© 2024 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW, Washington, DC 20433 Telephone: 202-473-1000; Internet: www.worldbank.org This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Nothing herein shall constitute or be considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Rights and Permissions Attribution—Please cite the work as follows: World Bank. 2024. Ethiopia Poverty and Equity Assessment. Washington, DC: World Bank. Translations—If you create a translation of this work, please add the following disclaimer along with the attribution: This translation was not created by The World Bank and should not be considered an official World Bank translation. The World Bank shall not be liable for any content or error in this translation. Adaptations—If you create an adaptation of this work, please add the following disclaimer along with the attribution: This is an adaptation of an original work by The World Bank. Views and opinions expressed in the adaptation are the sole responsibility of the author or authors of the adaptation and are not endorsed by The World Bank. Third-party content—The World Bank does not necessarily own each component of the content contained within the work. The World Bank, therefore, does not warrant that the use of any third-party-owned individual component or part contained in the work will not infringe on the rights of those third parties. The risk of claims resulting from such infringement rests solely with you. If you wish to reuse a component of the work, it is your responsibility to determine whether permission is needed for that reuse and to obtain permission from the copyright owner. Examples of components can include, but are not limited to, tables, figures, or images. All queries on rights and licenses should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; e-mail: pubrights@worldbank.org. Cover design & layout: Cybil Nyaradzo Maradza ii CURRENCY EQUIVALENTS Exchange Rate Effective as of June 28, 2024 Currency Unit = ETB (Ethiopian Birr) ETB 57.74 = US$1.00 Fiscal Year = September to August ACRONYMS AND ABBREVIATIONS Acronym Definition ACLED Armed Conflict Location & Event Data ATT Average Treatment Effect on the Treated CHIRPS Climate Hazards Group InfraRed Precipitations with Stations CPI Consumer Price Index DTM Data Tracking Matrix ECHO European Civil Protection and Humanitarian Aid Operations ESPS Ethiopian Socioeconomic Panel Survey ETB Ethiopian Birr FAO Food and Agriculture Organization GAEZ Global Agro-Ecological Zoning GDP Gross Domestic Product HBS Household Budget Surveys HCES Household Consumption and Expenditure Survey HFPS High-Frequency Phone Surveys HoWStat Household Welfare Statistics Survey IPC Integrated Food Security Phase Classification IOM International Organization of Migration LMMIS Large- and Medium-sized Manufacturing Industries Survey LFS Labor Force Survey MPSE Mobile Populations Survey for Ethiopia PPP Purchasing Power Parity OCHA Office for the Coordination of Humanitarian Affairs REER Real Effective Exchange Rate RID Rural Income Diagnostics SCD Systematic Country Diagnostics SME Small and Micro Enterprise SNNP Southern Nations, Nationalities, and People’s SOE State Owned Enterprise WMS Welfare Monitoring Survey iii ETHIOPIA POVERTY AND EQUITY ASSESSMENT Regional Vice President : Victoria Kwakwa Country Director : Maryam Salim Regional Director : Hassan Zaman Practice Manager : Pierella Paci, Rinku Murgai Task Team Leaders : Obert Pimhidzai, Christina Wieser ACKNOWLEDGEMENTS The report was prepared by Obert Pimhidzai, Lead Economist, EECPV; Christina Wieser, Senior Economist, ESAPV; Cesar A. Cancho, Senior Economist, EAEPV; Wondimagegn Mesfin Tesfaye, Economist, EAEPV; Jeremy Aaron Lebow, Young Professional, HSPGE; Lulit Mitik Beyene, Senior Economist, EMFMD; Alemayehu A. Ambel, Senior Economist, DECLS; Tawanda Chingozha, Consultant, EAEPV; Fikirte Girmachew Abeje, Consultant, EAEPV; Musa Hasen Ahmed, Consultant, EAEPV and Manex Bule Yonis, Economist, DECLS. The report draws from the following background papers commissioned for the study. These background papers were authored by Wondimagegn Mesfin Tesfaye (Economist, EAEPV); Musa Hasen Ahmed (Consultant, EAEPV) and Tawanda Chingozha (Consultant, EAEPV) – Weather variability, risk preferences, and household welfare Lulit Mitik Beyene (Senior Economist, EMFMD) – Impacts of exchange rate adjustment in Ethiopia Cesar A. Cancho (Senior Economist, EAEPV) – Rightsizing Estimates of Economic Mobility in Ethiopia Alemayehu A. Ambel (Senior Economist, DECLS) - Poverty dynamics and mobility in Ethiopia from 2019-2022: Evidence from longitudinal data Jeremy Aaron Lebow (Young Professional, HSPGE) – Welfare impacts of conflict in Ethiopia Christina Wieser (Senior Economist, ESAPV) and Aisha Mohammed Abubakar (Consultant, EAEPV) – Has Ethiopia harnessed structural transformation in the labor market? Christina Wieser (Senior Economist, ESAPV) and Eleni Abraham Yitbarek (Consultant, EAEPV) – Impact of Food Inflation on Household Welfare in Ethiopia iv ACKNOWLEDGEMENTS The team gratefully acknowledges the technical collaboration with Ministry of Planning and Development and Ethiopian Statistical Service that enabled analysis of poverty trends. The team also benefited from feedback from peer reviewers: Arden Finn (Senior Economist, EMNPV), Rob Swinkels (Lead Country Economist, EAWDR), Kevin Carey (Program Manager, EFICT), and Hans Hoogeveen (Senior Economist, EAWPV) and other Word Bank staff involved in team consultations – Nistha Sinha (Senior Economist, EAEPV); Vinayak (Vinny) Nagaraj (Senior Economist, EAEM2); Samuel Mulugeta (Economist, EAEM2); Tehmina S. Khan (Lead Economist, Program Leader, EAEDR). This report was prepared with guidance from Pierella Paci (Practice Manager, EAEPV) and Tom Bundervoet (Lead Economist, EAEPV). v ETHIOPIA POVERTY AND EQUITY ASSESSMENT TABLE OF CONTENTS Acronyms and Abbreviations ............................................................................................................................................ iii Acknowledgements ................................................................................................................................................................. iv Executive Summary ................................................................................................................................................................. xii Introduction 1 Background ........................................................................................................................................................................... 1 Recap of poverty challenges: A story of strong but spatially uneven progress ............................. 2 Marrying traditional and non-traditional data for a nuanced understanding of poverty ........... 2 Report outline ...................................................................................................................................................................... 6 Part 1: The Evolution of Poverty and Inequality ......................................................................................... 6 Part 2: Deepening the Understanding of Drivers of Poverty ................................................................. 7 Part 3: Resetting the Agenda for Poverty Reduction ................................................................................ 7 Part 1: Trends, Patterns, and Drivers of Poverty and Inequality 8 Recent socioeconomic developments and implications for household welfare ............................... 9 Multiple shocks left no households untouched ............................................................................................ 9 Growth decelerated in the face of multiple crises, creating unfavorable labor market conditions ........................................................................................................................................................ 11 Job creation lost steam, leaving many people out of work ..................................................................... 12 Household consumption growth declined across the board .................................................................. 13 Implications for the trends and pattern of poverty ...................................................................................... 14 Poverty increased, undoing gains made at the beginning of the previous decade ............................... 14 Poverty increased in all regions except Addis Ababa and Dire Dawa and increased most in the lowlands ................................................................................................................................................................... 19 The poor are concentrated in populous highland areas and are predominantly rural and agricultural with low human capital ................................................................................................................... 21 The differences among the poor are defined by agriculture dependency, connectivity, drought exposure and access to land ................................................................................................................................. 24 vi TABLE OF CONTENTS Drivers of the increase in poverty ........................................................................................................................... 26 Poverty increased due to a generalized decline in welfare across the entire distribution ....... 26 Household income gains were wiped out by inflation, driving the increase in poverty ............. 28 Beyond shocks, low endowments and lack of opportunities contributed to poverty ................. 34 The interplay of shocks, low endowments, and poor opportunities has increased vulnerability .............................................................................................................................................. 35 Social safety nets helped mitigate impacts among the beneficiaries ................................................ 36 Part 2: Deepening the Understanding of Drivers of Poverty 37 Climate change and household welfare in Ethiopia ..................................................................................... 38 Ethiopia has experienced extreme weather events and significant weather variability in the past decade, mainly between 2016 and 2021 ............................................................................................. 38 Household long-term vulnerability to climate shocks is high, more so for the poor ..................... 39 Long-term exposure to climate shocks intensifies poverty .................................................................... 40 Exposure to weather shocks reduce future incomes by lowering agricultural productivity and reducing farmers’ market orientation ................................................................................................................ 43 Agricultural households employ a range of coping strategies and responses that enhance resilience to climate shocks .................................................................................................................................... 47 The impact of climate shocks can be moderated with the implementation of economic reforms .. 48 Conflict and household welfare in Ethiopia ....................................................................................................... 50 Escalation of the conflict in Tigray spread the conflict across the country, with increasing intensity ..................................................................................................................................................... 50 Recent conflict events have been concentrated in urban centers, translating into increased exposure among better off households ............................................................................................................ 51 Conflict had broader impacts ................................................................................................................................ 52 Conflict has driven up internal displacement for longer periods .......................................................... 53 Internally displaced people have limited access to services ................................................................. 55 The majority of IDPs prefer return, but preferences vary across regions ........................................ 57 Structural transformation and household welfare in Ethiopia .............................................................. 59 Sectoral and spatial shifts of labor from agriculture and rural areas have the potential to drive poverty reduction in Ethiopia ..................................................................................................................... 59 Many rural Ethiopians have moved out of agriculture in recent years ............................................... 61 Recent signs of labor shifting away from agriculture belie weaknesses in Ethiopia’s labor market.................................................................................................................................................................... 63 vii ETHIOPIA POVERTY AND EQUITY ASSESSMENT Stalling non-agriculture job creation reflects a faltering transformation in Ethiopia’s economic structure as the state-led growth model reached its limit ...................................................................... 67 While addressing macroeconomic imbalances has short-term costs, it will generate better- quality jobs and higher incomes .......................................................................................................................... 69 Spatial transformation has also been hampered by barriers - some structural - to the movement of labor ...................................................................................................................................................... 72 Part 3: Resetting the Agenda for Poverty Reduction 74 Summary of key challenges for poverty reduction ....................................................................................... 75 Poverty increased due to a combination of shocks and longstanding vulnerabilities in the development model ................................................................................................................................................... 75 Interaction of shocks and vulnerability left scars that make the continuation of the current model untenable .......................................................................................................................................................... 75 The rural-urban gaps persist due to untapped potential in the rural sector ................................... 76 Policy implications ................................................................................................................................................................ 76 Enhancing resilience to shocks ............................................................................................................................ 76 Enhance generation of agriculture surplus ..................................................................................................... 77 Eliminate structural impediments to job creation ....................................................................................... 78 References ................................................................................................................................................................................... 79 Annex 85 Annex 1: Monetary Poverty Measurement Methodology ..................................................................................... 85 Annex 2: Multidimensional Poverty Methodology .................................................................................................. 92 Annex 3: Additional Descriptive Statistics and Regression Results ................................................................ 94 Annex 4: Correlates of Job Quality for the Urban Wage Employed .................................................................. 101 List of Maps Map 1: Spatial incidence of climate shocks, locust invasions, and conflict in Ethiopia ......................... 10 List of Boxes Box 1. Data quality and representativeness of HoWStat 2021 .................................................................... 5 Box 2. Poverty estimation methodology in Ethiopia ......................................................................................... 16 Box 3: What do we know about welfare in Tigray? ............................................................................................ 20 viii TABLE OF CONTENTS Box 4: Estimating the welfare impact of inflation in Ethiopia ....................................................................... 30 Box 5. Estimation of the impact of conflict on household consumption growth ................................. 33 Box 6. Estimation of the impact of long-term environmental risks and household welfare................. 42 Box 7. Estimation of the impact of climate shocks on household risk preferences .......................... 45 Box 8: Following cohorts over time ........................................................................................................................... 62 Box 9: Economic and welfare impacts of exchange rate unification in Ethiopia ................................. 71 List of Tables Table 1: Description of data sources for the Poverty and Equity Assessment ......................................... 3 Table 2: Multidimensional poverty and non-monetary poverty indicators ................................................ 18 Table 3: Composition of the population, poor and non-poor by location, sector, and education level, 2021 ............................................................................................................................................................. 22 Table 4: Demographic characteristics of poor and non-poor households, 2021 ................................... 23 Table 5: Comparison of growth in nominal and real consumption per adult equivalent (Birr), 2015/16- 2021 .................................................................................................................................................... 30 Table 6: Estimated impacts of conflict exposure on household consumption growth ........................ 35 Table 7: Safety Net Coverage in 2021 ......................................................................................................................... 36 Table 8: Estimated impacts of drought exposure and agriculture potential on welfare ..................... 41 Table 9: Descriptive statistics on the conditions of households based on drought shock ................. 43 Table 10: Arrow-Pratt (AP) absolute and Downside (DS) risk aversion by weather shocks ................. 44 Table 11: Estimated impacts of conflict exposure on household consumption growth ........................ 53 Table 12: Relationship between site outcomes and site type and age .......................................................... 57 Table 13: Preference for integration by site type and age ................................................................................... 58 Table 14: Labor market outcomes of migrants .......................................................................................................... 60 Table 15: Impact of migration on factor markets and welfare in origin communities ............................ 60 Table 16: Employment and job growth of subsectors ............................................................................................ 65 List of Figures Figure 1: Share of rural households reporting loss in incomes (%): April–October 2020 .......................... 11 Figure 2: Consumer Price Index, year-on-year % change ................................................................................... 11 Figure 3: GDP growth rate by sector, (%) ..................................................................................................................... 12 Figure 4: Agriculture production growth by crop, (%) ............................................................................................ 12 Figure 5: Trends in key labor market outcomes in Ethiopia ................................................................................ 13 Figure 6: Consumption growth by household welfare ranking .......................................................................... 13 Figure 7: Trends in poverty headcount rates (%), 2010/11 - 2021 ............................................................... 15 Figure 8: Comparison of the population distribution by consumption levels ............................................ 17 ix ETHIOPIA POVERTY AND EQUITY ASSESSMENT Figure 9: Incidence of calorie deficiency (%), 2015/16 - 2021 .................................................................. 17 Figure 10: International comparisons in poverty rate trends: 2014-2021, (%) ..................................... 19 Figure 11: Regional trends in monetary poverty, 2015/16-2021 ................................................................. 20 Figure 12: Composition of the poor by subgroups (%), 2021 ......................................................................... 24 Figure 13: Household characteristics by subgroups of the poor, 2021 (% of household in the subgroup) ................................................................................................................................................... 24 Figure 14: Trends in inequality (Gini coefficient), 2010/11 - 2021 .............................................................. 26 Figure 15: Decomposition of poverty changes due to growth and inequality, 2015/16 – 21 (pp change) ......................................................................................................................................................... 26 Figure 16: Decomposition of Poverty Changes by Intra-Location Changes and Population Shifts Across Locations (pp change), 2015/16 – 2021 ............................................................................ 27 Figure 17: Decomposition of poverty changes in rural areas by changes within income types and shifts across income types, 2015/16 – 2021 .................................................................................. 28 Figure 18: Decomposition of poverty changes in urban areas by changes within employment sectors and shifts across sectors, 2015/16 – 2021 ...................................................................... 28 Figure 19: Impact of incremental inflation on household welfare ................................................................ 29 Figure 20: Rural households net market position by food crop, 2022 ....................................................... 31 Figure 21: Share of households selling at least one cereal crop, 2022 ....................................................... 32 Figure 22: Welfare losses from rising food prices, 2022 ................................................................................... 32 Figure 23: Changes in the distribution of precipitation and temperature, 1951-2020 ...................... 39 Figure 24: Drought evolution and trends, 1980-2021 ....................................................................................... 39 Figure 25: Poverty incidence and long-term drought exposure ..................................................................... 40 Figure 26: The share of the population and poor exposed to drought shocks ........................................ 40 Figure 27: Adoption of agriculture technologies and farming practices .................................................... 48 Figure 28: Impact of climate change scenarios on poverty rates ................................................................... 49 Figure 29: Evolution of conflict in Ethiopia ............................................................................................................... 50 Figure 30: Dispersion of conflict events before and after November 2020 .............................................. 51 Figure 31: Number of days with conflict events by Consumption Quintile ................................................ 51 Figure 32: Internal displacement trends .................................................................................................................. 54 Figure 33: Millions of IDPs by reason for displacement ..................................................................................... 54 Figure 34: Average age of sites by region (December 2022) ......................................................................... 55 Figure 35: Millions of IDPs by settlement type ...................................................................................................... 55 Figure 36: Trends in site service quality attributes ............................................................................................... 56 Figure 37: Preferred solution by region .................................................................................................................... 58 Figure 38: Percentage change in household welfare by sectoral transition types ................................ 59 Figure 39: Sectoral-employment shares ................................................................................................................... 61 Figure 40: Predictive margins for sectoral labor mobility by birth year cohort and location ............ 61 Figure 41: Skill levels and rural-to-urban migration ............................................................................................. 64 x TABLE OF CONTENTS Figure 42: Labor market transitions among rural population aged 18-64.................................................... 66 Figure 43: Trends in wage job quality in urban areas.............................................................................................. 66 Figure 44: Changes in economic structure and labor productivity ............................................................... 67 Figure 45: Trends in exchange rates and inflation in Ethiopia ........................................................................ 68 Figure 46: Number of workers in Large and Medium Scale Manufacturing Firms in Ethiopia ......... 69 Figure 47: Relationship between exchange rates and employment by manufacturing firms ......... 69 Figure 48: Correlation between commodity prices and parallel and official exchange rate in Ethiopia .......................................................................................................................................................... 70 Figure 49: Comparison of the population distribution by consumption levels ........................................ 70 Figure 50: Sectoral breakdown of change in output and employment ....................................................... 71 Figure 51: Marginal effects correlated with migration ........................................................................................ 73 xi ETHIOPIA POVERTY AND EQUITY ASSESSMENT EXECUTIVE SUMMARY INTRODUCTION updates the understanding of poverty and inequality in the country, using new data collected from 2021. Ethiopia has seen many changes since 2016, This data was collected amidst security concerns, which until now, has been the reference year for which posed challenges during the data collection data about the level and pattern of poverty in the process. Despite these challenges, data quality country. The narrative around poverty was that years checks have verified that the collected information is of high growth resulted in a significant reduction in reliable and representative of the country, excluding poverty, but by less than expected because growth areas that were inaccessible, such as Tigray. The was uneven between rural and urban areas which PEA updates statistics on poverty rates, inequality, received most of the gains from growth and there the poverty profile, and identifies the drivers of was a slow shift of labor from agriculture into the these trends (Part 1). It provides an in-depth fast-growing segments of the economy. Since understanding of the key drivers of poverty in the 2016, GDP per capita growth has decelerated—to country (Part 2) and charts the course for reducing 4.6 percent during 2016-2022 compared to nearly poverty in the years to come (Part 3). Below are 7.4 percent during 2010-2016—not least because some high-level messages drawn from the analysis of multiple crises, including a global pandemic, presented in the seven chapters of the report. droughts, locust infestation, conflict, and market Additional details are accessible in background shocks. This Poverty and Equity Assessment (PEA) papers accompanying the report. xii EXECUTIVE SUMMARY KEY FINDINGS Figure ES1. Consumer Price Index, year-on-year % change The poverty rate increased significantly to around 45 33 percent in 2021, due to a combination of many 40 factors that led to declining incomes across 35 30 the entire socio-economic spectrum. Nearly all 25 households experienced at least one major shock, 20 though the type of shocks were different depending 15 10 on where they lived. These shocks slowed the 5 economy’s pace of growth and severely worsened 0 Jul-16 Oct-16 Jan-17 Apr-17 Jul-17 Oct-17 Jan-18 Apr-18 Jul-18 Oct-18 Jan-19 Apr-19 Jul-19 Oct-19 Jan-20 Apr-20 Jul-20 Oct-20 Jan-21 Apr-21 Jul-21 Oct-21 Jan-22 the economic situation of households across the entire socio-economic spectrum, resulting in a General Food Nonfood significant increase in poverty even as inequality Source: Ethiopian Statistical Service. declined. The increase in poverty can be directly attributed to shocks but only partly, as it is also Against the background of these shocks, reflective of underlying structural weaknesses economic growth decelerated by 4 percentage in Ethiopia’s growth model. Even though social points during 2016-21 compared to the preceding protection and adaptive measures are crucial, the 5 years, and job creation stalled. Previously fast- shocks left scars that can only be patched with major growing segments of the economy saw the biggest structural reforms to put the country back on the decline in growth. The annual growth rate of the path for sustainable growth and poverty reduction. industry sector decelerated from an average of 26.8 percent during 2010-16, to 10.5 during 2016- 1. Multiple shocks have left no household 2022, while growth in the services sector declined untouched and worsened the economic by 1.2 percentage points across these two periods. situation across the country. The labor market situation deteriorated in many ways too. Labor force participation dropped from The multiple shocks experienced since 2018 86 percent in 2013 to 74 percent in 2021, driven affected nearly all households in the country. by a decline in rural areas, while unemployment About 91 percent of the population either nearly doubled to 9 percent in 2021 (Figure ES2). experienced droughts, locust infestation, floods, Overall, the economy created less than 1.2 million conflict, or a combination of them, which taken jobs during a period when the out of school, working together, covered much of the country though their age population expanded by 5 million, signaling a individual effects were localized. For instance, deficit of job creation. The industry sector shed half lowland areas experienced severe droughts for a million jobs and wage job creation by the private three consecutive years in some areas, at a time sector came to a standstill. In comparison, 8.5 when conflict escalated in Northen Ethiopia million jobs were created during 2005-13, of which because of the Tigray conflict (Figure ES1). At more than 4 million, including 1.7 million wage the same time, households were affected by the jobs, were created outside the agriculture sector. COVID-19 pandemic which caused severe income Agriculture production grew in aggregate, but losses in 2020. Rising inflation followed in 2021 this was insufficient to make up for an expanding and 2022. Food inflation went on to reach a 10- population. Over 2019-22, agriculture GDP per year peak of 42 percent in February 2022 and capita did not grow at all. Survey data shows a consumer prices cumulatively tripled between decline in incomes from crop cultivation in real 2016 and 2022. terms during 2019 and 2022. xiii ETHIOPIA POVERTY AND EQUITY ASSESSMENT Figure ES2. Trends in key labor market Figure ES3. Trends in poverty headcount rates outcomes in Ethiopia (%), 2010/11 - 2021 Labor force participation rates by location, (%) 37% 95% 33% 29% 30% 27% 27% 90% 25% 86% 85% 19% 16% 80% 75% 78% 74% 70% 65% National Urban Rural 60% Poverty line using CPI deflator 1999 2005 2013 2021 2011 2016 2021 Urban Rural National Source: Authors’ estimates based on HCES 2010/11, 2015/16 and Source: Authors’ estimates based on LFS 2005; 2013; 2021. HoWStat 2021. 2. Previous gains in poverty reduction came poor were affected more. Consumption growth undone owing to a deficit of growth across the cumulatively declined by 13.2 percent nationally. entire socio-economic spectrum. It declined by much more among better-off households and in urban areas in general. This The poverty rate increased by 9 percentage reduced inequality as measured by the Gini points between 2016 and 2021, setting poverty coefficient from 0.33 in 2016 to 0.27 in 2021, back to levels seen in 2010 (Figure ES3). The owing to a drop in inequality within urban areas poverty rate was 33 percent in 2021, compared and a reduction in the gap between urban and rural to 24 percent in 2015/16 and 31 percent in 2010 areas. Statistical decomposition methods suggest with the Tigray region excluded from estimates that rising poverty was entirely driven by declining across all years since data was not collected there consumption growth. Although consumption in 2021. Rural areas performed worse in terms declined by less in rural areas, poverty rates of the increase in poverty rates, which rose by increased more because many non-poor rural 10 percentage points compared to urban areas households were just above the poverty line in where poverty increased by 3 percentage points. 2016 - unlike in urban areas where consumption The poverty rate increased across all regions levels were significantly higher than the poverty except for Addis Ababa and Dire Dawa. In absolute line. The decline in welfare, while consistent with terms, poverty rates increased much more in rural the deterioration in the labor market, presents a areas compared to urban areas within the same statistical discrepancy with the growth narrative regions. Therefore, differences in living standards from the macro statistics which show that the in Ethiopia continue to be defined by a rural-urban economy still grew at a decent pace. These divide. Rural areas had a poverty rate of 37 percent differences could be due to how changes in the – hence almost double the poverty rate in urban price level over time are adjusted in national areas - and made-up 88 percent of the poor in accounts and household surveys (GDP deflator vs Ethiopia in 2021. This is 10 percentage points CPI), or the sources of growth but these do not fully more than rural areas’ share in the population. explain the difference. The increase in poverty owed to declining living 3. The reversal of households’ fortunes was standards across the entire welfare distribution, not only caused by shocks, but also reflective while inequality declined because the non- of limitations in the country’s growth model. xiv EXECUTIVE SUMMARY High inflation, shocks, and limited transformation than households in less FAA. Meanwhile, each day in the labor market were the primary drivers of an active conflict exposure directly contributed of the increase in poverty. Both household to a 3.3 percent decline in consumption among consumption and incomes doubled in nominal households in Afar, Amhara and Benishangul- terms, but declined in real terms because incomes Gumuz, which (excluding Tigray) are the other did not rise fast enough to compensate for the regions most impacted by Northern Ethiopia conflict. increase in prices. Empirical estimates suggest that In other words, a household living in a location that incremental food inflation between 2019 and 2022 experienced 7 days of battles, strikes and/or other reduced household consumption by 22 percent, forms of violent events would experience close reflecting the impact of the rise of the general to 17.5 percent decline in consumption. Lastly, price level on poverty (Figure ES4 and Figure ES5). having few people engaged in work – because of a However, food prices rose faster than non-food combination of shocks and other structural factors prices, yet rural areas were still affected by rising – reduced household consumption. For example, prices because most households are subsistence the consumption of a household with just one of two orientated, generally self-sufficient in the production adults working would be 47 percent less than the of cereals. Between 75 and 89 percent of producers consumption of a similar household in which both of maize, teff and sorghum are self-sufficient, adults are working. and net buyers of food outnumber net sellers of food among rural household. Recent exposure Not all the increase in poverty is explained by to droughts further reduced consumption by at the impact of recent shocks. Instead, shocks least 5 percent, on average. Consumption among amplified existing vulnerabilities and showed households in Favored Agriculture Areas (FAA) i.e., limits to the country’s growth model. For those with low long-term exposure to geographical example, data from the Urban Employment and and environmental risks – is 45 percent higher Unemployment Survey (UEUS) shows that job Figure ES4. Impact of incremental inflation on Figure ES5. Rural households net market position household welfare by food crop, 2022 Share of rural households (%) 100 -0.01 80 60 40 -0.18 -0.21 -0.21 -0.22 -0.22 20 -0.21 0 -0.29 -0.30 Poorest Poor Middle Rich Richest Poorest Poor Middle Rich Richest Poorest Poor Middle Rich Richest Poorest Poor Middle Rich Richest Poorest Poor Middle Rich Richest -0.34 -0.35 -0.38 Teff Barley Wheat Maize Sorghum Net seller Net buyer Self-sufficient Total Food Non-food Total (bottom 40 pct) Source: Authors’ estimates based on ESPS 2021/22. Notes: A household’s net market position is defined based on the net buyer National Rural Urban ratio (NBR) of a food item which is calculated as net production (production- Source: Authors’ estimates based on ESPS 2018/19; 2021/22. consumption) divided by total household consumption. It expresses the household Notes: The dependent variables are total consumption (total), food food production and consumption gap relative to a household’s expenditure. consumption (food), nonfood consumption (non-food), and total consumption Households are classified based on their NBR as follows: Net Buyers (NBR < for the bottom 40 percent (total bottom 40 pct). -0.05); Net Sellers (NBR>0.05) and Self-sufficient (-0.05 < NBR < 0.05). xv ETHIOPIA POVERTY AND EQUITY ASSESSMENT creation in the industry sector and services sectors unsustainable, exposing households more to the already stalled from 2016 onwards (Figure ES6). impact of market shocks while the government Less than 50,000 non-agriculture jobs were created could not afford to fully adjust social assistance in urban areas between 2016 and 2018. The Large- benefits to inflation. and Medium-sized Manufacturing Industries Survey (LMMIS) shows that employment among these Inflation ended up hurting rural farmers, instead types of enterprises started declining in 2018 of them benefiting from rising food prices due to before COVID-19 hit in 2020, while the growth the legacy of distortionary agriculture policies. rate in the industry sector had already halved. Years of policies emphasizing food security – Overall, the state-led development model relied for good reasons given the country’s history of on suppression of market incomes, especially in famine – succeeded in increasing food security but the agriculture sector, followed by redistribution distorted the market by relying on market controls that the government could no longer afford after that reduced prices for farmers and distorted input the COVID-19 pandemic and the costly conflict markets. This resulted in the underdevelopment in Tigray. of agricultural technologies in some key areas (e.g., wheat) and the suboptimal availability of Structural weaknesses inherent in the country’s inputs such as fertilizers. Together, these factors growth model had started to bind. For example, limited surplus generation and farmers’ market macroeconomic imbalances like exchange orientation. Rising inflation therefore contributed rate misalignment proved a key constraint for to welfare losses in rural areas because most rural businesses. Estimates suggest that a real exchange households do not produce a marketable surplus, rate appreciation of the magnitude observed in hence they were not positioned to take advantage Ethiopia during 2016–2021 could have reduced of rising food prices. Rural households also faced employment among manufacturing firms limited access to off-farm opportunities in part proportionate to their degree of export orientation. due to restrictive spatial policies that inhibited Reduced fiscal space after the Northern Ethiopia the efficient functioning of land and labor markets conflict broke out made many of the subsidies – the two factors of production most available to the poor. As a result, migration has been low in Ethiopia, limiting an important trigger for rural Figure ES6. Trends in key labor market outcomes in Ethiopia transformation and closing a channel for access to non-farm opportunities for rural youth who became Net job creation by sector, location, and gender, (Million people) unemployed instead, and a coping mechanism for 9 rural households to climate shocks. 8 6 4. Scarring from the shocks poses a danger to recovery. Millions 5 3 0 Job market recovery showed signs of being -2 scared from the effects of shocks. Evidence -3 from COVID-19 monitoring surveys show that Ethiopia Urban Rural Male Female Ethiopia Urban Rural Male Female employment in the industry sector failed to recover from the job losses at the onset of the pandemic 2005-13 2013-21 when 8 percent of workers lost their jobs and a third of household enterprises closed. Both the Urban Rural Total number of households operating household Source: Authors’ estimates based on LFS 2005; 2013; 2021. xvi EXECUTIVE SUMMARY enterprises and revenues of businesses still in Figure ES7. Internal displacement trends operation had not fully recovered by 2022. Job losses during COVID-19 were mostly experienced Panel A: Nov 2018 – Oct 2020 by women who made up nearly two-thirds of 5 workers who lost their jobs due to COVID-19. Thus, labor force participation rates declined 4 more among women compared to men and female 3 unemployment nearly doubled, from 7 percent in 2 2013 to 13 percent in 2021. Furthermore, the 1 number of people Not in Employment, Education or Training (NEET) increased by 5 million. Ethiopia 0 May-18 Sep-18 Jan-19 May-19 Sep-19 Jan-20 May-20 Sep-20 Jan-21 May-21 Sep-21 Jan-22 May-22 Sep-22 also lost market access for its garments and textiles exports to the US after its suspension to AGOA, which threatens recovery of jobs in this Source: Authors calculations using IOM DTM Site Assessments and sector which has been an important source of Emergency Site Assessments (which covered parts of Tigray, Amhara, and Afar throughout 2021). jobs for women. Notes: Each blue point in Panel A and red dot in Panel B represents a separate site assessment. Site Assessments between March-August 2022 were Exposure to climate and conflict shocks can have excluded from Panel A because they did not include Tigray. Key informants persistent effects beyond the year of the shock. are used to identify sites with a reported 20 or more IDP households, then site visits and focus group discussions are conducted to estimate the number and Agricultural households in Ethiopia have a high characteristics of IDPs in each site. Each Site Assessment (SA) round, which aversion to downside risks to output variability typically occurs 4 times a year, presents an estimated snapshot of the IDP that increases with exposure to droughts and lasts situation in the country. However, it is not necessarily a representative sample for up to two years after the drought occurs. This of IDPs because coverage of many sites is severely limited by inaccessibility due to conflict, sites with less than 20 IDP households are excluded, and risk aversion partly explains lower adoption of self-settled IDPs in urban areas are often missed. These numbers should marketed inputs and increased land allocation therefore be seen as estimates that are lower bounds. The November 2022 to cereal production among households who Site Assessment includes Tigray and was implemented through June 2023. previously faced hot temperatures. Thus, high Given limited data coverage, these should be seen as lower bounds. exposure to shocks could be a strong disincentive for households to commercialize. Both drought POLICY IMPLICATIONS and conflict reduce households’ productive capital in terms of human capital and physical assets. In the current context, the key priories for poverty Drought exposure increases the chance that at reduction are (i) strengthening households and least one child in a household will be stunted by the economy’s resilience to shocks, (ii) increasing 7 percent while by its nature, conflict destroys the generation of agriculture surplus, and (iii) people’s assets and disrupts livelihoods, not least addressing spatial and economic policy driven through displacement. The number of displaced structural impediments to job creation and access people in Ethiopia rose to 4.4 million people at to better economic opportunities. the height of the conflict (Figure ES6). Livelihood recovery among displaced populations is slow 1. Strengthen resilience to shocks. and conflict exposure has long lasting effects that can extend intergenerationally. Human The high vulnerability of households to shocks capital losses due to conflict also lower people’s and the impacts this had on poverty, necessitates lifetime productivity and earnings and reduce the need to strengthen households’ resilience to intergenerational economic mobility. shocks in three ways: xvii ETHIOPIA POVERTY AND EQUITY ASSESSMENT i. Slowing the onset or impact of shocks to promote agriculture commercialization. Most at entry – This requires three types of of these measures have been discussed in detail interventions. One set of interventions are in the Ethiopia Rural Income Diagnostics (World investments to increase productive assets of Bank, 2022b). They include: households and communities, which range from infrastructure investment in irrigation i. Reducing market distortions to trigger a and land structures, natural resource supply response – by eliminating marketing management and skills development to controls that blunt price signals to farmers such increase households’ adaptability. The other as export controls and marking restrictions for set of interventions focuses on developing commodities. and promoting the adoption of Climate Smart Agriculture technologies and strategies. The third ii. Increasing availability and adoption set of interventions focuses on prevention and of advanced agriculture inputs and preparedness, which includes enhancing early technologies – by liberating input markets to warning systems for households to take adaptive promote a greater role of the private sector measures to minimize the impact of shocks. in agriculture technology development, input ii. Reducing the impact on incomes once they production, and distribution to increase timely occur – through establishing/ expanding availability of the right type of inputs. mechanisms to finance crisis response (e.g., destocking and school feeding programs) in iii. Optimizing crop cultivation choices and response to droughts; expanding the coverage incentivizing production of commercial and range of consumption smoothing measures crops – through the adoption of a plurality such as shock responsive social safety nets of agriculture extension services and shifting and access to credit and; establishing market in messaging to encourage a shift towards mechanisms to moderate volatility (e.g., commercial crops and optimize crop cultivation warehousing receipts). choices to land suitability. iii. Facilitating faster and full recovery from Other sets of measures include those mitigating shocks – through investments for livelihood the impact of climate shocks discussed under restoration and reconstruction which applies the priority intervention to increase resilience to for both climate and conflict shocks (e.g., shocks, as these can also influence household infrastructure rehabilitation and re-stocking agriculture production decisions based on their and input support programs) and promoting impact on risk preferences. the adoption of insurance products (e.g., livestock insurance). 3. Eliminate structural impediments to job creation. 2. Increase generation of agriculture surplus. A more fundamental challenge for poverty The poor are increasingly concentrated in reduction is the lack of better economic rural areas, most of them in high agricultural opportunities. This is evidenced by the declining potential areas. Many rural poor were not able pace of job creation, with a net reduction of jobs to capitalize on rising food prices. This points to in the industry sector, stalled private sector wage the necessity of interventions to increase market job creation, and the exit from the labor market surplus generation among rural households and by people – women in particular – facing limited xviii EXECUTIVE SUMMARY opportunities in the context of stalling structural in the financial sector to direct more lending transformation and the general decline in the towards the private sector. quality of jobs as trends in the job quality index showed. Evidence also suggests that the poverty ii. Reducing barriers to entry and state impacts of climate change can be moderated with dominance in the economy – by promoting the implementation of structural reforms that market neutrality and reducing foreign entry enhance productivity under all climate change restrictions in markets with high potential for scenarios. The necessary reforms have been reorganizing agriculture value chains (e.g., discussed in detail in the Systematic Country permitting foreign entry into wholesale and Diagnosis for Ethiopia (SCD, 2024). They include retail markets). the following: iii. Reducing barriers to labor mobility – by i. Macro-fiscal stabilization – by eliminating eliminating burdensome administrative macro policy distortions that undermine private procedures for migrants (e.g., household investment. Key among them is addressing registration requirements) and reducing job the exchange rate misalignment, liberalizing search costs by enhancing job intermediation interest rates, and reducing state dominance and employment promotion services. xix ETHIOPIA POVERTY AND EQUITY ASSESSMENT INTRODUCTION BACKGROUND However, Ethiopia has been a growth champion in the two decades up to 2020. Its GDP increased Ethiopia is the second most populous country in by 6-fold to PPP$ 333 billion (US$126.7 million) in Africa, with a population of 118 million people, 2022, which ranks as the fifth largest economy in mainly rural (78 percent), young with nearly 39 Africa. This is also by virtue of its population size percent of them children under the age of 14 because its GDP per capita of PPP $2,698 (US$ 1027 years old, and ethnically diverse – with more in 2022) ranks it as the 19th in Africa. Ethiopia’s than 80 ethnic groups. It is landlocked, and growth has been achieved through a state-led primarily agrarian, though two-thirds of its land growth model, with heavy public investment – is comprised of semi-arid and arid areas that are that for a time made the country one of the fastest sparsely populated and carry less than 9 percent of growing in Africa. Ethiopia made concerted efforts the total population. However, it has wide ecological to industrialize, through investments in industrial diversity, with 6 traditional ecological zones that parks in particular, which saw the industry sector’s can be subdivided into 33 agro-ecological zones. share in GDP double to 29 percent during the last The country is situated in a volatile region - sharing decade. However, its exports remain dominated borders with Djibouti, Eritrea, Kenya, South Sudan, by agriculture commodities, with coffee as its Sudan, and Somalia – which has high exposure to largest export, though manufacturing exports also weather shocks and a history of famines. rose over time, driven by the garment and textile 1 INTRODUCTION sectors. This Poverty Assessment looks at how by differences between urban and rural areas poverty evolved in this context. rather than differences within regions. Inequality of opportunity for children also showed large. Recap of poverty challenges: A story of Regional variations and whether a household is strong but spatially uneven progress rural or urban, rather than the level of wealth of the household, explained differences in primary school The last Poverty Assessment for Ethiopia was completion rates, secondary school enrolment completed in 2020, based on data from the rates, and access to electricity within regions. Household Consumption and Expenditure Survey of 2015/16 (HCES 2015/16) which until now, Drawing on more recent official data, this remains the latest publicly available data for Poverty and Equity Assessment (PEA) Report official poverty estimates. The analysis showed updates the knowledge of the poverty situation that Ethiopia made significant progress in reducing in the country, given the many challenges poverty, but less than proportionate to the high pace the country has faced since 2016. The report of economic growth the country experienced. For the shows recent trends in poverty, and provides 16 years leading up to 2015, the economy grew by an in-depth understanding of the drivers of an annualized average of 9 percent, tripling in size. poverty in the country. The PEA also investigates In that time, the poverty headcount rate declined (i) how much and who has been impacted by from 44.2 to 23.5 percent, with a 6-percentage various shocks; (ii) the extent to which structural point drop during the last observed period of 2011- transformation has occurred and growth trickled 16. This translated to a growth elasticity of poverty down to the poor; and (iii) how key policy reforms of -0.46, implying that a percentage increase in in the government’s economic reforms impact growth translated to less than half a percentage household welfare. Most importantly, the PEA decline in poverty between 2011 and 2016. This is discusses what needs to be done to accelerate lower than the growth elasticity in other countries poverty reduction by improving the productive such as Tanzania and Uganda. capacity of poor households to become active contributors to growth while strengthening their Past growth was urban centered, leaving many resilience to shocks. people in low productivity agriculture and did not fully exploit opportunities in rural areas. During Marrying traditional and non- 2011 to 2016, poverty in urban areas decreased traditional data for a nuanced by 11 percentage points compared to 5 percentage understanding of poverty points in rural areas that had higher poverty rates to start with. The annual growth rates in consumption The report primarily relies on the Household in rural areas did not exceed 3 percent, even for Welfare Statistics Survey that was conducted the richest percentile while mean consumption in 2021 (HoWStat 2021). This survey is growth in urban areas was 5.9 percent per year representative of all regions, except Tigray, which and was always above 3 percent, even for the was not accessible due to security reasons (see poorest. Growth was lower for rural households Box 1). It is a combined version of the previous because growth in the agriculture sector was lower Welfare Monitoring Surveys (WMS) and the than in any other sector, yet labor has been slow Household Consumption and Expenditure Surveys to shift out of agriculture. The uneven pattern of (HCES) that we shall collectively refer to as HCES. growth resulted in an increase in inequality in The analysis of trends uses the previous two almost all regions (except Somali) from 2011 to rounds covering the beginning and mid part of the 2016. Inequality became increasingly determined past decade (HCES 2010/11 and HCES 2015/16), 2 ETHIOPIA POVERTY AND EQUITY ASSESSMENT which gives a picture of the socio-economic outcomes is done using the Large- and Medium- development in Ethiopia over the past decade. sized Manufacturing Industries Survey (LMMIS) These surveys are the official data sources for from 1997-2020. These traditional data sources are poverty and welfare monitoring and are the basis combined with several geospatial data in the analysis. for analyzing poverty trends. High frequency quasi-global rainfall data from the Climate Hazards Group InfraRed Precipitations Complementary data is used to inform specific with Stations (CHIRPS) is incorporated to measure topics. This includes panel data from the Ethiopia the impact of climate shocks. Enumeration areas Socio-Economic Panel Survey (ESPS) that was level proxies for suitability and productive potential conducted every two years since 2011/12. The of land are computed from the FAO GAEZ spatial most recent rounds were conducted in 2018/19 data and Galor & Ozak (2016) caloric suitability and 2021/22 with a refreshed sample. The index – both based on the application of machine analysis also uses the National Labor Force Surveys learning to satellite imagery. Incidence of conflict conducted in 2005, 2013, and 2021 alongside the uses data from the Armed Conflict Location & Event Urban Employment and Unemployment Survey Data Project (ACLED) and the IOM’s Displacement (UEUS) to provide a more nuanced description Tracking Matrix (DTM) data is used for analysis of of the labor force. Some analysis of employment internal displacement. Table 1: Description of data sources for the Poverty and Equity Assessment Data Source Data Type and Characteristics Household Welfare Statistics Survey Nationally representative data collected from July to June the following year (HoWStat 2021) in 2010-11 and 2015-16 and from January to December in 2021. The Survey in Household Consumption and Expenditure 2021 did not cover the Tigray Region. The data provides variables on: Household Survey and Welfare Monitoring Survey consumption and other household and individual characteristics, but does not HCES/WMS 2010/11, 2015/16 include agriculture production. Ethiopia Socioeconomic Panel Survey Nationally representative longitudinal data that was first implemented in three (ESPS) – 2011/12, 2013/14, 2015/16; 3 waves from 2011/12-2015/16 and then in two waves in 2018/19 and 2021/22 2018/19, 2021/22 with a refreshed sample. The data provides variables on: Household and individual characteristics; Land and agriculture production at plot level; Geographic characteristics such as population density, connectivity. Ethiopia National Labor Force Survey Nationally representative, cross-sectional survey data, collected over a one-month (LFS) – 2005, 2013, 2021 period. Due to conflict the 2021 survey excluded the Tigray region (which constitutes about 6 percent of the population). Urban Employment and Unemployment Representative of urban areas in each region of Ethiopia. Data is collected over a Survey (UEUS) one-month period. Large- and Medium-sized Manufacturing Covers all manufacturing firms in Ethiopia that engage ten persons or more and Industries Survey (LMMIS)- 1997-2020 use power-driven machines. Captures firm’s characteristics including production, imports/export status, workers by gender. 3 INTRODUCTION Data Source Data Type and Characteristics Displacement Tracking Matrix (DTM) Data collected from sites with 20 or more IDP households - that are identified through key informant interviews -during site visits and focus group discussions as part of Site Assessment (SA) to estimate the number and characteristics of IDPs in each site. Each round typically occurs 4 times a year. Data presents an estimated snapshot of the IDP situation in the country but is not a fully representative sample of IDPs because coverage of many sites is severely limited by inaccessibility due to conflict, IDPs living in sites with less than 20 IDP households are excluded, and self-settled IDPs in urban areas are often missed. The Climate Hazards Group InfraRed High frequency 30 years quasi-global rainfall at 6-hourly, daily, monthly, bi-monthly, Precipitations with Stations (CHIRPS) quarterly & annually. Based on triangulation of Earth based rainfall gauge data and satellite imagery at 0.050 resolution. Armed Conflict Location & Event Data Geocoded data of each conflict event reported in the country by a special team Project (ACLED); www.acleddata.com devoted to mapping conflict in Ethiopia using a combination of Ethiopian and international media reports and local informants (EPO 2021b). This provides an “indirect” measure of conflict based on household proximity to the event. Analysis in this report is restricted to explosions, strikes, battles, or violence against civilians. FAO GAEZ Data Geospatial data, at 9.3 x 9.3 km pixel size (resolution). The Data is extracted at the EA Level. Provides information on agriculture potential. Galor & Ozak (2016) Geospatial data – at 9.3 x 9.3 km resolution measuring potential variation in crop yields across space in calories per hectare per annum. https://ozak.github.io/Caloric-Suitability-Index/ Drought data Palmer Drought Severity Index (PDSI) captures long-term drought as it relies on temperature information and a physical model of water balance. The data covers 1980-2019 and covers the months of May - September (the main agriculture season in Ethiopia). https://hydrology.princeton.edu/getdata.php?dataid=7 FAO Agriculture Stress Index (ASI) measures the proportion of an area that experienced drought in each season (1984-2019), only covers crop areas, based on the Vegetation Health Index (VHI) which is derived from satellite imagery. https:// www.fao.org/giews/earthobservation/asis/index_1.jsp?lang=en 4 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Box 1. Data quality and representativeness of HoWStat 2021 The HoWStat 2021 survey, conducted from January to December 2021, excluded the Tigray region and other conflict-affected areas due to security concerns. This issue was also encountered by other surveys conducted in the country during the same period, such as the Ethiopia Socioeconomic Panel Survey and the Labor Force and Migration Survey, which were officially released by the Ethiopian Statistical Service. For HoWStat 2021, multiple quality checks were performed to ensure data reliability. First, a qualitative assessment of the distribution of households and enumeration areas by region indicated that the household sample was of good quality, except for the Tigray region and other conflict-affected areas, which were not included. In addition, selected data quality indicators that might suggest sub-optimal fieldwork efforts showed no significant concerns in the data collection process. These indicators included histograms and distribution tests (Kolmogorov-Smirnov test, Shapiro-Wilk W test) to detect clustering or peaks in the distribution of the number of households, household size, household composition, the number of food and non-food items, and the number of durables. Additional checks included assessing missing quantities, measurements, units, prices, and expenditures information, along with missing information on key socio-economic characteristics. None of the tests conducted revealed issues of missing values in these key variables. Finally, the distributions of key variables collected in HoWStat 2021 were compared with those from the 2015/16 Household Consumption Expenditure and Welfare Monitoring Surveys (HCES/WMS) at the regional level. The indicators assessed included household expenditure (e.g., the number of food and non-food items consumed), the number of durables/assets owned, household characteristics (e.g., land and livestock ownership, housing ownership), labor market indicators, and household member characteristics (e.g., the share of male respondents, household size, age, sex). The assessment indicates that the distribution of these variables is consistent across the 2015/16 and 2021 surveys by region. 5 INTRODUCTION REPORT OUTLINE drivers of monetary and non-monetary poverty and inequality in the country. The second part of The PEA comprises of seven chapters including the PEA unpacks the drivers of poverty in an in- this introductory chapter, divided into three major depth analysis focusing on shocks and structural parts. The first part of the report provides the basic impediments to poverty reduction. The third and poverty diagnostics to update knowledge of the final part of the report provides strategies for broader country and socio-economic developments poverty reduction in the country. An overview of and their implications for the trends, patterns, and chapters in these parts is provided below. Part 1: The Evolution of Poverty and Inequality This part of the report is aimed at filling knowledge gaps on the evolution of poverty and inequality in Ethiopia and the underlying drivers of the observed trends in poverty and inequality in the country. It will have the following three chapters: Chapter 2: Recent socioeconomic developments and implications for household welfare – to set the stage, this chapter provides a snapshot of recent socio-economic developments that have a strong bearing on the direction for poverty reduction. It links these developments to macroeconomic outcomes and the implications for household income and consumption growth patterns. Chapter 3: Trends and patterns of poverty and inequality – this describes the trends and patterns of monetary and non–monetary poverty. The analysis presented in this chapter uses the HoWStat 2021, which is the official source of poverty estimates recognized by the country, hence most of the cross-sectional and trend analyses will rely on this data. Chapter 4: Drivers of poverty – After presenting data on the trends of poverty and inequality, this chapter starts digging into the drivers of the observed trends focusing both on the determinants of household welfare more broadly and the determinants of being poor and moving in and out of poverty using a combination of statistical decomposition techniques and regression analysis. 6 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Part 2: Deepening the Understanding of Drivers of Poverty The section provides an in-depth look at the key drivers of poverty trends focusing on three key areas that have been or are expected to continue to be drivers of challenges and opportunities for poverty reduction and where there is scope for value addition from new analysis. The three topics selected are climate change, conflict, and structural transformation. Chapter 5: Welfare impact of shocks – Focus on Climate Change - the first deep dive examines the impact of climate-induced shocks and policy responses on household welfare. It depicts the incidence and exposure of households to climate change variability and extremes, shows the temporal and spatial distribution of climate and drought shocks, and the short-term welfare impacts of climate change-induced shocks and applies macro-micro simulations to estimate the long-term poverty impacts of different climate change scenarios. Chapter 6: Welfare impact of shocks: Focus on Conflict – the second deep dive explores the relationship between conflict, internal displacement, and poverty trends. It charts the pattern of conflict and its immediate impact on welfare. It puts a lens on internal displacement by characterizing the socioeconomic conditions of IDPs in Ethiopia, the characteristics of hosting communities and the specific challenges faced by IDPs in camps and hosting communities. Chapter 7: Structural transformation and household welfare – the third deep dive looks beyond the impact of shocks to investigate the structural factors behind the changes in poverty in Ethiopia. The chapter examines the changes in the labor market in Ethiopia to assess the pace and nature of structural transformation, whether this contributes to welfare improvements and what are the policy constraints to acceleration of the transformation. Part 3: Turning Tides for Poverty Reduction The last section of the report synthesizes the main findings from the previous chapters and presents the policy recommendations. Given the focus on shocks and structural transformation, the policy priorities focus on increasing the productive capacity of the poor, based on what needs to be done to (i) increase household employment or use of their labor assets, (ii) increase returns and eliminate distortions to households’ assets, and (iii) insure or protect households’ assets and returns. 7 PART 1 Trends, Patterns, and Drivers of Poverty and Inequality This section discusses the socio-economic developments in Ethiopia since 2016 when the last survey for measuring poverty was implemented. Using new data, the section updates trends on poverty and inequality and identifies key dimensions of welfare heterogeneity, for example across geographical regions. The section aligns evidence to answer a series of questions on who and where are Ethiopia’s poor, how poverty has changed and what are the driver of those changes? The section thus marshals new evidence and deploys statistical techniques to help understand the drivers of poverty changes in the country, especially the effects of recent and contemporary shocks on vulnerability and welfare and the potential implications on poverty reduction efforts in Ethiopia. In relation to this, the section highlights the welfare effects of inflation, conflict, and droughts, along with some underlying drivers of poverty in the country. 8 ETHIOPIA POVERTY AND EQUITY ASSESSMENT RECENT SOCIOECONOMIC expanded beyond Tigray to the neighboring DEVELOPMENTS AND regions of Amhara and Afar. Across the country, IMPLICATIONS FOR 3,153 conflict events occurred, resulting in nearly HOUSEHOLD WELFARE twenty thousand fatalities between November 2019 and the end of 2022. This is around 10 This chapter discusses the many changes in times greater than the number of conflict events Ethiopia and indeed the world, since 2016 when and fatalities in the same period before the war. the last survey to measure poverty in Ethiopia was This created large-scale displacements, and conducted. A new government came into power in vast humanitarian needs in Northern Ethiopia 2018 and launched a Home-Grown Economic Reform (ECHO, 2022). As of June 2021, about 5.5 million Agenda aiming to shift the development paradigm people in Tigray and neighboring Afar and Amhara from a state to private sector led development. (nearly 93 percent of the population in Northern Then the country was hit with a series of global and Ethiopia) were in high acute food insecurity. local shocks that have characterized much of the The conflict also had other economic impacts recent period – ranging from a global pandemic, extending beyond the conflict areas. The country droughts, pest invasions, and an escalation in received USD 1.5 billion less ODA in 2021, while conflict. Though some shocks like droughts and its spending on defense increased, squeezing its conflict were localized, they covered different parts fiscal space for spending on social sectors and of the country such that almost all households were capital investments. Ethiopia was also suspended affected by at least one such shock, not counting the from AGOA, which affected the competitiveness economy wide effects of the COVID-19 pandemic of its nascent garment industry. and inflation. The economy ended up growing at a much slower pace than during the first half of the Meanwhile, the lowland areas were ravaged decade and job creation stalled. The consequence by the compounded risks of droughts, crop was a decline in household welfare across the entire diseases, and pests. The country experienced socio-economic spectrum. a major drought in 2019 covering most parts of the country. This increased in severity in lowland Multiple shocks left no households areas with failed consecutive rainy seasons in untouched. three subsequent years, affecting nearly 7 million people in Oromia, SNNP, Southwest, and Somali This PEA covers the period between 2016 and (OCHA, 2022). Then there were two invasions of 2021, a time when households in Ethiopia desert locusts in 2020. The first invasion, between contended with a series of shocks - often January to May, was reported to have invaded 180- concurrently. Droughts, locust invasions, floods, 240 Woredas primarily in eastern and southern and conflict were localized, but taken together, Ethiopia. The second invasion started in late covered much of the country. Consequently, 91 September and peaked in October-November. It percent of the population in Ethiopia experienced was more severe than the first and ranks amongst at least one of these shocks during 2017-2021 the worst to date (Ilukor & Gourlay, 2021). Data (Map 1). Close to half of the population (48 percent) from a High Frequency Phone Survey (HFPS), experienced multiple shocks in this time. This poly- conducted by the World Bank at the time, shows crisis was unprecedented. that over half of all rural households and nearly 30 percent, experienced locusts in their kebele Local conflict is not new in the country, but it and on their farms, respectively, during the first escalated in 2020 with the eruption of war in locust invasion (World Bank, 2021). At the peak Northern Ethiopia. Throughout 2021, the conflict of the second invasion, 37 percent and 20 percent 9 PART 1: TRENDS, PATTERNS, AND DRIVERS OF POVERTY AND INEQUALITY IN ETHIOPIA of the rural households observed locusts in the both droughts and flooding at various times, while kebele and on their farm, respectively. The western other parts (Benishangul-Gumuz for example) parts of Ethiopia, on the other hand, were prone to experienced an uptick in conflict. Map 1. Spatial incidence of climate shocks, locust invasions, and conflict in Ethiopia Incidence of conflict events within 20km PDSI Drought Frequency: 2019-2021 Conflicts within 20 km (N) Drought Frequency 0 1 2-3 3-10 >10 <20 21-25 26-30 30-35 >35 Crop damages due to locusts: 2012-2021 Incidence of floods: 2016-2021 Crop damage % Flood Frequency <20 20-35 36-50 51-65 >65 0 1-2 3-7 8-15 >15 Source: Authors calculations from various geospatial data sources: Armed Conflict Location & Event Data Project (ACLED); www.acleddata.com; https:// hydrology.princeton.edu/getdata.php?dataid=7; The Climate Hazards Group InfraRed Precipitations with Stations (CHIRPS). The conflict and climate change-induced percent of businesses reported no revenues and shocks occurred when the whole country was 8 percent of wage employees lost their job in the suffering from the COVID-19 pandemic. Its first month of the pandemic (March or April 2020) health impacts in Ethiopia may not have been as (World Bank, 2020). These were accompanied by a devastating as it was in the developed world, but decline in domestic and international remittances. it had a major impact on the welfare of households A significant share of households reported losses through the loss of incomes due to disruptions of in incomes throughout 2020 (Figure 1). While the employment and remittances caused by COVID-19 economic impacts abated during the second half related restrictions (Harris et al., 2021; Yimer et of the year, the pandemic left permanent scars. al., 2020). Estimates from the HFPS show that Real wages for example, declined more the longer over 40 percent of businesses had closed, 32 the pandemic lasted - having initially declined 10 ETHIOPIA POVERTY AND EQUITY ASSESSMENT by 3 percent in April 2020 compared to April which had knock-on effects for agricultural 2019 before eventually falling by 17 percent by production and food security in Ethiopia. December 2020 compared to December 2019. Both employment in industry and the number Growth decelerated in the face of of households running a business never fully multiple crises, creating unfavorable recovered to their pre-pandemic levels. labor market conditions. Inflation, which accelerated in 2021, Amidst the multiple crises, Ethiopia’s great run compounded the crisis. After hovering around slowed down and changed its pattern. In real 20 percent in 2020, inflation rose to above 30 terms, annual GDP per capita growth declined percent in early 2022. Food inflation reached from 6.8 percent during 2011-16 to 4.4 percent 42 percent in February 2022- the highest rate during 2016-21. It has averaged around 3 percent recorded since 2011 - while non-food inflation since 2019. The pattern of growth changed. Annual stood at 23 percent (Figure 2). The rising prices of growth in the industry sector which was growing bread and cereals and oils and fats contributed to fast from a low base (i.e., a tenth of GDP in 2011), about 60 percent of food inflation during this time. declined to 12 percent during 2016-21, shaving Non-food inflation was mainly driven by (i) housing, off 17 percentage points from the previous 5-year water, electricity, gas & other fuels, (ii) furnishings, average growth rate. Because droughts were household equipment & routine maintenance, and localized, the agriculture sector still grew by 4.4 (iii) clothing & footwear. Cumulatively, consumer percent per annum during 2016-21, but this was prices tripled between 2016 and 2021. 2 percentage points less than during 2011-16. Its share of GDP is not declining as fast as in the past. The global fall out from Russia’s invasion of Only services grew at a similar pace compared to Ukraine later aggravated the situation. Within just the preceding five years (Figure 3). After growing a few weeks of the outbreak of the Russia-Ukraine at 13.4 percent and 7.9 percent per annum during war in February 2022, the global prices of wheat, 2011-16, investment and public consumption corn, fertilizer, and oil all soared to unprecedented expenditures respectively, grew at an annual average levels. Ethiopia was highly exposed as a primary of 1.3 percent and 2.6 percent, during 2016-21. The importer of wheat and oil, but also of fertilizer, growth in national savings similarly declined. Figure 1. Share of rural households reporting Figure 2. Consumer Price Index, year-on-year % loss in incomes (%): April–October 2020 change 51.6 45 40 44.8 35 36.5 30 32.3 33.7 25 25.7 20 15 10 5 0 Jul-16 Oct-16 Jan-17 Apr-17 Jul-17 Oct-17 Jan-18 Apr-18 Jul-18 Oct-18 Jan-19 Apr-19 Jul-19 Oct-19 Jan-20 Apr-20 Jul-20 Oct-20 Jan-21 Apr-21 Jul-21 Oct-21 Jan-22 R1 R2 R3 R4 R5 R6 General Food Nonfood Source: World Bank (2020). Source: Ethiopian Statistical Service. 11 PART 1: TRENDS, PATTERNS, AND DRIVERS OF POVERTY AND INEQUALITY IN ETHIOPIA Figure 3. GDP growth rate by sector, (%) Figure 4. Agriculture production growth by crop, between 2015/16 and 2020/21 (%) 30% 13.1 25% 11.1 20% 7.4 6.7 5.8 5.9 5.5 5.4 4.8 4.1 4.1 4.0 3.9 3.5 3.5 3.3 15% 2.4 2.3 2.1 1.7 1.5 0.7 0.7 10% (0.2) (0.4) (1.0) (1.9) (2.1) 5% Teff Barley Wheat Maize Sorghum Pulses Oilseeds Vegetables Root crops Fruits Chat Coffee Hops Sugarcane 0% 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 Agriculture Industry Services GDP Total output Output per capita Source: Authors’ estimates based on NBE annual reports; AgSS 2015/16, 2020/21. Growth in agriculture output, was smaller National Labor Force Survey conducted in 2021 when considering the increase in the country’s shows that there were over 10 million more people population in that time (Figure 3). Analysis of of working age in 2021 than in 2013 when the agriculture output for major food crops, shows that previous LFS was conducted. Yet, the number of agriculture production increased during 2016-21, employed people increased by less than 1.2 million especially for staple foods. The production of maize in contrast to the preceding 8-year period (2005- increased by close to 48 percent, wheat by 36 13) when the economy had created a net total percent and teff production by 23 percent during of 8.5 million jobs during 2005-13. The industry 2016-21 (Figure 4). The slowest growth was in the sector shed more than 600,000 jobs during 2013- production of sorghum, which is mostly grown in 21. Employment in agriculture declined too. Some the drought ravaged eastern parts of Ethiopia. In people in rural areas were engaged in the services, total, cereal production grew at an annual average the only sector that created the same number of of 4.5 percent during 2016-21, closely aligned to jobs during 2013-21 as it had in the 8 years before the agriculture GDP growth, but in per capita terms, then (Figure 4). cereal production grew at half that pace. Therefore, marketed surplus generation for most cereals – More people – mostly rural women - found including those where yields have increased themselves out of work completely or still in substantially – has not grown fast enough. In fact, search of work. The unemployment rate increased agriculture GDP per capita has not grown since from 5 percent in 2013 to 9 percent in 2021, 2019 when the most recent drought set in. while the labor force participation rate (LFPR) in that period declined by 12 percentage points in Job creation lost steam, leaving many 2021 in contrast to a steady increase in previous people out of work. periods. These changes mostly occurred in rural areas and were borne by women. The net decline The economic impacts of the multiple crisis, in industry and agriculture employment among along with other factors, was reflected in the women accounted for all the decline in the number labor market which lost steam. Data from the last of jobs in these sectors (Figure 5). 12 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Figure 5. Trends in key labor market outcomes in Ethiopia Net job creation by sector, location, Unemployment rates by location, (%) Labor force participation rates by and gender, (Million people) location, (%) 9 95% 21% 8 18% 90% 6 17% 86% 15% 85% Millions 5 3 80% 78% 0 9% 75% -2 74% -3 6% 6% 70% 5% Ethiopia Urban Rural Male Female Ethiopia Urban Rural Male Female 3% 3% 65% 2% 1% 60% 2005-13 2013-21 1999 2005 2013 2021 1999 2005 2013 2021 Urban Rural Total Urban Rural National Urban Rural National Source: Authors’ estimates based on LFS 2005; 2013; 2021. Figure 6. Consumption growth by household welfare ranking Rural Ethiopia Urban Excluding Addis Ababa Addis Ababa 6 6 6 4 4 4 Annual growth rate % Annual growth rate % Annual growth rate % 2 2 2 0 0 0 -2 -2 -2 -4 -4 -4 -6 -6 -6 -8 -8 -8 -10 -10 -10 -12 -12 -12 1 10 20 30 40 50 60 70 80 90 100 1 10 20 30 40 50 60 70 80 90 100 1 10 20 30 40 50 60 70 80 90 100 Consumption PAE percentiles Consumption PAE percentiles Consumption PAE percentiles Growth income Growth income Growth income 95% confidence bounds 95% confidence bounds 95% confidence bounds Mean growth rate Mean growth rate Mean growth rate Source: Authors’ estimates based on HCES 2010/11, 2015/16 and HoWStat 2021. Household consumption growth together, and converted to an annual value. This declined across the board. is then divided by the household size adjusted for demographic composition using an equivalence Household consumption – the preferred scale, to arrive at annual consumption per adult measure of welfare in Ethiopia – declined for equivalent amount as the measure of welfare. The both the rich and the poor but more so for the approach for estimating welfare using HoWStat rich than the poor. To arrive at a measure of 2021 is comparable to that applied in HCES welfare, households’ total consumption of food, 2015/16. By this measure, average household frequently purchased non-food items, actual rent welfare declined by 2.1 percent nationally between paid by renters, or the amount homeowners think 2016 and 2021. In a complete reversal from the they would pay if they rented their dwellings and trend observed at the beginning of the last decade, the use value of durable goods they own, are added the richest 20 percent of the population lost more 13 PART 1: TRENDS, PATTERNS, AND DRIVERS OF POVERTY AND INEQUALITY IN ETHIOPIA than the poorest 20 percent, though all groups show that the economy in general still grew – at a were generally worse off in 2021 than in 2015/16. slower pace, but still high in comparison to other countries. This presents an inconsistency with the Rural areas appeared more resilient than urban microdata on rising unemployment and deteriorating households. The consumption per adult equivalent household consumption. There are several possible declined in rural areas by 2 percent on average, sources of this discrepancy, but none are definitive. which was only slightly worse for people ranked One is that real consumption and GDP growth are in the middle than for everyone else (Figure 6). In estimated using different deflators, therefore the urban areas outside Addis Ababa, consumption difference between the CPI, which is used to deflate declined by 5.2 percent on average and by up to consumption, and the GDP deflator, which rose more 11 percent for people in the richest two quintiles. slowly than CPI, accounts for some of the difference In both rural and urban areas outside Addis Ababa, in the household consumption growth and GDP people in all parts of the socioeconomic ranking growth. Growth in agriculture output would have were at best not better off in 2021 than they were been expected to benefit some rural households, in 2015/16. But in Addis Ababa, the poorest 25 which does not seem to be the case. That growth percent of households registered some gains, declined across the entire welfare spectrum, rules while those in the top half of the socioeconomic out the possibility that welfare deteriorated because rankings experienced significant losses. gains from growth were unevenly distributed. Thus, the period up to 2021, was characterized by In what follows, the report discusses the reversals of fortunes for households in Ethiopia, implications of this on trends in poverty and but that also presents a puzzle or a discrepancy inequality in Ethiopia. The reports builds on this with macro statistics. From the microeconomic to analyze drivers of the changes in poverty in the perspective, households endured a series of social, next two sections of Part 1 of this report. In Part 2, climate, health, and economic shocks around the the report will dig deeper into the discussion of the same time, fewer jobs were created, and more major shocks and their implications for household people were out of work than before. This would welfare in two deep dive chapters focused on align with the data showing that households ended conflict and climate shocks, then go in-depth on up worse in monetary welfare terms in 2021 than issues affecting the labor market in another deep they were in 2015/16. However, macro statistics dive focused on structural transformation. 14 ETHIOPIA POVERTY AND EQUITY ASSESSMENT IMPLICATIONS FOR THE of the increase depends on how one adjusts the TRENDS AND PATTERN poverty line for price changes over time (Figure OF POVERTY 7). In one approach, preferred by the government, the poverty line is adjusted by computing the cost This section explores the poverty implications of the of the poverty food basket using prices from the decline in consumption following the deterioration HoWStat 2021 to obtain a food poverty line which in the economic situation and multiple shocks is scaled up by the non-food share in consumption experienced in Ethiopia. It presents trends in poverty at the time the poverty basket was derived in and considers how the patterns of poverty changed 2010. This gives a total poverty line of 17,753 per over time. This shows that poverty increased in all annum per adult equivalent in December 2021 areas of Ethiopia except Addis Ababa and Dire Dawa prices and a poverty rate of 28 percent in 2021 by 10 percentage points and validated by worsening compared to 26 percent in 2015/16, excluding in other welfare measures including non-monetary Tigray from the calculations since it was not indicators. However, the distributional pattern of covered in the HoWStat 2021. Another approach poverty has remained the same as before, defined by uses the national CPI to update the official poverty rural-urban disparities including within regions, and line derived in 2015/16 in December 2021 prices, by the geographical distribution of the population giving a poverty rate of 33 percent in 2021 – an 8 across regions rather than the gaps in the incidence percentage points increase over the poverty rate in of poverty across regions. With most of the population 2015/16 excluding Tigray (Figure 7). This is more in Ethiopia living in moisture-reliable areas, so too than the incidence of poverty in 2010 (30 percent). are most of the poor, reflecting how the narrative of The analysis presented in the rest of this report poverty in Ethiopia is in part about the unrealized is based on this second approach and compares agricultural potential. poverty across time excluding Tigray in previous surveys. A detailed description of the methodology Poverty increased, undoing gains made and justification of this choice is provided in Box 2 at the beginning of the previous decade. and in Annex 1 and the methodological note (World Bank, 2024). Despite survey data not being available Poverty increased between 2016 and 2021 to for Tigray, Box 3 summarizes the welfare situation around where it was in 2010. The magnitude in Tigray and projected welfare trends. Figure 7. Trends in poverty headcount rates (%), 2010/11 - 2021 37% 33% 29% 30% 29% 30%29% 32% 25% 27% 27% 26% 28% 27% 19% 17% 15% 16% National Urban Rural National Urban Rural Poverty line using CPI deflator Poverty line based on re-costed original basket 2011 2016 2021 Source: Authors’ estimates based on HCES 2010/11, 2015/16 and HoWStat 2021. 15 PART 1: TRENDS, PATTERNS, AND DRIVERS OF POVERTY AND INEQUALITY IN ETHIOPIA Box 2. Poverty estimation methodology in Ethiopia The methodological choice for the 2021 poverty assessment hinged on two options: adopting a new methodology based on 2021 data or replicating the 2016 methodology. The second option was selected because the year 2021 was marked by crisis, making it an atypical year to anchor a new poverty basket. Notably, there was a decline in caloric intake in 2021 compared to 2016, coupled with an uptick in the prevalence of caloric inadequacy, indicating a rise in calorie-based poverty. More importantly, some parts of the country, mainly the Tigray region and conflict-affected areas, were not covered by the survey for security reasons. For this reason, the government plans to conduct a new survey from July 2024 to June 2025, with an improved consumption module and national coverage, laying the groundwork for a new poverty line. Therefore, replicating the 2016 methodology to the greatest extent possible is the more viable option, which offers the advantage of consistency, enabling the comparison of poverty trends over time. Replicating the 2016 methodology requires a consistent approach to generating the nominal consumption aggregate, spatial price deflators to account for regional price differences and adjusting for price changes over time, and updating the poverty line. Spatial price indices at the reporting level or strata were calculated using a Laspeyres weighted price food and non-food spatial deflators based on survey unit prices following the same approach documented in previous government poverty reports (MoFED, 2002). An alternative approach matching the HoWStat 2021 data to reporting levels for HECS 2016 to use spatial deflators published at the strata level in 2016 was not chosen because regional CPI trends suggest that relative prices across regions changed between 2016 and 2021. Furthermore, the 2021 based deflators yield monetary poverty rates rankings across strata that are more consistent with the non-monetary indicators rankings. Within-survey temporal price adjustments were done using national CPI with December 2021 as the reference month. The poverty line for 2021 is obtained using two alternatives – either updating the 2015 poverty line using CPI deflators or by re-costing the original poverty basket in average 2021 prices. The poverty line is anchored in a poverty basket through a selection of a food bundle commonly consumed by the poor, ensuring it meets a set minimum caloric intake (2,200 kcal). The composition of the food basket has remained unchanged since its determination in 1996. The poverty line was revised in 2011 by re-costing the items in the original food basket at prevailing prices and applying a non-food allowance component based on the non-food shares. The official poverty line in 2015/16 inflated the 2011 poverty line using the GDP deflator which produced a smaller increase in the poverty line relative to changes reflected in the CPI. To properly account for changes in the poverty basket since 2016, the poverty line in 2021 can either be adjusted by the CPI or derived by directly costing the existing food basket and apply an adjustment for the non-food component, which would entail revising the 2016 line also to have a consistent approach. The CPI based and direct costing approach yields poverty lines of Birr18,964 and 17,753 in 2021. The government prefers the second line, arguing that the CPI does not accurately reflect rural price changes since price data is collected mainly from urban markets. However, because the poverty basket is defined in broad groups, the direct costing of the food basket involves using a composite price for each group hence the preference for using the CPI based poverty line as the basis for analysis in this poverty assessment. 16 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Poverty increased more in rural areas than scale) – are deemed to consume inadequate in urban areas because many non-poor rural calories and classified as caloric deficient. The households were just above the poverty line in share of people in such circumstances increased 2015/16, hence highly vulnerable to falling into from 26 percent in 2015/16 to 31 percent in poverty. The poverty rate in rural areas increased 2021 (Figure 9). They increased more in urban from 27 percent in 2015/16 to 37 percent in 2021 areas (11 percentage points) than in rural areas compared to a 3 percentage points increase to 19 (2 percentage points), reflecting the larger decline percent in urban areas. The decline in household in food expenditure among urban households (23 consumption in rural areas, though smaller, pushed percent) than rural households (15 percent). Food many rural households to fall around the poverty line, poverty rates doubled because of this (see Annex resulting in a higher increase in poverty compared 3, Table A.3.1). to urban areas where many non-poor households started off with much higher consumption. The big Human development outcomes either deteriorated decline in consumption in urban areas left most of or did not statistically change. The share of adults them close, but still above the poverty line. This is with secondary education and above barely changed demonstrated in Figure 8 which also shows that and declined in urban areas. Thus, education households at all consumption levels were no better attainment levels remain low. Just 7 percent of adults off in 2021 compared to 2015/16 in both rural and in the country and just 2 percent in rural areas, had urban areas. Therefore, the poverty rate increased secondary education or above in 2021, while 78 in 2021 irrespective of the poverty line drawn. percent of adults nationally (86 percent in rural areas) have no formal education or have incomplete primary The caloric deficiency rate, which abstracts education. Child nutrition outcomes worsened, from the issue of price adjustments over time, reflected in the increase in under 5 children stunting also points to an increase in poverty. The caloric rates from 42 percent in 2015/16 to 45 percent in deficiency is based on an estimation of the calorie 2021 (Table 2). This increase was observed in both intake from the reported quantities of items rural and urban areas. Both health and education consumed by households. Households consuming outcomes deteriorated as one can imagine with less than the minimum daily calorie intake of 2,200 the disruptions caused by the COVID-19 pandemic per person - adjusted for demographic differences on schooling and health (Harris et al., 2021) and in minimum calorie requirements using an adult considering the nutritional impact of reduced food male as a benchmark (i.e., using the adult equivalent consumption reported above. Figure 8. Comparison of the population Figure 9. Incidence of calorie deficiency (%), distribution by consumption levels 2015/16 - 2021 .0004 30.5% 29.9% 30.7% 28.2% .0003 26.4% Density .0002 18.6% .0001 0 0 20,000 40,000 60,000 80,000 100,000 Per adult consumption (real) Black short dash: poverty line National Urban Rural 2016 urban 2021 urban 2016 rural 2021 rural 2016 2021 Source: Authors’ estimates based on HCES 2015/16; HoWStat 2021. 17 PART 1: TRENDS, PATTERNS, AND DRIVERS OF POVERTY AND INEQUALITY IN ETHIOPIA The improvement in access to some public quality of housing, and ownership of durable services, especially in rural areas, was the most assets in rural areas. The multi-dimensional notable positive welfare gain between 2016- poverty index combines three dimensions – 21. The share of households with access to piped education, health, and living standards. The first water and those connected to the electricity grid two dimensions deteriorated. The living standards both increased by 7 percentage points nationally. dimension improved due to the increase in access These improvements were most notable in rural to electricity, improved water, improved housing areas. Access to the electricity grid, for example, quality and asset ownership. Because of these doubled in rural areas to around 15 percent in 2021, gains, the multi-dimensional poverty index declined though it marginally declined in urban areas to 88 by three percentage points to 69 percent in 2021. percent in 2021 compared to 91 percent in 2016. This was driven by improvements in rural areas, More than a third of rural households had access where the improvements in access to services were to piped water in 2021, reflecting a 9 percentage notable, while multi-dimensional poverty increased points increase in rural access to piped water in urban areas (Table 2). compared to 2016. Housing quality (improved roofing and floor materials), along with ownership The weight of evidence, however, points to an of durable goods such as televisions and phones, overall increase in poverty. The incidence of also improved significantly in rural areas. These may poverty based on monetary poverty increased reflect earlier gains before multiple crises set in and and household consumption declined at all levels that households did not resort to asset disposals. of consumption which points to declining poverty irrespective of where the poverty line is drawn. The multidimensional poverty index improved Stripping the influence of prices in the valuation because of the increased access to public services, of consumption, the caloric deficiency shows that Table 2. Multidimensional poverty and non-monetary poverty indicators 2016 2021 National Rural Urban National Rural Urban Multi-dimensional poverty headcount 72% 82% 28% 69% 77% 38% Share of adults with no education or incomplete primary education 79% 90% 45% 78% 86% 54% Share of adults with secondary and above education 8% 2% 27% 7% 2% 20% Stunted child 42% 45% 30% 45% 48% 34% Severely stunted child 22% 23% 13% 25% 27% 17% Access to piped water 41% 25% 92% 48% 34% 87% Access to electric grid 27% 7% 91% 34% 15% 88% Television 14% 1% 54% 19% 5% 60% Telephone 53% 42% 89% 68% 61% 88% Improved roof 61% 50% 96% 72% 64% 96% Improved floor 11% 2% 43% 18% 6% 54% Source: Authors’ estimates based on WMS 2015/16 and HoWStat 2021. Notes: Multidimensional poverty computation is based on the Global MPI methodology by Alkire et al. (2021). The MPI covers health, education, and living standard dimensions, and ranges from 0 to 1, higher value representing a high level of deprivation. Multidimensional poverty headcount represents the share of people with a deprivation score of one-third or higher (see Annex 2 for detailed methodology). 18 ETHIOPIA POVERTY AND EQUITY ASSESSMENT more households failed to meet their minimum the poverty rate barely changed during 2015-20. consumption needs. Shifting away from monetary There are some exceptions, like Burkina Faso, that to non-monetary indicators shows a lack of still registered significant improvements in poverty improvement in human development outcomes. The over a comparable period. positive gains in access to piped water, electricity, and means of communication lay the potential for Poverty increased in all regions except future improvements but at present, households Addis Ababa and Dire Dawa and are in an economic pinch. increased most in the lowlands. Ethiopia’s experience of rising poverty during this A reversal in living standards was experienced in period is not unusual. Other African countries like all regions except in Addis Ababa. The incidence Kenya, Nigeria, and Cameroon, also either registered of poverty increased the most in Gambella, Somali, an increase in poverty or did not see any gains over and former SNPP regions (now Sidama, Southwest, a comparable period (Figure 10). For perspective, Southwest Ethiopia, and Central Ethiopia regions) the poverty rate based on the international extreme where poverty increased by 11–17 percentage poverty line of PPP $2.15 per person per day points (Figure 11). These are also the areas most increased from 27 percent in 2015 to 32 percent in exposed to drought shocks – particularly hit hard 2021 in Ethiopia, while it increased from 29 percent by the 2019 drought. This relationship between in 2015 to 32 percent in 2021 in Kenya. In Nigeria, drought shocks and various measures of poverty will be explored further in Chapter 5. The poverty Figure 10. International comparisons in poverty rate in the Amhara region – which experienced a rate trends: 2014-2021, (%) surge in conflict - increased by 5 percentage points 70 overall, primarily driven by the 11-percentage point 66 increase in poverty in urban areas where conflict 52 was concentrated. A similar pattern is observed in 43 42 45 the Afar Region. This poverty and conflict nexus shall 36 be explored further in Chapter 6. Evidence of the 32 32 31 33 27 29 changes in poverty in Benishangul is inconclusive 25 and tainted by the fact that zones that were poorer in 2015/16 were not covered in 2021 due to conflict. Poverty in the most urbanized regions either did not change (in Harari and Dire Dawa) or Ethiopia (2015) Ethiopia (2021 Rwanda (2016) Ghana (2016) Uganda (2016) Uganda (2019) Tanzania (2018) Kenya (2015) Kenya (2021) Malawi (2016) Malawi (2019) Nigeria (2015) Kenya (2018) Nigeria (2020) declined (in Addis Ababa). The increase in poverty in urban areas is thus driven by increasing poverty in regional capitals, medium cities, and small towns in other regions. These urban areas had previously Source: Compilations of various World Bank Reports. experienced some of the fastest decline in poverty. 19 PART 1: TRENDS, PATTERNS, AND DRIVERS OF POVERTY AND INEQUALITY IN ETHIOPIA Figure 11. Regional trends in monetary poverty, 2015/16-2021 0.50 0.45 0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 Afar Urban Afar Rural Amhara Urban Amhara Rural Oromia Urban Oromia Rural Somali Urban Somali Rural Benishangul-Gumuz Urban Benishangul-Gumuz Rural SNNP Urban SNNP Rural Sidama Urban Sidama Rural Gambella Urban Gambella Rural Harari Urban Harari Rural Addis Ababa Dire Dawa Urban Diredawa Rural 2016 2021 Source: Authors’ estimates based on HCES 2015/16 and HoWStat 2021. Notes: The EAs that were inaccessible in Benishangul-Gumuz in 2021 appear to have been replaced with EAs from areas that were less poor in 2015/16, which could bias the results. The SNNP region has since split into several regions, which are regrouped together in this Poverty Assessment when comparing regional trends over time. Box 3. What do we know about welfare in Tigray? The conflict caused the deaths of tens of thousands, created large-scale displacements, and vast humanitarian needs in Tigray. The number of deaths and internally displaced people due to conflict increased drastically (for more details, see Part 2 on Conflict and household welfare in Ethiopia). The conflict further led to widespread crop and livelihood losses, the destruction of the local economy, and impairment of market activities, access to services, and humanitarian assistance. The conflict in Tigray increased poverty, caused high levels of food insecurity, and large-scale displacement. The conflict erupted in November 2020 at the peak of the main agricultural season (Meher) harvest period when many households had not yet harvested their crops. As the conflict spread into Amhara and Afar, agriculture planting and later harvest were also interrupted in the conflict areas. Lack of access to agricultural inputs could have reduced yields too. Non-agriculture activities were also interrupted by the closure and damage of shops and industrial parks leading to forgone incomes not only during the conflict period but for some time after the conflict. In Tigray, the delivery of social assistance transfers was also suspended and support through the largest social protection programs—the PSNP and UPSNP—were only restarted in April 2023, 2.5 years after the conflict started. High inflation due to increased scarcity and impairment of market functioning further eroded household purchasing power. All this contributed to a loss in income for households, most deeply among poor households who depended heavily on agriculture and social assistance. Despite survey data not being available for Tigray through HoWStat 2021, conservative estimates based on a simulation exercise for 2020 and 2021 suggest that due to disruptions in livelihoods, households in Tigray lost 46 percent of their income in 2020 and 20 ETHIOPIA POVERTY AND EQUITY ASSESSMENT 38 percent in 2021. The poorest quintile lost a greater share (53 percent) of their income in 2020. The variation of losses within Tigray was large. Those around Mekelle lost close to 70 percent of their income in 2020, compared to an estimate of around a third of income lost by households in North-Western Tigray in that period. Though income losses remained high in Tigray zones in 2021 more broadly – ranging from 30 percent in Western Tigray to 55 percent in Central Tigray. Year Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5 Tigray 2020 53% 41% 42% 47% 44% 2021 39% 35% 40% 40% 36% The factors driving household income losses differed by region and socio-economic status. Among the poor, household income losses were largely driven by losses in agriculture incomes due to lost harvest and planting opportunities, while losses in business incomes drove losses in income among the top two quintiles. With high conflict intensity in urban areas in 2020, losses in business incomes accounted for a greater share of income losses in Tigray. Increases in poverty were accompanied by increased acute hunger. As of June 2021, about 5.5 million people in Tigray and neighboring Afar and Amhara (61 percent of the population in northern Ethiopia) were facing high levels of acute food insecurity, resulting from the cascading effects of conflict, including population displacements, movement restrictions, limited humanitarian access, loss of harvest and livelihood assets, and dysfunctional or non-existent markets (WFP, 2022b). Calorie deficiency and food poverty rates The poor are concentrated in populous increased everywhere except in rural Amhara, highland areas and are predominantly while multidimensional poverty increased in rural and agricultural with low all regions except rural Amhara, Oromia, and human capital. Harari. Rural Amhara, where calorie deficiency rates declined was less affected by droughts. Multi- Geographical differences in poverty rates in dimensional poverty remains highest in terms of Ethiopia remain primarily defined by rural-urban both its incidence and severity, in pastoral regions differences rather than regional differences of Somali (88 percent) and Afar (86 percent), a (Figure 11). Poverty rates in the three predominantly factor predominantly driven by poor access to urban regions ranged between 8 percent (in Harari) services and lack of ownership of durable assets and 14 percent (in Dire Dawa) with Addis Ababa at that require electricity connections (see World 12 percent. Setting aside Benishangul-Gumuz for Bank, 2023). This low access to services such as reasons stated above, the poverty rates in all other, electricity (grid) and access to piped water in rural more rural regions, ranged between 32 percent areas – despite recent improvements - combined (Afar and Oromia) and 38 percent (in the group with low education levels explains the high of regions previously forming the SNNP region), incidence of multi-dimensional poverty outside hence the gap between them is small. However, the predominantly urban regions of Ethiopia (See there is a wide gap in poverty rates between rural Annex 3, Table A.3.3). and urban areas within these regions. Poverty rates 21 PART 1: TRENDS, PATTERNS, AND DRIVERS OF POVERTY AND INEQUALITY IN ETHIOPIA in urban areas in all these regions were 21 percent Afar which together cover 61 percent of Ethiopia’s or less, except for Gambella (28 percent) and the land mass, only account for 9 percent of the poor, group of regions that previously formed the SNPP because they make up only 9 of the population region (27 percent). Thus, poverty rates in urban between them (see Annex 3, Table A.3.2). The areas across most regions are higher but closer geographical distribution of the poor thus remains to the incidence of poverty in Addis Ababa, with a like the pattern in 2016. bigger gap between these urban areas and rural areas within the same regions. The poor are predominantly agricultural but most of them reside in high agricultural potential areas Poverty in Ethiopia is therefore predominantly than in drought-prone areas. Pastoral ecological rural, with the distribution of poor people largely zones made up just 5 percent of the poor while the defined by the population size given the narrow drought-prone highlands and lowlands ecological differences in poverty rates across regions. Rural zones each accounted for 13 percent of the poor. areas accounted for 88 percent of all the poor As such, two-thirds of the poor reside in moisture people in the country, which is higher than their reliable highland areas that also account for 65 share of the population in 2021 (78 percent) but percent of Ethiopia’s population (Table 3). Poverty unchanged from 2016. Across regions, Oromia in Ethiopia is therefore largely a story of unrealized accounts for the largest share of poor people (39 agriculture potential given the geographical location percent), followed by Amhara (25 percent) with of most of the poor, combined with the fact that these two regions accounting for 64 percent of 80 percent of the poor are employed in agriculture. poor people in the country, hence proportionate That a significant share of the non-poor (60 percent) to their population. Pastoral regions of Somali and works in agriculture demonstrates the potential for reducing poverty among agricultural households. Table 3. Composition of the population, poor and non-poor by location, sector, and education Poor households in Ethiopia have more children level, 2021 than the workers needed to support them and are poorly educated. A typical poor household Pop. Poor Non- has 6 people, about half of them children under Share poor 15. The dependency ratio of 1.26 suggests that Urban 22% 12% 27% workers in poor households support more people Rural 78% 88% 73% than workers in non-poor households which have Drought prone highlands 16% 13% 17% a dependency ratio of 0.84 and have smaller Drought prone lowlands 9% 13% 7% households (4 people) with fewer children (Table Moisture reliable lowlands 3% 4% 3% 4). Though slightly more educated than poor Moisture reliable highlands 65% 65% 66% people, most non-poor people still have lower Pastoral 6% 5% 7% levels of education. Half of them have no formal Head employed in agriculture 73% 85% 67% education – but an even higher share of the poor Household head works in 27% 12% 26% (61 percent) has no formal education. Just 19 service/industry percent of the non-poor have completed primary Completed primary education 16% 7% 20% education (13 percent) or above (6 percent) though or above this is almost triple the share of poor people with Incomplete primary education 20% 19% 21% complete primary education or above. This lack Has no formal education 64% 74% 59% of skills could be a barrier to access to better economic opportunities. This will be discussed Source: Authors’ estimates based on HoWStat 2021. further in Chapter 7. 22 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Poor households face significant disparities in Market and public service accessibility is a key access to public services and living conditions. driver of poverty reduction, yet the poorest Half of the households in the poorest quintile households often face isolation from these critical experience overcrowding (four or more people per resources. The average distance to food markets for room), compared to just 8 percent in the richest the poorest is 7km, compared to just 4km for the quintile. The construction quality of homes also richest. The richest households have better access varies, with poorer families more likely to live in to climate-resilient roads that are usable throughout houses made of substandard materials. Sanitation the year. Financial services present another divide. facilities are accessible to 37 percent of the richest Poor households have significantly less access to families, but only 14 percent of the poorest. While institutions such as banks and microfinance than 53 percent of the richest families have electricity, their richer counterparts. Educational services this is true for only 14 of the poorest. Access to exhibit similar disparities on access to secondary safe water access is more equitable, yet still schools which is more limited for the poorest favors the richest quintiles (81 percent) over the compared to the richest. Access to primary school poorest (59 percent). Asset ownership further access is generally equitable. Likewise, primary illustrates the divide. Less than 1 percent of the healthcare facilities are accessible to all income poorest quintile own refrigerators, cars, bicycles, groups, but hospitals are on average 36km away or computers, and there is a significant gap in the from the poorest household compared to the ownership of cell phones (47.5 percent vs. 74.1 average of 23km for the richest. percent), televisions (3.7 percent vs. 33 percent), and radios (9.7 percent vs. 20.4 percent) when The poorest households are disproportionately compared to the richest quintile. The only assets affected by climatic shocks and food shortage, the poorest households are more likely to own while the richest are more often impacted by than richer households are farm implements, conflict. Objective measures of climate shocks reflecting the higher incidence of poverty among such as the Palmer Drought Severity Index agriculture households. (PDSI), indicate more severe conditions for the poorest households. Market shocks, which include fluctuations such as increases in food prices, Table 4. Demographic characteristics of poor higher prices for agricultural inputs, and decreases and non-poor households, 2021 in output prices, are slightly more common among the richest quintile (18 percent) than the poorest Poor Non- poor (15 percent). Contrary to expectations, insecurity, Male headed household 80% 75% violence, and conflict related shocks tend to impact Age of household head (years) 46.4 43.8 the richest households more than the poorest, Household size 5.92 4.11 suggesting that such conflicts are urban centered. Adult equivalent 4.95 3.40 The prevalence of employment and health shocks Number of working members 2.92 2.33 is relatively even across different levels of welfare. Dependency ratio 1.26 0.84 These patterns will be discussed in more detail in Number of members below 15 years 2.84 1.60 Part 2 of the report. of age Number of members above 64 years 0.17 0.18 Social assistance programs, such as the Rural of age Productive Safety Net Program (PSNP) and the Urban Productive Safety Net and Jobs Program Source: Authors’ estimates based on HoWStat 2021. (PSNJP), along with humanitarian aid, primarily 23 PART 1: TRENDS, PATTERNS, AND DRIVERS OF POVERTY AND INEQUALITY IN ETHIOPIA cover the poorest. Approximately, 16 percent between groups. The outcome has been modified of households in the lowest income quintile are to create four distinct subgroups of the poor enrolled in or receive transfers from PSNP, either (Figure 12). The first group of the poor–rural through public works or direct support. In contrast, agricultural households in non-farm transition– 4.5 percent of the wealthiest quintile households are concentrated in Oromia, SNNP, and Amhara benefit from PSNP. Humanitarian aid shows a regions. The second group–crop-livestock mixed broader distribution, benefiting 15 percent of rural households–also seems to be concentrated households in the lowest expenditure quintile more in Oromia, SNNPR, and Amhara regions. The and 6.6 percent in the highest quintile. Despite third group–pastoral, drought-prone, and remote the higher frequency of direct cash transfers from rural households–is concentrated in Oromia, the government through safety nets and aid among Amhara, and Somali regions. The fourth group – the poorest and most vulnerable households, the urban poor–is concentrated in Oromia, Amhara, coverage is modest (10.3 percent) considering the and SNNPR. high poverty incidence in the country. Figure 12. Composition of the poor by subgroups (%), 2021 The differences among the poor are defined by agriculture dependency, Agricultural in non-farm connectivity, drought exposure and 13 transition access to land. 39 16 Crop-livestock mixed The poor are not a homogenous group and can be classified into four distinct groups. These groups Pastoral, drought-prone, are identified using cluster analysis – a statistical and remote technique that partitions individuals into groups 32 Urban poor to maximize the similarity of individuals within each group while maximizing the dissimilarity Source: Authors’ estimates based on HoWStat 2021. Figure 13. Household characteristics by subgroups of the poor, 2021 (% of household in the subgroup) 100% 80% 60% 40% 20% 0% Share of members with education Household has electricity Has improved toilet Has non-farm enterprise Has agri. and non-agri. wage Lives in highland areas Lives in lowland or pastoral areas All-weather road within 5 kms Health facility within 5kms Agricultural non-farm transition Crop-livestock mixed Pastoral, drought-prone, and remote Urban poor Source: Authors’ estimates based on cluster analysis using HoWStat 2021. 24 ETHIOPIA POVERTY AND EQUITY ASSESSMENT The constraints faced by these groups of the engages in mixed farming systems that combine poor are different. For the rural poor, the primary crop cultivation with livestock and are concentrated constraints include limited education, lack of in moisture-reliant agroecological zones. The third non-farm income opportunities, few assets, low group of the poor consists of pastoralists and connectivity, and remoteness (Figure 13). The first rural households in drought-prone areas. These group of the poor is in areas with high agricultural groups are often more isolated and suffer from the potential and tends to have more diversified lowest level of human capital and asset ownership employment and livelihoods. Although reliant on crop (excluding livestock). The last group of the poor, production, they benefit from a broader economic the urban poor, presents a different group. They base, being less remote and better connected. In represent urban households with limited access contrast, the second group of the poor, although to formal employment, with most finding informal in areas with good agriculture potential, primarily work in the service sector. 25 PART 1: TRENDS, PATTERNS, AND DRIVERS OF POVERTY AND INEQUALITY IN ETHIOPIA DRIVERS OF THE INCREASE beneficiaries, but limited coverage limits its broader IN POVERTY impacts. Beyond shocks, low human capital and low access to better income generating opportunities This section analyzes the drivers behind the are underlying contributors to poverty. observed increase in poverty between 2015/16 and 2021. It starts by decomposing changes Poverty increased due to a generalized in poverty into the contribution of growth and decline in welfare across the entire inequality, changes within locations, employment/ distribution. income types, and shifts across them, and then applies advanced statistical techniques to estimate Inequality declined between 2015/16 and 2021 the impact of various shocks on household welfare as the non-poor faced worse welfare losses than and poverty. The primary driver of the observed the poor. The Gini coefficient declined from 0.33 increase in poverty between 2015/16 and 2021 was in 2015/16 to 0.29 in 2021 (Figure 14). Inequality a backslide of welfare across the entire population declined because households in all parts of the rather than increases in inequality, which declined. distribution became worse, but with consumption This welfare back slide reflected a lack of income declining more among the non-poor. Behind this growth mainly because incomes did not rise fast decline was a drastic reduction of inequality in urban enough to keep up with inflation – a finding that is areas (9 Gini points), to a similar level of inequality than validated by more sophisticated statistical analysis. in rural areas in 2021 (Gini of 0.27) and a narrowing The reduction in agriculture and self-employment of the gap between rural and urban areas. In fact, the earnings, respectively, contributed the most to share of the Gini coefficient that can be attributed to the increase in poverty in rural areas and urban differences between urban and rural areas fell from 29 areas. Rural households failed to capitalize on percent in 2015/16 to 21 percent in 2021 (see Annex rising prices because only a few of them produce 3, Table A.3.5), while the gap in mean consumption a marketable surplus – a legacy of the country’s between urban areas and rural areas declined from 76 agricultural policies. Analysis of poverty transitions percent to 42 percent in the same period. Inequality using panel data shows that exposure to shocks also declined across all regions, with a significant also contributed to the increase in poverty. Safety reversal in regions like Somali where inequality had nets helped reduce the depth of poverty among been rising fast (see Annex 3, Table A.3.5). Figure 14. Trends in inequality (Gini coefficient), Figure 15. Decomposition of poverty changes due 2010/11 - 2021 to growth and inequality, 2015/16 – 21 (pp change) 15 11.3 11.4 9.3 13.0 0.38 9.2 0.37 10 Change in Poverty 0.33 3.5 5 2.2 Headcount 0.30 0.29 0.28 0.29 0.27 0.27 0 -5 -2.0 -10 -9.5 -15 Growth Redistribution Total change Growth Redistribution Total change Growth Redistribution Total change National Urban Rural 2010/11 2015/16 2020/21 National Urban Rural Source: Authors’ estimates based on HCES 2015/16; HoWStat 2021. Notes: Growth-inequality decompositions follows methodology in Datt & Ravallion (1992). 26 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Therefore, the increase in poverty observed Figure 16. Decomposition of poverty changes by intra-location changes and population shifts in recent years was due to negative growth in across locations (pp change), 2015/16 – 2021 households’ welfare rather than changes that led to a more inequitable distribution of welfare. 10 Change in poverty headcount (pp change) The change in poverty between 2015/16 and 2021 8 can be decomposed into the contribution of growth 6 and changes in inequality (Figure 15). The growth 4 contribution suggests that the poverty rate would have 2 increased by 11.3 percentage points if consumption 0 for every household had declined at the same rate -2 -4 as the decline in average consumption during this -6 period. The redistribution contribution suggests that -8 poverty would have declined by 2 percentage points, -10 if only inequality had reduced as observed, without 2000-2005 2005-2011 2011-2016 2016-2021 the accompanying decline in average consumption. The final poverty rate decline of 9.3 percent is hence Interaction effect Population-shift effect primarily because of declining growth, marginally Urban Rural tampered by this decline being steepest among Source: Authors’ estimates based on HCES 2015/16; HoWStat 2021. people that were better off to begin with. These Notes: Methodology follows Ravallion & Huppi (1991). off-setting contributions explain the small drop in poverty in urban areas where average consumption increased slightly less there. Poverty increased by declined more, but the smaller-than-average more in lowland areas, but they only contributed to decline in consumption for households close to the 10 percentage points of the increase in poverty owing poverty line meant that urban poverty increased to their smaller population shares. less. Meanwhile, in rural areas, consumption for households close to the poverty line declined by Reduced incomes from on-farm activities were slightly more than the rural average, amplifying the the primary drivers of increasing poverty in rural overall increase in rural poverty. areas, but with variations across ecological zones. The increase in poverty was driven by From a spatial perspective, most of the increase in declining consumption among households that poverty was driven by increasing poverty in rural rely on on-farm incomes across the highlands, areas and in the highlands. The larger share of the lowlands, and pastoralist areas (Figure 17). In population in the countryside and the higher increase highland and lowland areas, rising poverty was in poverty among rural household’s accounts for 90 primarily driven by an increase in poverty among percent of the increase in poverty during 2015/16-21. crop growers. While Lowlands households show a This reversed the trend observed since 2000, when similar profile to the highlands, switching across rural poverty declines drove the national poverty different livelihood styles was the second largest declines, although urban areas had also contributed contributor to the poverty increase in lowland areas to poverty reduction more than proportionally to their – accounting for close to 15 percent of the increase population share (Figure 16). The contribution of in poverty in there – and third was rising poverty population shifts between rural and urban areas was among herders. Meanwhile, rising poverty among negligible, as always has been the case in Ethiopia. mixed crop and livestock farmers was the second Most of the increase in poverty rural areas is driven by largest contributor to rising poverty in highland the Highlands areas, because the highlands host over areas. Rising poverty among the herders mainly 80 percent of the rural population, although poverty contributed to the increase in pastoral poverty. 27 PART 1: TRENDS, PATTERNS, AND DRIVERS OF POVERTY AND INEQUALITY IN ETHIOPIA Figure 17. Decomposition of poverty changes in Figure 18. Decomposition of poverty changes rural areas by changes within income types and in urban areas by changes within employment shifts across income types, 2015/16 – 2021 sectors and shifts across sectors, 2015/16 – 2021 14 Agriculture 0.63 Self-employed Industry 12 0.29 Change in poverty headcount, 10 Services 0.39 8 Agriculture 0.11 (pp change) Salaried Industry 0.28 6 (private) Services 0.26 4 Public Sector 0.10 2 Other 0.44 0 Population-shift effect 0.53 -2 Interaction effect 0.61 Highlands Lowlands Pastoralist 0.00 0.50 1.00 Interaction effect Population-shift effect Others Contribution to Urban Poverty Increase 2016-2021 Crops & Liv. Livestock Crops Rural (Percentage points) Source: Authors’ estimates based on HCES 2015/16; HoWStat 2021. Source: Authors’ estimates based on HCES 2015/16; HoWStat 2021. Notes: Methodology follows Ravallion and Huppi (1991). Notes: Methodology follows Ravallion and Huppi (1991). In urban areas, declining earnings from self- moved to historically lower earning self- employment are behind most of the decline employment which was also suffering from the in living conditions during 2015/16-2021. The impacts of the COVID-19 pandemic. urban self-employed contributed to 1.3 out of 3.6 percentage points increase in the urban Household income gains were wiped poverty rate while increasing poverty among out by inflation, driving the increase households dependent on salaried employment in poverty. in the private sector contributed to 0.65 percentage points (Figure 18). Public sector The deficit in real household consumption growth workers contributed the least to the increase in that drove the observed increase in poverty was poverty, due to them being located safely above in part due to high inflation. Mean household the poverty line rather than being cushioned consumption per adult equivalent doubled in from the decrease in their consumption levels. nominal terms between 2015/16 and 2021. The population-shift effect and interaction- However, this primarily reflected inflation rather effect (altogether, larger than any other sector) than real gains. Nationally, mean consumption signal that workers switched to sectors or forms declined by close to 22 percent in real value of employment characterized by higher poverty (Table 5). The real value of non-food consumption rates. This is consistent with findings from the declined by more than food consumption in urban high frequency surveys monitoring the impacts areas, reflecting that households relying more of COVID-19 (Ambel et al., 2021). Household on market purchases of both food and non-food enterprises closed and the surviving ones goods sacrificed non-food spending to try to meet were less profitable. Employment in industry their food subsistence needs. Consequently, the declined and peopled switched to more self- share of food in total spending rose by more than employment, casual employment, and family 7 percentage points in urban areas to almost 53.8 work after the COVID-19 pandemic. Poverty percent in 2021. In contrast, food consumption in urban areas thus increased because people declined by slightly more than non-food 28 ETHIOPIA POVERTY AND EQUITY ASSESSMENT consumption among rural households – for whom the other enumeration areas of the country in the food production is the primary source of income. Ethiopian Socioeconomic Panel Survey (ESPS) Rural households would have to sell some of their 2018/19 -2021/22 (Box 4). The results show food production even if they are not in surplus to that total consumption and food consumption finance non-market purchases. The food share in of households in areas more severely hit by food total consumption among rural households slightly inflation declined by about 21 percent more than declined to 54.3 percent, which is a statistically the rest, with impacts of similar magnitude in both insignificant change, hence food and non-food urban and rural areas (Figure 19). The largest expenditure composition was unchanged. impact is observed in non-food consumption in urban areas, where the decline was close to 40 Incomes did not rise fast enough to catch percent larger in areas that were more severely up with rising consumer prices. This can be hit. Restricting the analysis only to households that seen from a comparison of growth in different belong to the bottom 40 percent of the country sources of income reported by households in shows that the differentiated impacts of inflation the Ethiopia Socio-Economic Surveys (ESS) were even larger among them. These results are conducted between 2015/16 and 2022 (Annex consistent with the patterns observed for real 3, Figure A.3.1). Crop incomes almost doubled in consumption growth in the HoWStat 2021 data nominal terms, while non-agriculture wage and presented earlier. That means the increase in the self-employment nominal incomes respectively, general price level contributed to the observed increased by 73 percent and 43 percent. But decline in welfare. when corrected for inflation, their real values declined by between 53 and 43 percent. Income Figure 19. Impact of incremental inflation on household welfare from transfers also declined, as social assistance benefit amounts were not fully adjusted for inflation. Only livestock incomes appear to have -0.01 increased, which should be interpreted with some caution since the nomadic population that was -0.18 more exposed to drought is not covered in the -0.21 -0.21 -0.21 -0.22 -0.22 ESS, yet they account for the lion’s share of the -0.29 -0.30 national livestock herd. -0.34 -0.35 -0.38 Not only was the overall impact of inflation Total Food Non-food Total (bottom negative, but the impact was uneven across 40 pct) regions of the country. This is shown from causal estimates based on a statistical method National Rural Urban Source: Authors’ estimates based on ESPS 2018/19; 2021/22. (difference in difference approach) that uses Notes: The dependent variables are total consumption (total), food comparisons between enumeration areas which consumption (food), nonfood consumption (non-food), and total consumption faced above average food price increases with for the bottom 40 percent (total bottom 40 pct). 29 PART 1: TRENDS, PATTERNS, AND DRIVERS OF POVERTY AND INEQUALITY IN ETHIOPIA Table 5. Comparison of growth in nominal and real consumption per adult equivalent (Birr), 2015/16- 2021 Nominal consumption Real consumption Location 2016 2021 Change (%) 2016 2021 Change (%) Total consumption National 14,723 29,688 101.7 38,594 31,714 -21.7 Urban 22,135 43,793 97.8 58,024 40,663 -42.7 Rural 12,422 24,820 99.8 32,564 28,625 -13.8 Food consumption National 7,619 15,679 105.8 19,973 17,161 -16.4 Urban 10,277 21,875 112.9 26,941 21,860 -23.2 Rural 6,794 13,540 99.3 17,810 15,539 -14.6 Nonfood consumption National 7,104 14,010 97.2 18,621 14,553 -28 Urban 11,858 21,918 84.8 31,084 18,803 -65.3 Rural 5,628 11,280 100.4 14,754 13,086 -12.7 Source: Authors’ estimates based on HCES 2015/16 and HoWStat 2021. Box 4. Estimating the welfare impact of inflation in Ethiopia Wieser and Yitbarek (2024) use the last two waves of the Ethiopian Socioeconomic Panel Survey (ESPS) and difference-in-difference approach to investigate the differentiated impact of high food inflation on household’s welfare in rural and urban Ethiopia. They classify enumeration areas (EAs) in the country into areas that faced above average price increases and areas that faced below average prices. Unit prices from the from the ESPS 2018/19 and 2021/22 rounds are used to create a Laspeyres price index, which is a fixed-base index that uses the quantities of goods and services in a base period as weights. If the Laspeyres price index of the EA was higher than the national median in 2019, households in that EA are classified as high inflation (H) households but are classified as a low-inflation (L) households if the EA’s Laspeyres index is less than or equal to the national median. A similar categorization was done using the Paasche index for robustness checks. The differences in inflation across regions can be attributed to differences in local prices and household consumption baskets (inflation was measured using increases in a Laspeyres index) assuming that the changes in outcomes for households in areas with high and low inflation would have been comparable in the absence of food inflation. This is estimated from a regression using equation 1, where γite is the observed welfare outcome for the ith households living in EA e at time t, measured through consumption expenditure per adult equivalent per year, monetary poverty (based on the bottom 20 percent and 40 percent of the consumption expenditure), nutrition security and multidimensional poverty, post is a dummy variable that takes the value 1 for 2021 or 0 for 2019, Hi is a dummy variable that takes the value 1 if a household enumeration area Laspeyres price index was higher than the national median or 0 otherwise. post*H is an interaction term. εi is the error term, Xit is a vector of control variables which include the gender, age, education, marital status, and occupation sectors of the household head, γs and δt are the EA fixed effects and the time fixed effects. β1 is the difference-in difference estimator which is computed by comparing the first differenced values of the outcome variables for the high and low-inflation household groups as in equation 2: Standard errors are clustered at the EA level to allow for within-cluster correlation. 30 ETHIOPIA POVERTY AND EQUITY ASSESSMENT The results of this analysis show that high food inflation affects households differently based on their consumption level and, for non-food consumption, based on their location (urban or rural). Household consumption, both for food and non-food items, was more negatively impacted in above-average food inflation regions in the country. Households living in regions with above-average food inflation experienced a decrease of 21.4 percent, 21.5 percent, and 34 percent in total consumption, food consumption, and non-food consumption, respectively, compared to households in below-average food inflation. The impact of above-average inflation was more significant in urban households where households experienced a decline of 20.8 percent and 37.6 percent in food and non-food consumption, respectively, compared to households in regions where inflation was below average. Figure 20. Rural households net market position about three-quarters of the total area cultivated by food crop, 2022 (Taffesse et al., 2011). Households – especially those less well-off - are mostly self-reliant for most Share of rural households (%) 100 of these crops. Just a small share of households 80 produced a marketable surplus across all crops 60 in 2022 (Figure 20). For instance, less than 10 40 percent of households produced a large enough 20 surplus of teff which has a greater share of net sellers than other crops. Those that participated 0 more intensively in the market are more likely to Poorest Poor Middle Rich Richest Poorest Poor Middle Rich Richest Poorest Poor Middle Rich Richest Poorest Poor Middle Rich Richest Poorest Poor Middle Rich Richest be net buyers instead and hence are negatively Teff Barley Wheat Maize Sorghum impacted by food prices increase, especially for Net seller Net buyer Self-sufficient maize for which between 21 and 30 percent of households are net buyers. Moreover, even when Source: Authors’ estimates based on ESPS 2021/22. Notes: A household’s net market position is defined based on the net buyer there is evidence of crop diversification among ratio (NBR) of a food item which is calculated as net production (production- Ethiopian rural households (Tesfaye, 2022) so consumption) divided by total household consumption. It expresses the household that they can protect their consumption, most food production and consumption gap relative to a household’s expenditure. households still need to go to the market to Households are classified based on their NBR as follows: Net Buyers (NBR < -0.05); Net Sellers (NBR>0.05) and Self-sufficient (-0.05 < NBR < 0.05). complement their diets and buy non-food goods and services, hence suffering the negative impacts Farming incomes did not keep up with rising of the prices increases in the period. prices even though food prices rose faster than non-food prices because most rural households Indeed, estimates of the short run impact of rising do not produce a marketable surplus needed cereal prices suggest that the aggregate welfare to capitalize on rising food prices. This is impact among rural households is negative. evident from looking at rural households’ market The short run impact of rising food prices can be position for five major cereals (teff, wheat, maize, estimated using a net benefit ratio (NBR) which is sorghum, and barley) that are the core of Ethiopia’s calculated for each crop by taking a household’s agriculture and food economy, accounting for production net of its consumption and then dividing 31 PART 1: TRENDS, PATTERNS, AND DRIVERS OF POVERTY AND INEQUALITY IN ETHIOPIA Figure 21. Share of households selling at least Figure 22. Welfare losses from rising food one cereal crop, 2022 prices, 2022 .5 -.1 Proportionate change in consumption Share of household selling food (%) .4 -.15 .3 -.2 .2 .1 -.25 8 9 10 11 12 8 9 10 11 Log of consumption expenditure Log of per capita total expenditure (2019) Source: Authors’ estimates based on ESPS 2021/22. Notes: The net buyer ratio (NBR) of a food item is calculated as net production (production-consumption) divided by total household consumption. it by a household’s total expenditure. Net producers country in East Africa whose nominal protection will have a positive NBR, which is higher the greater rates for maize protection (at -47 percent in 2017) their surplus over their consumption of the crop. suggest that there is a price disincentive faced by The opposite is true for net buyers. On average a farmers. Other distortions in the input markets third of households are net producers of at least result in suboptimal availability of inputs such as one cereal food (Figure 21), but most are also net fertilizers, while the focus on self-sufficiency of the buyers of other crops. Therefore, the net benefit past contributed to sub-optimal land use choices ratio is aggregated over crops to determine the net (see World Bank, 2022). market position of households across many cereals they consume. The NBR is found to be negative for Conflict had a direct negative welfare impact. households across the entire welfare distribution The exact quantification of the welfare impacts in 2022, with steeper welfare declines among the of the recent conflict are complicated by the fact poorest households (Figure 22). Shimeles and that data could not be collected from the areas Woldemichael (2023) found results consistent that experience the most intense conflict. That with these patterns in a previous period-rising conflict exposure is not random - conflict events food prices only benefit land-rich rural households may occur near urban areas with more resources and that a rise in agricultural food prices increases available to capture, near strategic locations, or overall poverty rates in rural areas of Ethiopia. disenfranchised populations where recruitment is easier – means the correlation between conflict The limited benefit of rising food prices to and household welfare may not reflect the causal rural households reflects a legacy of Ethiopia’s effect of conflict. Best estimates are obtained from agriculture policies. To combat food insecurity, using the MPS 2021 and the ESPS 2018/19 and the government focused on food self-sufficiency 2021/22 household data geographically matched achieved with heavy state interventions in both to geo-coded conflict events information from input and output markets. This achieved results ACLED, to exploit the variation in distance to in terms of increasing yields for specific crops like regional borders to indirectly infer (i.e., instrument maize, but other distortions such as export bans for) household exposure to conflict since border limited supply response to prices changes. For disputes and ethnic tensions along borders have long periods of time, the domestic price for maize fueled much of the conflict in Ethiopia (Box 5). In was significantly below the international price. Amhara and Afar, the conflict spilled across the Estimates by FAOSTAT, put Ethiopia as the only border with Tigray. 32 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Box 5. Estimation of the impact of conflict on household consumption growth The impact of conflict exposure on household consumption growth is estimated using the ESPS 2019 and 2022 rounds by geographically matching households to conflict events from September 2019 to April 2022, which includes the intense country-wide spike in conflict that accompanied the war in Tigray. Since Tigray was not included in the 2022 survey due to security concerns, the analysis of household consumption does not include households in Tigray but focuses on the extensive conflict that occurred outside Tigray. An instrumental variable (IV) strategy based on the distance to the nearest internal regional border is used to estimate the impact since conflict events may occur for various reasons – for example, near urban areas with more resources available to capture, near strategic locations, or where recruitment is easier – which may be correlated with household characteristics that drive both household welfare and conflict. Distance to regional borders is a useful variation to use because border disputes and ethnic tensions along borders have fueled much of the conflict in Ethiopia over the past decade, while in Amhara and Afar, the conflict has spilled over across the border with Tigray. This is a common instrument used in the economics literature studying conflict (Hönig, 2021; Rohner et al., 2013; Serneels & Verpoorten, 2015). The required assumption is that distance to the nearest regional border affects changes in household welfare only through its effect on increased violence exposure after controlling for various geographic and household characteristics. Variable type Variables Source Conflict exposure Days with conflict events within proximity (km) of household ACLED Baseline geographic Region, nearest regional border, pre-period violent events within ESPS 2019 controls 20km, urban, Zone capital, Woreda town, distance to nearest Zone capital, distance to nearest Woreda town, whether there is a weekly market, distance to weekly market, type of road access, distance to asphalt road Baseline demographics Household size, age and education composition, household head gender ESPS 2019 Baseline welfare Log consumption per adult equivalent, log income per capita, has ESPS 2019 controls crops, has livestock, income shares of crops and livestock, farm area, tropical livestock units, dietary diversity, months of food insecurity in past year Additional controls Whether the household moved between waves (8% of the sample) ESPS 2022 This IV strategy is most appropriate in the northern regions of Amhara, Afar, and Benishangul-Gumuz borders, and we focus our analysis on these areas. In these areas, border distance is a strong predictor of conflict because of various regional border disputes (such as in Afar along the Somali-Afar border), ethnic tensions along borders (such as between ethnic Gumuz and ethnic Amhara in Benishangul-Gumuz close to the Amhara border), and the spillover from the Tigray crisis in Amhara and Afar. The excluded region with the most conflict is Oromia, where Oromo-nationalist groups largely drove conflict and was less correlated with distance to borders, making it difficult to include in the IV analysis with reliable results. The regression controls for region fixed effects and various geographic and household characteristics (see Table above). 33 PART 1: TRENDS, PATTERNS, AND DRIVERS OF POVERTY AND INEQUALITY IN ETHIOPIA The welfare impact in high conflict affected consumption and a higher likelihood of poverty. For areas has been severe. Analysis using ESPS data example, the consumption per adult equivalent for households in Afar, Amhara and Benishangul- for a household with 2 adults and 4 children is Gumuz, shows that for every additional day 10 percent lower than a household with 3 adults of violent conflict within 10km, household and 3 children but similar in other characteristics. consumption growth was 5.2 percent slower than Furthermore, consumption is much less when a the average without conflict. It was 9 percent greater share of these households’ members slower if there was a fatality (Table 6). These has no formal education. The consumption of estimates are based on data that did not cover a household with just one of three adults never some of the areas that were most intensively having been to school would be 39 percent more affected by conflict. Estimates using the MPS than the consumption of a similar household 2021 which covered Afar zones affected by the with two adults without any formal education Northen Ethiopia conflict, suggest that each among the three adults. The high dependency additional day a conflict event within 20km to ratio and low educational attainment among poor the household between November 2020 and the households is thus a contributing factor to poverty survey date, was associated with a 22.1 percent in Ethiopia. decrease in consumption expenditure. The closer the conflict event the greater the impact. The Limited access to better economic opportunities incidence and impacts of conflict will be further further limited the income generating potential discussed in Chapter 6. of households due to both underutilization and lowering returns to labor. Households with Beyond shocks, low endowments and lack a high share of unemployed adults have lower of opportunities contributed to poverty. consumption and are more likely to be poor. The consumption per capita of a household having one Low human capital endowments reduce the of three adults in a household not working will be productive potential of households, lowering 40 percent lower than when two of the three adults their welfare. A key factor is the availability and are working, all other factors being the same. quality of labor among households. A typical The more adults working in services sectors, the poor household in Ethiopia has 6 household higher the consumption per capita consumption members with 3 working-age people, hence a of the household. Recent trends that saw more high dependency ratio than non-poor households people becoming unemployed and dropping out that typically have 4 people with at least two of the labor force altogether, would have reduced working-age people. Estimates suggest that a household consumption and increased poverty. higher dependency ratio is associated with lower This will be discussed further in Chapter 7. 34 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Table 6. Estimated impacts of conflict exposure on household consumption growth Explanatory variable Proximity threshold Proximity threshold 10km 20km 30km 10km 20km 30km Days with Violent Events -0.052* -0.025* -0.012* (0.028) (0.014) (0.006) Days with Fatalities -0.090** -0.040* -0.020* (0.045) (0.021) (0.010) Controls and region fixed effects Yes Yes Yes Yes Yes Yes Observations 1027 1027 1027 1027 1027 1027 Source: Authors’ estimates based on ESPS 2018/19 and 2021/22; Armed Conflict Location & Event Data Project (ACLED); www.acleddata.com Notes: The outcome variable is the change in log of household consumption per adult equivalent between ESS 2019-2022 (spatially adjusted and adjusted for inflation). Event days are the number of days with violent events - explosions/strikes and violence against civilians excluding political arrests - within 10, 20, or 30 km of the household in between ESPS waves. Fatality days are days where at least one fatality occurred. Population weights adjusted for attrition. Standard errors clustered at the EA level. However, some of the improvements in access to close to half of households living in poverty in public services might have helped moderate the 2022 were not poor in 2019 (Ambel, 2024). While increase in poverty. The consumption per adult the period covered by this panel study is shorter equivalent of households with access to electricity than the inter-survey period used for the rest of is at least 20 percent higher than those without the chapter, it still provides strong evidence of access to electricity. At the same time, household mobility in and out of poverty in Ethiopia during a consumption is higher and the likelihood of period in which multiple economic shocks affected being poor is less, the closer the household is to households’ livelihoods. Out of the 33 percent of an all-weather road. Proximity to a food market the country identified as poor in 2022, almost half seems to affect only absolute and food poverty. (16 percent) were also poor in 2019, while the The positive impact of improved connectivity remaining (17 percent) were households that fell has been established in other studies estimating into poverty between 2019 and 2022. Among the the causal relationship between road expansion non-poor, in contrast, there is less mobility, as out programs and household welfare (Nakamura et of the 67 percent of the country identified as non- al., 2020). The observed positive impact suggested poor in 2022, almost three-fourths (49 percent) that recent improvements in access to electricity were also non-poor in 2019. The remaining fourth and roads could have helped offset the negative (18 percent) are households who escaped poverty impacts of some of the factors discussed above. between 2019 and 2022. An analysis of household characteristics shows that the chronically poor (i.e., The interplay of shocks, low poor in both 2019 and 2022) tend to have lower endowments, and poor opportunities educational achievement and lower access to basic has increased vulnerability. services; and are located further away from the local markets, compared to those who were poor Panel data evidence also highlights how in only one period. This signals the importance of exposure to economic shocks plays a role in human capital, access to services, and connectivity poverty dynamics in the country, as movements to markets to escape poverty, though, at the same out and into poverty have been common in time, they are not silver bullets that guarantee recent years. Results from the ESPS show that escaping from poverty permanently. 35 PART 1: TRENDS, PATTERNS, AND DRIVERS OF POVERTY AND INEQUALITY IN ETHIOPIA Social safety nets helped mitigate lowland areas) (IFPRI and IDS, 2022). In urban impacts among the beneficiaries. areas, the UPSNP had positive impacts in Addis Ababa by increasing public employment, improving Social assistance programs partly helped reduce local amenities, and increasing private sector wages the depth of poverty among beneficiaries. across neighbourhood (AE) by 18.6 percent (Franklin Beneficiaries of the PSNP program tend to be et al., 2024). In terms of targeting performance, the poorer – evidenced by the negative correlation UPSNP’s reliance on a combination of geographic and between being a PSNP beneficiary and household community-based targeting performs reasonably consumption. However, estimates using methods well and UPSNP beneficiaries are poorer than the that compare PSNP beneficiaries with non-PSNP poorest urban residents (Wieser et al., 2021). beneficiaries who are similar in circumstances show that the consumption of PSNP program beneficiaries Given the positive impacts and good targeting is at least 10 percent higher. This suggests that the performance, safety nets in Ethiopia should be PSNP program reduces the depth of poverty, given expanded as they are well placed to compensate that most beneficiaries are still poor. However, the existing beneficiaries in cases of future shocks program’s limited coverage of the poor limits its (such as a sharp rise in inflation). The geographical broader impacts. coverage of both rural and urban safety nets program should be expanded to cover a larger share Indeed, social safety nets in Ethiopia have played of the poor. In 2021, the UPSNP covered 4.6 percent a pivotal role in increasing food security and of the urban and PSNP 11.8 percent of the rural reducing the depth of poverty. Evidence suggests population. Together, they covered 10.3 percent that PSNP significantly contributed to poverty of Ethiopians (Table 7) which explains the limited reduction until 2016 but progress stalled since. In coverage of the poorest 20 percent households by 2015/16, at the zonal level, a one percent annualized social safety nets programs. Expanding safety net increase in PSNP coverage was associated with a programs and better aligning regional caseloads 0.1 percent annualized decrease in the poverty to needs can support the poor in responding rate (World Bank, 2020). Though few beneficiaries to economic shocks. Moreover, increasing the graduate from the program, the PSNP shows a adequacy of benefits, especially in rural areas positive and significant impact on the welfare of which are supported under the PSNP which has a households by reducing the number of months of lower adequacy of benefits than the UPSNP which food shortage and increasing livestock ownership supports urban areas—is vital to support the poor in highland areas (with no significant impact in in times of hardship. Table 7. Safety net coverage in 2021 % of population % of households All Poorest quintile All Poorest quintile National 10.3% 16.8% 9.2% 16.1% Rural 11.8% 17.8% 11.0% 17.2% Urban 4.6% 8.4% 4.0% 8.1% Number of population (thousands) Number of households (thousands) National 10,011 3,273 1,967 515 Rural 9,025 3,092 1,748 485 Urban 986 181 219 31 Source: Authors’ estimates based on HoWStat 2021. Notes: Lowest quantile based on ‘total or post-transfer’ per capita consumption aggregate. 36 ETHIOPIA POVERTY AND EQUITY ASSESSMENT PART 2 Deepening the Understanding of Drivers of Poverty The section provides an in-depth look at the key drivers of poverty trends. This focuses on three key areas— weather variability in Chapter 5, conflict in Chapter 6, and structural transformation in Chapter 7 — that have been or are expected to continue to be drivers of challenges and opportunities for poverty reduction and where there is scope for value addition from new analysis. An overview of the welfare impacts of these factors has been presented in the previous chapter. The three chapters in this section will elaborate on the welfare impacts of these factors, first by providing additional context on key trends, a nuanced discussion of the incidence or impacts across the welfare distribution, highlighting the channels for long-term welfare impacts and putting into focus some issues that had so far not been extensively analyzed. 37 PART 2: DEEPENING THE UNDERSTANDING OF DRIVERS OF POVERTY IN ETHIOPIA CLIMATE CHANGE AND (temperature and rainfall) and its variability based HOUSEHOLD WELFARE on a comparison of successive climatology periods IN ETHIOPIA through shifts in mean as well as the spread (width) of the variability are provided in Figure 23. The Ethiopia is one of the most vulnerable countries to distribution of temperature variability shows that climate variability and climate change. Frequent years are becoming hotter and/or more intense extreme weather events, such as droughts and temperatures are occurring more frequently floods, exacerbate problems like soil erosion, (Figure 24a). Average temperatures in Ethiopia have deforestation, desertification, and biodiversity increased by 0.25 °C per decade since 1960, and loss. This chapter examines the impact of climate- the average number of ‘hot days’ (the hottest 10 induced shocks and policy responses on household percent of days annually) increased by 20 percent. welfare. The chapter starts by depicting the trends Temperature increases have also led to increased of and exposure of households to climate change evapotranspiration and reduced soil moisture; variability and extremes, shows the temporal and higher rates of warming have been observed in the spatial distribution of climate and drought shocks, central regions and highland areas. Strong variability and the long-term welfare impacts of climate change- makes long-term precipitation trends for Ethiopia induced shocks – focusing in weather variability. It difficult to determine, however, an overall decline finds that poor households are more vulnerable to has been observed in the last three to four decades, weather variability risks, though vulnerability among with significant year-to-year volatility (World Bank, the non-poor is also high. The increased exposure 2021). Rains appear to be less but intense, reflecting to weather variability risks results in significantly frequent droughts and floods in recent decades lower welfare – a 16 percent decline in consumption. (Figure 24b). The combination of less but intense Recent exposure to droughts shocks increase rainfall and hotter and more intense temperatures households aversion to downward output risks which in recent decades suggest that weather variability disincentives households from adopting marketed and climatic shocks have been more common in agriculture inputs and increases households’ bias Ethiopia in the last few decades. towards expanding areas under cultivation of cereals more than other crops. Household coping strategies Indeed, the change in climate has increased to drought shocks include conservation agriculture, the frequency of extreme events and climate- soil and water conservation, improved seeds, and related shocks. The country experienced droughts organic fertilizer. Looking ahead to 2050, the chapter in the 1983-1984, 1991-1992, 1998-2000, and shows that climate change could increase poverty 2002 periods; and in 2008, 2012, and 2015/16. by an additional 2 percentage points above the no According to the EM-DAT database, 11 drought climate change scenario, but the impacts of climate events and 10 floods have been reported in the change can be moderated with the implementation country since 1990. Estimates show that about 5 of structural reforms. million people are exposed to an average drought and 0.25 million people to an average flood event Ethiopia has experienced extreme every year (World Bank, 2019). Long-term drought weather events and significant weather trends assessed using the Palmer Drought Severity variability in the past decade, mainly Index (PDSI) suggest that drought shocks were between 2016 and 2021. severe and covered most of the country (Figure 24a). The FAO Agricultural Stress Index (ASI) Ethiopia is one of the most vulnerable countries indicates a higher share of crop areas affected by to climate change and variability. The trends drought in recent periods including 2015, 2016, and changes in the distribution of climate and 2021 (Figure 24b). Overall, drought shocks 38 ETHIOPIA POVERTY AND EQUITY ASSESSMENT proxied by the Palmer Drought Severity Index 1990s, then a slight increase, and especially a (PDSI) show a downward trend until the mid- decrease in overall volatility. Figure 23. Changes in the distribution of precipitation and temperature, 1951-2020 Mean temperature Precipitation Annual rainfall and temperature 1.25 0.005 trends, 1991-2021 1 Rainfall (mm) Temperature 0.004 1,100 25 Distribution Distribution 0.75 0.003 1,000 24 0.5 0.002 24 0.25 900 23 0.001 800 0 23 21 22 23 24 0 Average Mean Surface Air 750 1,000 1,250 1,500 700 22 1991 1994 1997 2000 2003 2006 2009 2012 2015 2018 2021 Temperature (°C) Precipitation (mm) 1951-1980 1971-2000 1991-2020 1951-1980 1971-2000 1991-2020 Rainfall Temperature Source: Authors based on Climate Change Knowledge Portal Ethiopia. Notes: Each bell-shaped distribution represents a 30-year climatology interval. Figure 24. Drought evolution and trends, 1980-2021 Panel A: Palmer Drought Severity Index (PDSI) Panel B: FAO Agriculture Stress Index (ASI) 2 60 1 50 40 Measure Measure 0 -1 30 20 -2 10 0 1980 1981 1982 1983 1984 1985 1986 1987 1988 1990 1991 1992 1993 1994 1995 1996 1997 1998 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 1984 1985 1986 1987 1988 1990 1991 1992 1993 1994 1995 1996 1997 1998 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 Year 2022 Year Variable: Variable: Belg Season FAO ASI Belg Season PDSI Meher Season PDSI Annual PDSI Meher Season FAO ASI Annual FAO ASI Source: CHIRPS; FAO: https://www.fao.org/giews/earthobservation/asis/index_1.jsp?lang=en; https://hydrology.princeton.edu/getdata.php?dataid=7; Notes: The Palmer Drought Severity Index (PDSI) captures long-term drought as it relies on temperature information and a physical model of water balance using values from 1980 to 2019. Values of PDSI > 0 denote no drought conditions while PDSI < 0 denotes drought conditions. FAO ASI is based on the Vegetation Health Index (VHI) [VHI < 40 = Drought] covering the period 1984-2019. ASI looks at the percentage of an area affected by drought for a given season. Household long-term vulnerability to climate is negative for at least 3 months during the Meher shocks is high, more so for the poor. season. A comparison of the PSDI and the incidence of poverty across zones in 2021 shows that poverty The poor have also been exposed to recent rates are greater in zones that experienced more severe weather shocks more than nonpoor households. droughts, particularly in 2019 (Figure 25). However, only Households’ exposure to drought shocks can be 27 percent of the poor were exposed to a drought shock determined by overlaying household enumeration between 2019-21 for example (Figure 26a). This areas with a monthly measure of drought severity – implies that there were more poor people who were the Palmer Drought Severity Index (PDSI). A household not recently exposed to droughts than the number of is considered exposed to a drought shock if the PSDI poor people who recently experienced drought. 39 PART 2: DEEPENING THE UNDERSTANDING OF DRIVERS OF POVERTY IN ETHIOPIA Figure 25. Poverty incidence and long-term drought exposure Panel A: Palmer Drought Severity Panel B: Absolute poverty incidence, Panel C: Palmer Drought Severity Index, 2019 2021 Index, 2021 Conflicts within 20 km (N) 0 1 2-3 3-10 >10 Source: Authors’ estimates based on HoWStat 2021, HCES 2015/16, ESPS 2018/19 and 2021/22 and geospatial data. Figure 26. The share of the population and poor exposed to drought shocks Panel A: Share of population and the poor exposed to droughts Panel B: Share of households by long-term drought exposure and welfare quintile, (%) 100 11.0 57 18.4 12.2 16.3 14.4 21.9 25.9 22.3 54 53 53 80 50 50 47 46 47 89.0 43 60 87.8 Percentage 81.6 85.6 78.1 83.7 77.7 40 74.1 20 0 Share of Poor(2019 PDSI) Share of Population (2019 PDSI) Share of Poor (2020 PDSI) Share of Population (2020 PDSI) Share of Poor (2021 PDSI) Share of Population (2021 PDSI) Share of Poor (2019- 2021 PDSI) Share of Population (2019-2021 PDSI)I) Poorest 2 3 4 Richest quintile quintile Not exposed to drought Exposed to drought High drought frequency Low drought frequency Source: Authors’ estimates based on HoWStat 2021, HCES 2015/16, ESPS 2018/19 and 2021/22. But over the long term, most households in showing that the impacts of climate change are Ethiopia have been highly vulnerable to climate regressive - falling more heavily on the poor than risks, more so for the poorest households. on the rich (Skoufias, 2012) and that the incidence Household locations can be grouped into high of poverty is higher in areas of high environmental drought frequency and low drought frequency areas risk (Narloch & Bangalore, 2018). based on whether they experienced a drought shock above the median number of droughts recorded Long-term exposure to climate shocks across enumeration areas in Ethiopia, defined using intensifies poverty. the PSDI as above. By this classification, around 54 percent of households in the poorest quintiles live Long-term exposure to weather variability — in high drought frequency areas, compared to about proxied by drought frequency and heat stress— 46 percent among the richest quintile (Figure 26b). have welfare-reducing effects. Spatial differences This is consistent with findings from other studies in long-term weather variability risk exposure are 40 ETHIOPIA POVERTY AND EQUITY ASSESSMENT captured using the concept of Favored Agriculture damage or loss of assets such as livestock or Areas (FAA) which combines geographical risks grains, coupled with asset liquidation as a form of (poor soil quality) and environmental risks consumption smoothing, can deplete household (droughts, floods, heat stress, and disease resources (Carter et al., 2007). A socio-economic likelihood) using climate variables for the 1980 study of pastoral and agropastoral populations – 2019 period (Box 6). In 2021, household in Ethiopia for example, showed that drought consumption in favored agriculture areas was 43 increased livestock mortality by 60 percent (World percent more than in less favored agriculture areas, Bank, 2023). Climate shocks can also lead to a after accounting for the influence of institutional/ redistribution of disease burden, disproportionately location factors such as market access, land affecting children who may be forced out of school tenure regime, livelihood system, infrastructure, or suffer long-term health issues. For instance, and household-level factors such as education rainfall variability has been associated with reduced and crop choice (Table 8). There is also a negative human capital formation (Alderman et al., 2006; relationship between each weather variability Hoddinott & Kinsey, 2001). Recent data indicates shock (drought and heat stress frequency) and that households subjected to drought shocks are household consumption. Estimates suggest that more likely to experience health shocks than those a one percent increase in drought frequency and not affected by drought shocks (50 percent vs. heat stress reduces per capita consumption by 37.8 percent), indicating that exposure to drought 15.5 and 6.2 percent, respectively. heightens health risks. The broader impact on labor markets is negative. Disasters can reduce the Climate change-induced shocks intensify marginal productivity of labor, leading to reduced poverty by eroding productive assets and labor demand and compromised production impairing human capital accumulation. Direct (Mueller & Quisumbing, 2011). Table 8. Estimated impacts of drought exposure and agriculture potential on welfare (1) (2) (3) (4) Favored Agriculture Areas (FAAs) 0.434 *** 0.320 *** (0.060) (0.082) Log PDSI Drought Frequency -0.155 *** -0.048 (0.031) (0.033) Log Heat Stress Frequency -0.062 *** -0.040 *** (0.012) (0.012) Log Caloric Suitability Index 0.046 *** 0.107 *** (0.010) (0.011) Agroecology fixed effects No Yes No Yes Rural fixed effects No Yes No Yes Controls No Yes No Yes Observations 6,680 6,680 6,503 6,503 Source: Authors’ estimates based on HCES 2015/16 and HoWStat 2021 and geospatial data. Notes: The dependent variable is the log of daily per capita consumption. The PDSI is the Palmer Drought Severity Index that captures long-term drought exposure (1980-2019). The Caloric Suitability Index (CSI) is based on Galor & Özak (2016) and captures soil quality. Long-term heat stress frequency (1980 - 2019) is based on Baquie and Fuje (2020) who defines the experience of an episode of heat in any of the Meher months if the recorded/estimated temperature is above the long-term 98th percentile. The FAAs index is computed at the EA level using these three bio-physical variables: PDSI, Heat Stress, and CSI. The control variables included in the regressions are distance to market, distance to road, livelihood systems, water and sanitation use, reliance on solid fuel for cooking, and livestock ownership. 41 PART 2: DEEPENING THE UNDERSTANDING OF DRIVERS OF POVERTY IN ETHIOPIA Box 6. Estimation of the impact of long-term environmental risks and household welfare The long-term association between climate shocks and welfare is assessed using survey data and novel geospatial data. The survey data used are the Ethiopia Socioeconomic Panel Survey (ESSP) 2018/19 and 2021/22 collected under the LSMS-ISA program of the World Bank in collaboration with the Ethiopian Statistical Service (ESS). The other data used are the 2015/16 Household Consumption Expenditure and Welfare Monitoring Survey (HCES/WMS) and the 2020/21 HoWStat survey data. The survey data are merged with weather and other geospatial datasets extracted from the Climate Hazards Group InfraRed Precipitations with Stations (CHIRPS); FAO (GAEZ Data and Agriculture Stress Index) and PSDI database. To estimate the impact of the most recent shocks, the team linked geographic data to the latest household survey. For the case of droughts, this was done at the enumeration area level using the official geo-variables module. In respect of floods, a greater disaggregation could be made both in temporal and spatial terms. Daily precipitation estimates from NASA’s IMERG project (which were formatted as a 0.1-degree square raster layer) were transformed into vectors of the same dimensions and merged with households based on their GPS coordinates. Key variables are defined as follows: • Long-term Drought Frequency (1980 - 2019) is measured using the Palmer Drought Severity Index (PDSI) which captures shocks that occurred in at least 3 months during the May-Sept Meher season long-term drought frequency. Using the PDSI, an Enumeration Area (EA) is assumed to have experienced drought for a particular season if PDSI was below zero for any of the months of May - September (the main agriculture season in Ethiopia). • Long Term Frequency of Heat Stress (1980 - 2019) - Following Baque and Fuje (2020), we assume that an EA experienced an episode of Heat in any of the Meher months if the recorded/estimated temperature is above the long-term 98th percentile. • Galor & Ozak Caloric Suitability Index (CSI) is time-invariant and captures agriculture potential. The LFAAs (Less Favored Agriculture Areas) Index is therefore computed at the EA level using these three bio-physical variables. Assuming climate shocks are exogenous, an OLS regression is used to estimate their impact on welfare outcomes. In the baseline regressions, the association is estimated by including the climate shock variable of interest only. In subsequent specifications, additional control variables are added progressively. The additional control variables included are agro-ecological regions, location (urban, rural), biophysical factors (proxied by agro-ecological regions as well as drought, heat stress, soil quality), human and institutional factors (proxied by distances to food markets, livestock markets, all-weather roads and banks or financial institutions), access to basic sanitation and solid fuel use for cooking dummies and livestock holding. The specifications also include different livelihood system dummies including agro-pastoral, arid pastoral oasis, eastern highlands maize mixed, highland barley livestock mixed, highland perennial, highland sorghum chat mixed, highland teff mixed, highland wheat livestock, lake fish-based, livestock maize mixed, lowland maize mixed, lowland sesame mixed, rift valley fish-based, sorghum mixed, and western highland maize based on Amede et al., 2017. 42 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Price fluctuations, particularly during drought by drought (Table 9). The data further shows that periods like those in 2011, 2016/17, and 2020- drought-affected households reported reduced 2022, can trigger food inflation, worsening utilization of improved seeds (17.3 percent vs. poverty, and food insecurity. Households typically 30.2 percent), chemical fertilizer (34.4 percent vs. face shocks in the form of increased food costs, 44 percent), and improved animal breeds (0.17 which can severely impact their ability to meet percent vs. 0.09 percent). The evidence suggests basic needs. These shocks disproportionately that drought amplifies economic hardships affect poor households, who depend heavily on by reducing the ability to invest in essential subsistence agriculture, more than their non- agricultural inputs. poor counterparts. However, in Ethiopia, studies show that the price impact of drought shocks is considered moderate due to the mitigating effects of infrastructure and social safety nets (Hill & Fuje, 2020). The results on the impact of PSNP on household consumption showed that the program increases consumption among beneficiaries by at least 10 percent more than comparable non-PSNP beneficiary households. Exposure to weather shocks reduce future incomes by lowering agricultural productivity and reducing farmers’ market orientation. Recurrent climate shocks have been shown to diminish agricultural productivity and income. Adverse weather conditions could have an immediate impact on agricultural production. A basic comparison shows that households affected by droughts experienced 3.5 times larger crop value losses compared to households not affected Table 9. Descriptive statistics on the conditions of households based on drought shock No drought Drought Mean difference Crop value lost 41.1 143.4 -102.3*** Improved seed use 0.30 0.17 0.13*** Chemical fertilizer use 0.44 0.34 0.10*** At least one improved animal 0.17 0.09 0.08*** Health shock 0.38 0.50 -0.12*** Employment shock 0.02 0.02 0.00 Participation in off-farm activities 0.31 0.24 0.07*** Source: Authors’ estimates based on ESPS 2018/19 and 2021/22. Notes: Self-reported drought shock (*** p0.01, ** p0.05, * p0.1). 43 PART 2: DEEPENING THE UNDERSTANDING OF DRIVERS OF POVERTY IN ETHIOPIA Table 10. Arrow-Pratt (AP) absolute and downside (DS) risk aversion by weather shocks (1) (2) Arrow-Pratt 0.126*** (0.002) 0.127*** (0.002) Downside risk 0.564*** (0.002) 0.563*** (0.002) Drought (t - 1) 0.001** (0.000) Drought (t - 2) 0.003*** (0.001) Drought (t - 3) -0.000 (0.000) Arrow-Pratt * Drought (t - 1) -0.009*** (0.003) Arrow-Pratt * Drought (t - 2) -0.009* (0.005) Arrow-Pratt *Drought (t - 3) -0.001 (0.002) Downside risk * Drought (t – 1) 0.007** (0.003) Downside risk *Drought (t - 2) 0.004 (0.006) Downside risk *Drought (t - 3) -0.001 (0.003) Observations 1,660 1,660 Source: Authors’ estimates based on ESPS 2018/19 and 2021/22. Notes: The Arrow-Pratt and downward risk preferences are estimated using the moment approach designed by Antle (1983, 1987). Details are provided in Box 7. Bootstrap standard errors for (300 replications) are in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Drought is a binary (0/1) variable. Farmers are less willing to take output variability showing less willingness to take risks that could risks, in particular the downward output risks. lead to lower outputs (Table 10). This is consistent Households in Ethiopia exhibit a general aversion with findings from similar studies (see Mulungu et to output variability and a specific reluctance to al. (2023) and Di Falco and Vieider (2022)). The take risks that could result in lower yields (Table influence of drought on risk preferences among 10). The positive and statistically significant farmers persists for up to two years and becomes Absolute Pratt (AP) coefficient underscores insignificant by the third year, suggesting that the farmers’ general risk aversion, particularly impact of a drought shock on risk preferences regarding output variability. The downside lasts beyond the drought year but wanes over risk (DS) aversion coefficient further reveals a time. Some studies indicate that households may reluctance to accept risks that could lead to yields change their risk tolerance in response to policy falling below a certain threshold. Farmers who and production shocks (Bozzola & Finger, 2021), have recently experienced a drought shock are hence the waning effect of a current drought more cautious about potential negative outcomes, shock on risk preferences. 44 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Box 7. Estimation of the impact of climate shocks on household risk preferences The impact of adverse rainfall shocks on farmers’ risk preferences is assessed by estimating the Arrow- Pratt and downward risk preferences using the moment approach designed by Antle (1983, 1987). This approach avoids potential estimation problems arising from directly inserting the risk parameters into a production function. Because of the approach’s flexibility, agricultural technology and risk parameters can be structurally estimated with no restrictions on the utility function (Mulungu et al., 2023). In this approach, the higher moments of the production distribution reflect the risk exposure. The first moment stands for the average expected yield, while the second and third moments, respectively, represent the output variability and the downside risk. Assume that a farmer produces output y=g(x;υ) using a set of inputs X={s,f}, where f and s stand for fertilizer and improved seed, respectively, and υ stands for uncertainty factors. The farmer chooses input X to maximize her expected utility, which is a function of the first three moments of the output distribution. max Eu (y)= U (τ1(X), τ2 (X), τ3 (X)) (1) fs Where τ1(X) is the first-order moment of the production distribution which stands for the expected yield and is given by E(y(X)). The second and third moments of the production distributions are represented by τ2(X) and τ3(X). They are the square and cubic of the error terms. i.e., τ2(X) = ( y(X) � E( y(X)))2 and τ3(X) = ( y(X) � E( y(X)))3. We can determine the conditions for each farmer to make the optimum use of the inputs by computing the first-order partial derivatives of equation 1 with respect to the inputs (improved seed and fertilizer) and equating them with zero. This implies that the mean output is a linear summation of the contributions of each input to output variance and skewness. Therefore, the third-order Taylor series expansion of the expected utility function around the expected output can be used to estimate our paraments of interest, the Arrow-Pratt and downside risk aversion coefficients. This can be done by approximating and rewriting u'' (τ1) � 2θ2 ≈ AP = � and 6θ3 ≈ DS � u''' (τ1) u' u' the first partial derivatives as in equations 2 and 3, to estimate the AP and DS coefficients, where ε f and ε f are the usual random errors. We follow Mulungu et al. (2023), Bozzola and Finger (2020), h h and others, and imposed the restrictions that APf = APs = AP and DSf = DSs = DS to refrain from input- specific risk aversion parameters. Besides, we standardized the dependent and input variables to have a mean of 0 and a standard deviation of 1 to make the marginal effects comparable. ϑτ1h 1 ϑτ2h 1 ϑτ3h f = APf DSf +ε (2) ϑfh 2 ϑfh 6 ϑfh h ϑτ1h 1 ϑτ2h 1 ϑτ3h s = APs DSs +ε (3) ϑsh 2 ϑsh 6 ϑsh h The above theoretical model can be empirically solved by estimating the effects of the inputs on the first three moments of the output distribution and then estimating the risk aversion parameters AP and DS using the 45 PART 2: DEEPENING THE UNDERSTANDING OF DRIVERS OF POVERTY IN ETHIOPIA marginal expected effects of these inputs. Thus, to estimate the impact of the inputs on the first three moments of the production distributions, we first predict the mean, variance, and skewness. The first central moments of the production distribution can be given by equation 4: y = g(X, V, υ), = τ1 (X, V; φ) + ε (4) Where E[ g(X, υ) | X, V ] = τ1 (X, V; φ ). X and V are the vectors of the main inputs and other controls. The control variables used in the analysis include the gender of the household head, the size of agricultural land, access to agricultural extension service, distance to market, and share of adults in the household. ε stands for the error term. Using the same functional form, the second (variance) and third (skewness) central moments of output can be represented as follows: E {[g(X, V, υ), � τ1 (X, V; φ)2 |X,V} = E [ε2 |X,V ] = τ2 (X, V; ϑ) (5) E {[g(X, V, υ), � τ1 (X, V; φ)3 |X,V} = E [ε3 |X,V ] = τ3 (X, V; ρ) (6) The parameter vectors of the mean, variance, and skewness equations are denoted, respectively, by φ,ϑ and ρ. To estimate the impacts of the inputs on the three moments of the production distribution, we follow prior studies and use the quadratic production function, which considers the input levels, their squared terms, and the interaction of the two inputs along with some control variables. Their functional form is given below using equations 7 to 9. yht = φ0 + φ1 fht + φ2 (fht)2 + φ3 sht + φ4 (sht)2 + φ5 (fht sht) + δht + (T * r)t + εht (7) (εht)2 = ϑ0 + ϑ1 fht + ϑ2 (fht)2 + ϑ3 sht + ϑ4 (sht)2 + ϑ5 (fht sht) + δht + (T * r)t + εht (8) (εht)3 = ρ0 + ρ1 fht + ρ2 (fht)2 + ρ3 sht + ρ4 (sht)2 + ρ5 (fht sht) + δht + (T * r)t t + εht (9) yht is output in kilogram and (εht)2 (εht)3 are the variance and skewness of the output distribution, respectively. δh and (T * r)t are households fixed effects and the interaction of region dummy and survey period. Once equations 7 to 9 are estimated, the marginal effects of the production distribution moments with respect to the inputs are predicted. These predictions are then used as inputs to estimate the AP and DS risk aversion coefficients. Hence, we solve the system of Equations (2) and (3) using a three-stage least squares model that helps us account for the correlation between the equation error terms. We estimate the above equations using a fixed effects method. This framework provides some key advantages that help us strengthen our identification. It enables us to account for any time-invariant unobserved household and community-level characteristics, such as soil fertility and other environmental conditions that have an impact on the types of crops a community plant and its risk-coping mechanisms. We also included household, and community-level controls, and the interactions of regional dummies with the survey period to account for any regional policy changes or other shocks, such as conflict, price, and other shocks, as well as other state-level policy changes that might affect demand and supply of commercial inputs and farm households' response to drought shock. This analysis is based on data from the Ethiopian Socioeconomic Panel Survey (ESPS) 2018/19 and 2021/22 data. ESPS compiles data on a variety of topics from a panel of households using five modules—households, 46 ETHIOPIA POVERTY AND EQUITY ASSESSMENT community, and three agricultural modules. The rainfall data comes from the Climate Hazards Group Infrared Precipitation with Stations (CHIRPS). CHIRPS provides gridded datasets from 1981 to the present at various temporal resolutions, with a spatial resolution of 0.05° (Funk et al., 2015). For analysis, we considered rainfall conditions during the agricultural season of the country. We integrate our climate data with the ESPS using the geographic identifiers included in the ESPS dataset. We calculate the weather shocks by using the standard definition of rainfall anomalies established in the relevant publications. More precisely, we use the Standard Precipitation Index (SPI) as our weather condition indicator. We begin by calculating the historical average and standard deviation of growing season rainfall using rainfall data from the years between 1981 and 2020. The difference between the annual rainfall and the historical mean is then calculated. The difference is then normalized by dividing it by the historical standard deviation. Following the related works of Mulungu et al. (2023) and others, we chose SPI less than negative 0.5 to represent drought conditions. The change in household risk preference Dhakal et al., 2022; Sardar et al., 2021; Tesfaye influences their technology adoption or et al., 2021), while enhancing the resilience innovations. Farmers’ decisions are informed by of vulnerable farmers while reducing and/or their understanding of climate, which is built upon removing greenhouse gas emissions (FAO, 2020; historical and present weather patterns which Torquebiau, 2017; van Wijk et al., 2020). The influences their expectations about future weather uptake of agricultural technologies and farming conditions, thereby affecting their decisions related practices is influenced by exposure to climate to the use of commercial inputs, crop choices, and shocks such as drought. For instance, the use land allocation (Ahmed et al., 2023; Cui, 2020; Cui of yield-increasing technologies like inorganic & Xie, 2022; Jagnani et al., 2021). Farmers might fertilizer is more prevalent in areas with low avoid adopting new agricultural inputs or practices frequency drought areas and in favored than that could increase their risk of yield variability and less-favored agriculture areas (48 vs 35 percent downside risk (Dercon & Christiaensen, 2011). On in 2022). Conversely, in areas where droughts the flip side, they may be more inclined to adopt are a common occurrence and in non-FAA, production techniques that reduce the likelihood farmers are more likely to adopt risk-reducing of negative yield deviations to minimize the risk of and resource-conserving practices such as adverse outcomes (Bozzola & Finger, 2021). conservation agriculture compared to other agriculture practices (Figure 27). Agricultural households employ a range of coping strategies and responses that Further analysis indicates that asset sales are a enhance resilience to climate shocks. common risk-coping strategy among households in high drought frequency areas to buffer against Agricultural households in Ethiopia utilize a the adverse welfare effects of climate shocks. wide range of strategies to cope with climatic However, the data do not show variation in the shocks. Prominent agricultural technologies use of ex-post shock coping mechanisms such and improved practices (innovations) such as as reliance on savings, social insurance (help improved seeds, irrigation, organic fertilizer, from government or social networks), and dietary soil and water conservation, and conservation changes, based on the level of drought exposure. agriculture have been shown to positively Diversification in the form of income, crop variety, affect both agricultural productivity and welfare and labor is a cornerstone of resilience to climate (Di Falco et al., 2011; (Di Falco et al., 2011; shocks. However, income diversification strategies, 47 PART 2: DEEPENING THE UNDERSTANDING OF DRIVERS OF POVERTY IN ETHIOPIA including increasing the number of income sources appear to vary with the degree of exposure to and the shift towards non-farm income, do not climate shocks in recent years in Ethiopia. Figure 27. Adoption of agriculture technologies and farming practices Panel A: Low vs high drought frequency Panel B: Favored Agriculture Areas (FAA) vs non-FAA Soil and water 56% Soil and water 58% conservation 57% conservation 53% Organic 43% Organic 50% fertilizer 49% fertilizer 39% Improved 31% Improved 25% seeds 22% seeds 28% Inorganic 40% Inorganic 43% fertilizer 42% fertilizer 36% High drought frequency Low drought frequency FAA Non FAA Source: Authors based on ESPS 2021/22. Instead, cash transfers, food aid, and insurance larger impacts in rural areas and highland areas. are pivotal in helping households with the The long-term impacts of climate change are adverse effects of climate shocks. Extensive modelled under wet-warm and dry-hot climate studies show that cash transfers help farmers scenarios, in the absence of structural reforms move out of poverty, accumulate assets, and (constrained growth - CG) or with structural improve overall well-being in the aftermath of reforms implemented (REF), yielding four shocks (Abay et al., 2022; Porter & Goyal, 2016). scenarios. Poverty is expected to decline as GDP is Despite this, cash transfers do not always result in projected to expand, translating into consumption lifting households out of poverty, a phenomenon growth, but the expected impacts vary spatially termed the “social protection paradox.” This and depending on the climate change scenario. was observed in Latin America and Kenya, The dry-hot, constrained growth scenario yields where cash transfers alone were insufficient the largest poverty impacts by 2050, with around for poverty alleviation (Ikegami et al., 2017). a 2 percentage points increase in poverty in rural While its effectiveness on asset growth has areas and the highland climatic zones. Under the been questioned, as seen after the 1998-2000 wet-warm scenario, this rate is 1.2 percentage Ethiopian drought (Carter et al., 2007), analysis points and 1.5 percentage points in rural areas and from recent data in Ethiopia finds evidence of the highland climatic zones, respectively (Figure 28). positive welfare impacts of social assistance. Increases in poverty rates are, in all cases, The impact of climate shocks can be expected to be smaller under the structural moderated with the implementation of reforms (REF) scenario than in the constrained economic reforms. growth (CG) scenario. The microsimulation results show that the poverty impacts of climate change Climate change will lead to increases in are lower under the REF scenario, with lower monetary poverty under all the different impacts on poverty as time passes, signaling that scenarios modelled in the next 25 years, with the policies implemented under this scenario could 48 ETHIOPIA POVERTY AND EQUITY ASSESSMENT largely counter the impacts of climate change in income-generation opportunities for households, the mid-term (Figure 28). Under the wet-weather either through salaried work, self-employment, scenario, the impact of climate change on poverty or agricultural production. These increases in would be negligible under the reforms compared to household earnings will counter the increases in a percentage point increase under the constrained poverty expected due to climate change under the growth scenarios. The increase in poverty by 2050 CG scenario. Poverty rates are expected to decline will be at least 1 percentage point less if reforms further with climate actions that involve adaptation are implemented than without them under the actions to cope with the impacts of climate change hot-dry scenarios. The REF scenario includes and strengthen resilience to climate shocks and reforms that will accelerate growth and improve mitigation measures, over and above implementing productivity, which will translate into more structural reforms. Figure 28. Impact of climate change scenarios on poverty rates Panel A: Location Panel B: Climate zone 2.2 2.0 Increase in poverty rate 1.8 (Percentage points) 1.6 1.4 1.2 1. 0 0.8 0.6 0.4 0.2 0.0 CG REF CG REF CG REF CG REF CG REF CG REF CG REF CG REF CG REF CG REF CG REF CG REF 2030 2040 2050 2030 2040 2050 2030 2040 2050 2030 2040 2050 Wet-warm Dry-hot Wet-warm Dry-hot National Urban Rural Highland Temperature Lowland Source: World Bank simulations using MFMOD inputs and HCES 2016. Notes: The chart shows the increase in poverty rates ($2.15 in 2017 PPP) due to Dry-hot and Wet-warm climate change scenarios, with respect to the CG and REF baseline scenarios (no climate change, no adaptation). 49 PART 2: DEEPENING THE UNDERSTANDING OF DRIVERS OF POVERTY IN ETHIOPIA CONFLICT AND HOUSEHOLD 2020, fighting erupted between the Ethiopian WELFARE IN ETHIOPIA National Defense Forces (ENDF) and the Tigray People’s Liberation Front (TPLF) which continued This chapter examines the impact of conflict for two years until a peace treaty was signed in on household welfare. It analyzes the trends November 2022. By then, the war had spread far and incidence of conflict, quantifies the welfare into the neighboring regions of Amhara and Afar. impacts, and zooms in on Internally Displaced People (IDPs). The analysis shows that conflict The onset of the war in Tigray in November 2020, spread more broadly across the country after the marked a drastic increase in violent events and escalation of conflict in Tigray which spilled to conflict fatalities across Ethiopia. Data from other regions resulting in a 10-fold increase in the the Armed Conflict Location & Event Data Project number of conflict events and fatalities recorded (ACLED) shows the frequency and location of during 2020-22 compared to the two years before. conflict across Ethiopia. It shows the prevalence of The number of IDPs rose to an all-time high of 4.4 conflict events (which include battles, explosions, million during December 2021 and January 2022, and violence against civilians) in the years leading of which 2.4 million were driven by the conflict up to the crisis, as well as the sharp increase in in Northern Ethiopia. Estimates show that for violence after November 2020 (Figure 29). Across every additional day of violent conflict within 20 the country, 3,153 conflict events were recorded, kilometers, household consumption grew by 2.5 resulting in 19,113 fatalities between November percent slower, or 4 percent slower if there was a 2020 and the end of 2022. This is around 10 times fatality. Those displaced live in camps or hosting greater than the number of conflict events and communities lacking public services, especially fatalities in the same period length before the war. water and electricity. The majority of IDPs—and nearly all displaced from Tigray—prefer to return, Figure 29. Evolution of conflict in Ethiopia followed by local integration among those who have 3,000 25 been displaced longer. 2,500 200 Escalation of the conflict in Tigray 2,000 Violent Events Fatalities 150 spread the conflict more across the 1,500 country, with increasing intensity. 100 1,000 500 50 Ethiopia has grappled with internal conflict and ethnic tensions in the past. In the years leading up 0 0 Jan-17 Aug-17 Mar-18 Oct-18 May-19 Dec-19 Jul-20 Feb-21 Sep-21 Apr-22 Nov-22 to the recent conflict in Northern Ethiopia, various disputes caused death and displacement in various regions across the country. Examples include the Fatalities Violent Events contestation over the Oromo-Somali regional Source: Authors’ calculations using data from Armed Conflict Location & border, over the administrative designation of the Event Data Project (ACLED); www.acleddata.com Konso woreda in SNNP region, and between Gedeo Notes: Violent events include battles, explosions/remote violence, and and Guji Oromo tribes in West Guji. In November violence against civilians. 50 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Figure 30. Dispersion of conflict events before and after November 2020 Panel A: Nov 2018 – Oct 2020 Panel B: Nov 2020 – Oct 2022 Panel C: Nov 2020 – Oct 2022 1,200 1,000 800 600 400 200 0 Afar Amhara Ben. Gumuz Gambella (2 l) Oromia SNNP Sidama Somali South West Tigray Source: Authors calculations based on ACLED data: www.acleddata.com. Each dot represents a conflict event. During the Tigray war, conflict increased across the Figure 31. Number of days with conflict events by consumption quintile entire country—not only in Tigray and bordering regions—with various causes indirectly related 10 to the conflict in Tigray. The increase in conflict events was not restricted to the North (Figure 30). 8 During the 2-year period of the Tigray conflict, 37 percent of all conflict events happened outside of Tigray, Amhara, and Afar, the majority in the region 6 of Oromia, Ethiopia’s largest region by population. The conflict events in the rest of the country had 4 many causes indirectly related to the war in Tigray. 1 2 3 4 5 First, increased ethnic tensions fueled conflict, for Consumption Quintile example, between ethnic Gumuz militia and ethnic Amhara in the Benishangul-Gumuz region (IGC, Source: Authors’ calculations using ESPS 20118/19; Armed Conflict Location 2022). Second, various long-standing disputes & Event Data Project (ACLED); www.acleddata.com. Notes: Events days count days when a conflict event is registered within 20 over internal regional boundaries between ethnic km of the household between September 2019 and December 2022. Distance militias intensified during the war, for example, in is relative to a household geocoded location in 2019. Conflict events include Afar along the Somali border (EPO, 2023). Third, battles, remote violence, and violence against civilians. Population weighted sapped federal resources emboldened ethno- means and 95% confidence intervals are presented. nationalist groups in Oromia, who participated in clashes with local and federal troops, violence and 2022. Households with high consumption in against civilians, and formed alliances with the the 2019 ESS experienced more days with a conflict TPLF against the Ethiopian government (BBC, event within 20km of the household from the end 2021; EPO, 2021; IGC, 2022). of the survey through 2022 (Figure 31). However, the positive relationship between consumption and Recent conflict events have been conflict exposure disappears when considering concentrated in urban centers, regional welfare differences (Annex 3, Table A.3.6) translating into increased exposure as the household pattern was driven by differences among better-off households. across regions rather than within regions. In other words, the regions with the lowest consumption On average, across Ethiopia, better-off areas had also had the lowest conflict exposure (particularly the highest exposure to conflict between 2019 in Somali and SNNP). Within regions, there is also 51 PART 2: DEEPENING THE UNDERSTANDING OF DRIVERS OF POVERTY IN ETHIOPIA no relationship between household demographic days, experience a corresponding consumption characteristics and the probability of experiencing decrease of 17.5 percent. Projections based on conflict. the estimated impact of conflict on consumption growth in Afar, Amhara and Benishangul-Gumuz Conflict was more likely to occur near Zone suggest that the poverty rate—based on the capitals and in places where conflict occurred international extreme poverty line—in this sample in the recent past. Households in Zone capitals in 2022 was 9.2 percentage points higher than it experienced 4.9 more days with violent conflict would have been in the absence of conflict. The events relative to rural areas in the same region effect of conflict is concentrated on food rather and 4 more days relative to other urban areas than non-food consumption or the diversity of (Annex 3, Table A.3.6). For every conflict event food (Table 11). This is corroborated by estimates that occurred in the same length of time before based on the MPS 2021 which show a large September 2019, the number of subsequent significant adverse impacts of conflict exposure conflict events increased by 0.36 days. This is on caloric intake and food poverty among pastoral consistent with how conflict events often persist communities. In addition to disruptions to harvests in the same location over time in many different and livestock losses, conflict reduced households’ settings worldwide (Bazzi et al., 2022). The access to food through supply chain disruptions higher conflict exposure in urban centers—where and sharp price increases. households are on average better off than in rural areas—partly explains the higher conflict exposure Conflict has long-term impacts on household among better off households. This is backed by welfare. Loss of assets and the effects caused multivariate analysis showing that the correlation by forced displacement, migration and killings, between initial household consumption level is injuries, or the recruitment of fighters that change both weaker and statistically insignificant when household composition and alter household the household location and past events are labor supply, are other channels through which accounted for. conflict impacts household welfare. Conflict can also affect human capital development. For Conflict had broader impacts. instance, conflict in childhood can lead to poorer lifetime health (Akresh et al., 2012; Bundervoet Conflict exposure has an immediate adverse et al., 2009). The number of conflict days is impact on monetary poverty among severely found to lead to an increase in the prevalence affected communities. The median household of child wasting by 19.3 percentage points in Ethiopia (excluding households in Tigray) among pastoral communities (World Bank, experienced 1 conflict event day within 20km of its 2022b). These negative health effects can location between September 2019 to April 2022, extend intergenerationally, leading the original corresponding to a consumption decrease of 2.5 victims’ children to also have worse outcomes. percent based on the estimates presented earlier. Therefore, human capital losses due to conflict It means that households with higher conflict lower people’s lifetime productivity and earnings exposure—at the 75th percentile—of 7 event and reduce intergenerational economic mobility. 52 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Table 11. Estimated impacts of conflict exposure on household consumption growth Explanatory variable 1 2 3 4 Log Log food Long non-food Dietary consumption consumption consumption Diversity Days with Violent Events -0.025* -0.034** -0.009 -0.025 (0.014) (0.016) (0.026) (0.043) Region fixed effects Yes Yes Yes Yes Geographic and household controls Yes Yes Yes Yes Observations 1,027 1,027 1,027 1,027 Source: Authors’ estimates based on ESPS 2018/19 and 2021/22; Armed Conflict Location & Event Data Project (ACLED); www.acleddata.com. Notes: The outcome variables measure is the change between 2019 and 2022 by geographically matching ESPS households to conflict events during September 2019 to April 2022, which includes the intense country-wide spike in conflict that accompanied the war in Tigray. The analysis does not include households in Tigray because the region was not covered in the 2021/22 survey, therefore analysis focuses on the extensive conflict that occurred outside Tigray. Event days are the number of days with violent events - explosions/strikes and violence against civilians excluding political arrests - within 10, 20, or 30 km of the household in between ESPS waves. Fatality days are days where at least one fatality occurred. Population weights adjusted for attrition. Standard errors clustered at the EA level. Dietary diversity is the number of food groups consumed in the past week (range 0 to 12). Conflict has driven up internal of the country between 2019 and 2022. Before displacement for longer periods. November 2020, there were around 100,000 IDPs reported in Tigray, corresponding to around Internal displacement increased drastically 5 percent of IDPs in Ethiopia. By mid-2021, the across the country during the Tigray conflict. A number of IDPs in Tigray increased to over 2 year into the conflict, the number of IDPs rose to 4.5 million (Figure 33a, Panel A), close to half of the million people, with at least 2.4 million driven by national number of IDPs. All displacement in the conflict in Northern Ethiopia according to lower Tigray over this period was attributed to conflict. bound estimates from the Displacement Tracking The number of IDPs in Tigray fell to around 1 Matrix (DTM) of the International Organization for million by the end of 2022, but still accounted Migration (IOM) based on the December 2021 for one quarter of the total number of IDPs in the through January 2022 observation round. At country. In the rest of the country, the number the time, this represented the “highest annual of IDPs displaced due to conflict increased displacement figure ever recorded for a single gradually from around 1 million at the start of country” (IOM, 2023). At the end of the conflict 2020 to 1.9 million at the end of 2022, while in November 2022, this number was still at 4.4 the number of IDPs displaced due to climate million (Figure 32a). Displacement was widely increased from 0.5 million to 1.2 million over the disbursed across the country and mostly driven same period (Figure 33b, Panel B). The increase by conflict (Figure 32b). in conflict-induced displacement occurred in almost all regions of the country. However, it Internal displacement in Tigray increased is particularly notable in Amhara, Afar, Somali, suddenly and drastically during the conflict and Oromia, as well as a sudden increase in in Tigray, while it steadily doubled in the rest Benishangul-Gumuz in mid-2022. 53 PART 2: DEEPENING THE UNDERSTANDING OF DRIVERS OF POVERTY IN ETHIOPIA Figure 32. Internal displacement trends Panel A: Nov 2018 – Oct 2020 Panel B: Location of IDP Sites, Nov 2022- Jun 2023 5 4 3 2 1 0 May-18 Sep-18 Jan-19 May-19 Sep-19 Jan-20 May-20 Sep-20 Jan-21 May-21 Sep-21 Jan-22 May-22 Sep-22 Source: Authors calculations using IOM DTM Site Assessments and Emergency Site Assessments (which covered parts of Tigray, Amhara, and Afar throughout 2021). Notes: Each blue point in Panel A and red dot in Panel B represents a separate site assessment. Site Assessments between March-August 2022 were excluded from Panel A because they did not include Tigray. Key informants are used to identify sites with a reported 20 or more IDP households, then site visits and focus group discussions are conducted to estimate the number and characteristics of IDPs in each site. Each Site Assessment (SA) round, which typically occurs 4 times a year, presents an estimated snapshot of the IDP situation in the country. However, it is not necessarily a representative sample of IDPs because coverage of many sites is severely limited by inaccessibility due to conflict, sites with less than 20 IDP households are excluded, and self-settled IDPs in urban areas are often missed. These numbers should therefore be seen as estimates that are lower bounds. The November 2022 Site Assessment includes Tigray and was implemented through June 2023. Given limited data coverage, these should be seen as lower bounds. Figure 33. Millions of IDPs by reason for displacement Panel A: Tigray Panel B: Rest of Ethiopia 2.5 4 2 3 1.5 2 1 1 0.5 0 0 Nov-19 Apr-20 Sep-20 Feb-21 Jul-21 Dec-21 May-22 Oct-22 Nov-19 Apr-20 Sep-20 Feb-21 Jul-21 Dec-21 May-22 Oct-22 Climate Conflict Other Climate Conflict Other Source: Authors calculations using IOM DTM Site Assessments and Emergency Site Assessments. Notes: Each site reported the most common reason for displacement in the site. Climate shocks were also a contributing “Other ” category in Figure 33) for 7 percent factor to internal displacement outside the of IDPs. Displacement due to climate shocks Tigray region. At the end of 2022, conflict increased sharply towards the end of 2021. was the primary cause of displacement for This increase is mostly reflecting the increase 67 percent of IDPs in Ethiopia, climate for 26 in drought events in Somali, Oromia, and Afar percent of IDPs, and social tension (the largest regions (IOM, 2023). 54 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Figure 34. Average age of sites by region communities increased steadily from 37 percent (December 2022) at the end of 2019 to 49 percent at the end of 50 44 2022 (Figure 35). In Tigray, IDPs initially settled 40 38 into camps and collective centers, but by June 2021 33 34 35 30 32 the majority were hosted in host communities. To 29 30 22 characterize the socioeconomic conditions of IDPs 20 in Ethiopia, we look at the characteristics of IDP 10 settlements over time. 0 Internally displaced people have limited Afar Amhara Ben. Gumuz Gambella Oromia SNNP Sidama Somali Tigray access to services. Source: Authors’ calculations using IOM DTM Site Assessments and Emergency The quality of housing, electricity, and water Site Assessments. services in IDP sites across Ethiopia is lacking - a persistent trend since 2019. Outside of Tigray, In December 2022, the average displacement almost 80 percent of IDPs were in sites where site had been open for 35 months, highlighting over three-quarters of IDPs did not have access the long-term nature of internal displacement in to electricity, and over 50 percent are in sites where Ethiopia. This duration is relatively constant across there were complaints about water quality. Just regions (Figure 34). It is lowest in Benishangul- under half of IDPs were in sites where roughly one- Gumuz, where there was a surge in displacement quarter were in shelters that did not protect them in mid-2022. Across the other regions with the most from the weather, while overcrowded and open- IDP sites – Tigray, Amhara, Oromia, and SNNP – the air shelters were less common, with few changes average site age varies from 32 to 35 months, and since 2019 (Figure 36). Multiple observations of it is 44 months in Somali. Almost 20 percent of IDP the same displacement site over time make it sites across the country – and 42 percent in Somali possible to measure changes in site services over - had been open for at least 5 years. time. This reveals a gradual increase in the quality of shelters, a decrease in the quality of water Recently, IDPs are more likely to settle in services, and no change in the share with access host communities. The share of IDPs in host to electricity (Table 12). Figure 35. Millions of IDPs by settlement type Panel A: Tigray Panel B: Rest of Ethiopia 2.5 4 2 3 1.5 2 1 1 0.5 0 0 Nov-19 Feb-20 May-20 Aug-20 Nov-20 Feb-21 May-21 Aug-21 Nov-21 Feb-22 May-22 Aug-22 Nov-22 Nov-19 Feb-20 May-20 Aug-20 Nov-20 Feb-21 May-21 Aug-21 Nov-21 Feb-22 May-22 Aug-22 Nov-22 Camp/Center/Settlement Host community Camp/Center/Settlement Host community Source: Authors calculations using IOM DTM Site Assessments and Emergency Site Assessments. Notes: Each site reported the most common reason for displacement in the site. 55 PART 2: DEEPENING THE UNDERSTANDING OF DRIVERS OF POVERTY IN ETHIOPIA In Tigray, on the other hand, some IDP camps IDPs lived in sites where electricity infrastructure evolved during the crisis to meet needs while and water service or tap infrastructure was non- others worsened. At the onset of the crisis (in existent, destroyed, or mostly not functioning. As December 2020), 63 percent of IDPs were in sites the crisis continued, the share of Tigrayan IDPs in with low-quality shelters (this time defined as “self- sites with low-quality or open-air shelters reduced constructed and below standard”), 58 percent were to below 10 percent, but shelter overcrowding in sites with over-crowded shelters, and 52 percent increased, and by August 2021, more than 90 were in sites where over a quarter were sleeping in percent of sites had dysfunctional electricity and the open air (Figure 36a, Panel A). Nine out of ten water services. Figure 36. Trends in site service quality attributes Panel A: Tigray (ESAs) – share of sites Panel B: Rest of Ethiopia (SAs) – share of sites 1 1 0.8 0.8 0.6 0.6 0.4 0.4 0.2 0.2 0 0 Nov-19 Feb-20 May-20 Aug-20 Nov-20 Feb-21 May-21 Aug-21 Nov-21 Feb-22 May-22 Aug-22 Nov-19 Feb-20 May-20 Aug-20 Nov-20 Feb-21 May-21 Aug-21 Nov-21 Feb-22 May-22 Aug-22 Shelters: Low Quality Shelters: Open Air Shelters: Low Quality Shelters: Open Air Shelters: Overcrowded Poor Electricity Shelters: Overcrowded Poor Electricity Poor Water Poor Water Source: Authors calculations using IOM DTM Site Assessments and Emergency Site Assessments. Notes: Quality indicators are binary variables measuring service quality at the site level, and the definitions vary between Tigray (ESAs) and the rest of Ethiopia (SAs). Shelters are “low quality” if over 25 percent are self-constructed and below standard (Tigray) or over 25 percent do not protect from weather (rest). Shelters are “open air” if over 25 percent are outside or in open space (Tigray and rest). Shelters are “overcrowded” if over 25% are within 3 meters (Tigray) or over 25 percent are overcrowded or congested (rest). Poor electricity indicates that electricity infrastructure is non-existent, destroyed, or mostly not functioning (Tigray) or over 75 percent do not have access to electricity (rest). Poor water indicates that water service or tap infrastructure is non-existent, destroyed, or mostly not functioning (Tigray) or if there are complaints about drinking water (rest). 56 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Table 12. Relationship between site outcomes and site type and age Explanatory variables Outcome variables Key variable Other Controls Number Low-Quality Poor Poor of IDPs Shelters Electricity Water Host Community Region and Round Fixed Effects, -3509.80 -0.34*** -0.21*** 0.05 Site controls (2438.61) (0.03) (0.03) (0.04) Years Since Site Site and Round Fixed Effects 542.14 -0.05*** 0.01 0.03** Opened (330.97) (0.02) (0.02) (0.01) Number of 21,504 13,741 15,722 13,525 observations Number of Sites 3,899 3,211 3,315 2,988 Source: Authors’ calculations using IOM DTM Site Assessments and Emergency Site Assessments. Notes: Data includes all DTM sites since Nov 2019 excluding Tigray. Tigray is excluded from this analysis because of the more limited coverage and different survey questions asked in the ESAs. Site fixed effects are not included in the regression with host community since the type of settlement typically does not change over time. Site controls include the site age and the primary reason for displacement. Shelters are low quality if over 25% of IDPs are not protected from weather, poor electricity indicates that over 75% do not have access to electricity, and poor water indicates that there are complaints about drinking water. Sample size changes as some variables are not available in some rounds. Sites are weighted by number of IDPs. Standard errors clustered at the site level. * p < 0.10, ** p < 0.05, *** p < 0.01 Relative to camps or settlements, IDP sites in This points to disparities between IDPs and host host communities tend to have higher-quality communities in access to public services. housing and services and a lower preference for integration. Table 12 shows that, after controlling The majority of IDPs prefer return, but for the region, the age of the site and the cause preferences vary across regions. of displacement, host community sites are 34 percentage points more likely to have high-quality Return and integration are the preferred solutions shelters, 21 percentage points more likely to have for IDPs, but preferences vary across regions. The broad electricity access. majority of IDPs—54 percent nationally in 2022— live in sites where return is the preferred solution. Low access to services reflects challenges to Local integration is the second preferred solution. service provision faced by hosting communities On average, 32 percent of IDPs in 2022, were in a and disparities between IDPs and hosting site where the majority preferred local integration as communities. Urban non-hosting areas have better opposed to return or relocation, the least preferred housing conditions in terms of dwelling conditions option. These preferences vary significantly across such as roof, floor, and toilet facilities compared to regions. In Tigray, almost all IDPs are interested in hosting areas – i.e., areas where households in the returning home, while in Somali and Afar over 70 ESPS 2021/22 data lived within 5 kilometers IDP percent of IDPs are in a site where the preferred site location in DTM data). Hosting areas also have solution is local integration (Figure 37). Only in lower access to improved drinking water sources, Amhara, is a considerable number of IDPs (37 in rural and urban areas. Thus, poor indicators on percent) living in sites where most people prefer housing among IDPs partly reflect less than ideal to be relocated. The longer IDPs are displaced, conditions among hosting communities. On the the more likely they are to prefer local integration. other hand, IDP hosting areas have better access to For each additional year of displacement, the electricity from grid connections in rural and urban preference for integration over return or relocation areas, yet these are the most lacking among IDPs. increases by 6 percentage points. However, IDPs 57 PART 2: DEEPENING THE UNDERSTANDING OF DRIVERS OF POVERTY IN ETHIOPIA in host communities are 19 percent less likely to in rural areas. Unlike urban areas, rural IDP prefer local integration (Table 13). hosting areas have lower welfare levels and fewer assets such as agricultural land, livestock, However, there are opportunities presented and farm equipment. Household welfare level in host communities that increase the appeal measured in per capita consumption expenditure for integration. IDP hosting areas have a better- is lower in rural hosting areas compared to educated population compared to non-hosting non-hosting areas. On the other hand, in urban areas. The share of household members with areas, welfare levels are similar between hosting secondary education is higher in hosting areas and non-hosting areas. Households in rural compared to non-hosting areas, both in rural and hosting areas cultivate less than a hectare of urban areas. The share of the population engaged agricultural land on average, which is smaller in wage employment is also higher in hosting areas, than non-hosting areas. In addition, ownership whereas employment in agricultural activities is of livestock in tropical livestock units (TLU) and lower in hosting areas, specifically in urban areas. farm implements is lower for households in rural Also, rural hosting areas have lower shares of hosting areas compared to households in non- people engaged in self-employment. hosting areas. Consistent with monetary and non-monetary welfare indicators, food security, Integration of IDPs into host communities measured by dietary diversity score, is lower in is likely more effective in urban areas than rural hosting areas. Figure 37. Preferred solution by region Table 13. Preference for integration by site type and age 100% Explanatory variable Site attribute 80% Host Years Since Site 60% Community Opened Prefer Integration -0.19*** 0.06*** 40% (0.03) (0.01) 20% Region Fixed Effects Yes No Site Fixed Effects No Yes 0% Site controls Yes No Afar Amhara Ben. Gumuz Gambella Oromia SNNP Sidama Somali Tigray National Round Fixed Effects Yes Yes N 16,634 16,634 Number of Sites 3,434 3,434 Local Integration Relocation Return Source: World Bank staff calculations based on ESS 2011/12, 2013/14, 2015/16, 2018/19. Notes: (a) The preferred durable solution indicates the preference for most IDPs in the site, for the most recent data available in August 2022 (Nov 2022 in Tigray). (b) Data includes all DTM sites since Nov 2019 excluding Tigray. Tigray is excluded from this analysis because of the more limited coverage and different survey questions asked in the ESAs. Site fixed effects are not included in the regression with host community since the type of settlement typically does not change over time. Site controls include the site age and the primary reason for displacement. Shelters are low quality if over 25% are not protected from weather, poor electricity indicates that over 75% do not have access to electricity, and poor water indicates that there are complaints about drinking water. Sample size changes as some variables are not available in some rounds. Sites are weighted by number of IDPs. Standard errors clustered at the site level. * p < 0.10, ** p < 0.05, *** p < 0.01. 58 ETHIOPIA POVERTY AND EQUITY ASSESSMENT STRUCTURAL TRANSFORMATION jobs. Migration is a channel for moving labor from AND HOUSEHOLD WELFARE agriculture into non-agriculture which increases IN ETHIOPIA household consumption while being a catalyst for rural economic transformation, but this too, has been This section examines the changes in the labor inhibited by a combination of prohibitive costs and market in Ethiopia to assess the pace and nature of administrative barriers faced by liquidity constrained structural transformation, whether this contributes migrants in destination areas. to welfare improvements, and the policy constraints to the acceleration of the transformation. The Sectoral and spatial shifts of labor from analysis reveals that movements of labor from agriculture and rural areas have the agriculture to non-agriculture sectors - services in potential to drive poverty reduction particular – are linked to increases in household in Ethiopia. consumption. Such shifts appear to have happened with a decline in agriculture employment and a rise Households shifting to more productive sectors in services employment, but this is overshadowed by experience higher consumption levels and the fact that more people have left the labor market greater growth in welfare compared to those altogether, resulting in a growing share of people who remain in their initial sectors. This is observed not in employment, education, or training, signifying when comparing growth rates in consumption that the economy generated far fewer jobs than among households who stayed within sectors the growing number of people who need a job. The using the Ethiopia Socio-economic Panel Survey stalling job creation – especially wage employment (ESPS) panel data for 2019 and 2022. Despite in the private sector – reflected a stagnation in the similar consumption levels in 2019, households that country’s economic structure as the state-led growth stayed in the same sector experienced a 3 percent model reached its limit which was laid bare by the decrease in consumption per capita by 2022, while recent poly-crisis. Distortions and macroeconomic households who made a progressive transition had imbalances have widened in recent years, especially a consumption growth rate of 9 percent and those in the forex market, hindering job creation. Evidence making a regressive transition a welfare loss of 12 presented shows that addressing these imbalances— percent (Figure 38). Those who reallocated labor aligning the exchange rate for example—will away from agriculture saw the biggest welfare increase household consumption and create more improvements. This pattern is driven by rural areas. Figure 38. Percentage change in household welfare by sectoral transition types 35% 12% 13% 10% 9% 9% 4% 4% 3% 7% -8% -8%-12% -4% -6% -9% -12% -30% to Agriculture to Industry to Services to Agriculture to Industry to Services to Agriculture to Industry to Services Agriculture Industry Services 2014 vs 2016 2019 vs 2022 Source: Authors’ estimates based on ESPS 2013/14; 2015/16; 2018/19; 2021/22. 59 PART 2: DEEPENING THE UNDERSTANDING OF DRIVERS OF POVERTY IN ETHIOPIA Migration is a powerful channel for shifting rural In addition, remittances that migrants send labor from agriculture into non-agriculture sectors, back home contribute to households’ income facilitating rural economic transformation and and become a source of investment and a increasing household consumption (World Bank, coping mechanism during shocks, improving 2022c). Rural-urban migrants are equally as likely resilience and preventing households from to be economically active, employed, and in non- falling back into poverty (World Bank, 2022c). agriculture work as urban residents, but are more On average, remittances from urban migrants likely to be in wage employment (Table 14), though were equivalent to one-third of the receiving a disproportionate share of wage-employed rural- household’s consumption per capita in 2016, urban migrants is engaged in low-quality domestic and more than double that among the poorest wage work (17 percent compared with 5 percent quintiles. Tracking migrants and non-migrants among non-migrant urban residents). In addition to after five years in 18 villages in Ethiopia, de being a pathway to better opportunities, migration Brauw, Mueller, and Woldehanna (2018) also reduces surplus labor in rural areas which raises find positive impacts on real consumption levels labor productivity and output per worker (Table 15). among migrants, demonstrating the net positive This, in turn, raises household consumption and benefits of migration. facilitates economic mobility (World Bank, 2022c). Table 14. Labor market outcomes of migrants Table 15. Impact of migration on factor markets and welfare in origin communities Labor market Rural: Non- Urban: Non- Rural-urban indicator migrant migrant migrant Statistics ATT SE Active 74% 71% 74% Cultivated land (ha per capita) 0.035** 0.001 Unemployed 6% 18% 20% Sector of employment Land rented out 0.01** 0.001 Agriculture 79% 13% 13% Industry 3% 17% 20% Family labor supply 69.4** 0.695 Services 18% 70% 68% (days per capita) Employment type Wage 4% 45% 53% employee Value of crop harvest 1621.5** 107.4 Self-employed 58% 42% 35% (Birr per capita Unpaid family 37% 11% 10% Employer 0% 1% 0% Consumption per capita 3848.1** 36.9 Others 1% 1% 1% Source: Authors’ estimates from the LFS 2021 (Table 15) and ESPS for 2018/19 and 2021/22 (Table 16). Notes: Results in Table 15 are based on a household level regression using a propensity score ‘Nearest Neighbor’ matching estimator. A migrant household has at least one-person aged 10 years and above who moved out of the household except for health and natural disaster displacement reasons. The model includes household demographics, human and social capital, liquidity constraints, drought presence, and market accessibility which are controlled at village level and woreda level, respectively. Outcome variables are; (a) Cultivated land per capita - the area per hectare that the household utilizes for crop production; (b) Land rented out - the share of households renting/sharing out agricultural land; (c) Family labor supply - the total number of days household members spent on the household planting and harvesting activities; (d) Value crop harvest - Ethiopia birr value of the total production; (e) Welfare is measured using spatially adjusted per adult total consumption expenditure. ** results are statistically significant at the 5 percent level. 60 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Many rural Ethiopians have moved out decline was greater among women, whose share of agriculture in recent years. of employment in the agriculture sector declined by 19 percentage points, and among older adults. There has been a shift—driven by women The increasing share of employment in the service in rural areas—away from agricultural sector correspond with the declining trend in employment accompanied by continued agricultural employment in rural areas (Figure 39). expansion of employment in the service sector. Industry’s employment share however declined The share of agricultural employment declined overall, driven by a decline in urban areas from by 16 percentage points between 2005 and 24 percent in 2013 to 17 percent in 2021. The 2021, which was wholly driven by the decline service sector has thus become the primary driver in agricultural employment in rural areas. This of job creation. Figure 39. Sectoral-employment shares By location By gender Age group 2021 2021 2021 Non-Youth Female Urban 2013 2013 2013 2005 2005 2005 2021 2021 2021 Youth Rural Male 2013 2013 2013 2005 2005 2005 0 0.5 1 0 0.5 1 0 0.5 1 Agriculture Industry Services Agriculture Industry Services Agriculture Industry Services Source: Authors’ estimates based on LFS 2005, 2013, 2021. Notes: Data from Tigray were removed from the analysis of LFS 2005 and 2013 as LFS 2021 did not include Tigray. More details on methodology and data can be found in Wieser and Abubakar (2024). Figure 40. Predictive margins for sectoral labor mobility by birth year cohort and location Rural: Urban: .2 .2 Predictive margins Predictive margins .1 .1 0 0 -.1 -.1 -.2 -.2 1949 1959 1969 1979 1989 1998 1949 1959 1969 1979 1989 1998 Birth year cohorts: 1949-1998 Birth year cohorts: 1949-1998 Agriculture Industry Services Agriculture Industry Services Source: Authors’ estimates based on LFS 2005, 2013, 2021. Notes: The intergenerational effects of sectoral employment were estimated using a fixed effects estimator on the cohort-level pseudo panel. Details on the methodology can be found in Box 2 and Wieser and Abubakar (2024). Data from Tigray were removed from the analysis of LFS 2005 and 2013 as LFS 2021 did not include Tigray. More details on methodology and data can be found in Wieser and Abubakar (2024). 61 PART 2: DEEPENING THE UNDERSTANDING OF DRIVERS OF POVERTY IN ETHIOPIA Youth and women are more likely to work generations in the shift of labor towards non- in non-agriculture sectors than the older agricultural sectors. The effects are most visible generation. Evidence suggests that the global among cohorts born after 1963 in rural and urban decline in agricultural employment is influenced areas and accelerated for those born around by the entry of new cohorts into the labor market 1978 in rural areas (Figure 40). Driven by younger (Porzio et al., 2022), with these cohorts exhibiting cohorts aged 15 to 20 years across periods, larger different sectoral preferences than existing ones dips in the share of younger agricultural workers (Hobjin and Schoellman, 2019). Tracking the labor are accompanied by substantial upticks in non- market outcomes using age cohort pseudo-panels agricultural employment, particularly in services. to predict labor market outcomes based on age The sectoral reallocation of labor is thus driven and gender cohort attributes (Box 8), reveals by the entry of new cohorts into non-agricultural a widening margin between older and younger sectors in rural and urban areas. Box 8. Following cohorts over time To better understand the impact of age on sectoral labor mobility decisions—for example, when, over the job life cycle, workers decide to reallocate labor out of or into agriculture—Wieser and Abubakar (2024) evaluate the intergenerational effects of sectoral employment, considering important determinants of sectoral choice using a fixed effects estimator on the cohort-level pseudo panel. A pseudo panel is an approach to create panel data from cross-sectional data by aggregation to follow cohorts over time. We define cohorts by grouping observations by their birth year, gender, and residential status, and then create a panel at the cohort level. Using this technique allows for examining sectoral mobility by tracking changes in the average share of sectoral employment across generations, along with adopting panel data techniques to assess the determinants of sectoral shares across generations. Following Guillerm (2017), the following fixed effects linear model is typically used with panel data: Yit = αi + βXit + εit , with i=1, …, N; t=1, …, T. (1) where Yit measures the dependent variable of person i in time t and α is the individual fixed effect. The term Xit is a vector of time-varying independent variables, and β, the associated vector of parameters to be estimated. Here, εit is the residual term capturing anything else that the model does not consider. In pooled cross-sectional data estimation, ignoring the fixed effect leads to biased estimators of the effect of the independent variables if these variables are correlated with the fixed effect. Because the same individuals are not observed over time as found in panel data, a pseudo panel is constructed from the repeated cross- sections such that cohorts can be observed over time rather than individuals. If c is the cohort at time t, then a pseudo‑panel model that is estimated in practice is as follows: Y*ct = α*c + βX*ct + ε*ct , with c=1, …, N; t=1, …, T. (2) Such that for each dependent or independent variable v, v*ct, = E (vit|i ∈ c, t). Considering that the true values of Y*ct and X*ct are unknown, empirical approximations are based on intra-cohort means: 1 1 Yct nct ∑i∈c,t ∑it and Xct nct ∑i∈c,t Xit (i.e., the means of observed values for the individuals belonging to the cohort). The smaller the size of the cohort, the greater the errors in measuring the empirical means for Y*ct and X*ct. For this reason, around 100 individuals per cohort were considered sufficient to reduce sampling errors (Verbeek & Nijman, 1992, 1993). 62 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Using one-year intervals for the birth year resulted in 264 cohorts across 66 birth-year groups, subdivided equally between men and women and between rural and urban locations over time. Including cohorts that appear in at least two survey rounds, reducing the sample to 50 birth-year groups born between 1949 and 1998 brought the sample down to 200 cohorts. To assess the generational effects of sectoral mobility at the cohort level, the mean level for continuous variables was included in addition to survey year dummies: age and age squared, household size, dependency ratio, and number of other employed adult household members. For non-continuous variables, cohort level shares were used, namely, the share of recent migrants, levels of educational attainment and skills, type of employment, household headship, marital status, and town size categories. Considering that gender and location of residence are time-invariant variables used to create cohorts, these are absorbed by the fixed effects parameter α* in equation (2). In urban areas, however, the demographic skilled industrial sectors, signaling the reallocation structure influences the sectors of work, for of labor from traditional, often low-productivity women in particular. Those belonging to a large rural agriculture to comparable low-skilled roles household with a high dependency ratio have in emerging urban sectors. a lower probability of working in industry and services compared to agriculture. This trend is Recent signs of labor shifting away driven by urban female workers, whose decisions from agriculture belie weaknesses in may be partly influenced by the constraints of Ethiopia’s labor market. balancing caregiving responsibilities with sectoral preferences. Also, the greater sense of communal Recent shifts belie the loss of workers as many living and stronger social support, more prevalent people, particularly women, stopped working. The in rural settings than urban areas, can influence a sectoral shift in employment composition is largely worker’s decision to seek employment outside or explained by the undesirable fact that millions left within agriculture. the labor market altogether in rural areas, which is evident in the rising number of NEET (not employed, Higher-skilled migrants, like other high-skilled in education, or training) in rural areas of 5 million workers in general, tend to work in non- during 2013-21. Moreover, the unemployment rate agriculture sectors. Compared to low-skilled among women and youth increased to 8 percent for workers, medium- and high-skilled workers are each group in rural areas and to 29 percent for both less likely to work in agriculture and industrial young people and women. Though the service sector sectors. There are larger marginal effects among net job growth of 1.7 million in rural areas matched high-skilled rural-to-urban migrants (45 percent) the 1.9 million decline in agriculture employment compared to similarly high-skilled non-migrants during 2013-21, the number of service jobs created (37 percent), with a similar gap observed among during this period was just the same as jobs created medium-skilled rural-urban migrants and non- in the sector during 2005-13 and fell significantly migrants (30 vs 24 percent). Many of the high- below the increase in the out of school, working- skilled rural-to-urban migrants get jobs in high- age population in rural areas. The increase in wage skilled services (Figure 41). On the other hand, employment also does not appear to be significant low-skilled workers who move into urban areas when comparing changes during 2013-21 rather are likelier to work in low-skilled agriculture or low- than the entire 2005-21 period. 63 PART 2: DEEPENING THE UNDERSTANDING OF DRIVERS OF POVERTY IN ETHIOPIA Figure 41. Skill levels and rural-to-urban migration 60% 45.4% 40% 37.0% 30.1% 24.0% 20% 0% -0.2% -5.7% -20% -13.3% -23.8% -24.4% -23.7% -22.0%-23.4% -40% Non-migrants Rural-urban migrants Non-migrants Rural-urban migrants Medium skilled High skilled Agriculture Industry Services Source: Authors’ estimates based on ESPS 2013/14; 2015/16; 2018/19; 2021/22. Notes: To determine whether the decline in low-skilled agricultural employment in rural areas and the corresponding increase in low-skilled urban agriculture is influenced by rural-to-urban migrants, we interacted occupational skill levels with a person’s rural-to-urban migration status for the urban sample Job creation has stalled, lagging the growing transitions between employment, unemployment, number of people needing jobs. Job creation and inactivity in the rural labor market can be decelerated from 34 percent between 2005 estimated by tracking individuals in the ESPS and 2013 to 10 percent between 2013 and 2018/19 and 2021/22 panel surveys to reveal 2021 (Table 16). While there were 3.2 million the correlates of the movement of people out of more jobs in 2021 compared to 2013, this was the labor force. This data corroborates the trend of underwhelming compared to the estimated 1.2 increasing economic activity observed in the LFS million new entrants into the labor market every 2021. The transition matrices confirm that women year and falls short of Ethiopia’s job creation target are more likely than men to become inactive in of 2 million jobs each year (MoF, 2020). The share the labor market. Almost half of unemployed of services in employment only increased because females in 2019 became inactive in 2022, while people left agriculture to become unemployed 30 percent of women employed in agriculture in or economically inactive but not because of 2019 went into inactivity (Figure 42). Estimates accelerated job creation in the service sector. controlling for multiple factors show that older Evidence from the cohort analysis points to the individuals, those with more than secondary new service job expansion being driven by labor education, exposed to recurrent weather shocks market entrants who found work in the services (hence limited work in agriculture), and those sector instead of existing, older, workers moving living in remote locations (measured by low levels from agriculture into services. of market accessibility) are more likely to become inactive. This broadly signifies that rising inactivity The lack of economic opportunities drove the is driven by the lack of suitable opportunities in rise in economic inactivity. The probabilities of rural areas. 64 Table 16. Employment and job growth of subsectors Employment Jobs Growth Number Share (%) 2013-2005 2021-2013 2005 2013 2021 2005 2013 2021 Number Share (%) Number Share (%) Total 23,933,681 31,984,367 35,209,845 100 100 100 8,050,686 100 3,225,478 100 Total Agriculture 18,747,856 22,676,005 22,283,795 78.4 70.9 63.3 3,928,149 48.8 -392,210 -12.2 Crops 2,017,929 5,079,065 4,731,206 8.4 15.9 13.4 3,061,136 38.0 -347,859 -10.8 Animals 607,989 1,782,599 3,414,581 2.5 5.6 9.7 1,174,610 14.6 1,631,982 50.6 Mixed farming 16,020,623 14,553,625 12,274,257 67 45.5 34.9 -1,466,998 -18.2 -2,279,368 -70.7 Other (e.g., hunting/gathering/forestry) 33,610 111,571 84,817 0.1 0.3 0.2 77,961 1.0 -26,754 -0.8 Agricultural support 67,705 1,149,146 1,778,935 0.3 3.6 5.1 1,081,441 13.4 629,789 19.5 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Total Industry 1,687,525 2,642,747 2,285,953 7.1 8.3 6.5 955,222 11.9 -356,794 -11.1 Food processing 560,978 259,535 365,942 2.3 0.8 1 -301,443 -3.7 106,407 3.3 Non-food manufacturing 694,281 1,377,231 780,022 2.9 4.3 2.2 682,950 8.5 -597,209 -18.5 Mining and extractives 59,387 116,274 184,562 0.2 0.4 0.5 56,887 0.7 68,288 2.1 Construction 334,774 717,795 726,421 1.4 2.2 2.1 383,021 4.8 8,626 0.3 65 Public Utilities 38,104 171,911 229,006 0.2 0.5 0.7 133,807 1.7 57,095 1.8 Total Services 3,483,283 6,665,615 10,640,096 14.6 20.8 30.2 3,182,332 39.5 3,974,481 123.2 Food trade 462,430 1,280,495 1,618,960 1.9 4 4.6 818,065 10.2 338,465 10.5 Non-food trade 784,843 775,105 703,063 3.3 2.4 2.0 -9,738 -0.1 -72,042 -2.2 Transport 113,825 318,772 547,244 0.5 1 1.6 204,947 2.5 228,472 7.1 Hospitality - food 612,598 375,054 336,956 2.6 1.2 1.0 -237,544 -3.0 -38,098 -1.2 Hospitality - accommodation 38,394 24,999 13,142 0.2 0.1 0.0 -13,395 -0.2 -11,857 -0.4 Information & communication 44,189 56,010 90,536 0.2 0.2 0.3 11,821 0.1 34,526 1.1 Finance & real estate 40,055 125,046 276,673 0.2 0.4 0.8 84,991 1.1 151,627 4.7 Professional/technical 28,799 128,822 180,131 0.1 0.4 0.5 100,023 1.2 51,309 1.6 Administrative & support 17,474 131,822 264,196 0.1 0.4 0.8 114,348 1.4 132,374 4.1 Public administration & defense 273,890 266,961 363,312 1.1 0.8 1.0 -6,929 -0.1 96,351 3.0 Education & health 325,238 837,187 1,243,855 1.4 2.6 3.5 511,949 6.4 406,668 12.6 Arts, entertainment & recreation 10,450 36,785 39,926 0.0 0.1 0.1 26,335 0.3 3,141 0.1 Community & family-oriented 681,914 2,288,843 4,891,159 2.9 7.2 13.9 1,606,929 20.0 2,602,316 80.7 Extraterritorial organizations 49,185 19,713 70,945 0.2 0.1 0.2 -29,472 -0.4 51,232 1.6 Source: Authors’ estimates based on LFS 2005, 2013, 2021. Notes: Analysis excludes Tigray in LFS 2005; 2013 for compatibility with LFS 202. See Wieser and Abubakar (2024) for more details. PART 2: DEEPENING THE UNDERSTANDING OF DRIVERS OF POVERTY IN ETHIOPIA Figure 42. Labor market transitions among rural population aged 18-64 Males (share of population) Female (share of population) Inactive Inactive Unemployed Unemployed Non-agriculture Non-agriculture Agriculture Agriculture Agriculture Non-Agriculture Agriculture Non-Agriculture Unemployed Inactive Unemployed Inactive Source: Authors’ estimates from the ESS 2018/19 and 2021/22. The quality of wage jobs in Ethiopia has been and 1.4 during the past decade, reflecting a generally consistently low and has deteriorated further in low quality of jobs (Figure 43). The lowest job quality recent years. Following an approach discussed by was registered in 2021, signifying a drop from 2018. Hovhannisyan et al. (2022), the quality of jobs in Service sector jobs, which offer better job stability, Ethiopia is assessed for wage-employed workers working conditions, and wages than agriculture and in urban areas (where wage employment is industry, are consistently higher than the overall concentrated) based on three dimensions: (i) income average in Ethiopia, with better jobs found in finance, (wage earnings), (ii) job stability (employment by a education, health, and other professional subsectors. formal entity and having a permanent contract), and However, the job quality in the service sector and (iii) working conditions (excessive work hours and the industry sector, has been on a downward whether people would like to work more, occupational trajectory since 2013. The job quality in agriculture, safety and skills match). These are combined to which has been on the rise since the beginning of generate a job quality index with a minimum and the past decade, also experienced a sharp drop in maximum value of 0 and 3, respectively. The index for 2021. The overall job quality of urban wage jobs thus urban wage jobs in Ethiopia has ranged between 1.2 deteriorated across all sectors in 2021. Figure 43. Trends in wage job quality in urban areas Sectoral comparison Gender comparison 1.9 1.7 1.6 1.6 1.5 1.5 1.5 1.5 1.4 1.5 1.4 1.4 JQI JQI 1.3 1.3 0.9 1.2 1.2 1.2 1.2 1.2 1.1 1.1 1.2 1.1 0.4 1 2005 2010 2011 2013 2014 2015 2016 2018 2020 2021 2010 2011 2013 2014 2015 2016 2018 2020 Total Agriculture Industry Services Men Women Source: Authors’ estimates based on UEUS 2010-2020. Notes: The job quality index (JQI) for urban Ethiopia describes the quality of jobs in seven indicators across three dimensions and sums the number of successes across dimensions with equal weighting and ranges from 0 to 3. It yields a score of 0 if a wage worker has a job that does not meet the criteria for quality in all dimensions and a maximum score of 3 if a wage worker’s job meets the requirements for quality across all three dimensions. For more details on methodology, see Wieser and Abubakar (2024). 66 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Job quality disparities persist by gender, age, and rural areas, this is only a 2 percentage point and 0.5 level of education. Adopting an approach inspired percentage point increase in absolute shares), while by Brummund et al. (2018) and Hovhannisyan et al. in urban areas, there was a 14 percent decrease (2022)¹ to better understand the characteristics that in medium-skilled jobs and a 6 percent increase determine the quality of jobs shows that women, in high-skilled jobs. In addition to human capital youth, and the less educated are less likely to secure development and occupational shift patterns, these higher-quality jobs. In contrast, those in the public differences may be explained by labor mobility sector with advanced skills work in better-quality between rural and urban areas. jobs. Gender differences in the quality of jobs are largely explained by sector. Working women Stalling non-agriculture job creation increasingly move from elementary occupations reflects a faltering transformation in into professional, technical, clerical, and skilled Ethiopia’s economic structure as the agriculture roles. At the same time, men account state-led growth model reached its limit for increases in the share of machinery operators in the industrial sector. The economy’s structure has remained broadly unchanged since 2018. Having initially declined by Jobs in rural Ethiopia are characterized by slightly more than 11 percentage points during 2010- low-skilled employment. Agricultural activities 18, agriculture’s share in GDP has since declined by are mainly described by subsistence farming. less than 2 percentage points while both the industry Consequently, 85 percent of rural employment in and service sectors’ shares increased by about 1 2021 was low-skilled, while in urban Ethiopia, only percentage point (Figure 44a). Consistent with this one-third of employment is low-skilled. There has lack of a further shift in the economic structure, net been some evidence of progress in skill upgrades job creation in non-agriculture sectors declined. in the Ethiopian workforce, albeit with noticeable The industry sector lost jobs, owing to the non- differences between urban and rural areas. Between food manufacturing subsector shedding close to 2013 and 2021, the share of medium-skilled 600,000 jobs during 2013-21, after having created workers in rural areas grew by 17 percent and that almost 700,000 jobs in the preceding 8-year period. of high-skilled workers by 33 percent (though given The contribution of between sector shifts to labor the low share of medium and high-skilled workers in productivity also declined (Figure 44b). Figure 44. Changes in economic structure and labor productivity Panel A. Change in sectoral GDP Panel B. Change in labor productivity 43.7 45.2 45.9 45.5 46.6 47.0 39.7 38.8 39.2 40.0 40.0 39.6 40.0 2016-20 10.2 10.4 11.5 13.0 13.8 15.0 23.7 25.9 27.0 28.1 28.1 29.3 28.9 2011-16 46.1 44.4 43.1 42.0 40.2 38.7 37.5 36.4 34.9 33.3 32.7 32.5 32.4 2006-11 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 -200 0 200 400 600 800 1,000 Within agriculture Within industry Agriculture Industry Services Within services Between sectors Source: Authors’ estimates based on NBE annual reports; World Bank, 2024. ¹ The pooled cross-sectional regression model includes an individual’s job quality about gender, age, educational attainment, occupational skill level, public or private sector employment, and employment subsector. Additionally, regional and year dummies are included to account for unobserved effects related to time and location. The specification and regression results are presented in Annex 4. 67 PART 2: DEEPENING THE UNDERSTANDING OF DRIVERS OF POVERTY IN ETHIOPIA The limited change in Ethiopia’s economic rationing that hindered economic activity, structure is a manifestation of the limitations particularly in the private sector. The black- of Ethiopia’s state-led growth model, exposed market premium increased from around 20-30 during multiple crises in recent years (World percent during the second half of the last decade Bank, 2024). Ethiopia relied on a state led growth to more than 100 percent now. Inflation has risen in model focused on improving intermediate inputs tandem with the parallel exchange rate (Figure 45a). (roads, energy infrastructure, industrial parks, Though the official exchange rate was devalued agricultural inputs) to stimulate domestic and significantly in 2021, the real effective exchange foreign private investment. This was financed rate has been appreciating which undermined the through domestic financial repression and competitiveness of exports (Figure 45b). exchange rate controls, enabling the state to commandeer domestic and external financial Widening imbalances, in the forex market, for resources that were mostly channeled to State- example, inhibited job creation, especially in the Owned Enterprises (SOEs). This created an manufacturing sector, which started losing jobs. enormous cost by limiting the accumulation of Analysis of data from the Large- and Medium-sized private surpluses, crowding out private investment, Manufacturing Industries Survey (LMMIS) shows undermining external competitiveness (World that manufacturing employment started to decline Bank, 2022a) and suppressing market incomes in 2018 (Figure 46), corroborating the trend including for farmers (World Bank, 2022b). observed in the labor force surveys. Statistical Declining exports and reduced capital inflows models show that a Real Effective Exchange contributed to mounting external imbalances which Rate (REER) depreciation would have created coupled with declining agricultural productivity more jobs and increased wages and productivity and the impact of persistent droughts and locust in proportionate to manufacturing firms’ export infestations, led to a recent slowdown in Ethiopia’s exposure without a significant impact in relation to growth rates (World Bank, 2024). the firms’ import exposure (Figure 47). The rents from an overvalued exchange rate to a degree, Foreign exchange market distortions and improved employment proportionately to a firm’s shortages aggravated these external challenges, import exposure, but at only slightly more than a resulting in widespread foreign exchange fifth of the impact of the REER appreciation. Figure 45. Trends in exchange rates and inflation in Ethiopia Panel A: Official and Parallel Exchange Rates and CPI Panel B: Real and Nominal Effective Exchange Rates Exchange rate of Birr to US$ 430 120 200 380 100 330 150 CPI index 280 80 2007=100 230 60 100 180 40 130 50 80 20 Aug-13 Sep-14 Oct-15 Nov-16 Dec-17 Jan-19 Feb-20 Mar-21 Apr-22 May-23 0 2000 2005 2010 2015 2020 General CPI Official exchange rate Nominal Effective Exchange Rate Parallel exchange rate Real Effective Exchange Rate Source: NBE reports, Ethiopia Statistical Service (ESS) and Darvas (2021). 68 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Figure 46. Number of workers in large and Figure 47. Relationship between exchange rates medium scale manufacturing firms in Ethiopia and employment by manufacturing firms 300 .04 Change in employment Number of workers .03 (in thousands) 200 .02 100 .01 -.2 -.1 0 .1 .2 0 ∆REER 1995 2000 2005 2010 2015 2020 Import share<=0.25 Total Employment Production workers Import share>0.25 & Import share<0.5 Administrative Import share>0.5 & Import share<0.75 Source: Authors’ estimates based on LMMIS. Notes: Predicted values of changes in employment based on a regression model following Nucci and Pozzolo (2010), using the LMMIS firm level data. The regression model includes the REER, the black-market premium, firm’s export exposure (share of exports in revenue) and import exposure (share of imports in input costs), interactions between the import/export exposure variables with the REER, as the main explanatory variables. Other firm characteristics and sector fixed effects are also included. While addressing macroeconomic of edible oil and wheat until 2021, for example, imbalances has short-term costs, it appeared to track the official rate more closely than will generate better-quality jobs and the parallel market, while the prices of maize and higher incomes. beef, whose distribution is not controlled, appear to track the parallel market exchange rate more Reforms to the forex regime in Ethiopia have been closely (Figure 48). obstructed by concerns that an exchange rate realignment would raise inflation. Analysis from Delaying forex reforms means avoiding short-term a CGE modeling accounting for the dual exchange adverse impacts but also losing out on medium- rate and preferential access to forex for some term benefits. Exchange alignment is projected preferred commodities provides tentative evidence to increase economic growth by 2 percent above supporting this notion (see Box 9). Compared to the the BAU scenario by 2030. Consumption of rural business-as-usual (BAU) scenario, the CPI would households would be 1.2 percent higher, driven be 5.4 percent higher in 2024 and 6.7 percent by by rising agriculture exports, and that of urban 2030. This can be attributed to the structure of households by 3.2 percent, as they are the primary imports in Ethiopia. Unlike other countries which recipients of remittances and rely more on the service had large forex distortions such as Zimbabwe, sector which will grow faster. Net exports would be not all key commodities in Ethiopia track the higher as increased exports offset the impact of parallel market rates because commodities on the higher import input costs (Figure 49). However, some preferred list in Ethiopia account for more than manufacturing sectors that rely heavily on imported half of imports. These have usually been imported intermediate inputs, like the textile and garment through SOEs (e.g., wheat, fuel, and edible oil) with sectors, would become less competitive. In the short lower prices partly passed to consumers using term, agricultural output would decline but would a combination of controls over the distribution pick up in the medium term while manufacturing channels and retail prices or subsidies. The price contracts throughout (Figure 50). 69 PART 2: DEEPENING THE UNDERSTANDING OF DRIVERS OF POVERTY IN ETHIOPIA Figure 48. Correlation between commodity prices and parallel and official exchange rate in Ethiopia Addis Ababa Addis Ababa Prices (Birr) Exchange rate Prices (Birr) Exchange rate 600 100 60 120 80 100 400 40 80 60 60 200 40 20 40 20 20 0 0 0 0 Dec-16 Jun-17 Dec-17 Jun-18 Dec-18 Jun-19 Dec-19 Jun-20 Dec-20 Jun-21 Dec-21 Jun-22 Dec-22 Jun-23 Dec-16 Jul-17 Feb-18 Sep-18 Apr-19 Nov-19 Jun-20 Jan-21 Aug-21 Mar-22 Oct-22 May-23 Injera Maize Wheat (traditional) Beef Edible oil local Coffee beans Bus fare (within town) Official exchange rate Parallel exchange rate Official exchange rate Parallel exchange rate Source: Authors’ estimates from ESS and NBE publications. Figure 49. Comparison of the population distribution by consumption levels Macro-economic impact of exchange rate devaluation Sectoral breakdown of change in export (% change from BAU) 25 Net exports 15 5 Investment -5 Private investment -15 -25 Public investment -35 Maize Pulses Oil seeds Vegetables Roots Fruits Coffee & khat Flower & oth. cash crops Livestock Meat & meat products Grain mill products Bakery Beverage & tobacco Textile Leather Cement & mineral prod. Electricity TnD Electricity Transport Other services Rural households’ consumption Urban households’ consumption Urban household’s consumption GDP at market prices 0.00 0.50 1.00 1.50 2.00 2030 2024 2024 2030 Source: Authors’ estimates based on HCES 2015/16 and HoWStat 2021. Notes: Estimates are based on analysis using the Mitigation, Adaptation and New Technologies Applied General Equilibrium (MANAGE) model with 82 sectors, 85 products, 2 labor categories (skilled, semi and low-skilled), 8 tax accounts (commodity and direct taxes), 4 economic agents (households, enterprises, government, and rest of the world), 2 household categories (rural and urban) and 2 investment accounts (private and government). The model applies a uniform, exogenously calculated exchange rate premium, added to the official exchange rate to obtain the parallel market rate applied to products with restricted access to forex. The 70 percent surrender requirement is considered by distributing exporters’ exports earnings for only 30 percent of exports, while the remaining is received by agents in the rest of the world and importers with priority access to forex (see Box 9 for additional details). 70 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Figure 50. Sectoral breakdown of change in output and employment Sectoral breakdown of change in output (% change from BAU) Sectoral breakdown of change in employment (% change from BAU) 5.7 4.5 3. 0 3.2 1.1 0.9 0.7 0.7 0.7 0.7 0.2 1.7 1.2 1. 0 -0.1 -0.1 -0.7 -0.5 All sectors Agriculture Manufacturing Services Skilled Semi-low skilled -0.9 -1.2 Agriculture Manufacturing Services Total 2024 2025 2024 2030 Source: Authors’ estimates based on HCES 2015/16 and HoWStat 2021. Notes: Estimates are based on analyzed using the MANAGE CGE model as described in Figure 48 above and Box 9 . Exchange rate alignment would catalyze services sector - where the job quality is higher sectoral shifts in employment towards higher – with a job growth of 3 percent compared quality jobs mirroring shifts in the economic to the BAU scenario, while labor moves away structure. Compared to the BAU scenario, output from manufacturing (0.7 percent decline) and would be 5.7 and 1.2 percent higher in the service agriculture (0.1 percent decline). More skilled and agriculture sectors, respectively, and lower jobs are created compared to low and semi- in the industry sector in 2030. The overall net skilled ones (Figure 50). In a nutshell, addressing employment effect is small but positive (0.5 the exchange rate imbalance would promote percent higher by 2030). But this is a result of structural transformation in the labor market and a stronger sectoral shift of labor towards the increase households’ incomes. Box 9. Economic and welfare impacts of exchange rate unification in Ethiopia The analysis uses the MANAGE-WB model (Mitigation Adaptation and New Technologies Applied General Equilibrium model of the World Bank) a recursive dynamic single-country computable general equilibrium (CGE) model. MANAGE-WB has been extended to incorporate two exchange rates (official and parallel) and the capital control mechanisms in place in Ethiopia. Demand for foreign exchange is only partially satisfied by supply at the official exchange rate. Excess demand is met via a parallel market for foreign exchange. The spread between the two rates is introduced in the model and calibrated to reflect its evolution between 2018 and 2023 and maintained at 2023 level for the rest of the period. Foreign exchange surrender requirements and rationing with preferential access for selected products are also introduced in the MANAGE-WB model. Goods and services have been classified in the model into two categories according to their priority in accessing foreign exchange at the official rate. Priority items are subsidized by an implicit tax/tariff on non-preferential imports and the surrender requirements. The 70 percent surrender rule on foreign exchange revenue from exports and inward remittances implies that only 30 percent of this inflow is received at the parallel rate. The rents generated by the premium between the official 71 PART 2: DEEPENING THE UNDERSTANDING OF DRIVERS OF POVERTY IN ETHIOPIA and parallel exchange rates are, in part, received by imports with preferential access to foreign exchange. A significant share of the rents is distributed to the rest-of-the-world thereby capturing capital flight resulting from the avoidance of capital account restrictions through under- and over invoicing of trade transactions. The baseline scenario assumes that the dual exchange rate system will continue with a progressive depreciation of the official rate. The counterfactual simulates a unification via a depreciation of the official rate and appreciation of the parallel rate. The unification is simulated via a 52.6 percent depreciation of the official rate and an appreciation of the parallel rate by 20 percent. The rationale is that initially, the official rate would depreciate to the level of the parallel rate. It is expected that the unified rate would then appreciate to settle at a level 20 percent lower that the parallel rate. The MANAGE model for Ethiopia is calibrated to annual data and therefore assumes that the new ‘equilibrium’ rate would be reached during the first year of the unification, that is in 2024. The transition and associated out of equilibrium effects are not accounted for. The unification is accompanied by the removal of capital controls. The FX surrender requirements, FX rationing, and preferential access to FX for selected products are lifted and associated capital flight is eliminated. This implies that a parallel market to evade capital controls would not persist. A prudent macroeconomic framework is implemented by imposing model closures that avoid the deterioration of the current account and fiscal balances with the unification. Accordingly, we assume that government expenditure is adjusted to avoid a larger fiscal deficit. Hence, we maintain the deficit-to-GDP ratio at the baseline level and allow government consumption to adjust. Government investment share of GDP is kept at the baseline level while tax rates are exogenous. With this approach, we can capture the net effect on government debt denominated in foreign currency. Similarly, the CAB-to-GDP ratio is fixed at the baseline level. This implies that exports will need to expand to finance imports. Investment is savings driven. Spatial transformation has also been (Figure 51) show that households with access to hampered by barriers - some structural a larger credit amount and those that received - to the movement of labor. cash transfers, were respectively, more than 15 and 7.5 percentage points more likely to send Rural-urban migration—a channel for moving out a migrant. The migration decision is closely labor into non-agriculture sectors—has also not correlated with whether a household is already happened fast enough to offset a growing rural receiving remittances either from a rural, or population. The rural population increased by 24.4 urban area or from abroad, which is also a sign percent between 2013 and 2021, notwithstanding for meaningful social capital, since the household the increase in rural-to-urban migration of 20.8 relates to the outside world other than just in percent. The pattern of rural-to-urban and intra- the home place. Estimates comparing similar rural migration shares to total internal migration households in 2011/12 in terms of their likelihood has remained similar (World Bank, 2022c). of having a migrant in the second visit (2013/14) or Moreover, cross-regional mobility has decelerated the third visit (2015/16) show that the probability in recent years compared with the trend observed of having at least one person from a rural household before 2013. Thus, migration largely remains an migrate in the subsequent period increases by 7.5 intra-regional phenomenon. percentage points if a household already had a migrant. Lastly, having poorly educated members Low financial, human, and social capital are a or a smaller number of working-age members also constraint to migration. Multivariate estimates reduces the possibility that any of them will migrate. 72 ETHIOPIA POVERTY AND EQUITY ASSESSMENT This shows that lack of capital – financial, human, Figure 51. Marginal effects correlated with migration and social capital – are important constraints for migration in Ethiopia. Sectoral breakdown of change in output (% change from BAU) The prohibitive costs of migration imposed by challenges in destination areas increase the Age of HH head barriers to migration for liquidity-constrained HH headed by female households. Evidence shows that the returns to HH size: Age 0-7 migration could be high, but the migration process is fraught with difficulties. Qualitative studies HH size: Age 7-17 (Bundervoet, 2018) suggest that migrants find HH size: Age 18-65 that the job search process is more difficult than HH average years of education anticipated and that they face administrative HH access to credit barriers in obtaining kebele IDs, inhibiting their HH receive cash transfer access to government services. Many typically find the transition to urban life onerous, and females HH own livestock face additional challenges. These factors make Drought index (SPEI) integration into destination areas more difficult -10 0 10 20 and costly for migrants in the immediate term. Probability of being with They create a barrier to migration for those from migrant (percentage points) liquidity-constrained rural households, who also lack social networks. Source: Adopted from World Bank (2022c). 73 PART 3 Turning Tides for Poverty Reduction The previous chapters discussed the recent socio-economic developments, demonstrated how shocks and existing structural weaknesses have contributed to the increase in poverty since 2016 and highlighted some of the key challenges for poverty reduction. This section discusses what needs to be done to accelerate poverty reduction by improving the productive capacity of poor households to become active contributors to growth while strengthening their resilience to shocks. 74 ETHIOPIA POVERTY AND EQUITY ASSESSMENT SUMMARY OF KEY CHALLENGES Shocks laid bare underlying structural FOR POVERTY REDUCTION weaknesses in the country’s growth model, which amplified the increase in poverty. First, Poverty increased due to a the state-led rural development policy focused combination of shocks and on increasing food security because of the longstanding vulnerabilities in the country’s history of food insecurity. Though yields development model. improved, farmers’ incentives and input markets were distorted in favor of staple food production, Past gains in poverty reduction were undone hampering farmers’ market orientation and during the second half of the last decade. generation of marketable surpluses. Therefore, The poverty rate is estimated to have increased only a few households could benefit from rising by between 4 and 10 percentage points during prices, and they were outnumbered by net food 2016-2021. This increase in poverty was broad, buyers. Second, structural transformation in and experienced in all regions except for Ethiopia’s Ethiopia stalled as the mounting distortions two major urban regions – Addis Ababa and Dire created by the state-led growth model increased Dawa. Poverty increased because of a general macro-economic imbalances that undermined decline in consumption across the entire socio- competitiveness and crowded out private sector economic spectrum. On average, consumption development. Job creation during 2013-21 cumulatively declined by 13 percent during 2016- declined by 24 percentage points compared to 21. The decline was deeper among better-off the preceding period, unemployment increased, households and in urban areas, but poverty still and labor force participation declined, leaving 5 increased more in rural areas because many non- million working-age Ethiopians not in employment, poor rural households were just above the poverty education, or training. The rising share of people line to begin with. Inequality declined because of out of work reduced consumption, while stalled the greater decline in consumption among the non- sectoral transition posed forgone opportunities for poor people making everyone poorer. income growth. Lastly, barriers to migration have limited labor mobility from agriculture to non- The country experienced multiple shocks, which agriculture sectors, and in so doing, stalled the were the immediate cause of the decline in economic transformation of rural areas. welfare and rising poverty. Though localized, the combination of drought shocks that was prolonged Interaction of shock and vulnerability in low land areas, conflict that intensified in the left scars that make the continuation of Northern Tigray, along with locust invasions and the current model untenable. flooding across the country, meant that more than 90 percent of the population experienced a The experience of shocks has long lasting effects covariate shock of one form or another. At least 48 that are amplified by the country’s structural percent experienced more than one such shock. weaknesses. Employment in the industry sector Additionally, there were economy wide shocks from which had already peaked before COVID-19 the impacts of the COVID-19 pandemic and rising struck, has not recovered from the effects of the inflation. Estimates suggest that rising inflation pandemic that saw women bear a greater burden reduced household consumption by 21 percent, of job losses, pushed labor to more vulnerable drought exposure by 9 percent and exposure to forms of employment and depressed earnings of conflict in Northern Ethiopia by up to 17.5 percent household enterprises. The overvaluation of the among the most exposed. exchange rate, financial repression and the large 75 PART 3: TURNING TIDES FOR POVERTY REDUCTION SOE footprint limit prospects for recovery from the POLICY IMPLICATIONS pandemic. Meanwhile, a combination of market volatility and climate shocks increase households’ In the current context, the key priorities incentive towards self-sufficiency, where public for poverty reduction are (i) strengthening extension services are already biased towards households’ and the economy’s resilience to and limit the adoption of high-risk high reward shocks, (ii) increasing the generation of agriculture agriculture technologies whose development and surplus and (iii) addressing spatial and economic availability are constrained by government controls. policy driven structural impediments to job creation and access to better economic opportunities. The rural-urban gaps persist due to untapped potential in the rural sector. Enhance resilience to shocks. The rural-urban divide in welfare continued to The high vulnerability of households to shocks persist. Poverty increased in rural areas during and the impacts this had on poverty necessitates 2015/16 - 21 by up to 10 percentage points strengthening households’ resilience to shocks in compared to less than 3 percentage points in urban three ways: areas. The incidence of poverty in rural areas is now double the poverty rate in urban areas. This i. Slowing down the onset or impact of shocks gap primarily reflects within regions disparities at entry – Household’s vulnerability to shocks between rural areas and the urban centers whose is in part driven by their limited ability to cope poverty rates are higher but close to Addis Ababa. or minimize the direct impact of these shocks Rural areas accounted for 88 percent of poor on their incomes. Addressing this requires three people in the country, though their share had types of interventions. One set of interventions declined to 78 percent by 2021. is investments to increase the productive assets of households and communities, which range Poverty remains a predominantly rural from infrastructure investment in irrigation and phenomenon because the huge agricultural land structures, natural resource management, potential is untapped. Most of the poor in Ethiopia and skills development to increase households’ (68 percent) reside in moisture reliable areas, adaptability. The other set of interventions is while pastoral areas that account for another developing and promoting the adoption of climate- 7 percent of the poor host a lion’s share of the smart agriculture technologies and strategies. The country’s livestock. Yet poverty increased more third set of interventions focuses on prevention in rural areas because of the reduction in real and preparedness, which includes enhancing early crop incomes despite the rising food prices which warning systems for households to take adaptive should have benefited farmers if they produced measures to minimize the impact of shocks. a marketable surplus as happened in countries like Cambodia during the 2008 – 2010 food ii. Reducing the impact of shocks on incomes price increases. Furthermore, because of high once they occur – The impact of shocks on dependency ratios and large household sizes, most households is exacerbated by limited access to gains in agriculture production have been absorbed or suboptimal response options to shocks. The by the growing population. coverage of social safety nets for example is 76 ETHIOPIA POVERTY AND EQUITY ASSESSMENT limited in Ethiopia, though this has been shown Ethiopia Rural Income Diagnostics (World Bank, to be an effective instrument for mitigating the 2022b). They include: impacts of shocks. The high livestock mortality due to droughts that have been observed in pastoral iv. Reducing market distortions to trigger a areas (World Bank, 2023) is another example supply response – Surplus generation has to some of how the limited options for households to extent been limited by a weak supply response, respond or pre-empty the shocks lead to worse given Ethiopia’s distortionary policies that until outcomes. Therefore, there is a need to establish/ recently, suppressed domestic prices below expand mechanisms to finance crisis response international prices for some crops like maize, e.g., destocking and school feeding programs in with Ethiopia emerging as one of the countries response to droughts; expanding the coverage and whose nominal protection rates for maize suggest range of consumption smoothing measures such its policies offer a price disincentive to farmers. as shock responsive social safety nets and access The government can increase the efficiency of to credit; and establishing market mechanisms to market mechanisms and trigger a supply response moderate volatility e.g., warehousing receipts. by eliminating marketing controls that blunt price signals to farmers such as export controls and iii. Facilitating faster and full recovery from marking restrictions for commodities. shocks – Households affected by shocks, such as the massive loss of livestock due to drought, v. Increasing availability and adoption of or displacement and destruction of assets due to advanced agriculture inputs and technologies conflict, need support to restore their livelihoods – Another driver of low surplus generation is low and recover from crisis. This can be done through productivity growth for crops such as teff, coffee, investments for livelihood restoration and and beans – in part due to distortions in input reconstruction which applies to both climate and markets which biased agriculture technology conflict shocks e.g., infrastructure rehabilitation development towards certain crops (e.g., maize) and re-stocking and input support programs; and than others and led to suboptimal availability of promoting the adoption of insurance products (e.g., inputs as discussed in the Ethiopia Rural Income livestock insurance). Diagnostics (World Bank, 2022b). Adoption of some agriculture technologies – such as improved Enhance generation of agriculture surplus seed varieties – is low, as statistics presented in Chapter 4 suggests. Liberating input markets Poor people in high agriculture potential areas to promote a greater role of the private sector have not been able to capitalize on rising food in agriculture technology development, input prices due to limited market surplus generation production, and distribution is needed to increase and limited market participation. This points to the timely availability of the right type of inputs. the necessity of interventions to increase market surplus generation among rural households vi. Optimizing crop cultivation choices and and promote agriculture commercialization, incentivizing production of commercial crops thus re-orienting Ethiopia’s rural development – Furthermore, surplus generation can be and agriculture policies from a food security maximized by optimizing crop choices to the land focus towards a more transformation agenda as suitability. Analysis presented in the Rural Income envisaged under the next phase of the Agriculture Diagnostics (World Bank, 2022b) show a crop bias and Rural Development Policies. Most of these in extension services messaging, while exposure to measures have been discussed in detail in the price and weather shocks and limited connectives 77 PART 3: TURNING TIDES FOR POVERTY REDUCTION disincentive commercial orientation of households. of exchange rate misalignment analyzed in this The adoption of a plurality of agriculture extension report shows. Eliminating macro policy distortions services and shifting in messaging to encourage a that undermine private investment is therefore shift towards commercial crops and optimize crop necessary for poverty reduction in Ethiopia. Key cultivation choices to land suitability will therefore among the macro-economic distortions – as be crucial for increasing commercial orientation identified in more detailed diagnostics such as and surplus generation by households. Other sets the Country Economic Memorandum (World Bank, of measures include those mitigating the impact 2022a) and the Systematic Country Diagnostics of climate shocks discussed under the priority (World Bank, 2024) - is addressing the exchange intervention to increase resilience to shocks, as rate misalignment, liberalizing interest rates, and these can also influence household agriculture reducing state dominance in the financial sector production decisions based on their impact on risk to direct more lending towards the private sector. preferences (World Bank, 2022b). viii. Reducing barriers to entry and state Eliminate structural impediments to dominance in the economy – The stalling job job creation. creation in Ethiopia in part reflects stalling private sector wage job creation during 2013-21. Beyond A more fundamental challenge for poverty the macro-economic distortions, private sector reduction is the lack of better economic growth is limited by restrictions to entry and a opportunities. This is evidenced by the declining playing field stacked in favor of state-owned pace of job creation, with a net reduction of jobs enterprises which was the core part of the state- in the industry sector, the exit from the labor led growth model (World Bank 2024). As noted in market by people – women in particular – facing other studies, private sector wage job creation can limited opportunities in the context of stalling be stimulated by promoting market neutrality and structural transformation and the general decline reducing foreign entry restrictions in markets with in the quality of jobs as trends in the job quality high potential for reorganizing agriculture value index showed. Analysis in the report showed that chains (e.g., permitting foreign entry into wholesale eliminating macroeconomic distortions such as and retail markets). the exchange rate misalignment, can promote structural transformation and household income ix. Reducing barriers to labor mobility – Evidence growth. Other studies identify state intervention in presented earlier suggests that rural-urban product and financial markets as key constraints to migration is a critical channel for rural workers to the creation of a vibrant private sector that creates access non-farm economic opportunities, while better quality jobs. This calls attention to the need at the same time, catalyzing rural transformation. for reforms that promote job creation in an economy However, this channel is limited in Ethiopia that is competitive and private sector led, while by the high cost of migration in part driven by enhancing women’s economic empowerment. The administrative barriers that migrants face in their necessary reforms have been discussed in detail destinations and job search costs. Eliminating in the Systematic Country Diagnosis for Ethiopia burdensome administrative procedures for (World Bank, 2024). They include the following: migrants (e.g., household registration requirements) and reducing job search costs by enhancing job vii. Macro-fiscal stabilization – Macro-economic intermediation and employment promotion services distortions undermine competitiveness of the will be critical for linking rural workers to better private sector and job creation as the example economic opportunities in urban areas. 78 ETHIOPIA POVERTY AND EQUITY ASSESSMENT REFERENCES Abay, K. A., Abay, M. H., Berhane, G., & Chamberlin, J. (2022). Social protection and resilience: The case of the productive safety net program in Ethiopia. Food Policy, 112 (August), 102367. https://doi. org/10.1016/j.foodpol.2022.102367 Ahmed, M. H., Tesfaye, W. M., & Gassmann, F. (2023). Early growing season weather variation, expectation formation and agricultural land allocation decisions in Ethiopia. Journal of Agricultural Economics, 74(1), 255–272. Akresh, R., Lucchetti, L., & Thirumurthy, H. (2012). Wars and child health: Evidence from the Eritrean– Ethiopian conflict. Journal of Development Economics, 99(2), 330–340. Alderman, H., Hoddinott, J., & Kinsey, B. (2006). Long term consequences of early childhood malnutrition. Oxford Economic Papers, 58(3), 450–474. Alkire, S., Kanagaratman, U., & Suppa, N. (2021). The Global Multidimensional Poverty Index (MPI) 2021, OPHI MPI Methodological Note 51, Oxford Poverty and Human Development Initiative, University of Oxford. Amare, M., & Balana, B. (2023). Climate change, income sources, crop mix, and input use decisions: Evidence from Nigeria. Ecological Economics, 211, 107892. Ambel, A. A. (2024). Poverty dynamics and mobility in Ethiopia from 2019-2022: Evidence from longitudinal data. Mimeo. Ambel, A. A., Sosa, L. C. M., Kassa, W. Y., Tsegay, A. H., & Wieser, C. (2021). Monitoring COVID-19 Impacts on Households in Ethiopia: How COVID-19 is Affecting households - Results from the High-Frequency Phone Surveys of Households from April 2020 through January 2021 (English). Antle, J. M. (1983). Testing the stochastic structure of production: a flexible moment-based approach. Journal of Business & Economic Statistics, 1(3), 192–201. Antle, J. M. (1987). Econometric estimation of producers’ risk attitudes. American Journal of Agricultural Economics, 69(3), 509–522. Aragón, F. M., Oteiza, F., & Rud, J. P. (2021). Climate change and agriculture: Subsistence farmers’ response to extreme heat. American Economic Journal: Economic Policy, 13(1), 1–35. 79 REFERENCES Baquie, S., & Fuje, H. (2020). Vulnerability to poverty following extreme weather events in Malawi. World Bank. Available online: https://elibrary.worldbank.org/doi/abs/10.1596/1813-9450-9435. Policy Research Working Paper - World Bank. Bazzi, S., Blair, R. A., Blattman, C., Dube, O., Gudgeon, M., & Peck, R. (2022). The promise and pitfalls of conflict prediction: evidence from Colombia and Indonesia. Review of Economics and Statistics, 104(4), 764–779. BBC. (2021). Oromo Liberation Army: On the ground with Ethiopian fighters. BBC News. https://www.bbc. com/news/av/world-africa-59095778. Bozzola, M., & Finger, R. (2021). Stability of risk attitude, agricultural policies, and production shocks: evidence from Italy. European Review of Agricultural Economics, 48(3), 477–501. Brummund, P. W., Mann, C., & Rodríguez-Castelán, C. (2018). Job Quality and Poverty in Latin America. Review of Development Economics, 22(4), 1682–1708. Bundervoet, T. (2018). “Internal Migration in Ethiopia Evidence from a Quantitative and Qualitative Research Study.” Washington, DC: World Bank Group. https://openknowledge. worldbank.org/ handle/10986/32097. Bundervoet, T., Verwimp, P., & Akresh, R. (2009). Health and civil war in rural Burundi. Journal of Human Resources, 44(2), 536–563. Carter, M. R., Little, P. D., Mogues, T., & Negatu, W. (2007). Poverty Traps and Natural Disasters in Ethiopia and Honduras. World Development, 35(5), 835–856. https://doi.org/10.1016/j.worlddev.2006.09.010 Chau, T. N., & Scrimgeour, F. (2023). Will climate change jeopardize the Vietnamese target of maintaining farmland for food security? A fractional multinomial logit analysis of land use choice. Agricultural Economics, 54(4), 570–587. Cui, X. (2020). Climate change and adaptation in agriculture: Evidence from US cropping patterns. Journal of Environmental Economics and Management, 101, 102306. Cui, X., & Xie, W. (2022). Adapting agriculture to climate change through growing season adjustments: Evidence from corn in China. American Journal of Agricultural Economics, 104(1), 249–272. Cui, X., & Zhong, Z. (2024). Climate change, cropland adjustments, and food security: Evidence from China. Journal of Development Economics, 167(December 2023), 103245. https://doi.org/10.1016/j. jdeveco.2023.103245 Datt, G., & Ravallion, M. (1992). Growth and redistribution components of changes in poverty measures: A decomposition with applications to Brazil and India in the 1980s. Journal of Development Economics, 38(2, April), 275–259. https://doi.org/https://doi.org/10.1016/0304-3878(92)90001-P de Brauw, A., Mueller, V., & Woldehanna, T. (2018). Does internal migration improve overall well-being in Ethiopia? Journal of African Economies, 27(3), 347–365. https://doi.org/10.1093/jae/ejx026 Deaton, A., & Zaidi, S. (2002). Guidelines for Constructing Consumption Aggregates For Welfare Analysis. In World Bank Living Standards Measurement Study Working Paper (Vol. 135). Dercon, S., & Christiaensen, L. (2011). Consumption risk, technology adoption and poverty traps: Evidence from Ethiopia. Journal of Development Economics, 96(2), 159–173. https://doi.org/10.1016/j. jdeveco.2010.08.003 80 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Dhakal, C., Khadka, S., Park, C., & Escalante, C. L. (2022). Climate change adaptation and its impacts on farm income and downside risk exposure. Resources, Environment and Sustainability, 10. https://doi. org/10.1016/j.resenv.2022.100082 Di Falco, S., Veronesi, M., & Yesuf, M. (2011). Does Adaptation to Climate Change Provide Food Security? A Micro-Perspective from Ethiopia. American Journal of Agricultural Economics, 93(3), 829–846. https://doi.org/10.1093/ajae/aar006 Di Falco, S., & Vieider, F. M. (2022). Environmental adaptation of risk preferences. Economic Journal, 132(648), 2737–2766. ECHO. (2022). ECHO (European Commission’s Directorate-General for European Civil Protection and Humanitarian Aid Operations) Factsheet – Ethiopia. EPO. (2021). West and Kellem Wollega. Ethiopia Peace Observatory. https://epo.acleddata.com/west- and-kellem-wollega. EPO. (2023). Afar-Somali Regions Border Conflict. Ethiopia Peace Observatory. https://epo.acleddata. com/afar-somali-border-conflict. FAO. (2020). Food aid supports climate-adaptive investments by farmers in sub-Saharan Africa. Franklin, S., Imbert, C., Abebe, G., & Mejia-Mantilla, C. (2024). Urban Public Works in Spatial Equilibrium: Experimental Evidence from Ethiopia. American Economic Review, 114(5), 1382-1414. Funk, C., Peterson, P., Landsfeld, M., Pedreros, D., Verdin, J., Shukla, S., Husak, G., Rowland, J., Harrison, L., Hoell, A., & Michaelsen, J. (2015). The climate hazards infrared precipitation with stations—a new environmental record for monitoring extremes. Scientific Data, 2, 150066. Galor, O., & Özak, Ö. (2016). The Agricultural Origins of Time Preference. American Economic Review, 106(10), 3064–3103. https://doi.org/10.2139/ssrn.2486206 Grace, K., Hertrich, V., Singare, D., & Husak, G. (2018). Examining rural Sahelian out-migration in the context of climate change: An analysis of the linkages between rainfall and out-migration in two Malian villages from 1981 to 2009. World Development, 109, 187–196. Gray, C., & Mueller, V. (2012). Drought and population mobility in rural Ethiopia. World Development, 40(1), 134–145. Guillerm, M. (2017). Pseudo‑panel methods and an example of application to Household Wealth data. Economie et Statistique, 491(1), 109-130. Harris, D., Baird, S., Ford, K., Hirvonen, K., Jones, N., Kassa, M., Meyer, C., Pankhurst, A., Wieser, C., & Woldehanna, T. (2021). The Impact of COVID-19 in Ethiopia: Policy Brief. He, X., & Chen, Z. (2022). Weather, cropland expansion, and deforestation in Ethiopia. Journal of Environmental Economics and Management, 111, 102586. Hill, R., & Fuje, H. (2020). What is the Impact of Weather Shocks on Prices? : Evidence from Ethiopia. What Is the Impact of Weather Shocks on Prices? : Evidence from Ethiopia, September. https://doi. org/10.1596/1813-9450-9389 Hoddinott, J., & Kinsey, B. (2001). Child growth in the time of drought. Oxford Bulletin of Economics and Statistics, 63(4), 409–436. Hönig, T. (2021). The Legacy of Conflict: Aggregate Evidence from Sierra Leone. UNU-WIDER Working 81 REFERENCES Paper, 104/2021. (UNU-WIDE104/2021; UNU-WIDER Working Paper). https://doi.org/10.35188/ UNU-WIDER/2021/044-3. Hovhannisyan, S., Montalva-Talledo, V., Remick, T., & Rodríguez-Castelán, C., Stamm, K. (2022). Global job quality: Evidence from wage employment across developing countries. Working Paper, 10134. Hussain, A., Memon, J. A., & Hanif, S. (2020). Weather shocks, coping strategies and farmers’ income: A case of rural areas of district Multan, Punjab. Weather and Climate Extremes, 30, 100288. IGC. (2022). Finding a path to peace in Ethiopia’s Tigray region. International Crisis Group. https://www. crisisgroup.org/africa/horn-africa/ethiopia/167-finding-path-peace-ethiopias-tigray-region. Ikegami, M., Carter, M. R., Barrett, C. B., & Janzen, S. (2017). Poverty traps and the social protection paradox. In The economics of poverty traps (pp. 223-256). University of Chicago Press. In The economics of poverty traps (pp. 223–256). University of Chicago Press. Ilukor, J., & Gourlay, S. (2021). Locus Invasion in Ethiopia: Scope & Impact - Evidence from the World Bank-Supported High-Frequency Phone Surveys. IFPRI & IDS (2022). The Productive Safety Net Programme IV End-line Evaluation (2021) – Main Findings. Mimeo. IOM. (2023). Ethiopia national displacement report 15. International Organization for Migration. Jagnani, M., Barrett, C. B., Liu, Y., & You, L. (2021). Within-season producer response to warmer temperatures: Defensive investments by Kenyan farmers. The Economic Journal, 131(633), 392–419. Kumar, S., Mishra, A. K., Pramanik, S., Mamidanna, S., & Whitbread, A. (2020). Climate risk, vulnerability and resilience: Supporting livelihood of smallholders in semiarid India. Land Use Policy, 97, 104729. Little, P. D., Stone, M. P., Mogues, T., Castro, A. P., & Negatu, W. (2006). ‘Moving in place’: drought and poverty dynamics in South Wollo, Ethiopia. Journal of Development Studies, 42(2), 200–225. Mancini, G., & Vecchi, G. (2022). On the Construction of a Consumption Aggregate for Inequality and Poverty Analysis (Issue March). http://documents.worldbank.org/curated/en/099225003092220001/ P1694340e80f9a00a09b20042de5a9cd47e MoFED. (2002). Poverty Profile of Ethiopia: Analysis based on the 1999/00 HICE and WM Survey Results. MoFED. (2012). Ethiopia’s Progress Towards Eradicating Poverty: An Interim Report on Poverty Analysis Study (2010/11) (Issue March). Mueller, V., & Quisumbing, A. (2011). How resilient are labour markets to natural disasters? The case of the 1998 Bangladesh flood. Journal of Development Studies, 47(12), 1954–1971. Mulungu, K., Kimani, M. E., & Sarr, M. (2023). Do farmers’ risk preferences remain stable in the face of climatic shocks? Evidence from smallholder farmers in Zambia. Applied Economics, 00(00), 1–17. https://doi.org/10.1080/00036846.2023.2177599 Nainggolan, D., Abay, A. T., Christensen, J. H., & Termansen, M. (2023). The impact of climate change on crop mix shift in the Nordic region. Scientific Reports, 13(1), 2962. Nakamura, S., Bundervoet, T., & Nuru, M. (2020). Rural Roads, Poverty, and Resilience: Evidence from Ethiopia. Journal of Development Studies, 56(10), 1838–1855. https://doi.org/10.1080/00220388. 2020.1736282 82 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Narloch, U., & Bangalore, M. (2018). Themultifaceted relationship between environmental risks and poverty: New insights from Vietnam. Environment and Development Economics, 23(3), 298–327. https://doi.org/10.1017/S1355770X18000128 OCHA. (2022). Ethiopia Drought Update - UN Office for the Coordination of Humanitarian Affairs. Opondo, D. O. (2013). Erosive coping after the 2011 floods in Kenya. International Journal of Global Warming, 5(4), 452–466. PDC. (2018). Poverty and Economic Growth in Ethiopia (1995/96-2015/16). Planning and Development Commission (PDC), Federal Democratic Republic of Ethiopia, December, Addis Ababa. (Issue December). Porter, C., & Goyal, R. (2016). Social protection for all ages? Impacts of Ethiopia’s Productive Safety Net Program on child nutrition. Social Science & Medicine, 159, 92–99. https://doi.org/10.1016/j. socscimed.2016.05.001 Ravallion, M., & Huppi, M. (1991). Measuring changes in poverty: A methodological case study of Indonesia during an adjustment period. World Bank Economic Review, 5(1), 57–82. https://doi. org/10.1093/wber/5.1.57 Rohner, D., Thoenig, M., & Zilibotti, F. (2013). Seeds of distrust: Conflict in Uganda. Journal of Economic Growth, 18, 2021/044-3. https://doi.org/10.35188/UNU-WIDER/2021/044-3. Sardar, A., Kiani, A. K., & Kuslu, Y. (2021). Does adoption of climate-smart agriculture (CSA) practices improve farmers’ crop income? Assessing the determinants and its impacts in Punjab province, Pakistan. Environment, Development and Sustainability. https://doi.org/10.1007/s10668-020- 01049-6 Serneels, P., & Verpoorten, M. (2015). The impact of armed conflict on economic performance: Evidence from Rwanda. Journal of Conflict Resolution, 59(4), 555–592. Shimeles, A., & Woldemichael, A. (2023). Rising food prices and household welfare in Ethiopia: evidence from micro data. African Development Bank Group, Working Paper Series, (182). Skoufias, E. (2012). The Poverty and Welfare Impacts of Climate Change. In The Poverty and Welfare Impacts of Climate Change (Issue January). https://doi.org/10.1596/978-0-8213-9611-7 Taffesse, A. S., Dorosh, P., & Asrat, S. (2011). Crop Production in Ethiopia: Regional Patterns and Trends”, Working Paper No. 0016, Development Strategy and Governance Division, International Food Policy Research Institute, Ethiopia Strategy Support Program II, Ethiopia. March 2011. (0016). Tesfaye, W., Blalock, G., & Tirivayi, N. (2021). Climate-Smart Innovations and Rural Poverty in Ethiopia: Exploring Impacts and Pathways. American Journal of Agricultural Economics, 103(3), 878–899. https://doi.org/10.1111/ajae.12161 Tesfaye, W. M. (2022). Crop diversification and child malnutrition in rural Ethiopia: Impacts and Pathways. Food Policy, 113. https://doi.org/10.1016/j.foodpol.2022.102336 Torquebiau, E. (2017). Climate-smart agriculture: Making agriculture climate-friendly | Climate-smart agriculture: Pour une agriculture climato-compatible. Cahiers Agricultures. van Wijk, M. T., Merbold, L., Hammond, J., & Butterbach-Bahl, K. (2020). Improving Assessments of the Three Pillars of Climate Smart Agriculture: Current Achievements and Ideas for the Future. 83 REFERENCES In Frontiers in Sustainable Food Systems (Vol. 4). Frontiers Media S.A. https://doi.org/10.3389/ fsufs.2020.558483 Verbeek, M., & Nijman, T. (1992). Can cohort data be treated as genuine panel data? Empirical Economics, 17(1), 9–23. Verbeek, M., & Nijman, T. (1993). Minimum MSE estimation of a regression model with fixed effects from a series of cross‑sections. Journal of Econometrics, 59(1–2), 125–136. Wieser, C., & Abubakar, A. (2024). Has Ethiopia harnessed structural transformation in the labor market? Mimeo. Wieser, C., Franklin, S., & Hari, S. (2021). Effectiveness of the Urban Productive Safety Net Project. World Bank Group. Wieser, C., & Yitbarek, E. (2024). Impact of Food Inflation on Household welfare in Ethiopia. Mimeo. Worku, I. H., Dereje, M., Minten, B., & Hirvonen, K. (2017). Diet transformation in Africa: the case of Ethiopia. Agricultural Economics (United Kingdom), 48(March), 73–86. https://doi.org/10.1111/ agec.12387 World Bank. (2019). Disaster Risk Profile: Ethiopia. https://reliefweb.int/report/ethiopia/disaster-risk- profile-ethiopia World Bank. (2020). Ethiopia Poverty Assessment: Harnessing Continued Growth for Accelerated Poverty Reduction. World Bank. (2021). Climate Risk Profile: Ethiopia (2021). World Bank. (2022a). Ethiopia’s great transition: the next mile - A Country Economic Memorandum, Washington, DC: World Bank. www.worldbank.org World Bank. (2022b). Ethiopia - Rural Income Diagnostics Study: Leveraging the Transformation in the Agri-Food System and Global Trade to Expand Rural Incomes. Washington, D.C. : World Bank Group. https://documentsinternal.worldbank.org/search/33891468. World Bank. (2022c). Voluntary Migration in Ethiopia: In Search for Work and Better Opportunities. Washington, D.C.: World Bank Group. https://documentsinternal.worldbank.org/search/33889248. World Bank. (2023). The State of Welfare of Pastoral Communities in Ethiopia – Forthcoming. World Bank. (2024). Ethiopia Poverty Measurement Methodology Note 2021. Mimeo. Yimer, F., Alemayehu, M., & Taffesse, S. (2020). The Short-run Impact of the COVID-19 Crisis on Poverty in Ethiopia. 84 ETHIOPIA POVERTY AND EQUITY ASSESSMENT ANNEX ANNEX 1: MONETARY POVERTY MEASUREMENT METHODOLOGY Data for poverty measurement This poverty assessment is based on the Household Welfare Statistics (HoWStat) survey conducted in 2020/21 by the Ethiopian Statistical Service (ESS). The ESS, previously known as the Central Statistics Agency (CSA), has a long history of conducting national household surveys to regularly track poverty and living standards, a practice initiated in 1995/96. The Household Income and Consumption Expenditure Survey and Welfare Monitoring Survey (HCES/WMS) stand out as the most significant among these, with its inception in 1995/96 and subsequent iterations in 1999/2000, 2004/05, 2010/11, and 2015/16. The HCES and WMS were combined into a single survey named Household Welfare Statistics (HoWStat) in 2021. HoWStat 2021 was conducted over a full year from January to December 2021 by randomly assigning households to each of the 12 months at the enumeration areas (EA) level to account for seasonal effects. The HoWStat 2021 survey covered all rural and urban areas of the country, regardless of area type, except for the Tigray region due to ongoing conflict. The survey utilized a two-stage stratified sampling technique to draw representative samples. The country was first stratified into nine regional states and two city administrations. Then each regional state was further stratified into three broad categories namely, rural, major urban centers, and other urban area categories. However, Harari region and Dire Dawa City Administration were stratified into rural and urban categories, while Addis Ababa has only an urban category, but is stratified by sub-city. In most cases, each sub-stratum was a survey domain or reporting level for which the major findings of the survey could be reported. In this way, the HoWStat 2021 survey has 45 reporting levels. Enumeration Areas (EAs) are the Primary Sampling Units (PSUs) and the households as the Secondary Sampling Unit (SSU). In total, the sample size of the HoWStat 2021 survey is 38,828, which is higher than the previous surveys. Methodological Approach The methodological choice for the 2021 poverty assessment hinged on two options: adopting a new methodology based on 2021 data or replicating the 2016 methodology. The new methodology based on 2021 data would entail generating a new consumption aggregate, a new poverty basket, and a new poverty line based on international best practices. On the other hand, the replication of the 2016 methodology would mean using the same approach to construct the consumption aggregate as in 2016 and making a temporal price adjustment using an appropriate deflator to inflate the 2016 poverty lines to 2021 prices. Upon careful consideration, it was evident that adopting a new methodology based on 2021 data may not be the optional path. The year 2021 was marked by crisis, making it an atypical year to anchor a new poverty basket. Notably, there was a decline in caloric intake in 2021 compared to 2016, coupled with an uptick in the prevalence of caloric inadequacy, indicating a rise in calorie-based poverty. More importantly, there is inadequate coverage of the country, mainly the Tigray region and conflict- affected areas for security reasons. The fact that 2021 appears to be an unusual year makes the 2021 data inappropriate for establishing a new poverty estimation methodology. For this reason, the government 85 ANNEX 1: MONETARY POVERTY MEASUREMENT METHODOLOGY plans to conduct a new survey from July 2024 to June 2025, with an improved consumption module and national coverage, laying the groundwork for a new poverty line. Therefore, the more viable option is the replication of the 2016 methodology to the greatest extent possible. This approach offers the advantage of consistency, enabling the comparison of poverty trends over time. It relies on an established methodology with published figures from 2016 (PDC, 2018; World Bank, 2020), facilitating the analysis of poverty changes. Replicating the 2016 methodology requires a consistent approach to generating the nominal consumption aggregate, generating spatial price deflators to account for regional price differences, and updating/deflating price changes over time. The main challenge with this replication lies in the incomplete information regarding the creation of spatial deflators in 2016. Consumption Aggregate Construction Welfare measurement in Ethiopia is based on consumption. The debate over the choice of income and consumption as indicators of welfare—highlighted by Mancini & Vecchi (2022)—continues, with consumption often being favored as a more reliable indicator of current living standards, especially in developing countries where formal labor market participation and income data may be less available. Official poverty rates in Ethiopia are calculated based on household budget surveys that capture consumption or expenditure on food and non-food items, which form consumption aggregates. The construction of the consumption aggregate is structured around four main components: (1) food, (2) non-food nondurables, (3) durables, and (4) housing. The food component corresponds to the value of food consumed by households acquired through different means (purchased in the marketplace, home- produced, received in-kind or as gifts, and food consumed away from home). The non-food component captures the value of non-food items that are consumed by households and have a short lifespan, such as clothing, personal care products, and household supplies. The durables component includes the use value of durable goods, such as furniture, appliances, and vehicles, in a reference period. The housing component captures the cost of housing for households, either the rent for tenants or an estimated value of what homeowners would pay if they were to pay rent for their dwellings. The nominal consumption aggregate is constructed as the sum of the value of the four components. It is finally adjusted to December 2021 using within survey temporal deflators and for spatial variations in the cost of living using spatial price deflators. To ensure consumption reflects household welfare, adjustments are made for the varying caloric needs of household members based on age and gender. This is achieved by applying an adult equivalent scale to household consumption. Food aggregates The food component of household consumption encompasses various sources of food acquisition, including (1) market purchases; (2) own production; (3) in-kind receipts or gifts; and (4) food away from home. The HoWStat 2021 survey detailed the acquisition and consumption of 669 different food and non-alcoholic beverage items, collecting data biweekly using “last three days” and “last four days” recall periods for the first and second visits within a week. The survey also asked about the acquisition and consumption of 11 items of alcoholic beverages, cigarettes, and tobacco. However, following the recommendation of Deaton and Zaidi (2002) and the latest COICOP classification, these items are excluded from the food aggregate. Data on food consumed away from home, an increasingly relevant component of food consumption in Ethiopia (Worku et al., 2017), is collected within the same recall period and separate line items for breakfast, lunch/dinner, and drinks within the consumption module. 86 ETHIOPIA POVERTY AND EQUITY ASSESSMENT HoWStat 2021 also includes school meal data estimated by the ESS team based on data obtained from the Ministry of Education (MoE). While not a separate category, school meal expenditure is mapped to the different items in the survey. The food consumption aggregate is obtained by adding the consumption of food and non-alcoholic beverages from all the sources valued at survey prices. The HoWStat 2021 survey asks households to report the quantity of each food item consumed during the reference period then the monetary value of these quantities is estimated using local market prices – reported by households first or from prices from the market survey module when households do not report a price - forming the basis of the food consumption aggregate. It follows a consumption rather than acquisition approach. While one can conclude that household-level prices are used in the survey, the ESS also emphasizes that these prices are reasonably comparable to the local market prices, hence the quantities consumed are valued at prevailing prices in the enumeration area. The monetary valuation of food away from home consumption follows the same criteria used to value the consumption from food purchased. All values were standardized and converted to a monthly period, i.e., December 2021. The annual food expenditure aggregation is in line with the best practice suggested by the Mancini and Vecchi (2022) guidelines. Non-food non-durables Non-food nondurable categories considered in the construction of the welfare aggregate do not differ from best practices. Items classified under the non-food nondurables include (i) clothing and footwear, (ii) housing, water, electricity, gas, and other fuels (excluding actual rent and imputed rent), (iii) furnishings, household equipment and routine maintenance such as construction materials only for repairs and maintenance, (iv) transport and communication excluding purchase of vehicle and communication equipment, (v) personal care, (vi) recreation, sport, and culture, (vii) restaurants and accommodation services (excluding food consumed away from home), (viii) insurance (financial services are excluded), and (ix) other miscellaneous goods and services frequently purchased and consumed. Also included under the non-food nondurables category is alcohol, tobacco, and narcotics (e.g., chat). The HoWStat 2021 survey included data on transportation allowances for civil servant workers (calculated at 2837.51 Birr per person per year with several assumptions and rates). This data was not present in the 2015/16 survey; hence they were omitted from the non-food consumption aggregates to maintain comparability. As per the 2015/16 approach and as suggested by Mancini and Vecchi (2022), health and education expenditures are part of the total welfare aggregate. While the best practice in computing non-food aggregates is to exclude “lumpy” and relatively infrequent expenditures that do not reflect regular household consumption— such as expenses associated with special or extraordinary occasions (such as funerals or weddings), high-value jewelry, transfers to other households, donations to religious or charitable organizations, financial expenses—the 2016 consumption aggregates did not explicitly document the exclusion of these expenditures. To ensure comparability over time and mainly due to the lack of documented exclusion in 2016, these types of expenditures have been included in the 2021 consumption aggregates. The non- food nondurable consumption aggregate is computed using annualized estimates for consumption items measured with a monthly or 3 months recall period. Durables component The use value of durables is calculated by taking the full value of durable purchases in the survey as was done in 2016. The valuation of durable goods within consumption surveys is a nuanced process that requires careful consideration of the long-term benefits these goods provide as the utility offered 87 ANNEX 1: MONETARY POVERTY MEASUREMENT METHODOLOGY by durable goods often goes beyond the scope of a single survey cycle (Mancini & Vecchi, 2022). As such, it is the service flow—or the use value—of durable goods that contributes to welfare, rather than the initial purchase price. The information on durable goods and their characteristics is inconsistent between the 2015/16 and 2021 household surveys. The latter survey includes data on the ownership of household assets and their characteristics (including quantities, purchase price, age, and estimated current value). In contrast, the 2015/16 survey lacks detailed information on household assets, and neither the government report nor the previous poverty assessment provides the value of durables for that period. This discrepancy hinders the application of best practices for calculating the use value or service flow from durables for 2016. Consequently, it is not feasible to compute consumption flows from durables using the 2015/16 survey data, which impedes the comparison of the value of durables across survey periods. For the 2021 calculations, the survey reports the full purchase value of durable goods, which is used to estimate the durables component, consistent with the method used in 2016. The items considered “durables” are identified by mapping those listed in the expenditure section with those in the asset module and include (i) furniture and furnishing, carpets and other floor coverings, (ii) major kitchen, laundry, cleaning, and household appliances (such as air conditioners, washing machines, dryers, freezers, stoves, microwaves, among others), (iii) major tools and household equipment, (iv) information, communication, and recreation equipment (such as televisions, radios, DVD players, VCRs, computers, printers, faxes, among others), and (v) purchase of vehicles. Farm implements, other productive assets, and jewelry (considered an investment) are excluded from the computation of the “durables” component. Despite deviating from international best practices that recommend using the service flow of durables, this method was adopted due to the limitations of the available data and to maintain comparability in the estimation approach with 2016. Housing component Implicit rent based on the self-assessment method is used to compute the housing component for non-renters. The housing component of the consumption aggregate reflects the value of housing services used by households within a specific reference period. For renters, the housing component of the consumption aggregate is straightforwardly calculated as the actual rent paid. One method to estimate the housing component is to use self-reported imputed rent (implicit rent), which involves asking homeowners and non-market tenants to estimate the rent they would pay for their housing units. For non- renters, the housing component is estimated by soliciting their response to the question “How much rent would you charge monthly if you were to rent this accommodation?” or “If you were to pay rent for this dwelling, how much would you pay per month?”. This method of self-assessment yields an implicit rent for owner-occupied housing, aligning with the methodology established in 2016. For subsidized housing, the actual rent paid is used. The ESS incorporates imputed rent into the expenditure data but introduced an adjustment for kebele rented houses in Addis Ababa and other urban areas in the 2021 survey which was not present in the 2016 survey. This adjustment is based on rental values from housing agencies and kebele house rental values. Another approach is to calculate rental equivalences using econometric models, such as hedonic regression or the Heckman selection model. However, these methods require a developed rental market and consistent rent information across all population groups to be fully effective. The rental market in Ethiopia is not developed and a greater share of renters are concentrated in urban areas. Moreover, this is not the approach used in 2016. 88 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Adjustment for spatial and temporal price variations Accurate comparisons of consumption levels across households require using consistent monetary units that reflect the same individual utility. The price level of commodities and services varies both over time (inflation) and across geographical locations (cost-of-living differences). Two households with identical nominal expenditures but facing different price levels will not be able to purchase the same quantity of goods; hence, their nominal expenditures do not equate to the same standard of living. To make meaningful welfare comparisons across individuals, it is necessary to account for these spatial and temporal differences. This is achieved by applying spatial and temporal priced deflators, which adjust the nominal figures to account for price disparities at different reporting periods and locations. Within-survey spatial price adjustments were made to address differences in prices across locations. Spatial price indices at the reporting level or strata were calculated using the methodology used for spatial price adjustment in 2016 and before (PDC, 2018). Calculating spatial price indices at the reporting level (region) ensures a robust sample size of households, encompassing both urban and rural areas, which reduces the variance of mean prices and decreases the risk of insufficient household data for certain food or nonfood items. Laspeyres weighted price food and non-food spatial deflators, utilized in the previous government poverty reports (MoFED, 2002) are calculated and used for the spatial price adjustment. While the calculation of money metric utility requires that the nominal aggregate be deflated by a Paasche price index, the Laspeyres index is the deflator to use if the analyst prefers to work with the welfare ratio approach to measurement. However, these price indexes are of independent interest beyond their roles in deflating expenditures, simply for measuring prices (Deaton & Zaidi, 2002). The reporting level (or regional) relative price index—Laspeyres (relative to the national average) price index for food and nonfood items is calculated using the reporting level (regional) average price of each item, the average national price, and the national budget share of the item determined in 2011. The poverty basket was published in previous poverty estimation reports by the government and used in the 2015/16 poverty measurement. The basket encompasses both food and nonfood groups. The values of the Laspeyres deflator calculated using the price is consistent with one calculated using detailed list of items (CPI items). However, a comparison with the Paasche index shows that results are sensitive. An alternative approach to the spatial price deflator is to use the 2016 deflators, which necessitates matching the HoWStat 2021 data to reporting levels for HECS 2016 to use spatial deflators provided at the strata level in 2016. This is not an appropriate option since there are changes in prices across regions between 2016 and 2021 that should be reflected in the spatial price index. The Laspeyres spatial deflator based on 2021 prices is selected over adopting the 2016 deflators because it also yields monetary poverty rates more consistent with the non-monetary indicators. The reporting levels (strata) were ranked based on poverty rates using the two spatial deflator options, as well as non-monetary poverty indicators such as MPI, education, health, food budget share, food security, and access to water and sanitation. The Spearman’s rank correlation coefficient of the ranking on poverty estimates based in each spatial deflating approach on one hand, and the non-monetary poverty rankings on the other, indicates that the alignment between monetary and non-monetary indicator rankings is stronger when we use the 2021 prices that when the 2015/16 spatial deflators are used. Within-survey temporal price adjustments were done using national CPI. After adjusting the nominal consumption aggregate for variation in cost of living across space, it is adjusted for price variations over 89 ANNEX 1: MONETARY POVERTY MEASUREMENT METHODOLOGY time. This step is crucial because the survey was conducted over a 1-year period from January 2021 to December 2021, during which the annual inflation rate reached 27 percent. Specifically, the food inflation rate was 31 percent, while the non-food inflation rate was 20 percent. The high inflation rates during the survey’s implementation period underscore the necessity of temporal price adjustments. Without such adjustments, the calculated poverty levels could be significantly skewed. To address this, a temporal price deflator is applied using the official national food and non-food CPI deflators reported by the ESS with December 2021 as the reference month. These CPIs were recalculated excluding data from the Tigray region which were not correctly collected during the survey period due to insecurity reasons. The practice of using official food and non-food CPIs from the ESS aligns with the methodology employed in previous government reports and the most recent poverty assessment (World Bank, 2020). The sensitivity of poverty estimations for changes in temporal price deflators is assessed by comparing temporal deflators based on official CPI data and one computed using survey prices. The calculation of the survey-prices-based temporal deflator is based on the CPI basket and involves mapping the expenditure items from the survey (for 2011, 2016, and 2021) to the CPI basket items and using similar CPI weights (from December 2016 in this case). The survey-price-based temporal deflator is calculated consistently with the regional monthly CPI deflator for 2021 by taking national CPI weights and using regional prices for each item for each month. Based on this, we generate food and non-food deflators using December 2021 as the base period. The resulting temporal price deflators are consistent with the figures obtained using the official CPI data. Poverty lines The poverty line for 2021 is obtained using two alternatives: by updating the 2015 poverty line using CPI deflators and by re-costing the original poverty basket in average 2021 prices. The method used by the government for defining and measuring poverty since 1995/96 has been consistently detailed in previous government reports (MoFED, 2002, 2012; PDC, 2018). While alternative approaches exist for defining poverty lines, poverty assessments in Ethiopia predominantly utilize the cost of basic needs (CBN) method. This method begins by defining a food poverty line through a selection of a food bundle commonly consumed by the poor, ensuring it meets a set minimum caloric intake (2,200 kcal). Since its determination in 1996, the composition of the food basket has remained unchanged. To derive the overall poverty line, a food poverty line was first derived by costing the food basket at average national prices during the survey period. A non-food poverty line is added to the food poverty line to make up an overall poverty line by dividing the food poverty line by the food budget share of the poorest 25 percent to get the total poverty line (MoFED, 2002). Accordingly, the poverty lines for the first three surveys (1995/96, 1999/00, 2004/05) were determined to be Birr 1,075 deflated to 1995/96 constant prices. The poverty line was revised in 2011 by re-costing the items in the original food basket at prevailing prices and doing a similar adjustment for non-food consumption. The food and total poverty line were 1,985 and 3,781 Birr per adult equivalent per year, respectively (in December 2010 prices). For the 2015/16 poverty measurement, the 2011 poverty line was inflated using the GDP deflator, resulting in a poverty line of 7,184 Birr per adult equivalent per year in December 2015 prices (World Bank, 2020). The 2021 poverty line can be obtained using two alternative approaches: updating the 2015 poverty line using CPI and re-costing the existing poverty basket in 2021 prices. Under the first alternative, the food and non-food poverty lines from December 2015 are updated upwards using the national food and non-food CPI for December 2015 and December 2021. The second alternative involves costing the existing food poverty basket using average national 2021 prices (this gives the food poverty line) and adjusting for non-food allowance by dividing the food poverty line by the food budget share of the poorest 25 percent. The total of the two gives the national poverty line. Table A.1.1. summarizes the poverty lines in each survey year since 2010/11. 90 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Table A.1. Food and total (absolute) poverty lines for Ethiopia (in Birr and average prices) Year 2010/11 2015/16 2020/21 Source or method Published Published in Re-costed CPI deflator Re-costed CPI deflator in 2011 2016 poverty basket based poverty basket based Food poverty line per adult per year 1,985 3,781 4,458 3,941 11,060 10,903 Total poverty line per adult per year 3,781 7,184 7,615 7,442 17,753 18,964 Source: NPC (2017) and World Bank Staff Calculations based on HoWStat 2020/21. 91 ANNEX 2: MULTIDIMENSIONAL POVERTY METHODOLOGY Multidimensional poverty is calculated using the Global MPI methodology by Alkire et al. (2022) and available indicators in WMS 2015/16 and HoWStat 2021 datasets. The Multidimensional Poverty Index (MPI) covers three dimensions, health, education, and living standards, with nine indicators as depicted in Table A.2. A deprivation score is assigned to each individual based on his or her household’s deprivation status based on the provided definitions, taking the value 1 if the household is deprived, and 0 otherwise. After adding the deprivation scores of each indicator to get the total deprivation score, people with a deprivation score of one-third (33 percent) or higher are identified as multidimensionally poor. In addition, those with a deprivation score of one-fifth (20 percent) or higher are vulnerable to multidimensional poverty, and people with a 50 percent or higher deprivation score are in severe multidimensional poverty. Hence, the incidence of multidimensional poverty (poverty headcount, H) is estimated as the ratio of the number of multidimensional poor people (q) to the total population (n). Moreover, the intensity of poverty, which measures average deprivation among multidimensionally poor is calculated as: q ∑i gi A= where si -- the deprivation score of the ith multidimensional poor individual. Then, the q multidimensional poverty index is generated as a product of poverty headcount and intensity (MPI = H * A ) which ranges from 0 to 1, higher value representing a high level of deprivation. Finally, the contribution of dimensions d and indicators j to the MPI are calculated as: q q ∑jϵd ∑1 cij ∑1 cij Contd = /MPI and ContIndj = /MPI n n Where, cj, censored headcount of indicator j refers to the proportion of people who are multidimensionally poor and deprived in indicator j. Table A.2. Multidimensional poverty dimensions and indicators Dimensions and Definition of deprivation Weights indicators Health 1/3 Nutrition Any child under age 5 years for whom there is nutritional information is undernourished. 1/3 Child is undernourished if their z-score for either height-for-age (stunting) or weight- for-age (underweight) is below minus two standard deviations from the median of the reference population. Education 1/3 Years of schooling No household member aged ‘school entrance age + six years or older’ has completed at 1/6 least six years of schooling. In Ethiopia, entry age is 7 years for primary school. School attendance Any school-aged child not attending school up to the age at which he/she would 1/6 complete class eight. The age to complete 8th grade is 14 years. 92 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Dimensions and Definition of deprivation Weights indicators Living Standards 1/3 Housing At least one of the three housing materials for roof, walls and floor are inadequate: the floor 1/18 is of natural materials and/or the roof and/or walls are of natural or rudimentary materials.1 Asset The household does not own more than one of these assets: radio, television, telephone, 1/18 animal cart, bicycle, motorbike, or refrigerator, and does not own a car or truck. Electricity The household has no electricity (grid). 1/18 Drinking water The household lacks access to improved sources of drinking water.2 1/18 Sanitation The household lacks access to improved sanitation facilities. 1/18 Cooking fuel The household cooks with firewood, charcoal, dung/manure, crop residue/leaves, sawdust. 1/18 Source: Alkire et al. (2022). Notes: (1) Floor (mug/dung, bamboo/reed/wood planks); roof (thatch, wood and mud, bamboo/reed, plastic cover) or; Walls (wood and mud, wood and thatch, wood, stone, stone and mud, blocks (unplastered), parquet or polished wood, chip wood, bamboo/reed, plastic). (2) Improved drinking water sources include piped water, protected wells/spring/boreholes, rainwater, and packaged or delivered water. (3) Improved sanitation facilities are flush/pour flush toilets connected to piped sewer systems, septic tanks, or pit latrines; pit latrines with slabs (including ventilated pit latrines), and composting toilets, and not shared with other households. 93 ANNEX 3: ADDITIONAL DESCRIPTIVE STATISTICS AND REGRESSION RESULTS Table A.3.1. Total and food poverty and calorie deficiency by location and region, 2016 and 2021 Total poverty Food poverty Calorie deficiency Total Urban Rural Total Urban Rural Total Urban Rural 2016 2021 2016 2021 2016 2021 2016 2021 2016 2021 2016 2021 2016 2021 2016 2021 2016 2021 National 25.0% 33.1% 16.1% 18.6% 27.1% 37.2% 27.3% 32.6% 17.3% 17.6% 29.6% 36.8% 26.4% 30.5% 18.6% 29.9% 28.2% 30.7% Region: Afar 28.7% 31.8% 11.6% 20.7% 32.6% 34.8% 47.7% 25.6% 12.7% 17.8% 55.8% 27.7% 37.0% 24.7% 43.3% 26.6% 35.5% 24.2% Amhara 28.8% 34.2% 12.7% 20.1% 31.8% 37.6% 34.4% 33.1% 14.3% 17.4% 38.2% 36.9% 53.3% 32.4% 17.0% 32.4% 60.1% 32.3% Oromia 25.6% 32.3% 16.2% 15.9% 27.1% 35.6% 22.7% 34.4% 13.4% 17.4% 24.2% 37.9% 21.3% 31.1% 22.1% 29.4% 21.2% 31.5% Somali 22.7% 33.5% 25.3% 17.5% 22.3% 36.3% 28.3% 29.1% 31.1% 14.6% 27.8% 31.7% 16.8% 32.2% 16.3% 28.7% 16.9% 32.8% Benishangul-Gumuz 28.2% 29.1% 18.4% 19.6% 30.6% 32.1% 25.6% 35.0% 20.6% 23.9% 26.9% 38.6% 26.3% 33.5% 18.0% 35.6% 28.3% 32.8% SNNP 20.9% 37.6% 16.6% 27.1% 21.7% 40.1% 27.0% 36.5% 17.6% 29.1% 28.6% 38.1% 7.7% 29.5% 17.2% 30.0% 6.0% 29.3% Sidama 26.3% 41.9% 13.5% 29.2% 29.3% 45.6% 29.0% 31.2% 17.1% 25.1% 31.8% 33.0% 28.6% 21.4% 24.7% 27.9% 29.5% 19.5% 94 Gambella 24.3% 36.4% 17.5% 27.5% 27.8% 41.4% 19.1% 30.6% 13.6% 21.5% 22.0% 35.9% 10.9% 27.1% 13.0% 27.1% 9.9% 27.1% Harari 7.5% 8.1% 6.7% 6.7% 8.5% 9.8% 7.3% 15.1% 5.8% 8.6% 9.0% 23.3% 9.7% 26.3% 9.9% 33.4% 9.5% 17.4% Addis Ababa 18.0% 12.5% 18.0% 12.5% - - 23.8% 7.8% 23.8% 7.8% - - 14.9% 28.2% 14.9% 28.2% - - Dire Dawa 15.8% 14.2% 11.4% 14.8% 24.0% 13.1% 14.6% 20.4% 11.4% 15.4% 20.4% 29.5% 11.3% 25.4% 10.9% 29.8% 12.1% 17.4% Source: World Bank staff calculations using HCES 2011, 2016, and HoWStat 2021. Notes: Sidama was part of SNNP before 2021. The poverty rates for SNNP in 2011 and 2016 are calculated by including/excluding Sidama. ETHIOPIA POVERTY AND EQUITY ASSESSMENT Table A.3.2. Composition of the population and the poor by location, region, and agroecology, 2021 Total Urban Rural Population Share Population Share Population Share Share of poor Share of poor Share of poor National 100% 100% 21.9% 12.3% 78.1% 87.7% Region: Afar 2.0% 1.9% 0.4% 0.3% 1.6% 1.7% Amhara 23.9% 24.6% 4.7% 2.8% 19.2% 21.8% Oromia 40.0% 39.0% 6.8% 3.3% 33.2% 35.7% Somali 6.5% 6.6% 1.0% 0.5% 5.6% 6.1% Benishangul -Gumuz 1.2% 1.1% 0.3% 0.2% 0.9% 0.9% SNNP 16.6% 18.9% 3.1% 2.6% 13.5% 16.3% Sidama 4.4% 5.6% 1.0% 0.9% 3.4% 4.7% Gambella 0.5% 0.6% 0.2% 0.2% 0.3% 0.4% Harari 0.3% 0.1% 0.2% 0.0% 0.1% 0.0% Addis Ababa 4.0% 1.5% 4.0% 1.5% - - Dire Dawa 0.5% 0.2% 0.3% 0.2% 0.2% 0.1% Agroecological zone: Drought prone highlands 15.8% 13.4% 3.7% 1.8% 12.1% 11.5% Drought prone lowlands 9.1% 12.5% 1.0% 0.5% 8.2% 12.0% Moisture reliable lowlands 3.4% 3.9% 0.6% 0.5% 2.8% 3.4% Moisture reliable highlands 65.4% 65.0% 15.3% 8.7% 50.2% 56.3% Pastoral 6.2% 5.2% 1.3% 0.7% 4.9% 4.5% Source: World Bank staff calculations using HoWStat 2021. 95 Table A.3.3. Multidimensional poverty results by rural/urban and regions 2016 National Rural Urban Afar Amhara Oromia Somali Benishan- SNNP Sidama Gambella Harari Addis Dire gul- Gumuz Ababa Dawa % of people deprived in 49% 57% 15% 62% 53% 51% 62% 47% 46% 44% 20% 31% 6% 34% Years of schooling School attendance 25% 30% 4% 29% 14% 31% 43% 18% 28% 24% 10% 13% 2% 11% Nutrition 30% 34% 16% 48% 28% 31% 35% 31% 32% 30% 20% 25% 10% 17% Electricity 78% 93% 10% 73% 81% 81% 90% 81% 81% 74% 73% 19% 1% 31% Sanitation 97% 99% 88% 98% 99% 99% 99% 100% 94% 100% 96% 94% 72% 91% Drinking water 40% 48% 3% 39% 39% 44% 46% 20% 40% 50% 20% 17% 0% 9% Housing 98% 100% 89% 96% 99% 99% 97% 99% 99% 99% 98% 88% 77% 66% Cooking fuel 95% 100% 76% 100% 98% 98% 100% 100% 98% 99% 100% 73% 28% 82% Asset 71% 80% 28% 77% 79% 68% 89% 70% 72% 70% 73% 34% 7% 44% MPI 0.43 0.50 0.14 0.54 0.42 0.46 0.55 0.39 0.44 0.44 0.25 0.27 0.06 0.23 Headcount 72% 82% 28% 80% 73% 76% 87% 64% 72% 75% 48% 47% 13% 40% 96 Intensity 60% 61% 50% 67% 58% 61% 63% 60% 61% 58% 52% 56% 44% 57% Severity 50% 58% 15% 66% 47% 52% 67% 40% 52% 48% 25% 32% 4% 23% Vulnerability 16% 16% 16% 10% 19% 16% 9% 28% 17% 11% 39% 6% 5% 12% Contribution to MPI Health 23% 22% 39% 30% 22% 23% 21% 26% 24% 23% 27% 31% 55% 25% Education 28% 29% 21% 28% 26% 30% 32% 28% 28% 26% 19% 27% 13% 30% Living Standards 48% 49% 40% 42% 51% 48% 47% 46% 48% 51% 54% 42% 31% 45% Years of schooling 19% 19% 17% 19% 21% 18% 19% 20% 17% 17% 13% 19% 12% 23% School attendance 10% 10% 5% 9% 5% 11% 13% 8% 11% 9% 6% 8% 1% 7% Nutrition 23% 22% 39% 30% 22% 23% 21% 26% 24% 23% 27% 31% 55% 25% Electricity 8% 9% 2% 7% 9% 8% 8% 8% 8% 8% 9% 4% 1% 7% Sanitation 9% 9% 11% 8% 10% 9% 9% 9% 9% 10% 10% 10% 11% 10% Drinking water 5% 5% 1% 4% 5% 5% 4% 3% 5% 6% 4% 3% 0% 2% Housing 9% 9% 11% 8% 10% 9% 9% 9% 9% 10% 11% 10% 11% 9% Cooking fuel 9% 9% 10% 8% 10% 9% 9% 9% 9% 10% 11% 9% 7% 9% Asset 8% 8% 5% 7% 9% 7% 8% 7% 8% 8% 9% 6% 2% 8% ANNEX 3: ADDITIONAL DESCRIPTIVE STATISTICS AND REGRESSION RESULTS 2021 National Rural Urban Afar Amhara Oromia Somali Benishan- SNNP Sidama Gambella Harari Addis Dire gul- Gumuz Ababa Dawa % of people deprived in 45% 51% 22% 63% 43% 48% 68% 42% 41% 35% 19% 27% 10% 24% Years of schooling School attendance 31% 37% 12% 47% 24% 34% 58% 28% 33% 28% 20% 19% 1% 14% Nutrition 29% 31% 21% 45% 28% 30% 40% 35% 27% 23% 30% 24% 12% 26% Electricity 70% 86% 13% 78% 75% 72% 88% 66% 69% 71% 70% 18% 2% 28% Sanitation 85% 88% 71% 94% 89% 85% 84% 93% 84% 78% 84% 78% 59% 66% Drinking water 31% 39% 5% 54% 34% 31% 42% 19% 34% 24% 14% 16% 0% 9% Housing 95% 99% 81% 95% 98% 96% 94% 98% 99% 95% 96% 72% 61% 40% ETHIOPIA POVERTY AND EQUITY ASSESSMENT Cooking fuel 95% 100% 76% 99% 97% 97% 99% 99% 99% 97% 98% 68% 21% 73% Asset 68% 78% 32% 80% 76% 65% 88% 60% 70% 76% 77% 36% 11% 39% MPI 0.41 0.47 0.19 0.58 0.40 0.43 0.59 0.40 0.40 0.35 0.31 0.25 0.07 0.23 Headcount 69% 77% 38% 86% 69% 72% 88% 67% 68% 62% 57% 42% 16% 43% Intensity 60% 61% 51% 68% 58% 60% 68% 60% 58% 56% 55% 60% 44% 54% Severity 45% 53% 19% 71% 45% 47% 70% 44% 44% 37% 34% 29% 5% 22% 97 Vulnerability 17% 18% 14% 6% 20% 17% 8% 23% 20% 25% 29% 10% 7% 11% Contribution to MPI Health 24% 22% 36% 25% 24% 24% 22% 29% 22% 22% 32% 32% 58% 38% Education 30% 31% 26% 31% 28% 31% 35% 28% 31% 30% 20% 29% 14% 24% Living Standards 46% 47% 38% 43% 49% 45% 42% 42% 47% 48% 48% 39% 28% 39% Years of schooling 18% 18% 16% 18% 18% 18% 19% 17% 17% 16% 10% 17% 12% 15% School attendance 13% 13% 9% 13% 10% 13% 16% 12% 14% 13% 10% 12% 2% 9% Nutrition 24% 22% 36% 25% 24% 24% 22% 29% 22% 22% 32% 32% 58% 38% Electricity 8% 8% 3% 7% 8% 8% 8% 7% 8% 8% 8% 3% 1% 6% Sanitation 8% 8% 9% 8% 9% 8% 7% 9% 8% 8% 9% 8% 9% 8% Drinking water 4% 4% 1% 5% 4% 4% 4% 2% 4% 4% 2% 3% 0% 2% Housing 9% 9% 10% 8% 9% 9% 8% 9% 10% 10% 10% 9% 9% 7% Cooking fuel 9% 9% 10% 8% 9% 9% 8% 9% 10% 10% 10% 9% 5% 10% Asset 7% 8% 6% 7% 8% 7% 7% 6% 8% 9% 9% 6% 3% 7% Source: World Bank staff calculations using HoWStat 2021. ANNEX 3: ADDITIONAL DESCRIPTIVE STATISTICS AND REGRESSION RESULTS Table A.3.4. Poverty profile Expenditure Monetary Multidimensional quintiles poverty poverty Non- Non- Poorest Q2 Q3 Q4 Richest Poor Poor Poor Poor Household characteristics Adult equivalent 5.2 4.6 4.1 3.5 2.6 4.9 3.4 4.1 3.3 Dependency ratio 1.3 1.1 1.1 0.9 0.6 1.3 0.8 1.2 0.6 Dwelling characteristics Access to electricity (%) 14.3 22.3 26.6 36.3 52.9 17.5 39.3 10.4 64.0 Improved water source (%) 58.9 63.1 65.7 71.9 80.8 60.6 73.2 52.6 92.4 Improved toilet facility (%) 13.8 16.7 22.2 26.5 37.2 14.4 29.0 11.0 43.8 Livelihoods Agricultural land ownership (%) 91.4 88.8 84.6 78.0 63.7 90.5 74.8 93.2 60.3 Land use right (%) 89.5 89.3 87.2 83.1 72.1 89.4 80.3 91.6 71.0 Livestock ownership (TLU) 3.0 3.3 3.2 2.8 2.0 3.1 2.6 3.4 1.9 Non-farm enterprise ownership (%) 9.0 11.1 14.3 17.2 20.1 9.6 17.3 10.7 21.3 Agriculture main income source (%) 83.1 80.4 75.6 67.1 49.9 82.1 63.5 84.9 46.6 Non-agriculture main income source (%) 13.4 16.2 20.5 27.9 43.7 14.5 31.4 11.7 46.9 Labor market Head employed (%) 89.2 90.7 90.6 90.1 89.8 90.0 90.1 90.8 89.1 Members employed (%) 66.3 70.2 70.7 71.0 74.0 68.2 71.9 71.8 69.8 Members employed in agriculture (%) 81.2 77.0 71.8 63.6 46.6 79.9 60.1 82.4 43.5 Members employed in industry (%) 3.8 4.3 4.6 5.6 6.1 3.8 5.5 2.9 7.8 Members employed in service (%) 15.0 18.7 23.6 30.8 47.4 16.3 34.4 14.7 48.7 Shocks Market shock (%) 14.9 15.2 16.6 17.1 17.9 15.2 17.0 17.4 15.5 Health shock (%) 5.8 5.4 5.0 5.2 5.5 5.7 5.3 5.9 4.8 Food shortage (%) 8.2 5.8 4.9 4.6 3.5 7.1 4.3 6.7 2.9 Days of conflict within 20km radius 8.1 9.2 10.9 12.0 15.2 8.6 12.8 7.6 17.0 Years with PDSI < 0 24.7 24.0 23.5 22.9 22.3 24.4 22.9 24.0 22.3 Proximity to public services, Km Food market 7.0 6.1 5.7 5.1 4.1 6.6 4.9 7.0 3.2 Livestock market 10.0 8.6 8.8 8.0 6.7 9.4 7.7 10.4 5.2 All-weather road 7.1 4.8 4.8 3.5 2.6 6.2 3.6 6.2 1.7 Bank 22.2 19.1 19.6 16.5 13.1 20.9 16.1 23.2 9.7 Primary school 2.5 2.2 2.3 2.2 1.7 2.3 2.0 2.7 1.3 Secondary school 11.5 9.7 9.9 8.9 7.0 10.7 8.5 11.8 5.5 Health post, clinic, center 8.3 6.9 6.6 6.0 4.8 7.8 5.7 8.1 3.8 Safety nets and aid SafetyNet program (PSNP) 15.8 10.9 10.7 8.3 4.5 13.8 7.6 12.4 5.0 Humanitarian aid 15.0 15.0 13.2 10.5 6.6 14.9 10.0 15.0 6.4 Other aid 7.6 6.6 6.6 4.8 3.5 6.9 4.9 6.5 4.1 Source: World Bank staff calculations using HoWStat 2021. PDSI stands for Palmer Drought Severity Index. 98 ETHIOPIA POVERTY AND EQUITY ASSESSMENT Table A.3.5. Inequality indicators and Pyatt’s inequality decomposition HCES 2015/16 HoWStat 2021 Gini by residency National 0.33 0.29 Urban 0.38 0.29 Rural 0.28 0.27 Gini by region Afar 0.33 0.26 Amhara 0.34 0.27 Oromia 0.3 0.29 Somali 0.26 0.26 Benishangul-Gumuz 0.34 0.3 SNNP 0.32 0.3 Sidama 0.31 0.26 Gambella 0.34 0.31 Harari 0.35 0.26 Addis Ababa 0.36 0.27 Dire Dawa 0.37 0.27 Percentile ratios p90/p10 4.25 3.63 p90/p50 1.99 1.94 p10/p50 0.47 0.53 p75/p25 1.9 1.96 Generalized Entropy Indices GE(a) GE(-1) 0.2 0.15 GE(0) 0.18 0.13 GE(1) 0.2 0.14 GE(2) 0.29 0.17 Atkinson indices A(0.5) 0.09 0.07 A(1) 0.16 0.13 A(2) 0.29 0.23 Pyatt’s Inequality decomposition Between 28.6 20.9 Overlap 14.7 19.6 Within 56.7 59.5 Source: Authors’ estimates based on HCES 2015/16 and HoWStat 2021. Notes: Pyatt’s inequality decomposition follows methodology from Pyatt, G. (1976). 99 ANNEX 3: ADDITIONAL DESCRIPTIVE STATISTICS AND REGRESSION RESULTS Figure A.3.1. Income growth by source of income, 2016 vs 2022 237% 169% 97% 126% 73% 91% 43% 11% -23% -35% -43% -53% -26% -11% -37% -75% Crop Livestock Agricultural wage Non-agricultural wage Self-employment Transfers Other income Total income Crop Livestock Agricultural wage Non-agricultural wage Self-employment Transfers Other income Total income Nominal income Real income (in 2022 prices) Source: Authors’ estimates based on ESPS 2015/16 and 2021/22. Notes: Income growth calculated only for those households with income. Table A.3.6. Regression of conflict exposure on 2019 characteristics (1) Event Days (2) Event Days (3) Event Days (4) Event Days Log Consumption 1.53*** 0.31 0.06 -0.27 (0.45) (0.40) (0.41) (0.38) Female Household Head 0.15 0.02 (0.53) (0.54) Household Size 0.00 -0.04 (0.12) (0.11) Share Finished Primary -0.30 -1.37 (0.91) (0.83) Landed -1.42 -0.29 (1.17) (1.08) Zone Capital - Other Urban -3.96* (2.17) Rural -4.88** (1.97) Past Event Days 0.36*** (0.11) Past Fatalities -0.03 (0.02) Region Fixed Effects No Yes Yes Yes N 6,768 6,768 6,768 6,768 Sample Mean 6.38 6.38 6.38 6.38 Source: Authors' estimates based on ESS 2019. Notes: All household characteristics are measured in ESS Wave 4 which was completed in August 2019. The outcome is the number of days with conflict events (battles, remote violence, and violence against civilians) within 20km between September 2019-December 2022 (ACLED). Past conflict covers a period of the same length through August 2019. Population weights applied. Standard errors clustered at the EA level. * p < 0.10, ** p < 0.05, *** p < 0.01 100 ETHIOPIA POVERTY AND EQUITY ASSESSMENT ANNEX 4: CORRELATES OF JOB QUALITY FOR THE URBAN WAGE EMPLOYED We estimate a pooled cross-sectional regression model using a generalized specification of the following form: JQI = α + βX + γJob + ρRegion + δYear + εi (1) where JQI measures the level of job quality of an urban wage worker and α is the intercept. The term X is a vector of individual characteristics such as age gender and education, and β, the associated vector of parameters to be estimated. The term Job is a vector of job characteristics such as job sector, occupational skill levels and detailed subsectors of employment, while γ is the associated vector of parameters to be estimated. We also control for Region and Year fixed effects, along with the respective ρ and δ vectors of parameters to the estimated. Here, ε is the residual term capturing anything else that the model does not take into account. Table A.4.1. Correlates of job quality in urban areas (1) (2) (3) All Men Women Female -0.086*** Youth -0.146*** -0.141*** -0.135*** Educational attainment (Ref=No education) Less than primary -0.175*** -0.239*** -0.103*** Completed primary -0.104*** -0.153*** -0.069*** Completed secondary 0.060*** 0.046*** 0.046** Higher education 0.426*** 0.409*** 0.434*** Job sector (Ref=Public) Private -0.347*** -0.318*** -0.379*** Occupational skill level (Ref=Low) Medium -0.015* 0.106*** -0.134*** High 0.531*** 0.590*** 0.498*** Sector of employment (Ref=Primary agriculture) Agricultural services 0.089* 0.105* 0.049 Manufacturing 0.156*** 0.154*** 0.127*** Mining and extractives 0.147** 0.190*** 0.029 Construction 0.076*** 0.091*** 0.014 Public utilities 0.190*** 0.173*** 0.162*** Trade 0.103*** 0.097*** 0.096** Transport 0.272*** 0.258*** 0.249*** Hospitality 0.022 -0.072* 0.052 Information & communication 0.097*** 0.088** 0.105** Finance & real estates 0.337*** 0.323*** 0.333*** Professional/technical 0.200*** 0.190*** 0.210*** Administrative & support 0.110*** 0.085** 0.119** Government 0.211*** 0.204*** 0.214*** Education & health 0.125*** 0.144*** 0.079** Arts, entertainment & recreation -0.019 -0.052 -0.004 Community/household activities -0.260*** -0.176*** -0.324*** Extraterritorial organizations 0.117*** 0.128*** 0.087 Region (Ref= Oromia) Tigray 0.040*** 0.049*** 0.025 101 ANNEX 4: CORRELATES OF JOB QUALITY FOR THE URBAN WAGE EMPLOYED (1) (2) (3) All Men Women Afar -0.062*** -0.052** -0.077*** Amhara 0.002 -0.004 0.013 Somali 0.058*** 0.081*** 0.014 Benishangul-Gumuz 0.042** 0.028 0.058** SNNPR -0.029*** -0.037*** -0.017 Gambella 0.127*** 0.135*** 0.096*** Harari 0.060*** 0.060*** 0.051** Addis Ababa 0.129*** 0.121*** 0.130*** Dire Dawa 0.021 0.040* -0.016 Year of survey (Ref=2010) year=2011 -0.167*** -0.205*** -0.109*** year=2014 -0.030** -0.048*** 0.0004 year=2015 -0.001 -0.024 0.031 year=2016 -0.077*** -0.103*** -0.039* year=2018 -0.110*** -0.137*** -0.071*** year=2020 -0.164*** -0.237*** -0.067*** Constant 1.354*** 1.319*** 1.305*** Observations 101,664 59,730 41,934 Notes: Authors’ estimations using Ethiopia UEUS 2010-2020. The table reports estimate for working age adults between 15 and 64 years in wage employment. * p < 0.10, ** p < 0.05, *** p < 0.01. 102