian Spices 130Minerals found in Ethiopia include gold, platinum, niobium, tan- Sub-Sector, Advance Consulting. talum, nickel, copper, chrome, manganese, limestone, sandstone, 128 YSO (Yahia Sayed Omar) YSO Import-Export and ESEF (Ethiopian gypsum, clay, lignite, opal, oil shale, laterite iron ore, bentonite, clay, Spice Extraction Factory). perlite, diatomite, potash, halite, and oil and gas. Chapter 9: Economic Development   125 Table 9.2: Major exploration projects in lowland areas of Ethiopia Type of mineral Region Locality and zone Precious, base, and rare metals BG Baruda-Bulen, Ablarus, Bahu-Anjakoya, Yabanja and West Tangoy localities; Metekel zone Gold and base metals BG Shungu and Nazali Gold and base metals BG Sherkole locality; Asossa zone Gold and base metals BG Oda Godere locality; Mengi and Oda Godere woredas; Asossa and Kamashi zones Gold and base metals BG Asossa and Kurmuk Gold and associated minerals SNNP Dizi, Sheka, and Surma woredas Gold and associated minerals Oromia Haramfama and GenaleKorcha, Meda Welabu, Adolana, Oddo Shakiso woredas, Bale and Guji zones Gold and base metals Oromia Okote, Borena, and Guji zones Gold Somali Haramsam and Hasamite localities; Liben zone Gold, silver, and base metals Afar Tendaho Precious and base metals Afar Gabala and Gira localities Potash Afar Musley and Crescent localities Potash Afar Bada Potash Afar Dallol Natural gas Somali Ogaden Source: Ministry of Mines. Figure 9.2: Contribution of extractive sector exports 2013–35 (ETB M) scenarios 3,000 2,500 2,000 1,500 1,000 500 0 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2024 Conservative Probable Possible Source: Ministry of Mines. new gold projects, while the fall in revenue in 2030 line with the projections of GTP II, the government is linked to depletion of resources at the Midroc has set a high growth target of 5 million interna- LegeDembi Mine. As new production comes on line tional visitors in 2025 with revenues of ETB 180 bil- for gold in 2033, revenue picks up again. lion in 2025.131 The scale of the potential opportunity for the Tourism lowlands is not significant (including the Danakil Tourism makes substantial contributions by creat- Depression, Great Rift Valley Lakes, protected ing considerable employment and creating demand from micro, small-scale, and medium-scale enter- 131Ministry of Culture and Tourism (2015). Sustainable Tourism Master prises for the production of goods and services. In Plan 2015–2025. 126   Part 2: Promoting Resilience and Development in the Lowlands wildlife areas, etc.). Attracting tourists will require Governance and Stability investment in tourist infrastructure, marketing, For the growth in these sectors to materialize and and stability, as well as stronger local engagement, markets to be reached, the key economic growth legal boundaries, and management plans to ensure drivers need to work for the benefit of the lowland that people in the lowlands benefit. population. This growth will be critically dependent Further research is needed on tourist-linked on good governance and stability. The Growth and employment potential for the lowlands. The evi- Transfromation Plan II (GTP II) of the Ethiopian Gov- dence from Afar confirms that a significant amount ernment highlights the importance of improving of tourism development in Ethiopia is conducted capacity at all levels of government administration with little engagement by or for local people; this is and addresses: “malpractices and governance prob- likely to link to low levels of human capital. lems are still prevalent in land use and administration, justice service, and public service deliveries in Govern- Key Drivers of Growth ment institutions.” Weak governance is a deterrent The framework in Figure 9.3 shows the key eco- to investors; investors will choose areas with a nomic growth sectors that are of significance to capable, engaged local administration. the lowlands. The four key drivers of economic growth132 described below were identified as the Infrastructure key determinants of growth for the lowland periph- Improving connectivity through improved infra- eries. This growth will be critically dependent on the structure and access to markets is critical. The underlying political and economic conditions and lowland population has low levels of access to requires specific policies to enable the growth gen- all-weather paved roads. Excluding Oromia and erated by the markets to be inclusive and benefit SNNP where the regional-level picture is likely to the lowland population and pastoralists. be skewed by better-developed highland areas, the Government policies strengthening the enabling lowlands are at a clear disadvantage: per 1,000 km, environment will be a precondition for successful road density in Afar (12 km), BG (17 km), Gam- engagement of the growth sectors with the eco- bella (16 km), and Somali (9 km) is significantly nomic growth drivers. lower than in Amhara (35 km) or Tigray (32 km).133 Figure 9.3: Sectors and drivers for economic growth in the lowlands Growth sectors Key economic growth drivers Markets Agri/ Industrial Agroforestry Urbanization Urban centers Rural markets livestock parks Growth Local cross- Extractives Manufacturing Cross-border Major export corridors boundary trade markets markets Tourism Other services Enabling environment Governance Access to finance decentralization, External stability Infrastructure Social inclusion privatization FDI 132Figure 9.3 is a schematic representation. The identification and 133Ethiopian Road Authority estimates in “Existing Situation and selection of the key drivers were discussed with MOFEC and DFID in Diagnostic Final Report, 2015” FDRE Ministry of Urban Development, December 2018. Housing and Construction. Chapter 9: Economic Development   127 Crucially, most of the existing road network is (v) corporate and environmentally responsible poli- unpaved—in higher rainfall regions of BG and cies and incentives are in place. Gambella this means that entire communities can ­ be cut off during the rainy seasons. Access to Finance Beyond roads, arguably one of the most signifi- Limited access to finance is a key barrier to cant infrastructure gaps is the shortage of border economic growth, limiting start-up capital and posts that would allow formalized cross-border business expansion. There are 35 Micro Finance trade, especially in Somali and Afar regions. Only Institutions (MFIs) of which the five largest are seven posts have been set up across the entire state owned. The state-owned MFIs are largely length of the Somali region border with Somalia/ restricted to their own regional base, and compe- Kenya, namely at Awbere, Tug Wajale, Hartishek, tition in the industry is very limited. They are also Gashamo, Ferfer, Dollo Ado, and Moyale. Currently, especially influenced by their respective regional the only open posts are Moyale and Tug Wajale. governments, which virtually own the MFI and provide below market-rate funds. There are major Market Development regional variations with stronger coverage in Somali Proactive government engagement is needed to than in Afar. support the development of market infrastructure Overall, financial inclusion is low in Ethiopia and improvements in productivity in some key compared to the rest of East Africa, particularly value chains, notably livestock, but also to facilitate in the lowland areas. Figure 9.4 shows the skewed the aggregation and development of the market for regional distribution of access to commercial or other value chains, notably spices. MFI bank accounts (and, inversely, of the unbanked population) between the highlands and the low- Privatization will play a key role in market efficiency lands (the latter highlighted in bold). The share of There is a long legacy of state ownership in Ethio- population with access to a bank account is over pia. State firms usually operate in a monopolistic 20 percent for women and 30 percent for men in environment, limiting the incentive to innovate and Addis Ababa, Tigray, and Amhara. By contrast, in resulting in higher prices in some sectors (e.g., tele- Somali, Oromia, SNNP, Benishangul-Gumuz (BG), coms and logistics) but lower prices in others (e.g., and Afar, it is below 10 percent for women, and energy). Nevertheless, the government has shown 20 percent for men.134 Although gender disparities a commitment to private investment by attract- exist throughout the country, the relative disadvan- ing Foreign Direct Investment (FDI), particularly tages for women are even more marked in some of in manufacturing businesses in textiles, apparel, the lowland areas, notably BG, Afar, and Gambella. pharmaceuticals, and tanneries. Outside of the A significant portion of the population in the low- industrial parks and apart from petty trading, the lands use Iqub, a rotating saving and credit associ- concept of private-led commercialization is still ation where each member agrees to regularly pay a underdeveloped, and the state is seen as the main small sum into a common pool so that each person, actor for economic development. in rotation, can receive one large payout. These are Privatization and increased competition, along often formed by people sharing the same employer, with continued foreign investment, could be major trade association, or business, or members of a game changers for the lowlands. In particular, community. These ‘clubs’ are unregulated, and the opening up the telecoms (especially if accompanied leader is in control of any net surplus of savings by financial sector reforms) and transport logistics over loan disbursements, which is rarely placed in a sectors could have a major impact on lowland areas dedicated account when it is banked. in addressing the critical constraints of access A recent study135 by the Association of Ethio- to finance and high transport costs. But this pri- pian Micro Finance Institutions (AEMFI) indicates vatization process will need to ensure that: (i) the that substantial cash holdings can be captured regulatory framework is transparent and encour- by Financial Institutions (FIs) with the correct ages competition on a level playing field; (ii) inves- tors have some form of political risk insurance; 134 DHS (2016). (iii) local communities are engaged; (iv) there is an 135AEMFI Occasional Paper by Storrow/Gobezie/Figge (2015) ‘Overview emphasis on local job creation through backward of Practical Challenges in Local Saving Mobilization by Ethiopian Micro­ linkages or value addition through processing; and finance Institutions’. 128   Part 2: Promoting Resilience and Development in the Lowlands Figure 9.4: Access to a bank account 2016 Addis Ababa Tigray Harari Oromia Amhara Gambella SNNP Dire Dawa Somali Benishangul-Gumuz Afar 0 10 20 30 40 50 60 70 Women Men Source: DHS. ­ arketing techniques. This is important because m between economic advancement and the provision the culture of holding cash savings at home, or of new bank branches or increased mobile financial buried outside for security, is reducing Ethiopia’s services that follow the intermediary principle of liquidity. This cash is currently ‘dead’ in terms of leveraging local savings to support local lending. potential for investment. Unless there is a concerted effort to ensure The National Financial Inclusion Strategy financial institutions expand their networks, (NFIS)136 seeking, inter alia, to develop strategies to increase their footprint in the lowlands, and develop capture these ‘lost’ savings notes that there are few new instruments to cope with pastoralists’ mobility, financial services access points in Afar, BG, Somali, access to finance will remain a key constraint to and Gambella regions, with only around 80 bank smallholders expanding their herds. branches to serve a population of 8 million. It also Any support to financial inclusion will also require notes that there is no detailed information about support to financial literacy and entrepreneurship. the location of financial service access points, and The evidence from NGO and government initia- this hinders the outreach planning of financial tives are that the most successful interventions institutions. are those that combine economic measures (e.g., Access to financing for pastoralists is complicated access to finance, market information, and link- by a general preference for and trust in holding sav- ages) with social measures (e.g., training, education, ings in livestock, which is thought to be more advan- and work to address discriminatory attitudes). tageous than savings in banks. However, even if this Some models of mobile money have potential, barrier could be overcome, banks in Ethiopia have no particularly for pastoralists who struggle to reach legal framework to provide loans for pastoralists as static banking facilities. Restrictions within the they do not accept livestock as collateral. As far as financial sector are preventing wider uptake, and pastoralists are concerned, livestock is their main so initiatives on agent-based mobile banking such asset; therefore, banks need to come up with innova- as BelCash and HelloCash cannot operate at scale tive loan procedures to accommodate pastoralists. until the licensing of service providers is relaxed The lowlands in general are in need of an and/or the telecom sector liberalized. New creative increased financial services footprint. Completion instruments to support access to working capital of Geographic Information Systems (GIS) access financing for pastoralists need to be piloted. points mapping would facilitate planning for improvement. In general, there is a direct correlation Urbanization Ethiopia is urbanizing very rapidly from a low base, 136 National Financial Inclusion Strategy (2017). and the pace of urbanization is estimated to have Chapter 9: Economic Development   129 been rapidly increasing in recent years. Urbaniza- productivity. Hence, while the opportunity for urban tion is particularly low, however, in the lowlands manufacturing at scale in many lowland urban but is beginning to pick up pace. By 2025 the centers may be limited, there is significant scope urban population in Ethiopia is expected to reach for greater backward linkages from lowland areas 27–30 percent of the total population, and by 2050 to urban manufacturing hubs from improved trans- to rise to 38 percent of the population (86 mil- port connectivity. lion out of total projected national population of 226 million), or over three times the current level. Growth Corridors In the lowlands the population is projected to rise The development of transport corridors that link to 35 million by 2050 of which 11 million will live in the key centers of economic activity in the country urban areas.137 to major ports is an important goal in the govern- Urban centers and the largest cities are con- ment’s GTP II strategy. The economic development centrated in the highlands. The only city of any corridor (EDC) framework approach views trans- size in the lowlands is Dire Dawa (269,000). Any port corridors not only as a means of transporting substantial urban population growth is likely to be goods, services, and people, but also as a tool for associated with those centers located on the main stimulating social and economic transformation in transport, economic corridors, and trade routes, the wider region. They do this by: (i) concentrating e.g., transport connections to the ports of Djibouti transport and industrial infrastructure investment and Berbera linking Dire Dawa to the highland cities in a defined geographic area; (ii) removing non- of Harar and Jijiga. Other ‘oasis city’ growth centers physical barriers to the flow of goods, services, and are projected for Gode in the Somali Region and people; (iii) promoting integrated governance and Semera in Afar. management of trans-boundary resources; and There is a strong correlation between the level (iv) using the corridor as a concept to ‘crowd-in’ pri- of urbanization and economic value added. Effec- vate sector investment’. tive, long-term economic development will involve The transformation of a transport corridor to an (i) movement of labor from low productivity agricul- economic development corridor follows four stages, tural jobs to high productivity manufacturing jobs given careful planning, political commitment, and and services mainly in urban areas, and (ii) broader stakeholder involvement: urban economic agglomeration benefits, which boost incomes and grow investment capital. Manu- Stage 1: Basic transport corridor: characterized facturing is unlikely to play a significant role in the by infrastructure (e.g., a road, rail, or waterway) lowlands in the short to medium term, but back- physically linking two points of economic impor- ward linkages and opportunities for labor migration tance across a national border. will be important. Stage 2: Multimodal transport corridor: char- Improved road infrastructure, rural-to-urban acterized by the integration of modes of trans- migration, and secondary city and market town port (e.g., combined road and rail) and the limited development will all assist in the intensification of provision of logistics facilities, such as dry ports urban development. Urbanization, in turn, will assist or warehousing along the route designed to facili- with a reduction in poverty in both urban and rural tate the movement of traded goods. areas. Stage 3: Functional logistics corridor: char- The way that cities interact with their hinter- acterized by a clearly defined institutional lands is critical in driving more inclusive growth. arrangement, such as a committee of the elected Cities act as a stimulus to nearby agricultural representatives of cities along the corridor or development, providing ready access to local and a special purpose vehicle which oversees and/ wider markets. Processing industries that draw on or implements the corridor’s regulations and and promote local agricultural outputs often follow. cross-border trade agreements; and plans for the Efficient farming that feeds the growing urban pop- establishment and operation of related services, ulation and responds to changing tastes for higher such as storage, warehousing, trucking, insur- value produce can stimulate improved agricultural ance, and freight management. Stage 4: Economic development corridor (EDC): 137Project based on Central Statistical Agency Population Projection of characterized by multimodal transportation Ethiopia for All Regions at Woreda Level from 2014–2017. infrastructure, the attraction of domestic 130   Part 2: Promoting Resilience and Development in the Lowlands Map 9.3: Transport corridors for the Ethiopian lowlands in the IGAD region Source: IGAD Regional Infrastructure Master Plan, 2018. investment and FDI, and the generation of is required. As cargo shifts from the road to the economic activity and benefits in surrounding Addis-Djibouti railway line, the transport corridor regions through a clearly defined spatial develop- may no longer require the same level of supplemen- ment initiative (SDI) focused on proactive urban- tary services, and the lowland “truck stop” economy ization and economic growth. suffers.138 A balance is needed between reducing transit time and costs through these corridors to The right enabling environment is essential for promote overall trade and growth and create local a transport corridor to become an EDC. At pres- economic opportunities. ent, more than 90 percent of Ethiopia’s trade The Djibouti Corridor has the highest poten- flows through the port of Djibouti, and the Djibouti tial, as it reaches the greatest number of lowland Corridor has benefited from the most significant residents, includes the emerging urban cluster of investment. The government has ambitious plans Dire Dawa-Harar-Jijiga, and will continue to be the to develop transport corridors to reduce reliance on primary corridor for Ethiopian trade. There is signif- the port of Djibouti, including the Berbera Corridor, icant untapped potential to increase trade between LAPSSET Corridor, Port Sudan Corridor, Massawa/ the western lowlands and South Sudan. Assab Corridor, and Mogadishu Corridor. These cor- The Berbera Corridor has been given high priority ridors are illustrated in Map 9.3. by the government, given its 19 percent stake in the Potential development corridors will traverse the port, and as such is likely to receive future invest- lowland regions facilitated by transport connec- ments. New access to Eritrea’s ports in Massawa tions and have the potential for a positive impact and Assab could lead to greater opportunities along on lowland economies if two conditions are met. that corridor, though large freight volumes are only Firstly, intervention in ‘last-mile’ connectivity and likely once the ports have been upgraded; short- warehousing and logistics are necessary to connect term opportunities exist in burgeoning cross-­ border remote areas to the primary corridor infrastructure trade in items scarce in Eritrea. The LAPSSET and to enable linkages between agrarian producers in peripheral areas to market their produce in the 138 Centre for Rural Development (SLE) Berlin (2017), Ethiopia’s Arid highlands and internationally. Secondly, manage- and Semi-Arid Lowlands: Towards Inclusive and Sustainable Rural ment of some unintended negative consequences Transformation, SLE DISCUSSION PAPER 03/2017–E. Chapter 9: Economic Development   131 Corridor has significantly reduced travel time has introduced the Simplified Trade Regime, i.e., between Addis and Nairobi, providing new opportu- a simplified customs clearance procedure for its nities for bilateral trade. This route passes through member states for goods originating from the the emerging urban cluster around Awassa, which COMESA region. Small-scale cross-border traders will boost the prospects for an economic develop- benefit from completing simplified customs docu- ment corridor in this region with potential knock-on ments and simplified COMESA certificates of origin. benefits for the SNNP lowlands. The long-term IGAD142 member states adopted a policy frame- impact of the corridor will be reliant on the success work on informal cross-border trade that is aimed of the planned port in Lamu and the ability of the at improving cross-border security governance and Kenyan government to secure investment for its seeks to achieve a common understanding and development, as well as associated infrastructure. attitudinal shift regarding informal cross-­ border trade, as well as the standardization of IGAD mem- Cross-Border Trade ber state’s definitions of informal cross-border Cross-border trade has existed for centuries and trade. This is yet to be implemented. plays a crucial economic, political, and social role More border crossing points would generally open for pastoralists in the Horn of Africa. This is largely the opportunity for many areas of the lowlands, and an unrecorded activity but is a key driver of poten- of critical importance is the adoption of an inte- tial growth for the lowland regions. Informal trade grated border management strategy. The absence of is more important than formal trade along the formal and regulated border crossing infrastructure Somaliland border with Ethiopia, with an estimated has worked to the detriment of pro-poor growth in value between US$200–300 million, which is four to the lowlands, especially in relation to the livestock five times the formal trade.139 Livestock is a signifi- sector, whereby a buoyant illegal and informal trade cant element of cross-border trade with Somaliland. in the South and East has enabled rents to be cap- The income from cross-border trade can be an tured by a minority of powerful traders. Formalizing important source of savings and capital for start- cross-border trade with improved transport and bor- ups and can strengthen alternative livelihood activ- der infrastructure could generate more revenue at ities.140 Cross-border trade creates employment the federal level and may open up opportunities for and business opportunities in related areas, e.g., many lowland areas, including widening the opportu- pharmaceuticals for livestock, restaurants and nity to livestock producers in particular. bars, commodity sellers, chat traders, and holding ground and loading ramp facility providers.141 Industrial Parks (IP) The prospect of formalizing cross-border trade The development of industrial parks is central to should be pro-poor but faces a number of chal- the government’s policy of transitioning from an lenges. Currently, the benefits from informal trade agro-based economy to an industrial one. Ethi- are captured by elites, and although there has been opia’s industrialization strategy recognizes that an increased volume of recorded exports from the state intervention is necessary to support the region (e.g., Somaliland, Djibouti, and Ethiopia), development of new value chains for both export no substantial improvement is evident in the live- and import substitution. Neither the cement sector lihoods of poorer pastoralists. The current policy (import substitution) nor floriculture (export) would favors formalizing larger scale livestock exports to have grown without strong state support. international markets, a process from which poorer The IP policy will stimulate the demand for women pastoralists are largely excluded. employees who comprise the majority of employ- Trade facilitation is limited, largely due to unco- ees in the textile, leather, and agroprocessing sec- ordinated border management practices. COMESA tors. GTP II envisages creating 750,000 jobs in the medium- and large-scale manufacturing industries 139 by 2020, with 60 percent of the low and medium Ethiopia Ministry of Agriculture (2012): Ethiopia Country Program- ming Paper to End Drought Emergencies in the Horn of Africa, Ethiopia; skilled, and 30 percent of the high skilled jobs Catley, et al., 2011 cited in MoA 2012. expected to be held by women. Low levels of human 140 The Little study (2010) found about 25 percent of livestock traders in the Ethiopia/Somalia/Kenya cross-border trade were involved in sell- ing staple foods, most of which were unofficially imported from Soma- 142 IGAD (2018) Policy Framework on the Informal Cross-Border Trade & lia and purchased with revenues from the livestock trade. Cross-Border Security Governance Nexus; Enhancing Cross-Border Coop- 141 Yacob A., and Catley A. (2010): Pastoral Livestock Commercialization eration and Cross-Border Economic Exchanges in the IGAD Region, V. 17, in Ethiopia: Driving economic growth or driving destitution? January 2018. 132   Part 2: Promoting Resilience and Development in the Lowlands capital are a constraint to labor mobility and full Focusing on addressing constraints to inte- participation in the labor market for lowlanders. The grating lowland resources into manufacturing low penetration of Technical and Vocational Educa- and agro-processing would create jobs and raise tion Training (TVET) in the lowlands could be consid- incomes, as well as boost productivity through ered to be a challenge, but quality of TVET education exposure to export markets. Key to this is continu- remains persistently low, and thus few factories see ing to improve human capital that would empower TVET education as a condition for employment. more people to take advantage of new job opportu- It is unlikely that industrial parks will be a sil- nities being created throughout the country and in ver bullet for the lowlands. Apart from Dire Dawa, the lowlands, and to improve connectivity to urban industrial parks are not generally being developed in centers and export markets. Other policy changes lowland areas. Investors will invest where they have may incentivize FDIs to source more inputs from suitable labor supply, water, electricity, connectivity, local inputs in a way that does not undermine their conducive climatic conditions (to save cost on air global competitiveness. conditioning), and good governance. The lowlands are sparsely populated and often poorly connected. Enabling Social Inclusion Huge investment would be required, and geographic Ethiopia’s recent growth trajectory has been rapid, isolation and climate may deter investors in any but unless the next phase of growth adopts a trans- case. There is an initiative by the government to formative approach to addressing gender and social promote integrated agro-industrial parks (IAIPs) inclusion, there is a risk of leaving a large proportion in Semera (Afar), and the opportunity around the of the population in the lowlands behind. There is bamboo sector in BG has been highlighted. also an economic imperative, since inequalities and Improving general connectivity without building social norms may hinder the ability of parts of the industrial parks would be a more cost-effective population to engage in formal economic activities. way of encouraging investors. Two other key fea- Recent IMF analysis suggests that eliminating gen- tures apply to the lowlands: (i) the need for detailed der gaps in both educational attainment and the research on the market opportunity of different rate of formal employment could increase Ethiopian value chains, and (ii) to apply a process of adapta- GDP over time by over 24 percent. tion and learning based on the successful experi- A transformative approach is required that ence of the cement and flower industries. The key seeks to change gender norms, attitudes toward message here is that the needs of the potential disability, and the imbalance of power between investor should be the focus of government support men and women and combines social interventions and facilitation. alongside economic interventions (Figure 9.5). This Figure 9.5: Human capital and empowerment Low literacy of Transformation Improved women, poor literacy: education education of of girls and girls women Patriarchal Early marriage, Slow erosion of culture—boy patriarchal Delayed marriage, FGM, women increased control preference property of culture: changing social norms over fertility family Barriers to Women’s economic economic inclusion empowerment Unfair burden of Lack of access/ Resources Contributions to work, domestic control over devoted to family: household income, and productive assets limits improved health increased voice duties access to and education and role in finance decision making Chapter 9: Economic Development   133 is a long-term process to break the social norms to veterinary services, access to irrigation, relevant underpinning the barriers to inclusion. In the long finance and insurance instruments, and health and term the most fundamental routes to economic education services. and social inclusion are to address geographic and While highland areas have experienced faster gender inequalities in education and provide access and more diverse growth, this study has shown to training. that there is considerable growth potential in the Affirmative action to target girls, women, and lowlands, building on the comparative advantage of people with disabilities is a necessary component the area, in terms of livestock and niche agricultural in education, training, access to agricultural and products like spices, bamboo, and extractives, as livestock inputs, and finance to support small busi- well as exploiting the location of these areas along nesses. Similarly, cooperative approaches have important trade and logistical routes to neighboring been shown to be successful in both supporting countries. Much of the economic disadvantage of small business and providing access to finance the lowlands is related to poor infrastructure, poor and building the stepping-stone from participation transport access to the region, and high transport in Village Savings and Loans Associations (VSLA) costs. Improving connectivity through improved through rural SACCOs to MFIs. Women are often infrastructure and access to markets is critical, and unable to approach an MFI because of a lack of introducing greater competition in the freight and property or land to offer as collateral, the short logistics sector is central to lowering costs. Under- time frame within which they may be required to standing and reducing barriers to economic migra- repay a loan, or for cultural reasons. Therefore, this tion to take new jobs (e.g., in industrial parks) and gradual step of combining social measures with develop the financial system to facilitate cheap, economic ones is needed. efficient remittance flows would also improve inte- gration with the rest of Ethiopia. Conclusions There are some new critical sources of growth Historically, the lowlands have been peripheral143 in the lowlands. The clearest growth opportunity is to the Ethiopian state: geographically, politically, extractives, though local employment opportunities and increasingly economically isolated from mar- are likely to be limited and there are well-known kets, finance, and industrialization opportunities. risks to the broader tradeable sector from the The government needs to take a strategy forward effects of Dutch disease.145 The main benefit from for the lowlands of Ethiopia, which has a long-term extractives to lowland areas would therefore be perspective and recognizes the diversity of oppor- how revenues are redistributed to lowland areas. tunity and challenges in the area. At the forefront Other investment opportunities similarly would of any strategy should be the need to build the exploit lowland resources (e.g., bamboo). The simple resilience of the lowland population. message on financial inclusion is that all financial Pastoralism is not a residual livelihood system institutions need a greater footprint in the lowlands that needs to be phased out, which, at times, is the to address the very low levels of financial inclusion tone of the GTP II.144 Instead, it needs to be seen as and the increasing gender gap between men and a key feature of the lowland economy with enor- women. Poor access to financing and low levels of mous economic potential, as well as the livelihood financial literacy were common themes from this of choice for a large proportion of the population. study’s regional consultation. Mobility is part of this livelihood and not just a Proactive government engagement will be critical short-term coping strategy. Pastoralism will con- to support the development of market infrastruc- tinue to be the main livelihood for large parts of the ture and improvements in productivity in some key lowlands for the foreseeable future. It follows that a value chains, notably livestock, but also to facilitate range of policies to support the pastoralist commu- the aggregation and development of the market nity are needed rather than policies to settle them, for other value chains, notably spices. On livestock including access to markets for livestock, access a first step would include range management, the development of cold chain storage, additional border posts, and veterinary and phytosanitary 143 Markakis (2011) Ethiopia The Last Two Frontiers. services. 144 Looking forward, further emphasis will be given to ensure the ben- efits to the pastoralists building on the achievements gained so far and through enhancing voluntary sedentary farming (crop farming) 145 The negative impact on an economy of sharp inflows of foreign practices. currency. 134   Part 2: Promoting Resilience and Development in the Lowlands There are considerable opportunities identified Improved access to social services, particularly from the development of growth corridors. The education and health, would empower more of the absence of a formal and regulated border-crossing population to take advantage of economic oppor- infrastructure has worked to the detriment of pro- tunities in the lowlands but also elsewhere in the poor growth in the lowlands, especially in relation to country. In addition, the cultural change of moving the livestock sector whereby a buoyant illegal and from a rural nomadic experience to an urbanized, informal trade in the South and East has enabled industrialized lifestyle and work pattern will require rents to be captured by a minority of powerful trad- long-term support. Addressing inclusion and gender ers. Formalizing cross-border trade with improved equality is a slow process and will require time to transport and border infrastructure opens up change social norms. A transformative approach, opportunities for many lowland areas and could cognizant of the cultural differences on the ground, widen the opportunities for livestock producers. is required to surmount the barriers to economic Underpinning the economic potential of the inclusion in the lowlands. This will require significant lowlands is the need to develop its human capital. investment and long-term planning. Chapter 9: Economic Development   135 Poverty and Vulnerability in the Ethiopian Lowlands References Abate, T., A. Ebro, and L. Nigatu. 2010. Traditional rangeland Barrett, C. B., and Constas, M. A. 2014. “Toward a Theory of resource utilisation practices and pastoralists’ perceptions Resilience for International Development Applications,” on land degradation in south-east Ethiopia. Tropical PNAS, Vol. 11, no. 40, pp. 14625–14630. Grasslands 44: 202–212. Berhane, G., Hirvonen, K., and Hoddinott, J. 2016. The Adugna, E., and M. Sileshi. 2013. “Double Marginalized Implementation of the Productive Safety Nets Programme, Livelihoods: Invisible Gender Inequality in Pastoral Societies” 2014: Highlands Outcomes Report (2015). Addis Ababa: Societies, 3, 104–116; doi:10.3390/soc3010104. file:///C:/ Ethiopia Strategy Support Program, International Food Users/wb17224/Downloads/Double_Marginalized_ Policy Research Institute. Livelihoods_Invisible_Gender_I.pdf Berhane, G., Hirvonen, K., Hoddinott, J., Kuman, N., Agricultural Transformation Agency. 2017. Ethiopian National Tafesse, A. S., . . . Strickland, A. 2015. The Implementation of FMD Control Strategy and Program Design—Final Design the Productive Safety Nets Programme and the Household Document. Asset Building Programme in the Ethiopian Highlands, Agricultural Transformation Agency. 2018. Beef and Dairy 2014: Program Performance Report. Addis Ababa: Ethiopia Cluster Strategies. Strategy Support Programme, International Food Policy Aklilu, Y. 2009. Livestock and Animal Product Trade and Their Research Institute. Importance for Pastoral Economic Growth in Ethiopia. Berhanu, W. 2017. “Assessment of Vulnerability to Persistent Aklilu, Y., and Catley, A. 2009. Livestock exports from Deprivation: Evidence from a Peripheral Pastoralist pastoralist areas: an analysis of benefits by pastoral wealth Population in Ethiopia.” AERC, Final Research Report. group. A report for the IGAD/FAO Livestock Policy Initiative. Brocklesby, et al. 2010. IDS Working Paper 340 ‘Raising Voice— Feinstein International Centre, Tufts University, Addis securing a livelihood: the role of diverse voices in developing Ababa. https://wikis.uit.tufts.edu/confluence/display/FIC/ a secure likelihood in pastoralist areas in Ethiopia’ March. Livestock+Exports+from+the+Horn+of+Africa Cabot-Venton, et al. 2012. The Economics of Early Response Alkire, S., and Foster, J. E. 2011. ‘Counting and multidimensional and Disaster Resilience: Lessons from Kenya and Ethiopia, poverty measurement’, Journal of Public Economics, vol. 95 DFID. (7–8), pp. 476–487, doi:10.1016/j.jpubeco.2010.11.006 Cabot-Venton, C. 2018a. Economics of Resilience to Drought Association of Ethiopian Microfinance Institutions (AEMFI). in Ethiopia, Kenya and Somalia. USAID: USAID Center for 2015. Overview of Practical Challenges in Local Saving Resilience. Available from: https://www.usaid.gov/sites/ Mobilisation by Ethiopian Microfinance Institutions. default/files/documents/1867/Summary_Economics_of_ Atlas of Ethiopian Livelihoods. 2010. “The Livelihoods Resilience_Final_Jan_4_2018_BRANDED.pdf Integration Unit. Government of Ethiopia: Disaster Risk Cabot-Venton, C. 2018b. Economics of Resilience to Drought: Management and Food Security Sector.” A joint Government Ethiopia Analysis, USAID: USAID Center for Resilience. of Ethiopia and USAID Ethiopia Document. MoARD, USAID. Available from: https://www.usaid.gov/sites/default/files/ Bacha, K., Melesse, K., Hunde, A., and Regassa, T. 2014. “Socio- documents/1867/Ethiopia_Economics_of_Resilience_Final_ Economic Status of Parents and Children’s Schooling Jan_4_2018_BRANDED.pdf in Gambella Region, South West Ethiopia.” Ethiopian Carter, M. R., and Barrett, C. B. 2006. “The Economics of Journal of Education and Science Special Issue Vol 9 SP 1. Poverty Traps and Persistent Poverty: An Asset-Based Accessed at https://www.ju.edu.et/ejes/sites/default/files/ Approach” The Journal of Development Studies, Vol. 42, No. 2, document%281%29.pdf on 15.12.2018 pp. 178–199. Bachewe, F., F. Yimer, and B. Minten. 2017. Agricultural price Carter, M. R., and Barrett, C. B. 2013. “The Economics of evolution in drought vs. non-drought affected areas in Poverty Traps and Persistent Poverty: Empirical and Policy Ethiopia: An Updated Assessment Using National Producer Implications” The Journal of Development Studies, Vol. 49, Data (January 2014 to January 2017). Ethiopia Strategy No. 7, pp. 976–990. Support Program Working Paper 106. Addis Ababa: IFPRI. Catley, A. 2017. “Pathways to Resilience in Pastoralist Areas: Barrett, C. B. 2009. “Measuring Food Insecurity.” Science 327: A Synthesis of Research in the Horn of Africa.” Boston: 825. Feinstein International Center, Tufts University. 136 Catley, Andy, and Alula Iyasu. 2010. Moving Up or Moving Out? Dyer, C., and Engdasew, Z. 2016. “Situation Analysis: Education A Rapid Livelihoods and Conflict Analysis in Mieso-Mulu in the Pastoralist areas of Ethiopia: Report on Exploratory Woreda, Shinile Zone, Somali Region, Ethiopia. Feinstein Survey for Revisiting the 2008 Pastoralist Education International Center, April. Strategy.” British Council. Unpublished. Centre for Rural Development (SLE) Berlin. 2017. Ethiopia’s Arid Elias, Eyasu. 2014. Environmental rights and Pastoral and Semi-Arid Lowlands: Towards Inclusive and Sustainable livelihoods: The case of Borena and Kaarrayu pastoralists in Rural Transformation. Ethiopia. Journal of Environment and Earth Science www.iiste Cervigni, R., and Michael Morris (eds). 2016. Confronting .org, ISSN 2224-3216 (paper) ISSN 2225-0948 (online) Drought in Africa’s Drylands: Opportunities for Enhancing Vol. 4, No. 21, 2014. Resilience. Africa Development Forum Series. Washington, Ellis, J. E., and Swift, D. M. 1988. Stability of African pastoral DC. World Bank, doi: 10.1596/978-14648-0817-3 ecosystems: alternative paradigms and implications for Chaudhuri, S. 2002. Empirical methods for assessing development. Journal of Range Management 41 (6): 458–459. household vulnerability to poverty. New York: Mimeo, Endale, K., Pick, A., and T. Woldehanna. 2019. Financing Social Department of Economics, Columbia University. Protection in Ethiopia: A long term perspective, OECD, Chaudhuri, S. 2003. Assessing vulnerability to poverty: February. Concepts, empirical methods and illustrative examples. ESA. 2018. European Space Agency (ESA) CCI Land Cover—S2 New York: Mimeo, Department of Economics, Columbia prototype Land Cover 20m map of Africa 2016. Retrieved University. September 30, 2019, from http://2016africalandcover20m Chaudhuri, S., Jalan, J., and Suryahadi, A. 2002. Assessing .esrin.esa.int/ household vulnerability to poverty from cross-sectional Ethiopia Civil Servant Survey. 2016. Bureaucracy Lab, data: A methodology and estimates from Indonesia. Governance Global Practice, World Bank Group. Discussion paper no. 0102-52. New York: Columbia Ethiopia Ministry of Agriculture. 2012. Ethiopia Country University. Planning Paper to End Drought Emergencies in the Horn of Cloudburst Group. 2017. Land use rights, land governance Africa. institutions and tenure security indicators in pastoral EWRD (Early Warning and Response Directorate). 2014. community: evidence from a baseline study in the Afar Hotspot (Woredas of Concern) Classification Guideline. Addis region, Ethiopia, USAID. Ababa: Disaster Risk Management and Food Security Conciliation Resources Political Unsettlement and the Somali Sector of the Federal Ministry of Agriculture. Regional State of Ethiopia. 2017. FEWS NET. 2016. Famine Early Warning Systems Network Coppock, D. L. 1994. “The Borana Plateau of Southern Ethiopia: Ethiopia. www.fews.net/ethiopia Synthesis of Pastoral Research, Development, and Change. Feyissa, D. 2010. A National Perspective on the Conflict in 1980–1991.” System Study No. 5. International Livestock Gambella. Center for Africa. Addis Ababa, Ethiopia. Feyissa, D. 2013. “Centering the Periphery? The Praxis of Coppock, D., S. Layne, S. Desta, S. Tezera, and G. Gebru. 2011. Federalism at the Margins of the Ethiopian State.” Center for “Community Building Helps Pastoral Women Transform Federal Studies, Addis Ababa University, Vol. 1, No. 1, June. Impoverished Communities in Ethiopia.” Science 334(9). Foster, J., Greer, J., and Thorbecke, E. 1984. A class of Cossins, N. J., and M. Upton. 1988. Options for improvement of decomposable poverty measures. Econometrica, 52, the Borana pastoral system. Agric. Systems 27:251–278. 761–765. de Haan, C. (ed.). 2016. Prospects for Livestock-Based Fratkin, Elliot. 2001. East African Pastoralism in Transition: Livelihoods in Africa’s Drylands. Washington, DC: World Maasai, Boran, and Rendille Cases. African Studies Review, Bank. Vol. 44, No. 3 (Dec.), pp. 1–25. Degife, A. W., Zabel, F., and Mauser, W. 2018. Assessing land Fratkin, Elliot, and Robin Mearns. 2003. Sustainability and use and land cover changes and agricultural farmland Pastoral Livelihoods: Lessons from East African Maasai and expansions in Gambella Region, Ethiopia, using Landsat 5 Mongolia. Human Organization, Vol. 62, No. 2, 2003. and Sentinel 2a multispectral data. Heliyon, 4(11), e00919. Funk, C. C., J. Rowland, G. Eilerts, E. Kebebe, N. Biru, L. White, http://doi.org/10.1016/j.heliyon.2018.e00919 and G. Galu. 2012. A climate trend analysis of Ethiopia, Deininger, K., Byerlee, D., Lindsay, J., Norton, A., Selod, H., US Geological Survey. and Stickler, M. 2011. Rising Global Interest in Farmland. Gebremedhin, B., Woldehanna, M., Flintan, F., Wieland B., Washington D.C.: World Bank. http://doi.org/10.1596/ and Poole, J. 2017. Baseline survey report for the Regional 978-0-8213-8591-3 Pastoral Livelihoods Resilience Project in Ethiopia. ILRI Desta, S., and L. Coppock. 2002. “Cattle Population Dynamics project report. Nairobi, Kenya: International Livestock in the Southern Ethiopian Rangelands, 1980–1997.” Journal Research Institute (ILRI). of Range Management 55(5): 439–451. Getachew, Kassa Nigussie. 2001. Among the Pastoral Afar in Desta, S., W. Berhanu, A. Sebsibe, and E. Guerne Bleich. 2011. Ethiopia: Tradition, Continuity and Socioeconomic Change. “Regional Initiative in Support of Vulnerable Pastoralists and Addis Ababa: OSSREA. Agro-Pastoralists. Assessment of Cross Border Livestock Gezahegn, Amaha Kassahun. 2006. Characterization of Trade in the Ethiopia-Somali Region.” Food and Agriculture Rangeland Resources and Dynamics of the Pastoral Organization of the United Nations, Addis Ababa, Ethiopia. Production Systems in the Somali Region of Eastern Devereux, Stephen. 2006. Vulnerable Livelihoods in Somali Ethiopia. Submitted in partial fulfillment of the requirement Region, Ethiopia. Institute of Development Studies, for the degree of Doctor of Philosophy in the faculty of Research Report 57. Natural and Agricultural Sciences, Department of Animal, DevSeed. 2018. https://developmentseed.org/ Wildlife and Grassland Sciences (Grassland Science), University of the Free State, Bloemfontein. References  137 Gunther I., and K. Harttgen. 2009. “Estimating Household Krauer, J., Fries, M., Ursula, G., Hodel, E., Kassawmar, T., Vulnerability to Idiosyncratic and Covariate Shocks: A Novel Zeleke, G., and Aragi, Y. 2015. EthioGIS—II Data Catalog. Methods Applied in Madagascar,” World Development, Vol. 37, Addis Ababa. No. 7, pp. 1222–1234. Lind, J., et al. 2018. Targeting Social Transfers in Pastoralist Habte, A. 2003. Ethnic Federalism in Ethiopia: Background, Societies—Ethiopia’s Productive Safety Net Programme Present Conditions and Future Prospects. Revisited, IFPRI. Harbeson, J. W. 1978. Territorial and development politics in the Little, P. 2010. Unofficial Cross-border Trade in Eastern Africa. Horn of Africa: the Afar of the Awash valley. African Affairs Little, Peter D., Dejene Negassa Debsu, and Waktole Tiki. 77 (309): 479–498. 2015. How pastoralists perceive and respond to market Hatfield, R., and Davies, J. 2006. Global Review of the opportunities: The case of the Horn of Africa. Food Policy 49 Economics of Pastoralism. World Initiative for Sustainable (2014) 389–397. Pastoralism. Little, Peter D., McPeak, John G., Barrett, Christopher B., Headey, D., Taffesse, A. S., and You, L. 2014. Diversification and and Kristjanson, Patti. 2011. “Challenging Orthodoxies: Development in Pastoralist Ethiopia. World Development, 56, Understanding Poverty in Pastoral Areas of East Africa.” 200–213. https://doi.org/10.1016/j.worlddev.2013.10.015 Economics Faculty Scholarship. Paper 83. Hill, R. V., and Porter, C. 2017. “Vulnerability to Drought Livestock Master Plan. 2015. Roadmaps for Growth and and Food Price Shocks: Evidence from Ethiopia,” World Transformation. Development, Vol. 96, pp. 65–77. Markakis, J. (ed.) 1993. Conflict and the Decline of Pastoralism Holtland, Gerrit. 2011. The Role of Cooperatives, SACCOs in the Horn of Africa, London: Macmillan. and Traders in Value Chain Development in Borana Zone. Markakis, J. 2013. Ethiopia The Last Two Frontiers. Research Report, Trócaire CAFOD SCIAF, January. Markakis, John. 2011. Ethiopia: The Last Two Frontiers. NY: Hox, J. 2010. Multilevel Analysis: Techniques and Applications (2nd James Currey. edition) Routledge. McPeak, John, Peter Little, Mercedes Stickler, and Heather HRD (Humanitarian Requirement Document). 2015. Ethiopia Huntington. 2015. Pastoral Land Tenure at the Margins of 2016: Joint Government and Humanitarian Partners’ Document. Intensive and Extensive Land Use: Baseline Survey Results Addis Ababa: Federal Democratic Republic of Ethiopia. From a USAID Customary Land Rights Recognition Project Accessed from: http://www.unicef.org/ethiopia/ECO_ in Southern Ethiopia. Paper prepared for presentation at Ethiopia_HRD_2016.pdf the “2015 World Bank Conference on Land and Poverty” The Human Rights Report. 2015. Country Reports on Human World Bank—Washington DC, March 23–27, 2015. Rights Practices for 2015. United States Department of Mekuyie, M., Jordaan, A., and Melka, Y. 2018. Land-use State, Bureau of Democracy, Human Rights and Labor. and land-cover changes and their drivers in rangeland- Hurni, H. 1982. Soil Conservation Research Project. Bern. dependent pastoral communities in the southern Afar Retrieved from https://scholar.google.com/scholar? Region of Ethiopia. African Journal of Range and Forage cluster=16246864023628728797&hl=en&oi=scholar Science, 35(1), 33–43. http://doi.org/10.2989/ Hurni, H. 1998. Agroecological Belts of Ethiopia: Explanatory notes 10220119.2018.1442366 on three maps at a scale of 1:1,000,000 in association with The Mussa, Mohammed. 2004. Case Study on the Pastoral Affairs Ministry of Agriculture, Ethiopia. Bern. Retrieved from https:// Standing Committee of Ethiopia. For the NRI/PENHA library.wur.nl/isric/fulltext/isricu_t4d384c8b_001.pdf Research Project on Pastoralist Parliamentary Groups, Hurni, H., Tato, K., and Zeleke, G. 2005. The Implications of funded by DFID’s Livestock Production Programme and the Changes in Population, Land Use, and Land Management CAPE Unit, African Union’s Interafrican Bureau of Animal for Surface Runoff in the Upper Nile Basin Area of Ethiopia. Resources (AU-IBAR). May. BioOne Complete, 25(2), 147–155. http://doi.org/10.1659/ Mwangi, E., and Swallow, B. 2005. Invasion of prosopis juliflora 0276-4741(2005)025[0147:TIOCIP]2.0.CO;2 and local livelihoods: Case study from the Lake Baringo area of Hurni, K. 2008. Major Agroecological Zones of Ethiopia. Bern. Kenya (ICRAF Working Paper No. 3). Nairobi. ICG. 2019. Managing Ethiopia’s Unsettled Transition, Report National Bank of Ethiopia. 2017. National Financial Inclusion No. 269. Strategy. IFPRI. 2016. Economy-wide impacts of the Productive Safety Niang, I., O. C. Ruppel, M. A. Abdrabo, A. Essel, C. Lennard, Net Programme (PSNP). ESSP Research Note 57: IFPRI, J. Padgham, and P. Urquhart. 2014. Africa. In: Climate Washington DC. Change 2014: Impacts, Adaptation, and Vulnerability. Intergovernmental Authority on Development (IGAD). 2018. Part B: Regional Aspects. Contribution of Working Group II Policy Framework on the Informal Cross-Border Trade and to the Fifth Assessment Report of the Intergovernmental Cross-Border Security Governance Nexus; Enhancing Cross- Panel on Climate Change [Barros, V. R., C. B. Field, Border Cooperation and Cross-Border Economic Exchanges D. J. Dokken, M. D. Mastrandrea, K. J. Mach, T. E. Bilir, in the IGAD Region. M. Chatterjee, K. L. Ebi, Y. O. Estrada, R. C. Genova, International Crisis Group (ICG). 2009. “Ethiopia: Ethnic B. Girma, E. S. Kissel, A. N. Levy, S. MacCracken, Federalism and Its Discontents.” Africa Report No. 153, P. R. Mastrandrea, and L. L. White (eds.)]. Cambridge Nairobi/Brussels. University Press, Cambridge, United Kingdom and New York, Kassawmar, T., Zeleke, G., Bantider, A., Gessesse, G. D., and NY, pp. 1199–1265. Abraha, L. 2018. A synoptic land change assessment of Nigussie A., D. Hoag, and T. Alemu. 2014. Women’s workload Ethiopia’s Rainfed Agricultural Area for evidence-based and role in livestock production in pastoral and agro- agricultural ecosystem management. Heliyon, 4(11), e00914. pastoral communities of Ethiopia: The case of Afar. African http://doi.org/10.1016/J.HELIYON.2018.E00914 Journal of Agricultural Economics and Rural Development ISSN: 138  References 2141–5091 Vol. 2 (4), pp. 138–146, June. Available online at Institute for Environment and Development, Issue Paper, www.internationalscholarsjournals.org April. Oba, Gufu. 2009. Harnessing Pastoralists’ Indigenous Range Tegegn, G. G. 2008. Experiences on Prosopis Management: Management Knowledge for Drought-Resilient Livelihood Case of Afar Region (Publication from Farm Africa). Systems in the Horn of Africa—September. https://pdfs.semanticscholar.org/6aac/ Ohta, I., and Gebre Yintiso. 2005. Displacement risks in Africa. 4627d3399c7424f673dd375a7f81c7bb5c1f.pdf Kyoto, Japan: Kyoto University Press. Townsend, R. M. 1994. “Risk and Insurance in village India” Omzigt, Dirk-Jan. 2018. A Study on Lowland Resilience in Econometrica, vol. 62, No. 3, pp. 539–591. Ethiopia—Building a More Developmental Approach to UNDESA (United Nations Department of Economic and Social Responding to Shocks, May UNDP. Affairs). 2017. World Population Prospects: The 2017 Revision. Oumer, A., and Z. Ali. 2007. ‘Handicraft Production in Somali, New York: United Nations. https://population.un.org/wpp/ Afar and South Omo’ in: Ridgewell and Flintan (eds.) (2007) DataQuery/ Gender and Pastoralism, Vol. 2: Livelihoods and Income UNDP. 2018. Human Development Report, Human Development in Ethiopia. SOS Sahel Ethiopia, Addis Ababa. Development Indicators: Ethiopia. http://www.hdr.undp.org/ Owitti, Ojulu. 2015. Gender Differences and Relations in Rural en/countries/profiles/ETH Household Livelihoods of Gog District, Anywaa Zone, United Nations Population Division, World Population Gambella Region, South Western Ethiopia. International Prospects. 2019. https://population.un.org/wpp/DataQuery/ Journal of Gender & Women's Studies. 3. 10.15640/ijgws USAID. 2009. Agricultural Value Chain Financing in Kenya: .v3n1a7 Assessment of Potential Opportunities for Growth. Pankhurst, A. (ed.). 2017. Changes and Transformation in USAID. 2013. (AGP—Livestock Market Development Project) Twenty Rural Communities in Ethiopia: Selected aspects Value Chain Analysis for Ethiopia: Meat and Live Animals, and implications for policy, Addis Ababa. Hides, Skins and Leather, Dairy. PFE, IIRR, and DF. 2010. Pastoralism and land: Land tenure, USAID. 2017. (Feed the Future) Value Chain Analysis: Live administration and use in pastoral areas of Ethiopia. Animals and Meat. Rahmato, Dessalegn. 2007. Customs in Conflict: Land Tenure USAID. 2017. Afar Regional Overview & Change Analysis. Issues among Pastoralists in Ethiopia. Forum for Social Van Der Beken, C. 2018. The Challenge of Reform within Studies. Addis Ababa. November. Ethiopia’s Constitutional Order. Reda, Kelemework Tafere. 2013. Pastoral Customary Laws and Vaughan, S. 2012. Revolutionary democracy as an Practices on Land Administration and Management of Land uncompromising political strategy: Revolutionary Resources in Afar Regional State. Ethiopia—Strengthening democratic state-building: party, state and people in the Land Administration Program (ELAP). Addis Ababa, Ethiopia. EPRDF’s Ethiopia. Regassa, S., Givey, C., and Castillo, G. E. 2010. The Rain Doesn’t Vaughan, S. 2017. Innovation, ‘technology transfer’ and positive Come On Time Anymore: Poverty, vulnerability, and climate social change: models and realities of transformation. variability in Ethiopia. Retrieved from http://lib.icimod.org/ Vaughan, S., and K. Tronvoll. 2003. “The Culture of Power in record/14668/files/6371.pdf Contemporary Ethiopian Political Life” Sida studies No. 10. Said, Ali. 1994. Pastoralism and the State Policies in Middle https://www.researchgate.net/publication/44833437_The_ Awash Valley: The Case of the Afar, Ethiopia. African Arid Culture_of_Power_in_Contemporary_Ethiopian_Political_Life Lands Working Paper Series No. 1/94. Yang, W. C., R. Seager, M. A. Cane, and B. Lyon. 2015. The Sandford, Stephen. 1983. Management of Pastoral Annual Cycle of East African Precipitation. Journal of Climate Development in the Third World Overseas Development 28 (6): 2385–404. doi:10.1175/Jcli-D-14-00484.1 Institute, London, and John Wiley and Sons, Chichester. Young, H., Behnke, R., Sulieman, H., Robinson, S., with Save the Children. 2016. International Briefing Note on Mohamed, A. 2016. Risk, Resilience, and Pastoralist Migration for Employment. David Gray. Mobility. Feinstein International Center, August 2016. Shigute, Z., Strupat, C., Burchi, F., Alemu, G., and Arjun S. Williams, A. P., and C. Funk. 2011. “A Westward Extension of the Bedi. 2017. The Joint Effects of a Health Insurance and a Warm Pool Leads to a Westward Extension of the Walker Public Works Scheme in Rural Ethiopia. Institute of Labor Circulation, Drying Eastern Africa.” Climate Dynamics 37 Economics Discussion Paper 10939. IZA, Bonn. (11–12): 2417–35. doi:10.1007/s00382-010-0984-y. SNV Ethiopia and Target Business Consultants. 2017. Inventory World Bank. 2017. World Development Indicators. Washington, of Dairy Policy, Ethiopia. DC. https://data.worldbank.org/country Tache, Boku, and Gufu Oba. 2010. Is Poverty Driving Borana Zeleke, G., and Hurni, H. 2001. Implications of Land Use and Herders in Southern Ethiopia to Crop Cultivation? Human Land Cover Dynamics for Mountain Resource Degradation Ecology: 38:639–649. in the Northwestern Ethiopian Highlands. BioOne Complete, Tari, Daoud, and James Pattison. 2014. Evolving Customary 21(2), 184–192. http://doi.org/10.1659/0276-4741(2001) Institutions in the Drylands: An opportunity for devolved 021[0184:IOLUAL]2.0.CO;2 natural resource governance in Kenya? International References  139 Poverty and Vulnerability in the Ethiopian Lowlands Annex 1: Environment and Land Use Tables Table A.1: Area by region, lowland/highland and lowland AEZ (km2) Lowland AEZ     Lowland Highland Total Dry Berha Dry Kolla Moist Berha Moist Kolla Addis Ababa — 539 539 Afar 94,384 773 95,157 41,568 52,817 — — Amhara 40,092 115,284 155,376 — 18,255 — 21,837 Benishangul-Gumuz 46,233 3,797 50,030 — 21,131 — 25,102 Dire Dawa 705 349 1,054 — 705 — — Gambella 29,579 1,129 30,708 — 2,128 17,238 10,213 Hareri 120 250 371 — 120 — — Oromia 141,970 180,877 322,848 992 107,495 96 33,387 SNNPR 61,100 47,445 108,545 8,619 25,881 — 26,600 Somali 304,952 7,861 312,814 92,337 212,615 — — Tigray 24,472 27,366 51,838 65 23,762 — 644 Total   743,607 385,672 1,129,280 143,581 464,909 17,335 117,783 Source: WB calculations based on Hurni (1982) and WLRC. 140 Table A.2: Average annual rainfall by region, lowland/highland, and lowland AEZ (mm/year) Lowland AEZ Lowland Highland Total Dry Berha Dry Kolla Moist Berha Moist Kolla Mean annual precipitation 1982–2017 Afar 348.6 548.4 350.2 227.5 444.0 Amhara 1,042.8 1,091.7 1,079.1 888.4 1,172.0 Benishangul-Gumuz 1,220.1 1,448.8 1,237.5 1,074.3 1,342.9 Gambella 1,135.3 1,810.2 1,160.1 1,297.8 985.0 1,355.1 Oromia 749.5 1,207.3 1,006.0 397.5 628.4 974.7 1,149.3 SNNPR 1,161.7 1,389.3 1,261.2 742.3 997.5 1,457.3 Somali 330.8 473.7 334.4 267.4 358.3 Tigray 937.8 716.6 821.0 412.2 928.4 1,337.6 Total   627.2 1,148.0 804.9 285.3 553.2 984.9 1,283.2 Mean annual precipitation 1982–1999 Afar 349.0 544.0 350.6 225.0 446.6 Amhara 1,026.4 1,083.3 1,068.6 875.2 1,152.8 Benishangul-Gumuz 1,194.2 1,423.4 1,211.6 1,042.4 1,322.0 Gambella 1,116.4 1,795.3 1,141.3 1,266.1 967.7 1,336.1 Oromia 757.1 1,206.2 1,008.7 395.3 634.4 965.6 1,162.2 SNNPR 1,120.3 1,372.8 1,230.7 702.5 954.7 1,416.7 Somali 336.9 477.3 340.5 270.6 365.8 Tigray 921.7 709.3 809.6 435.9 912.5 1,309.8 Total   624.1 1,142.2 800.8 284.3 553.0 967.7 1,267.9 Mean annual precipitation 2000–2017 Afar 348.3 552.6 349.9 230.0 441.3 Amhara 1,059.3 1,100.1 1,089.6 901.5 1,191.2 Benishangul-Gumuz 1,246.0 1,474.2 1,263.4 1,106.1 1,363.8 Gambella 1,154.2 1,825.2 1,178.8 1,329.4 1,002.2 1,374.1 Oromia 742.0 1,208.4 1,003.3 399.7 622.4 984.0 1,136.5 SNNPR 1,203.1 1,405.8 1,291.7 782.1 1,040.2 1,497.9 Somali 324.6 470.2 328.2 264.3 350.8 Tigray 953.8 724.0 832.5 389.1 944.2 1,365.3 Total   630.4 1,153.8 808.9 286.4 553.4 1,002.1 1,298.6 Source: WB calculations and WLRC using CHIRPS. Excludes Addis Ababa, Dire Dawa, and Hareri. Annex 1: Environment and Land Use Tables   141 Table A.3: Rainfall trend by region, lowland/highland, and lowland AEZ (mm/year)     Lowland AEZ     Lowland   Highland   Total   Dry Berha Dry Kolla Moist Berha Moist Kolla Precipitation trend 1982–2017 Afar 0.657 1.217 0.661 0.716 0.610 Amhara 3.003 2.285 2.470 2.442 3.471 Benishangul-Gumuz 4.227 4.207 4.226 4.772 3.768 Gambella 2.987 3.035 2.989 4.823 2.708 3.077 Oromia –0.590 0.989 0.295 0.615 –0.596 1.314 –0.610 SNNPR 5.432 3.142 4.431 5.176 5.310 5.633 Somali –0.264 0.055 –0.256 0.114 –0.429 Tigray 2.476 1.551 1.988 –1.859 2.442 4.180 Total   0.935 1.702 1.197 0.595 0.492 2.700 2.836 Precipitation trend 1982–1999 Afar 5.456 10.842 5.500 3.148 7.273 Amhara 11.279 11.283 11.282 11.477 11.113 Benishangul-Gumuz 8.392 7.789 8.346 10.761 6.398 Gambella 6.725 8.406 6.787 8.077 6.428 6.946 Oromia –2.194 3.688 1.101 0.325 –2.874 4.107 –0.097 SNNPR 8.739 7.481 8.189 7.350 7.883 10.022 Somali 0.525 1.964 0.562 1.394 0.148 Tigray 9.225 9.504 9.372 6.949 9.162 11.770 Total   2.911 6.880 4.265 2.254 2.117 6.415 6.326 Precipitation trend 2000–2017 Afar 0.055 –3.994 0.022 0.923 –0.628 Amhara 1.791 1.407 1.507 –0.679 3.856 Benishangul-Gumuz 8.178 8.975 8.239 6.201 9.843 Gambella 4.591 5.933 4.641 9.424 3.753 5.000 Oromia 2.517 3.517 3.077 3.157 2.110 0.327 3.814 SNNPR 7.131 6.692 6.939 7.555 6.096 8.002 Somali 1.492 0.836 1.476 1.641 1.428 Tigray –0.118 –1.994 –1.109 –6.244 –0.190 3.157 Total   2.472 2.875 2.610 1.795 1.701 3.734 6.152 Source: WB calculations and WLRC using CHIRPS. Annual trends are obtained by regressing annual total rainfall as a function of year on a pixel-by- pixel basis. The reported values are an average of the estimated slope coefficient for a given classification (weighted by area). Excludes Addis Ababa, Dire Dawa, and Hareri. 142   Annex 1: Environment and Land Use Tables Table A.4: Key short-term agro-meteorological trends by seed location Location Drought cluster Rainfall Max temp. stress days ID Region Elevation AEZ (mm/year) (degrees/year) (# days/year) 1 Afar 329 Dry <500 m (Dry Berha) 6.4 –0.3 –1.8 2 Afar 821 Dry 500–1,500 m (Dry Kola) 17.6 –0.3 –1.1 3 Afar 912 Dry 500–1,500 m (Dry Kola) 4.0 –0.4 –1.3 4 Somali 1,007 Dry 500–1,500 m (Dry Kola) 9.9 0.1 –0.9 5 Somali 384 Dry <500 m (Dry Berha) 3.7 –0.2 1.0 6 Somali 906 Dry 500–1,500 m (Dry Kola) –12.7 –0.4 1.6 7 Oromia 1,430 Dry 500–1,500 m (Dry Kola) –7.4 –0.3 –0.4 8 SNNP 594 Dry <500 m (Dry Berha) 3.6 –0.1 –0.7 9 Gambella 451 Moist <500 m (Moist Berha) 9.5 –0.7 –0.8 10 Benishangul-Gumuz 1,366 Moist 500–1,500 m (Moist Kola) –3.1 –0.8 –1.5 11 Amhara 673 Dry 500–1,500 m (Dry Kola) –15.0 –0.8 –0.1 12 Tigray 906 Dry 500–1 500 m (Dry Kola)   –13.5 –0.8 –0.2 Source: Trend estimates based on analysis by aWhere. Annex 1: Environment and Land Use Tables   143 Table A.5: Average NDVI by region, lowland AEZ, and lowland/highland (index) Lowland AEZ     Lowland Highland Total   Dry Berha Dry Kolla Moist Berha Moist Kolla Mean NDVI 2000–2017 Afar 0.152 0.278 0.153 0.113 0.182 Amhara 0.395 0.357 0.366 0.347 0.435 Benishangul-Gumuz 0.512 0.545 0.515 0.483 0.537 Gambella 0.635 0.835 0.642 0.600 0.592 0.714 Oromia 0.402 0.511 0.463 0.136 0.378 0.561 0.487 SNNPR 0.525 0.587 0.552 0.397 0.479 0.610 Somali 0.235 0.259 0.235 0.208 0.246 Tigray 0.317 0.284 0.299 0.099 0.314 0.433 Total   0.324 0.454 0.369 0.191 0.302 0.592 0.535 Mean NDVI 2000–2009 Afar 0.153 0.281 0.154 0.113 0.185 Amhara 0.403 0.359 0.371 0.353 0.445 Benishangul-Gumuz 0.517 0.550 0.519 0.486 0.543 Gambella 0.622 0.835 0.630 0.594 0.575 0.708 Oromia 0.404 0.518 0.468 0.136 0.379 0.550 0.493 SNNPR 0.526 0.594 0.556 0.389 0.477 0.617 Somali 0.233 0.260 0.234 0.209 0.244 Tigray 0.324 0.283 0.302 0.101 0.321 0.439 Total   0.325 0.459 0.371 0.191 0.302 0.575 0.541 Mean NDVI 2010–2017 Afar 0.150 0.273 0.151 0.114 0.178 Amhara 0.384 0.353 0.361 0.338 0.422 Benishangul-Gumuz 0.506 0.539 0.509 0.479 0.529 Gambella 0.654 0.836 0.661 0.609 0.618 0.724 Oromia 0.399 0.502 0.457 0.137 0.376 0.578 0.480 SNNPR 0.523 0.577 0.547 0.407 0.481 0.601 Somali 0.237 0.258 0.237 0.208 0.249 Tigray 0.307 0.284 0.295 0.096 0.305 0.425 Total   0.324 0.447 0.366 0.192 0.302 0.618 0.528 Source: WB calculations and WLRC using MODIS 250 m, 16-day composite (MOD13Q1). The dataset used to generate these estimates and for subsequent trend/regression analysis has been cleaned using a filtering routine to remove contaminated or low-quality pixels (i.e., cloud, shadow, etc.) was implemented along with a water mask to remove outlier observations and pixels which might skew results. Excludes Addis Ababa, Dire Dawa, and Hareri. 144   Annex 1: Environment and Land Use Tables Table A.6: NDVI trend by region, lowland/highland, and lowland AEZ (%) Lowland AEZ Lowland Highland Total Dry Berha Dry Kolla Moist Berha Moist Kolla NDVI trend 2000–2017 Afar –0.11% –0.16% –0.11% 0.01% –0.20% Amhara –0.15% –0.03% –0.06% –0.16% –0.14% Benishangul-Gumuz 0.02% 0.02% 0.02% 0.04% 0.01% Gambella 0.30% 0.02% 0.29% 0.13% 0.39% 0.18% Oromia 0.02% –0.07% –0.03% 0.13% 0.03% 0.29% –0.01% SNNPR 0.02% –0.09% –0.03% 0.19% 0.06% –0.08% Somali 0.07% –0.03% 0.07% –0.01% 0.10% Tigray –0.16% 0.01% –0.07% –0.25% –0.17% –0.06% Total   0.02% –0.05% 0.00% 0.01% 0.02% 0.39% –0.03% NDVI trend 2000–2009 Afar –0.34% –0.66% –0.34% –0.13% –0.50% Amhara –0.14% –0.07% –0.09% –0.21% –0.08% Benishangul-Gumuz 0.00% –0.02% 0.00% 0.00% 0.01% Gambella 0.18% 0.05% 0.18% 0.09% 0.21% 0.16% Oromia 0.22% 0.08% 0.14% 0.39% 0.22% 0.10% 0.20% SNNPR 0.22% 0.01% 0.13% 0.49% 0.29% 0.07% Somali 0.04% –0.03% 0.04% 0.05% 0.04% Tigray –0.18% –0.21% –0.20% –0.10% –0.18% –0.06% Total   0.03% 0.00% 0.02% 0.03% 0.01% 0.21% 0.08% NDVI trend 2010–2017 Afar 0.40% 1.00% 0.41% 0.43% 0.38% Amhara 0.61% 0.63% 0.63% 0.40% 0.78% Benishangul-Gumuz 1.38% 1.27% 1.37% 1.49% 1.29% Gambella 1.27% 0.15% 1.23% 0.33% 1.62% 0.87% Oromia 0.15% 0.34% 0.26% –0.18% 0.04% 1.53% 0.50% SNNPR –0.26% 0.17% –0.07% –0.88% –0.36% 0.04% Somali 0.22% –0.15% 0.21% 0.05% 0.29% Tigray 0.55% 0.61% 0.58% –0.47% 0.54% 0.80% Total   0.33% 0.43% 0.37% 0.10% 0.28% 1.62% 0.65% Source: WB calculations and WLRC using MODIS 250 m, 16-day composite (MOD13Q1). Values in the table represent trends as percentage change for a given study period (weighted by area). These values are computed by regressing on a pixel-by-pixel the full collection of NDVI observations against time to estimate the slope coefficient and expressed as an elasticity. Excludes Addis Ababa, Dire Dawa, and Hareri. Annex 1: Environment and Land Use Tables   145 Table A.7: Average EVI by region, lowland AEZ, and lowland/highland (index)   Lowland AEZ     Lowland Highland   Total   Dry Berha Dry Kolla Moist Berha Moist Kolla Mean EVI 2000–2017 Afar 0.088 0.161 0.089 0.065 0.106 Amhara 0.224 0.195 0.203 0.193 0.250 Benishangul-Gumuz 0.287 0.303 0.288 0.269 0.301 Gambella 0.380 0.487 0.384 0.350 0.349 0.439 Oromia 0.219 0.297 0.263 0.089 0.203 0.334 0.273 SNNPR 0.309 0.357 0.330 0.234 0.274 0.367 Somali 0.140 0.157 0.140 0.132 0.143 Tigray 0.184 0.166 0.175 0.066 0.183 0.242 Total   0.187 0.262 0.213 0.119 0.171 0.349 0.310 Mean EVI 2000–2009 Afar 0.089 0.163 0.089 0.065 0.107 Amhara 0.228 0.197 0.205 0.196 0.255 Benishangul-Gumuz 0.291 0.307 0.292 0.273 0.306 Gambella 0.370 0.488 0.375 0.345 0.335 0.435 Oromia 0.219 0.301 0.265 0.088 0.203 0.327 0.276 SNNPR 0.309 0.363 0.333 0.226 0.272 0.373 Somali 0.138 0.157 0.139 0.132 0.141 Tigray 0.188 0.166 0.176 0.067 0.186 0.247 Total   0.187 0.265 0.214 0.118 0.170 0.335 0.314 Mean EVI 2010–2017 Afar 0.087 0.159 0.088 0.066 0.104 Amhara 0.218 0.193 0.199 0.189 0.242 Benishangul-Gumuz 0.281 0.298 0.282 0.264 0.295 Gambella 0.396 0.485 0.399 0.357 0.371 0.446 Oromia 0.218 0.292 0.260 0.089 0.204 0.348 0.269 SNNPR 0.309 0.350 0.326 0.244 0.278 0.359 Somali 0.141 0.156 0.142 0.132 0.145 Tigray 0.179 0.166 0.172 0.064 0.177 0.235 Total   0.187 0.258 0.211 0.119 0.171 0.371 0.305 Source: WB calculations and WLRC using MODIS 250 m, 16-day composite (MOD13Q1). The dataset used to generate these estimates and for subsequent trend/regression analysis has been cleaned using a filtering routine to remove contaminated or low-quality pixels (i.e., cloud, shadow, etc.) and was implemented along with a water mask to remove outlier observations and pixels which might skew results. Excludes Addis Ababa, Dire Dawa, and Hareri. 146   Annex 1: Environment and Land Use Tables Table A.8: EVI trend by region, lowland/highland, and lowland AEZ (%) Lowland AEZ Lowland Highland Total Dry Berha Dry Kolla Moist Berha Moist Kolla EVI trend 2000–2017 Afar –0.14% –0.15% –0.14% –0.02% –0.23% Amhara –0.17% –0.05% –0.08% –0.17% –0.16% Benishangul-Gumuz –0.06% –0.03% –0.06% –0.05% –0.07% Gambella 0.37% –0.03% 0.36% 0.15% 0.52% 0.17% Oromia 0.05% –0.06% –0.01% 0.11% 0.07% 0.33% –0.02% SNNPR 0.02% –0.12% –0.04% 0.23% 0.09% –0.11% Somali 0.08% –0.04% 0.08% –0.01% 0.12% Tigray –0.17% –0.01% –0.08% –0.25% –0.17% –0.14% Total   0.02% –0.06% –0.01% 0.00% 0.03% 0.52% –0.06% EVI trend 2000–2009 Afar –0.42% –0.71% –0.42% –0.17% –0.61% Amhara –0.22% –0.13% –0.15% –0.32% –0.14% Benishangul-Gumuz –0.14% –0.11% –0.14% –0.16% –0.13% Gambella 0.16% –0.18% 0.15% 0.14% 0.21% 0.07% Oromia 0.21% 0.04% 0.12% 0.35% 0.23% 0.14% 0.13% SNNPR 0.17% –0.08% 0.07% 0.45% 0.25% 0.01% Somali 0.05% –0.04% 0.05% 0.03% 0.06% Tigray –0.21% –0.26% –0.23% –0.14% –0.21% –0.13% Total   0.00% –0.05% –0.02% 0.00% –0.01% 0.21% –0.01% EVI trend 2010–2017 Afar 0.12% 0.83% 0.13% 0.10% 0.14% Amhara 0.43% 0.63% 0.58% 0.27% 0.56% Benishangul-Gumuz 1.29% 1.22% 1.29% 1.35% 1.25% Gambella 1.42% 0.21% 1.38% 0.24% 1.78% 1.06% Oromia 0.10% 0.35% 0.24% –0.14% –0.03% 1.37% 0.53% SNNPR –0.33% 0.18% –0.10% –1.12% –0.47% 0.07% Somali 0.12% –0.26% 0.11% –0.06% 0.19% Tigray 0.40% 0.53% 0.47% –0.65% 0.40% 0.58% Total   0.23% 0.43% 0.30% –0.08% 0.17% 1.78% 0.63% Source: WB calculations and WLRC using MODIS 250 m, 16-day composite (MOD13Q1). Values in the table represent trends as percentage change for a given study period (weighted by area). These values are computed by regressing on a pixel-by-pixel the full collection of NDVI observations against time to estimate the slope coefficient and expressed as an elasticity. Excludes Addis Ababa, Dire Dawa, and Hareri. Annex 1: Environment and Land Use Tables   147 Table A.9: Land use and land cover (LULC) classes and definitions LULC class Description Forest An area covered with natural and plantation tree species with more than 30% crown cover, 5 meters height at maturity, and greater than 0.5 ha in area. Woodland The land occupied by woody species with open crown cover. Shrub and bush Shrubs and bushes. Cropland A land cover with different crops and physical conservation structures on it. Grassland A land used for pasture and occupied by grass species. Physical and biological measures done on grassland categorize it as grassland. Bare land A land without any vegetation. It includes rocky surfaces, sandy surfaces, roads, etc. Swamp and marsh area An area occupied with a water body and having aquatic grass and herb species for more than nine months in the year. Water body Water body includes lakes, big ponds and dam areas. Settlement Small and big towns with dense houses. It does not include rural settlements. Source: WLRC. Table A.10: Summary of LULC by highland/lowland (1986 and 2016) Highland Lowland National Area (1,000 ha) Change Area (1,000 ha) Change Area (1,000 ha) Change Land use category 1986 2016 % 1986 2016 % 1986 2016 % Forest 5,915 5,478 –7% 5,064 3,318 –34% 10,979 8,795 –20% Woodland 5,666 3,692 –35% 30,688 22,919 –25% 36,355 26,611 –27% Shrub/bush 5,534 4,464 –19% 19,436 23,355 20% 24,970 27,819 11% Cropland 13,847 17,962 30% 2,234 4,691 110% 16,081 22,653 41% Grassland 5,278 3,798 –28% 7,106 9,392 32% 12,385 13,189 6% Bare land 1,416 2,154 52% 9,102 10,007 10% 10,518 12,161 16% Wetland 133 60 –55% 478 298 –38% 611 359 –41% Water body 471 497 5% 241 349 45% 711 846 19% Settlements   57 239 321% 11 32 179% 68 270 297% Source: WB calculations based on data from WLRC, ESA, and DevSeed. Table A.11: Lowland LULC (1986, 2000, and 2016) Area (1,000 ha) Change (%) Land use category 1986 2000 2016 1986–2000 2000–2016 1986–2016 Forest 5,064 3,954 3,318 –22% –16% –34% Woodland 30,688 29,635 22,919 –3% –23% –25% Shrub/bush 19,436 20,164 23,355 4% 16% 20% Cropland 2,234 3,232 4,691 45% 45% 110% Grassland 7,106 8,638 9,392 22% 9% 32% Bare land 9,102 8,002 10,007 –12% 25% 10% Wetland 478 407 298 –15% –27% –38% Water body 241 268 349 11% 30% 45% Settlements   11 — 32 — — 179% Source: WB calculations based on data from WLRC, ESA, and DevSeed. 148   Annex 1: Environment and Land Use Tables Table A.12: Lowland forest (1986 and 2016) Area (1,000 ha) Change (%) Region 1986 2016 1986–2016 Afar 124.3 119.8 –4% Amhara 142.9 125.2 –12% Benishangul-Gumuz 555.9 542.4 –2% Dire Dawa 2.6 0.7 –74% Gambella 535.8 511.4 –5% Hareri 2.1 0.3 –87% Oromia 1,177.6 695.1 –41% SNNPR 1,564.9 866.6 –45% Somali 898.3 423.8 –53% Tigray   59.3 32.4 –45% Total   5,063.6 3,317.7 –34% Source: WB calculations based on data from WLRC, ESA, and DevSeed. Table A.13: Lowland cropland (1986, 2000, and 2016) Area (1,000 ha) Change (%) Region 1986 2000 2016 1986–2000 2000–2016 1986–2016 Afar 27.7 45.7 147.1 65% 222% 431% Amhara 463.1 615.7 716.2 33% 16% 55% Benishangul-Gumuz 205.9 370.2 519.3 80% 40% 152% Dire Dawa 10.8 9.9 11.6 –8% 17% 7% Gambella 27.3 39.0 106.8 43% 174% 292% Hareri 6.2 6.0 7.0 –3% 16% 12% Oromia 588.9 947.1 1,378.5 61% 46% 134% SNNPR 414.5 446.7 853.8 8% 91% 106% Somali 147.1 250.7 410.7 70% 64% 179% Tigray   342.7 500.8 540.4 46% 8% 58% Total   2,234.1 3,231.9 4,691.3 45% 45% 110% Source: WB calculations based on data from WLRC, ESA, and DevSeed. Table A.14: Lowland grazing and forage* (1986 and 2016) Table A.15: Lowland woodland (1986 and 2016) Area (1,000 ha) Change (%) Area (1,000 ha) Change (%) Region 1986 2016 1986–2016 Region 1986 2016 1986–2016 Afar 3,006.6 2,786.8 –7% Afar 571.1 342.7 –40% Amhara 3,204.6 2,826.6 –12% Amhara 1,503.5 1,082.0 –28% Benishangul-Gumuz 3,844.9 3,545.5 –8% Benishangul-Gumuz 2,801.7 1,767.7 –37% Dire Dawa 49.3 47.2 –4% Dire Dawa 17.4 14.1 –19% Gambella 2,185.2 2,139.0 –2% Gambella 667.3 665.6 0% Hareri 3.6 3.9 7% Hareri 2.1 1.0 –54% Oromia 12,053.4 11,631.5 –4% Oromia 7,679.1 5,225.2 –32% SNNPR 3,939.6 4,136.5 5% SNNPR 2,035.6 1,935.7 –5% Somali 27,015.8 26,872.4 –1% Somali 14,554.3 11,487.5 –21% Tigray   1,928.2 1,676.1 –13% Tigray   856.2 397.7 –54% Total   57,231.3 55,665.4 –3% Total   30,688.3 22,918.9 –25% Source: WB calculations based on data from WLRC, ESA, and DevSeed. Source: WB calculations based on data from WLRC, ESA, and DevSeed. *Grazing and forage land combines three LULC categories: woodland, shrub and bush, and grassland. Annex 1: Environment and Land Use Tables   149 Table A.16: Lowland shrub and bush (1986 and 2016) Table A.19: Lowland swamp and marsh (1986 and 2016) Area (1,000 ha) Change (%) Area (1,000 ha) Change (%) Region 1986 2016 1986–2016 Region 1986 2016 1986–2016 Afar 1,471.7 1,476.4 0% Afar 81.0 19.9 –75% Amhara 1,237.3 1,299.0 5% Amhara 5.5 0.3 –94% Benishangul-Gumuz 767.3 1,241.2 62% Benishangul-Gumuz 0.1 0.9 643% Dire Dawa 15.7 17.1 9% Dire Dawa — — — Gambella 847.2 789.7 –7% Gambella 202.7 182.0 –10% Hareri 1.1 2.0 93% Hareri — — — Oromia 3,242.7 4,883.3 51% Oromia 78.9 20.6 –74% SNNPR 1,451.5 1,527.5 5% SNNPR 27.1 45.6 68% Somali 9,467.4 11,300.5 19% Somali 82.6 28.9 –65% Tigray   934.7 818.4 –12% Tigray   0.0 0.0 –3% Total   19,436.5 23,355.0 20% Total   478.0 298.2 –38% Source: WB calculations based on data from WLRC, ESA, and DevSeed. Source: WB calculations based on data from WLRC, ESA, and DevSeed. Table A.17: Lowland grassland (1986 and 2016) Table A.20: Lowland waterbody (1986 and 2016) Area (1,000 ha) Change (%) Area (1,000 ha) Change (%) Region 1986 2016 1986–2016 Region 1986 2016 1986–2016 Afar 963.8 967.7 0% Afar 72.9 78.3 7% Amhara 463.9 445.7 –4% Amhara 0.7 12.8 1,614% Benishangul-Gumuz 275.9 536.6 94% Benishangul-Gumuz 0.0 0.4 3,145% Dire Dawa 16.2 16.1 0% Dire Dawa — — — Gambella 670.7 683.7 2% Gambella 1.1 2.4 119% Hareri 0.5 0.9 91% Hareri — — — Oromia 1,131.7 1,523.1 35% Oromia 69.3 152.9 121% SNNPR 452.5 673.3 49% SNNPR 84.9 98.6 16% Somali 2,994.1 4,084.5 36% Somali 11.7 3.1 –74% Tigray   137.3 460.1 235% Tigray   0.1 1.1 1,618% Total   7,106.4 9,391.5 32% Total   240.7 349.5 45% Source: WB calculations based on data from WLRC, ESA, and DevSeed. Source: WB calculations based on data from WLRC, ESA, and DevSeed. Table A.18: Lowland bare land (1986 and 2016) Table A.21: Lowland settlement (1986 and 2016) Area (1,000 ha) Change (%) Area (ha) Change (%) Region 1986 2016 1986–2016 Region 1986 2016 1986–2016 Afar 6,125.7 6,284.0 3% Afar 235.0 2,482.0 956% Amhara 190.4 322.7 70% Amhara 2,029.0 5,490.0 171% Benishangul-Gumuz 15.8 13.1 –17% Benishangul-Gumuz 739.0 1,614.0 118% Dire Dawa 7.1 8.2 16% Dire Dawa 717.0 2,769.0 286% Gambella 4.8 14.9 211% Gambella 937.0 1,348.0 44% Hareri 0.0 0.9 343.5% Hareri 32.0 20.0 –38% Oromia 225.6 313.6 39% Oromia 3,415.0 5,000.0 46% SNNPR 78.6 104.3 33% SNNPR 466.0 4,765.0 923% Somali 2,337.3 2,750.8 18% Somali 2,449.0 5,583.0 128% Tigray   116.5 194.4 67% Tigray   374.0 2,671.0 614% Total   9,101.7 10,006.9 10% Total   11,393.0 31,742.0 179% Source: WB calculations based on data from WLRC, ESA, and DevSeed. Source: WB calculations based on data from WLRC, ESA, and DevSeed. Table A.22: EHAIA land concession details Average Number area Total area Median Standard Maximum Region of leases (ha) (ha) (ha) deviation (ha) (ha) Benishangul-Gumuz 446 549 244,930 325 1,112 20,029 Gambella 520 800 416,252 528 865 9,990 SNNPR   20 3,485 69,694 1,001 4,648 18,516 Total   986 741 730,876 501 1,242 20,029 Source: Ethiopian Horticulture and Agriculture Investment Agency (EHAIA). Figures reflect all types of concession and cover the majority of land concessions issued between 2015 and 2017 in these three regions. Information on land concessions from other regions was not provided. Table A.23: Large farms and concessions by region Region Area (ha) % of lowland area Afar 84,889 0.90% Amhara 16,998 0.42% Benishangul-Gumuz 254,448 5.50% Gambella 437,932 14.81% Oromia 24,595 0.17% SNNPR 134,369 2.20% Somali 12,542 0.04% Tigray   44,060 1.80% Total   1,009,833 1.36% Source: WB calculations using information on land concessions from Ethiopian Horticulture and Agriculture Investment Agency (EHAIA) plus DevSeed. Areas reflect areas identified as land concessions (from EHAIA data), DevSeed, or where these datasets overlap. Table A.24: Prosopis expansion in Afar 2000–2016/17 2000 2016/17 % Change 2000–2016/17 Prosopis High Prosopis High Prosopis High present infestation present infestation present infestation Area (ha) 828,411 41,529 1,173,300 306,409 42% 638% % of lowland area   8.8% 0.4% 12.4% 3.2% Source: WLRC. Annex 1: Environment and Land Use Tables   151 Poverty and Vulnerability in the Ethiopian Lowlands Annex 2: Methodology for Measuring Household Vulnerability The method used to estimate vulnerability to pov- Substituting (2) and (3) into (1) yields the regression erty is based on Gunther and Hartgen (2009) who equation integrated a two-level hierarchical model (e.g., Hox, g 00 1 g 01Zj 1 (g 10 1 g 11Zj)Xij lncij 5  2010) into Chaudhuri’s (2002, 2003) method to 1 u0j 1 u1jXij 1 eij (4) estimate vulnerability from cross-sectional or short panel data resolving the problem of missing lengthy Or panel data (see, e.g., Chaudhuri, Jalan, & Suryahadi, g 00 1 g 01Zj 1 (g 10 1 u1j)Xij 1 lncij 5  2002, for applications). The two key features of g 11ZjXij 1 u0j 1 eij (4a) the model are (i) the error term in the consump- tion regression or the unexplained variance in the Equation (4a) implies that the hierarchical model consumption of otherwise identical households is as specified amounts to a random coefficient (or decomposed into household-specific and commu- random slopes) model. Thus, there are three errors nity (woreda)-specific shocks to household con- terms to be estimated u0j, u1j, and eij, whereby cap- sumption; and (ii) the variance of these two types of tures the idiosyncratic shocks u0j and u1j Xij capture shocks is then modelled as a function of observable the community covariate shocks. Specifically, household and community (woreda) characteristics. u0j captures the direct effect of covariate shocks Specifically, let i = 1, ..., N denote household at affecting the intercept of each community and thus level one and j = 1, ..., J denote communities (wore- all households in the same community in the same das) at level two, with households being nested manner, u1j Xij and captures the indirect effect of within communities (woredas). Consumption of covariate shocks. In practice, a likelihood ratio test household i in community j is specified as may be used to test whether a random coefficient model is more appropriate than a simple random lncij 5 b 0j 1 b 1jXij 1 eij (1) effects model, where only the direct effect of the The coefficients (both constant term and slopes) community covariate shocks is taken into account. of each community are assumed to be affected by community observed (Z) and community unob- Step 1: Equation (4a) above is estimated using Sta- served factors denoted by u0j and u1j ta’s command for mixed-effects maximum likeli- hood regression. b 0j 5 g 00 1 g 01Zj 1 u0j (2) b 1j 5 g 10 1 g 11Zj 1 u1j (3) 152 2 and u 2 of Step 2: Regress squared residuals eij There are two main caveats associated with the 0j Equation (4a) from step 1 above on Xij and Zj method that are important to bear in mind: 2 5 e2 5 q 1 q Z 1 q X Z • σ e • The cross-sectional variance is assumed to esti- ij ij 0 1 j 3 ij j 2 5 u2 5 t 1 t Z mate future intertemporal variance in consump- • σ u 0j 0 1 j 0j tion. This implicitly assumes that the variance 2 • σ u0j1eij 5 (u0j 1 eij )2 1 p0 1 p1Zj 1 p3XijZj in consumption of a particular household is con- stant over time. Assuming consumption is log-normally distributed, • There is no measurement error in consumption one can estimate the probability of consumption and/or no time invariant of unobserved fixed falling below the poverty line z and using a thresh- effects. If there is measurement error, this can old for this probability (e.g., 50%) to define who is lead to an overestimate of the variance of con- vulnerable to poverty (i.e., a household is vulnerable sumption, especially to an overestimation of the if υij > 0.5). The poverty threshold is set at 20 per- 8 impact of idiosyncratic shocks on consumption. cent, which is equivalent to having a 50 percent In principle, this can be addressed by checking probability of falling below the poverty line in any the robustness of the vulnerability estimates given year. under alternative assumptions about the propor- lnz – lncij tion of the estimated variance of consumption 8 υij = P(lncij  lnz | X, Z) = f a a 8 that is due to measurement error. x _ 2 σu0j1eij ® 2 If instead of σ u0j1eij , in the denominator, we use 8 2 → vulnerability to poverty from covariate or • σ u0j woreda level shocks 8 2 → vulnerability to poverty from idiosyncratic • σ eij shocks Annex 2: Methodology for Measuring Household Vulnerability   153