ETHIOPIA POVERTY ASSESSMENT Harnessing Continued Growth for Accelerated Poverty Reduction OVERVIEW © 2020 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW, Washington, DC 20433 Telephone: 202-473-1000; Internet: www.worldbank.org SOME RIGHTS RESERVED This work is a product of the staff of The World Bank 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 This work is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) http:// creativecommons. org/licenses/by/3.0/igo. Under the Creative Commons Attribution license, you are free to copy, distribute, transmit, and adapt this work, including for commercial purposes, under the following conditions: ATTRIBUTION—Please cite the work as follows: World Bank (2020). Ethiopia Poverty Assessment: Harnessing Continued Growth for Accelerated Poverty Reduction. 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 re-use a component of the work, it is your responsibility to determine whether permission is needed for that re-use 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. PHOTO CREDITS Cover World Bank/Sylwia Pecio page 25 Tim Weg/Shutterstock.com Page 4 World Bank/Binyam Teshome page 27 Tim Weg/Shutterstock.com Page 9 World Bank/Sylwia Pecio page 30 World Bank/Gelila Woodeneh page 10 World Bank/Chris Terry page 33 World Bank/Sylwia Pecio page 14 World Bank/Candace Gebre page 34 World Bank/Chris Terry page 17 World Bank/Chris Terry page 36 World Bank/Sylwia Pecio Cover and interior design: Hanna S. Chang ETHIOPIA POVERTY ASSESSMENT Harnessing Continued Growth for Accelerated Poverty Reduction OVERVIEW HARNESSING CONTINUED GROWTH FOR ACCELERATED POVERTY REDUCTION 1 Acknowledgements This report was prepared by a core team consisting of Tom The report benefitted from inputs from officials of the Plan- Bundervoet (Task Team Leader, Poverty and Equity Practice), ning and Development Commission, the Ministry of Finance, Arden Finn (Co-Task Team Leader, Poverty and Equity GP), and the Central Statistics Agency. The team is indebted to Shohei Nakamura (Economist, Poverty and Equity GP), and the Central Statistics Agency (CSA) for making available Berhe Mekonnen Beyene (Economist, Poverty and Equity the datasets on which this report is based and to Professor GP), under the overall guidance of Pierella Paci (Practice Tassew Woldehanna for sharing the final consumption ag- Manager, Poverty and Equity GP), Nataliya Mylenko (Pro- gregate. Comments and inputs by participants at the 16th gram Leader), and Carolyn Turk (Country Director, AFCE2). Ethiopian Economics Association International Conference Nora Dihel, Zerihun Kelbore and Samuel Mulugeta contrib- (July 2018) are greatly appreciated. uted to the macro-economic analysis presented in the intro- The team thanks the report’s peer reviewers for thoughtful duction. Thomas Sohnesen (Consultant, Poverty and Equity inputs and comments. The peer reviewers were Kathleen GP), Lucian Bucur Pop (Senior Social Protection Specialist, Beegle (Lead Economist, GTGDR), Ruth Hill (Lead Economist, Social Protection and Jobs GP), Daisy Demirag (Consultant, Poverty and Equity GP), Johan Mistiaen (Program Leader, Social Protection and Jobs GP), Judith Sandford (Consul- AFCE2), Margaret Grosh (Senior Advisor, GSJD1), and tant, Social Protection and Jobs GP), and Abu Yadetta (Se- Alemayehu Seyoum Tafesse (Senior Research Fellow, IFPRI). nior Social Protection Specialist, Social Protection and Jobs The team also thanks Emily Schmidt and Mekamu Kedir at GP) contributed to Chapter 6 on social protection. Chapter IFPRI for making available spatial data used in this report. 7 was prepared in collaboration with Maude Cooper, Saori Iwamoto and Rewa Misra (Georgetown University) with guid- ance from Jacobus Cilliers (Georgetown University). Manex Bule Yonis (Consultant, Poverty and Equity GP) contributed to data analysis for several chapters in this report. Contents 1. Introduction______________________________________________________________________________ 5 2. Poverty in Ethiopia_______________________________________________________________________ 7 2.1 Trends in Poverty Reduction_____________________________________________________________ 7 2.2 Non-monetary Dimensions of Poverty Reduction__________________________________________ 11 2.3 Characteristics of the Poor_____________________________________________________________ 13 2.4 Geographic Distribution of the Poor_____________________________________________________ 15 2.5 Characteristics of the Extreme Poor_____________________________________________________ 16 2.6 Agriculture and Poverty________________________________________________________________ 18 3. Special Topics in Poverty Analysis_______________________________________________________ 20 3.1 Household Poverty Dynamics and Economic Mobility _____________________________________ 20 3.2 Urban Poverty in Ethiopia ______________________________________________________________ 23 3.3 Poverty and Social Protection___________________________________________________________ 28 3.4 Inequality of Opportunity in Ethiopia_____________________________________________________ 30 4. Perspectives on Continued Poverty Reduction___________________________________________ 33 HARNESSING CONTINUED GROWTH FOR ACCELERATED POVERTY REDUCTION 3 4 ETHIOPIA POVERTY ASSESSMENT OVERVIEW 1. Introduction The poverty headcount in Ethiopia is falling. The share was mainly driven by services, which explained 42 percent of of the population below the national poverty line decreased the expansion in GDP between 2011 and 2016. Agriculture from 30 percent in 2011 to 24 percent in 2016. This decrease contributed 25 percent to the growth in GDP over the same was achieved in spite of the fact that the 2015/16 survey was period, while industry accounted for 34 percent (Figure 1). conducted during the severe El-Nino drought. The observed Economic output has been shifting from agriculture to reduction in poverty is robust to the use of alternative defla- industry, but employment shares have not changed as tors. The fall in the poverty headcount was driven mainly by much. Agriculture’s share in GDP decreased from 46 percent Ethiopia’s strong economic growth over that period. in 2011 to 38 percent in 2016, while the share of industry Ethiopia has continued to pursue its developmental rose from 14 percent to 24 percent over the same period. state model. The approach is characterized by a strategic Services remained fairly constant at 39 to 40 percent of GDP. focus on agriculture and industrialization, coupled with large Changes in sectoral employment shares have been less dra- public infrastructure investments facilitated by heterodox matic. Agriculture’s share of employment modestly decreased macro-financial policies. Economic growth in Ethiopia has to 74 percent in 2013 (year of the latest Labor Force Survey), remained exceptionally strong. GDP grew at an average rate down from 78 percent in 2005. Most workers shifted towards of almost 10 percent per year between 2011 and 2016, re- services (employment share of 18 percent in 2013) and, to a sulting in a 39 percent increase in per capita GDP levels. This lesser extent, industry (share of 9 percent in 2013). Figure 1 SERVICES  AND INDUSTRY HAVE BEEN DRIVING GROWTH Sectoral contribution to GDP growth Agriculture Industry Services GDP 12 11.3 10.6 10.3 10.4 10.2 10 5.2 3.2 3.6 8.1 8 4.7 6.3 6 4.4 1.9 3.9 4.2 4 3.1 2.1 2 4.1 2.9 3.1 2.6 2.2 2.3 0.9 0 2010/11 2011/12 2012/13 2013/14 2014/15 2015/16 Source: National Planning Commission HARNESSING CONTINUED GROWTH FOR ACCELERATED POVERTY REDUCTION 5 Ethiopia’s performance in converting economic growth This poverty assessment focuses on the evolution to poverty reduction has been relatively weak. Between of poverty and other social indicators in Ethiopia be- 1997 and 2016, a one percent increase in per capita GDP tween 2011 and 2016. It uses data from a variety of sourc- was associated with a 0.33 percent decrease in poverty es, mainly the Household Consumption and Expenditure rates. Among a sample of comparators, only Mozambique Survey (HCES), the Welfare Monitoring Surveys (WMS), the and Rwanda had a lower “poverty-elasticity of growth” (Figure Ethiopia Socioeconomic Survey (ESS) and the Demographic 2). The semi-elasticity, however, which measures the percent- and Health Surveys (DHS), to observe trends in monetary age point change in poverty for a one percent change in per and non-monetary dimensions of living standards and to ex- capita GDP, was lowest in Ethiopia. Between 1997 and 2016, amine the drivers of these trends, with a special focus on a one percent increase in per capita GDP was accompanied government programs. The aim of the poverty assessment by only a 0.19 percentage point reduction in poverty, less is to provide policymakers and development partners with than a quarter of Tanzania’s semi-elasticity. Research shows information and analysis that can be used to improve the that countries with low levels of initial development tend to effectiveness of their poverty reduction and social programs. have lower growth-poverty elasticities. It is possible that the baseline level of development in Ethiopia was so low that growth has increased incomes of the poor but not yet to the level of pushing them above the poverty line. Figure 2 BUT  THE RATE AT WHICH GROWTH HAS TRANSMITTED TO POVERTY REDUCTION IN ETHIOPIA IS AMONG THE LOWEST Growth elasticity of poverty for Ethiopia and its comparators, 1997 - 2016 Tanzania Burkina Faso Uganda -0.19 Ethiopia -0.33 Mozambique Rwanda -1.00 -0.90 -0.80 -0.70 -0.60 -0.50 -0.40 -0.30 -0.20 -0.10 0.00 Semi-elasticity Elasticity Note: The elasticities are estimated by taking the first and last years between 1997 and 2016 when data on poverty is available. Myanmar is not included because poverty data is available only for 2015. Source: World Development Indicators. World Bank staff calculations. 6 ETHIOPIA POVERTY ASSESSMENT OVERVIEW 2. Poverty in Ethiopia 2.1 TRENDS IN POVERTY REDUCTION Poverty declined much more in urban areas than in The depth and severity of poverty also decreased, but rural areas. Although the poverty headcount fell from 30 again the gains accrued mostly to the urban population. percent in 2011 to 24 percent in 2016, these gains were not The depth of poverty, which measures how far on average spread evenly throughout the country. Economic expansion the consumption of the poor is from the poverty line (also translated into strong household consumption growth for the called the “poverty gap”), modestly dropped at the national urban population, but the impact for the rural population was level. However, when examined in more detail, this change very small in comparison. Consumption of urban households represents a sharp decrease in the depth of poverty for ur- grew at 6 percent per year on average, while for rural house- ban areas and a weak one in rural areas. Similarly, the sever- holds it was less than 1 percent. As a result of increased ity of poverty, another measurement of the average poverty consumption, the poverty rate for urban Ethiopia decreased gap for the poor that attaches more weight to the poorest, from 26 percent in 2011 to 15 percent in 2016, a drop of 11 showed no improvement for rural areas despite a strong percentage points (Figure 1) In contrast, poverty decreased decrease for urban areas. At the regional level, poverty se- in rural areas by only 4 percentage points in that time, from verity decreased strongly in Afar, Benishangul-Gumuz and 30 percent to 26 percent. Consequently, poverty became Gambella, and in the city administrations (Addis Ababa and somewhat more concentrated in rural areas. Close to 90 Dire Dawa). Conversely, poverty severity in Harari increased percent of the poor lived in rural areas in 2016, compared to sharply from a low base. a rural population share of 80 percent. Figure 3 POVERTY  DECREASED IN BOTH RURAL AND URBAN AREAS Poverty headcount rate based on the national poverty line, 2011 and 2016 2011 2016 35 29.6 30.4 30 Percentage poor 25.7 25.6 25 23.5 20 14.8 15 10 5 0 National Urban Rural Source: HCES, 2011; 2016. World Bank staff calculations. HARNESSING CONTINUED GROWTH FOR ACCELERATED POVERTY REDUCTION 7 The bottom 10 percent have not experienced any real Weaker growth in rural areas has meant that the rela- consumption growth since 2005. A previous World Bank tive contribution of urban areas to poverty reduction poverty assessment showed that consumption among the is increasing. One third of poverty reduction from 2011 to bottom 15 percent actually contracted between 2005 and 2016 was attributable to urban areas, more than doubling 2011, both in rural and urban areas. This pattern continued its share in the previous period (Figure 6). Population shifts into the latest period for the rural population but not for the from rural to urban areas did not contribute to poverty re- urban population. For the country as a whole, growth for the duction because rural-to-urban migration, while increasing, bottom 15 percent was not statistically different from zero is still relatively weak. After a period of stagnation in per from 2011 to 2016, in contrast to the top of the distribution capita GDP between 2000 and 2005, strong and sustained where growth rates reached a maximum of just under 6 per- economic growth began to drive robust consumption growth cent per year. This uneven pattern was driven by rural areas, and poverty reduction at the household level, mainly in urban in which the bottom 20 percent of the consumption distri- areas (Figure 7). The contribution of urban areas to poverty bution experienced zero or negative consumption growth reduction is expected to further increase in coming years as (Figure 4). In contrast, growth across the urban consumption improved rural education levels and land scarcity speed up distribution was always above 3 percent per year, even for rural-urban migration and the ongoing reforms create more the poorest, and was increasingly strong towards the upper job opportunities in the urban private sector. end of the distribution (Figure 5). Because the bulk of the Ethiopian population live in rural areas, the national pattern of growth closely resembles the rural pattern. Figure 4 WELFARE  OF THE POOREST Figure 5 …WHILE  GROWTH WAS 20 PERCENT IN RURAL STRONG ACROSS THE AREAS DID NOT INCREASE URBAN WELFARE BETWEEN 2011 AND 2016… DISTRIBUTION Average annual growth rates of rural Average annual growth rates of rural consumption by percentile between consumption by percentile between 2011 and 2016 2011 and 2016 Rural Urban 8 8 −2 −1 0 1 2 3 4 5 6 7 −2 −1 0 1 2 3 4 5 6 7 Annual mean growth rate (%) Annual mean growth rate (%) 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Percentiles Percentiles 95% confidence bounds 95% confidence bounds Source: HCES 2011; 2016. World Bank staff calculations. Source: HCES 2011; 2016. World Bank staff calculations. 8 ETHIOPIA POVERTY ASSESSMENT OVERVIEW Figure 6 THE  CONTRIBUTION OF Figure 7 SUSTAINED  ECONOMIC URBAN AREAS TO POVERTY GROWTH HAS LIFTED REDUCTION IS INCREASING MANY URBAN HOUSEHOLDS OUT Rural-urban decomposition of the reduction in poverty OF POVERTY GDP per capita and urban poverty rates, 0 40 300 Poverty rate (%) Change in poverty headcount 2000-2016 -1 35 -2 36.9 250 35.1 0 40 30 300 Poverty rate (%) Per capita GDP (2000=100) -3 Change in poverty headcount -1 200 -4 35 25 -2 36.9 250 25.7 35.1 -5 -3 30 20 150 200 -6 -4 25 15 25.7 -7 -5 20 150 100 14.8 -6 10 -8 15 -7 100 50 5 14.8 -9 -8 10 50 -10 -9 5 0 0 -10 2005-2011 2011-2016 0 2000 2005 2011 0 2016 2005-2011 2011-2016 2000 2005 2011 2016 Rural Urban Population shift Poverty rate (urban) GDP per capita Rural Urban Population shift Poverty rate (urban) GDP per capita Note: The population shift effect estimates the change in poverty due to a shift in population from rural to urban areas. Source: HCES 2011; 2016. World Bank staff calculations. Source: HCES 2011; 2016. World Bank staff calculations. HARNESSING CONTINUED GROWTH FOR ACCELERATED POVERTY REDUCTION 9 Inequality increased slightly. The Gini coefficient rose than rural consumption and grew much more quickly. The from 0.30 in 2011 to 0.33 in 2016, but remains relatively low share of total inequality that can be explained by differences in comparison to other countries in the region. The rise in in welfare between urban and rural areas doubled to reach inequality is mainly due to an increasing disparity between ru- 29 percent in 2016 (Figure 8). ral and urban areas. Urban consumption was already higher Figure 8 INEQUALITY  INCREASED DUE TO THE INCREASING GAP BETWEEN URBAN AND RURAL AREAS Decomposition of the Gini coefficient into a between rural-urban component and a within-component 20.1 15.0 Relative contribution to Overlap Gini (%) 55.9 65.0 Within Between 29.1 14.9 2011 2016 Source: HCES 2011; 2016. World Bank staff calculations. 10 ETHIOPIA POVERTY ASSESSMENT OVERVIEW 2.2 NON-MONETARY DIMENSIONS OF POVERTY REDUCTION Non-monetary dimensions of household welfare im- (cement, bricks, wood, etc…) nonetheless remained rare. proved alongside poverty reduction. The share of Access to water and sanitation services improved consider- households with a television, mobile phone, or refrigerator ably as well. In 2016, only 35 percent of Ethiopians used an increased from 2011 to 2016, as did the use of improved unimproved water source, down from 46 percent in 2011. housing materials and access to electricity (Figure 10). The Nonetheless, a third of households still do not have access share of Ethiopian households living in a home with an im- to any toilet facility. proved roof rose dramatically, but having improved walls Figure 9 ASSET  HOLDINGS AND LIVING CONDITIONS IMPROVED BETWEEN 2011 AND 2016 Selected household characteristics in 2011 and 2016, % of households with asset/characteristic A. Household Durables B. Housing/Energy 70 70 70 70 70 70 70 70 60 60 60 60 60 60 2016 2016 2011 2011 60 60 2016 2016 2011 2011 50 50 2016 2016 2011 2011 50 50 2016 2016 2011 2011 50 50 50 50 Percent Percent Percent Percent 40 40 40 40 Percent Percent Percent Percent 40 40 40 40 30 30 30 30 30 30 30 30 20 20 20 20 20 20 20 20 10 10 10 10 10 10 10 10 0 0 0 0 0 0 0 0 ile ile e ne ge gene ne rigfrig or or cra r im im rt rt or or rt rt ot ot le le e e ov ov ll ll ov ov ll ll oooo vrov of of s/ of of ce ce s/ s/ nt nt ce ce el el nt nt tri tri eleele ty ty tri tri r or ity ity r cr c g g ng ng M MelTel on on on on atlo to pr pr wawa pr pr wawa cl cl fo fo kinkin ot ot ca ca M M yc yc ec ec o o icirici ho fri friho ho ReRe ratrat p ro ro il ro ro AcAc tiletile eme AcAc to to meme f c c ki ki r ra si si si isi cy cy obob on ty ri ri c c AnAn l ca ty ty o o oooo M M php ReRe p p M M al al vi vi Im Im ed ed Im Im ed ed nie e Im Im ed ed r: r: ed ed t ci ct ct v ce ss ss ce ce or or aa El El s t s t ctrc ci ci co co r: r: cem le le e m ilie s s e e ov ov bv Te Te oboe im ile oooo /e pr pr AnfA Fl Fl pro Fl Fl tilets T Im Im ec ec fo C. Sanitation El El city D. Water 100 100 100 100 100 100 100 100 80 80 38.3 38.3 32.3 32.3 80 80 35.3 35.3 80 80 38.3 38.3 32.3 32.3 80 80 46.3 46.3 35.3 35.3 46.3 46.3 Percent Percent Percent Percent 60 60 60 60 11.7 11.7 Percent Percent Percent Percent 60 60 60 60 11.7 11.7 9 9 40 40 43.5 43.5 52.2 52.2 40 40 9 10.19 17 17 10.1 17 17 40 40 43.5 43.5 52.2 52.2 40 40 10.1 10.1 23.3 23.3 21.1 21.1 20 20 4.6 4.6 20 20 21.1 21.1 20 20 1 1 20 20 23.3 23.3 4.6 4.6 10.5 10.5 1 10.8 10.8 1 14.3 14.3 2.8 10.5 10.5 2.8 2.8 2.8 10.8 10.8 11.1 11.1 0 0 0 0 11.1 11.1 14.3 14.3 0 2.8 2.8 0 2.8 2.8 0 0 2011 2011 2016 2016 2011 2011 2016 2016 2011 2011 2016 2016 2011 2011 2016 2016 toilet toilet Flush Flush Improved Improved pit latrine pit latrine PipedPiped waterwater - compound - compound waterwater PipedPiped - outside - outside compound compound Flush toilet Flush Composting Composting toilet toilet toilet Improved pit pit Improved Unimproved latrine pit Unimproved latrine pit latrine latrine Piped water Piped Protected - water Protectedcompound - compound well/tubewell PipedPiped water well/tubewell Protected - outside water Protected spring - outside spring compound compound Composting Composting OtherOther toilet unimproved toilet unimproved Unimproved No toilet toilet toilet Unimproved No toilet pit facility latrine pit facility latrine Protected Protected improved OtherOther well/tubewell well/tubewell improved Protected Protected Unimproved spring spring sources Unimproved sources unimproved OtherOther unimproved No toilet toilet toilet facility No toilet facility OtherOther improved improved Unimproved Unimproved sources sources Source: DHS, 2011; 2016 HARNESSING CONTINUED GROWTH FOR ACCELERATED POVERTY REDUCTION 11 Human development indicators for the country im- percent of births taking place in a health facility and less than proved but nonetheless remain low. The rate of delivery 40 percent of children being fully immunized. Both net enrol- in a health facility more than doubled, while the share of fully ment in primary school and the completion rate went up, as immunized children increased by 14 percentage points and well as gross enrolment in secondary school. These numbers stunting rates fell by 6 percentage points (Figure 10). Infant also remained weak however, with only one in three people and child mortality rates decreased accordingly. These de- aged 15 to 24 having completed primary school. velopment indicators nonetheless remained low, with only 26 Figure 10 CHILDREN’S  MORTALITY DECREASED, AND THEIR HEALTH IMPROVED BETWEEN 2011 AND 2016 Selected health and education variables for children in 2011 and 2016 A. Child and infant mortality rates Immunization, health facility delivery and stunting B.  100 100 90 80 80 # of deaths per 1000 70 60 60 Percent 50 44.4 40 40 38.5 38.4 30 26.2 24.3 20 20 9.9 10 0 0 Infant Child Under ve Health facility Fully Stunted mortality mortality mortality rate delivery immunized children rate rate children 2011 2016 2011 2016 Source: DHS, 2011; 2016 80 71.8 70 62.4 60 50 Percent 40 32.7 30.7 27.6 30 21.5 20 10 0 Net primary Primary school Gross school completion secondary enrolment (15-24) school enrolment 2011 2016 12 ETHIOPIA POVERTY ASSESSMENT OVERVIEW 2.3 CHARACTERISTICS OF THE POOR Ethiopia has a traditional poverty profile. As in most household has 1.4 dependents for every working-age adult, low-income countries in the world, the poor tend to live in compared to 1.1 for non-poor households. Given the limited rural areas in large households with high dependency rates, resources available to poor households, having dependents and that are headed by an older and little-educated person. creates relatively more financial strain. Households with less The poor mainly engage in agriculture and casual labor for than 0.5 dependents per working-age adult have an average their livelihoods, are relatively isolated from key infrastructure, poverty rate of 16 percent, while households with 2 or more and have more limited access to basic services. Rapid urban dependents for working-age adult have poverty rates in ex- poverty reduction in Ethiopia has meant that the poor are cess of 30 percent (Figure 11). While dependency rates have increasingly concentrated in rural areas. decreased since 2011, the poorest are lagging. The depen- dency rate in the bottom quintile remained constant at 1.5 Poor households tend to have more children and high- dependents per working-age adult, reflecting the persistently er dependency ratios. In both rural and urban areas, the high total fertility rate of 6.4 for this cohort. average poor household includes about 1.5 more members than non-poor households. In rural areas, the average poor Figure 11 CHARACTERISTICS  OF THE POOR The poor are largely uneducated A.   ive in households with high dependency rates B. L Poverty rate by education of household head Poverty rate by household dependency rate 35 35 31.3 31.3 30 28.4 28.4 30 30 30 25.7 25.7 25 25 22.1 22.1 25 25 22.7 22.7 20.1 20.1 20 20 20 20 15.6 15.6 13.4 13.4 12.8 12.8 15 15 15 15 10 10 10 10 4.9 4.9 5 5 3.3 3.3 5 5 0 0 0 0 0.5-1 0.5-1 0-0.5 0-0.5 1.5-2 1.5-2 2- 1-1.5 1-1.5 2- pl d on im n y ary y da y ry nd y co y y ry ry ar imar on ar co ar ar ar pr tio da da i at pr m m se nd nd Dependency Dependency rate rate e ca on on uc i se ri te pr e o et u p c et ec ec ec ed se pllete et te m oe pl e s t-s -s e pl le o e otse st N co N omp m p om t C le Pe Po Com m pl p co o m om c In C In In co C In Source: HCES 2011; 2016. Farming Self-employed Farming World Bank staff nonfarm Self-employed calculations. Wage nonfarm Wage nonfarm nonfarm 30 30 26.6 26.6 100 100 4.7 4.7 6.3 25 25 6.3 5.7 5.7 10 10 11.5 11.512.7 12.711.4 11.4 20.1 20.1 28.7 28.7 15.3 15.3 20 20 Poverty rate Poverty rate 80 80 15.9 15.9 15 15 13.9 13.9 60 60 25.7 25.7 10 10 40 83.8 40 83.8 81 81 82.9 82.9 74.7 74.7 5 5 20 20 45.5 45.5 0 0 0 0 Female Male Male Male Male Female Female Female Q1 Q1 Q2 Q2 Q3 Q3 Q4 Q4 Q5 Q5 Urban Urban Rural Rural Quintiles Quintiles of consumption of consumption expenditures expenditures HARNESSING CONTINUED GROWTH FOR ACCELERATED POVERTY REDUCTION 13 A lack of education is strongly correlated with pover- a household head that completed primary school doubled ty. Poverty is highest among households with a head who from 10 percent in 2011 to 21 percent in 2016, although few never went to school and decreases with each extra level of rural Ethiopians attain that much education. The significant education (Figure 11). A complete cycle of primary education increase in education returns in rural areas may reflect the seems to have the biggest returns in terms of poverty reduc- adoption of agricultural technology such as improved seeds, tion. Households with a head who has completed primary fertilizers and herbicides. Such innovations are mainly taken school have poverty rates that are less than half of those up by more educated farmers. of households with heads who never went to school. There The poor tend to live in more remote and badly-con- is also a strong relationship between household welfare and nected areas. Relative to the rural non-poor, the rural poor educational attainment of the household’s youth, suggesting live further away from roads, health facilities, and urban cen- a high degree of intergenerational transmission of poverty. ters. For instance, 57 percent of the poor live more than three Returns to education in terms of consumption have kilometers away from an all-weather road, compared to 45 increased, both in urban and in rural areas. Whereas percent of the non-poor.1 Similarly, 43 percent of the poor live in urban areas post-secondary education had a 45 percent more than three kilometers away from a health facility, com- return in terms of household consumption in 2011 (rela- pared to 34 percent of the non-poor. In general, better con- tive to a household with an uneducated head), it had a 64 nectivity is correlated with lower poverty. While connectivity percent return in 2016. This may reflect the increase in real has increased in recent years, large parts of rural Ethiopia hourly wages in urban Ethiopia as well as the modest de- remain poorly connected. crease in unemployment. In rural areas, the reward for having 1  HCES, WMS, 2016. 14 ETHIOPIA POVERTY ASSESSMENT OVERVIEW 2.4 GEOGRAPHIC DISTRIBUTION OF THE POOR Poverty rates do not vary much by region, but do vary by agro-ecological classification based on altitude, rainfall, and agro-ecological zone. In contrast to many countries, there predominant livelihoods. The drought-prone lowlands, which is no strong regional concentration of poverty in Ethiopia. Dif- include the eastern and southern parts of Oromia and the ferences in consumption levels between regions explained southern parts of SNNPR (but do not include pastoral areas a mere two percent of total inequality in 2016. Regional of Afar and Somali), had the highest poverty rate in 2016, contributions to overall poverty thus largely reflect regional at 32 percent. The depth and severity of poverty was also population shares (Table 1). However, some disparities in highest in this zone. In contrast, the drought-prone highlands poverty are evident when considering the “five Ethiopias”– an had the lowest poverty rate (21 percent). Table 1 POVERTY  RATES, POVERTY SHARES, AND POPULATION SHARES BY REGION AND AGRO-ECOLOGICAL ZONE, 2016 POVERTY RATE POVERTY SHARE POPULATION SHARE BY REGION Tigray 27.0% 6.6% 5.8% Afar 23.6% 1.9% 1.9% Amhara 26.1% 25.5% 23.0% Oromia 23.9% 38.3% 37.8% Somali 22.4% 5.5% 5.8% Benishangul Gumuz 26.5% 1.3% 1.1% SNNPR 20.7% 17.5% 19.9% Gambella 23.1% 0.4% 0.4% Harari 7.1% 0.1% 0.3% Addis Ababa 16.8% 2.6% 3.6% Dire Dawa 15.4% 0.3% 0.5% BY AGRO-ECOLOGICAL ZONE Moisture-reliable highlands 23.6% 58.5% 58.4% Drought-prone highlands 20.8% 19.9% 22.5% Moisture-reliable lowlands 25.4% 4.7% 4.3% Drought-prone lowlands 31.7% 7.5% 4.7% Pastoral areas 21.9% 6.9% 7.4% Note: Poverty share denotes the contribution of the region to overall poverty. Source: HCES, WMS, 2016. World bank staff calculations. The persistent notion that the pastoral areas of Ethi- health, and other social indicators tend to be much worse in opia are the poorest of the country is not confirmed the pastoral regions (Afar and Somali). Other regions with a by the data. The pastoral areas, which cover most parts of significant pastoral population, such as Oromia, also tend to Afar and Somali region, have typically had average or below perform below average on human development outcomes, average poverty rates. Nonetheless these areas are signifi- reflecting the difficulty of providing public services in low-den- cantly lagging on human development outcomes. Education, sity areas with mobile populations. HARNESSING CONTINUED GROWTH FOR ACCELERATED POVERTY REDUCTION 15 2.5 CHARACTERISTICS OF THE EXTREME POOR The characteristics of the extreme poor are like those extreme poor are more likely to be located in SNNPR and of the poor, only more severe. The extreme poor are those Somali regions, in contrast to the fairly even distribution of in the bottom 10 percent of the consumption distribution. the poor throughout all regions. As discussed above, this cohort have not experienced any Extremely poor households experienced some im- real consumption growth since 2005. Whereas the poor are provements in non-monetary welfare, despite not ex- characterized by large households, high dependency rates, periencing any real growth in consumption. The share and a lack of education, the extreme poor have still larger of the extreme poor living in a house with an improved roof households, even higher dependency rates, and even less increased sharply from a low base, from 1 percent in 2011 to education. Compared to the poor, extreme poor are also 10 percent in 2016 (Figure 12). Net primary school enrolment more likely to be rural and more isolated from markets. The reached 66 percent in 2016, up from 56 percent in 2011. extreme poor are also lagging behind in Ethiopia’s fertility Despite this increase in enrollment, completion of primary transition. While the Total Fertility Rate decreased substan- school remained unchanged at about 17 percent of the 15 tially in the third, fourth, and fifth wealth quintiles, it decreased to 24 age cohort. only modestly in the second quintile (a decrease of 0.6 in 16 years) and did not change at all in the bottom quintile. The Figure 12 LIVING  CONDITIONS OF THE BOTTOM 10 PERCENT IMPROVED BETWEEN 2011 AND 2016 Trends in selected indicators from the bottom 10 percent, 2011 and 2016 2011 2016 100 80 60 % 40 20 0 Children Improved Improved Net primary Primary school fully water roof school completion immunized (%) source (%) material (%) enrolment (%) (15-24, %) Source: WMS 2011, 2016; DHS, 2011, 2016. World Bank staff calculations. 16 ETHIOPIA POVERTY ASSESSMENT OVERVIEW Some development indicators for the extreme poor improved water source. Only 7 percent of children were born showed no improvement. The share of extremely poor in a health facility, although this represents a sharp increase children stunted or wasted remained stable at a high level, from a low base (Table 2). These weak indicators highlight despite an improvement in reported food shortages. In 2016, the enormous efforts that lie ahead in making access to basic a mere one in four extremely poor children had received all public services less dependent on location and wealth. basic vaccinations, and about one in four had access to an Table 2 A  MIXED PICTURE OF PROGRESS FOR THE EXTREME POOR Means of selected variables for the bottom 10 percent, 2011 and 2016 MEAN 2011 2016 DIFFERENCE Household head literate (%) 32.6 35.7 3.1 Household size 7.3 7.2 -0.1 Dependency ratio 1.38 1.41 0.03 Cumulative fertility 3.6 3.7 0.1 Births in health facility (%) 2.7 7 4.3 Children stunted (%) 47.1 45.2 -1.9 Children wasted (%) 13.9 14.4 0.5 Average annual household expenditures per AE (2015 Birr) 3,827 3,762 -65 Daily calorie intake per AE 1,633 1,777 144 Food shortage (%) 31 20 -11 Source: WMS, 2011; 2016. World Bank staff calculations. The food gap is only calculated for those households who reported a food shortage. Differences in bold are statistically significant. HARNESSING CONTINUED GROWTH FOR ACCELERATED POVERTY REDUCTION 17 2.6 AGRICULTURE AND POVERTY Most Ethiopians are engaged in agriculture. Overall, Although the contribution of urban areas to overall poverty 70 percent of household heads in Ethiopia have their main reduction is increasing, the rural nature of Ethiopia means occupation in agriculture or livestock, but this increases to that the agricultural sector remains crucial. Poverty fell fastest 84 percent for households in the bottom quintile (Figure 13). in the zones that had the strongest agricultural growth be- Beyond agriculture, poor households engage in non-farm tween 2000 and 2016.2 Two thirds of the reduction in poverty self-employment but do not have the education or skills nec- from 2011 to 2016 can be explained by agriculture (Figure essary to access non-farm wage employment. This pattern is 14). Notably, this is less than the share it contributed over widespread throughout the income distribution, and the oc- the previous period. Changes within the services sector ac- cupational structure of the bottom 80 percent of households counted for about 15 percent of poverty reduction between is fairly similar to the bottom 20 percent. The top quintile is 2011 and 2016. The role of structural transformation – shifts remarkably different however, with more than half of house- in the population out of agriculture and into manufacturing holds having a main occupation outside of agriculture. or services – was very limited over the last period, reflecting the familiar “growth without structural transformation” narra- Most of the poverty reduction took place in the agri- tive for Ethiopia. Nonetheless, the importance of this factor is cultural sector, due to its dominance of the economy. likely to increase in the future. Figure 13 AGRICULTURE  REMAINS THE MOST COMMON OCCUPATION, ESPECIALLY FOR THE POOR Main occupation of household head, by quintile, 2016 Farming Self-employed nonfarm Wage nonfarm 100 4.7 5.7 6.3 10 90 11.5 12.7 11.4 15.3 28.7 80 70 60 25.7 50 40 83.8 81 82.9 74.7 30 20 45.5 10 0 Q1 Q2 Q3 Q4 Q5 Quintiles of consumption expenditures Source: HCES, WMS, 2016. 2  Zones are administrative entities below the regions and above the woredas. 18 ETHIOPIA POVERTY ASSESSMENT OVERVIEW Figure 14 AGRICULTURE  REMAINS THE LARGEST CONTRIBUTOR TO POVERTY REDUCTION Sectoral decomposition of changes 2005 to 2016 Change in poverty headcount 2011-2016 2005-2011 -10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 Pop. Shift Other Serv. Const. Manu. Agri. Source: HCES 2005, 2011, 2016. World Bank staff calculations. Cash crops were more important for poverty reduction cash crops there was significant movement in prices, with than cereals. In the latest period (2011-2016), a strong shift the price of khat relative to the price of coffee rising by more was observed away from the production of cereal crops in than 40 percent in 10 years (Figure 15). West and East Ha- favor of cash crops. The expansion of cash crop cultivation rarge zones, where khat is cultivated, both experienced very was especially strong for coffee, khat, oil seeds such as ses- sharp drops in poverty between 2011 and 2016. Despite the ame and noog (guizotia abyssinica), as well as kocho. The immediate economic gains, a focus on cash crops raises the zones that showed particularly strong shifts towards cash exposure of farmers to market fluctuations, given that the crops were in Oromiya (Jimma, West Hararge and East Ha- impact of crop price changes on revenue is about twice as rarge), SNNP (Sidama) and Harari. Within the category of large for cash crops as for cereals. Figure 15 THE  RELATIVE PRICE OF KHAT OVER COFFEE JUMPED BY 40 PERCENT OVER A DECADE The ratio of khat prices to coffee prices between 2007 and 2017 2 Price ratio (2007=1) 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 07 08 09 10 11 12 13 14 15 16 17 Year Source: Figure provided by IFPRI. HARNESSING CONTINUED GROWTH FOR ACCELERATED POVERTY REDUCTION 19 3. Special Topics in Poverty Analysis 3.1 HOUSEHOLD POVERTY DYNAMICS AND ECONOMIC MOBILITY Although the poorest households were no better-off in poverty, who remained trapped in poverty, and the reasons 2016 compared to 2012 (in terms of consumption), in why. Crucially, it also allows us to determine the profiles of most cases these were not the same households. The those who were chronically poor. cross-sectional nature of the Household Consumption and Poorer households experienced better growth rates. Expenditure Survey (HCES) data considered in the last chap- For the rural and small-town population, comparing each in- ter makes it impossible to assess whether the households dividual household’s consumption in 2012 to its consump- that were poor in 2016 were the same ones that were poor tion in 2016 shows that the households that were initially the in 2011. However, longitudinal surveys allow for the analysis poorest experienced the fastest rate of consumption growth of the consumption dynamics of individual households. The (Figure 16). Although the consumption growth of the baseline Ethiopian Socioeconomic Survey (ESS) is a longitudinal sur- poor was impressive in percentage terms, it was modest in vey that interviewed a representative sample of households real ETB terms and for many of the poor it was not enough in rural areas and small towns of Ethiopia in 2012, and then to lift them over the poverty line. In contrast to the poor, the interviewed them again in 2014 and 2016. Larger towns and consumption of rural and small-town households that were cities were also covered in the last two rounds. Exploiting the initially in the upper part of the distribution actually contracted. time dimension of the data allows us to see who escaped Figure 16 THE  BASELINE POOR GREW FASTEST BETWEEN 2012 AND 2016 Growth rate of consumption conditional on decile in 2012 (non-anonymous quasi-GICs) 40% 34% 30% Annual percentage change 20% 11% 10% 8% 2% 1% 0% -1% -4% -5% -10% -8% -14% -20% Poorest 2 3 4 5 6 7 8 9 Richest 2012 consumption deciles Source: Own calculations from ESS 2012, 2014 and 2016. 20 ETHIOPIA POVERTY ASSESSMENT OVERVIEW Figure 17 CHRONIC  POVERTY IS Figure 18 MOST  OF THE CHRONIC HIGHER IN SNNPR AND POOR LIVE IN SNNPR AND TRANSITORY POVERTY AMHARA IN AMHARA Regional shares of each poverty Chronic and transitory poverty over the category 2012-2016 period, rural areas and small 100% towns 100% Chronic poor 80% Chronic poor 80% 60% 60% Transient poor Transient poor 40% 40% 20% 20% Never poor Never poor 0% 0 10 20 30 40 50 60 0% Rural and Amhara Oromia SNNPR Tigray Others 0Amhara 10 20 30 40 50 60 Rural and small towns Amhara Oromia SNNPR Tigray Others Oromia SNNPR Tigray Others small towns Chronic poor Transient poor Never poor Amhara Oromia SNNPR Tigray Others Chronic poor Transient poor Never poor Source: ESS 2012, 2014, 2016. World Bank staff calculations. Source: ESS 2012, 2014, 2016. World Bank staff calculations. Most of the longer-term poverty in Ethiopia is transi- Chronically poor households tend to be larger and tory in nature, but a considerable share of households have fewer resources, but are more likely to benefit are trapped in chronic poverty. 16 percent of people in from government assistance. Relative to the transitory rural areas and small towns were chronically poor over the poor and the never poor, the chronic poor have more chil- 2012 to 2016 period, and 31 percent experienced transito- dren, higher dependency rates, less land per adult, fewer ry poverty (Figure 17).3 Taken together, almost half of that assets and less education. Interestingly, chronically poor population experienced at least one spell of poverty in that households were not substantially more remote than house- time, reflecting the high extent of consumption variability and holds that were never poor, as measured by distance to the vulnerability in rural Ethiopia. Chronic poverty was mainly nearest road or population center of more than 20,000 peo- concentrated in SNNPR, where 30 percent of households ple. As would be expected with efficient targeting, the chron- were considered to be chronically poor from 2012 to 2016. ically poor are more likely to benefit from the government’s Most of the transitory poor were found in Amhara Region social protection programs. (Figure 18). 3 Chronic poverty is defined as those households whose average consumption expenditure over all three rounds of ESS was below the poverty line. Transitory poverty means the household’s average consumption expenditure over the three rounds of ESS was above the poverty line, but the household was poor in at least one round. HARNESSING CONTINUED GROWTH FOR ACCELERATED POVERTY REDUCTION 21 There was a large amount of mobility as measured by and households living in the drought-prone lowlands were consumption expenditure, both upwards and down- more likely to fall into poverty. wards. Transitions into poverty were far more likely in rural The analysis of the longitudinal data confirms many of areas than in small towns or urban areas. 26 percent of the the conclusions of the analysis of the cross-sectional initially non-poor population in rural areas had fallen into pov- data. Both the ESS and HCES show that (1) chronic pover- erty by 2016, compared to only 14 percent in towns and 4 ty is mainly concentrated in SNNPR and transitory poverty percent in cities.4 The same holds true in reverse, with upward in Amhara, (2) households in pastoral areas have done rela- mobility higher in towns and cities than in the rural hinterland. tively well, and (3) households in the drought-prone lowlands Although 58 percent of the rural population who were poor in have fared relatively poorly. Both datasets also show that 2012 had managed to escape poverty by 2016, this number female-headed households have higher consumption levels was 62 percent in towns and 69 percent in cities. Other fac- and were less likely to fall into poverty. Finally, they also agree tors associated with a higher probability of escaping poverty that urban areas performed better than rural areas between are a higher level of education of the household head and 2011 (or 2012) and 2016. living in pastoral areas (Figure 19). Male-headed households Figure 19 MORE  EDUCATED HOUSEHOLDS, HOUSEHOLDS HEADED BY WOMEN, AND PASTORALIST HOUSEHOLDS WERE MORE LIKELY TO EXIT POVERTY Probability of exiting poverty by baseline characteristics 80 70 Percentage 60 50 40 30 e 3 3 d nd e u. u. u. ay ra ia er P 3 d d t is PR al al W W W n n n N m ed ed ed ha th al gr la la la la M m PS ro N O P od n or gh gh w w Ti Am Fe o ry y io SN N O lo lo st ar N fo er hi hi a ns PS pa im nd e ne Ev ee e ne te bl bl Pr co ro d Ex lia Fr ro lia n −p Se la re −p re w ht e− ht e− Lo ug ur ug ur ro st ro st D oi D oi M M Household head Household Above average Below average Note: Dashed line is the average probability of exiting poverty of 57.91%. Source: ESS 2012, 2014, 2016. World Bank staff calculations. 4  “Initially” refers to the baseline poverty status in 2012. 22 ETHIOPIA POVERTY ASSESSMENT OVERVIEW 3.2 URBAN POVERTY IN ETHIOPIA Ethiopia is rapidly urbanizing from a low base. While 19.1 percent in 2016. Nearly 1 million people are added to its urbanization level is still among the lowest in Sub-Saha- the urban population every year. Ethiopia’s urban population ran Africa, the urban population has increased by 6.2 per- is projected to reach 42 million by 2032 and its population cent annually since 2011, much faster than rural population share to hit 30 percent by 2028.5 Urban population growth growth rate of 2.7 percent. Consequently, the share of Ethio- will take place mainly in small towns and secondary cities pians living in urban areas rose from 16.6 percent in 2011 to (Figure 20). Figure 20 SMALL  TOWNS AND SECONDARY CITIES WILL ACCOUNT FOR THE BULK OF URBAN POPULATION GROWTH Urban population trends and projections, 2007-2035 52.6 50 5.81 40 20.242 31.1 Less than 50,000 Population (million) 30 50,000 to 100,000 4.561 5.593 100,000 to 500,000 9.225 20 17.5 Addis Ababa 11.9 3.273 3.284 3.488 10 2.74 1.782 20.929 2.276 14.002 1.139 8.916 5.708 0 2007 2015 2025 2035 Source: Schmidt, E., Dorosh, P., Jemal, M.K., & Smart, J. (2018). Ethiopia’s spatial and structural transformation: Public policy and drivers of change. IFPRI/EDRI Strategy Support Program Working Paper 119. 5  World Bank. (2015b). Ethiopia Urbanization Review: Urban institutions for a middle-income Ethiopia. Washington, D.C.: World Bank. HARNESSING CONTINUED GROWTH FOR ACCELERATED POVERTY REDUCTION 23 Urban poverty has declined rapidly. The share of the with a larger population, except for Addis Ababa. The poverty urban population living below the national poverty line de- headcount ratios in towns of a population of less than 20,000 creased from 26 percent in 2011 to 15 percent by 2016. (20 percent) are higher than other bigger towns and cities. A Headcount ratios have declined by around 10 percentage third of Ethiopia’s urban poor population lives in these small points in Addis Ababa, major towns, medium towns, and towns. Given their large contribution to the total urban pop- small towns (Figure 21). Poverty gap and severity measures ulation, small and medium-sized towns accounted for over have also decreased across the board. The absolute number half of urban poverty reduction between 2011 and 2016 of poor also fallen, despite rapid urban population growth. (Figure 22). Population shifts across towns of different siz- Overall, urban poverty reduction between 2011 and 2016 es accounted for only a limited proportion of poverty reduc- has been robust and widespread. tion, since the population shares across Addis Ababa, major towns, medium towns, and small towns have changed only Poverty reduction was mainly driven by small and me- marginally since 2011, indicating limited inter-city migration. dium-sized towns. Poverty rates tend to be lower in cities Figure 21 STRONG  URBAN POVERTY REDUCTION ACROSS CITY SIZE Poverty trends by city size A. Poverty headcount ratio B. Poverty gap/severity 30 9 28.1 28.2 8 7.9 25 Poverty headcount ratio (%) 23.7 7 7.0 Poverty gap and severity 6.5 20.7 19.3 6 20 16.8 4.9 5 5.1 15 4.1 11.4 12.0 4 3.3 3.2 10 3 2.5 2.5 2.7 2.0 2 1.9 5 1.4 1.3 1 0.9 0 0 Poverty rate Poverty rate Poverty rate Poverty rate Severity Severity Severity Severity Gap Gap Gap AA Major towns Medium Small towns Gap towns AA Major towns Medium towns Small towns 2016 2011 2016 2011 Note: Major towns have populations greater than 100,000 (excluding Addis Ababa); medium towns have populations between 20,000 and 100,000; and small towns have populations less than 20,000. Source: HCES; 2016. World Bank staff calculations. 24 ETHIOPIA POVERTY ASSESSMENT OVERVIEW Figure 22 SMALL  AND MEDIUM TOWNS, TRADE AND AGRICULTURE, AND THE LOW- SKILLED ACCOUNTED FOR THE BULK OF URBAN POVERTY REDUCTION Contribution to urban poverty reduction, 2011-2016, percentage points Addis Ababa -2.4 Major towns -1.8 Medium towns -3.2 Small towns -2.9 Population shifts -0.5 -3.5 -3 -2.5 -2 -1.5 -1 -0.5 0 Note: the population shifts estimate the change in poverty due to a shift in population across towns (migration between towns). Source: HCES 2011; 2016. World Bank staff calculations. Although employment growth has been strong, private sector wage employment has contributed little to pov- erty reduction. 74 percent of the working-age population participated in the labor force in 2016, up from about 71 per- cent in 2011. Over the same period, unemployment decreased from 19.6 percent to 17.3 percent.6 The fact that unemploy- ment decreased even though more people are participating in the labor force indicates a solid pace of employment growth in urban Ethiopia. Female unemployment however remains high, with 1 in 4 women in urban Ethiopia unable to find a job. Real hourly wages increased by 10 percent from 2011 to 2016 in both the public and private sectors, which is not sur- prising given that private sector wages seem to be informally pegged to public sector wages. Nonetheless, wage levels in the private sector remain low and not sufficiently determined by their productivity. Real hourly wages increased most for the uneducated, which is consistent with the observation of strong urban poverty reduction among households headed by persons with low levels of education. 6  Urban Employment and Unemployment Surveys, 2011-2016. HARNESSING CONTINUED GROWTH FOR ACCELERATED POVERTY REDUCTION 25 Urban poverty reduction was mainly driven by house- This effect was especially important in raising consumption holds headed by self-employed workers. Returns to levels of the poorest households (Figure 24). The share of self-employment relative to wage-employment increased these urban households in which non-head household mem- across the welfare distribution between 2011 and 2016, bers engage in self-employment increased from 13 percent in except for the very poorest (Figure 23). Households with a 2011 to 19 percent in 2016. Although the returns to having self-employed head accounted for 53 percent of the reduction a self-employed household head were not particularly strong in urban poverty though they represent only 46 percent of the for poor urban households, take-up of self-employment by urban population. Furthermore, the reduction in poverty for non-head household members was a strong driver of con- households with a self-employed head was strongest when sumption changes over time. additional household members also took up self-employment. Figure 23 THE  PREMIUM OF SELF- Figure 24 TAKE  UP OF SELF- EMPLOYMENT OVER EMPLOYMENT WAS WAGE EMPLOYMENT MOST IMPORTANT FOR INCREASED OVER THE CONSUMPTION GROWTH URBAN CONSUMPTION OF THE POOREST URBAN DISTRIBUTION HOUSEHOLDS Returns to self-employment versus wage The effect of additional household employment in urban areas 2011-2016 members in self-employment in urban areas 2011 to 2016 .04 .05 .03 .04 Log difference Log difference .02 .03 .01 .02 .01 0 −.01 0 10 20 30 40 50 60 70 80 90 10 20 30 40 50 60 70 80 90 Consumption percentiles Consumption percentiles Premium: Self employment vs wage employment Effect of increasing share of self employment Source: HCES 2011, 2016. World Bank staff calculations. Source: HCES 2011, 2016. World Bank staff calculations. 26 ETHIOPIA POVERTY ASSESSMENT OVERVIEW Living conditions in medium and small towns have not The economic integration of rural migrants is better in changed much in recent years. Access to piped water is smaller towns than in the capital. Though still limited, ru- relatively good in Addis Ababa, covering nearly 90 percent ral to urban migration is expected to increase substantially of households. However, in major towns, medium towns, in the coming years. In Addis Ababa, recent migrants (less and small towns, only about 70 percent of households have than 3 years in the city) have substantially worse employment access to piped water and the situation has not changed outcomes relative to older migrants (between 3 and 10 years since 2011. Access to improved sanitation is much worse, in the city) and the resident population. However, as they stay with only a fifth of urban households having access to an in Addis longer, rural migrants increasingly find a chance to improved sanitation facility and no observable improvement work in public employment and private permanent jobs, and since 2011. Residents of small towns face the worst living their employment structure becomes more comparable to conditions, as the share of substandard housing in these 7 the resident population. In other major towns outside Addis, communities slightly increased from 2011 to 2016. Similarly, that economic integration of migrants is relatively smooth the share of small town households with access to improved compared to the capital city, and the employment status of sanitation and an improved solid waste management is still recent migrants is fairly similar to that of older migrants and less than 10 percent. On the other hand, the provision of the resident population. Social integration however seems electricity in medium and small towns is widespread, with to be more difficult. The education level of children of rural nearly 90 percent having access. migrants is substantially worse than that of the resident pop- ulation of the same age, even for migrants that have been in the city for long. 7 Defined as lack of access to piped water and improved sanitation, and overcrowding (more than three persons per room). HARNESSING CONTINUED GROWTH FOR ACCELERATED POVERTY REDUCTION 27 3.3 POVERTY AND SOCIAL PROTECTION The Productive Safety Net Program (PSNP) significant- are included in the PSNP, and hence there is no exclusion of ly contributed to poverty reduction. The PSNP provides the poor because of the selection of woredas. conditional (on work) or unconditional cash or food transfers The PSNP’s contribution to poverty reduction can be to targeted poor rural households during the lean season. At further increased. The analysis in this poverty assessment the zonal level, a one percent annualized increase in PSNP highlights three main issues. First, the number of beneficia- coverage was associated with a 0.1 percent annualized de- ries at regional level bears little relation to the prevalence of crease in the poverty rate. This implies that the PSNP was poverty or self-reported food insecurity, with the number of well targeted overall. Indeed, in 2016, 34 percent of PSNP beneficiaries exceeding the number of poor and food-inse- beneficiaries were in the bottom welfare quintile and over 60 cure people in certain regions and falling far short in others percent were drawn from the bottom 40 percent (Figure 25). (Figure 26). Second, geographical targeting (selection of PSNP targeting is progressive both in the highlands woredas) adds little to the PSNP’s targeting performance, and lowlands. While the share of beneficiaries that is drawn which is largely due to poverty and food-insecurity not being from the bottom quintile is substantially higher in the high- geographically concentrated in Ethiopia (Figure 27). Third, lands, the share that is drawn from the bottom 40 percent under-coverage remains an issue, with only 13 percent of is higher in the lowlands. Inclusion of households in the top Ethiopia’s poor covered by the PSNP in 2016. Better aligning quintile is higher in the highlands. On the regional level the regional caseloads to regional needs and expanding PSNP data show that, relative to what would be possible in case to more woredas but with smaller beneficiary numbers per of perfect targeting, Afar obtains the best targeting perfor- woreda are likely to increase PSNP’s coverage of the poor mance. This counter-intuitive outcome is explained by the and its contribution to poverty reduction. absence of first-stage woreda targeting – in Afar, all woredas Figure 25 MOST  OF PSNP BENEFICIARIES ARE IN THE LOWER CONSUMPTION QUINTILES Share of beneficiaries by quintile, 2011 and 2016 2011 2016 40 33 33.8 30 27 23.6 18.8 18.1 20 14.6 11.6 10 9.5 10 0 Q1 Q2 Q3 Q4 Q5 Quintiles of pre-transfer consumption expenditures Source: ESS 2012, 2014, 2016. World Bank staff calculations. 28 ETHIOPIA POVERTY ASSESSMENT OVERVIEW Figure 26 DISPARITIES  BETWEEN Figure 27 WOREDA  SELECTION DOES REGIONS’ SHARES IN NOT ADD MUCH TO OVERALL OVERALL CHRONIC TARGETING PERFORMANCE POVERTY AND REGIONS’ Decomposition of the targeting SHARES IN OVERALL PSNP differential into a “woreda-selection” CASELOAD and a “within-woreda household selection” component Regions’ contribution to national chronic Oromia poverty and Amhara Tigraycaseload, national PSNP SNNPR Somali Afar 2016 Selection of woredas Household selection 0.4% 2.2% 100% Oromia 2.9%SNNPR Tigray Amhara 7.0% Afar Somali 10 90% Selection of woredas 9 Household selection 80% 0.4% 2.2% 20.9% 100% Targeting differential 2.9% 7.0% 10 8 70% 90% 39.7% 80% 12.7% 20.9% 9 7 60% Targeting differential 70% 39.7% 8 6 50% 13.0% 12.7% 7 60% 5 8.4 40% 6 50% 33.6% 13.0% 4 4.9 23.7% 30% 5 8.4 40% 3 4.9 20% 33.6% 23.7% 4 30% 2 10% 21.7% 3 20% 18.2% 2 1 1.7 0% 10% 18.2% 21.7% 0.8 1 0 1.7 0% Share in national chronic Share in national PSNP 0.8 0 Poverty Food insecurity poverty Share in national bene ciaries chronic Share in national PSNP Poverty Food insecurity poverty bene ciaries Note: Targeting differential is the difference between coverage of the poor (food-insecure) and that of the non-poor (food-secure) Source: ESS 2012, 2014, 2016. World Bank staff calculations. Just like the PSNP, Humanitarian Food Aid (HFA) was lower, hinting at a recent exposure to a negative shock. There reasonably well-targeted in 2016. As per the design, are however substantial inclusion errors in HFA targeting, PSNP and HFA reach different types of households. PSNP with 30 percent of beneficiaries in the top two consumption households share many of the typical characteristics of the quintiles. These inclusion errors are due to HFA targeting in poor, including having few assets and livestock, being re- woredas where PSNP is not active. Further harmonizing the motely located, and having little education. HFA households PSNP and HFA is likely to improve performance and target- however are similar to the average household in rural areas, ing of the joint programs. but with the difference that their calorie intake is substantially HARNESSING CONTINUED GROWTH FOR ACCELERATED POVERTY REDUCTION 29 3.4 INEQUALITY OF OPPORTUNITY IN ETHIOPIA Inequality in welfare is partly the result of inequities in increased from 61 percent in 2011 to 71 percent in 2016. access to opportunities earlier in life. If, for instance, ed- There were also large increases in access to an improved ucation produces significant returns, an adult who had the water source and in the share of children that lived within opportunity to complete schooling when she was young will 5km of a health post. Despite these improvements, access have higher welfare levels than an otherwise comparable to electricity, health care, primary school completion and sec- person who did not have the opportunity to go to school. ondary enrollment remain low and unequal. Fewer than one The resulting inequality can be considered unfair because it in five children who were age-eligible for secondary school results from differences in circumstances early in life and not (15 to 18) were enrolled in 2016. The proportion of children differences in talent or hard work. In an equitable society, an with access to electricity improved marginally by about 3 per- individual’s circumstances at birth (such as being born a girl centage points but remained low at 20 percent in 2016. or a boy, in a rural or an urban area, in a poor or a better-off The rural poor are at risk of being left behind. The gaps household, etc.) should not influence the individual’s access between rural and urban areas and between richer and poor- to a set of important opportunities (such as education, health er households remain large. Access to electricity provides a care, clean water, etc.). Ensuring that there is equal opportu- striking example, reaching 90 percent for urban areas but re- nity of access to services is linked to more sustainable and maining less than 10 percent for rural areas (Figure 28). There inclusive economic growth. are also significant gaps in access to education when con- Equitable access to opportunities in Ethiopia is slow- sidering dimensions of wealth and geography. In 2016, half ly increasing but inequalities remain. The proportion of children aged 15 to 18 in households from the top con- of age-eligible children (7 to 14) enrolled in primary school sumption quintile had completed primary school, compared 30 ETHIOPIA POVERTY ASSESSMENT OVERVIEW Figure 28 LOCATION  AND HOUSEHOLD WEALTH ARE THE MAIN SOURCES OF INEQUITY IN ACCESS TO OPPORTUNITIES Coverage of basic opportunities, urban vs rural and poorest quintile vs richest quintile, 2016 100% 90% 80% Coverage of access 70% 60% 50% 40% 30% 20% 10% 0% Primary Completed Attending Electricity Improved Within 5km enrolled primary secondary water of health post Urban Rural Richest Poorest Note: Primary enrolled refers to children between 7 and 14 years of age. Completed primary and attending secondary refers to persons between 15 and 18-years-old. Access to electricity, improved water and a health post refers to children between 7 and 18-years-old. Source: WMS, HCES 2016. World Bank staff calculations. to less than 20 percent of children in the bottom consump- The effect of parental education on children’s education tion quintile. Almost 90 percent of age-eligible children living is still strong but was somewhat reduced. If children’s in Addis Ababa are enrolled in primary school, while the figure educational attainment is largely influenced by that of their for Somali is just over half. Secondary school enrollment also parents, it will contribute to a high transmission of poverty or demonstrates major disparities, with 40 percent of children prosperity across generations. The extent to which parental aged 15 to 18 in urban areas enrolled in 2016, compared to education influences children’s education in Ethiopia dimin- only 10 percent in rural areas. ished between 2011 and 2016, but remained strong and sig- nificant. Improvements in access to education were observed Analysis of the Human Opportunity Index confirms that for children with poorly educated parents in urban areas, as inequality of opportunity is mostly driven by differenc- well as children with relatively better educated parents in ru- es between rural and urban areas and differences in ral areas. There is still a large education-effect of living in an wealth. The household’s consumption quartile explained urban area, as the average urban child had completed 1.44 17 percent and 22 percent of inequality in secondary school more grades of education in 2016 than rural children, all else enrollment and access to an improved water source, respec- equal. Furthermore, although enrolment in primary school has tively. Two thirds of unequal access to electricity is explained become less dependent on parents’ education levels during by a household’s rural location, as is about half of unequal this period, enrolment in secondary school became even access to an improved water source. Differences by gender more dependent on parents’ education levels. play a relatively minor role in explaining inequalities across all seven outcome variables. HARNESSING CONTINUED GROWTH FOR ACCELERATED POVERTY REDUCTION 31 Household consumption levels have a large and grow- in secondary school for children in the bottom half of the dis- ing influence on whether the household’s children go to tribution was not different in 2011 and 2016. The effect of school. The probability of the poorest children being enrolled household welfare levels on children’s schooling is significant- in primary school did not significantly increase between 2011 ly stronger in rural areas, indicating greater scope for upward and 2016, but this increase was significant at the top of the mobility in urban areas (where access to schooling is far less distribution (Figure 29). Changes in the predicted probability dependent on household wealth). The implication of these re- of secondary school enrollment were concentrated in the top sults is that children of poor households and poorly-educated quartile of households, while the probability of being enrolled parents in rural areas are in danger of being left behind. Figure 29 ENROLMENT  IN SCHOOL BECAME MORE DEPENDENT ON HOUSEHOLD WELFARE BETWEEN 2011 AND 2016 School enrollment probabilities by household consumption, 2011 and 2016 Primary school enrollment Secondary school enrollment Probability of secondary enrollment .8 .8 Probability of primary enrollment .6 .6 .4 .4 .2 .2 0 0 6 7 8 9 10 11 6 7 8 9 10 11 Log of real consumption per adult equivalent Log of real consumption per adult equivalent 2011 2016 2011 2016 Source: Calculations from WMS 2010/11 and 2015/16, and HCES 2010/11 and 2015/16. 32 ETHIOPIA POVERTY ASSESSMENT OVERVIEW Perspectives on Continued Poverty 4.  Reduction The agricultural sector still holds the key for sustained force will try to make a living in the informal or semi-formal poverty reduction. Given its large share in employment and sector, in both wage employment and self-employment. The livelihoods, especially among the poor, the agricultural sector projected increase in the size of the agri-food sector as Ethi- will continue to drive national poverty reduction, though its opia urbanizes and urban incomes grow could be a large contribution will progressively decrease. The role of structural generator of employment for young people leaving the farm. transformation in promoting economic growth is still limited, Designing interventions and policies to boost the labor force and between 2000 and 2016 poverty fell fastest in areas that participation of women and youth could further increase the had the strongest agricultural growth. There is room to fur- labor market’s contribution to poverty alleviation. ther increase yields, mainly through promoting the use of im- The role of urban areas in poverty reduction will con- proved seeds, currently being used on only about 6 percent tinue to grow. Although Ethiopia is under-urbanized given of cultivated land. Although the recent shift towards cash its income level, urban areas of all sizes are growing sub- crop production and the rise in crop prices helped net pro- stantially and poverty in cities and towns is falling rapidly. The ducers, there are also potential losers from these changes. contribution of urban areas to poverty reduction will further Agricultural policy will need to be nimble enough to ensure increase in the coming years as rural to urban migration ac- that the effects of rising prices on vulnerable households are celerates and ongoing reforms create more job opportunities effectively mitigated. in the urban private sector. Given the close linkages between The agricultural sector in its present form will not be small towns and their surrounding rural hinterland, and the able to absorb the rapidly growing labor force. The lower skills requirements for jobs in these towns, investing in Ethiopian working-age population is projected to grow at small towns and removing any barriers to migration to those two million per year in the coming decade. The increasing towns holds significant promise for the creation of relatively scarcity of agricultural land in the highlands means that an low-skilled jobs and continued poverty reduction. In this re- ever-larger share of young people will not inherit enough land gard, it will also be important to better understand the chal- and will need to transition to livelihoods off the farm. Given lenges that rural migrants face in their new urban homes, the low education levels in rural areas, the bulk of the new- particularly with respect to access to public services and comers will not qualify for modern wage employment in the rights to formally establish and operate businesses. formal economy. This implies that most of the growing labor HARNESSING CONTINUED GROWTH FOR ACCELERATED POVERTY REDUCTION 33 Spreading the benefits of growth will also require in- performance and coverage. Going forward, the safety net vestment in services and infrastructure in small towns. should be flexible enough to scale up or down depending on Small towns are expected to add much of the urban popu- the particular state of national and local economies. lation in the future, and they will gain in importance as local Finally, improving the welfare of the poorest will require centers of demand and employment for the surrounding rural more investments in the human capital of children. In areas. Though small towns have been important for urban an ideal scenario, children of extremely poor households poverty reduction, access to key services and amenities is would accumulate more education and be able to move out not keeping up. Investments in small towns will be required and diversify into more productive activities, breaking the to improve living standards and to reduce migration pressure intergenerational transmission of poverty. This however is on the bigger cities. not taking place. Education gains between 2011 and 2016 Well-functioning and well-targeted safety nets will re- were concentrated among the children of better-educated main essential. In 2016, only 13 percent of Ethiopia’s poor parents in rural areas and less-educated parents in urban were covered by the PSNP. Increasing the coverage of the areas. The children of poor and poorly educated parents in poor will require either a scaling-up of the PSNP, improved rural areas lag on education, and the effect of household targeting to reduce the coverage of the non-poor, or a combi- wealth on child education has only strengthened since 2011. nation of both. Analysis suggests that scaling up the PSNP to Crucial investments need to be made in providing access to more woredas, while reducing beneficiary numbers per wore- services in rural areas, so that children born in rural house- da to help manage the fiscal implications, would likely increase holds are afforded the same opportunities as those born in the program’s contribution to poverty reduction. Harmonizing urban areas. More equal access to key opportunities will be the PSNP and Humanitarian Food Aid (HFA) and revisiting the needed to harness Ethiopia’s strong economic growth for first-stage geographical selection would likely improve both faster poverty reduction. 34 ETHIOPIA POVERTY ASSESSMENT OVERVIEW