ETHIOPIA REGIONAL POVERTY REPORT Promoting Equitable Growth for All Regions ETHIOPIA REGIONAL © 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 POVERTY REPORT 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 Promoting Equitable Growth for All Regions 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 Regional Poverty Report: Promoting Equitable Growth for All Regions. 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. Cover Photo: World Bank PROMOTING EQUITABLE GROWTH FOR ALL REGIONS 1 Contents Acknowledgements _________________________________________________________________________ 7 Executive Summary _________________________________________________________________________ 8 The Way Forward on Continued Poverty Reduction and Narrowing Regional Gaps_____________________ 11 Section I: Overview of Regional Poverty Changes in Ethiopia, 2011–2016___________ 13 Monetary poverty and welfare___________________________________________________________________ 16 Non-monetary welfare _________________________________________________________________________ 25 Inequality of opportunity________________________________________________________________________ 29 Public expenditure _____________________________________________________________________________ 33 The Way Forward on Poverty Reduction and Narrowing Regional Gaps_______________________________ 35 Policy recommendations________________________________________________________________________ 35 Section II: Regional Analysis_______________________________________________________________ 37 Tigray________________________________________________________________________________________ 38 Introduction________________________________________________________________________________ 38 Poverty and inequality_______________________________________________________________________ 39 Non-monetary welfare ______________________________________________________________________ 45 Inequality of opportunity_____________________________________________________________________ 49 Public expenditure__________________________________________________________________________ 50 Afar ______________________________________________________________________________________ 53 Introduction________________________________________________________________________________ 53 Poverty and inequality_______________________________________________________________________ 54 Non-monetary welfare_______________________________________________________________________ 61 Inequality of opportunity_____________________________________________________________________ 65 Public expenditure__________________________________________________________________________ 66 Introduction________________________________________________________________________________ 69 Amhara______________________________________________________________________________________ 69 Introduction________________________________________________________________________________ 69 Poverty and inequality_______________________________________________________________________ 70 Non-monetary welfare ______________________________________________________________________ 76 Inequality of opportunity_____________________________________________________________________ 80 Public expenditure__________________________________________________________________________ 81 PROMOTING EQUITABLE GROWTH FOR ALL REGIONS 3 Oromia______________________________________________________________________________________ 84 Harari_______________________________________________________________________________________ 162 Introduction________________________________________________________________________________ 84 Introduction_______________________________________________________________________________ 162 Poverty and inequality_______________________________________________________________________ 85 Poverty and inequality______________________________________________________________________ 163 Non-monetary welfare ______________________________________________________________________ 92 Non-monetary welfare _____________________________________________________________________ 168 Inequality of opportunity_____________________________________________________________________ 96 Inequality of opportunity____________________________________________________________________ 172 Public expenditure__________________________________________________________________________ 97 Public expenditure_________________________________________________________________________ 173 Somali______________________________________________________________________________________ 100 Addis Ababa________________________________________________________________________________ 176 Introduction_______________________________________________________________________________ 100 Introduction_______________________________________________________________________________ 176 Poverty and inequality______________________________________________________________________ 101 Poverty and inequality______________________________________________________________________ 177 Non-monetary welfare _____________________________________________________________________ 108 Non-monetary welfare _____________________________________________________________________ 183 Inequality of opportunity____________________________________________________________________ 112 Inequality of opportunity____________________________________________________________________ 187 Public expenditure_________________________________________________________________________ 113 Public expenditure_________________________________________________________________________ 188 Benishangul-Gumuz_________________________________________________________________________ 116 Dire Dawa__________________________________________________________________________________ 191 Introduction_______________________________________________________________________________ 116 Introduction_______________________________________________________________________________ 191 Poverty and inequality______________________________________________________________________ 117 Poverty and inequality______________________________________________________________________ 192 Non-monetary welfare _____________________________________________________________________ 122 Non-monetary welfare _____________________________________________________________________ 199 Inequality of Opportunity____________________________________________________________________ 126 Inequality of opportunity____________________________________________________________________ 203 Public expenditure_________________________________________________________________________ 127 Public expenditure_________________________________________________________________________ 204 SNNP_______________________________________________________________________________________ 130 References_________________________________________________________________________________ 207 Introduction_______________________________________________________________________________ 130 Poverty and inequality______________________________________________________________________ 131 Annex A_____________________________________________________________________________________ 209 Non-monetary welfare _____________________________________________________________________ 138 Sectoral decompositions of changes in poverty __________________________________________________ 209 Inequality of opportunity____________________________________________________________________ 142 Annex B_____________________________________________________________________________________ 210 Public expenditure_________________________________________________________________________ 143 Constructing the Human Opportunity Index for Ethiopia ___________________________________________ 210 Gambella___________________________________________________________________________________ 146 Introduction_______________________________________________________________________________ 146 Poverty and inequality______________________________________________________________________ 147 Non-monetary welfare _____________________________________________________________________ 154 Inequality of opportunity____________________________________________________________________ 158 Public expenditure_________________________________________________________________________ 159 4 ETHIOPIA REGIONAL POVERTY REPORT PROMOTING EQUITABLE GROWTH FOR ALL REGIONS 5 Acknowledgements This report was prepared by a core team consisting of The report benefitted from inputs from officials of the Planning Berhe Mekonnen Beyene (Economist, Poverty and Equity), and Development Commission, the Ministry of Finance, and Arden Finn (Economist, Poverty and Equity), Christina Wieser the Central Statistics Agency. The team is indebted to the (Senior Economist, Poverty and Equity) and Manex Bule Central Statistics Agency for making available the data- Yonis (Consultant, Poverty and Equity). As the report is a sets on which this report is based and to Professor Tassew complement to the national Ethiopia Poverty Assessment, Woldehanna for sharing the final consumption aggregate. there is some overlapping material, and the team would like The team thanks the report’s peer reviewers for thought- to acknowledge the contributions of Tom Bundervoet (Senior ful inputs and comments. The peer reviewers are Kebede Economist, Poverty and Equity), and Shohei Nakamura Feda (Senior Economist, Social Protection and Jobs), Nga (Economist, Poverty and Equity), who were core team mem- Thi Viet Nguyen (Economist, Poverty and Equity), and Hiroki bers of the national Poverty Assessment. The former also gave Uematsu (Senior Economist, Poverty and Equity). direct inputs to this report. The team benefited greatly from the inputs of the Ethiopia Poverty and Equity team including Wondimagegn Mesfin Tesfaye. This report was prepared un- der the overall guidance of Pierella Paci (Practice Manager, Poverty and Equity), Marius Vismantas (Program Leader, EAEDR), and Ousmane Dione (Country Director, AECE3). PROMOTING EQUITABLE GROWTH FOR ALL REGIONS 7 Executive Summary regions experienced significant reduction in rural poverty. The relatively more urbanized regions of Harari, Addis Ababa, and Inequality, measured by the Gini coefficient, increased in almost all regions from 2011 to 2016. The increase in Dire Dawa continued to experience strong poverty reduction inequality, generally a result of stronger growth for relatively Ethiopia’s economy has grown rapidly over the last welfare, inequality of opportunity and patterns in public despite their relatively lower poverty levels, underscoring the richer households, weakened the extent of poverty reduc- two decades, leading to significant national poverty expenditure. The main objective of this report is to create importance of urbanization for poverty reduction. The least tion. Of the predominantly rural regions, Tigray had the high- reduction, but there were significant differences in the regional poverty profiles of Ethiopia’s regions and better un- urbanized regions experienced the slowest poverty reduc- est inequality in 2016, with a Gini index of 39. This inequality pace and nature of poverty reduction across Ethiopia’s derstand why the regions experienced different poverty re- tion, with the exception of Somali. Oromia, Amhara, and hampered overall progress in poverty reduction. Strong rise eleven regions. Based on the national poverty line, poverty duction rates between 2011 and 2016. The study also aims Benishangul-Gumuz had urbanization rates in 2016 below in average consumption in Tigray was not shared equally, as decreased from 46 percent in 1996 to 24 percent in 2016 to analyze regional patterns and inequalities in non-monetary 20 percent and experienced very slow poverty reduction. On growth predominated for the relatively well-off households. nationwide. Poverty decreased in all the regions, particularly welfare, including health, education, ownership of consumer the other hand, more urbanized regions experienced faster Somali experienced a decrease in inequality between 2011 after 2005 when growth became more robust. Although re- durables, and access to basic services such as electricity than average poverty reduction. In the predominantly rural re- and 2016 and, with a Gini index of 26, had the lowest in- gional disparities remained fairly low, regional poverty rates and water. Furthermore, it discusses trends and allocation gions, only Afar and Somali, the regions with high initial pov- equality in 2016. Inequality has become increasingly deter- converged until 2011 only to begin diverging again in 2016. of public expenditure across regions, and attempts to deter- erty rates, experienced relatively stronger poverty reduction mined by differences between urban and rural areas rather mine how allocation of public funds affects regional variance compared to those with low initial poverty rates (Figure E1). than differences within regions, leading to a rise in the nation- Identifying the factors associated with quicker and in monetary and non-monetary welfare. Benishangul-Gumuz and SNNP had similar initial poverty al Gini coefficient. Inequality in terms of consumption rose more sustained poverty reduction can provide use- rates, but SNNP experienced much faster poverty reduction particularly in Gambella and Benishangul-Gumuz, muting ful insights into the mechanisms of poverty reduction In all regions, poverty declined at a much faster pace than more urban Benishangul-Gumuz. Amhara and Tigray poverty reduction. and how to accelerate it for lagging regions. This report in urban areas than in rural areas, exacerbating ur- had higher than national average poverty rates in 2011—with complements the national focus of the World Bank’s 2020 ban-rural differences from 2011 to 2016. All regions ex- Alongside increases in consumption, non-monetary Tigray also showing higher than national urbanization rates in Ethiopia Poverty Assessment by bringing the focus to region- cept Tigray, Somali, and Benishangul-Gumuz experienced welfare outcomes also improved, although notable 2016—but neither region experienced strong poverty reduc- al differences and trends. Each regional chapter in this report significant urban poverty reduction, but the reduction was differences across regions persist, and serious chal- tion between 2011 and 2016. may be read independently – the structure of these chapters particularly strong in urban Dire Dawa and Amhara. Only So- lenges remain. Primary school completion rates for youth, is the same, and covers monetary welfare, non-monetary mali and Southern Nations, Nationalities and People (SNNP) Figure E1 THE  POOREST AND MOST URBAN REGIONS EXPERIENCED THE Table E1 REGIONAL  MONETARY AND NON-MONETARY INDICATORS HIGHEST POVERTY REDUCTION BETWEEN 2011 AND 2016 NON-MONETARY MONETARY INDICATORS INDICATORS A. Poverty reduction and initial poverty level B. Poverty reduction and urbanization Pov. Urban Prim. school 50 Somali 50 Somali Pov. Pov. change Rural pov. pov. Gini completion Immunized Urbanization Dire Dawa Dire Dawa 45 45 reduction reduction 2011 2016 (ppt) 2016 2016 coeff. 2016 children 2016 rate 2016 Faster Faster Addis Addis 40 40 Rate of poverty reduction (%) Rate of poverty reduction (%) Tigray 31.8 27.0 -4.8 31.1 14.2 38.8 44.2 67.3 24.4 Ababa Ababa Harari Afar 36.1 25.6 -10.5 31.9 10.6 37.0 24.2 15.2 29.8 35 35 Afar SNNP Afar 30 SNNP 30 Amhara 30.5 26.1 -4.4 28.8 11.6 34.3 29.5 45.8 15.8 Gambella Gambella Oromia 28.7 23.9 -4.8 25.3 15.3 31.2 30.1 24.7 14.2 25 25 Somali 32.8 16.8 -16.0 16.3 19.4 26.2 18.6 21.8 19.1 20 20 Oromia Oromia Tigray B.-G. 28.9 26.5 -2.4 28.7 17.7 35.6 39.2 57.4 19.7 15 Tigray 15 Amhara Amhara SNNP 29.6 20.7 -8.9 21.9 14.4 32.7 32.4 46.9 15.2 10 10 Ben.-Gumuz Ben.-Gumuz Higher Higher Gambella 32.0 23.0 -9.0 26.4 16.6 36.0 45.4 41.1 34.0 5 5 poverty poverty Harari 11.1 7.1 -4.0 8.5 6.0 37.1 47.4 42.2 54.7 0 0 25 27 29 31 33 35 37 0 20 40 60 80 100 A.A. 28.1 16.8 -11.3 16.8 37.8 66.1 89.2 100.0 Poverty rate in 2011 (%) Urbanization rate in 2016 (%) Dire Dawa 28.3 15.4 -12.9 23.3 11.1 39.4 50.0 75.9 64.7 Note: The blue lines represent rate of poverty reduction between 2011 and 2016 at national level while the orange lines represent the National 29.6 23.5 -6.1 25.6 14.8 32.9 32.7 38.5 19.5 national poverty rate in 2011 and national urbanization rate in 2016 in Panel A and Panel B, respectively. Harari has been omitted from Figure E1 for presentational purposes, as it had by far the lowest poverty rate in 2011 (11.1%). Poverty in Harari fell by 36% (to 7.1%) Note: The 2016 poverty and inequality values, and urbanization rates don’t include households sampled from new zones in Afar and So- between 2011 and 2016. mali. Numbers in bold are those above national average. Source: HCES, 2011; 2016. World Bank staff calculations. Source: HCES, 2011, 2016 ; WMS; 2016; DHS, 2016. World Bank staff calculations. 8 ETHIOPIA REGIONAL POVERTY REPORT EXECUTIVE SUMMARY 9 for example, are still low in all regions; two-thirds of those Inequality of opportunity for children shows large re- THE WAY FORWARD ON CONTINUED POVERTY REDUCTION AND aged 15 to 24 completed primary education in the capital gional variations. Circumstances such as location, gender, city of Addis Ababa, while only one in five completed primary and household wealth play a role in determining primary school NARROWING REGIONAL GAPS education in Somali. The predominantly urban regions, partic- completion rates, secondary school enrollment rates, and ac- The regional poverty study shows that consumption Ethiopia’s rural safety net program has helped reduce ularly the capital, have better outcomes in general. Amongst cess to electricity. Within each region, the location of house- growth alone is not sufficient for continued poverty re- rural poverty. This calls for continued investment and fur- the mainly rural regions, Gambella performs best with respect holds generally explains these differences—that is, whether duction. The type of growth matters. While measures lead- ther improvement in the program to accelerate poverty re- to indicators such as education and child stunting. Tigray rural or urban—rather than level of wealth of the household. ing to continued broad consumption growth are a necessary duction in rural areas by (a) expanding it to regions with low performs best in indicators on health, including infant and Public expenditure per capita increased in all regions pre-condition, they need to couple with specific measures to coverage but high poverty rates, (b) improving targeting of under-five mortality, immunization of children, and health fa- between 2011 and 2016, but large variations between ensure consumption growth for the poorest, most vulnerable the poorest beneficiaries within regions, and (c) expanding cility delivery. Despite strong reduction in monetary poverty, regions persist. In 2016, Gambella had the highest expen- households, including those in rural areas. the number of woredas included in the program. the predominantly pastoral regions of Afar and Somali lag diture of 5,127 Birr per capita, while SNNP had the lowest in non-monetary welfare. This may signal a special need to About 80 percent of Ethiopians live in rural areas that Continued investment in human capital, especially ed- with 1,888 Birr per capita. This ranking is also reflected in focus on poverty reduction in pastoral regions, especially if depend on agriculture. The two city administrations of Ad- ucation, in rural areas is required. Investments in hu- spending per capita on selected “pro-poor” sectors (agri- human capital accumulation continues to lag other regions. dis Ababa and Dire Dawa, and the small, mostly urban region man capital will help narrow regional and rural-urban gaps culture and rural development, water resources, health, and of Harari, are exceptions. Given the largely agrarian nature of within regions, promote equality of opportunities for children Ownership of household durables, access to electrici- education), with higher spending often reflected in better out- Ethiopia’s rural economy, poverty reduction requires contin- irrespective of location and household wealth, and reduce ty, and access to improved water generally increased. comes. Gambella, for example, has the highest per capita ued effort and investment in improving agricultural practices medium and long-term poverty. Unsurprisingly, the predominantly urban regions of Addis spending on health and education, and this is mirrored in and value chains. Continued expansion of improved seeds Ababa, Dire Dawa, and Harari perform relatively better than all its strong performance in non-monetary outcomes. Somali, and fertilizers, as well as improving access to agricultural other regions. However, large differences prevail even among on the other hand, spent less on health and education than product markets, can help decrease poverty. Poor farmers the more rural regions. Access to electricity, for example, is most regions, and it also shows the lowest performance on require special attention, because they lack complementary 42 percent in Gambella but only 12 percent in Somali. human capital outcomes. endowments such as land, social-networks, or education, and are therefore less likely to benefit from productivity-based interventions. Another way to decrease rural poverty is to encourage off-farm employment. Specifically, this means investing in small towns and better integrating them with rural areas to boost rural growth. Investing in small town infrastructure and services continues the strong urban poverty reduction trend in Ethiopia, a large part of which has been attributed to small towns. Combined with relaxing constraints to rural-to urban migration, investing in small towns is likely to be a sustainable poverty reduction strategy. 10 ETHIOPIA REGIONAL POVERTY REPORT EXECUTIVE SUMMARY 11 SECTION I Overview of Regional Poverty Changes in Ethiopia, 2011–2016 Monetary poverty and welfare Non-monetary welfare Inequality of opportunity Public expenditure The Way Forward on Poverty Reduction and Narrowing Regional Gaps Policy recommendations 12 ETHIOPIA REGIONAL POVERTY REPORT SECTION I: OVERVIEW OF REGIONAL POVERTY CHANGES IN ETHIOPIA, 2011-2016 13 Ethiopia’s economy has grown rapidly over the last two decades, leading to significant national pov- Box 1  ata sources, consumption aggregation and poverty D erty reduction. Poverty decreased in all regions, particularly after 2005 when growth was more robust. measurement Regional poverty rates converged until 2011, and then began to diverge again, though regional disparities remained fairly small as of 2016. In 1996, poverty in the poorest region (Tigray) was 82 percent higher than Most of the analysis presented in this report is based on the two most recent rounds (2010/2011 and 2015/2016) of the in the least poor region (Somali) (56 percent versus 31 percent).1 In 2011, that gap between the poorest and twinned household living standard surveys (HCES and WMS). HCES covers monetary welfare, while WMS deals with least poor regions had fallen to 26 percent (36 percent in Afar versus 29 percent in Oromia). Poverty further non-monetary welfare. The sample is representative at the urban and rural levels for all regions, and Dire Dawa city ad- decreased in 2016 in all regions, although the rate of poverty reduction varied, resulting in a slight divergence ministration. In Addis Ababa there is no rural population, and the sample is representative at sub-city level. of poverty rates. The poverty and inequality analysis is based on consumption expenditure from HCES. Consumption of food and non- Identifying factors associated with quicker and more Some main findings include: food items is included, regardless of whether these items are purchased on the market, come from own production, or sustained poverty reduction can provide insights into were received as gifts. For own-consumption and gifts, the quantities consumed are valued at prevailing prices in the ● Urbanization has been critical for poverty reduc- enumeration area. Although consumption is expressed on an annual basis, the reference period used during data collec- the mechanisms of poverty reduction and how to ac- tion, and growth in consumption in urban areas was tion varies based on the nature of the consumption items. For example, information on food and food-related items was celerate it for lagging regions. This report complements much faster than in rural areas. The predominantly asked twice a week using the “last three days” and “last four days” as reference periods (households are visited twice the national focus of the World Bank’s 2020 Ethiopia Poverty urban regions experienced strong poverty reduction, during the HCES). For housing rent, durable goods, clothing, health and education expenditures, and so forth, the “last Assessment by bringing the focus to regional differences and even if their initial poverty levels where low, underscor- 3 months” and “last 12 months” were used as references. Imputed rent is included in the consumption aggregate.2 To trends. Each regional chapter in this report may be read inde- ing the importance of urbanization for poverty reduction capture the effect of seasonal variations, the data were collected over a 12-month span (Hamle 1 to Sene 30, i.e. July 8 pendently – the structure of these chapters is the same and between 2011 and 2016. In most of the regions pover- to July 7), by randomly allocating sampled households to different months. covers monetary welfare, non-monetary welfare, inequality of ty reduction largely took place in urban areas, meaning opportunity and patterns in public expenditure. that urban-rural differences were exacerbated during To adjust for price variations across time and space, spatial and temporal price deflators are used. First, nominal con- The main objective of this report is to profile pover- the 2011 to 2016 period. In addition, those regions with sumption is adjusted for price differences across reporting levels, by using the spatial deflators provided by FDRE (2012, ty in each of Ethiopia’s regions and better understand initial high poverty rates, for example Afar and Somali, 2017). Second, spatially-deflated consumption levels are expressed in December prices (December 2010 and December why they experienced different poverty reduction rates experienced stronger poverty reduction. 2015), by using the food and non-food Consumer Price Indexes provided by the Central Statistical Agency. The food and between 2011 and 2016. The study also aims to analyze non-food Consumer Price Indexes are also used to bring the December 2010 consumption expenditure (2010/11 HCES) ● Urban-rural differences translated into a higher to December 2015 prices. Finally, to adjust for variations in household size and composition, consumption expenditure is regional pattens and inequalities in non-monetary welfare, in- overall inequality, as measured by the Gini coefficient divided by the officially-used adult-equivalent scales, which are based on calorie requirements and vary by age and sex. cluding health, education, ownership of consumer durables, in Ethiopia. Inequality is increasingly determined by dif- This exercise should result in a consumption aggregate that can be compared through space and time. The consumption and access to basic services such as electricity and water. ferences between urban and rural areas, rather than dif- aggregates used in this poverty assessment are the official ones used by the National Planning Commission. Furthermore, it discusses trends and allocation of public ex- ferences within them. Increase in inequality weakened penditure across regions, and attempts to determine how potential poverty reduction in most regions. The poverty rates presented in this report are based on the national poverty line. The poverty line is based on a food bas- much the allocation of public funds affects regional variance ket that is required to achieve the minimum daily calorie requirement – 2,200 kilo-calories per adult in Ethiopia – and ad- in monetary and non-monetary welfare. ● Non-monetary welfare improved between 2011 justed upwards to include non-food consumption. The food basket was determined in 1996 and has not been changed and 2016, although differences across regions are since. Accordingly, the poverty line in 1996 was 1,075 Birr per adult equivalent per year in 1996. In 2011, the poverty The analysis in this report draws on a number of differ- notable. The predominantly urban regions, particular- line was updated by costing the items in the original food basked at prevailing prices and doing a similar adjustment ent data sources from the 2011 to 2016 period. These ly the capital city Addis Ababa, performed better on for non-food consumption. The updated poverty line was 3,781 Birr per adult equivalent per year (in December 2010 include the Household Consumption Expenditure Survey non-monetary outcomes. The predominantly pastoral prices). For the most recent poverty measurement the 2011 poverty line was inflated using the GDP deflator, resulting (HCES), Welfare Monitoring Survey (WMS), and Demo- regions of Afar and Somali lag in non-monetary out- in a poverty line of 7,184 Birr per adult equivalent per year in December 2015 prices. The World Bank’s 2020 Ethiopia graphic and Health Survey (DHS). The report also uses data comes, even though they made some of the fastest Poverty Assessment tested the sensitivity of poverty patterns to the choice of deflator and found that poverty patterns from the Ministry of Finance (MoF) to analyze regional public progress in reducing monetary poverty. Long-term pov- at national and regional levels hardly changed when alternative deflators are applied. spending patterns and composition. erty reduction may be difficult to maintain in these areas, This chapter first presents an overview of the regional especially if human capital accumulation continues to All temporal comparisons in this poverty report exclude a number of zones in Somali and Afar regions. In 2011, the analyses on monetary welfare, non-monetary welfare, lag all other regions. HCES covered only two zones in Afar and three in Somali. In 2016, the coverage was improved – five zones and eight inequality of opportunity, and public spending. It sum- zones were covered respectively in Afar and Somali. While the 2016 data are more representative of the two regions, to marizes the main results emerging from the detailed regional maintain comparability, when comparison is made between 2011 and 2016, households sampled from the new zones analyses presented in section two of this report, and highlights are excluded from the analysis. policy directions to further reduce poverty and narrow gaps between regions and between urban and rural areas. Sources: FDRE;2012, 2017; Central Statistical Agency; 2018; World Bank; 2020. 1 This comparison excludes the two city administrations (Addis Ababa and Dire Dawa) as well as the small region of Harari which is 2 Imputed rent was calculated by the Central Statistics Agency (CSA) and was included in the consumption aggregate that was shared also predominantly urban. with the World Bank team. 14 ETHIOPIA REGIONAL POVERTY REPORT SECTION I: OVERVIEW OF REGIONAL POVERTY CHANGES IN ETHIOPIA, 2011-2016 15 MONETARY POVERTY AND WELFARE Table 2 MEDIAN HOUSEHOLD CONSUMPTION INCREASED IN ALL REGIONS EXCEPT AFAR AND AMHARA Average household consumption increased in all re- Despite adverse circumstances related to the 2015/16 Regional median annual consumption per adult equivalent in December 2015 prices gions except Somali, driven mostly by urban areas. 3 El Nino drought, poverty continued to decrease overall Harari experienced the highest consumption growth (in- between 2011 and 2016 in Ethiopia. The share of the pop- TOTAL URBAN RURAL crease of 59 percent), followed by Dire Dawa (increase of 55 ulation living below the national poverty line decreased from 2011 2016 % change 2011 2016 % change 2011 2016 % change percent), and Gambella (increase of 34 percent; Table 1). The 30 percent in 2011 to 24 percent in 2016. In contrast to ear- National 9,520 10,657 11.9 10,750 14,230 32.4 9,331 9,944 6.6 increase in consumption within regions was largely driven by lier periods, poverty reduction was especially strong in urban urban areas, with Tigray being the only region in which the areas and fairly weak in rural areas. The increasing disparity in Tigray 9,308 10,749 15.5 13,786 15,665 13.6 8,604 9,705 12.8 percentage change was greater in its rural areas. While the household welfare between rural and urban areas increased Afar 9,031 8,503 -5.9 11,317 15,339 35.5 8,277 7,892 -4.7 increase in urban consumption is statistically significant in all overall inequality, although it remains low compared to other Amhara 9,395 9,219 -1.9 10,289 16,095 56.4 9,301 8,758 -5.8 regions and cities, except for urban Somali, the change in countries in the region. For both rural and urban areas, the Oromia 9,748 10,993 12.8 10,758 14,090 31.0 9,615 10,894 13.3 rural consumption between 2011 and 2016 is only statisti- better-off generally experienced faster consumption growth. Somali 9,197 10,195 10.9 11,052 12,143 9.9 8,868 10,128 14.2 cally significant in Tigray, Gambella, and Harari. Looking at Monetary poverty decreased in most regions, particu- Ben.-Gumuz 9,671 10,641 10.0 11,640 14,659 25.9 9,506 9,971 4.9 median rather than mean consumption, we find that median larly in urban areas. Poverty rates decreased significantly SNNP 9,278 9,972 7.5 10,308 14,089 36.7 9,169 9,692 5.7 consumption decreased in Amhara and Afar, pointing to dif- in all regions expect for Tigray, Benishangul-Gumuz, and Ha- Gambella 9,134 11,382 24.6 10,304 13,862 34.5 8,837 10,210 15.5 ferent dynamics at different parts of the welfare distribution in rari, where the decrease in poverty rates is not statistically those regions (Table 1). Harari 11,255 16,739 48.7 12,448 18,392 47.8 10,638 15,607 46.7 Addis Ababa 10,377 12,718 22.6 10,377 12,718 22.6 - - - Dire Dawa 9,610 12,203 27.0 9,540 15,876 66.4 9,733 11,280 15.9 Source: HCES, 2011; 2016. World Bank staff calculations. Table 1  EAN HOUSEHOLD CONSUMPTION SAW THE GREATEST INCREASES IN M HARARI, DIRE DAWA AND GAMBELLA Regional mean annual consumption per adult equivalent in December 2015 prices significant (Table 3).4 Poverty reduction was especially strong Regions with the highest poverty rate in 2011, and TOTAL URBAN RURAL in the city administrations of Addis Ababa and Dire Dawa, as those relatively more urban, experienced the strongest 2011 2016 % change 2011 2016 % change 2011 2016 % change well as in the regions that were poorest in 2011, Somali and poverty reduction. Afar and Somali, which were the two National 11,009 12,500 13.5 13,901 18,649 34.2 10,434 11,014 5.6 Afar. Poverty decreased strongly in urban areas of all sizes, poorest regions in 2011, experienced among the highest Tigray 11,630 14,108 21.3 17,691 20,536 16.1 10,074 12,038 19.5 including small towns. Although urbanization in Ethiopia is poverty reduction between 2011 and 2016. But Tigray and relatively low — even compared to other African countries Amhara, where poverty rates were higher than the national Afar 10,641 12,902 21.2 13,945 18,645 33.7 9,298 10,512 12.5 — the annual urban population growth rate of six percent average in 2011, did not experience strong poverty reduction. Amhara 10,944 12,340 12.8 14,325 21,879 52.7 10,464 10,557 0.9 is fast. So far, urban growth has primarily been the outcome Meanwhile, the predominantly urban regions of Addis Ababa, Oromia 10,947 12,022 9.8 13,891 18,080 30.2 10,504 11,022 4.9 of natural population growth. Over the last two decades the Dire Dawa, and Harari experienced strong poverty reduction Somali 10,565 11,714 10.9 12,942 14,470 11.8 10,004 9,242 10.6 scale of rural-to-urban migration has been small, though this even if they had poverty rates lower than the national average Ben.-Gumuz 11,435 13,373 17.0 15,124 18,524 22.5 10,832 12,112 11.8 is starting to change. Urbanization is expected to continue in 2011. Those regions least urbanized in 2016 experienced SNNP 10,725 12,204 13.8 13,391 18,049 34.8 10,414 11,157 7.1 growing rapidly, particularly in small towns which are relative- the slowest poverty reduction, with the exception of Somali. ly poorer, and constitute a large share of the total urban pop- Oromia, Amhara, and Benishangul-Gumuz all have 2016 ur- Gambella 10,334 13,855 34.1 12,477 17,945 43.8 9,325 11,745 26.0 ulation. A detailed analysis on urbanization and poverty can banization rates under 20 percent, and all experienced very Harari 13,264 21,059 58.8 15,344 24,028 56.6 11,397 17,479 53.4 be found in the most recent Ethiopia Poverty Assessment slow or negligible poverty reduction (Figure 1). Addis Ababa 12,831 16,237 26.5 12,831 16,237 26.5 - - - (World Bank, 2020). By contrast, poverty reduction in rural Dire Dawa 11,268 17,428 54.7 11,617 20,718 78.3 10,532 11,393 8.2 areas only occured in Afar, Somali, and SNNP and increased Note: The increase in mean consumption is statistically significant in all regions except for Somali. For rural areas, the increase in mean in rural Dire Dawa, but this is only marginally significant from consumption is statistically different from zero only in Tigray, Gambella and Harari. a statistical point of view. Source: HCES, 2011; 2016. World Bank staff calculations. 4 Most of the analysis presented in the Ethiopia Poverty Assessment is based on detailed consumption data included in the Household In 2011, the HCES covered only two zones in Afar and three in Somali. In 2016, the coverage was improved – five zones and eight 3  Consumption Expenditure Surveys (HCES) (2010/11 and 2015/16), with price deflators used to adjust for spatial and temporal price zones were covered respectively in Afar and Somali. While the 2016 data is more representative of pastoral areas and the two regions, variations. The poverty rates discussed here are based on the national poverty line derived from a 2,200 kilo-calorie food basket and to maintain comparability households sampled from the new zones are excluded from the analysis. adjusted upwards to include non-food consumption. 16 ETHIOPIA REGIONAL POVERTY REPORT SECTION I: OVERVIEW OF REGIONAL POVERTY CHANGES IN ETHIOPIA, 2011-2016 17 Table 3 POVERTY DECREASED IN MOST REGIONS, ESPECIALLY IN URBAN AREAS 5 Figure 1 THE POOREST AND MOST URBAN REGIONS EXPERIENCED THE HIGHEST POVERTY REDUCTION BETWEEN 2011 AND 2016 Percentage of people below the national poverty line, by region and urban versus rural TOTAL URBAN RURAL A. Poverty reduction and initial poverty level B. Poverty reduction and urbanization 2011 2016 % change 2011 2016 % change 2011 2016 % change 50 Somali 50 Somali National 29.6 23.5 -20.6 25.7 14.8 -42.4 30.4 25.6 -15.8 Dire Dawa Dire Dawa 45 45 reduction reduction Faster Faster Tigray 31.8 27.0 -4.8 13.7 14.2 0.5 36.5 31.1 -5.3 Addis Addis 40 40 Rate of poverty reduction (%) Rate of poverty reduction (%) Ababa Ababa Afar 36.1 25.6 -10.5 23.7 10.6 -13.1 41.1 31.9 -9.2 Harari 35 35 Afar SNNP Afar Amhara 30.5 26.1 -4.4 29.2 11.6 -17.6 30.7 28.8 -1.9 30 SNNP 30 Gambella Gambella Oromia 28.7 23.9 -4.8 24.8 15.3 -9.5 29.3 25.3 -4.0 25 25 Somali 32.8 16.8 -16.0 23.1 19.4 -3.7 35.1 16.3 -18.8 20 20 Oromia Ben.-Gumuz 28.9 26.5 -2.4 21.3 17.7 -3.6 30.1 28.7 -1.4 Oromia Tigray 15 Tigray 15 Amhara Amhara SNNP 29.6 20.7 -8.9 25.8 14.4 -11.3 30.0 21.9 -8.1 10 10 Ben.-Gumuz Ben.-Gumuz Gambella 32.0 23.0 -9.0 30.7 16.6 -14.1 32.5 26.4 -6.1 Higher Higher 5 5 poverty poverty Harari 11.1 7.1 -4.0 11.7 6.0 -5.7 10.5 8.5 -2.0 0 0 Addis Ababa 28.1 16.8 -11.3 28.1 16.8 -11.3 25 27 29 31 33 35 37 0 20 40 60 80 100 Dire Dawa 28.3 15.4 -12.9 34.9 11.1 -23.8 14.2 23.3 9.1 Poverty rate in 2011 (%) Urbanization rate in 2016 (%) Note: Changes in bold are statistically significant at the 5 percent level. Note: The blue lines represent rate of poverty reduction between 2011 and 2016 at national level while the orange lines represent national poverty rate in 2011 and national urbanization rate in 2016 in Panel A and Panel B, respectively. Harari has been omitted Source: HCES, 2011; 2016. World Bank staff calculations. from Figure 1 for presentational purposes, as it had by far the lowest poverty rate in 2011 (11.1%). Poverty in Harari fell by 36% (to 7.1%) between 2011 and 2016. Source: HCES, 2011; 2016. World Bank staff calculations. Inequality, as measured by Gini coefficient, increased in Poverty trends in terms of depth and severity are more all regions except Somali, which weakened the extent complex. The depth of poverty—which measures how far of poverty reduction in most regions. With a Gini index of average consumption of the poor is from the poverty line 39, Tigray has the highest inequality in 2016 among the pre- (also called the poverty gap)—modestly dropped nationally, Figure 2 INEQUALITY INCREASED IN ALL REGIONS EXCEPT SOMALI BETWEEN dominantly rural regions; this is consistent with the fact that a sharp decrease in urban areas and a mild decrease in rural 2011 AND 2016 Tigray did not experience statistically significant poverty re- areas. The severity of poverty— which measures the average duction between 2011 and 2016, despite significant growth poverty gap for the poor but attaches more weight to the Gini coefficient, 2011 and 2016 in consumption. Somali experienced a decrease in inequality, poorest— decreased strongly in urban areas, but remained 2011 2016 and with a Gini index of 26, had the lowest inequality in 2016; constant in rural areas (Figure 3). Changes were mixed at 40 this helped reduced poverty in Somali between 2011 and the regional level. Poverty severity decreased strongly in Afar, 2016, despite insignificant consumption growth (Figure 2). Benishangul-Gumuz, and Gambella, and in the city adminis- 30 Gini coefficient These patterns show that while consumption growth is im- trations of Addis Ababa and Dire Dawa. In Harari, however, it portant, it does not necessarily lead to strong poverty reduc- increased sharply from a low base. 20 tion unless growth is shared and benefits the poorest people. 10 0 Afar Tigray Amhara Oromia Somali Benishangul- Gumuz SNNP Gambella Harari Addis Ababa Dire Dawa National 5  In 2011, the HCES covered only two zones in Afar and three in Somali. In 2016, the coverage was improved – five zones and eight Source: HCES; 2011, 2016. World Bank staff calculations. zones were covered respectively in Afar and Somali. While the 2016 data is more representative of pastoral areas and the two regions, to maintain comparability households sampled from the new zones are excluded from the analysis. If households from the new zones of Afar and Somali are included in the 2016 analysis, total (urban and rural combined) poverty becomes 23.6 percent and 22.4 percent and rural poverty will be 26.5 percent and 22.3 percent in Afar and Somali, respectively. Urban poverty in Afar remains the same (10.6 percent) while it went up to 22.9 percent in Somali. 18 ETHIOPIA REGIONAL POVERTY REPORT SECTION I: OVERVIEW OF REGIONAL POVERTY CHANGES IN ETHIOPIA, 2011-2016 19 Table 4 THE POVERTY SHARES OF REGIONS LARGELY REFLECT THEIR Figure 3 REGIONAL VARIATIONS IN THE TREND IN POVERTY SEVERITY POPULATION SHARES 8 Poverty severity based on national poverty line, 2011 and 2016 Poverty rates, poverty shares, and population shares by region and agro-ecological zone, 2016 2011 2016 POVERTY RATE POVERTY SHARE POPULATION SHARE 0.04 Tigray 27.0% 6.6% 5.8% 0.035 Afar 23.6% 1.9% 1.9% 0.03 0.025 Amhara 26.1% 25.5% 23.0% 0.02 Oromia 23.9% 38.3% 37.8% 0.015 Somali 22.4% 5.5% 5.8% 0.01 Benishangul-Gumuz 26.5% 1.3% 1.1% 0.005 SNNP 20.7% 17.5% 19.9% 0 Gambella 23.1% 0.4% 0.4% Urban Rural Tigray Afar Amhara Oromia Somali Benishangul-G. SNNP Gambella Harari Addis Ababa Dire Dawa Harari 7.1% 0.1% 0.3% Addis Ababa 16.8% 2.6% 3.6% Dire Dawa 15.4% 0.3% 0.5% Urban/Rural Regional Source: HCES, WMS, 2016; World bank staff calculations. Source: HCES; 2011, 2016. World Bank staff calculations. Half of the urban poor live in the Oromia region and also declined, despite rapid urban population growth. In 2016 Addis Ababa. Urban poverty in Ethiopia declined rapidly be- an estimated 2.4 million people lived below the poverty line in In contrast to many countries, there is no strong re- tween 2011 and 2016. The share of the urban population liv- urban areas, the majority of which were located in the Oromia gional concentration of poverty in Ethiopia. Poverty ing below the national poverty line decreased from 26 percent region and Addis Ababa (Figure 4). rates among the largely rural regions vary from 21 percent in in 2011 to 15 percent by 2016. The absolute number of poor SNNP to 27 percent in Tigray (Table 4).6 Poverty in the two city administrations and the largely urban region of Harari is significantly lower at 15 percent in Dire Dawa, 17 percent in Addis Ababa, and 7 percent in Harari. Given the small regional Figure 4 HALF OF THE URBAN POOR ARE LOCATED IN OROMIA AND ADDIS ABABA disparities, the poverty shares of regions largely reflect their population shares, with the most populated regions account- Regional share of the urban poor, 2016 ing for the bulk of the poor. The Oromia region accounts for Other regions, 5% 38 percent of the poor in Ethiopia, Amhara for 26 percent, and SNNP for 18 percent. In contrast to conventional wisdom, the Tigray, 7% Photo: World Bank/Gelila Woodeneh pastoral regions of Afar and Somali are not poorer in terms Somali, 7% of monetary poverty than the average: poverty rates in 2016 Oromia, 29% amounted to 22 percent in Somali and 24 percent in Afar.7 Amhara, 15% Addis Ababa, SNNP, 15% 22% Source: HCES; 2011, 2016. World Bank staff calculations. 6 The largely rural regions are Tigray, Afar, Amhara, Oromia, Somali, Benishangul-Gumuz, SNNP, and Gambella. 7 However, the pastoral areas have the highest rate of vulnerability to poverty which measures the proportion of people who are likely to 8 The results for Afar and Somali include households sampled from the new zones, as described above. The results reported in Table 1 be poor during a shock (WB & DFID, 2019). exclude these zones and thus report different 2016 poverty rates for Afar and Somali. 20 ETHIOPIA REGIONAL POVERTY REPORT SECTION I: OVERVIEW OF REGIONAL POVERTY CHANGES IN ETHIOPIA, 2011-2016 21 Data does not confirm the persistent notion that pas- to Afar, Somali’s high level of PSNP and HFA coverage does toral Ethiopians are poorest. Pastoral areas, which cover not explain the strong poverty reduction: simulated pre-ben- Figure 6 STRONG SHIFTS FROM CEREALS TO CASH CROPS IN THE WEST AND most parts of Afar and Somali region, have typically had av- efit poverty rates in Somali are similar to actual post-benefit EAST OF THE COUNTRY erage or below-average poverty rates. Somali and Afar also poverty rates. This either means that benefits were directed Changes in the predominant kind of crop cultivation as measured by revenue per capita at the experienced exceptionally strong poverty reduction between to households already above the poverty line, or that benefits zonal level, 2011 to 2016 2011 and 2016, dropping 16 percentage points in Somali went to extremely poor households who remained below the and 11 percentage points in Afar (Table 5). This finding is poverty line despite the assistance. at odds with the common perception that pastoral regions Parts of Oromia, SNNP, and Harari exhibited large Tigray are the most destitute, and the belief that the 2015/16 El movements into cash crop production between 2011 Change to cereal crop production Nino drought was particularly devastating in the pastoral and 2016. Growth in cash crop cultivation was especially No change lowlands. Although Afar was heavily affected by the drought, Change to cash crop production strong for coffee, khat, and oil seeds such as sesame, nueg, NA the Productive Safety Net Program (PSNP) was scaled up Afar and kocho. The zones that showed particularly strong shifts and complemented by Humanitarian Food Aid (HFA). Overall, towards cash crops were in Oromia (Jimma, West Hararge, Amhara over half the Afar population received either PSNP or HFA and East Hararge), SNNP (Sidama), and Harari. Figure 6 (or both), according to the 2016 HCES. This high level of coverage translated into higher consumption expenditures shows changes in the predominant type of crop revenue at Benishangul-Gumuz the zonal level. This is determined by which crop type had the and lower poverty rates. A simulation that removed benefits higher revenue per capita. Zones in green represent those in Dire Dawa (both in-cash and in-kind) from the consumption aggregate Harari which predominant revenue per capita switched from cereals suggests that poverty rates in 2016 in Afar would have been Addis Ababa to cash crops between 2011 and 2016. In the eastern parts much higher without PSNP and HFA – 34 percent instead of Oromia, this shift was largely towards khat, following the of the actual poverty rate of 24 percent (Figure 5). In Somali, trend of increasing khat prices relative to the price of coffee. Gambella Oromia the picture is less straightforward: one zone in northern So- Red zones represent a switch from cash crops to cereals, mali was heavily affected by the drought, while the largest Somali while zones in yellow did not change (Figure 6). part of Somali had unusually wet soil moisture. In contrast SNNP N Figure 5 PSNP AND HFA WERE HIGHLY EFFECTIVE IN AFAR Actual poverty rates in 2011 and 2016; simulated pre-benefit poverty rate 2016 0 100 200 300 400 500 km Poverty rate - 2011 Poverty rate - 2016 Pre-bene t poverty rate - 2016 40 36.1 Source: Calculations from Agricultural Sample Surveys, 2011 and 2015. 34.2 32.8 30 23.6 22.4 22.9 20 10 0 Afar Somali Note: The pre-benefit poverty rate is an estimate of what poverty would have been in 2016 in the absence of PSNP and HFA. Source: HCES; 2011, 2016. World Bank staff calculations. 22 ETHIOPIA REGIONAL POVERTY REPORT SECTION I: OVERVIEW OF REGIONAL POVERTY CHANGES IN ETHIOPIA, 2011-2016 23 NON-MONETARY WELFARE Box 2  ata gaps and future research: Areas for poverty related D Although non-monetary indicators improved between tend to be much worse in the pastoral regions (Afar and So- 2011 and 2015, challenges remain and regional inequi- mali). Other regions with a significant pastoral population, such policies and interventions ties persist. The pastoral regions are not worse off in terms of as Oromia, also tend to perform below average on human While the availability of data and evidence base relating to poverty and welfare is in general good in Ethiopia, monetary poverty but they lag significantly on human develop- development outcomes, reflecting the difficulty of providing there are areas for improvement both in terms of increasing frequency and coverage of official data and ment outcomes. Education, health, and other social indicators public services in low-density areas with mobile populations. additional regionally focused studies. We have outlined a few areas below. Improving frequency of welfare data. The HCES and WMS are the main data sources used to monitor welfare in Table 5 PASTORAL REGIONS DO NOT HAVE HIGHER RATES OF MONETARY Ethiopia and have been conducted roughly every five years between 1996 and 2016. With five welfare and poverty POVERTY, BUT LAG ON HUMAN DEVELOPMENT OUTCOMES surveys over a period of 20 years, Ethiopia compares favorably to other IDA countries. There is however room to Non-monetary indicators of welfare, 2016 further strengthen the welfare monitoring system. The HCES and WMS are currently implemented every five years. This may not be frequent enough to confidently inform effective policy-making in a fast-growing and changing coun- UNDER FIVE YOUTH (15-24) try with frequent weather shocks. To illustrate, the results of the latest 2015/16 survey may suggest that progress MORTALITY RATE FULLY IMMUNIZED HEALTH FACILITY PRIMARY SCHOOL (PER THOUSAND) CHILDREN (%) DELIVERY (%) COMPLETION (%)10 in rural Ethiopia has slowed substantially, and that renewed efforts are needed to uplift rural areas. This survey was however implemented amidst the El Niño drought, so the rural numbers may reflect the short-term impact of a se- 2011 2016 2011 2016 2011 2016 2011 2016 vere shock and as such mask a positive long-term trend. Having more frequent poverty measurements would allow National 88.0 67.0 24.3 38.5 10.0 26.2 27.6 32.7 for distinguishing longer-term trends from short-term fluctuations, as well as a closer monitoring of living standards. Tigray 84.7 48.9 58.9 67.3 11.6 56.9 34.8 44.2 Afar 108.2 108.7 8.6 15.2 6.8 14.7 17.7 31.1 Making welfare data available at a more disaggregated level. It is also important to note that the HCES and WMS are representative at regional, urban and rural areas. While this is sufficient to make national and to some Amhara 81.0 59.9 26.3 45.8 10.2 27.1 24.6 29.5 extent regional analysis, a detailed regional analysis requires a more disaggregated data which is representative at Oromia 89.7 69.5 15.6 24.7 8 18.8 25.8 30.1 a lower level, for example, at zonal or even woreda level. If expanding the whole HCES/WMS to a lower level is not 9 Somali 90.4 79.6 16.6 21.8 7.6 17.9 11.6 22.6 feasible, an alternative is to collect a separate disaggregated data on a selected socio-economic indicators. The Ben.-Gumuz 120.0 82.5 23.6 57.4 9.1 25.7 28.2 39.2 second option will not only help to monitor the trends of main socio-economic indicators, but it will also help to SNNP 94.5 70.6 24.1 46.9 6.2 25.5 20.4 32.4 impute poverty at a lower level in combination with the main HCES data using survey to survey imputation methods. Gambella 117.7 78.2 15.5 41.1 27.6 45 35.1 45.4 Evidence on lack of growth at the bottom of the distribution. While there has been strong consumption growth Harari 95.1 76.3 34.1 42.2 32.4 50.2 51.1 47.4 in most of the regions between 2011 and 2016, there was in general no growth at the bottom, particularly in rural Addis Ababa 41.1 38.3 78.7 89.2 82.7 96.6 74.6 66.1 areas. This pattern is also observed at national level during the same period and earlier periods (2005 -2011 and Dire Dawa 72.1 73.4 58.6 75.9 39.8 56.2 57.5 50.0 2011 and 2016). This contributed for slower poverty reduction in light of the relatively stronger consumption growth and if it continues in the future it will make future poverty reduction even harder. A study focusing at understanding Note: Under-5 mortality rate is expressed as number of deaths per 1,000 live births. Delivery in health facility refer to women who gave birth in the 5 years prior to the survey. better why those at the bottom of the welfare distribution are not benefiting from the fast growth is important for future poverty reduction policies. Source: DHS, 2016 (for health indicators); WMS, 2016 (for primary school completion); World Bank staff calculations. Study focusing on regional poverty policies and strategies and expenditure patterns. In order to device pol- icies that can lead to further poverty reduction and minimize future divergence, it is imperative to understand how regions differ in their poverty reduction strategies and public spending decisions. A regionally focused study on how Health indicators improved in general between 2011 percent) and Oromia (25 percent). Similarly, health facility regions differ in their socio-economic strategies and public spending decisions can provide an important insight as and 2016, but Afar and Somali still lag. With 109 deaths delivery improved in all regions, but regional variations are to why some regions do better at reducing poverty while others struggle. per 1000, Afar had the highest under-five mortality rate, fol- still large. Afar, Somali, and Oromia recorded less than 20 lowed by Benishangul-Gumuz (83), and Somali (80). The percent health facility delivery in 2016. Addis Ababa had the percentage of fully immunized children increased in all re- best health indicators in 2016 overall, while Tigray performed gions, but remains very low in Afar (15 percent), Somali (22 better among the predominantly rural regions (Table 5). 10 In 2011, the WMS covered only two zones in Afar and three in Somali. In 2016, coverage improved – five zones and eight zones were covered respectively in Afar and Somali. While the 2016 data is more representative of pastoral areas and the two regions, to maintain comparability, households sampled from the new zones are excluded from the analysis. If households from the new zones of Afar and 9 While the sample in some of the large zones could be large enough to be representative, this may not apply to many of the smaller Somali are included in the 2016 analysis, youth primary school completion rate in Afar and Somali will respectively be 24.2 percent zones. and 18.6 percent. 24 ETHIOPIA REGIONAL POVERTY REPORT SECTION I: OVERVIEW OF REGIONAL POVERTY CHANGES IN ETHIOPIA, 2011-2016 25 Primary school completion for youth aged 15-24 im- Differences in human capital indicators are also large proved in most regions, but it is generally low. Even in within regions. Figures 7 and 8 shows zonal fertility and Figure 8 ADULT LITERACY RATE WAS LOW IN MOST ZONES IN SOMALI AND Addis Ababa, only two-thirds of youth completed primary adult literacy rates, respectively, in 2016. Results show signif- AFAR IN 2016 education in 2016. In Harari and Dire Dawa, 47 percent and icant variations within some regions. For example, in Oromia, Adult literacy rate at zone level in 2016 50 percent of youth completed primary education in 2016, total fertility rate ranges from two children per woman in Bore- respectively. With a 45 percent youth primary school com- na to nine in southwest Showa. Similarly, adult literacy rates pletion rate, Gambella performed better among the primarily vary from 18 percent in Degehabur to 63 percent in Liben. 0 to 10 rural regions. Somali, with just 23 percent, had the lowest Tigray 10 to 20 school completion rate (Table 5). 20 to 30 30 to 40 40 to 50 Afar 50 to 60 60 to 70 Amhara 70 to 80 Figure 7 ALL ZONES IN SOMALI HAD FERTILITY RATES OF SIX AND ABOVE IN 2016 80 to 90 Total fertility rate at zone level in 2016 90 to 100 Benishangul-Gumuz NA Dire Dawa Harari Tigray Addis Ababa 0 to 2 2 to 4 4 to 6 Gambella Oromia Afar 6 to 8 8 to 10 Somali Amhara NA SNNP Benishangul-Gumuz N Dire Dawa Harari Addis Ababa 0 100 200 300 400 500 km Gambella Oromia Source: HCES; 2016. World Bank staff calculations. Somali SNNP Despite the severity of the 2015/16 El Nino drought, Overall production remained higher than two years earlier in N almost all regions experienced a decrease in self-re- 2013/14.12 Food security indicators in the WMS tell a similar ported food insecurity. Failure of two consecutive rainy story: the share of Ethiopians who experienced a food short- seasons in 2015 led to a sharp increase in humanitarian age in the 12 months prior to the survey decreased from 22 0 100 200 300 400 500 km requirements, with more than 10 million additional Ethiopi- percent in 2011 to 10 percent in 2016 (Table 6). The food ans needing humanitarian food aid, on top of the chronical- gap—the number of months a household experienced food ly food-insecure PSNP caseload of 8 million. 11 Government shortages—remained the same, but given that the gap only Source: DHS; 2016. World Bank staff calculations. and development partners mounted a large-scale humanitar- refers to households who actually experienced food short- ian response, which was credited with averting loss of life. ages, it also decreased on aggregate. At the regional level, Overall, chronic malnutrition decreased between 2011 and only Benishangul-Gumuz experienced a significant increase 2016, acute malnutrition remained unchanged, and overall in self-reported food insecurity, but it started from a low base. agricultural production in 2015/16 decreased only marginally. 11  Based on the 2016 Humanitarian Requirements Document 12 Based on the Ethiopian Demographic and Health Surveys (2011 and 2016) and the Agricultural Sample Surveys (2011-2016) 26 ETHIOPIA REGIONAL POVERTY REPORT SECTION I: OVERVIEW OF REGIONAL POVERTY CHANGES IN ETHIOPIA, 2011-2016 27 Table 6 AS A SIGN OF PROGRESS, FOOD SHORTAGES ARE INCREASINGLY LESS INEQUALITY OF OPPORTUNITY PREVALENT Although inequality of monetary outcomes remains Regional differences in primary school enrollment and Food shortage in the last 12 months -% yes- and number of months of shortage, 2005-2011-2016 relatively low in Ethiopia, economic growth in the last completion are striking. Almost 90 percent of age-eligible decade has not been equally shared. Ensuring equal op- children living in Addis Ababa are enrolled in primary school, 2011 2016 portunity of access to services, and investing in the produc- while only half of these children in Somali are enrolled (Fig- Food shortage (%) Food gap (months) Food shortage (%) Food gap (months) tivity and welfare of all citizens, is linked to more sustainable ure 9). Somali lags far behind Afar, the next lowest region. National 21.6 3.3 10.2 3.3 and inclusive economic growth. Access to basic services im- The patterns for enrollment in secondary education (for those Tigray 13.2 3 11.9 2.5 proved overall, but large disparities remain between groups. aged between 15 and 18) are similar, but at much lower ab- Afar 7.7 5.2 9 3.8 Rural versus urban differences and gaps in household wel- solute levels (Figure 10). The mean secondary enrollment rate Amhara 23.2 3.1 10.4 3 fare are the main sources of access inequality. Children living for all age-eligible Ethiopian children was under 20 percent Oromia 16 3.1 10.5 3.6 in cities and children from households in the top consump- (Figure 11). Again, a strong regional discrepancy is evident, tion quintile have far better access to key services compared with the rate for the city administration of Addis Ababa sur- Somali 30.3 4.4 6 3 to rural or poorer children. passing 35 percent, in contrast to rates below 15 percent in Benishangul-Gumuz 5.6 2.1 8.5 2.8 Somali and Oromia. SNNP 35 3.4 12.6 3.2 Gambella 31.6 2.6 3.8 1.2 Harari 8 3.2 0 NA Addis Ababa 7.8 4 1.1 3.6 Figure 9 PRIMARY SCHOOL ENROLLMENT CRUCIALLY DEPENDS ON THE Dire Dawa 13.5 1.6 7.7 2 REGION WHERE ONE IS BORN Note: Food gap is only calculated for those households that report a food shortage. Coverage of access to primary education by circumstance variables, 2016 Source: WMS, 2005; 2011; 2016 Location Gender Region Consumption Rural accessibility quintile quintile Photo: World Bank/Chris Terry Source: HCES, 2016; WMS, 2016. World Bank staff calculations. 28 ETHIOPIA REGIONAL POVERTY REPORT SECTION I: OVERVIEW OF REGIONAL POVERTY CHANGES IN ETHIOPIA, 2011-2016 29 Very strong regional gaps were also evident in spatial with all predominantly rural regions having less than 20 per- LOCATION AND HOUSEHOLD WEALTH DETERMINES PRIMARY SCHOOL distribution of health posts and hospitals. About two- cent of families living within 5 km of a hospital, compared to Figure 10 COMPLETION thirds of Ethiopian households live within 5 km of one of the almost complete access in Addis Ababa (Figure 13). Almost country’s approximately 17,000 health posts. Regional differ- all urban households are located within the 5 km threshold. Completion of primary school by circumstance variables, children 15-18, 2016 ences are stark, however, ranging from almost full access in Rates of access to improved water sources also demon- Addis Ababa and 90 percent in Dire Dawa to around 50 per- strate similar regional disparity, with Oromia and Somali only cent in Somali and Afar (Figure 12). The gap between regions slightly above 50 percent compared to almost 100 percent is even more striking when considering access to hospitals, access in Addis Ababa (Figure 14). Figure 12 LOCATION DETERMINES PROXIMITY TO A HEALTH POST Coverage of having a health post within 5km by circumstance variables, 2016 Location Gender Region Consumption Rural accessibility quintile quintile Source: HCES, 2016; WMS, 2016. World Bank staff calculations. Figure 11 LOCATION AND HOUSEHOLD WEALTH DETERMINE ENROLLMENT IN SECONDARY SCHOOL Completion of primary school by circumstance variables, children 15-18, 2016 Location Gender Region Consumption Rural accessibility quintile quintile Source: HCES, 2016; WMS, 2016. World Bank staff calculations. Location Gender Region Consumption Rural accessibility quintile quintile Source: HCES, 2016; WMS, 2016. World Bank staff calculations. 30 ETHIOPIA REGIONAL POVERTY REPORT SECTION I: OVERVIEW OF REGIONAL POVERTY CHANGES IN ETHIOPIA, 2011-2016 31 PUBLIC EXPENDITURE Figure 13 LOCATION DETERMINES PROXIMITY TO A HOSPITAL The volume and nature of public investment deter- Public expenditure per capita increased in all regions Access to a hospital within 5km, children 7-18, 2016 mines availability and quality of basic services, in- between 2011 and 2016, although with large varia- cluding education and health. Public investment can also tions, except in Afar. In Afar, per capita public expenditure have important implications for poverty reduction. Ethiopia’s dropped slightly between 2011 and 2016, although it was federal system gives regions autonomy on how to spend above average at the beginning of the period. On the other funds coming both from the federal budget and from region- hand, Oromia, which had the lowest per capita expenditure, al revenue. Understanding how much each region spends is experienced the biggest growth: per capita expenditure in therefore essential to understanding why some regions are Oromia increased about three-fold between 2011 and 2016. more successful at reducing poverty. We analyzed regional In 2016, Gambella had the highest per capita expenditure of public spending patterns from 2011 to 2016 using Ministry 5,127 Birr while SNNP had the lowest with 1,888 Birr. Per of Finance (MoF) data. However, lack of regional spending capita expenditure for most of the regions is close to the data over long periods, and limited details of how regions national average though the two city administrations (Addis spend their budget, limit the depth of our analysis. Still, some Ababa and Dire Dawa) and the smaller regions like Harari and interesting patterns emerged, which we summarize below. Gambella spend significantly higher per capita (Figure 15).13 Figure 15 PER CAPITA PUBLIC EXPENDITURE INCREASED OVER TIME Location Gender Region Consumption Rural accessibility Per capita real public expenditure, 2011 and 2016 quintile quintile Source: HCES, 2016; WMS, 2016. World Bank staff calculations. 2011 2016 6000 5000 Figure 14 LOCATION LARGELY DETERMINES ACCESS TO IMPROVED WATER In 2016 Birr 4000 Access to an improved water source by circumstance variables, children 7-18, 2016 3000 2000 1000 0 Afar Tigray Amhara Oromia Somali Benishangul- Gumuz SNNP Gambella Harari Addis Ababa Dire Dawa National Source: MoF; World Bank staff calculations. Location Gender Region Consumption Rural accessibility quintile quintile 13 The budget for most of the public spending comes from the Federal Government though regions also cover part of their spending Source: HCES, 2016; WMS, 2016. World Bank staff calculations. from internal revenue. Regions differ in their capacity of raising internal revenue and that partly explains the variations in public expen- diture per capita (for a detailed discussion on this refer WB & DFID, 2019) 32 ETHIOPIA REGIONAL POVERTY REPORT SECTION I: OVERVIEW OF REGIONAL POVERTY CHANGES IN ETHIOPIA, 2011-2016 33 THE WAY FORWARD ON POVERTY REDUCTION AND NARROWING Figure 16 ADDIS ABABA AND TIGRAY SPEND THE LEAST ON ADMINISTRATION REGIONAL GAPS Expenditure share by sectors (%) in 2016 The regional analysis shows that consumption growth is 2011 and 2016 Tigray registered one of the highest consump- Economic Social Administration Other not enough for continued poverty reduction— the type tion growth rates among the predominantly rural regions, it did 100% of growth matters. If growth is accompanied by increased not record significant poverty reduction. Somali, on the other 80% inequality and a growing gap between rural and urban areas, hand, saw large poverty reduction despite low growth in con- 60% poverty reduction is likely to be weak. On the other hand, even sumption. Thus, while measures to continue to grow con- small consumption growth could lead to strong poverty reduc- sumption are necessary, they should be designed in such a 40% tion if inequality does not increase. For example, while between way that the poorest and those living in rural areas also benefit. 20% 0% Afar Tigray Amhara Oromia Somali Benishangul- Gumuz SNNP Gambella Harari Addis Ababa Dire Dawa National POLICY RECOMMENDATIONS Following are some specific suggestions for future policy and boosting demand for agricultural outputs and leads to Source: MoF; World Bank staff calculations. direction. poverty reduction (Bachewe et al., 2015; World Bank, 2020). Continue investing in agriculture. Most of the population Ease migration from rural to urban areas, particularly to Figure 17 GAMBELLA SPENDS THE HIGHEST ON PRO-POOR SECTORS lives in rural areas and depends mostly on agriculture in all nearby small towns. Urbanization, even in small towns, has regions except the urban-dominated ones, which account proved an effective poverty reduction strategy, and relaxing the Per capita expenditure on selected pro-poor sector in 2016 for less than ten percent of the national population. The ag- constraints to migrate to nearby towns is likely to reduce pov- Education Health Agriculture & Rural Dev. Water Resources riculture sector can be improved by continued investment in erty. Migrants also face challenges in their destinations includ- 2000 past interventions that improved productivity and reduced ing limited access to services and employment opportunities 1500 poverty. Better access to modern agricultural inputs (includ- and addressing that will further help to reduce poverty.15 In 2016 Birr ing improved seeds and chemicals like fertilizers, pesticides Strengthen existing social protection programs. The ru- 1000 and herbicides) and markets have in the past contributed ral safety-net has been effective in reducing poverty and Ethi- to agricultural growth which in turn led to poverty reduction 500 opia should expand this to regions where coverage is very (Bachewe et al., 2015). However, there should also be target- low, along with improving beneficiary targeting within regions. 0 ed interventions focusing on poorer communities and farm- Given that poverty is not highly geographically concentrat- Afar Tigray Amhara Oromia Somali Benishangul- Gumuz SNNP Gambella Harari Addis Ababa Dire Dawa National ers who, due to lack of complementary endowments such as ed, expanding the woredas covered, while allocating fewer land, education, and social network, are less likely to benefit beneficiaries within each woreda, may improve targeting and from such interventions. hence help reduce poverty. The urban safety net effect on Source: MoF; World Bank staff calculations. Take additional policy measures in non-agricultural in- poverty reduction is not yet known because it started more terventions. While continuing investment in agriculture is im- recently, but it is expected to contribute to poverty reduction. portant, the working population is increasing and access to There are also important differences in terms of how Per capita expenditure on combined “pro-poor” sectors Continue investment in human capital, particularly land is limited. There are also farmers who are less likely to the regions allocate their spending. The share of spend- (agriculture and rural development, water resources, education, narrow inequality between and within re- benefit from any investments, meaning that limiting support to ing on economic sectors in 2016 is highest in Harari (44 health and education) is highest in Gambella and lowest gions and contribute to medium and long-term pover- agriculture only will likely curtail poverty reduction in rural areas. percent) and lowest in Dire Dawa (14 percent). Dire Dawa in SNNP. Gambella also spends the most on health and edu- ty reduction. Lack of education is strongly correlated with spends the largest share (33 percent) on the social sector, cation, consistent with the fact that Gambella performs better Encourage off-farm employment opportunities by in- poverty; even having some years of education is associated while Somali spends the lowest (18 percent).14 The share in many non-monetary indicators. Somali, on the other hand, tegrating small towns with rural areas. Investing in small with significantly lower poverty compared to not having any of administration-related expenses is high in general, but is spends less on health and education compared to most re- towns in infrastructure and services is important both to con- education. Devising policies or interventions to keep chil- highest in Gambella (53 percent), and lowest in Addis Ababa gions, consistent with its lower performance on human capital tinue strong urban poverty reduction, a big part of which is dren from poor rural families in school for longer, potentially (14 percent) and Tigray (16 percent) (Figure 16). indicators. Further detailed study on how the regions spend attributed to small towns, and to boost growth in rural areas through income incentives, will be particularly crucial to share their budget is needed to better understand possible links with linked to small towns. Investing in towns improves agricul- the benefits of growth more widely. public expenditure and poverty reduction (Figure 17). tural productivity by improving availability of modern inputs 14 Social sector includes education, health, culture and sports, and labor and social affairs while economic sector includes agriculture and rural development, water resources, trade and industry, mines and energy, transportation and communication, construction. 15 For a detailed discussion on this refer to the most recent Ethiopia Poverty Assessment (World Bank, 2020). 34 ETHIOPIA REGIONAL POVERTY REPORT SECTION I: OVERVIEW OF REGIONAL POVERTY CHANGES IN ETHIOPIA, 2011-2016 35 SECTION 2 Regional Analysis Photo: World Bank/Chris Terry 36 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 37 POVERTY AND INEQUALITY Food and nonfood consumption in the Tigray region did not increase as much between 2011 and 2016. While it increased significantly between 2011 and 2016, both increased by 36 percent in the rest of the country, in Tigray it in rural and urban areas. Mean consumption per adult in- rose only by 16 percent. On the contrary, rural consumption creased by 21 percent in Tigray alone compared to a 13 per- grew faster in Tigray than in the other regions combined – 20 cent increase for all other regions combined. Though urban percent versus five percent (Table 1). Tigray consumption in Tigray is larger than the rest of the country, it Table 1 CONSUMPTION INCREASED BETWEEN 2011 AND 2016 BOTH IN TIGRAY AND THE REST OF THE COUNTRY Mean annual household consumption per adult equivalent in December 2015 prices, 2011 and 2016 INTRODUCTION TOTAL URBAN RURAL This chapter focuses on the evolution of socio-economic and health indicators such as poverty, inequali- 2011 2016 % change 2011 2016 % change 2011 2016 % change ty, access to basic services, and education in the Tigray region between 2011 and 2016. Using data from Tigray 11,630 14,108 21.3* 17,691 20,536 16.1* 10,074 12,038 19.5* a variety of sources, mainly the twinned household living standards surveys (HCES and WMS)16 and the National 10,968 12,397 13.0* 13,595 18,495 36.0* 10,455 10,953 4.8* Demographic and Health Survey (DHS), the chapter begins by documenting changes in monetary poverty excl. Tigray and living standards in the region. It then moves to an analysis of non-monetary welfare before turning to Note: * represents changes that are statistically significant. challenges of inequality of opportunity. The final section discusses changes in sectoral allocations of social Source: HCES, 2011; 2016; World Bank staff calculations. spending in the region from 2011 and 2016. The chapter aims to provide policy makers with information and analysis that will be useful in improving the effectiveness of poverty reduction programs in the region. Despite the significant increase in consumption, pov- and rural Tigray, indicating that the poor did not benefit from erty in Tigray did not fall between 2011 and 2016. The consumption growth. In other regions combined, poverty de- Some of the main findings include: change in poverty is not statistically significant both in urban creased, particularly in urban areas (Table 2). ● The poverty rate in Tigray stood at 27 percent in 2016, as was the case for human capital indicators such as compared to 32 percent in 2011. This change was not, health and education. Tigray in general has better hu- Table 2 POVERTY DID NOT DECREASE IN TIGRAY BUT IT DID IN THE REST OF however, statistically significant in either urban or rural man capital indicators than the rest of the country. This THE COUNTRY parts of the region. is also reflected in a higher score on the Human Oppor- tunity Index than the national average – implying higher Percentage of people below the national poverty line ● The pattern of consumption growth in Tigray was in- equality of access to key basic services for children in equality-increasing, with growth in both rural and ur- TOTAL URBAN RURAL Tigray compared to the rest of Ethiopia. ban areas being strongly concentrated at the top of % pt. % pt. % pt. the distribution. There was no significant consumption ● Average per capita public expenditure in Tigray is higher 2011 2016 change 2011 2016 change 2011 2016 change growth at the bottom of the distribution in Tigray, and than in the rest of the country, while the share of ex- Tigray 31.8% 27.0% -4.8 13.7% 14.2% 0.4 36.5% 31.1% -5.3 by 2016 inequality in the region was higher than in the penditure on administration is lower. Per capita expendi- National 29.4% 23.2% -6.2* 26.6% 14.6% -12.0* 30.0% 25.3% -4.7* rest of Ethiopia. tures on “pro-poor” sectors including health, education, excl. Tigray agriculture and rural development, and water resources Note: * represents changes that are statistically significant. ● Non-monetary indicators including ownership of con- are also higher in Tigray. Source: HCES, 2011; 2016; World Bank staff calculations. sumer durables and access to drinking water improved, In Tigray, there was no consumption growth at the bot- and gets stronger further up the distribution. The fact that tom of the distribution between 2011 and 2016. Both consumption growth is positive for most of the distribution in urban and rural areas, consumption growth for the bot- is a departure from the result observed at national level – no tom one-third of the distribution was not statistically different growth for most of the distribution (Figure 1).17 from zero. After that, consumption growth becomes positive 16 Household Consumption Expenditure Survey (HCES) and Welfare Monitoring Survey (WMS) are the two official household surveys collected by CSA every five year and used to monitor welfare. HCES covers monetary welfare and WMS deals with non-monetary welfare like health and education. Both surveys are collected from the same sample of households. 17  See Ethiopia Poverty Assessment: World Bank (2020). 38 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 39 The rise in inequality between 2011 and 2016 has more was statistically significant. However, there was nevertheless Figure 1 GIC FOR TIGRAY URBAN AND RURAL to do with the overall pattern of growth than it does an increase in inequality, as noted. The “between” share of with increasing gaps between urban and rural parts of inequality – the part of inequality resulting from differences in Tigray. Average consumption was already significantly high- average consumption between urban and rural areas – de- Urban Rural er in urban than in rural areas in 2011. The gap decreased creased from 32 percent in 2011 to 29.6 percent in 2016. in 2016 because higher rates of growth in rural compared to On the other hand, the “within” share of inequality – the part urban areas. Indeed, as shown earlier, although there was a resulting from changes in distribution within urban and rural small increase in poverty in urban Tigray, while rural Tigray areas – was quite consistent in both periods (Figure 2). saw a small decrease in poverty – neither of these changes Figure 2 INCREASED INEQUALITY IS THE OUTCOME OF THE OVERALL GROWTH PATTERN Decomposition of the Gini coefficient into a between rural-urban component and a within-component Overlap Within Between 100% Relative contribution to Gini (%) 90% 14.3 17.7 80% 53.7 Source: HCES, 2011; 2016; World Bank staff calculations. 70% 52.7 60% 50% 40% The lack of consumption growth for the poorest led 90th percentile and consumption at the 10th percentile in- 30% 20% 32.0 29.6 to significant increase in inequality In Tigray between creased from 3.7 in 2011 to 5.5 in 2016, reflecting strong 10% 2011 and 2016. The Gini index increased from 34 to 39 consumption growth at the top and lack of growth at the 0% between 2011 and 2016. The ratio of consumption at the bottom (Table 3). 2011 2016 Source: HCES, 2011; 2016; World Bank staff calculations Table 3 INEQUALITY INCREASED BETWEEN 2011 AND 2016 Increased inequality between 2011 and 2016 in Tigray for the rest of the country. However, the poverty-increasing 2011 2016 contributed to the lack of poverty reduction. Figure 5 effect of increasing inequality in Tigray is more than three Gini index p90/p10 ratio Gini index p90/p10 ratio decomposes the change in poverty into a growth and re- times what it was in the rest of the country over the period Tigray 34.4 3.7 38.8 5.5 distribution component using the Datt-Ravallion decomposi- (7.6 points versus 2.4 points). Another way of interpreting tion technique. It indicates which part of the poverty change this is that if there were no changes in the distribution be- National 29.8 3.6 32.9 4.3 can be attributed to: (i) the change in average consumption tween 2011 and 2016, poverty would have decreased by Source: HCES, 2011; 2016; World Bank staff calculations. (growth effect), and (ii) the change in the redistribution of con- 12.5 and 8.5 percent in Tigray and the rest of the country, sumption; that is, changes in inequality (redistribution effect). respectively, instead of the 4.8 and 6.1 percent we actually The growth effect on poverty in Tigray is negative (poverty-re- observe. In other words, if the consumption distributions in ducing), and is significantly larger than in the rest of Ethiopia. 2011 were maintained in 2016, poverty reduction could have The redistribution effect is positive not only for Tigray but also been stronger in Tigray than in other regions (Figure 3). 40 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 41 In Tigray, households that primarily depend on crop from the rest of the country in that poverty for households Figure 3 RISING INEQUALITY LED TO INSIGNIFICANT POVERTY REDUCTION and livestock production have the highest poverty rate. relying on crop and livestock as their main income source Those who depend on remittances, salaried employment, is twice (48 percent) more than other regions combined (24 Datt-Ravallion decomposition of poverty change between 2011 and 2016 and services have the lowest poverty rates. Tigray diverges percent) (Figure 5). Tigray National excl. Tigray Contribution to change in poverty 10.0 Figure 5 THOSE WHO DEPEND ON MIXED FARMING HAVE THE HIGHEST 7.6 5.0 POVERTY RATE 0.0 2.4 Poverty rates by household’s main income source, 2016 -4.8 -5.0 -6.1 Tigray National excl. Tigray -8.5 -10.0 -12.5 50% 45% -15.0 40% Poverty rate Poverty reduction Consumption growth Increase in inequality 35% 30% 25% Source: HCES, 2011; 2016; World Bank staff calculations. 20% 15% 10% 5% Average consumption in Tigray remains low for the 31,535 Birr per year on average; reflecting the strong con- 0% first four quintiles. The average annual consumption sumption growth at the top of the distribution. Tigray closely n il t es or g ns ck en ta tio rin lab vic io to re per adult equivalent in 2016 was 4,804 Birr for the poor- mirrors consumption patterns for the rest the country, though m uc tu at es oy er d al ac od on an liv -s su pl est quintile and 15,685 Birr for the fourth quintile. Average the richest quintile in Tigray is significantly better off than the uf pr /d em d ale Ca an lf an ns op Se -m les ied consumption shows a sharp increase from the fourth to the richest quintile in the rest of the country (Figure 4). io p cr ns ro ho lar lf - lf -c Pe Se -w fifth (richest) quintile, with the richest 20 percent consuming Sa Se lf Se lf Se Main source of household income Source: HCES, 2016; World Bank staff calculations. Figure 4 HOUSEHOLD CONSUMPTION REMAINS FAIRLY LOW EXCEPT FOR THE TOP QUINTILE Average household expenditures per capita per year, 2016 Tigray National excl. Tigray 35,000 Average annual consumption 30,000 25,000 20,000 15,000 10,000 5,000 0 Poorest 2 3 4 Richest Consumption quintiles Source: HCES, 2011; 2016; World Bank staff calculations. 42 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 43 Educational attainment of household heads correlates mere four percent for those with post-secondary education. NON-MONETARY WELFARE strongly with poverty rates. With a poverty rate of 30 per- Tigray closely mirrors the trend for the rest of the country, cent, poverty in Tigray is highest among households headed with the exception that in Tigray having a household head The increase in consumption expenditure was accom- proportionally more households reported food shortages in by an individual with no education or who has not completed with some primary education is not associated with lower panied by improvements in non-monetary welfare. In Tigray (12 percent) than rest of the country in 2016 (10 per- primary education. Increasing education of the household poverty compared to those headed by a person with no ed- Tigray, the percentage of households that reported a food cent). For those who faced food shortage, the average num- head leads to decreasing poverty rates, reaching eight per- ucation (Figure 6). shortage in the last 12 months before the interview was sig- ber of months with a food shortage decreased from three to cent for those who completed secondary education, and a nificantly lower in 2011 than the percentage for the rest of 2.5 in Tigray, while it remained 3.3 for the rest of the country the country (13 versus 22 percent). However, Tigray did not (Table 4). experience much improvement in this regard, and as a result Figure 6 EDUCATION IS A MAIN CORRELATE OF POVERTY Table 4 FOOD SECURITY IMPROVED BETWEEN 2011 AND 2016 Poverty rates by educational attainment of the household head, 2016 Incidence of food shortages and average duration of food shortages – food gap Tigray National excl. Tigray 35% 2011 2016 30% 30% 28% Food shortage (%) Food gap (months) Food shortage (%) Food gap (months) 30% 25% 22% Tigray 13.2 3.0 11.9 2.5 Poverty rate 21% 20% National excl. Tigray 22.1 3.3 10.3 3.3 15% 12% 13% 13% Source: HCES, 2011; 2016; World Bank staff calculations. 10% 8% 6% 4% 3% 5% 0% Ownership of household durables and access to elec- combined. The proportion of households with an improved No education Incomplete Complete Incomplete Complete Post-secondary primary primary secondary secondary tricity also improved in Tigray, though still at low levels. roof, access to electricity, and who use electricity for cooking Ownership of household durables such as television, mobile also improved. But there was no significant change in other Education level of household head phone, refrigerator, animal cart, and motorcycle increased housing quality indicators, such as the quality of wall and between 2011 and 2016, and the percentage of households floor (Figure 7). Source: HCES, 2016; World Bank staff calculations. owning these durables are more than for the other regions 44 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 45 Figure 7 OWNERSHIP OF CONSUMER DURABLES AND HOUSING CONDITIONS Figure 8 ACCESS TO IMPROVED WATER SOURCES INCREASED BETWEEN IMPROVED BETWEEN 2011 AND 2016 2011 AND 2016 Household durables, housing conditions and access to energy in 2011 and 2016, % of households Access to water in 2011 and 2016, % of households with access/characteristics with asset/characteristic Tigray National excl. Tigray A. Household Durables Piped water - compound Piped -outside Protected well/tube well Protected spring 2011 2016 Other improved Unimproved sources 70 100% 100% 60 23.07 80% 31.45 80% 36.1 2.47 47.39 50 3.47 60% 60% 12.46 40 38.8 31.82 9.44 40% 40% 15.33 8.56 30 15.61 18.19 23.65 21.5 20% 20% 20 19.53 15.01 10.72 13.86 10 0% 0% 2011 2016 2011 2016 0 Mobile Mobile Refrigerator Refrigerator Animal Cart Motorcycle Animal Cart Motorcycle TV TV Source: DHS, 2011; 2016; World Bank staff calculations. Tigray National excl. Tigray Access to improved drinking water sources18 also im- proved in Tigray. The proportion of households with ac- cess to improved water sources in Tigray increased from 69 percent to 77 percent between 2011 and 2016, mirroring B. Housing/energy improvements in the rest of Ethiopia (Figure 8). 2011 2016 Improvements were also observed in health and edu- 80 cation, although challenges remain. Infant, child and un- 70 der-five mortality rates in Tigray decreased respectively from 60 60 (per thousand) to 32, from 27 to 18, and from 85 to 49 50 between 2011 and 2016. Delivery at health facility increased 40 from 12 to 57 percent, while the percentage of fully immu- 30 nized children increased from 59 to 67 percent. The stunting 20 rate decreased from 51 percent to 39 percent. Net primary 10 enrollment, primary school completion, and gross secondary 0 enrollment rates also improved. The rest of the country also experienced similar improvement, although Tigray in general Improved wall Floor: tiles/cement Access to electricity Electricity for cooking Improved wall Floor: tiles/cement Access to electricity Electricity for cooking Improved roof Improved roof experienced better outcomes (Figure 9). Tigray National excl. Tigray Source: DHS, 2011; 2016; World Bank staff calculations. Improved drinking water sources include piped water, protected well, protected spring, rainwater and bottled water. 18  46 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 47 INEQUALITY OF OPPORTUNITY Figure 9 CHILDREN’S MORTALITY DECREASED, AND THEIR HEALTH IMPROVED BETWEEN 2011 AND 2016 2011 2016 To assess inequality of opportunity19 in Tigray and the rest of the rest of the country at 28 percent. Tigray significantly out- 100 Ethiopia, we focus on six key outcomes: (i) primary school performs the rest of Ethiopia in secondary school enrollment 84.72011 in 2011 Selected health and education variables and 2016 2016 88.3 enrollment, (ii) primary school completion, (iii) secondary rates. 28 percent of age-eligible pupils in Tigray are enrolled 80 100 88.3 68.3 school enrollment, (iv) access to electricity; (v) access to in secondary schooling, while in the rest of the country the 59.6 A. Child and84.7 59.2rates (per 1000) infant mortality 60 80 an improved water source; and (vi) access to a health post rate is under 17 percent. Differences between Tigray and the 48.9 49.2 2011 2016 68.3 40 59.6 59.2 within 5 km. Circumstance variables 20 of the household in- rest of Ethiopia across the other key outcomes show that 100 60 31.6 30.9 26.7 49.2 clude rural/urban location, gender, religion, agro-ecological Tigray generally outperforms the rest of the country. 84.7 48.9 20 88.3 20 17.9 40 80 31.6 30.9 zone, consumption quintile of the child’s household, and 26.7 68.3 The higher access and coverage rates in Tigray trans- 0 59.6 17.9 59.2 20 the household’s score on the rural accessibility index (RAI).21 20 60 Infant mortality Child mortality Under 48.9 five 49.2 Infant mortality Child mortality Under five late into higher HOI scores compared to the rest of Figure 10 presents the coverage, Human Opportunity Index 40 rate rate mortality rate rate rate mortality rate Ethiopia. For each of the key outcomes, except living within 0 31.6 30.9 (HOI), and Dissimilarity Index (D-Index) across these key out- Infant mortality 26.7 Child mortality Under five Infant mortality Child mortality Under five Tigray National excl. 20 Tigray 5km of a health post, Tigray’s HOI index is higher than in the 20 rate 17.9 rate mortality rate rate rate mortality rate come variables in 2016. rest of the country. The difference in the HOI index is particu- Tigray National excl. Tigray 0 In terms of coverage, the proportion of age-eligible larly large for primary school enrollment and completion and Infant mortality Child mortality Under five Infant mortality Child mortality Under five children (aged 7 to 14) enrolled in primary school was access to improved water. Although Tigray’s HOI for elec- rate rate mortality rate rate rate mortality rate significantly higher in Tigray than the rest of the coun- tricity access is about double that of the rest of the country, Tigray National excl. Tigray try (80 versus 71 percent). Primary school completion rates it should be remembered that electricity access rates in the Source: DHS, 2011; 2016; World Bank staff calculations in Tigray are around 46 percent, while they are much lower in region are still low, at around 36 percent. 2011 2016 80 B. Immunization and health facility delivery (percent) 67.3 2011 2016 80 56.9 58.9 60 67.3 51.4 Figure 10 TIGRAY PERFORMS WELL ON PRIMARY SCHOOL ENROLLMENT, WHILE 56.9 58.9 39.3 43.9 38.3 UNEQUAL ACCESS TO INFRASTRUCTURE REMAINS A CHALLENGE 60 40 51.42011 2016 36.2 80 24.1 43.9 Coverage, HOI, and D-index for different outcomes, 2016 39.3 21.8 38.3 40 67.3 36.2 20 11.6 56.9 58.9 9.8 60 51.4 24.1 21.8 Tigray National excl. Tigray 20 0 11.6 43.9 39.3 9.8 40 Health facility Fully Stunted Health facility Fully 36.2 38.3 Stunted 90 Coverage HOI D-index 0 delivery immunized children delivery immunized children 80 24.1 21.8 Health facility children Fully Stunted Health facility children Fully Stunted 20 70 Percent/score 11.6 delivery immunized children delivery 9.8 immunized children Tigray National excl. Tigray 60 children children 0 50 Health facility Tigray Fully Stunted National Health facility excl. Tigray Stunted Fully 40 Source: DHS, 2011; delivery immunized children 2016; World Bank staff calculations delivery immunized children 30 children children 20 Tigray C. Education (percent) National excl. Tigray 10 2011 2016 0 100 alt ter alt ter alt ter st st st ec ry ool ec ry ool ec ry ool l l l m ve ess m ve ess m ve ess nd sc l co ry ool co ry ool ity oo ity oo ty oo oo po po po he wa he a he a 2011 2016 a h a h a h ric sch ric sch ici ch im ch im ch im ch 5k pro cc 5k pro cc 5k pro cc of d w of d w 80.2 nd sc nd sc h h h s pr ry s pr ry s pr ry s a a a of d 80 73.7 co ry 100 71.3 a a a a a a En let rim En let rim En let rim 61.6 r t t t 80.2 m m m m dp p p 60 80 73.7 I I I ro ed ro ed ro ed se se se m ed m ed El El El 71.3 Co rolle 44.2 45.8 2011 2016 Co ll Co ll d d d o o in in in 61.6 lle lle lle r r p p p ith ith ith En En En 40 34.8 100 60 31.9 29.9 W W W 23.8 45.8 27.1 44.2 21.3 80.2 20 80 40 73.7 34.8 71.3 31.9 Source: HCES, 2011; 2016; World bank staff calculations. 27.1 29.9 23.8 61.6 21.3 0 60 20 Net primary Primary school 44.2 Gross 45.8 Net primary Primary school Gross 40 enrolment completion 34.8 secondary enrolment completion secondary 0 31.9 29.9 (15-24) school (15-24) 27.1 school A full description of the methodology and variables used in the construction of this index in Ethiopia can be found in the technical 19  Net primary Primary school Gross 23.8 Net primary Primary school Gross 21.3 enrolment enrolment appendix. 20 enrolment completion secondary enrolment completion secondary (15-24) Tigray school (15-24) National excl. Tigray school 20  These are generally thought of as variables over which a person does not have control, such as gender or region of birth. 0 enrolment enrolment Source: HCES, 2011; 2016; World Bank staff calculations. 21 Accessibility was also tested using a market accessibility variable, and the results were broadly consistent. Net primary Primary school Tigray Gross Primary Net primary National school excl. Tigray Gross enrolment completion secondary enrolment completion secondary 48 (15-24) school (15-24) REGIONAL ETHIOPIA school POVERTY REPORT SECTION II: REGIONAL ANALYSIS 49 enrolment enrolment The D-index for completed primary school and second- the country. The D-index was highest for access to electricity, ary school enrollment in Tigray is about the same as in exemplifying the rural-urban divide in access to these ser- Figure 12 SPENDING SHARE OF ECONOMIC SECTOR INCREASED the rest of the country, indicating that circumstances vices in the region, and more broadly throughout the country. Expenditure share by sectors (%), 2011 to 2016 play a similar role in determining these outcomes in As noted earlier, a decomposition of this index (not shown) this region than they do in the rest of Ethiopia. The D-in- reveals that access depends far more on whether a house- Tigray National excl. Tigray dex was lowest for enrollment in primary education, driven hold is in an urban or rural area than on the level of wealth of by the fact that enrollment rates for age-eligible children was the household. For example, even the poorest urban house- Administration Economic Social Other already relatively high. However, unequal coverage of infra- holds tend to have access to electricity, while even the rich- structure access remains a challenge in Tigray and the rest of est rural households do not generally have access. 2016 2016 2016 13.2 2016 26.3 15.6 30.8 17.8 4.9 PUBLIC EXPENDITURE 34.6 18 32 39.5 In the previous sections, we discussed trends in monetary on “pro-poor” sectors are also higher in Tigray. These results 2016 29.6 2016 2016 23.7 2016 and non-monetary welfare between 2011 and 2016, and are consistent with the fact that Tigray is better off in terms of 27.5 30.6 27.8 28.3 inequality of opportunities focusing on 2016 both in Tigray monetary consumption and non-monetary welfare. But pover- and the rest of the country. In this section, we discuss the ty is higher in Tigray, suggesting that expenditures on pro-poor Note: The inner circles present the 2011 values while the outer one presents the 2016 values. trends, volume, and allocation of public expenditures to iden- sectors have not yet succeeded in reaching the poorest. Source: MoF; World Bank staff calculations. tify variation across regions and explore how these link to Per capita public expenditure increased between 2011 variations in welfare. We use Ministry of Finance (MoF) data. and 2016, and is significantly higher in the Tigray re- We express all the expenditure values in 2016 prices using a gion than the rest of the country for all years. Per cap- non-food CPI deflator. We first present total per capita pub- higher in Tigray than in the rest of the country, the share of Per capita spending24 in selected “pro-poor” functional ita public spending in the Tigray region increased from Birr lic expenditure and then discuss the expenditure allocation own (regional) revenue out of total expenditure was also high- activities increased between 2011 and 2016. In the Tigray 1,907 in 2011 to Birr 2,437 in 2016 (in constant 2016 Birr), by sector. Finally, we present expenditure on selected “pro- er in Tigray compared to other regions combined. Of total region, the increase was particularly strong for water resourc- while it increased from 1,042 to 1,968 in other regions com- poor” sectors, which include agriculture and rural develop- expenditure in Tigray, 39 percent was covered by own rev- es, where per capita spending increased from ETB107 to bined. However, spending per capita grew faster in the rest ment, water resources, health, and education. enue on average over the years between 2012 to 2018, as ETB 171. The rest of the country also experienced per capita of the country. On average, per capita public expenditure in- opposed to 25 percent for other regions combined.22 spending increase on pro-poor functional activities (Figure Average per capita public expenditure in Tigray is larg- creased at an average rate of 5 percent annually in Tigray 13A). Per capita expenditure in all pro-poor sectors is higher er than the rest of the country, while share of expen- compared to 15 percent in other regions combined (Figure Public expenditure growth in Tigray was accompanied in Tigray than the rest of the country; per capita expenditure diture on administration is lower. Per capita expenditures 11). While public expenditure per capita was significantly by a spending shift toward economic sectors.23 Eco- on education is particularly higher in Tigray, reflecting the bet- nomic sector spending, as share of the total expenditure, ter education outcomes. increased from 18 to 31 percent. The shares of social and administration sectors did not change much. For the rest of In Tigray, the total share of spending on pro-poor activ- Figure 11 PER CAPITA PUBLIC EXPENDITURE INCREASED OVER TIME the country, spending on economic sector increased from ities remained stable between 2011 and 2016, while it Per capita real public expenditure, 2011 to 2016 24 percent to 28 percent, while spending share on the social declined in the rest of the country. The share of spending sector decreased from 32 to 28 percent. Administration-re- on the water sector in Tigray increased from 5 percent to Tigray National excl. Tigray lated spending significantly dropped from 40 percent to 31 9 percent, while that of education sector dropped from 21 percent, while in Tigray it did not decrease much but is rela- percent to 19 between 2011 and 2016. The spending shares 2,400 tively smaller, accounting for 16 percent in 2016 (Figure 12). of agriculture and health did not change much. In other re- 2,000 gions combined, the share of education dropped from 22 In 2016 Birr percent to 17, while the shares of the other pro-poor sectors 1,600 remained the same (Figure 13B). 1,200 800 400 2011 2012 2013 2014 2015 2016 22  World Bank and DFID, 2019. Source: MoF; World Bank staff calculations 23 Social sector includes education, health, culture and sports, and labor and social affairs while economic sector includes agriculture and rural development, water resources, trade and industry, mines and energy, transportation and communication, construction. Per capita spending refers to total spending in a particular sector divided by the population of the region. For example, per capita 24  spending on education refers to total spending on education divided by the total population of the region and not spending per student. 50 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 51 Figure 13 EXPENDITURE SHARE ON WATER RESOURCES INCREASED IN TIGRAY BETWEEN 2011 AND 2016 Real per capita expenditure and share of total expenditure (%) for selected sectors, 2011 and 2016 A. Real per capita expenditure in 2016 price 2011 2016 Afar 500 450 400 350 In 2016 Birr 300 250 INTRODUCTION 200 150 This chapter focuses on the evolution of socio-economic and health indicators, such as poverty, inequal- 100 50 ity, access to basic services, and education, in the Afar region between 2011 and 2016. Using data from 0 a variety of sources, mainly the twinned household living standards surveys (HCES and WMS)25 and the Agriculture and Rural Development Water Resources Education Health Agriculture and Rural Development Water Resources Education Health Demographic and Health Survey (DHS), the chapter begins by documenting changes in monetary poverty and living standards in the region. It then moves to analyze non-monetary welfare, before turning to the chal- lenges of inequality of opportunity. The final section discusses changes in the sectoral allocations of social spending in the region between 2011 and 2016. The chapter aims to provide policy makers with information and analysis that will be useful in improving the effectiveness of poverty reduction programs in the region. Tigray National excl. Tigray Some of the main findings include: ● The poverty rate in Afar dropped from 36 percent in the case for human capital indicators such as health and B. Expenditure share (%) 2011 to 26 percent in 2016. Poverty decreased in Afar education. But Afar lags in many of the non-monetary faster than in the rest of Ethiopia but in rural areas there indicators, particularly in health indicators. This is also Agriculture & Rural Dev. Water Resources Education Health was no statistically significant poverty reduction. reflected in a lower score on Human Opportunity Index 45 than the national average – implying lower equality of ac- 40 ● Consumption growth in Afar was strong only in urban 8.1 cess to key basic services for children in Afar compared 7.2 6.5 areas while rural areas didn’t experience significant con- 35 7.9 to the rest of Ethiopia. 30 sumption growth. The gap between rural and urban 25 18.5 22.2 areas further widened and that led to an increase in in- ● Average per capita public expenditure in Afar is larger 20 21.1 17.3 equality. By 2016 inequality in the region was higher than than in the rest of the country, though Afar spends a sig- 15 in the rest of Ethiopia. nificantly higher proportion on administration. Afar spends 10 8.6 5.5 5.4 4.8 more on agriculture and rural development and less on 5 ● Non-monetary welfare including ownership of consumer 6 6.5 6.9 6.8 education compared to all other regions combined. 0 durables and access to drinking water improved, as was 2011 2016 2011 2016 Tigray National excl. Tigray Note: The sum does not add up to 100% because it includes only shares of selected “pro-poor” sectors. Source: MoF; World Bank staff calculations. 25 Household Consumption Expenditure Survey (HCES) and Welfare Monitoring Survey (WMS) are the two official household surveys collected by CSA every five year and used to monitor welfare. HCES covers monetary welfare and WMS deals with non-monetary welfare like health and education. Both surveys are collected from the same sample of households. 52 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 53 POVERTY AND INEQUALITY Consumption growth in urban Afar was positive and the bottom, but this decreased further up the distribution, stable for most of the consumption distribution, while reaching zero or negative in the middle and then becom- Food and nonfood consumption in the Afar region in- combined. The increase in consumption can be mainly at- growth in rural areas varied. In urban areas, consumption ing positive at the top (Figure 1). The fact that there was no creased significantly between 2011 and 2016, though tributed to urban areas, where consumption increased by growth was positive and strong across the consumption dis- significant poverty reduction in rural areas implies that con- only urban areas experienced the full benefit of this more than one-third between 2011 and 2016, a comparable tribution except for the poorest, where it was either relatively sumption for the poor was so low in 2011 that growth was increase. Mean consumption per adult increased by 21 increase as for the rest of the country. Change in consump- weak or not statistically different from zero. Rural areas, on not enough to significantly reduce poverty. percent in Afar compared to 14 percent for all other regions tion in rural Afar was not statistically significant (Table 1). the other hand, experienced strong consumption growth at Table 1 CONSUMPTION INCREASED BETWEEN 2011 AND 2016 BOTH IN AFAR AND THE REST OF THE COUNTRY Figure 1 GICS FOR AFAR URBAN AND RURAL Mean annual household consumption per adult equivalent in December 2015 prices, 2011 and 2016 Urban Rural TOTAL URBAN RURAL 2011 2016 % change 2011 2016 % change 2011 2016 % change Afar 10,641 12,902 21.2* 13,945 18,645 33.7* 9,298 10,463 12.5 National excl. 11,010 12,495 13.5* 13,901 18,649 34.2* 10,439 11,020 5.6* Afar Note: * represents changes that are statistically significant. Source: HCES, 2011; 2016; World Bank staff calculations. The increase in consumption in Afar led to a decrease from 24 to 11 percent), while we see no significant differ- in poverty, but only in urban areas of the country. The ence in rural poverty between 2011 and 2016. The decrease proportion of people below the national poverty line in Afar in poverty of about 11 percentage points was faster in Afar decreased from 36 percent in 2011 to 26 percent in 2016. compared to other regions combined, where poverty de- In urban areas, poverty decreased by more than half (falling creased by 6 percentage points (Table 2). Source: HCES, 2011; 2016; World Bank staff calculations. Table 2 POVERTY DECREASED IN AFAR FASTER THAN IN THE REST OF THE COUNTRY 26 Percentage of people below the national poverty line TOTAL URBAN RURAL % pt. % pt. % pt. 2011 2016 change 2011 2016 change 2011 2016 change Photo: World Bank/Sylwia Pecio Photo: World Bank/Sylwia Pecio Afar 36.1% 25.6% -10.5* 23.7% 10.6% -13.1* 41.1% 31.9% -9.2 National excl. 29.6% 23.4% -6.1* 25.7% 14.6% -11.0* 30.3% 25.5% -4.8* Afar Note: * represents changes that are statistically significant. Source: HCES, 2011; 2016; World Bank staff calculations. 26 In 2011, the HCES covered only two zones in Afar. In 2016, the coverage was improved – five zones were covered respectively in Afar. While the 2016 data is more representative of pastoral areas and the two regions, to maintain comparability households sampled from the new zones are excluded from the analysis. If households from the new zones are included in the 2016 analysis, overall poverty in Afar becomes 23.6% and rural poverty becomes 26.5% while urban poverty remains 10.6%. 54 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 55 Relatively stronger and stable consumption growth in between 2011 and 2016. The ratio of consumption at the 90th Given the increase in inequality between 2011 and on poverty in Afar is negative (poverty-reducing). The redistri- urban as opposed to rural areas, where consumption percentile and consumption at the 10th percentile increased 2016, poverty reduction in Afar can be entirely attribut- bution effect is positive not only for Afar but also for the rest was low or zero for most of the distribution, helped in- from 3.8 in 2011 to 4.5 in 2016, reflecting strong consumption ed to increases in consumption. Figure 3 decomposes of the country. If there were no changes in the distribution crease inequality. The Gini index increased from 31 to 37 growth at the top relative to the bottom (Table 3). the change in poverty into a growth and redistribution com- between 2011 and 2016, poverty would have decreased by ponent using the Datt-Ravallion decomposition technique. It 19 and 9 percent in Afar and the rest of the country, respec- indicates which part of the poverty change can be attributed tively, instead of the 13 and 6 percent we actually observe. Table 3 INEQUALITY INCREASED BETWEEN 2011 AND 2016 to: (i) the change in average consumption (growth effect), and In other words, if the consumption distributions in 2011 were (ii) the change in the redistribution of consumption; that is, maintained in 2016, poverty reduction could have been even changes in inequality (redistribution effect). The growth effect stronger in Afar compared to the rest of Ethiopia (Figure 3). 2011 2016 Gini index p90/p10 ratio Gini index p90/p10 ratio Afar 30.5 3.8 37.0 4.5 National 29.8 3.6 32.9 4.3 Figure 3 CONSUMPTION GROWTH DROVE POVERTY REDUCTION Source: HCES, 2011; 2016; World Bank staff calculations. Datt-Ravallion decomposition of poverty change between 2011 and 2016 Afar National excl. Afar The rise in inequality between 2011 and 2016 is mainly – the part of inequality resulting from differences in average Contribution to change in poverty due to the increasing gap between urban and rural ar- consumption between urban and rural areas – increased from 10.0 eas. Average consumption was already significantly higher in 31 percent in 2011 to 37 percent in 2016. On the other hand, 5.0 6.4 urban than rural areas in 2011. The gap further increased in the “within” share of inequality – the part resulting from chang- 2.6 0.0 2016, with urban consumption showing strong growth and ru- es in distribution within urban and rural areas – diminished -5.0 -6.0 ral consumption also growing, but at about one-third the rate from 50 percent to 46 percent (Figure 2). -8.6 of urban areas. As a result, the “between” share of inequality -10.0 -12.5 -15.0 -18.9 -20.0 -25.0 Figure 2 INEQUALITY INCREASED DUE TO THE INCREASING GAP BETWEEN Poverty reduction Consumption growth Increase in inequality URBAN AND RURAL AREAS Source: HCES, 2011; 2016; World Bank staff calculations. Decomposition of the Gini coefficient into a between rural-urban component and a within-component Overlap Within Between Relative contribution to Gini (%) 100% Average consumption in Afar remains low for the first per year on average, reflecting strong consumption growth at 90% 18.9 16.6 four quintiles. The average annual consumption per adult the top of the distribution. Afar closely mirrors consumption 80% 70% 49.8 46.4 equivalent in 2016 was 5,546 Birr for the poorest quintile patterns for the rest the country, though the richest quintile in 60% and 14,030 Birr for the fourth quintile. Average consumption Afar outstrips the richest quintile in the rest of the country by 50% shows a sharp increase from the fourth to the fifth (richest) around 3,200 Birr per year, on average (Figure 4). 40% 30% quintile, with the richest 20 percent consuming 29,070 Birr 37.0 20% 31.2 10% 0% 2011 2016 Source: HCES, 2011; 2016; World Bank staff calculations. 56 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 57 Educational attainment of household heads strongly of the household head leads to decreasing poverty rates, Figure 4 HOUSEHOLD CONSUMPTION REMAINS FAIRLY LOW EXCEPT FOR THE negatively correlates with poverty. With a poverty rate reaching zero in households where the head has completed TOP QUINTILE of 30 percent, poverty in Afar is highest among households secondary or post-secondary education. Afar closely mirrors headed by a person with no education. Increasing education the trend for the rest of the country (Figure 6). Average household expenditures per capita per year, 2016 Afar National excl. Afar Figure 6 EDUCATION IS A MAIN CORRELATE OF POVERTY Average annual consumption 35,000 30,000 Poverty rates by educational attainment of the household head, 2016 25,000 Afar National excl. Afar 20,000 15,000 35% 30% 30% 28% 10,000 25% 22% Poverty rate 5,000 0 20% Poorest 2 3 4 Richest 15% 12% 12% 13% 13% Consumption quintiles 10% 6% Source: HCES 2016; World Bank staff calculations. 3% 3% 5% 0% 0% 0% No education Incomplete Complete Incomplete Complete Post-secondary In Afar, households that primarily depend on livestock particularly high for those who depend on livestock produc- primary primary secondary secondary production and casual labor are the poorest. With 31 tion. Those who depend on wholesale and retail and services Education level of household head percent of the people living below the poverty line, poverty is have the lowest poverty rates (Figure 4). Source: HCES 2016; World Bank staff calculations. Figure 5 CASUAL LABORERS AND LIVESTOCK PRODUCERS HAVE THE HIGHEST Poverty reduced at a faster rate in Afar than in the rest growth was strong and positive throughout the bottom 40 POVERTY RATES of Ethiopia. Figure 9 shows that of the 11 percentage point percent of households (see Figure 2) in both urban and rural Poverty rates by household’s main income source, 2016 reduction in poverty in Afar, 7 points are attributed to rural Afar (Figure 7). areas and 4 to urban areas. This is not surprising given that Afar National excl. Afar 35% 30% Figure 7 POVERTY REDUCTION IN AFAR LOOKED A LOT LIKE POVERTY 25% REDUCTION IN THE REST OF ETHIOPIA Poverty rate 20% Rural-urban decomposition of poverty changes 2011 to 2016 15% Rural Urban Population shift Interaction 10% Afar National excl. Afar 5% 0 Change in poverty headcount 0% -2 Casual labor Self - Self - Salaried Self - services livestock keeping wholesale and retail employment -4 Main source of household income -6 -8 Source: HCES 2016; World Bank staff calculations. Note: Only main income groups with at least 100 observations in Afar are included. -10 -12 Source: HCES, 2011; 2016; World Bank staff calculations. 58 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 59 When decomposing poverty reduction by sector, we Afar over the period. The importance of this shift is illustrated NON-MONETARY WELFARE see that Afar differs substantially from the rest of Ethi- by how quickly the share of the population in Afar relying on opia. While over 60 percent of poverty reduction in the rest agriculture fell over the period. In 2011, 68 percent of house- Increase in consumption expenditure was accompa- food shortage dropped from five percent to zero. There was of Ethiopia can be explained by improvements in the agricul- holds in Afar relied primarily on agriculture or livestock as the nied by improvements in non-monetary welfare. In Afar, lower food shortage in Afar compared to other regions com- tural sector, the corresponding figure for Afar is only 39 per- primary source of income, while in 2016 this had fallen to the percentage of households that reported a food shortage bined, where the percentage of households who faced food cent. The other major variable explaining poverty reduction less than half. At the same time, the share of households in the last 12 months before the interview decreased from shortage dropped from 22 to 10 between 2011 and 2016; the in Afar is the so-called “population shift” effect. This variable, relying primarily on services increased substantially – from 21 eight percent in 2011 to five percent in 2016. For those who average month with food shortage remained three (Table 4). often used as an indicator of structural transformation (prox- percent to 34 percent – in strong contrast to the rest of the faced food shortage, the average number of months with a ied by shift in population from agriculture to manufacturing or country (Figure 8). services) explained about 36 percent of poverty reduction in Table 4 FOOD SECURITY IMPROVED BETWEEN 2011 AND 2016 Incidence of food shortages and average duration of food shortages – food gap Figure 8 THE AGRICULTURE SECTOR REMAINS THE LARGEST CONTRIBUTOR TO 2011 2016 POVERTY REDUCTION IN AFAR Food shortage (%) Food gap (months) Food shortage (%) Food gap (months) Sectoral decomposition of poverty changes 2011 to 2016 Afar 7.7 5.2 4.6 0.3 National excl. Afar 21.7 3.2 10.4 3.2 Agriculture Manufacturing Construction Services Other Population shift Interaction Source: HCES, 2011; 2016; World Bank staff calculations. Afar National excl. Afar 0 Change in poverty headcount -2 Ownership of household durables and access to elec- The proportion of households with an improved wall, im- -4 tricity also improved in Afar, though they remain at low proved roof, and improved floor also increased slightly, while -6 levels. Ownership of television improved from 15 percent to access to electricity did not improve. The percentage of those 18 percent, while mobile phone ownership increased from 24 who use electricity for cooking also improved, though still low -8 percent to 58 percent between 2011 and 2016. But own- at only two percent. Access to energy and housing quality -10 ership of refrigerators, animal carts, and motorcycles hardly improved (Figure 9). -12 changed in Afar, while it improved in the rest of the country. Source: HCES, 2011; 2016; World Bank staff calculations. Photo: World Bank/Gelila Woodeneh 60 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 61 Figure 9 OWNERSHIP OF CONSUMER DURABLES AND HOUSING CONDITIONS Figure 10 ACCESS TO IMPROVED WATER SOURCES INCREASED BETWEEN IMPROVED BETWEEN 2011 AND 2016 2011 AND 2016 Household durables, housing conditions and access to energy in 2011 and 2016, % of households Access to water in 2011 and 2016, % of households with access/characteristics with asset/characteristic Afar National excl. Afar A. Household Durables Piped water - compound Piped -outside Protected well/tube well Protected spring 2011 2016 Other improved Unimproved sources 70 100% 100% 60 53.5 43.21 46.26 35.11 50 80% 80% 40 60% 2.26 60% 11.83 30 4.71 9.1 0.81 17.11 40% 8.99 23.67 40% 10.15 20 15.04 23.35 21.05 10 20% 24.34 20% 19.89 0 14.18 10.94 Mobile Mobile Refrigerator Refrigerator Animal Cart Motorcycle Animal Cart Motorcycle 0% 0% TV TV 2011 2016 2011 2016 Source: DHS, 2011; 2016; World bank staff calculations. Afar National excl. Afar Access to improved drinking water sources27 also im- mortality increased from 49 to 63 and under-five mortality B. Housing/energy proved in Afar. The proportion of households with access hardly changed. Delivery at health facility increased from 7 2011 2016 to improved water sources in Afar increased from 54 percent to 15 percent, while the percentage of fully immunized chil- to 65 percent between 2011 and 2016, mirroring improve- dren increased from 9 percent to 15 percent. The stunting 70 ments for the rest of Ethiopia (Figure 10). rate decreased from 50 percent to 41 percent. Net primary 60 enrollment, primary school completion, and gross second- 50 Improvements were also observed in health and ed- ary enrollment rates also improved.28 The rest of the country ucation, although challenges remain. Child mortality 40 improved in all indicators. Afar lags in all health-related indi- rate decreased from 63 per thousand to 49, while infant 30 cators (Figure 11). 20 10 0 Improved wall Floor: tiles/cement Access to electricity Electricity for cooking Improved wall Floor: tiles/cement Access to electricity Electricity for cooking Improved roof Improved roof Afar National excl. Afar 27  Improved drinking water sources include piped water, protected well, protected spring, rainwater and bottled water Source: DHS, 2011; 2016; World bank staff calculations. 28 In 2011, the WMS covered only two zones in Afar. In 2016, the coverage was improved – five zones were covered respectively in Afar. While the 2016 data is more representative of pastoral areas and the two regions, to maintain comparability households sampled from the new zones are excluded from the analysis. If households from the new zones are included in the 2016 analysis, the values of the education indicators become lower, for example, gross secondary enrollment becomes 31 percent instead of 42 percent. 62 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 63 INEQUALITY OF OPPORTUNITY Figure 11 CHILDREN’S MORTALITY DECREASED, AND THEIR HEALTH IMPROVED BETWEEN 2011 AND 2016 2011 2016 To assess inequality of opportunity29 in Afar and the rest of the HOI scores for these key outcomes are also low (see 120 108.7 Ethiopia, we focus on six key outcomes: (i) primary school the Coverage and HOI panels in Figure 14). The D-index for Selected health and education 108.22011 in 2011 variables and 2016 2016 100 enrollment, (ii) primary school completion, (iii) secondary these variables – measuring how much these outcomes dif- 120 87.8 108.2 108.7 school enrollment, (iv) access to electricity; (v) access to fer according to circumstances – is correspondingly high, al- 80 A. Child and infant mortality rates (per 1000) 66.6 100 62.6 62.7 59.3 87.8 an improved water source; and (vi) access to a health post though similar to the rest of the country. This implies strong 60 48.6 49.2 2011 201647.9 80 within 5 km. Circumstance variables 30 of the household in- differences between households in which children complete 120 62.6 62.7 66.6 40 108.2 108.7 59.3 30.3 clude rural/urban location, gender, religion, agro-ecological primary school, enroll in secondary school, and have access 60 48.6 49.2 47.9 19.6 100 20 87.8 zone, consumption quintile of the child’s household, and to electricity and those that do not. Decomposing this dif- 40 30.3 80 0 19.6 66.6 the household’s score on the rural accessibility index (RAI).31 ference (not shown) reveals that this is largely explained by 20 62.6 62.7 59.3 60 Infant mortality Child mortality Under five Infant mortality Child mortality Under five Figure 12 presents the coverage, Human Opportunity Index rural-urban differences in access rates, both within Afar and 48.6 49.2 47.9 0 rate rate mortality rate rate rate mortality rate 40 30.3 (HOI), and Dissimilarity Index (D-Index) across these key out- in the rest of Ethiopia. Infant mortality Child mortality Afar Under five Infant mortality Child National mortality excl. Afar Under five rate rate mortality rate rate 19.6 rate mortality rate come variables in 2016. 20 About one in six children age-eligible for secondary Afar National excl. Afar 0 In terms of coverage, Afar lags the rest of Ethiopia on school (15 to 18) were enrolled in 2016. This indicates Infant mortality Child mortality Under five Infant mortality Child mortality Under five all key outcomes, except for electricity access and ac- that educational opportunity is quite different for the older co- rate rate mortality rate rate rate mortality rate cess to an improved water source. While primary school hort of children (15-18) than the younger cohort (7-14). More Afar National excl. Afar completion and secondary school enrollment are low all over than two-thirds of age-eligible primary school children in Afar Source: DHS, 2011; 2016; World Bank staff calculations Ethiopia, they are particularly low in Afar. This means that are enrolled, yet just 29 percent of children in the region (and 2011 2016 60 B. Immunization and health facility delivery (percent) 50.2 2011 2016 50 44.4 60 41.1 38.8 38.3 Figure 12 AFAR PERFORMS WELL ON ACCESS TO IMPROVED WATER, WHILE 40 50.2 50 44.4 UNEQUAL ACCESS TO INFRASTRUCTURE REMAINS A CHALLENGE 30 41.1 2011 201626.4 40 24.4 38.8 38.3 60 Coverage, HOI, and D-index for different outcomes, 2016 20 14.7 15.2 30 50.2 26.4 50 10 24.4 6.8 8.6 44.4 10 41.1 Afar National excl. Afar 20 14.7 15.2 38.8 38.3 40 10 0 8.6 10 6.8 80 30 Health facility Fully Stunted facility Health26.4 Fully Stunted Coverage HOI D-index 24.4 0 delivery immunized children delivery immunized children 70 20 Health 14.7 facility children 15.2 Fully Stunted Health facility children Fully Stunted 60 Percent/score 8.6 10 10 delivery 6.8 immunized children delivery immunized children Afar children National excl. Afar children 50 0 40 Health facility Afar Fully Stunted National Health facility excl. Afar Fully Stunted 30 Source: DHS, 2011; delivery immunized children 2016; World Bank staff calculations delivery immunized children children children 20 10 Afar C. Education (percent) National excl. Afar 2011 2016 0 alt ter alt ter alt ter st st st ec ry ool ec ry ool ec ry ool in Imp y ac ol in mp y ac ol in mp y ac ol m ve ess m ve ess m ve ess nd sc l co ry ool co ry ool 80 oo po po po cit ho cit ho cit ho 71.6 71.8 he wa he a he a a h a h a h im ch im ch im ch 2011 2016 5k ro c 5k ro c 5k ro c of d w of d w nd sc nd sc tri sc tri sc tri sc h h h pr ry s pr ry s pr ry s of d 62.4 co ry 80 a a a a a a 60 53.4 71.6 71.8 En let rim im En let rim pr m dp p 62.4 I I ro ed ro ed ro ed 42.3 se se se m d m d El El El Co rolle e e 60 En et Co oll Co oll 2011 2016 d d d 40 53.4 32.7 lle lle lle pl r r p p 31.1 30.7 ith ith ith En En En 27.7 80 W W W 71.6 22.9 42.3 71.8 21.5 40 17.7 20 31.1 62.4 32.7 30.7 27.7 Source: HCES, WMS 2011; 2016; World Bank staff calculations. 60 53.4 22.9 21.5 17.7 20 0 42.3 40 Net primary Primary school Gross Net primary Primary school Gross 31.1 32.7 30.7 enrolment completion secondary enrolment completion 27.7 secondary 0 (15-24) school 22.9 (15-24) school 21.5 A full description of the methodology and variables used in the construction of this index in Ethiopia can be found in the technical 29  Net primary Primary school 17.7 Gross Net primary Primary school Gross 20 enrolment enrolment appendix. enrolment completion secondary enrolment completion secondary (15-24) Afar school (15-24) National excl. Afar school 30  These are generally thought of as variables over which a person does not have control, such as gender or region of birth. 0 enrolment enrolment Source: WMS, 2011; 2016; World Bank staff calculations. 31  Accessibility was also tested using a market accessibility variable, and the results were broadly consistent. Net primary Afarschool Primary Gross Net primary National Primary school excl. Afar Gross enrolment completion secondary enrolment completion secondary 64 (15-24) school school (15-24) REGIONAL ETHIOPIA POVERTY REPORT SECTION II: REGIONAL ANALYSIS 65 enrolment enrolment in the rest of the country) complete primary school, according Unequal infrastructure coverage remains a challenge to 2016 data. For enrollment in secondary schooling, both in Afar and the rest of the country. The D-index was Figure 13 PER CAPITA PUBLIC EXPENDITURE INCREASED OVER TIME Afar and the rest of the country have very low rates, with Afar highest for access to electricity, exemplifying the rural-ur- Per capita real public expenditure, 2011 to 2016 faring slightly worse (16 percent compared to 17 percent). ban divide in access to these services in the region, and throughout the country. As noted, a decomposition of this The D-index for completed primary school and sec- Afar National excl. Afar index (not shown) reveals that access depends far more on ondary school enrollment in Afar is higher than in the whether a household is in an urban or rural area than on level rest of the country, indicating that circumstances play 2,800 of household wealth. For example, in urban areas, even the more of a role in determining outcomes in Afar. The 2,400 In 2016 Birr poorest households tend to have access to electricity, while 2,000 D-index was lowest for enrollment in primary education in in rural areas even the richest households do not generally 1,600 Afar, driven by the fact that enrollment rates for age-eligible have access. 1,200 children was already relatively high. 800 400 2011 2012 2013 2014 2015 2016 PUBLIC EXPENDITURE Source: MoF; World Bank staff calculations. In previous sections, we discussed monetary and non-mon- Per capita public expenditure increased between 2011 etary welfare trends between 2011 and 2016, and inequality and 2016 in Afar, and generally higher than the rest Figure 14 SPENDING SHARE ON ECONOMIC SECTOR INCREASED of opportunities focusing on 2016, both in Afar and the rest of the country, particularly for earlier years. Per capita of the country. In this section, we discuss the trends, volume, public spending in the Afar region decreased from Birr 2,573 Expenditure share by sectors (%), 2011 to 2016 and allocation of public expenditures to identify variations per person in 2011 to Birr 2,281 per person in 2016 (in con- across regions and how these variations may link to welfare. stant 2016 Birr), with the lowest per capita expenditure of Birr Afar National excl. Afar We use data MoF data. We express all expenditure values in 2,130 recorded in 2012. However, spending per capita grew Administration Economic Social Other 2016 prices using the non-food CPI deflator. We first present in the rest of the country, leading almost to a convergence in total per capita public expenditure and then discuss expen- 2016. On average, per capita public expenditure decreased 2016 2016 14.4 2016 diture allocation by sector. Finally, we present expenditure on at an average rate of one percent annually in Afar while it 21.3 29.3 select “pro-poor” sectors, which include agriculture and rural increased at 14 percent annually in other regions combined 20.7 2016 7.4 development, water resources, health, and education. (Figure 13). 37.3 44.6 32.1 21.7 57.6 Although average per capita public expenditure in Public expenditure growth in Afar accompanied a 23.3 2016 2016 2016 Afar is larger than the rest of the country, the region spending shift away from administration and towards 34.2 28.0 28.3 spends a significantly higher proportion on adminis- economic sectors.32 Economic sector spending, as share tration. While the share of administration remained high in of the total expenditure, increased from 22 to 34 percent, Note: The inner circles present the 2011 values. Afar, a shift in favor of social and economic sectors occurred while the share of social sector spending hardly changed. Source: MoF; World Bank staff calculations. between 2011 and 2016. Spending on agriculture and rural For the rest of the country, spending on economic sectors in- development increased significantly in Afar, and per capita creased from 23 to 28 percent, while spending share on the spending for the sector in 2016 was almost four times higher social sector decreased from 32 to 28 percent. Administra- than the rest of the country. Per capita spending on educa- tion-related spending decreased both in Afar and the rest of from Birr 179 to Birr 632 between 2011 and 2016. The rest of spending on agriculture and rural development water and tion was significantly lower in Afar compared to other regions the country, but it is still large, particularly in Afar, accounting the country also witnessed per capita spending on pro-poor health sectors in Afar increased by almost three-fold from 8 combined, while per capita spending on health is modestly for 45 percent in 2016 (Figure 14). functional activities. Per capita spending in 2016 is higher in percent to 22 percent between 2011 and 2016.34 The share higher. These results are not consistent with the non-mone- Afar compared to other regions combined for all pro-poor of water resources dropped from nine percent to five per- Per capita spending33 in selected “pro-poor” functional activities except education, where per capita spending is cent while the shares of education and health did not change tary welfare, where Afar fares lower in most health indicators, activities increased between 2011 and 2016. In the Afar higher in the rest of the country (Figure 15A). much. In other regions combined, the share of education while being about the same as the rest of the country in ed- region, the increase was particularly strong for agriculture dropped from 22 percent to 18 while the shares of the other ucation indicators. and rural development, where per capita spending increased In Afar, the total share of spending on pro-poor ac- activities did not change much (Figure 15B). tivities also increased between 2011 and 2016, while it contracted in the rest of the country. The share of 32 Social sector includes education, health, culture and sports, and labor and social affairs while economic sector includes agriculture and rural development, water resources, trade and industry, mines and energy, transportation and communication, and construction. 33 Per capita spending refers to total spending in a particular sector divided by the population of the region. For example, per capita spending on education refers to total spending on education divided by the total population of the region and not spending per 34 It is remarkable that rural poverty in Afar did not decrease despite high spending on agriculture and rural development, suggesting student. that it has not yet benefited the very poor. 66 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 67 Figure 15 EXPENDITURE SHARE ON AGRICULTURE AND RURAL DEVELOPMENT INCREASED IN AFAR BETWEEN 2011 AND 2016 Real per capita expenditure and share of total expenditure (%) for selected sectors, 2011 and 2016 A. Real per capita expenditure in 2016 price 2011 2016 Amhara 500 450 400 350 In 2016 Birr 300 250 INTRODUCTION 200 150 This chapter focuses on the evolution of socio-economic and health indicators such as poverty, inequality, 100 50 access to basic services, and education, in the Amhara region between 2011 and 2016. Using data from 0 a variety of sources, mainly the twinned household living standards surveys (HCES and WMS)35 and the Agriculture and Rural Development Water Resources Education Health Agriculture and Rural Development Water Resources Education Health Demographic and Health Survey (DHS), the chapter begins by documenting changes in monetary poverty and living standards in the region. It then moves to an analysis of non-monetary welfare before turning to challenges of inequality of opportunity. The final section discusses changes in sectoral allocations of social spending in the region from 2011 to 2016. The chapter aims to provide policy makers with information and analysis that will be useful in improving the effectiveness of poverty reduction programs in the region. Afar National excl. Afar Some of the main findings include: ● Although the poverty rate in Amhara decreased from and education. Amhara’s Human Opportunity Index was B. Expenditure share (%) 31 percent to 26 percent between 2011 and 2016, this in general similar to the rest of the country – implying change was not statistically significant. that children in Amhara had a similar level of equality of Agriculture & Rural Dev. Water Resources Education Health access to key basic services as children elsewhere in 50 ● Consumption growth in urban Amhara was strong while Ethiopia. 45 there was no consumption growth in rural Amhara. The 7.6 40 8.1 gap between rural and urban areas widened further, ● Average per capita public expenditure in Amhara is sim- 35 7.2 12.2 7.8 leading to increase in inequality. ilar to the rest of the country, though Amhara spends 30 22.4 12.8 17.5 a significantly higher proportion on administration. Per 25 4.9 ● Non-monetary welfare including ownership of consum- 20 capita expenditures on health and education are also 21.7 er durables and access to drinking water improved, as 15 9.0 higher in Amhara. 10 5.4 5.6 was the case for human capital indicators such as health 5 7.7 6.8 6.4 0 2011 2016 2011 2016 Afar National excl. Afar Note: The sum does not add up to 100% because it includes only shares of selected “pro-poor” sectors. Source: MoF; World Bank staff calculations. 35 Household Consumption Expenditure Survey (HCES) and Welfare Monitoring Survey (WMS) are the two official household surveys collected by CSA every five year and used to monitor welfare. HCES covers monetary welfare and WMS deals with non-monetary welfare like health and education. Both surveys are collected from the same sample of households. 68 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 69 POVERTY AND INEQUALITY Consumption growth in urban Amhara was positive for distribution, reaching eight percent at the top one-third. In the entire consumption distribution, while there was no rural areas, on the other hand, consumption growth is not Food and nonfood consumption in the Amhara region attributed to urban areas, where consumption increased by growth in rural areas across the distribution. In urban significantly different from zero for most of the distribution increased significantly between 2011 and 2016, though more than 50 percent, higher than the increase of 30 percent areas, consumption growth was positive and strong across except the middle, where growth is negative (Figure 1). only urban areas experienced the full benefit of this at the national level. Rural areas, however, were left behind. the consumption distribution, and increases further up the increase. Mean consumption per adult increased by 13 Consumption in rural areas remained largely the same be- percent in Amhara compared to 14 percent for all other re- tween 2011 and 2016, compared to increases of seven per- gions combined. The increase in consumption can be solely cent for the rest of the country (Table1). Figure 1 GIC FOR AMHARA URBAN AND RURAL Table 1 CONSUMPTION INCREASED BETWEEN 2011 AND 2016 BOTH IN AMHARA Urban Rural AND THE REST OF THE COUNTRY Mean annual household consumption per adult equivalent in December 2015 prices, 2011 and 2016 TOTAL URBAN RURAL 2011 2016 % change 2011 2016 % change 2011 2016 % change Amhara 10,944 12,340 12.8* 14,325 21,879 52.7* 10,464 10,557 0.9 National excl. 11,029 12,551 13.8* 13,806 17,874 29.5* 10,423 11,167 7.1* Amhara Note: * represents changes that are statistically significant. Source: HCES, 2011; 2016; World Bank staff calculations. The increase in consumption in Amhara helped de- and 2016. The decrease in poverty of 4 percentage points crease poverty, but less so compared to the rest of the was slower in Amhara compared to other regions combined, country. The proportion of people below the national pov- 36 where poverty decreased by 7 percentage points (Table 2). erty line in Amhara decreased by 4 percentage points from The smaller reduction in poverty in Amhara is driven by the Source: HCES, 2011; 2016; World bank staff calculations. 31 percent in 2011 to 26 in 2016. In urban areas, poverty lack of progress in rural areas, while urban areas performed decreased by more than a half (falling from 29 to 12 percent), better than the regional average. while we see no difference in poverty rates between 2011 The lack of consumption growth in rural areas with ratio of consumption at the 90th percentile and consumption strong growth in urban areas, particularly for the top of at the 10th percentile increased from 3.5 in 2011 to 4.1 in the distribution, led to an increase in inequality. The Gini 2016, reflecting strong consumption growth at the top and Table 2 POVERTY DECREASED IN AMHARA BUT LESS THAN THE REST OF THE index increased from 29 to 34 between 2011 and 2016. The lack of growth at the bottom (Table 3). COUNTRY Percentage of people below the national poverty line Table 3 INEQUALITY INCREASED BETWEEN 2011 AND 2016 TOTAL URBAN RURAL % pt. % pt. % pt. 2011 2016 2011 2016 change 2011 2016 change 2011 2016 change Gini index p90/p10 ratio Gini index p90/p10 ratio Amhara 30.5% 26.1% -4.4* 29.2% 11.6% -17.6* 30.7% 28.8% -1.9 Amhara 29.6 3.5 34.3 4.1 National excl. 29.3% 22.6% -6.7* 24.9% 15.3% -9.6* 30.3% 24.5% -5.8* Amhara National 29.8 3.6 32.9 4.3 Note: * represents changes that are statistically significant. Source: HCES, 2011; 2016; World Bank staff calculations. Source: HCES, 2011; 2016; World Bank staff calculations. 36 The relatively slower poverty reduction in Amhara could partly be due the higher rural population share (84% for Amhara versus 81% for the rest of the country in 2016). Poverty reduction was low in rural areas in general. 70 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 71 The rise in inequality between 2011 and 2016 is main- inequality – the part resulting from changes in distribution – to: (i) the change in average consumption (growth effect), and reduction could have been stronger in Amhara than in the ly due to the increasing gap between urban and rural within urban and rural areas diminished from 70 percent to (ii) the change in the redistribution of consumption; that is, rest of the country (Figure 3). areas. Average consumption was already significantly high- 55 percent (Figure 2). changes in inequality (redistribution effect). The growth ef- Average consumption in Amhara remains low for the er in urban than in rural areas in 2011. The gap further in- fect on poverty in Amhara is negative (poverty-reducing). The Given increase in inequality between 2011 and 2016, first four quintiles. The average annual consumption per creased in 2016, with urban consumption showing strong redistribution effect is positive not only for Amhara but also poverty reduction in Amhara can be entirely attributed adult in 2016 was 5,068 Birr for the poorest quintile and growth and rural consumption hardly changing. As a result, for the rest of the country. If there were no changes in the to increases in consumption. Figure 5 decomposes the 13,102 Birr for the fourth quintile. Average consumption the “between” share of inequality – the part of inequality re- distribution between 2011 and 2016, poverty would have change in poverty into a growth and redistribution compo- shows a sharp increase from the fourth to the fifth (richest) sulting from differences in average consumption – between decreased by 11.2 and 8 percent in Amhara and the rest nent using the Datt-Ravallion decomposition technique. It quintile, with the richest 20 percent consuming 26,648 Birr urban and rural areas increased from 13 percent in 2011 to of the country, respectively, instead of the 4.4 and 6.5 per- indicates which part of the poverty change can be attributed per year on average, reflecting strong consumption growth at 35 percent in 2016. On the other hand, the “within” share of cent we actually observe. In other words, if the consumption the top of the distribution. Amhara closely mirrors consump- distributions in 2011 had been maintained in 2016, poverty tion patterns for the rest the country (Figure 4). Figure 2 INEQUALITY INCREASED DUE TO THE INCREASING GAP BETWEEN URBAN AND RURAL AREAS Figure 4 HOUSEHOLD CONSUMPTION REMAINS FAIRLY LOW EXCEPT FOR THE Decomposition of the Gini coefficient into a between rural-urban component and a within-component TOP QUINTILE Overlap Within Between Average household expenditures per capita per year, 2016 Relative contribution to Gini (%) 100% Amhara National excl. Amhara 90% 17.4 9.4 80% 55.2 30,000 69.7 Average annual consumption 70% 60% 25,000 50% 40% 20,000 30% 35.4 15,000 20% 10% 10,000 12.8 0% 2011 2016 5,000 Source: HCES, 2011; 2016; World Bank staff calculations 0 Poorest 2 3 4 Richest Consumption quintiles Source: HCES, 2011; 2016; World Bank staff calculations. Figure 3 CONSUMPTION GROWTH DROVE POVERTY REDUCTION Datt-Ravallion decomposition of poverty change between 2011 and 2016 Amhara National excl. Amhara In Amhara, households that primarily depend on man- which includes small manufacturing businesses such furni- ufacturing, casual labor, and crop production have the ture shops, bakery, and injera making. Countrywide, 55 per- Contribution to change in poverty 8.0 highest poverty rates. Those who depend on salaried em- cent of Ethiopians work informally (Figure 5). 6.0 6.8 ployment and wholesale and retail have the lowest poverty 4.0 2.0 Educational attainment of the household head strongly rates. Amhara diverges from the rest of the country in that 0.0 1.5 negatively correlates with poverty rates. With a pover- poverty rates for households relying on manufacturing as -2.0 -4.4 ty rate of 31 percent, poverty in Amhara is highest among -4.0 their main income source are more than double (32 percent) -6.0 -6.5 households headed by someone with no education. Increas- -8.0 of the other regions combined (14 percent). This likely is due -8.0 ing education of the household head leads to decreasing -10.0 -11.2 to the fact that proportionally more of those who depend on poverty rates, reaching merely three percent in households -12.0 manufacturing in Amhara work in the informal sector relative -14.0 headed by someone who completed secondary education, to the rest of the country. In Amhara, 75 percent of the those Poverty reduction Consumption growth Increase in inequality and zero for those with post-secondary education. Amhara who primarily depend on manufacturing work informally, closely mirrors the trend for the rest of the country (Figure 6). Source: HCES, 2011; 2016; World Bank staff calculations. 72 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 73 Poverty reduction in Amhara differed from the rest of much higher. This pattern is different from the rest of the Figure 5 CASUAL LABORERS, CROP PRODUCERS AND MANUFACTURERS HAVE Ethiopia in that it was driven mainly by changes in ur- country, in which rural areas account for the largest share THE HIGHEST POVERTY RATES ban areas of the region. Figure 9 shows that of the 4.4 of poverty reduction. This is not surprising, however, given percentage point reduction in poverty, in Amhara only 1.6 the evidence presented in Figure 2 showing that growth was Poverty rates by household’s main income source, 2016 percentage points are attributed to rural areas and 2.2 per- strong and positive throughout urban Amhara, but was very cent to urban areas, despite the rural population share being weak over most of the rural distribution (Figure 7). Amhara National excl. Amhara 35% 30% 25% Figure 7 POVERTY REDUCTION IN AMHARA WAS DRIVEN BY CHANGES IN URBAN Poverty rate 20% AREAS, UNLIKE IN THE REST OF ETHIOPIA 15% Rural-urban decomposition of poverty changes 2011 to 2016 10% Rural Urban Population shift Interaction 5% Amhara National excl. Amhara 0% 0 Change in poverty headcount or n g ns il t es ck -1 en ta tio rin lab vic io to re m uc tu at es er oy d al ac on od -2 an liv -s su pl uf /d pr em d Ca ale an lf an ns Se p -3 -m les ro io ied p ns -c ro ho lar lf -c Pe Se -w lf -4 Sa Se lf Se lf Se -5 Main source of household income -6 -7 Source: HCES, 2011; 2016; World Bank staff calculations. Source: HCES, 2011; 2016; World Bank staff calculations. Figure 6 EDUCATION IS A MAIN CORRELATE OF POVERTY Poverty rates by educational attainment of the household head, 2016 Agriculture is the largest contributor to poverty reduc- service sector. As shown in Figure 10, the contribution of tion in both Amhara and the rest of the country, indi- manufacturing to poverty reduction was extremely limited Amhara National excl. Amhara cating that structural change has not yet materialized. in Amhara and the rest of the country. Structural transfor- Improvements in the welfare of households that had agricul- mation, proxied by the shift in population from agriculture to 35% ture as a main income source explains two-thirds of poverty manufacturing or services, played a very minor role, explain- 31% 30% 28% reduction in Amhara between 2011 and 2016 . The other ing less than 5 percent of poverty reduction between 2011 25% 22% major contributor to poverty reduction in the region was the and 2016 (Figure 8). Poverty rate 21% 20% 16% 15% 13% 13% 9% 10% 6% 3% 4% 5% 0% 0% No education Incomplete Complete Incomplete Complete Post-secondary primary primary secondary secondary Education level of household head Source: HCES, 2011; 2016; World Bank staff calculations. 74 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 75 Figure 8 THE AGRICULTURE SECTOR REMAINS THE LARGEST CONTRIBUTOR TO Figure 9 OWNERSHIP OF CONSUMER DURABLES AND HOUSING CONDITIONS POVERTY REDUCTION IN AMHARA IMPROVED BETWEEN 2011 AND 2016 Sectoral decomposition of poverty changes 2011 to 2016 Household durables, housing conditions and access to energy in 2011 and 2016, % of households with asset/characteristic Agriculture Manufacturing Construction Services Other Population shift Interaction A. Household Durables Amhara National excl. Amhara 2011 2016 0 Change in poverty headcount 60 -1 -2 50 -3 40 -4 30 -5 -6 20 -7 10 Source: HCES, 2011; 2016; World Bank staff calculations. 0 Mobile Mobile Refrigerator Refrigerator Animal Cart Motorcycle Animal Cart Motorcycle TV TV NON-MONETARY WELFARE Amhara National excl. Amhara Improvements in non-monetary welfare accompanied percent in 2011 to 10 percent in 2016. However, for those increase in consumption expenditure. In Amhara, the who faced food shortage, the average number of months percentage of households that reported a food shortage in with a food shortage stayed the same. The results for the B. Housing/energy the last 12 months before the interview decreased from 23 whole country are similar (Table 4). 2011 2016 80 Table 4 FOOD SECURITY IMPROVED BETWEEN 2011 AND 2016 70 Incidence of food shortages and average duration of food shortages – food gap 60 50 40 2011 2016 30 Food shortage (%) Food gap (months) Food shortage (%) Food gap (months) 20 Amhara 23.2 3.1 10.4 3.0 10 National excl. Amhara 21.0 3.3 10.4 3.3 0 Improved wall Floor: tiles/cement Access to electricity Electricity for cooking Improved wall Floor: tiles/cement Access to electricity Electricity for cooking Improved roof Improved roof Source: WMS, 2011; 2016; World Bank staff calculations. Ownership of household durables and access to elec- below the rest of the country. The proportion of households tricity also improved in Amhara, although they remain with an improved roof, access to electricity, and who use at low levels. Ownership of household durables such as electricity for cooking also improved. But there was no signif- televisions, mobile phones, refrigerators, animal carts, and icant improvement in other housing quality indicators such as Amhara National excl. Amhara motorcycles increased between 2011 and 2016, although the quality of wall and floor (Figure 9). the percentage of households owning these durables is Source: DHS, 2011; 2016; World Bank staff calculations. 76 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 77 Access to improved drinking water sources37 also im- respectively from 62 per thousand to 48, from 21 to 12, and proved in Amhara. The proportion of households with ac- from 81 to 60 between 2011 and 2016. Delivery at health Figure 11 CHILDREN’S MORTALITY DECREASED, AND THEIR HEALTH IMPROVED cess to improved water sources in Amhara increased from facility increased from 10 to 27 percent, while the percentage BETWEEN 2011 AND 2016 2011 2016 54 percent to 66 percent between 2011 and 2016, mirroring of fully immunized children increased from 26 to 46 percent. 100 90.2 variables Selected health and education 81.0 2011 in 2011 and 2016 2016 improvements seen for the rest of Ethiopia (Figure 10). The stunting rate decreased from 52 percent to 46 percent. 80 100 Net primary enrollment, primary school completion, and 90.2 68.7 61.7 Improvements were also observed in health and edu- A. Child and 59.9mortality infant 81.0 58.5rates (per 1000) gross secondary enrollment rates also improved. The rest of 60 80 48.1 48.1 68.7 cation, although challenges remain. Infant mortality rates 2011 2016 the country experienced similar improvements (Figure 11). 61.7 59.9 58.5 40 33.6 and child and under-five mortality rates in Amhara decreased 60 100 90.2 48.1 48.1 20.5 81.0 21.7 20 40 80 12.3 33.6 68.7 61.7 20.5 59.9 58.5 21.7 0 20 60 12.3 48.1 Infant mortality Child mortality Under five 48.1 Infant mortality Child mortality Under five 0 40 rate rate mortality rate rate 33.6rate mortality rate Figure 10 ACCESS TO IMPROVED WATER SOURCES INCREASED BETWEEN 2011 Infant mortality Child mortality Under five Infant mortality mortality Childexcl. Amhara 20.5 National 21.7AmharaUnder five AND 2016 20 rate rate 12.3 mortality rate rate rate mortality rate Amhara National excl. Amhara Access to water in 2011 and 2016, % of households with access/characteristics 0 Infant mortality Child mortality Under five Infant mortality Child mortality Under five rate rate mortality rate rate rate mortality rate Amhara National excl. Amhara Amhara National excl. Amhara Piped water - compound Piped -outside Protected well/tube well Protected spring Source: DHS, 2011; 2016; World Bank staff calculations Other improved Unimproved sources 2011 2016 60 B. Immunization and health facility delivery (percent) 100% 100% 52.0 2011 2016 43.89 35.56 47.2 35.04 50 45.8 46.3 60 42.4 80% 80% 52.0 36.9 36.4 40 50 45.8 46.3 27.1 42.4 30 26.3 2011 201626.0 60% 10.9 60% 12.03 40 23.7 36.9 36.4 11.72 8.05 60 24.24 14.54 20 27.1 52.0 30 26.3 9.9 26.0 40% 13.04 40% 9.09 50 10.2 45.8 46.3 23.7 24.19 22.07 10 42.4 20.76 20 36.9 36.4 20% 18.15 20% 40 10.2 9.9 0 10 11.16 15.33 30 Health 27.1 facility Fully 26.3 Stunted facility Health26.0 Fully Stunted 10.6 11.14 23.7 0% 0% 0 delivery immunized children delivery immunized children 2011 2016 2011 2016 20 Health children children 10.2 facility Fully Stunted 9.9 facility Health Fully Stunted 10 delivery immunized children delivery immunized children Source: DHS, 2011; 2016; World Bank staff calculations. Amhara National excl. Amhara children children 0 Health facility Amhara Fully Stunted excl. Amhara Stunted National Fully Health facility Source: DHS, 2011; delivery immunized 2016; World Bank children staff calculations delivery immunized children children children Amhara C. Education (percent)National excl. Amhara 2011 2016 100 81.0 2011 2016 80 100 69.0 69.3 81.0 60.2 60 80 69.0 2011 2016 69.3 60.2 33.7 32.0 40 100 60 29.5 26.6 28.5 24.6 22.6 81.0 18.0 20 80 33.7 32.0 40 69.0 29.5 26.6 69.3 28.5 24.6 22.6 18.0 60.2 0 60 20 Net primary Primary school Gross Net primary Primary school Gross enrolment completion secondary enrolment completion 33.7 secondary 32.0 0 40 29.5 26.6 28.5 Net primary (15-24) 24.6 Primary school school Gross Net primary (15-24) Primary school school Gross 22.6 enrolment completion enrolment 18.0 secondary enrolment completion enrolment secondary 20 (15-24) Amhara school National(15-24) school excl. Amhara 37  Improved drinking water sources include piped water, protected well, protected spring, rainwater and bottled water. enrolment 0 2011; 2016; World Bank staff calculations. enrolment Source: WMS, Net primary Primary school Amhara Gross Primary Net primary National school excl. Amhara Gross enrolment completion secondary enrolment completion secondary 78 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS (15-24) school (15-24) school 79 enrolment enrolment INEQUALITY OF OPPORTUNITY in the rest of Ethiopia complete secondary school, accord- Unequal coverage of infrastructure access remains a ing to 2016 data. As for enrollment in secondary schooling, challenge in Amhara and the rest of the country. The To assess inequality of opportunity38 in Amhara and the school completion, secondary school enrollment, and ac- both Amhara and the rest of the country have very low rates, D-index was highest for access to electricity, exemplifying rest of Ethiopia, we focus on six key outcomes: (i) primary cess to electricity all exhibit very low rates, and also very low with Amhara faring slightly worse (16 percent compared to the rural-urban divide in access to these services in Amha- school enrollment, (ii) primary school completion, (iii) second- HOI scores nationwide (see the Coverage and HOI panels in 18 percent). ra, and broadly throughout the country. As noted earlier, de- ary school enrollment, (iv) access to electricity; (v) access to Figure 12). The D-index for these variables – measuring how composition of this index (not shown) reveals that access The D-index for completed primary school and sec- an improved water source; and (vi) access to a health post much these outcomes differ according to circumstances – is depends far more on whether a household is in an urban ondary school enrollment in Amhara is higher than the within 5 km. Circumstance variables39 of the household in- correspondingly high. This implies that there are strong dif- or rural area than on the level of household wealth. For ex- rest of the country, indicating that circumstances play clude rural/urban location, gender, religion, agro-ecological ferences between households in which children complete ample, in urban areas even the poorest households tend to more of a role in determining these outcomes in this zone, consumption quintile of the child’s household, and primary school, enroll in secondary school, and have access have access to electricity, while in rural areas even the richest region. The D-index was lowest for enrollment in primary the household’s score on the rural accessibility index (RAI).40 to electricity, and those that do not. Decomposing this dif- households generally do not have access. education, driven by the fact that enrollment rates for age-el- Figure 12 presents the coverage, Human Opportunity Index ference (not shown) reveals that this is largely explained by igible children was already relatively high. (HOI), and Dissimilarity Index (D-Index) across these key out- rural-urban differences in access rates, both within Amhara come variables in 2016. and in the rest of Ethiopia. The proportion of enrolled primary age-eligible children Fewer than one in five children age-eligible for second- (aged 7 to 14) was significantly higher in Amhara than ary school (aged 15 to 18) were enrolled in 2016. This PUBLIC EXPENDITURE in the rest of the country (81 versus 69 percent). Differ- indicates that educational opportunity is quite different for the In the previous sections, we discussed monetary and Although average per capita public expenditure in ences between Amhara and the rest of Ethiopia across the older cohort of children (aged 15-18) than the younger co- non-monetary welfare trends between 2011 and 2016, and Amhara is similar to the rest of the country, Amhara other key outcomes were relatively muted. Access to primary hort (aged 7-14). Just 29 percent of children in Amhara and inequality of opportunities focusing on 2016, both in Amha- spends a significantly higher proportion on administra- ra and the rest of the country. In this section, we discuss tion. Although the share of administration remained high in the trends, volume, and allocation of public expenditures to Amhara, spending shifted in favor of social and economic Figure 12 AMHARA PERFORMS WELL ON PRIMARY SCHOOL ENROLLMENT, WHILE identify variation across regions and how these variations link sectors between 2011 and 2016. Improvements in per cap- UNEQUAL ACCESS TO INFRASTRUCTURE REMAINS A CHALLENGE to welfare. We use MoF data. We express all expenditure ita expenditure on pro-poor sectors were stronger in Amara values in 2016 prices using the non-food CPI deflator. We compared to other regions combined, except for agriculture Coverage, HOI, and D-index for different outcomes, 2016 first present total per capita public expenditure and then dis- and rural development. Per capital expenditures in health cuss the expenditure allocation by sector. Finally, we present and education in 2016 were significantly higher than the rest Amhara National excl. Amhara expenditure on selected “pro-poor” sectors, which include of the country. These results contradict the fact that poverty 90 agriculture and rural development, water resources, health, in Amhara is higher than the rest of the country, which could Coverage HOI D-index 80 and education. be because higher spending is not yet reflected in welfare 70 improvements. Percent/score 60 50 40 30 Figure 13 PER CAPITA PUBLIC EXPENDITURE INCREASED OVER TIME 20 10 Per capita real public expenditure, 2011 to 2016 0 Amhara National excl. Amhara alt ter alt ter alt ter st st st ec ry ool ec ry ool ec ry ool l a ol in Imp y ac ol m ve ess m ve ess m ve ess nd sc l co ry ool co ry ool ty oo o po po po cit ho cit ho se ma ho he wa he a he a h a h a h ici ch im ch im ch 5k pro cc 5k ro cc 5k ro c of d w of d w nd sc nd sc tri sc tri sc c h h h s pr ry s pr ry s pr ry s a of d co ry y 2,000 a a a a a a En let rim En let rim En let rim r p i t m m m dp p p I I ro ed ro ed ro ed se se m d m d El El El In 2016 Birr 1,600 Co rolle e e Co roll Co roll d d d in in lle lle lle p p p ith ith ith En En En W W W 1,200 Source: HCES, WMS 2011; 2016; World Bank staff calculations. 800 400 2011 2012 2013 2014 2015 2016 A full description of the methodology and variables used in the construction of this index in Ethiopia can be found in the technical 38  appendix. Source: MoF; World Bank staff calculations. 39  These are generally thought of as variables over which a person does not have control, such as gender or region of birth. 40  Accessibility was also tested using a market accessibility variable, and the results were broadly consistent. 80 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 81 Figure 14 SPENDING SHARE ON ECONOMIC SECTOR INCREASED Figure 15 EXPENDITURE SHARE ON WATER RESOURCES AND HEALTH INCREASED IN AMHARA BETWEEN 2011 AND 2016 Expenditure share by sectors (%), 2011 to 2016 Real per capita expenditure and share of total expenditure (%) for selected sectors, 2011 and 2016 Amhara National excl. Amhara A. Real per capita expenditure in 2016 price Administration Economic Social Other 2016 2011 2016 4.1 2016 16.7 2016 26.3 400 2016 6 2016 7.5 28.2 50.8 33.8 31.7 42.8 32.9 300 In 2016 Birr 15.1 25.8 2016 2016 200 2016 26.8 30.2 21.4 100 Note: The inner circles present the 2011 values. Source: MoF; World Bank staff calculations. 0 Agriculture and Rural Development Water Resources Education Health Agriculture and Rural Development Water Resources Education Health Per capita public expenditure in Amhara increased be- Per capita spending42 in selected “pro-poor” activities tween 2011 and 2016 as was the case for the rest of increased between 2011 and 2016, both in terms of per Ethiopia combined. Public spending in the Amhara region capita and share of total expenditure. In Amhara, the in- increased from Birr 1,262 per person in 2011 to Birr 1,938 crease was particularly strong for water resources and health Amhara National excl. Amhara per person in 2016 (in constant 2016 Birr), with the lowest sectors; per capita spending on water and health sectors in- per capita expenditure of Birr 1,155 recorded in 2012. Per creased from Birr 31 to Birr 104 and from Birr 78 to Birr 173, capita spending in 2011 was higher in Amhara, but it grew respectively. The rest of the country also witnessed per cap- B. Expenditure share (%) faster in the rest of the country, leading to a convergence in ita spending increases on pro-poor functional activities, but 2016. On average, per capita public expenditure increased the rate of change is faster in Amhara, except for agriculture Agriculture & Rural Dev. Water Resources Education Health at an average rate of ten percent annually in Amhara com- and rural development. In Amhara, such growth has resulted 50 pared to 15 percent in other regions combined (Figure 13). in a higher per person spending on the sectors compared to 45 40 8.8 other regions combined in 2016 (Figure 15A). 8.9 Public expenditure growth in Amhara accompanies a 35 7.5 5.5 22.2 spending shift away from administration and towards In Amhara, the share of spending on pro-poor activities 30 21.9 25 21.8 16.2 social and economic sectors.41 Social and economic sec- also increased between 2011 and 2016, while it con- 20 tor spending, as share of the total expenditure, increased tracted in the rest of the country. The share of spending 15 from 15.1 to 21.4 percent and from 28.2 to 31.7 percent, re- on water and health sectors in Amhara increased from 2.2 10 6.5 5.7 2.2 5.3 spectively. For the rest of the country, spending on economic and 5.5 percent in 2011 to 5.3 and 8.9 percent in 2016, re- 5 7.1 7 6 5.9 sector increased from 25.8 to 30.2 percent, while spending spectively. Moreover, the share of the education sector in the 0 2011 2016 2011 2016 share on the social sector decreased from 32.9 to 26.8 per- total spending remained stable at 22 percent between 2011 Amhara National excl. Amhara cent. Administration-related spending decreased both in Am- and 2016, whereas it fell from 22.2 to 16.2 percent during the hara and the rest of the country, but it is still large, particularly same period in other regions combined (Figure 15B). in Amhara, accounting for 42.8 percent in 2016 (Figure 14). Note: The sum does not add up to 100% because it includes only shares of selected “pro-poor” sectors. Source: MoF; World Bank staff calculations. 41 Social sector includes education, health, culture and sports, and labor and social affairs while economic sector includes agriculture and rural development, water resources, trade and industry, mines and energy, transportation and communication, construction. 42 Per capita spending refers to total spending in a particular sector divided by the population of the region. For example, per capita spending on education refers to total spending on education divided by the total population of the region and not spending per student. 82 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 83 POVERTY AND INEQUALITY Food and nonfood consumption in the Oromia region be solely attributed to urban areas, where consumption in- increased significantly between 2011 and 2016, al- creased by 30 percent, modestly lower than the increase of though only urban areas experienced the full benefit 36 percent for the rest of the country. Rural areas, however, of this increase. Mean consumption per adult increased by were left behind. Consumption in rural areas remained largely 10 percent in Oromia, compared to a 16 percent increase for the same between 2011 and 2016 (Table 1). Oromia other regions combined. The increase in consumption can Table 1 CONSUMPTION INCREASED BETWEEN 2011 AND 2016 BOTH IN OROMIA AND THE REST OF THE COUNTRY Mean annual household consumption per adult equivalent in December 2015 prices, 2011 and 2016 INTRODUCTION TOTAL URBAN RURAL This chapter focuses on the evolution of socio-economic and health indicators such as poverty, inequality, 2011 2016 % change 2011 2016 % change 2011 2016 % change access to basic services, and education, in the Oromia region between 2011 and 2016. Using data from Oromia 10,947 12,022 9.8* 13,891 18,080 30.2* 10,504 11,022 4.9 a variety of sources, mainly the twinned household living standards surveys (HCES and WMS)43 and the National excl. 11,048 12,810 15.9* 13,906 18,877 35.7* 10,384 11,009 6.0* Demographic and Health Survey (DHS), the chapter begins by documenting changes in monetary poverty Oromia and living standards in the region. It then moves to an analysis of non-monetary welfare before turning to Note: * represents changes that are statistically significant. challenges of inequality of opportunity. The final section discusses changes in social spending from 2011 to Source: HCES, 2011; 2016; World Bank staff calculations. 2016. The chapter aims to provide policy makers with information and analysis that will be useful in improv- ing the effectiveness of poverty reduction programs in the region. The increase in consumption in Oromia led to a de- while rural poverty rates remained about the same between Some of the main findings include: crease in poverty, but less so compared to the rest of 2011 and 2016. The decrease in poverty of 5 percentage the country.44 The proportion of people below the national points was slower in Oromia compared to the rest of Ethiopia, ● The poverty rate in Oromia decreased from 29 percent to education. Despite the improvement, Oromia in general poverty line in Oromia decreased by 5 percentage points from where poverty decreased by 7 percentage points (Table 2). 24 percent between 2011 and 2016. However, poverty lags from the rest of the country in non-monetary indica- 29 percent in 2011 to 24 in 2016. In urban areas, poverty reduction in rural Oromia wasn’t statistically significant. tors. This is also reflected in a lower score on the Human decreased by about 40 percent, falling from 25 to 15 percent, Opportunity Index than the national average – implying ● Consumption growth in Oromia was strong only in ur- lower equality of access to key basic services for chil- ban areas while rural areas did not experience signifi- dren in Oromia compared to the rest of Ethiopia. cant consumption growth. The gap between rural and POVERTY DECREASED IN OROMIA BUT LESS THAN THE REST OF THE Table 2 urban areas further widened and that led to an increase ● Average per capita public expenditure in Oromia is COUNTRY in inequality smaller than the rest of the country. Oromia also spends Percentage of people below the national poverty line lower per capita on “pro-poor” sectors which includes ● Non-monetary welfare including ownership of consumer health, education, agriculture and rural development, durables and access to drinking water improved, as was TOTAL URBAN RURAL and water resources. the case for human capital indicators such as health and % pt. % pt. % pt. 2011 2016 change 2011 2016 change 2011 2016 change Oromia 28.7% 23.9% -4.8* 24.8% 15.3% -9.5* 29.3% 25.3% -4.0 National excl. 30.2% 23.2% -7.0* 26.1% 14.3% -11.8* 31.1% 25.8% -5.3* Oromia Note: * represents changes that are statistically significant. Source: HCES, 2011; 2016; World Bank staff calculations. 43 Household Consumption Expenditure Survey (HCES) and Welfare Monitoring Survey (WMS) are the two official household surveys collected by CSA every five year and used to monitor welfare. HCES covers monetary welfare and WMS deals with non-monetary 44 The relatively slower poverty reduction in Oromia could partly be due the higher rural population share (86% for Oromia versus 81% welfare like health and education. Both surveys are collected from the same sample of households. for the rest of the country in 2016). Poverty reduction was low in rural areas in general. 84 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 85 Consumption growth in urban Oromia was positive hand, consumption growth was not significantly different from The rise in inequality between 2011 and 2016 is mainly “between” share of inequality – the part of inequality result- for the entire consumption distribution, while growth zero for most of the distribution, except for the middle, where due to the increasing gap between urban and rural ar- ing from differences in average consumption between urban in rural consumption varied and was generally low. In consumption grew by almost four percent annually, and at eas. Average consumption was already significantly higher in and rural areas – increased from 13 percent in 2011 to 23.5 urban areas, consumption growth was positive and mostly the top. For the poorest rural quintile, growth was negative urban than in rural areas in 2011. The gap further increased percent in 2016. On the other hand, the “within” share of uniform across the consumption distribution, although it was explaining the lack of significant poverty reduction in rural in 2016, with urban consumption showing strong growth inequality – the part resulting from changes in distribution relatively weaker at the bottom. In rural areas, on the other Oromia (Figure 1). and rural consumption also growing, but only at about one- within urban and rural areas – diminished from 70 percent to sixth the rate of urban areas (see Table 1). As a result, the 64 percent (Figure 2). Figure 1 GICS FOR OROMIA URBAN AND RURAL Figure 2 INEQUALITY INCREASED DUE TO THE INCREASING GAP BETWEEN Urban Rural URBAN AND RURAL AREAS Decomposition of the Gini coefficient into a between rural-urban component and a within-component Overlap Within Between Relative contribution to Gini (%) 100% 90% 17.2 12.8 80% 63.7 69.8 70% 60% 50% 40% 30% 20% 23.5 10% 13.0 0% 2011 2016 Source: HCES, 2011; 2016; World Bank staff calculations. Source: HCES, 2011; 2016; World Bank staff calculations. The lack of consumption growth in rural areas com- percentile and consumption compared to the 10th percentile Given the increase in inequality between 2011 and 2016, redistribution effect is positive for Oromia and for the rest of bined with strong growth in urban areas to increase increased from 3.5 in 2011 to 4.4 in 2016, reflecting strong poverty reduction in Oromia can be entirely attributed the country. If there were no changes in the distribution be- inequality. The Gini index increased from 28 to 31 be- consumption growth at the top and lack of growth at the bot- to increases in consumption. Figure 5 decomposes the tween 2011 and 2016, poverty would have decreased by tween 2011 and 2016. The ratio of consumption at the 90th tom (Table 3). change in poverty into growth and redistribution components 6.8 and 10.2 percentage points in Oromia and the rest of the using the Datt-Ravallion decomposition technique. The tech- country, respectively, instead of the 4.8 and 6.8 percentage nique indicates which part of the poverty change can be at- points we actually observe. In other words, if the consump- Table 3 INEQUALITY INCREASED BETWEEN 2011 AND 2016 tributed to: (i) the change in average consumption (growth tion distributions in 2011 were maintained in 2016, pover- effect), and (ii) the change in the redistribution of consump- ty reduction could have been stronger in Oromia, though it tion – that is. changes in inequality (redistribution effect). The would still have lagged the rest of Ethiopia (Figure 3). 2011 2016 growth effect on poverty is negative (poverty-reducing). The Gini index p90/p10 ratio Gini index p90/p10 ratio Oromia 28.3 3.5 31.2 4.4 National 29.8 3.6 32.9 4.3 Source: HCES, 2011; 2016; World Bank staff calculations. 86 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 87 In Oromia, households that primarily depend on mixed Ethiopia in that the poverty rate for households relying on Figure 3 CONSUMPTION GROWTH DROVE POVERTY REDUCTION farming (crop and livestock) and casual labor have the manufacturing as their main income source are less than half highest poverty rates. Those who depend on services, (11 percent) of the poverty rate for other regions combined Datt-Ravallion decomposition of poverty change between 2011 and 2016 remittances, manufacturing, and salaried employment have (24 percent) (Figure 5). the lowest poverty rates. Oromia diverges from the rest of Oromia National excl. Oromia Contribution to change in poverty 6.0 4.0 2.0 3.4 1.9 Figure 5 MIXED FARMERS AND CASUAL WORKERS HAVE THE HIGHEST 0.0 POVERTY RATES -2.0 -4.0 -4.8 Poverty rates by household’s main income source, 2016 -6.0 -6.8 -6.8 -8.0 Oromia National excl. Oromia -10.2 -10.0 -12.0 35% Poverty reduction Consumption growth Increase in inequality 30% 25% Poverty rate Source: HCES, 2011; 2016; World Bank staff calculations. 20% 15% 10% Average consumption in Oromia remains low for the per year on average, reflecting strong consumption growth at 5% first four quintiles. The average annual consumption per the top of the distribution. Oromia closely mirrors consump- 0% adult in 2016 was 4,781 Birr for the poorest quintile and tion patterns for the rest the country, though the jump from or n g ns il t es ck en ta tio rin lab vic io to re m 12,632 Birr for the fourth quintile. Average consumption the fourth to the fifth quintile in Oromia is smaller than the uc tu at es er oy d al ac on od an liv -s su pl uf /d shows a sharp increase from the fourth to the fifth (richest) same jump for the rest of the country (Figure 4). pr em d Ca ale an lf an ns Se p -m les ro io ied p quintile, with the richest 20 percent consuming 23,583 Birr ns -c ro ho lar lf -c Pe Se -w lf Sa Se lf Se lf Se Main source of household income Figure 4 HOUSEHOLD CONSUMPTION REMAINS FAIRLY LOW EXCEPT FOR THE Source: HCES 2016; World Bank staff calculations. TOP QUINTILE Average household expenditures per capita per year, 2016 Oromia National excl. Oromia Educational attainment of household heads strongly or have some post-secondary education. Oromia closely negatively correlates with poverty. With a poverty rate of mirrors the trend for the rest of the country, with the excep- 30,000 29 percent, households headed by someone with no educa- tion that in Oromia households headed by someone who has Average annual consumption 25,000 tion have the highest poverty in Oromia. Increasing education completed primary education have a lower poverty rate (14 20,000 of the household head decreases poverty rates, reaching percent) than those headed by someone with incomplete merely three and four percent, respectively, in households secondary education (17 percent) (Figure 6). 15,000 headed by someone who completed secondary education 10,000 5,000 0 Poorest 2 3 4 Richest Consumption quintiles Source: HCES 2016; World Bank staff calculations. 88 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 89 When decomposing poverty reduction by sector, Oro- agriculture to manufacturing or services) explained less than Figure 6 EDUCATION IS A MAIN CORRELATE OF POVERTY mia closely matches what happened in the rest of the 1 percent of poverty reduction in Oromia over the period. country. Improvements in the agricultural sector explain over Although this effect has been be important in other regions, Poverty rates by educational attainment of the household head, 2016 60 percent of poverty reduction in the rest of Ethiopia; the it is not a factor in Oromia. There was a slight fall in the share corresponding figure for Oromia is close at 58 percent. Gains of the population in Oromia relying on agriculture over the Oromia National excl. Oromia for households working in the services sector also contribut- period, but much smaller than the rest of the country. In 35% ed to poverty reduction in Oromia, but the effect was not as 2011, 80 percent of households in Oromia relied primarily 29% 28% pronounced as the rest of the country. The so-called “pop- on agriculture as the primary source of income. In 2016 this 30% ulation shift” effect had no significant impact on poverty re- had fallen slightly to 77 percent. At the same time, the share 25% 22% 22% Poverty rate duction in Oromia. This variable, often used as an indicator of of households relying primarily on services increased by less 20% 17% structural transformation (proxied by shift in population from than one percentage point (Figure 8). 14% 15% 12% 11% 10% 7% 5% 4% 3% 4% 0% Figure 8 THE AGRICULTURE SECTOR REMAINS THE LARGEST CONTRIBUTOR TO No education Incomplete Complete Incomplete Complete Post-secondary POVERTY REDUCTION IN OROMIA primary primary secondary secondary Education level of household head Sectoral decomposition of poverty changes 2011 to 2016 Source: HCES 2016; World Bank staff calculations. Agriculture Manufacturing Construction Services Other Population shift Interaction Oromia National excl. Oromia 0 Change in poverty headcount Poverty reduction in Oromia was very similar to the and 1.2 percentage points to urban areas. The contribution -1 rest of Ethiopia, albeit poverty reduced more slowly in of rural areas would have been even stronger had there not -2 the region between 2011 and 2016. Figure 7 shows that been negative growth for the very poorest households in -3 of the almost 5 percentage points of reduction in poverty in Oromia (as shown in Figure 2). -4 Oromia, 3.6 percentage points are attributed to rural areas -5 -6 -7 -8 Figure 7 POVERTY REDUCTION IN OROMIA LOOKED A LOT LIKE POVERTY Source: HCES; 2011, 2016. World Bank staff calculations. REDUCTION IN THE REST OF ETHIOPIA Rural-urban decomposition of poverty changes 2011 to 2016 Rural Urban Population shift Interaction Oromia National excl. Oromia 0 Change in poverty headcount -1 -2 -3 -4 -5 -6 -7 --8 Source: HCES, 2011; 2016; World Bank staff calculations. 90 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 91 NON-MONETARY WELFARE Figure 9 OWNERSHIP OF CONSUMER DURABLES AND HOUSING CONDITIONS The increase in consumption expenditure in the region 2016. However, for those who faced food shortage, the aver- IMPROVED BETWEEN 2011 AND 2016 was accompanied by improvements in non-monetary age number of months with a food shortage increased from Household durables, housing conditions and access to energy in 2011 and 2016, % of households welfare. In Oromia, the percentage of households that re- 3.1 to 3.6. The percentage of households who faced food with asset/characteristic ported food shortage in the last 12 months before the inter- shortage was significantly higher for the rest of the country in view decreased from 16 percent in 2011 to 11 percent in 2011, but the difference disappeared in 2016 (Table 4). A. Household Durables 2011 2016 Table 4 FOOD SECURITY IMPROVED BETWEEN 2011 AND 2016 60 Incidence of food shortages and average duration of food shortages – food gap 50 40 2011 2016 Food shortage (%) Food gap (months) Food shortage (%) Food gap (months) 30 Oromia 16.0 3.1 10.5 3.6 20 National 25.1 3.3 10.3 3.0 10 Source: WMS, 2011; 2016; World Bank staff calculations. 0 Mobile Mobile Refrigerator Refrigerator Animal Cart Motorcycle Animal Cart Motorcycle TV TV Ownership of household durables and access to elec- regions combined. The proportion of households with an im- tricity also improved in Oromia, although this remains proved roof, access to electricity, and who use electricity for Oromia National excl. Oromia low. Ownership of household durables such as televisions, cooking also increased. But there was no significant improve- mobile phones, refrigerators, animal carts, and motorcycles ment in other housing quality indicators such as the quality of increased between 2011 and 2016, although the percentage wall and floor (Figure 9). B. Housing/energy of households owning these durables is slightly below other 2011 2016 70 60 50 40 30 20 10 0 Improved wall Floor: tiles/cement Access to electricity Electricity for cooking Improved wall Floor: tiles/cement Access to electricity Electricity for cooking Improved roof Improved roof Photo: World Bank/Sylwia Pecio Oromia National excl. Oromia Source: DHS, 2011; 2016; World Bank staff calculations. 92 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 93 Figure 10 ACCESS TO IMPROVED WATER SOURCES INCREASED BETWEEN 2011 Figure 11 CHILDREN’S MORTALITY DECREASED, AND THEIR HEALTH IMPROVED AND 2016 BETWEEN 2011 AND 2016 2011 2016 100 89.7 Selected health and education variables 87.0 Access to water in 2011 and 2016, % of households with access/characteristics 2011 in 2011 and 2016 2016 80 100 69.5 89.7 87.0 65.1 Oromia National excl. Oromia 60.8 57.0 A. Child and infant mortality rates (per 1000) 60 80 51.1 69.5 2011 201645.8 65.1 Piped water - compound Piped -outside Protected well/tube well Protected spring 57.0 34.6 60.8 40 60 100 51.1 28 89.7 45.8 87.0 Other improved Unimproved sources 19.4 20.3 20 34.6 40 80 69.5 28 100% 100% 65.1 19.4 60.8 20.3 52.94 37.24 42.38 33.98 0 57.0 20 60 51.1 Infant mortality Child mortality Under five 45.8 Infant mortality Child mortality Under five 80% 80% 0 40 rate 34.6rate mortality rate rate rate mortality rate 28 Infant mortality Child mortality Oromia19.4 Under five Infant mortality 20.3OromiaUnder five Child excl. National mortality 60% 60% 10.34 14.21 8.31 20 rate rate mortality rate rate rate mortality rate 20.22 40% 10.24 11.39 40% 13.39 Oromia National excl. Oromia 4.68 0 24.04 21.72 19.38 Infant mortality Child mortality Under five Infant mortality Child mortality Under five 20% 25.91 20% rate rate mortality rate rate rate mortality rate 13.86 15.7 Oromia National excl. Oromia 6.21 11.77 0% 0% Source: DHS, 2011; 2016; World Bank staff calculations 2011 2016 2011 2016 2011 2016 Source: DHS, 2011; 2016; World Bank staff calculations. B. Immunization and health facility delivery (percent) 49.4 50 46.8 41.4 2011 2016 39.8 40 36.5 49.4 50 46.8 32.1 30.6 41.4 39.8 Access to improved drinking water sources also im- 45 35 to 19, and from 90 to 70 between 2011 and 2016. Deliv- 30 40 24.7 36.5 2011 201632.1 proved in Oromia. The proportion of households with ac- ery at health facility increased from 8 to 19 percent, while the 18.8 30.6 49.4 20 50 30 15.6 46.8 24.7 11.4 cess to improved water sources in Oromia increased from percentage of fully immunized children increased from 16 to 41.4 8.0 18.8 39.8 10 40 20 36.5 47 percent to 63 percent between 2011 and 2016, mirroring 25 percent. The stunting rate decreased from 41 percent to 15.6 11.4 32.1 30.6 improvements in the rest of Ethiopia (Figure 10). 37 percent. Net primary enrollment, primary school comple- 0 30 8.0 10 24.7 tion, and gross secondary enrollment rates also improved. Health facility Fully Stunted Health facility Fully Stunted 18.8 Improvements were also observed in health and ed- 20 0 delivery immunized 15.6 children delivery immunized children The rest of the country also improved on these indicators, children children Health facility Fully Stunted 11.4 facility Health Fully Stunted ucation, although challenges remain. The infant mor- 8.0 and Oromia lags in most of the indicators (Figure 11). 10 delivery immunized children delivery immunized children tality rate, child, and under-five mortality rates in Oromia Oromia children National excl. Oromia children decreased, respectively, from 57 per thousand to 51, from 0 Health facility Oromia Fully Stunted excl. Oromia Stunted National Fully Health facility Source: DHS, 2011; delivery immunized 2016; World Bank children staff calculations delivery immunized children children children Oromia C. Education (percent)National excl. Oromia 2011 2016 80 73.6 69.0 2011 2016 65.9 80 57.2 73.6 60 69.0 65.9 57.2 60 40 2011 2016 34.1 33.5 30.1 28.6 80 25.8 26.2 73.6 20.1 22.3 69.0 34.1 40 20 30.1 65.9 33.5 26.2 28.6 57.2 25.8 60 20.1 22.3 20 0 40 Net primary Primary school Gross Net primary school Primary 34.1 Gross 33.5 enrolment 30.1 completion secondary enrolment completion 28.6 secondary 0 25.8 26.2 (15-24) school 20.1 (15-24) 22.3 school Net primary Primary school Gross Net primary Primary school Gross 20 enrolment completion enrolment secondary enrolment completion enrolment secondary (15-24) Oromia school (15-24) National excl. Oromiaschool 45 Improved drinking water sources include piped water, protected well, protected spring, rainwater and bottled water. enrolment 0 2011; 2016; World Bank staff calculations. enrolment Source: WMS, Net primary Primary school Oromia Gross Primary Net primary National school excl. Oromia Gross enrolment completion secondary enrolment completion secondary 94 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS (15-24) school (15-24) school 95 enrolment enrolment INEQUALITY OF OPPORTUNITY Figure 12 OROMIA LAGS BEHIND THE REST OF ETHIOPIA IN ALL SIX KEY To assess inequality of opportunity46 in Oromia and the rest children (15 to 18) were enrolled in secondary school in 2016. OUTCOMES of Ethiopia, we focus on six key outcomes: (i) primary school This indicates that education opportunity is quite different for Coverage, HOI, and D-index for different outcomes, 2016 enrollment, (ii) primary school completion, (iii) secondary the older cohort of children (15-18) than the younger cohort school enrollment, (iv) access to electricity; (v) access to (7-14) in the region. Oromia National excl. Oromia an improved water source; and (vi) access to a health post The HOI scores for Oromia are, like the coverage 80 within 5 km. Circumstance variables47 of the household in- Coverage HOI D-index scores, lower than the rest of Ethiopia. A higher HOI clude rural/urban location, gender, religion, agro-ecological 70 score is associated with a more equitable access to a key zone, consumption quintile of the child’s household, and 60 Percent/score service or outcome. For a given coverage rate, the lower the the household’s score on the rural accessibility index (RAI).48 50 HOI score the more unequal the access to that key outcome Figure 12 presents the coverage, Human Opportunity Index 40 is across different groups. In Oromia, for example, the cover- (HOI), and Dissimilarity Index (D-Index) across these key out- 30 age rate of access to electricity is 20 percent. However, this come variables in 2016. 20 is heavily skewed towards the urban population in the region. 10 In terms of coverage, Oromia lags the rest of Ethio- The very low HOI score of 9 for electricity reflects the strong 0 pia on all key outcomes except for electricity access difference in access rates across the population in the region. alt ter alt ter alt ter st st st ec ry ool ec ry ool ec ry ool l l l m ve ess m ve ess m ve ess nd sc l co ry ool co ry ool ity oo ty oo ty oo oo po po po and access to improved water source. Although primary he wa he a he a a h a h a h ric sch ici ch ici ch im ch im ch im ch 5k pro cc 5k pro cc 5k pro cc of d w of d w nd sc nd sc h h h The dissimilarity index, or D-index, shows how much s s pr ry s pr ry s pr ry s a a a of d school completion and secondary school enrollment are low co ry a a a a a a inequality in access there is between groups – a high- En let rim En let rim En let rim r r in Ethiopia, they are particularly low in Oromia. This means t t t m m m m dp p p I I I ro ed ro ed ro ed se se se er index means less equal access across groups. The m ed m ed El El El Co rolle that the HOI scores for these key outcomes are also low Co ll Co roll d d d o in in in lle lle lle r p p p D-index for completed primary school and secondary school ith ith ith En En En (see the Coverage and HOI panels in Figure 14). The D-index W W W enrollment in Oromia is the same as in the rest of Ethiopia. for these variables – measuring how much these outcomes This indicates that circumstances such as location, gender, Source: HCES, WMS 2011; 2016; World Bank staff calculations. differ according to circumstances – is correspondingly high, and household income do not matter more in Oromia than although similar to the rest of the country. This implies that in the rest of the country. The D-index scores for second- there are strong differences between households in which ary school enrollment and electricity access in Oromia were children complete primary school, enroll in secondary school, higher than the rest of the country, implying greater inequality and have access to electricity, and those that do not. De- PUBLIC EXPENDITURE in the distribution of access. composing this difference (not shown) reveals that this is largely explained by rural-urban differences in access rates, In the previous sections, we discussed trends in monetary results are consistent with the poorer performance of Oromia Unequal coverage of infrastructure access variables both within Oromia and in the rest of Ethiopia. and non-monetary welfare between 2011 and 2016, and in- in non-monetary indicators including health and education. remains a challenge in Oromia and the rest of the equality of opportunities focusing on 2016, both in Oromia country. The D-index was highest for access to electricity, Per capita public expenditure increased between 2011 The primary school enrollment rate in Oromia is about and the rest of the country. In this section, we discuss trends, exemplifying the rural-urban divide in access to these ser- and 2016 and was lower in Oromia than the rest of the 4.5 percentage points lower than the rest of Ethiopia. volume, and allocation of public expenditures to identify vari- vices in the region, and broadly throughout the country. As country in all years except 2014. Public spending in the In the region, although 69 percent of age-eligible children ation across regions and possible links with welfare. We use noted earlier, a decomposition of this index (not shown) re- region increased from Birr 626 per person in 2011 to Birr were enrolled in primary school in 2016, only one quarter of MoF data. All the expenditure values are expressed in 2016 veals that access depends far more on whether a household 1681 per person in 2016 (in constant 2016 Birr). However, age-eligible children had completed primary school. These prices using the non-food CPI deflator. We first present total is in an urban or rural area than on the level of household spending per capita grew faster in Oromia than in other re- numbers are both lower than the rest of Ethiopia, in which per capita public expenditure and then discuss the expendi- wealth. For example, in urban areas even the poorest house- gions combined, which narrowed that gap in 2016. On aver- close to three-quarters of primary school-aged children are ture allocation by sector. Finally, we present expenditure on holds tend to have access to electricity, while in rural areas age, per capita public expenditure increased at an average enrolled, and almost one-third complete primary education. selected “pro-poor” sectors, which include agriculture and even the richest households do not generally have access. rate of 32 percent annually in Oromia compared to ten per- By comparison, fewer than one in six children age-eligible rural development, water resources, health, and education. cent in the rest of the country (Figure 13). Average per capita public expenditure in Oromia is Public expenditure growth in Oromia is accompanied significantly lower than the rest of the country. Though by a spending shift away from administration and the Oromia spends relatively less on administration, per capita social sector towards “others”.49 Social sector and admin- spending on “pro-poor” sectors is lower in Oromia. These istration spending, as a share of total expenditure, decreased A full description of the methodology and variables used in the construction of this index in Ethiopia can be found in the technical 46  appendix. 47  These are generally thought of as variables over which a person does not have control, such as gender or region of birth. 49 Social sector includes education, health, culture and sports, and labor and social affairs while economic sector includes agriculture 48  Accessibility was also tested using a market accessibility variable, and the results were broadly consistent. and rural development, water resources, trade and industry, mines and energy, transportation and communication, and construction. 96 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 97 increase in per capita spending on pro-poor activities, at an share on pro-poor sectors combined decreased from 65 per- Figure 13 PER CAPITA PUBLIC EXPENDITURE INCREASED OVER TIME even faster rate, in general, compared to Oromia. Per capita cent in 2011 to 38 percent in 2016 in Oromia. Decrease in the spending in each of the pro-poor sectors in 2016 was lower share of spending occurred in all four sectors. For example, Per capita real public expenditure, 2011 to 2016 in Oromia compared to the rest of the county (Figure 15A). the share of agriculture and rural development dropped from 10 percent to 6 percent, while that of education dropped Oromia National excl. Oromia However, the share of spending on pro-poor activities from 34 percent to 19 percent. For the rest of the country, decreased in Oromia between 2011 and 2016, while 2,800 the spending shares on each of the sectors remained about it hardly changed in the rest of Ethiopia. The spending 2,400 the same (Figure 15B). In 2016 Birr 2,000 1,600 1,200 800 Figure 15 EXPENDITURE SHARE ON PRO-POOR SECTORS DECREASED IN 400 OROMIA BETWEEN 2011 AND 2016 0 2011 2012 2013 2014 2015 2016 Real per capita expenditure and share of total expenditure (%) for selected sectors, 2011 and 2016 Source: MoF; World Bank staff calculations. A. Real per capita expenditure in 2016 price 2011 2016 from 49 percent to 29 percent, and from 24 percent to 15 rest of the country, but it is still large, particularly in the rest 400 percent, respectively. Spending share on economic sectors of the country. accounting for 42 percent in 2016 (Figure 14). 350 remained roughly the same, while the share of spending on 300 Per capita spending50 in selected “pro-poor” activities In 2016 Birr “other” activities jumped from zero to 30 percent. For the rest 250 increased between 2011 and 2016. In Oromia, the in- 200 of the country, spending on the economic sector increased crease was particularly strong for water resources and health 150 from 22 percent to 29 percent, while the spending share on sectors; per capita spending on water and health sectors 100 the social sector remained the same at 27 percent. Admin- increased from Birr 53 to 81 Birr and from Birr 93 to Birr 50 istration-related spending decreased both in Oromia and the 128, respectively. The rest of the country also witnessed an 0 Agriculture and Rural Development Water Resources Education Health Agriculture and Rural Development Water Resources Education Health Figure 14 EXPENDITURE SHARE ON ADMINISTRATION AND SOCIAL SECTOR DECREASED Expenditure share by sectors (%), 2011 to 2016 Oromia National excl. Oromia Oromia National excl. Oromia B. Expenditure share (%) Administration Economic Social Other Agriculture & Rural Dev. Water Resources Education Health 2016 2016 70 14.5 7.4 60 13.2 23.6 50 2016 2016 2016 9.2 2016 33.7 49.3 27.2 26.8 41.9 40 30.1 26.0 36.0 8.3 6.6 7.6 27.1 30 22.1 18.8 18.8 16.8 2016 20 29.4 7.5 2016 10 5.8 29.3 5.2 4.9 10.4 5.7 5.9 7.3 0 Note: The inner circles present the 2011 values. 2011 2016 2011 2016 Source: MoF; World Bank staff calculations. Oromia National excl. Oromia 50 Per capita spending refers to total spending in a particular sector divided by the population of the region. For example, per capita Note: The sum does not add up to 100% because it includes only shares of selected “pro-poor” sectors. spending on education refers to total spending on education divided by the total population of the region and not spending per Source: MoF; World Bank staff calculations. student. 98 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 99 POVERTY AND INEQUALITY Food and nonfood consumption in the Somali region urban Somali was slightly lower compared to urban areas in increased significantly only in urban areas between the rest of the country in 2011, and the gap widened in 2016 2011 and 2016. Mean urban consumption per adult in- due to difference in growth between 2011 and 2016. The creased by 12 percent in Somali, compared to a 34 percent increase in total and rural consumption in Somali were not increase for all other regions combined. Consumption in statistically significant (Table 1). Somali Table 1 CONSUMPTION INCREASED BETWEEN 2011 AND 2016 ONLY IN URBAN SOMALI Mean annual household consumption per adult equivalent in December 2015 prices, 2011 and 2016 INTRODUCTION TOTAL URBAN RURAL 2011 2016 % change 2011 2016 % change 2011 2016 % change This chapter focuses on the evolution of socio-economic and health indicators, such as poverty, inequality, Somali 10,565 11,714 10.9 12,942 14,470 11.8* 10,004 11,063 10.6 access to basic services, and education, in the Somali region between 2011 and 2016. Using data from a National excl. variety of sources, mainly the twinned household living standards surveys (HCES and WMS)51 and the De- Somali 11,018 12,522 13.6* 13,925 18,761 34.7* 10,442 11,013 5.5* mographic and Health Survey (DHS), the chapter begins by documenting changes in monetary poverty and Note: * represents changes that are statistically significant. living standards in the region. It then moves to an analysis of non-monetary welfare before turning to chal- Source: HCES, 2011; 2016; World Bank staff calculations. lenges of inequality of opportunity. The final section discusses changes in the sectoral allocations of social spending in the region between 2011 and 2016. The chapter aims to provide policy makers with information and analysis that will be useful in improving the effectiveness of poverty reduction programs in the region. However, Somali experienced strong poverty reduction, country it decreased from 30 percent to 24 percent. In rural particularly in rural areas. Poverty in Somali decreased Somali, poverty decreased by more than a half from 35 per- Some of the main findings include: faster than the rest of the country; in Somali, the proportion cent to 16 percent. Urban poverty did not decrease signifi- of people below the national poverty line dropped by almost cantly in Somali, indicating that the increase in consumption ● The poverty rate in Somali dropped from 33 percent to health and education indicators. This is also reflected a half from 33 percent to 17 percent, while in the rest of the did not benefit the urban poor (Table 2). 17 percent between 2011 and 2016. However, poverty in a lower score on the Human Opportunity Index than reduction in urban Somali was not statistically significant. the national average -implying lower equality of access to key basic services for children in Somali compared to ● Although consumption growth in Somali was weak Table 2 POVERTY DECREASED IN SOMALI MORE THAN THE REST OF THE the rest of Ethiopia. in general, those around the poverty line experienced COUNTRY 52 strong consumption growth, particularly in rural areas. ● Although average per capita public expenditure in So- Percentage of people below the national poverty line This led to a modest decrease in inequality, and by 2016 mali is similar to the rest of the country, it spends propor- inequality in the region was the lowest in Ethiopia. tionally higher on administration. Somali spends lower TOTAL URBAN RURAL on health and education relative to the other regions ● There were also some improvements in non-monetary % pt. % pt. % pt. combined. 2011 2016 change 2011 2016 change 2011 2016 change welfare, but Somali still lags behind the rest of the coun- try in most of the non-monetary indicators including Somali 32.8% 16.8% -16.0* 23.1% 19.4% -3.7 35.1% 16.3% -18.8* National excl. 29.5% 23.6% -5.9* 25.7% 14.4% -11.3* 30.3% 25.8% -4.4* Somali Note: * represents changes that are statistically significant. Source: HCES, 2011; 2016; World Bank staff calculations. 52 In 2011, the HCES covered only three zones in Somali. In 2016, the coverage was improved –eight zones were covered in Somali. While the 2016 data is more representative of pastoral areas and the two regions, to maintain comparability households sampled from 51 Household Consumption Expenditure Survey (HCES) and Welfare Monitoring Survey (WMS) are the two official household surveys the new zones are excluded from the analysis. If households from the new zones of Somali are included in the 2016 analysis, poverty collected by CSA every five year and used to monitor welfare. HCES covers monetary welfare and WMS deals with non-monetary rates in 2016 for Somali would be higher. Overall, urban and rural poverty rates become respectively 22.4 percent, 22.9 percent and welfare like health and education. Both surveys are collected from the same sample of households. 22.3 percent. 100 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 101 Consumption growth in Somali was weak both in ur- growth was not significantly different from zero for the bottom Lower inequality in both rural and urban areas com- households (as shown in the previous figure). As a result ban and rural areas. In urban areas, consumption growth 15 percent, and then becomes positive and rose quickly be- bined to decrease inequality overall in Somali between of these dynamics, the “between” share of inequality – the was negative for the very bottom quintile, and then zero up fore dropping again, becoming zero after the 70th percentile 2011 and 2016. Average consumption was higher in urban part of inequality resulting from differences in average con- to the 50th percentile. Consumption grew afterwards, but (Figure 1). than rural areas in 2011, but the gap was far smaller than the sumption between urban and rural areas – increased from not strongly. In rural areas, on the other hand, consumption rest of Ethiopia. The gap further increased slightly in 2016 14.5 percent in 2011 to 16.9 percent in 2016. On the other (see Table 1), with urban consumption growing by around 12 hand, the “within” share of inequality – the part resulting from percent and rural consumption growing by around 11 per- changes in distribution within urban and rural areas – dimin- cent. However, the key factor helping to reduce inequality in ished from 64.6 percent to 60.5 percent (Figure 2). Figure 1 GICS FOR SOMALI URBAN AND RURAL Somali was strong growth of consumption for the poorest Urban Rural Figure 2 OVERALL INEQUALITY DECREASED EVEN THOUGH THERE WAS A SMALL INCREASE IN THE GAP BETWEEN URBAN AND RURAL AREAS Decomposition of the Gini coefficient into a between rural-urban component and a within-component Overlap Within Between Relative contribution to Gini (%) 100% 90% 20.9 22.6 80% 70% 64.6 60.5 60% 50% 40% 30% 20% 10% 14.5 16.9 Source: HCES, 2011; 2016; World Bank staff calculations. 0% 2011 2016 Source: HCES, 2011; 2016; World Bank staff calculations. Because of weak consumption growth in urban ar- 10th percentile remained 3.7. The fact that inequality did not eas, and the lack of growth at the top in rural areas, increase between 2011 and 2016 is a departure from the inequality in Somali dropped slightly. The Gini index national pattern, where inequality increased; inequality is rel- Unlike other regions of Ethiopia, both growth in con- This means that the equalization of the distribution in Somali decreased from 29 to 26 between 2011 and 2016. The ra- atively lower in Somali compared to national inequality with a sumption and reduction in inequality contributed sig- was a significant driver of poverty reduction, while in the rest tio of consumption of the 90th percentile compared to the Gini index of 33 in 2016 (Table 3). nificantly to poverty reduction in Somali between 2011 of the country decrease in equality of the distribution caused and 2016. Figure 5 decomposes the change in poverty into poverty rates to rise. If there were no changes in the distribu- growth and redistribution components using the Datt-Ra- tion between 2011 and 2016, poverty in Somali would have Table 3 INEQUALITY INCREASED BETWEEN 2011 AND 2016 vallion decomposition technique. It indicates which part of decreased by only 5.8 percentage points. In other words, the poverty change can be attributed to: (i) the change in if the consumption distribution in 2011 was maintained in average consumption (growth effect), and (ii) the change in 2016, poverty reduction would have been weaker in Somali, 2011 2016 the redistribution of consumption or changes in inequality and it would have lagged the rest of Ethiopia. Instead, overall Gini index p90/p10 ratio Gini index p90/p10 ratio (redistribution effect). The growth effect on poverty in Soma- poverty reduction in Somali was about 10 percentage points Somali 28.6 3.7 26.2 3.7 li is negative (poverty-reducing). The redistribution effect is stronger than the rest of the country (Figure 3). National 29.8 3.6 32.9 4.3 negative for Somali, but positive for the rest of the country. Source: HCES, 2011; 2016; World bank staff calculations. 102 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 103 In Somali, households that primarily depend on casual lowest poverty rates. Somali diverges from the rest of the Figure 3 THE REDUCTION OF INEQUALITY DROVE POVERTY REDUCTION labor have the highest poverty rate, followed by those country in that variation in poverty rates across the different who rely on pensions and donations. Those who depend groups are relatively small (Figure 5). Datt-Ravallion decomposition of poverty change between 2011 and 2016 on salaried employment and wholesale and retail have the Somali National excl. Somali Contribution to change in poverty 5.0 3.5 Figure 5 CASUAL LABORERS AND THOSE WHO DEPEND ON PENSIONS/ 0.0 DONATIONS HAVE THE HIGHEST POVERTY RATES -5.0 -5.9 -5.8 Poverty rates by household’s main income source, 2016 -9.4 -10.2 -10.0 Somali National excl. Somali -15.0 -16.0 40% -20.0 35% Poverty reduction Consumption growth Increase in inequality 30% Poverty rate 25% Source: HCES, 2011; 2016; World Bank staff calculations. 20% 15% 10% 5% Average consumption in Somali remains low for the of the distribution. Somali closely mirrors consumption pat- 0% first four quintiles. Average annual consumption per adult terns for the rest the country, although the jump from the or g ck ns il t en ta in in 2016 was 5.317 Birr for the poorest quintile and 12,930 fourth to the fifth quintile in Somali is smaller than for the rest lab to io re ep m at es oy d ke al on liv an su Birr for the fourth quintile. Average consumption shows a of the country; the richest quintile of households in Somali pl ck /d em d Ca ale an ns to les sharp increase from the fourth to the fifth (richest) quintile, are, on average, about 5,000 Birr worse off than the richest es io ied p ns ro liv ho lar -c Pe lf- -w with the richest 20 percent consuming 21,238 Birr per year quintile in the rest of Ethiopia (Figure 4). Sa Se lf Se lf Se on average, reflecting strong consumption growth at the top Main source of household income Source: HCES 2016; World Bank staff calculations. Note: Only main income groups with at least 100 observations in Somali are included. Figure 4 HOUSEHOLD CONSUMPTION REMAINS FAIRLY LOW EXCEPT FOR THE TOP QUINTILE Average household expenditures per capita per year, 2016 Educational attainment of household heads strongly education of the household head leads to decreasing pover- Somali National excl. Somali negatively correlates with poverty rates. With a poverty ty rates, reaching merely four percent for those households 30,000 rate of 25 percent, poverty in Somali is highest among house- headed by someone who has at least completed secondary Average annual consumption holds headed by someone with no education. Increasing education (Figure 6). 25,000 20,000 15,000 10,000 5,000 0 Poorest 2 3 4 Richest Consumption quintiles Source: HCES 2016; World Bank staff calculations. 104 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 105 When decomposing poverty reduction by sector, So- percent of poverty reduction in Somali over the period. Al- Figure 6 EDUCATION IS A MAIN CORRELATE OF POVERTY mali closely matches the rest of the country, with ag- though this effect has been important in the rest of Ethiopia, riculture playing a dominant role. Improvements in the it is small in Somali. There was a slight fall in the share of the Poverty rates by educational attainment of the household head, 2016 agricultural sector explain about 90 percent of poverty re- population in Somali relying on agriculture over the period – duction in Somali, a considerably higher figure compared to much smaller than in the rest of the country. In 2011, 67 per- Somali National excl. Somali about 60 percent for the rest of Ethiopia. Gains for house- cent of households in Somali relied primarily on agriculture as 35% holds working in the services sector also contributed to the primary source of income. In 2016 this had fallen slightly 29% poverty reduction in Somali. The so-called “population shift” to 63 percent. At the same time, the share of households 30% 25% effect had no significant impact on poverty reduction in So- relying primarily on services also fell slightly, with households 25% 22% Poverty rate 19% mali. This variable, often used as an indicator of structural moving more towards construction and other primary sourc- 20% transformation (proxied by the shift in population from ag- es of work (Figure 8). 15% 13% riculture to manufacturing or services), explained around 3 9% 10% 8% 4% 5% 0% No education Incomplete Incomplete Secondary and Figure 8 THE AGRICULTURE SECTOR REMAINS THE LARGEST CONTRIBUTOR TO primary secondary above POVERTY REDUCTION IN SOMALI Education level of household head Sectoral decomposition of poverty changes 2011 to 2016 Source: HCES 2016; World Bank staff calculations. Agriculture Manufacturing Construction Services Note: Poverty rate for household heads with completed primary education omitted because of low sample size. Other Population shift Interaction Somali National excl. Somali 2 Change in poverty headcount Changes in rural areas were responsible for nearly all and a mere 0.3 percentage points to urban areas. The contri- 0 -2 poverty reduction in Somali between 2011 and 2016. bution of rural areas clearly reflects very strong growth at the -4 Figure 7 shows that of the 16 percentage point reduction in bottom of the rural distribution (as shown earlier in Figure 2). -6 poverty in Somali, 15.6 points are attributed to rural areas, -8 -10 -12 -14 -16 Figure 7 POVERTY REDUCTION IN SOMALI LOOKED DIFFERENT TO POVERTY -18 REDUCTION IN THE REST OF ETHIOPIA Source: HCES; 2011, 2016. World Bank staff calculations. Rural-urban decomposition of poverty changes 2011 to 2016 Rural Urban Population shift Interaction Somali National excl. Somali 0 Change in poverty headcount -2 -4 -6 -8 -10 -12 -14 -16 -18 Source: HCES, 2011; 2016; World Bank staff calculations. 106 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 107 NON-MONETARY WELFARE Figure 9 OWNERSHIP OF MOBILE PHONES INCREASED BETWEEN 2011 AND 2016 Improvements in non-monetary welfare generally ac- shortage, the average number of months with food shortage Household durables, housing conditions and access to energy in 2011 and 2016, % of households companied the increase in consumption expenditure decreased from four to three. For the rest of Ethiopia, the with asset/characteristic in Somali region. The percentage of households that re- proportion of households who faced food shortage dropped ported a food shortage in the last 12 months before the in- from 21 percent to 11 while the average number of months A. Household Durables terview decreased from 30 percent in 2011 to seven percent they faced food shortage did not change (Table 4). 2011 2016 in 2016 in the region. Furthermore, for those who faced food 70 60 Table 4 FOOD SECURITY IMPROVED BETWEEN 2011 AND 2016 50 Incidence of food shortages and average duration of food shortages – food gap 40 30 20 2011 2016 10 Food shortage (%) Food gap (months) Food shortage (%) Food gap (months) 0 Somali 30.3 4.4 7.2 3.3 Mobile Mobile Refrigerator Refrigerator Animal Cart Motorcycle Animal Cart Motorcycle TV TV National excl. Somali 21.4 3.2 10.5 3.2 Source: WMS, 2011; 2016; World bank staff calculations. Somali National excl. Somali However, ownership of household durables and access wall and floor also increased in Somali, while access to im- to electricity did not improve in Somali. While owner- proved roof and electricity deteriorated. Use of electricity B. Housing/energy ship of mobile phones increased significantly between 2011 for cooking is still extremely low. The rest of the country im- and 2016, ownership of televisions and refrigerators slightly proved in ownership of household durables, housing condi- 2011 2016 dropped, and that of animal carts and motorcycles hardly tions, and access to electricity (Figure 9). 70 changed. The proportion of households with an improved 60 50 40 30 20 10 0 Improved wall Floor: tiles/cement Access to electricity Electricity for cooking Improved wall Floor: tiles/cement Access to electricity Electricity for cooking Improved roof Improved roof Somali National excl. Somali Source: DHS, 2011; 2016; World Bank staff calculations. 108 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 109 Figure 10 THERE WAS NOT IMPROVEMENT IN ACCESS TO IMPROVED WATER Figure 11 CHILDREN’S MORTALITY DECREASED, AND THEIR HEALTH IMPROVED BETWEEN 2011 AND 2016 BETWEEN 2011 AND 2016 2011 2016 100 90.4 87.9 Access to water in 2011 and 2016, % of households with access/characteristics Selected health and education variables 2011 79.6 in 2011 and 2016 2016 80 100 90.4 87.9 66.4 Somali National excl. Somali 59.9 A. Child and infant 59.4rates (per 1000) 79.6mortality 60 53.6 80 47.5 2011 2016 66.4 Piped water - compound Piped -outside Protected well/tube well Protected spring 59.9 39 59.4 40 60 100 53.6 90.4 30.4 47.5 87.9 Other improved Unimproved sources 21 79.6 19.9 39 20 40 80 30.4 100% 100% 66.4 59.9 21 59.4 19.9 54.03 57.78 46.17 34.46 0 53.6 20 60 47.5 80% 80% Infant mortality Child mortality Under five Infant mortality Child mortality Under five 39 0 40 rate rate mortality rate rate rate mortality rate 30.4 60% 60% Infant mortality Child mortality Somali 21 Under five Infant mortality Child excl. National mortality 19.9Somali Under five 12.09 20 rate rate mortality rate rate rate mortality rate 1.22 9.2 16.95 40% 6.8 40% 10.02 Somali National excl. Somali 0.72 0.79 0 16.3 18.67 23.39 21.19 Infant mortality Child mortality Under five Infant mortality Child mortality Under five 20% 17.74 20% rate rate mortality rate rate rate mortality rate 17.39 11.15 14.58 Somali National excl. Somali 0% 4.43 4.16 0% Source: DHS, 2011; 2016; World Bank staff calculations 2011 2016 2011 2016 2011 2016 Source: DHS, 2011; 2016; World Bank staff calculations. B. Immunization and health facility delivery (percent) 50 44.7 2011 2016 39.2 38.8 40 50 33 44.7 27.4 26.6 39.2 38.8 30 40 24.5 Access to improved drinking water sources53 did not to 21, and from 90 to 80 between 2011 and 2016; but infant 21.8 33 2011 2016 17.9 16.6 27.4 improve in Somali. The proportion of households with ac- mortality increased from 54 percent to 60 percent, which is 20 50 30 26.6 24.5 44.7 21.8 10 cess to improved water sources in Somali decreased from a departure from the rest of the country. Delivery at health 10 7.6 17.9 16.6 39.2 38.8 40 20 46 percent to 42 percent between 2011 and 2016. For the facility increased from 8 to 18 percent, while the percentage 33 10 0 30 7.6 27.4 26.6 rest of the country, on the other hand, the proportion of of fully immunized children increased from 17 to 22 percent. 10 24.5 Health facility Fully 21.8 Stunted Health facility Fully Stunted households with access to drinking water increased from 54 The stunting rate decreased from 33 percent to 27 percent. 20 17.9 delivery immunized 16.6 children delivery immunized children 0 percent to 65 percent (Figure 10). Primary school completion and gross secondary enrollment Health facility children Fully Stunted Health children 10 facility Fully Stunted 7.6 rates also improved, while net primary school enrollment 10 delivery immunized children delivery immunized children Somali children National excl. Somali children Health and education indicators also improved, al- deteriorated.54 Somali lags the rest of Ethiopia combined in 0 though challenges remain. Child and under-five mortality Somali excl. Somali Stunted NationalFully most health and education indicators (Figure 11). Health facility Fully Stunted Health facility rates in Somali decreased respectively from 39 per thousand Source: DHS, 2011; delivery immunized 2016; World Bank children staff calculations delivery immunized children children children Somali C. Education (percent) National excl. Somali 2011 2016 80 72.6 2011 2016 62.5 58.5 80 60 72.6 62.5 42.7 58.5 60 40 2011 2016 32.9 28 30.9 80 42.7 72.6 22.6 25.6 21.7 40 20 62.5 32.9 30.9 11.65 28 9.9 58.5 25.6 60 21.7 22.6 20 0 11.65 42.7 9.9 40 Net primary Primary school Gross Net primary Primary school Gross 32.9 30.9 53  Improved drinking water sources include piped water, protected well, protected spring, rainwater and bottled water. enrolment completion 28 secondary enrolment completion secondary 0 22.6 25.6 Net primary (15-24) Primary school school 21.7 Gross Net primary (15-24) Primary school school Gross In 2011, the WMS covered only three zones in Somali. In 2016, the coverage was improved –eight zones were covered in Somali. 54  20 enrolment 11.65 completion enrolment secondary enrolment completion enrolment secondary While the 2016 data is more representative of pastoral areas and the two regions, to maintain comparability households sampled from 9.9 the new zones are excluded from the analysis. If households from the new zones of Somali are included in the 2016 analysis, the val- (15-24) Somali school National(15-24) excl. Somalischool ues of the education indicators become lower, for example, gross secondary enrollment becomes 21 percent instead of 26 percent. enrolment 0 2011; 2016; World Bank staff calculations. enrolment Source: WMS, Net primary Primary school Somali Gross Primary Net primary National school excl. Somali Gross enrolment completion secondary enrolment completion secondary 110 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS (15-24) school (15-24) school 111 enrolment enrolment INEQUALITY OF OPPORTUNITY Figure 12 SOMALI LAGS BEHIND THE REST OF ETHIOPIA IN ALL SIX KEY To assess inequality of 0pportunity55 in Somali, we focus on 2016. This indicates that educational opportunity is quite dif- OUTCOMES six key outcomes: (i) primary school enrollment, (ii) prima- ferent for the older cohort of children (15-18) than the young- Coverage, HOI, and D-index for different outcomes, 2016 ry school completion, (iii) secondary school enrollment, (iv) er cohort (7-14). access to electricity; (v) access to improved water source; The HOI scores for Somali are, like coverage scores, Somali National excl. Somali and (vi) access to a health post within 5 km. Circumstance lower than the rest of Ethiopia. A higher HOI score is as- 80 variables56 of the household include rural/urban location, Coverage HOI D-index sociated with more equitable access to a key service or out- 70 gender, religion, agro-ecological zone, consumption quintile come. For a given coverage rate, the lower the HOI score, 60 of the child’s household, and the household’s score on the Percent/score the more unequal the access to that key outcome is across 50 rural accessibility index (RAI).57 Figure 12 presents the cover- different groups. In Somali, the coverage rate of access to age, Human Opportunity Index (HOI), and Dissimilarity Index 40 electricity is 12 percent. However, this is heavily skewed to- (D-Index) across these key outcome variables in 2016. 30 wards the urban population in the region. The very low HOI 20 In terms of coverage, Somali lags the rest of Ethiopia score of 4 for electricity reflects strong differences in access 10 on all key outcomes, but in particular for key education across the population in the region. 0 variables. Although primary school completion and second- alt ter alt ter alt ter st st st ec ry ool ec ry ool ec ry ool l l l m ve ess m ve ess m ve ess nd sc l co ry ool co ry ool The dissimilarity index, or D-index, shows how much ity oo ty oo ty oo oo po po po ary school enrollment are already low in Ethiopia, they are he wa he a he a a h a h a h ric sch ici ch ici ch im ch im ch im ch 5k pro cc 5k pro cc 5k pro cc of d w of d w nd sc nd sc h h h s s pr ry s pr ry s pr ry s inequality in access there is between groups – a high- a a a of d co ry particularly low in Somali. This means that the HOI scores a a a a a a En let rim En let rim En let rim er index means less equal access across groups. The r r t t t for these key outcomes are also low (see the Coverage and m m m m dp p p I I I ro ed ro ed ro ed se se se m ed m ed El El El D-index for completed primary school and secondary school Co rolle HOI panels in Figure 12). The D-index for these variables – Co roll Co roll d d d in in in lle lle lle p p p ith ith ith En En En enrollment in Somali is very close to the scores for the rest of measuring how much these outcomes differ according to cir- W W W Ethiopia. This indicates that circumstances such as location, cumstances – is correspondingly high, although similar to the gender, and household income do not matter more in Somali Source: HCES, WMS 2011; 2016; World Bank staff calculations. rest of the country. This implies strong differences between than in the rest of the country. The D-index scores for primary households in which children complete primary school, en- school enrollment and electricity access in Somali were high- roll in secondary school, and have access to electricity, and er than in the rest of the country, implying greater inequality in those that do not. Decomposing this difference (not shown) reveals that this is largely explained by rural-urban differences the distribution of access to these key outcomes. PUBLIC EXPENDITURE in access rates, both within Somali and in the rest of Ethiopia. Unequal coverage of infrastructure access variables In the previous sections, we discussed trends in monetary and 2016, the spending share of the economic sector in- remains a challenge in Somali and the rest of the coun- and non-monetary welfare between 2011 and 2016, and in- creased, while that for the social sector decreased. There The primary school enrollment rate in Somali is almost try. The D-index was highest for access to electricity, exem- equality of opportunities focusing on 2016, both in Somali was a shift in favor of the economic sector. Per capita spend- 21 percentage points lower than in the rest of Ethio- plifying the rural-urban divide in access in the region, and and the rest of the country. In this section, we discuss trends, ing on health and education remain low relative to the rest pia. In Somali, although 52 percent of age-eligible children broadly throughout the country. As noted earlier, decompo- volume, and allocation of public expenditures to identify vari- of the country. These results speak to the fact that Somali were enrolled in primary school in 2016, only 24 percent of sition of this index (not shown) reveals that access depends ations across regions and possible links to welfare. We use lags in health and education indicators relative the rest of the age-eligible children had completed primary school. These far more on whether a household is in an urban or rural area MoF data. We express all expenditure values in 2016 prices country. numbers are both lower than the rest of Ethiopia, in which than it does on how the level of household wealth. For ex- using the non-food CPI deflator. We first present total per close to three-quarters of primary school-aged children are ample, in urban areas even the poorest households tend to Average per capita public expenditure in Somali in- enrolled and almost one-third complete primary education. capita public expenditure and then discuss the expenditure have access to electricity, while in rural areas even the richest creased between 2011 and 2016 as was the case for By comparison, only one in eight age-eligible children in So- allocation by sector. Finally, we present expenditure on se- households generally do not have access. rest of Ethiopia. Per capita public spending in the Somali mali (aged 15 to 18) were enrolled in secondary school in lected “pro-poor” sectors, which include agriculture and rural region increased from Birr 1038 per person in 2011 to Birr development, water resources, health and education. 2,026 per person in 2016 (in constant 2016 Birr). Average Though per capita public expenditure in Somali was spending per capita in Somali was similar to the rest of the similar to the other regions combined, Somali spends country in 2016 (Figure 13). more proportionally on administration. Between 2011 A full description of the methodology and variables used in the construction of this index in Ethiopia can be found in the technical 55  appendix. 56  These are generally thought of as variables over which a person does not have control, such as gender or region of birth. 57  Accessibility was also tested using a market accessibility variable, and the results were broadly consistent. 112 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 113 per capita spending increased from Birr 93 to Birr 426 be- In Somali, the share of spending on pro-poor activities Figure 13 PER CAPITA PUBLIC EXPENDITURE INCREASED OVER TIME 58 tween 2011 and 2016. The rest of the country also witnessed remained the same between 2011 and 2016, while it increase in per capita spending increment on pro-poor activi- contracted in the rest of the country. The share of spend- Per capita real public expenditure, 2011 to 2016 ties (Figure 15A). In 2016, per capita spending in all pro-poor ing on agriculture and rural development in Somali increased activities, except agriculture and rural development, were low- from 9 percent to 22 percent between 2011 and 2016, while Somali National excl. Somali er in Somali compared to the rest of the country (Figure 15A). spending on water resources dropped from 12 percent to 4 2,100 percent. The share of spending in the education and health 1,800 sectors also decreased (Figure 15B). In 2016 Birr 1,500 1,200 900 600 EXPENDITURE SHARE ON AGRICULTURE AND RURAL DEVELOPMENT Figure 15 300 0 INCREASED IN SOMALI BETWEEN 2011 AND 2016 2011 2012 2013 2014 2015 2016 Real per capita expenditure and share of total expenditure (%) for selected sectors, 2011 and 2016 Source: MoF; World Bank staff calculations. A. Real per capita expenditure in 2016 price 2011 2016 Public expenditure growth in Somali is accompanied but it is still large, particularly in Somali, where it accounted 450 by a spending shift away from administration and the for 46 percent of total public spending in 2016 (Figure 14). 400 social sector towards the economic sector.59 Economic 350 Per capita spending60 in selected “pro-poor” activities In 2016 Birr sector spending, as a share of total expenditure, increased 300 increased between 2011 and 2016. In Somali, per capita 250 from 25 to 35 percent, while the share of social sector spend- spending increased in all pro-poor activities except health, for 200 ing decreased from 24 percent to 18 percent. This mirrored which per capita spending decreased. The increase was par- 150 changes in the rest of the country. Administration-related 100 ticularly strong for agriculture and rural development, where 50 spending decreased both in Somali and the rest of the country, 0 Agriculture and Rural Development Water Resources Education Health Agriculture and Rural Development Water Resources Education Health Figure 14 SPENDING SHARES OF SOCIAL AND ECONOMIC SECTORS INCREASED Expenditure share by sectors (%), 2011 to 2016 Somali National excl. Somali Somali National excl. Somali Administration Economic Social Other 2016 2016 B. Expenditure share (%) 1.5 14.9 2016 Agriculture & Rural Dev. Water Resources Education Health 2016 28.6 18.2 23.8 50 2016 7.7 45 50.9 32.4 36.9 1.5 40 9.1 6.3 8 25.3 35 11.1 7.9 2016 2016 23.1 2016 30 13.3 22.8 34.5 28.5 28.0 17.8 25 4.3 20 22.1 12.1 15 Note: The inner circles present the 2011 values. 10 5 5.7 5 8.8 Source: MoF; World Bank staff calculations. 6.7 5.8 0 2011 2016 2011 2016 Somali National excl. Somali 58  Data for 2014 is missing for Somali. Note: The sum does not add up to 100% because it includes only shares of selected “pro-poor” sectors. 59 Social sector includes education, health, culture and sports, and labor and social affairs while economic sector includes agriculture and rural development, water resources, trade and industry, mines and energy, transportation and communication, construction. Source: MoF; World Bank staff calculations. Per capita spending refers to total spending in a particular sector divided by the population of the region. For example, per capita 60  spending on education refers to total spending on education divided by the total population of the region and not spending per student. 114 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 115 POVERTY AND INEQUALITY Food and nonfood consumption in the Benishan- percent increase for all other regions combined. The increase gul-Gumuz region increased significantly between in consumption can be solely attributed to urban areas, where 2011 and 2016, although only urban areas experienced consumption increased by more than 23 percent, lower than the full benefit. Mean consumption per adult increased 17 the national increase of 34 percent. Increase in consumption percent in Benishangul-Gumuz alone, compared to a 14 in rural areas was not statistically significant (Table 1). Benishangul-Gumuz Table 1 CONSUMPTION INCREASED BETWEEN 2011 AND 2016 BOTH IN BENISHANGUL-GUMUZ AND THE REST OF THE COUNTRY Mean annual household consumption per adult equivalent in December 2015 prices, 2011 and 2016 INTRODUCTION TOTAL URBAN RURAL 2011 2016 % change 2011 2016 % change 2011 2016 % change This chapter focuses on the evolution of socio-economic and health indicators such as poverty, inequality, Benishangul- access to basic services, and education, in the Benishangul-Gumuz region between 2011 and 2016. Using 11,435 13,373 17.0* 15,124 18,524 22.5* 10,832 12,112 11.8 Gumuz data from a variety of sources, mainly the twinned household living standards surveys (HCES and WMS)61 National excl. 11,004 12,490 13.5* 13,890 18,651 34.3* 10,429 11,001 5.5* and the Demographic and Health Survey (DHS), the chapter begins by documenting changes in monetary B-G poverty and living standards in the region. It then moves to an analysis of non-monetary welfare before Note: * represents changes that are statistically significant. turning to challenges of inequality of opportunity. The final section discusses changes in sectoral allocations Source: HCES, 2011; 2016; World Bank staff calculations. of social spending in the region between 2011 and 2016. The chapter aims to provide policy makers with information and analysis that will be useful in improving the effectiveness of poverty reduction programs in Decreasing poverty did not accompany the increase areas for most of the consumption distribution. The 40 the region. in consumption in Benishangul-Gumuz.62 The decrease percent at the bottom are being left behind in both urban and in poverty in Benishangul-Gumuz was not statistically signif- rural areas. In urban areas, consumption growth was largely Some of the main findings include: icant both in urban and rural areas, while poverty decreased not significantly different from zero up to the 40th percentile, in the rest of the country (Table 2). after which it becomes positive and strong. In rural areas, on ● The poverty rate in Benishangul-Gumuz stood at 27 education. Benishangul Gumuz has better human cap- the other hand, there was no consumption growth until the percent in 2016, compared to 29 percent in 2011. This ital indicators, particularly education indicators than the Consumption growth in urban Benishangul-Gumuz 60th percentile and this tailed off at the top of the distribution change was not, however, statistically significant in ei- rest of the country. This is also reflected in a higher score was mostly positive, while there was no growth in rural (Figure 1). ther urban or rural parts of the region. on the Human Opportunity Index than the national av- erage – implying higher equality of access to key basic ● The pattern of consumption growth in Benishangul-Gu- services for children in Benishangul-Gumuz compared muz was inequality-increasing, where growth at the to the rest of Ethiopia. Table 2 POVERTY DID NOT DECREASE IN BENISHANGUL-GUMUZ bottom was zero or weak in urban areas and generally Percentage of people below the national poverty line weak in rural areas. Rural-urban gap further widened, ● Although average per capita public expenditure in Ben- and by 2016 inequality in the region was higher than in ishangul-Gumuz is larger than in the rest of the country, it TOTAL URBAN RURAL the rest of Ethiopia spends proportionally higher on administration. Per cap- % pt. % pt. % pt. ita spending on health, education and agriculture and ru- 2011 2016 change 2011 2016 change 2011 2016 change ● Non-monetary welfare including ownership of consumer ral development are also higher in Benishangul-Gumuz. Benishangul- durables and access to drinking water improved, as was 28.9% 26.5% -2.4 21.3% 17.7% -3.6 30.1% 28.7% -1.4 Gumuz the case for human capital indicators such as health and National excl. 29.6% 23.4% -6.2* 25.7% 14.5% -11.2* 30.4% 25.5% -4.8* B-G Note: * represents changes that are statistically significant. Source: HCES, 2011; 2016; World Bank staff calculations. 61 Household Consumption Expenditure Survey (HCES) and Welfare Monitoring Survey (WMS) are the two official household surveys collected by CSA every five year and used to monitor welfare. HCES covers monetary welfare and WMS deals with non-monetary 62 The relatively slower poverty reduction in Benishangul-Gumuz could partly be due the higher rural population share (80% for Benis- welfare like health and education. Both surveys are collected from the same sample of households. hangul-Gumuz versus 75% for the rest of the country in 2016). Poverty reduction was low in rural areas in general. 116 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 117 The rise in inequality between 2011 and 2016 is main- from differences in average consumption between urban and Figure 1 GICS FOR BENISHANGUL-GUMUZ URBAN AND RURAL ly due to the increasing gap between urban and rural rural areas – increased from 14.8 percent in 2011 to 22.2 areas. Average consumption was already significantly high- percent in 2016. On the other hand, the “within” share of in- er in urban than in rural areas in 2011. The urban-rural gap equality – the part resulting from changes in distribution within Urban Rural further increased in 2016, with rural consumption growing at urban and within rural areas – diminished from 69 percent to about half the rate of urban areas (Table 1). As a result, the 59 percent (Figure 2). “between” share of inequality – the part of inequality resulting Figure 2 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 Overlap Within Between Relative contribution to Gini (%) 100% 90% 16.6 18.6 80% 70% 68.6 59.2 Source: HCES, 2011; 2016; World bank staff calculations. 60% 50% 40% 30% 20% Relatively stronger consumption growth in urban areas, the 90th percentile compared to the 10th percentile increased 10% 22.2 14.8 particularly at the top, compared to rural areas led to from 3.9 in 2011 to 4.6 in 2016, reflecting strong consump- 0% an increase in inequality. The Gini index increased from 32 tion growth at the top relative to the bottom (Table 3). 2011 2016 to 36 between 2011 and 2016. The ratio of consumption at Source: HCES, 2011; 2016; World Bank staff calculations INEQUALITY INCREASED BETWEEN 2011 AND 2016 Poverty reduction was not statistically significant in not only for Benishangul-Gumuz, but also for the rest of the Table 3 Benishangul-Gumuz between 2011 and 2016, and this country. In fact, the increase in inequality in Benishangul-Gu- 2011 2016 was largely because of the sharp increase in inequality. muz slowed overall poverty reduction more than in the rest Gini index p90/p10 ratio Gini index p90/p10 ratio Figure 3 decomposes the change in poverty into growth and of Ethiopia. If there were no changes in the distribution be- redistribution components using the Datt-Ravallion decompo- tween 2011 and 2016, poverty would have decreased by Benishangul-Gumuz 32.0 3.9 35.6 4.6 sition technique. It indicates which part of the poverty change 9.3 in Benishangul-Gumuz and 8.7 percent in the rest of the National 29.8 3.6 32.9 4.3 can be attributed to: (i) the change in average consumption country, instead of the 2.4 and 6.1 percent we actually ob- Source: HCES, 2011; 2016; World Bank staff calculations. (growth effect), and (ii) the change in the redistribution of con- serve. In other words, if the consumption distributions in 2011 sumption; that is, changes in inequality (redistribution effect). were maintained in 2016, poverty reduction could have been The growth effect on poverty in Benishangul-Gumuz is neg- significant in Benishangul-Gumuz, and indeed it would have ative (poverty-reducing). The redistribution effect is positive outperformed the rest of Ethiopia. 118 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 119 In Benishangul-Gumuz, households that primarily de- depend on salaried employment have the lowest poverty Figure 3 INCREASED INEQUALITY LED TO NO SIGNIFICANT POVERTY REDUCTION pend on casual labor and crop production have the rates. The results for the rest of the country are similar, where highest poverty rates. Poverty for those who depend on those who depend on casual labor and crop production are Datt-Ravallion decomposition of poverty change between 2011 and 2016 casual labor in Benishangul-Gumuz is higher than in the rest poorest (Figure 5). of the country (37 percent versus 29 percent). Those who Benishangul-Gumuz National excl. Benishangul-Gumuz Contribution to change in poverty 8.0 6.0 6.9 4.0 Figure 5 CASUAL LABORERS AND CROP PRODUCERS HAVE THE HIGHEST 2.0 2.6 POVERTY RATES 0.0 -2.0 -2.4 Poverty rates by household’s main income source, 2016 -4.0 -6.1 -6.0 -8.7 Benishangul-Gumuz National excl. Benishangul-Gumuz -8.0 -9.3 -10.0 40% -12.0 35% Poverty reduction Consumption growth Increase in inequality 30% Poverty rate 25% Source: HCES, 2011; 2016; World Bank staff calculations. 20% 15% 10% Average consumption in Benishangul-Gumuz remains increase from the fourth to the fifth (richest) quintile, with the 5% low for the majority of its population. Average consump- richest 20 percent consuming 28,046 Birr per year on aver- 0% tion in the bottom three quintiles in Benishangul-Gumuz is very age, reflecting strong consumption growth at the top of the Casual labor Self - crop production Self - wholesale and retail Salaried employment close to average consumption for this part of the distribution distribution. Benishangul-Gumuz closely mirrors consump- Main source of household income in the rest of Ethiopia. Average annual consumption per adult tion patterns for the rest the country, although the increase in 2016 was 5,295 Birr for the poorest quintile and 15,052 Birr from the fourth to the fifth quintile in Benishangul-Gumuz is Source: HCES 2016; World Bank staff calculations. for the fourth quintile. Average consumption shows a sharp larger than for the rest of the country (Figure 4). Note: Only main income groups with at least 100 observations in Benishangul-Gumuz are included. Figure 4 HOUSEHOLD CONSUMPTION REMAINS FAIRLY LOW EXCEPT FOR THE Educational attainment of household heads strongly poverty rates, reaching six percent for those households TOP QUINTILE negatively correlate with poverty rates. With a poverty headed by someone who completed secondary education, Average household expenditures per capita per year, 2016 rate of 34 percent, poverty in Benishangul-Gumuz is highest and merely one percent for households headed by someone among households headed by someone with no education. In- with post-secondary education. Benishangul-Gumuz closely Benishangul-Gumuz National excl. Benishangul-Gumuz creasing education of the household head leads to decreasing mirrors the trend for the rest of the country (Figure 6). 30,000 Average annual consumption 25,000 20,000 15,000 10,000 5,000 0 Poorest 2 3 4 Richest Consumption quintiles Source: HCES 2016; World Bank staff calculations. 120 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 121 Ownership of household durables and access to elec- percentage of households owning these durables are below Figure 6 EDUCATION IS A MAIN CORRELATE OF POVERTY tricity also improved in Benishangul-Gumuz, although other regions combined. The proportion of households with an remaining low. Ownership of household durables such as improved wall, improved roof, improved floor, access to elec- Poverty rates by educational attainment of the household head, 2016 televisions, mobile phones, refrigerator, animal carts, and mo- tricity, and who use electricity for cooking also improved; but torcycles increased between 2011 and 2016, although the the region lags the rest of the country (Table 7). Benishangul-Gumuz National excl. Benishangul-Gumuz 40% 34% 35% 30% 28% Figure 7 OWNERSHIP OF CONSUMER DURABLES AND HOUSING CONDITIONS 25% Poverty rate 25% 22% IMPROVED BETWEEN 2011 AND 2016 20% Household durables, housing conditions and access to energy in 2011 and 2016, % of households 15% 13% 13% 13% with asset/characteristic 10% 8% 2011 2016 6% 6% A. Household Durables 5% 3% 1% 60 2011 2016 0% 50 60 No education Incomplete Complete Incomplete Complete Post-secondary primary primary secondary secondary 40 50 Education level of household head 30 40 Source: HCES 2016; World Bank staff calculations. 20 30 10 20 0 10 Mobile Mobile Refrigerator Refrigerator Cart Cart Motorcycle Cart Cart Motorcycle TV TV NON-MONETARY WELFARE 0 Animal Animal Mobile Mobile Refrigerator Refrigerator Motorcycle Motorcycle TV TV Improvements in non-monetary welfare accompanied percent in 2016. For those who faced a food shortage, the av- Animal Animal increase in consumption expenditure. However, food se- erage number of months with a food shortage increased from Benishangul-Gumuz National excl. Benishangul-Gumuz curity didn’t improve though it is was low relative to rest of two to three. For other regions combined, the prevalence of the country. In Benishangul-Gumuz, the percentage of house- food shortage decreased but the average number of months Benishangul-Gumuz National excl. Benishangul-Gumuz holds that reported a food shortage in the last 12 months be- households faced food shortage remained the same between fore the interview increased from six percent in 2011 to nine 2011 and 2016 (Table 4). B. Housing/energy 2011 2016 80 2011 2016 Table 4 FOOD SECURITY IMPROVED BETWEEN 2011 AND 2016 80 60 Incidence of food shortages and average duration of food shortages – food gap 60 2011 2016 40 Food shortage (%) Food gap (months) Food shortage (%) Food gap (months) 40 20 Benishangul-Gumuz 5.6 2.1 8.5 2.8 National excl. B.G. 21.8 3.3 10.4 3.2 20 0 wall wall tiles/cement to electricity for cooking wall wall tiles/cement to electricity for cooking roof roof roof roof Source: HCES, 2011; 2016; World Bank staff calculations. 0 Improved Improved Improved Improved tiles/cement to electricity for cooking tiles/cement to electricity for cooking Improved Improved Improved Improved Electricity Electricity Access Access Floor:Floor: Floor:Floor: Electricity Electricity Access Access Benishangul-Gumuz National excl. Benishangul-Gumuz Source: DHS, 2011; 2016; WorldBenishangul-Gumuz Bank staff calculations. National excl. Benishangul-Gumuz 122 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 123 Figure 8 ACCESS TO IMPROVED WATER SOURCES INCREASED BETWEEN Figure 9 CHILDREN’S MORTALITY DECREASED, AND THEIR HEALTH IMPROVED 2011 AND 2016 BETWEEN 2011 AND 2016 2011 2016 120 120 Access to water in 2011 and 2016, % of households with access/characteristics 2011 in 2011 Selected health and education variables and 2016 2016 100 120 120 87.7 82.5 Benishangul-Gumuz National excl. Benishangul-Gumuz 80 72.9 A. Child and infant mortality rates (per 1000) 100 66.9 59 87.7 60 50.8 82.5 2011 201648.1 50.1 Piped water - compound Piped -outside Protected well/tube well Protected spring 80 72.9 66.9 120 120 40 34.1 59 30.4 Other improved Unimproved sources 60 50.1 50.8 48.1 100 19.7 20 87.7 40 34.1 82.5 30.4 100% 100% 80 72.9 43.39 17.82 46.35 35.37 0 19.7 66.9 20 59 80% 80% 60 Infant mortality 50.1 Child 50.8mortality Under five Infant mortality Child mortality Under five 7.43 48.1 0 rate rate mortality rate rate rate mortality rate 54.18 40 34.1 30.4 Infant mortality Child mortality Benishangul-Gumuz Under five mortality excl. InfantNational Child mortality Under five Benishangul-Gumuz 60% 60% 11.79 19.7 7.62 20 rate rate mortality rate rate rate mortality rate 9.05 16.6 40% 26.82 40% 9.96 Benishangul-Gumuz National excl. Benishangul-Gumuz 0 23.34 21.18 Infant mortality Child mortality Under five Infant mortality Child mortality Under five 20% 17.84 20% rate rate mortality rate rate rate mortality rate 11.83 14.33 Benishangul-Gumuz National excl. Benishangul-Gumuz 4.25 8.63 11.09 0% 0% Source: DHS, 2011; 2016; World Bank staff calculations 2011 2016 2011 2016 2011 2016 Source: DHS, 2011; 2016; World Bank Staff Calculations B. Immunization 57.4 and health facility delivery (percent) 60 48.62011 2016 50 42.7 44.4 60 57.4 38.3 38.3 40 48.6 50 44.4 Access to improved drinking water sources63 also respectively from 73 per thousand to 50, from 51 to 34, and 42.7 30 25.7 23.6 2011 2016 26.2 24.3 38.3 38.3 improved in Benishangul-Gumuz. The proportion of from 120 to 83 between 2011 and 2016. Delivery at health fa- 40 60 57.4 20 26.2 households with access to improved water sources in Ben- cility increased from 9 to 26 percent, while the percentage of 30 9.1 25.7 23.6 48.6 10 24.3 50 42.7 44.4 ishangul-Gumuz increased from 57 percent to 82 percent fully immunized children increased from 24 to 57 percent. The 10 20 38.3 38.3 40 10 between 2011 and 2016, representing a faster improvement stunting rate decreased from 49 percent to 43 percent. Net 0 9.1 10 compared to the rest of Ethiopia (Figure 8). primary enrollment, primary school completion, and gross 30 Health facility 25.7 Fully 23.6 Stunted Health26.2 facility Fully 24.3 Stunted 0 delivery immunized children delivery immunized children secondary enrollment rates also improved. Although the rest 20 children children Health and education indicators improved, although Health facility Fully Stunted Health 10 facility Fully Stunted of the country also improved on educational indicators, Ben- 9.1 10 delivery immunized children delivery immunized children challenges remain. The infant mortality, child, and un- Benishangul-Gumuz children Benishangul-Gumuz National excl.children ishangul-Gumuz outperformed in education (Figure 9). der-five mortality rates in Benishangul-Gumuz decreased 0 Benishangul-Gumuz Health facility Fully Stunted National Health facility excl. Benishangul-Gumuz Fully Stunted Source: DHS, 2011; delivery immunized 2016; World Bank children staff calculations delivery immunized children children children Benishangul-Gumuz C. Education (percent) National excl. Benishangul-Gumuz 2011 2016 78.3 80 71.7 67.4 2011 2016 62.3 78.3 80 60 71.7 67.4 62.3 39.2 37.2 60 40 2011 2016 32.6 28.1 30.7 78.3 27.6 80 39.2 24.2 71.7 21.4 67.4 37.2 40 20 62.3 32.6 30.7 28.1 27.6 60 24.2 21.4 20 0 40 Net primary Primary39.2 school 37.2 Gross Net primary Primary school Gross 32.6 30.7 enrolment completion 28.1 secondary enrolment completion 27.6 secondary 0 24.2 Net primary (15-24) Primary school school Gross Net primary (15-24) Primary school school 21.4 Gross 20 enrolment completion enrolment secondary enrolment completion enrolment secondary (15-24) Benishangul-Gumuz school (15-24) school National excl. Benishangul-Gumuz 63  Improved drinking water sources include piped water, protected well, protected spring, rainwater, and bottled water. 0 enrolment enrolment Source: WMS, 2011; 2016; World Bank staff calculations. Primary school Net primary Benishangul-Gumuz Gross Net primary Primary National excl. school Gross Benishangul-Gumuz enrolment completion secondary enrolment completion secondary 124 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS (15-24) school (15-24) school 125 enrolment enrolment INEQUALITY OF OPPORTUNITY Figure 10 BENISHANGUL-GUMUZ IS AHEAD OF IN MOST KEY OUTCOMES To assess inequality of opportunity64 in Benishangul-Gumuz in secondary school in 2016. This indicates that educational Coverage, HOI, and D-index for different outcomes, 2016 and the rest of Ethiopia, we focus on six key outcomes: (i) opportunity is quite different for the older cohort of children primary school enrollment, (ii) primary school completion, (iii) (15-18) than the younger cohort (7-14). Benishangul-Gumuz National excl. Benishangul-Gumuz secondary school enrollment, (iv) access to electricity; (v) ac- The HOI scores for Benishangul-Gumuz are, like cover- cess to an improved water source; and (vi) access to a health 100 age scores, generally higher than the rest of Ethiopia. A Coverage HOI D-index post within 5 km. Circumstance variables of the household 65 90 higher HOI score is associated with a more equitable access 80 include rural/urban location, gender, religion, agro-ecologi- Percent/score to a key service or outcome across socio-economic groups 70 cal zone, consumption quintile of the child’s household, and 60 within Benishangul-Gumuz. For a given coverage rate, the the household’s score on the rural accessibility index (RAI).66 50 lower the HOI score, the more unequal the access to that key 40 Figure 12 presents the coverage, Human Opportunity Index outcome is across different groups. In Benishangul-Gumuz, 30 (HOI), and Dissimilarity index (D-index) across these key out- the coverage rate of access to electricity is 22 percent. How- 20 come variables in 2016. ever, this is heavily skewed towards the urban population in 10 0 In terms of coverage, Benishangul-Gumuz is ahead the region, even though the share of the urban population is alt ter alt ter alt ter st st st ec ry ool ec ry ool ec ry ool l l l m ve ess m ve ess m ve ess nd sc l co ry ool co ry ool ity oo ty oo ty oo oo of the rest of Ethiopia on all key outcomes except for still low in the region. The very low HOI score of 11 for elec- po po po he wa he a he a a h a h a h ric sch ici ch ici ch im ch im ch im ch 5k pro cc 5k pro cc 5k pro cc of d w of d w nd sc nd sc h h h s s pr ry s pr ry s pr ry s a a a of d electricity access. Primary school completion and second- tricity reflects strong differences in access across the popu- co ry a a a a a a En let rim En let rim En let rim r r ary school enrollment are generally quite low in Ethiopia, but lation in the region. t t t m m m m dp p p I I I ro ed ro ed ro ed se se se m ed m ed El El El Co rolle they are slightly higher in Benishangul-Gumuz. This means Co ll Co roll d d d o in in in lle lle lle r The dissimilarity index, or D-index, shows how much inequal- p p p ith ith ith En En En that the HOI scores for these key outcomes are also higher W W W ity in access there is between groups – a higher index in the region than the rest of the country (see the Coverage means less equal access across groups. The D-index for Source: HCES, WMS 2011; 2016; World Bank staff calculations. and HOI panels in Figure 10). The D-index for these variables completed primary school and secondary school enrollment – measuring how much these outcomes differ according to in Benishangul-Gumuz is lower than the rest of Ethiopia. This circumstances – is correspondingly slightly lower, although indicates that circumstances such as location, gender and similar to the rest of the country. This implies that differenc- es between households in which children complete prima- household income matter less in Benishangul-Gumuz than PUBLIC EXPENDITURE in the rest of the country. The D-index scores for secondary ry school, enroll in secondary school, and have access to school enrollment and electricity access in Benishangul-Gu- In the previous sections, we discussed the trends in mone- education. Per capita expenditures on health, education, and electricity are less prevalent in Benishangul-Gumuz than in muz were also lower than the rest of the country, implying tary and non-monetary welfare between 2011 and 2016, and agriculture and rural development were significantly higher the rest of Ethiopia. Decomposing this difference (not shown) slightly less inequality in the distribution of access to these inequality of opportunities focusing on 2016, both in Benis- than in other regions combined. These results to some ex- reveals that this is largely explained by rural-urban differences key outcomes. hangul-Gumuz and the rest of the country. In this section, we tent raise question about the effectiveness of spending, given in access rates, both within Benishangul-Gumuz and in the discuss trends, volume, and allocation of public expenditures that results presented in earlier sections show that Benishan- rest of Ethiopia. Unequal coverage of infrastructure access variables to identify variation across regions and links with welfare. We gul-Gumuz lags on most health outcomes despite higher per remains a challenge in Benishangul-Gumuz and the The primary school enrollment rate in Benishangul-Gu- use MoF data. We express all expenditure values in 2016 capita spending. rest of the country. The D-index was highest for access muz is about 7 percentage points higher than in the prices using the non-food CPI deflator. We first present total to electricity, exemplifying the rural-urban divide in access Per capita public expenditure increased between 2011 rest of Ethiopia. In the region, although 78 percent of per capita public expenditure and then discuss the expendi- to these services in the region, and broadly throughout the and 2016, and was higher in the Benishangul-Gumuz age-eligible children were enrolled in primary school in 2016, ture allocation by sector. Finally, we present expenditure on country. As noted earlier, decomposition of this index (not region than the rest of the country for all years. Public only one-third of age-eligible children had completed prima- selected “pro-poor” sectors, which include agriculture and shown) reveals that access depends far more on whether a spending in the Benishangul-Gumuz region increased from ry school. These numbers are both higher than the rest of rural development, water resources, health, and education. household is in an urban or rural area than it does on level Birr 2,285 per person in 2011 to Birr 2,808 per person in Ethiopia, in which 71 percent of primary school-aged chil- of household wealth. For example, in urban areas even the Although per capita public expenditure in Benishan- 2016 (in constant 2016 Birr). Per capita public expenditure dren are enrolled and 29 percent complete primary edu- poorest households tend to have access to electricity, while gul-Gumuz is larger than the rest of the country, a increased at an average rate of six percent annually be- cation. By comparison, fewer than one in four age-eligible in rural areas even the richest households generally do not. significantly higher proportion is spent on administra- tween 2011 and 2016 in Benishangul-Gumuz. The per capita children in Benishangul-Gumuz (aged 15 to 18) were enrolled tion in Benishangul-Gumuz. However, between 2011 and spending growth rate in other regions combined was slightly 2016, spending shifted from administration towards social higher (14 percent), and as a result the gap between Benis- and economic sectors. Improvements in per capita expen- hangul-Gumuz and the rest of the country narrowed modest- A full description of the methodology and variables used in the construction of this index in Ethiopia can be found in the technical 64  diture on “pro-poor” activities were stronger in Benishan- ly (Figure 11). appendix. gul-Gumuz compared to the rest of the country, except for 65  These are generally thought of as variables over which a person does not have control, such as gender or region of birth. 66  Accessibility was also tested using a market accessibility variable, and the results were broadly consistent. 126 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 127 41 to Birr 91, respectively. Per capita spending in education In Benishangul-Gumuz, the share of spending on wa- Figure 11 PER CAPITA PUBLIC EXPENDITURE INCREASED OVER TIME activity slightly decreased from Birr 533 in 2011 to Birr 517 ter and health activities increased between 2011 and in 2016. The rest of the country also witnessed a per capita 2016, while the share spent on education contracted. Per capita real public expenditure, 2011 to 2016 spending increase on pro-poor activities, although the rate of The share of spending on water and health sectors in Benis- change is faster in Benishangul-Gumuz, except in education hangul-Gumuz increased from two and nine percent in 2011 Benishangule-Gumuz National excl. Benishangule-Gumuz (Figure 17A). Per capita spending in 2016 was significantly to three and ten percent in 2016, respectively. The share higher in Benishangul-Gumuz than the rest of the country of education and agriculture and rural development in total 3,000 2,500 for all pro-poor sectors except water resources (Figure 13A). spending decreased between 2011 and 2016 both in the In 2016 Birr 2,000 Benishangul-Gumuz region and in other regions (Figure 13B). 1,500 1,000 500 EXPENDITURE SHARE ON WATER RESOURCES AND HEALTH Figure 13 0 INCREASED IN BENISHANGUL-GUMUZ BETWEEN 2011 AND 2016 2011 2012 2013 2014 2015 2016 Real per capita expenditure and share of total expenditure (%) for selected sectors, 2011 and 2016 Source: MoF; World Bank staff calculations. A. Real per capita expenditure in 2016 price 2011 2016 Public expenditure growth in Benishangul-Gumuz is administration-related spending decreased both in Benishan- accompanied by a spending shift away from admin- gul-Gumuz and the rest of the country, it is still large in Benis- 550 500 istration and towards social and economic sectors.67 hangul-Gumuz, accounting for 42 percent in 2016 (Figure 1)2. 450 Social and economic sector spending, as a share of total 400 In 2016 Birr Spending per capita68 in selected “pro-poor” activities 350 expenditure, increased from 30 percent to 32 percent, and 300 increased between 2011 and 2016, except in education. 250 from 15 percent to 25 percent, respectively. For the rest of The increase was particularly strong for health and water re- 200 the country, spending on the economic sector increased 150 sources activities; per capita spending on health and water 100 from 23 to 29 percent, while spending share on the social activities increased from Birr 219 to Birr 295 and from Birr 50 sector decreased from 32 percent to 28 percent. although, 0 Agriculture and Rural Development Water Resources Education Health Agriculture and Rural Development Water Resources Education Health Figure 12 SPENDING SHARE OF SOCIAL AND ECONOMIC SECTORS INCREASED Expenditure share by sectors (%), 2011 to 2016 Benishangul-Gumuz National excl. Benishangul-Gumuz Benishangul-Gumuz National excl. Benishangul-Gumuz Administration Economic Social Other B. Expenditure share (%) 2016 2016 Agriculture & Rural Dev. Water Resources Education Health 1.4 14.3 2016 29.5 45 2016 29.8 40 8.0 7.3 8.5 10.1 32.1 2016 31.8 35 7.8 54.8 41.7 37.5 22.2 30 20.8 15.1 25 17.8 17.4 2016 2016 23.4 2016 24.7 27.7 28.5 20 15 10 1.7 3.1 5.6 5.6 5 9.2 8.8 Note: The inner circles present the 2011 values. 6.8 6.8 0 Source: MoF; World Bank staff calculations. 2011 2016 2011 2016 Benishangul-Gumuz National excl. Benishangul-Gumuz 67 Social sector includes education, health, culture and sports, and labor and social affairs while economic sector includes agriculture Note: The sum does not add up to 100% because it includes only shares of selected “pro-poor” sectors. and rural development, water resources, trade and industry, mines and energy, transportation and communication, construction. Source: MoF; World Bank staff calculations. Per capita spending refers to total spending in a particular sector divided by the population of the region. For example, per capita 68  spending on education refers to total spending on education divided by the total population of the region and not spending per student. 128 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 129 POVERTY AND INEQUALITY Food and nonfood consumption in the SNNP region where consumption increased by more than one-third. Con- increased significantly between 2011 and 2016, par- sumption increase in rural SNNP was not statistically signifi- ticularly in urban areas. Mean consumption per adult in- cant. Total consumption and urban consumption in the rest creased 14 percent in SNNP between 2011 and 2016. This of the country increased by the same magnitude (Table 1). SNNP increase in consumption was mainly driven by urban areas, Southern Nations, Nationalities, and Peoples’ Region Table 1 CONSUMPTION INCREASED BETWEEN 2011 AND 2016 BOTH IN SNNP AND THE REST OF THE COUNTRY Mean annual household consumption per adult equivalent in December 2015 prices, 2011 and 2016 INTRODUCTION TOTAL URBAN RURAL 2011 2016 % change 2011 2016 % change 2011 2016 % change This chapter focuses on the evolution of socio-economic and health indicators such as poverty, inequality, SNNP 10,725 12,204 13.8* 13,391 18,049 34.8* 10,414 11,157 7.1 access to basic services, and education, in the SNNP region between 2011 and 2016. Using data from a National excl. variety of sources, mainly the twinned household living standards surveys (HCES and WMS)69 and the De- 11,086 12,578 13.5* 13,981 18,764 34.2* 10,439 10,975 5.1* SNNP mographic and Health Survey (DHS), the chapter begins by documenting changes in monetary poverty and Note: * represents changes that are statistically significant. living standards in the region. It then moves to an analysis of non-monetary welfare before turning to chal- Source: HCES, 2011; 2016; World Bank staff calculations. lenges of inequality of opportunity. The final section discusses changes in the sectoral allocations of social spending in the region between 2011 and 2016. The chapter aims to provide policy makers with information The increase in consumption in SNNP led to a de- to 22 percent between 2011 and 2016. The decrease in and analysis that will be useful in improving the effectiveness of poverty reduction programs in the region. crease in poverty, more so than the rest of the coun- poverty was stronger in SNNP than in other regions com- Some of the main findings include: try.70 The proportion of people below the national poverty bined, particularly in rural areas; rural poverty in the rest of line in SNNP decreased by nine percentage points from 30 the country dropped from 30 percent in 2011 to 24 percent ● The poverty rate in SNNP decreased from 30 percent to the case for human capital indicators such as health and percent in 2011 to 21 percent in 2016. In urban areas, pov- in 2016, while rural poverty decreased from 31 percent to 27 21 percent between 2011 and 2016. Poverty reduction education. SNNP’s Human Opportunity Index was in erty decreased by more than 40 percent (falling from 26 to percent. Decrease in urban poverty in the rest of the country was strong both in urban and rural areas. general similar to the rest of the country – implying that 14 percent), while rural poverty decreased from 30 percent was comparable with that of SNNP (Table 2). children in SNNP had similar level of equality of access ● Consumption growth was strong only in urban areas to key basic services as children elsewhere in Ethiopia. where growth in rural areas was not statistically signif- Table 2 POVERTY DECREASED IN SNNP FASTER THAN IN THE REST OF THE icant. There were, however, consumption growth at the ● Average per capita public expenditure in SNNP is low- COUNTRY bottom both in rural and urban areas. The gap between er than in the rest of the country, and SNNP spends rural and urban areas further widened and that led to a proportionally higher on administration. Per capita ex- Percentage of people below the national poverty line modest increase in inequality. penditures on “pro-poor” sectors which includes health, TOTAL URBAN RURAL education, agriculture and rural development, and water ● Non-monetary welfare including ownership of consumer % pt. % pt. % pt. resources are also lower in SNNP. durables and access to drinking water improved, as was 2011 2016 change 2011 2016 change 2011 2016 change SNNP 29.6% 20.7% -8.9* 25.8% 14.4% -11.3* 30.0% 21.9% -8.1* National excl. 29.6% 24.1% -5.4* 25.6% 14.6% -11.0* 30.5% 26.6% -3.8* SNNP Note: * represents changes that are statistically significant. Source: HCES, 2011; 2016; World Bank staff calculations. 69 Household Consumption Expenditure Survey (HCES) and Welfare Monitoring Survey (WMS) are the two official household surveys collected by CSA every five year and used to monitor welfare. HCES covers monetary welfare and WMS deals with non-monetary 70 The relatively slower poverty reduction in SNNP could partly be due the higher rural population share (80% for SNNP versus 75% for welfare like health and education. Both surveys are collected from the same sample of households. the rest of the country in 2016). Poverty reduction was low in rural areas in general. 130 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 131 Consumption growth in urban SNNP was positive percentile. In rural areas, on the other hand, there was posi- The increasing gap between urban and rural areas part of inequality resulting from differences in average con- and strong for the entire consumption distribution, tive growth for the bottom 15th percentile, after which it be- helped increase inequality between 2011 and 2016. Av- sumption between urban and rural areas – increased from while it was weak and varied in rural areas. In urban comes zero before picking up again after the 20th percentile. erage consumption was already significantly higher in urban 8.7 percent in 2011 to 22.6 percent in 2016. On the other areas, consumption growth was positive and strong across It then started to decline and reached zero between the 60th than rural areas in 2011. The gap increased in 2016, with hand, the “within” share of inequality – the part resulting from the consumption distribution, increases as one moves up and 70th and again became positive at the top of the distri- urban consumption showing strong growth while rural con- changes in distribution within urban and rural areas – dimin- the distribution, reaching at least six percent after the 50th bution (Figure 1). sumption grew at about one-fifth the rate of urban areas (see ished from 77 percent to 64 percent (Figure 2). Table 1). As a result, the “between” share of inequality – the Figure 1 GICS FOR SNNP URBAN AND RURAL Figure 2 INEQUALITY INCREASED DUE TO THE INCREASING GAP BETWEEN Urban Rural URBAN AND RURAL AREAS Decomposition of the Gini coefficient into a between rural-urban component and a within-component Overlap Within Between Relative contribution to Gini (%) 100% 90% 14.6 13.9 80% 76.7 63.5 70% 60% 50% 40% 30% 20% 10% 22.6 8.7 0% 2011 2016 Source: HCES, 2011; 2016; World Bank staff calculations Source: HCES, 2011; 2016; World Bank staff calculations. Poverty reduction in SNNP can be entirely attributed The growth effect on poverty in SNNP is negative (pover- Strong growth in urban areas combined with the weak percentile compared to the 10th percentile increased from to increases in consumption, with the small increase ty-reducing). The redistribution effect is zero for SNNP, in or no growth in rural areas led to modest increase 4.1 in 2011 to 4.7 in 2016, reflecting stronger consumption in inequality not hindering poverty reduction. Figure 3 contrast to the rest of the country where it is positive (pov- in inequality. The Gini index increased from 30 to 33 be- growth at the top (Table 3). decomposes the change in poverty into growth and redis- erty-increasing). If there were no changes in the distribution tween 2011 and 2016. The ratio of consumption at the 90th tribution components using the Datt-Ravallion decomposi- between 2011 and 2016, poverty in the rest of the country tion technique. It indicates which part of the poverty change would have decreased by 8.7 percent. Given the negligible can be attributed to: (i) the change in average consumption changes in inequality in SNNP, overall poverty reduction (8.8 Table 3 INEQUALITY INCREASED BETWEEN 2011 AND 2016 (growth effect), and (ii) the change in the redistribution of con- percent) was the same as the reduction that could be at- sumption; that is, changes in inequality (redistribution effect). tributed to consumption growth. 2011 2016 Gini index p90/p10 ratio Gini index p90/p10 ratio SNNP 30.3 4.1 32.7 4.7 National 29.8 3.6 32.9 4.3 Source: HCES, 2011; 2016; World Bank staff calculations. 132 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 133 In SNNP, households that primarily depend on casu- the lowest poverty rates. SNNP diverges from the rest of the Figure 3 CONSUMPTION GROWTH DROVE POVERTY REDUCTION al labor and crop production have the highest poverty country in that poverty rates for households relying on crop rates. Those who depend on service self-employment, crop and livestock as their main income source are significantly Datt-Ravallion decomposition of poverty change between 2011 and 2016 and livestock, salaried employment, and remittances have lower than for other regions combined (Figure 5). SNNP National excl. SNNP Contribution to change in poverty 4.0 2.0 3.3 Figure 5 CASUAL LABORERS AND CROP PRODUCERS HAVE THE HIGHEST 0.0 0.0 POVERTY RATES -2.0 Poverty rates by household’s main income source, 2016 -4.0 -5.4 SNNP National excl. SNNP -6.0 -8.0 -8.8 -8.8 -8.7 35% -10.0 30% Poverty reduction Consumption growth Increase in inequality Poverty rate 25% 20% Source: HCES, 2011; 2016; World Bank staff calculations. 15% 10% 5% Average consumption per quintile in SNNP is almost the at the top of the distribution. Although SNNP closely mir- 0% same as in the rest of Ethiopia. Average annual consump- rors consumption patterns for the rest of the country, the in- or n g ns il s t es ck en ta ce tio rin lab vic io to tion per adult in 2016 was 4,509 Birr for the poorest quintile crease from the fourth to the fifth quintile in SNNP is smaller re m an uc tu at es er oy d al ac on itt od an liv -s su pl m uf and 13,312 Birr for the fourth quintile. Average consumption than the rest of the country; consumption in the top SNNP /d pr em d Ca Re ale an lf an ns Se p -m les ro io ied shows a sharp increase from the fourth to the fifth (richest) quintile was about 1,500 Birr per year less than in the rest of p ns -c ro ho lar lf -c Pe Se -w lf Sa quintile, with the richest 20 percent consuming 24,808 Birr Ethiopia (Figure 4). Se lf Se lf Se per year on average, reflecting strong consumption growth Main source of household income Source: HCES 2016; World Bank staff calculations. Figure 4 HOUSEHOLD CONSUMPTION REMAINS FAIRLY LOW EXCEPT FOR THE TOP QUINTILE Educational attainment of household heads strongly headed by someone with post-secondary education. SNNP Average household expenditures per capita per year, 2016 negatively correlates with poverty rates. With a poverty diverges from the rest of the country in that poverty for those SNNP National excl. SNNP rate of 25 percent, poverty in SNNP is highest among house- households headed by someone who has completed pri- holds headed by someone who has no education. Increasing mary/incomplete secondary education is not higher than for 30,000 education of the household head leads to decreasing pover- households headed by someone that has not completed Average annual consumption 25,000 ty rates, reaching merely five percent for those households secondary education (Figure 6). 20,000 15,000 10,000 5,000 0 Poorest 2 3 4 Richest Consumption quintiles Source: HCES 2016; World Bank staff calculations. 134 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 135 When decomposing poverty reduction by sector, ag- shift, often used as an indicator of structural transformation Figure 6 EDUCATION IS A MAIN CORRELATE OF POVERTY riculture was a bigger driver of poverty reduction in (proxied by the shift in population from agriculture to manu- SNNP than in the rest of the country. Improvements in facturing or services) explained around 8 percent of poverty Poverty rates by educational attainment of the household head, 2016 the agricultural sector explain around 56 percent of poverty reduction in SNNP over the period. Yet, this effect is less reduction in the rest of Ethiopia; the corresponding figure for important than in other regions. There was a slight fall in SNNP National excl. SNNP SNNP is higher at 74 percent. This is in line with the previous the share of the population in SNNP relying on agriculture 35% figure showing that rural areas drove reduction in poverty in over the period, but smaller than in the rest of the country. In 29% the region between 2011 and 2016. Gains for households 2011, 83 percent of households in SNNP relied primarily on 30% 25% working in the services or manufacturing sectors were min- agriculture as the primary source of income. In 2016 this fell 25% 21% 22% Poverty rate imal, with services in particular not as pronounced as in the to 75 percent. At the same time, the share of households re- 20% rest of the country. The so-called “population shift” effect had lying primarily on services increased by around 4 percentage 14% 15% a small impact on poverty reduction in SNNP. The population points (Figure 8). 9% 10% 10% 5% 5% 5% 3% 0% No education Incomplete Complete primary/ Complete Post-secondary primary Incomplete secondary secondary Figure 8 THE AGRICULTURE SECTOR REMAINS THE LARGEST CONTRIBUTOR TO POVERTY REDUCTION IN SNNP Education level of household head Sectoral decomposition of poverty changes 2011 to 2016 Source: HCES 2016; World Bank staff calculations. Note: The complete primary and incomplete secondary categories have been combined because of small sample sizes. Agriculture Manufacturing Construction Services Other Population shift Interaction SNNP National excl. SNNP 0 Change in poverty headcount Poverty reduction in SNNP was stronger than the rest to rural areas and 1.3 percent to urban areas. The contribu- -1 of Ethiopia, and was largely driven by improvements in tion of rural areas is driven by the high rural population share, -2 rural areas. Figure 7 shows that of the almost 9 percentage and the fact that the poorest rural households in the region -3 -4 point reduction in poverty in SNNP, 7.2 points are attributed experienced growth over the period (as shown in Figure 2). -5 -6 -7 -8 -9 Figure 7 POVERTY REDUCTION IN SNNP LOOKED A LOT LIKE POVERTY -10 REDUCTION IN THE REST OF ETHIOPIA Source: HCES; 2011, 2016. World Bank staff calculations. Rural-urban decomposition of poverty changes 2011 to 2016 Rural Urban Population shift Interaction SNNP National excl. SNNP 0 Change in poverty headcount -1 -2 -3 -4 -5 -6 -7 -8 -9 -10 Source: HCES, 2011; 2016; World Bank staff calculations. 136 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 137 NON-MONETARY WELFARE Figure 9 OWNERSHIP OF CONSUMER DURABLES AND HOUSING CONDITIONS Improvements in non-monetary welfare accompanied faced a food shortage, the average number of months with a IMPROVED BETWEEN 2011 AND 2016 increase in consumption expenditure. In SNNP, the per- food shortage hardly changed. Reduction in food shortage in Household durables, housing conditions and access to energy in 2011 and 2016, % of households centage of households that reported a food shortage in the SNNP was significantly stronger than in other regions com- with asset/characteristic last 12 months before the interview decreased from 35 per- bined (Table 4). cent in 2011 to 13 percent in 2016. However, for those who A. Household Durables 2011 2016 Table 4 FOOD SECURITY IN SNNP IMPROVED BETWEEN 2011 AND 2016 MORE 60 THAN THE REST OF THE COUNTRY 50 Incidence of food shortages and average duration of food shortages – food gap 40 2011 2016 30 Food shortage (%) Food gap (months) Food shortage (%) Food gap (months) 20 SNNP 35.0 3.4 12.6 3.2 10 National excl. SNNP 17.9 3.2 9.8 3.2 0 Mobile Mobile Refrigerator Refrigerator Animal Cart Motorcycle Animal Cart Motorcycle TV TV Source: WMS; 2011, 2016; World Bank staff calculations. Ownership of household durables and housing condi- between 2011 and 2016. The proportion of households with SNNP National excl. SNNP tions also improved in SNNP, although this remains low. an improved roof and improved floor also improved, while Ownership of household durables such as mobile phones, access to electricity did not change. The rest of the country refrigerators, animal carts, and motorcycles increased experienced similar improvements (Figure 9). B. Housing/energy 2011 2016 70 60 50 40 30 20 10 0 Improved wall Floor: tiles/cement Access to electricity Electricity for cooking Improved wall Floor: tiles/cement Access to electricity Electricity for cooking Improved roof Improved roof Photo: World Bank/Sylwia Pecio SNNP National excl. SNNP Source: DHS, 2011; 2016; World Bank staff calculations. 138 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 139 Figure 10 ACCESS TO IMPROVED WATER SOURCES INCREASED BETWEEN 2011 Figure 11 CHILDREN’S MORTALITY DECREASED, AND THEIR HEALTH IMPROVED AND 2016 BETWEEN 2011 AND 2016 2011 2016 100 94.5 Access to water in 2011 and 2016, % of households with access/characteristics 2011 in 2011 Selected health and education variables and 2016 2016 86.3 80 100 94.5 70.6 63.6 86.3 66.0 SNNP National excl. SNNP 58.0rates (per 1000) A. Child and infant mortality 60 80 48.9 45.0 70.6 63.6 2011 2016 66.0 Piped water - compound Piped -outside Protected well/tube well Protected spring 33.0 58.0 40 60 100 94.5 30.0 26.8 48.9 Other improved Unimproved sources 45.0 18.0 86.3 20 40 80 33.0 26.8 70.6 30.0 100% 100% 63.6 66.0 53.8 41.29 44.51 33.61 0 58.0 18.0 20 60 48.9 80% 80% Infant mortality 45.0 Child mortality Under five Infant mortality Child mortality Under five 0 40 rate 33.0rate mortality rate rate rate mortality rate 26.8 30.0 Infant mortality Child mortality SNNP Under five Infant mortality Child mortality National excl. Under five 60% 60% 10.48 18.0 SNNP 16.72 20 rate rate mortality rate rate rate mortality rate 9.15 18.11 40% 8.52 40% 10.22 SNNP National excl. SNNP 12.68 0 9.81 23.51 20.51 Infant mortality Child mortality Under five Infant mortality Child mortality Under five 20% 22.31 23.3 20% rate rate mortality rate rate rate mortality rate 12.36 16.38 SNNP National excl. SNNP 0% 5.41 6.0 0% Source: DHS, 2011; 2016; World Bank staff calculations 2011 2016 2011 2016 2011 2016 Source: DHS, 2011; 2016; World Bank Staff Calculations B. Immunization and health facility delivery (percent) 50 46.9 44.12011 44.5 2016 40 38.6 38.3 50 46.9 36.3 44.1 44.5 30 40 25.5 38.6 26.4 38.3 Access to improved drinking water sources71 also im- 64 per thousand to 45, from 33 to 27, and from 95 to 71 24.1 24.4 36.3 2011 2016 proved in SNNP. The proportion of households with access between 2011 and 2016. Delivery at health facility increased 20 50 30 46.9 26.4 25.5 24.1 24.4 44.5 44.1 10.9 to improved water sources in SNNP increased from 46 per- from 6 to 26 percent, while the percentage of fully immunized 38.6 38.3 10 40 20 6.2 36.3 cent to 59 percent between 2011 and 2016, mirroring im- children increased from 24 to 47 percent. The stunting rate 10.9 provements in the rest of Ethiopia (Figure 10). decreased from 44 percent to 39 percent. Net primary en- 0 30 10 6.2 25.5 26.4 24.1 24.4 Health facility Fully Stunted Health facility Fully Stunted rollment, primary school completion, and gross secondary delivery immunized children delivery immunized children 20 0 Health and education indicators also improved, al- children children enrollment rates also improved. The rest of the country expe- Health facility Fully Stunted 10.9 facility Health Fully Stunted though challenges remain. The infant, child, and un- 10 delivery 6.2 immunized children delivery immunized children rienced similar improvements (Figure 11). SNNP National excl. SNNP der-five mortality rates in SNNP decreased respectively from children children 0 Health facility SNNP Fully Stunted National Health facility excl. SNNP Stunted Fully Source: DHS, 2011; delivery immunized 2016; World Bank children staff calculations delivery immunized children children children SNNP C. Education (percent) National excl. SNNP 2011 2016 80 72.3 69.7 2011 2016 63.4 80 58.8 60 69.7 72.3 63.4 58.8 60 40 2011 35.9 2016 32.4 32.7 29.4 29.5 80 72.3 22.0 69.7 20.4 19.4 35.9 40 20 32.4 63.4 32.7 29.4 29.5 58.8 60 20.4 22.0 19.4 20 0 40 Net primary Primary school Gross 35.9 Net primary Primary school Gross 32.4 32.7 enrolment completion secondary enrolment completion 29.4 secondary 29.5 0 Net primary (15-24) 20.4 Primary school school Gross 19.4 Net primary (15-24) Primary school school 22.0 Gross 20 enrolment completion enrolment secondary enrolment completion enrolment secondary (15-24) SNNP school (15-24) National excl. SNNP school 71  Improved drinking water sources include piped water, protected well, protected spring, rainwater and bottled water. enrolment 0 2011; 2016; World Bank staff calculations. enrolment Source: WMS, Net primary Primary SNNPschool Gross Primary Net primary National school excl. SNNP Gross enrolment completion secondary enrolment completion secondary 140 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS (15-24) school (15-24) school 141 enrolment enrolment INEQUALITY OF OPPORTUNITY Figure 12 SNNP LAGS IN SOME INDICATORS BUT DOES BETTER WITH ACCESS TO To assess inequality of opportunity72 in SNNP and the rest of 18) were enrolled in secondary school in 2016. This indicates HEALTHCARE FACILITIES Ethiopia, we focus on six key outcomes: (i) primary school that educational opportunity is quite different for the older Coverage, HOI, and D-index for different outcomes, 2016 enrollment, (ii) primary school completion, (iii) secondary cohort of children (15-18) than the younger cohort (7-14). school enrollment, (iv) access to electricity; (v) access to The HOI scores for SNNP are, like coverage scores, SNNP National excl. SNNP an improved water source; and (vi) access to a health post similar to the rest of Ethiopia. A higher HOI score is as- 90 within 5 km. Circumstance variables73 of the household in- sociated with more equitable access to a key service or out- Coverage HOI D-index clude rural/urban location, gender, religion, agro-ecological 80 come. For a given coverage rate, the lower the HOI score, 70 zone, consumption quintile of the child’s household, and Percent/score the more unequal the access to that key outcome is across 60 the household’s score on the rural accessibility index (RAI).74 different groups. In SNNP, the coverage rate of access to 50 Figure 12 presents the coverage, Human Opportunity Index electricity is 20 percent. However, this is heavily skewed to- 40 (HOI), and Dissimilarity index (D-index)across these key out- wards the urban population in the region. The very low HOI 30 come variables in 2016. score of 9.7 for electricity reflects strong differential access 20 In terms of coverage, SNNP lags the rest of Ethiopia on 10 across the population in the region. 0 all key outcomes except for secondary school enroll- The dissimilarity index, or D-index, shows how much alt ter alt ter alt ter st st st ec ry ool ec ry ool ec ry ool l l l m ve ess m ve ess m ve ess nd sc l co ry ool co ry ool ity oo ty oo ty oo oo ment and access to healthcare facilities. Primary school po po po he wa he a he a a h a h a h ric sch ici ch ici ch im ch im ch im ch 5k pro cc 5k pro cc 5k pro cc of d w of d w nd sc nd sc inequality in access there is between groups – a high- h h h s s pr ry s pr ry s pr ry s a a a of d completion and secondary school enrollment are already low co ry a a a a a a er index means less equal access across groups. The En let rim En let rim En let rim r r in Ethiopia, and the same is true in SNNP. This means that t t t m m m m dp p p I I I D-index for completed primary school and secondary school ro ed ro ed ro ed se se se m ed m ed El El El Co rolle the HOI scores for these key outcomes are also low (see Co roll Co roll d d d in in in lle lle lle p p p enrollment in SNNP is significantly lower than the rest of Ethi- ith ith ith En En En the Coverage and HOI panels in Figure 14). The D-index for W W W opia. This indicates that circumstances such as location, gen- these variables – measuring how much these outcomes dif- der, and household income matter less in SNNP than in the Source: HCES, WMS 2011; 2016; World Bank staff calculations. fer according to circumstances – is correspondingly high, al- rest of the country. The D-index scores for secondary school though similar to the rest of the country. This implies strong enrollment and electricity access in SNNP were also lower differences between households in which children complete than in rest of the country, implying greater equality in the dis- primary school, enroll in secondary school, and have access to electricity, and those that do not. Decomposing this differ- tribution of access to these key outcomes in the region. PUBLIC EXPENDITURE ence (not shown) reveals that this is largely explained by ru- Unequal coverage of infrastructure access remains a In the previous sections, we discussed trends in monetary expenditure on “pro-poor” activities improved, although it ral-urban differences in access rates, both within SNNP and challenge in SNNP and the rest of the country. The D-in- and non-monetary welfare between 2011 and 2016, and in- remained lower compared to other regions combined. Per in the rest of Ethiopia. dex was highest for access to electricity, exemplifying the equality of opportunities focusing on 2016, both in SNNP capita expenditures on health were significantly lower than rural-urban divide in access to these services in the region, and the rest of the country. In this section, we discuss trends, the rest of the country. These results help explain some re- The primary school enrollment rate in SNNP is about 2 and broadly throughout the country. As noted, decomposi- volume, and allocation of public expenditures to identify vari- sults presented in earlier sections, which showed that SNNP percentage points lower than the rest of Ethiopia. In the tion of this index (not shown) reveals that access depends ation across regions and links with welfare. We use MoF lags in most non-monetary indicators. For example, the poor region, although 70 percent of age-eligible children were en- far more on whether a household is in an urban or rural area data. We express all expenditure values in 2016 prices using improvement of the region on health indicators could re- rolled in primary school in 2016, only 28 percent of age-eligi- than on the level of household wealth. For example, in urban the non-food CPI deflator. We first present total per capita flect the smaller amount of per capita public expenditure on ble children had completed primary school. These numbers areas even the poorest households tend to have access to public expenditure and then discuss expenditure allocation health, particularly on child mortality. are both slightly lower than the rest of Ethiopia, in which close electricity, while in rural areas even the richest households do by sector. Finally, we present expenditure on selected “pro- to three-quarters of primary school aged children are enrolled not (Figure 12). Per capita public expenditure in SNNP increased be- and almost one-third completed primary education. By com- poor” sectors, which include agriculture and rural develop- tween 2011 and 2016, although it is on average slightly parison, one in five age-eligible children in SNNP (age 15 to ment, water resources, health, and education. lower than the rest of the country. Public spending in the Average per capita public expenditure in SNNP is SNNP region increased from Birr 980 per person in 2011 slightly lower than the rest of the country, and the re- to Birr 1,888 per person in 2016 (in constant 2016 Birr). On gion spends a significantly higher proportion on admin- average, per capita public expenditure increased at an aver- istration. Although the share of spending on administration age rate of 14 percent annually in SNNP as well as in other remained high in SNNP, between 2011 and 2016 spending regions combined (Figure 13). shifted in favor of economic sectors. In SNNP, per capita A full description of the methodology and variables used in the construction of this index in Ethiopia can be found in the technical 72  appendix. 73  These are generally thought of as variables over which a person does not have control, such as gender or region of birth. 74  Accessibility was also tested using a market accessibility variable, and the results were broadly consistent. 142 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 143 Birr 303, consistent with improved education indicators. The However, the total share of spending in “pro-poor” ac- Figure 13 PER CAPITA PUBLIC EXPENDITURE INCREASED OVER TIME rest of the country also witnessed an increase in per capi- tivities decreased between 2011 and 2016. The share of ta spending, and the rate of change is faster compared to spending on education and agriculture and rural development Per capita real public expenditure, 2011 to 2016 SNNP. Such growth in the rest of the country has resulted in in SNNP decreased from 21 percent and eight percent in a higher pro-poor spending per person in 2016 compared to 2011 to 18 percent and seven percent in 2016, respectively. SNNP National excl. SNNP SNNP (Figure 15A). The share of the health sector in total spending remained at a 2,400 stable seven percent between 2011 and 2016 (Figure 15B). 2,000 In 2016 Birr 1,600 1,200 800 Figure 15 EXPENDITURE SHARE ON EDUCATION DECREASED IN SNNP BETWEEN 400 2011 AND 2016 0 Real per capita expenditure and share of total expenditure (%) for selected sectors, 2011 and 2016 2011 2012 2013 2014 2015 2016 Source: MoF; World Bank staff calculations. A. Real per capita expenditure in 2016 price 2011 2016 350 A spending shift away from administration and towards related spending decreased both in SNNP and the rest of 300 economic sectors accompanied public expenditure the country, but it is still large, particularly in SNNP where it 250 In 2016 Birr growth in SNNP.75 Between 2011 and 2016, economic sec- accounted for 49 percent in 2016 (Figure 14). 200 tor spending as a share of total expenditure increased from 150 Per capita spending76 on selected “pro-poor” activ- 15 to 20 percent, while it increased from 25 to 30 percent for ities increased between 2011 and 2016 in SNNP. The 100 the rest of the country. The share of spending in the social increase was particularly strong for education, where per 50 sector changed only slightly in SNNP, while it decreased for capita spending on education increased from Birr 234 to 0 the rest of the country (from 33 to 28 percent). Administration Agriculture and Rural Development Water Resources Education Health Agriculture and Rural Development Water Resources Education Health Figure 14 SPENDING SHARE OF ECONOMIC SECTOR INCREASED OVER TIME Expenditure share by sectors (%), 2011 to 2016 SNNP National excl. SNNP SNNP National excl. SNNP B. Expenditure share (%) Administration Economic Social Other Agriculture & Rural Dev. Water Resources Education Health 2016 2016 2.8 16.3 50 2016 45 1.9 25.9 40 8.3 2016 29.0 35 7.0 7.9 28.2 2016 8.4 7.2 22.4 54.6 32.5 33.7 30 48.8 21.0 17.2 25 18.1 14.5 20 2016 2016 25.4 2016 20.2 27.7 30.1 15 10 2.3 6.3 6.2 2.7 5 7.8 6.8 6.6 6.8 0 Note: The inner circles present the 2011 values. 2011 2016 2011 2016 Source: MoF; World Bank staff calculations. SNNP National excl. SNNP Note: The sum does not add up to 100% because it includes only shares of selected “pro-poor” sectors. Source: MoF; World Bank staff calculations. 75 Economic sector includes agriculture and rural development, water resources, trade and industry, mines and energy, transportation and communication, construction. Per capita spending refers to total spending in a particular sector divided by the population of the region. For example, per capita 76  spending on education refers to total spending on education divided by the total population of the region and not spending per student. 144 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 145 POVERTY AND INEQUALITY Food and nonfood consumption in the Gambella region where consumption increased by 44 percent compared to increased significantly both in urban and rural areas 34 percent in urban areas in other regions combined. Rural between 2011 and 2016. Mean consumption per adult in consumption in Gambella increased by 26 percent, which is Gambella increased by about one-third, compared with a 14 significantly higher than rural consumption growth of six per- percent increase for all other regions combined. The increase cent for the rest of the country (Table 1). Gambella in consumption in urban areas in the region was stronger, Table 1 CONSUMPTION INCREASED IN GAMBELLA MORE THAN THE REST OF THE COUNTRY BETWEEN 2011 AND 2016 Mean annual household consumption per adult equivalent in December 2015 prices, 2011 and 2016 INTRODUCTION TOTAL URBAN RURAL This chapter focuses on the evolution of socio-economic and health indicators such as poverty, inequality, 2011 2016 % change 2011 2016 % change 2011 2016 % change access to basic services, and education, in the Gambella region between 2011 and 2016. Using data from Gambella 10,334 13,855 34.1* 12,477 17,945 43.8* 9,325 11,745 26.0* a variety of sources, mainly the twinned household living standards surveys (HCES and WMS)77 and the National excl. Demographic and Health Survey (DHS), the chapter begins by documenting changes in monetary poverty 11,011 12,494 13.5* 13,912 18,655 34.1* 10,437 11,012 5.5* Gambella and living standards in the region. It then moves to an analysis of non-monetary welfare before turning to Note: * represents changes that are statistically significant. challenges of inequality of opportunity. The final section discusses changes in the sectoral allocations of Source: HCES, 2011; 2016; World Bank staff calculations. social spending in the region from 2011 to 2016. The chapter aims to provide policy makers with information and analysis that will be useful in improving the effectiveness of poverty reduction programs in the region. The increase in consumption in Gambella led to a de- from 31 to 17 percent), while the change in rural poverty was Some of the main findings include: crease in poverty, but not in rural areas. The proportion of not statistically significant. The decrease in overall poverty in people below the national poverty line in Gambella decreased the region was faster in Gambella compared to other regions ● The poverty rate in Gambella decreased from 32 per- as was the case for human capital indicators such as by nine percentage points from 32 percent in 2011 to 23 in combined (Table 2). cent to 23 percent between 2011 and 2016. However, health and education. Gambella in general has better 2016. In urban areas, poverty decreased by 46 percent (falling poverty reduction in rural Gambella was not statistically human capital indicators than the rest of the country. significant. This is also reflected in a higher score on the Human Opportunity Index than the national average – implying ● The pattern of consumption growth in Gambella was Table 2 POVERTY DECREASED IN GAMBELLA FASTER THAN THE REST OF THE higher equality of access to key basic services for chil- inequality-increasing, with urban consumption growing COUNTRY dren in Gambella compared to the rest of Ethiopia. relatively faster and rural consumption growth concen- Percentage of people below the national poverty line trated at the top. There was no significant consumption ● Average per capita public expenditure in Gambella is growth at the bottom of the distribution in rural Gambel- higher than in the rest of the country, but Gambella also TOTAL URBAN RURAL la, and by 2016 inequality in the region was higher than spends proportionally higher on administration. Per cap- % pt. % pt. % pt. in the rest of Ethiopia. ita expenditures on health and education are higher in 2011 2016 change 2011 2016 change 2011 2016 change Gambella. Gambella 32.0% 23.0% -9.0* 30.9% 16.6% -14.3* 32.5% 26.4% -6.2 ● Non-monetary indicators including ownership of con- National excl. sumer durables and access to drinking water improved, 29.6% 23.4% -6.1* 25.6% 14.6% -11.1* 30.4% 25.6% -4.8* Gambella Note: * represents changes that are statistically significant. Source: HCES, 2011; 2016; World Bank staff calculations. 77 Household Consumption Expenditure Survey (HCES) and Welfare Monitoring Survey (WMS) are the two official household surveys collected by CSA every five year and used to monitor welfare. HCES covers monetary welfare and WMS deals with non-monetary welfare like health and education. Both surveys are collected from the same sample of households. 146 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 147 Consumption growth in urban Gambella was positive pattern from the national level. In rural areas, there was zero The rise in inequality between 2011 and 2016 is mainly inequality – the part of inequality resulting from differences for the entire consumption distribution, while there growth for the bottom one-third of the distribution, but con- due to the increasing gap between urban and rural ar- in average consumption between urban and rural areas – was no growth at the bottom of the distribution in rural sumption growth becomes positive thereafter, and increases eas. Average consumption was already significantly higher increased from 24 percent in 2011 to 29 percent in 2016. areas. In urban areas, consumption growth was strongest moving up the distribution, reaching eight percent at the very in urban than in rural areas in 2011. The gap increased in Rising inequality within both urban and rural areas was also at the bottom of the distribution, which represents a different top (Figure 1). 2016, with urban consumption showing strong growth and evident, as the “within” share of inequality – the part resulting rural consumption also growing, but not nearly as quickly as from changes in distribution within urban and rural areas – in urban areas (Table 1). As a result, the “between” share of rose slightly from 47 percent to 48 percent (Figure 2). Figure 1 GICS FOR GAMBELLA: URBAN AND RURAL Urban Rural Figure 2 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 Overlap Within Between Relative contribution to Gini (%) 100% 90% 29.3 23.2 80% 70% 47.7 60% 46.8 50% 40% 30% 20% 29.1 23.9 10% 0% 2011 2016 Source: HCES, 2011; 2016; World bank staff calculations. Source: HCES, 2011; 2016; World Bank staff calculations Strong growth across the distribution in urban areas, of consumption at the 90th percentile compared to the 10th Given the increase in inequality between 2011 and 2016, redistribution effect is positive for Gambella, but also for the and at the top in rural areas, combined with lack of percentile increased from 3.6 in 2011 to 4.8 in 2016, re- poverty reduction in Gambella can be entirely attribut- rest of the country. If there were no changes in the distribu- growth at the bottom of the distribution in rural areas flecting strong consumption growth at the top, and lack of ed to increases in consumption. Figure 3 decomposes tion between 2011 and 2016, poverty would have decreased to significantly increase inequality. The Gini index in- growth at the bottom (Table 3). the change in poverty into growth and redistribution compo- by 18.8 percent in Gambella and 8.7 percent in the rest of the creased from 29 to 36 between 2011 and 2016. The ratio nents using the Datt-Ravallion decomposition technique. It country, instead of the 9.0 and 6.0 percent we actually ob- indicates which part of the poverty change can be attributed serve. In other words, if the consumption distribution in 2011 to: (i) the change in average consumption (growth effect), and were maintained in 2016, poverty reduction could have been INEQUALITY INCREASED BETWEEN 2011 AND 2016 (ii) the change in the redistribution of consumption; that is, far stronger in Gambella, and it would have outperformed the Table 3 changes in inequality (redistribution effect). The growth effect rest of Ethiopia by an even greater margin. on poverty in Gambella is negative (poverty-reducing). The 2011 2016 Gini index p90/p10 ratio Gini index p90/p10 ratio Gambella 28.9 3.6 36.0 4.8 National 29.8 3.6 32.9 4.3 Source: HCES, 2011; 2016; World Bank staff calculations. 148 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 149 In Gambella, households that primarily depend on services have the lowest poverty rates. The results for the Figure 3 CONSUMPTION GROWTH DROVE POVERTY REDUCTION crop and livestock production have the highest pov- rest of the country are similar (Figure 5). erty rates. Those who depend on salaried employment and Datt-Ravallion decomposition of poverty change between 2011 and 2016 Gambella National excl. Gambella Contribution to change in poverty 15.0 Figure 5 MIXED FARMERS AND CROP PRODUCERS HAVE THE HIGHEST POVERTY 10.0 RATES IN GAMBELLA 5.0 9.8 2.7 Poverty rates by household’s main income source, 2016 0.0 -5.0 -6.0 -9.0 -8.7 -10.0 Gambella National excl. Gambella -15.0 40% -18.8 -20.0 35% -25.0 30% Poverty rate Poverty reduction Consumption growth Increase in inequality 25% 20% Source: HCES, 2011; 2016; World Bank staff calculations. 15% 10% 5% 0% Average consumption in Gambella remains low for the of the distribution. Gambella closely mirrors consumption Self - Self - Self - Salaried employment first four quintiles. Average annual consumption per adult patterns for the rest the country, although the increase from crop production crop and livestock wholesale and retail in 2016 was 5,163 Birr for the poorest quintile and 15,101 the fourth to the fifth quintile in Gambella is larger than for Main source of household income Birr for the fourth quintile. Average consumption shows a the rest of the country; average consumption in the richest sharp increase from the fourth to the fifth (richest) quintile, quintile in Gambella is around 3,500 Birr more per year than with the richest 20 percent consuming 29,423 Birr per year in the rest of Ethiopia (Figure 4). Source: HCES 2016; World Bank staff calculations. on average, reflecting strong consumption growth at the top Educational attainment of household heads strongly percent for households headed by someone who has com- Figure 4 HOUSEHOLD CONSUMPTION REMAINS FAIRLY LOW EXCEPT FOR THE negatively correlates with poverty rates. With a poverty pleted secondary education or has some post-secondary TOP QUINTILE rate of 33 percent, poverty in Gambella is highest among education, respectively. Gambella mirrors the trend for the households headed by someone with no education. In- rest of the country, although poverty is generally higher in Average household expenditures per capita per year, 2016 creasing education of the household head leads to decreas- Gambella (Figure 6). ing poverty rates, with poverty rates of 16 percent and 12 Gambella National excl. Gambella 35,000 Average annual consumption 30,000 25,000 20,000 15,000 10,000 5,000 0 Poorest 2 3 4 Richest Consumption quintiles Source: HCES 2016; World Bank staff calculations. 150 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 151 Figure 6 EDUCATION IS A MAIN CORRELATE OF POVERTY Figure 7 POVERTY REDUCTION IN GAMBELLA LOOKED A LOT LIKE POVERTY REDUCTION IN THE REST OF ETHIOPIA Poverty rates by educational attainment of the household head, 2016 Rural-urban decomposition of poverty changes 2011 to 2016 Gambella National excl. Gambella Rural Urban Population shift Interaction 33% 35% Gambella National excl. Gambella 28% 0 Change in poverty headcount 30% -1 25% 22% 22% 21% Poverty rate -2 20% -3 16% -4 15% 13% 12% -5 10% -6 6% -7 5% 3% -8 0% -9 No education Incomplete Complete primary/ Complete Post-secondary -10 primary Incomplete secondary secondary Source: HCES, 2011; 2016; World Bank staff calculations. Education level of household head Source: HCES 2016; World Bank staff calculations. Note: The completed primary and incomplete secondary categories have been combined because of small sample sizes. Figure 8 THE AGRICULTURE SECTOR REMAINS THE LARGEST CONTRIBUTOR TO POVERTY REDUCTION IN GAMBELLA Sectoral decomposition of poverty changes 2011 to 2016 Poverty decreased in Gambella similarly to the rest of 4.8 points are attributed to rural areas and 4.0 percent to Ethiopia, with poverty reducing more quickly in the re- urban areas. The contribution of rural areas would have been Agriculture Manufacturing Construction Services gion between 2011 and 2016. Figure 7 shows that of the even stronger had there not been negative growth for the Other Population shift Interaction almost 9 percentage point reduction in poverty in Gambella, very poorest households in Gambella (as shown in Figure 2). Gambella National excl. Gambella Change in poverty headcount 0 -1 -2 -3 -4 -5 -6 -7 -8 -9 -10 Source: HCES; 2011, 2016. World Bank staff calculations. Photo: World Bank/Gelila Woodeneh 152 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 153 NON-MONETARY WELFARE Figure 9 OWNERSHIP OF CONSUMER DURABLES AND HOUSING CONDITIONS Improvements in non-monetary welfare accompanied For those who faced a food shortage, the average number IMPROVED BETWEEN 2011 AND 2016 the increase in consumption expenditure. In Gambella, of months with a food shortage decreased from 2.6 to 1.2. Household durables, housing conditions and access to energy in 2011 and 2016, % of households the percentage of people who live in households that report- For the rest of the country, the prevalence of food shortage with asset/characteristic ed a food shortage in the last 12 months before the interview decreased from 22 percent to 10 percent, while the average decreased from 32 percent in 2011 to four percent in 2016. months of food shortage did not change (Table 4). A. Household Durables 2011 2016 Table 4 FOOD SECURITY IMPROVED BETWEEN 2011 AND 2016 70 Incidence of food shortages and average duration of food shortages – food gap 60 50 2011 2016 40 Food shortage (%) Food gap (months) Food shortage (%) Food gap (months) 30 Gambella 31.6 2.6 3.8 1.2 20 National excl. Gambella 21.5 3.3 10.4 3.2 10 Source: WMS; 2011, 2016; World Bank staff calculations. 0 Mobile Mobile Refrigerator Refrigerator Animal Cart Motorcycle Animal Cart Motorcycle TV TV Ownership of household durables, housing conditions, The proportion of households with an improved roof, im- and access to electricity also improved in Gambella, proved floor, access to electricity, and who use electricity for Gambella National excl. Gambella although remaining low. Ownership of household dura- cooking also improved. But there was no significant improve- bles such as televisions, mobile phones, refrigerators, animal ment in the quality of walls. Similar changes were observed in carts, and motorcycles increased between 2011 and 2016. the rest of the country as (Figure 9). B. Housing/energy 2011 2016 70 60 50 40 30 20 10 0 Improved wall Floor: tiles/cement Access to electricity Electricity for cooking Improved wall Floor: tiles/cement Access to electricity Electricity for cooking Improved roof Improved roof Gambella National excl. Gambella Source: DHS, 2011; 2016; World Bank staff calculations. 154 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 155 Figure 10 ACCESS TO IMPROVED WATER SOURCES INCREASED BETWEEN 2011 Figure 11 CHILDREN’S MORTALITY DECREASED, AND THEIR HEALTH IMPROVED AND 2016 BETWEEN 2011 AND 2016 2011 2016 120 117.7 Access to water in 2011 and 2016, % of households with access/characteristics 2011 in 2011 Selected health and education variables and 2016 2016 100 87.9 120 117.7 78.2 Gambella National excl. Gambella 80 67.5 A. Child and infant mortality rates (per 1000) 67 100 59.2 87.9 60 53.9 2011 78.2 2016 48.1 Piped water - compound Piped -outside Protected well/tube well Protected spring 80 67.5 41.3 67 120 38.5 117.7 40 59.2 30.6 Other improved Unimproved sources 60 53.9 48.1 19.9 100 41.3 38.5 87.9 20 100% 100% 40 78.2 30.6 32.24 16.21 80 67.5 19.9 67 46.37 35.23 0 20 59.2 80% 80% 60 Infant mortality Child 53.9mortality Under five Infant mortality Child mortality Under five 19.1 48.1 0 rate 41.3 rate 38.5 mortality rate rate rate mortality rate 40 Infant mortality Child mortality Under five Infant mortality 30.6 Child mortality Under five 60% 12.43 60% Gambella National excl. Gambella 34.09 11.72 rate rate mortality rate rate 19.9 rate mortality rate 23.82 9.02 20 16.96 40% 40% 10.09 Gambella National excl. Gambella 0 23.29 21.09 Infant mortality Child mortality Under five Infant mortality Child mortality Under five 21.13 17.78 20% 20% rate rate mortality rate rate rate mortality rate 9.87 11.97 11.01 14.27 Gambella National excl. Gambella 0% 0% Source: DHS, 2011; 2016; World Bank staff calculations 2011 2016 2011 2016 2011 2016 Source: DHS, 2011; 2016; World Bank Staff Calculations B. Immunization and health facility delivery (percent) 50 45 44.5 41.1 2011 2016 38.5 38.4 40 50 45 44.5 41.1 30 27.5 27.3 26.2 38.5 38.4 40 23.5 24.3 Access to improved drinking water sources also im- 78 from 68 per thousand to 41, from 54 to 39, and from 118 2011 2016 proved in Gambella, and is better than in the rest of the to 78 between 2011 and 2016. Delivery at health facility in- 20 50 30 27.5 15.5 27.3 26.2 45 23.5 24.3 44.5 41.1 9.9 country. The proportion of households with access to im- creased from 28 to 45 percent, while the percentage of fully 38.5 38.4 10 40 20 15.5 proved water sources in Gambella increased from 68 percent immunized children increased from 16 to 41 percent. The 9.9 0 30 27.5 27.3 26.2 to 84 percent between 2011 and 2016, mirroring improve- stunting rate decreased from 27 percent to 24 percent. Net 10 23.5 24.3 Health facility Fully Stunted Health facility Fully Stunted ments in the rest of Ethiopia (Figure 10). primary enrollment, primary school completion, and gross delivery immunized children delivery immunized children 20 0 15.5 secondary enrollment rates also improved. The rest of the Health facility children Fully Stunted Health children 9.9 facility Fully Stunted Health and education indicators also improved, al- 10 delivery immunized children delivery immunized children country experienced similar improvements (Figure 11). Gambella excl. Gambella Nationalchildren though challenges remain. Infant mortality, child, and un- children 0 der-five mortality rates in Gambella decreased respectively Health facility Gambella Fully Stunted Fully GambellaStunted National excl. Health facility Source: DHS, 2011; delivery immunized 2016; World Bank children staff calculations delivery immunized children children children Gambella C. Education (percent) National excl. Gambella 2011 2016 100 86.6 2011 2016 80.3 80 100 71.7 86.6 62.3 80.3 56.2 60 80 71.7 45.4 47.9 2011 2016 62.3 40 35.0 56.2 32.6 100 60 27.5 30.6 86.6 45.4 47.9 80.3 21.3 20 80 40 35.0 71.7 32.6 30.6 27.5 62.3 21.3 0 60 20 56.2 Net primary Primary 45.4 school 47.9 Gross Net primary Primary school Gross 40 enrolment completion 35.0 secondary enrolment completion 32.6 secondary 0 30.6 Net primary (15-24) Primary school school Gross Net primary 27.5 (15-24) Primary school school Gross 21.3 enrolment completion enrolment secondary enrolment completion enrolment secondary 20 (15-24) Gambella school Gambella school (15-24) National excl. 78  Improved drinking water sources include piped water, protected well, protected spring, rainwater and bottled water. enrolment 0 2011; 2016; World Bank staff calculations. enrolment Source: WMS, Net primary Primary school Gambella Gross Net primary Primary school National excl. Gambella Gross enrolment completion secondary enrolment completion secondary 156 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS (15-24) school (15-24) school 157 enrolment enrolment INEQUALITY OF OPPORTUNITY Figure 12 GAMBELLA IS AHEAD OF THE REST OF ETHIOPIA IN ALL SIX KEY To assess inequality of opportunity79 in Gambella and the (aged 15 to 18) were enrolled in secondary school in 2016. OUTCOMES rest of Ethiopia, we focus on six key outcomes: (i) primary This indicates that educational opportunity is quite different Coverage, HOI, and D-index for different outcomes, 2016 school enrollment, (ii) primary school completion, (iii) second- for the older cohort of children (15-18) than the younger co- ary school enrollment, (iv) access to electricity; (v) access to hort (7-14). Gambella National excl. Gambella an improved water source, and (vi) access to a health post The HOI scores for Gambella are, like coverage scores, high- within 5 km. Circumstance variables80 of the household in- 100 er than the rest of Ethiopia. A higher HOI score is associated Coverage HOI D-index clude rural/urban location, gender, religion, agro-ecological 90 with more equitable access to a key service or outcome. For 80 zone, consumption quintile of the child’s household, and Percent/score a given coverage rate, the lower the HOI score, the more 70 the household’s score on the rural accessibility index (RAI).81 60 unequal the access to that key outcome is across different Figure 12 presents the coverage, Human Opportunity Index 50 groups. In Gambella, the coverage rate of access to electric- 40 (HOI), and Dissimilarity index (D-index)across these key out- ity is 28.5 percent. However, this is heavily skewed towards 30 come variables in 2016. the urban population in the region. The very low HOI score of 20 In terms of coverage, Gambella is ahead of the rest of Ethi- 13.7 for electricity reflects strong differential access between 10 0 opia on all key outcomes. Primary school completion and urban and rural populations in the region. alt ter alt ter alt ter st st st ec ry ool ec ry ool ec ry ool l l l m ve ess m ve ess m ve ess nd sc l co ry ool co ry ool ity oo ity oo ty oo oo secondary school enrollment are generally quite low in Ethi- po po po he wa he a he a a h a h a h ric sch ric sch ici ch im ch im ch im ch 5k pro cc 5k pro cc 5k pro cc of d w of d w nd sc nd sc The dissimilarity index, or D-index, shows how much in- h h h s pr ry s pr ry s pr ry s a a a of d opia but they are relatively better in Gambella. This means co ry a a a a a a equality in access there is between groups – a higher index En let rim En let rim En let rim r that the HOI scores for these key outcomes are also higher t t t m m m m dp p p I I I means less equal access across groups. The D-index for ro ed ro ed ro ed se se se m ed m ed El El El Co rolle in the region than the rest of the country (see the Coverage Co ll Co ll d d d o o in in in lle lle lle completed primary school and secondary school enrollment r r p p p ith ith ith En En En and HOI panels in Figure 14). The D-index for these variables W W W in Gambella is much lower than the rest of Ethiopia. This – measuring how much these outcomes differ according to indicates that circumstances such as location, gender and Source: HCES, WMS 2011; 2016; World Bank staff calculations. circumstances – is correspondingly slightly lower, although household income matter less in Gambella than in the rest similar to the rest of the country. This implies that differenc- of the country. The D-index scores for improved water and es between households in which children complete primary electricity access in Gambella were also lower than in rest school, enroll in secondary school, and have access to elec- tricity are less prevalent in Gambella than in the rest of Ethi- of the country, implying greater equality in distribution of ac- PUBLIC EXPENDITURE cess to these key outcomes. opia. Decomposing this difference (not shown) reveals that this is largely explained by rural-urban differences in access In the previous sections, we discussed trends in monetary spending in these sectors increased, except in education. Unequal coverage of infrastructure access remains a chal- rates, both within Gambella and in the rest of Ethiopia. and non-monetary welfare between 2011 and 2016, and in- Per capita spending on education in Gambella decreased, lenge in Gambella and the rest of the country. The D-index equality of opportunities focusing on 2016, both in Gambella but was higher than the average in other regions combined was highest for access to electricity, illustrating the rural-urban The primary school enrollment rate in Gambella is about 15 and the rest of the country. In this section, we discuss trends, in 2016. Per capita spending on health was also higher in divide in access to these services in the region, and broadly percentage points higher than the rest of Ethiopia. In the volume, and allocation of public expenditures to identify vari- Gambella. These results are consistent with results presented throughout the country. As noted, decomposition of this index region, although 87 percent of age-eligible children were ation across regions and links to welfare. We use MoF data. in earlier sections showing that the region improved faster in (not shown) reveals that access depends far more on wheth- enrolled in primary school in 2016, only 36 percent of age-el- We express all expenditure values in 2016 prices using the both monetary and most of non-monetary indicators com- er a household is in an urban or rural area than on the level igible children completed primary school. These numbers are non-food CPI deflator. We first present total per capita public pared to all other regions combined. of household wealth. For example, in urban areas even the both higher than the rest of Ethiopia, in which a little over expenditure and then discuss the expenditure allocation by poorest households tend to have access to electricity, while Per capita public expenditure increased between 2011 71 percent of primary school-aged children are enrolled and sector. Finally, we present expenditure on selected “pro-poor” in rural areas even the richest households generally do not. and 2016, and is higher in the Gambella region than the under one-third completed primary education. By compar- sectors, which include agriculture and rural development, wa- ison, just over one in four age-eligible children in Gambella rest of the country for all years with available data.82 ter resources, health, and education. Public spending in the Gambella region increased from Birr Per capita public expenditure in Gambella has been 3,711 per person in 2011 to Birr 5,127 per person in 2016 significantly larger than the rest of the country over (in constant 2016 Birr). Although per capita spending also in- the years, but Gambella spends a higher proportion creased in other regions combined, the gap with Gambella on administration. While the total spending share on “pro- remained high (Figure 13). poor” sectors decreased between 2011 and 2016, per capita A full description of the methodology and variables used in the construction of this index in Ethiopia can be found in the technical 79  appendix. 80 These are generally thought of as variables over which a person does not have control, such as gender or region of birth. 81  Accessibility was also tested using a market accessibility variable, and the results were broadly consistent. 82  The public expenditure data for Gambella is not complete in the years 2013 and 2014 and hence are omitted from the analysis. 158 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 159 capita spending increases on pro-poor functional activities, share of pro-poor sectors decreased from 42 percent to Figure 13 PER CAPITA PUBLIC EXPENDITURE INCREASED OVER TIME although the rate of change is faster in Gambella, except 35 percent between 2011 and 2016, mainly attributable to for the education sector, which declined from Birr 850 to decrease in the share of spending in the education sector, Per capita real public expenditure, 2011 to 2016 Birr 694 (Figure 16A). In Gambella, such growth has result- whose share dropped from 22 percent to 17 percent. How- ed in higher per capita spending on the activities compared ever, the share of spending on agriculture and rural devel- Gambella National excl. Gambella to other regions combined in 2016 (Figure 15A). opment increased from ten to 11 percent (Figure 15B). 5,400 4,800 In Gambella, the total share of spending on pro-poor 4,200 sectors declined between 2011 and 2016. The total In 2016 Birr 3,600 3,000 2,400 1,800 1,200 600 EXPENDITURE SHARE OF PRO-POOR SECTORS DECREASED IN 0 Figure 15 2011 2012 2013 2014 2015 2016 GAMBELLA BETWEEN 2011 AND 2016 Source: MoF; World Bank staff calculations. Real per capita expenditure and share of total expenditure (%) for selected sectors, 2011 and 2016 A. Real per capita expenditure in 2016 price 2011 2016 A slight spending shift away from administration and (from 32 to 28 percent). Administration-related spending towards economic sectors accompanied public expen- decreased both in Gambella and the rest of the country, but 800 diture growth in Gambella. 83 Between 2011 and 2016, it is still large, particularly in Gambella, where it accounted 700 economic sector spending as share of total expenditure for 53 percent in 2016 (Figure 14). 600 In 2016 Birr increased from 15 to 18 percent, while it increased from 23 500 Per capita spending84 in “pro-poor” activities increased 400 to 29 percent for the rest of the country. However, the share between 2011 and 2016, except for education. In the 300 of spending in the social sector decreased between 2011 Gambella region, the increase was particularly strong for 200 and 2016 from 31 to 29 percent, and the decrease is slightly agriculture and rural development activities; increasing from 100 lower compared to the decrease for the rest of the country Birr 345 to Birr 505. The rest of the country experienced per 0 Agriculture and Rural Development Water Resources Education Health Agriculture and Rural Development Water Resources Education Health Figure 14 SPENDING SHARE OF ECONOMIC SECTOR INCREASED Expenditure share by sectors (%), 2011 to 2016 Gambella National excl. Gambella Gambella National excl. Gambella Administration Economic Social Other B. Expenditure share (%) 2016 2016 Agriculture & Rural Dev. Water Resources Education Health 2016 0.3 14.3 2016 29.1 29.4 45 40 7.0 8.1 31.3 7.2 2016 31.8 37.6 35 7.8 54.2 52.8 23.4 7.4 22.1 30 14.5 25 17.4 2016 2016 23.3 2016 14.7 17.9 27.8 28.6 20 15 10 2.3 1.9 10.7 5.5 5.7 5 9.5 Note: The inner circles present the 2011 values. 6.8 6.7 0 Source: MoF; World Bank staff calculations. 2011 2016 2011 2016 Gambella National excl. Gambella 83 Economic sector includes agriculture and rural development, water resources, trade and industry, mines and energy, transportation Note: The sum does not add up to 100% because it includes only shares of selected “pro-poor” sectors. and communication, construction. Source: MoF; World Bank staff calculations. Per capita spending refers to total spending in a particular sector divided by the population of the region. For example, per capita 84  spending on education refers to total spending on education divided by the total population of the region and not spending per student. 160 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 161 POVERTY AND INEQUALITY Food and nonfood consumption in the Harari region rural consumption in Harari is a departure from the trend in increased significantly faster than the rest of the coun- the rest of the country, where consumption increased only try between 2011 and 2016. Mean consumption per adult by six percent (Table 1). The significantly higher consumption increased by about 59 percent in Harari alone, compared growth in rural Harari could be due to the close proximity of to a 13 percent increase for all other regions combined. Ur- rural residents to urban markets, in contrast to most rural Harari ban and rural consumption in Harari increased by about 57 percent and 53 percent, respectively. The strong increase in Ethiopians who live far from urban centers and infrastructure. Table 1 CONSUMPTION INCREASED IN HARARI FASTER THAN THE REST OF THE COUNTRY BETWEEN 2011 AND 2016 INTRODUCTION Mean annual household consumption per adult equivalent in December 2015 prices, 2011 and 2016 This chapter focuses on the evolution of socio-economic and health indicators such as poverty, inequality, TOTAL URBAN RURAL access to basic services, and education, in the Harari region between 2011 and 2016. Using data from a 2011 2016 % change 2011 2016 % change 2011 2016 % change variety of sources, mainly the twinned household living standards surveys (HCES and WMS)85 and the De- Harari 13,264 21,059 58.8* 15,344 24,028 56.6* 11,397 17,479 53.4* mographic and Health Survey (DHS), the chapter begins by documenting changes in monetary poverty and National excl. 11,002 12,476 13.4* 13,890 18,606 34.0* 10,432 11,004 5.5* living standards in the region. It then moves to an analysis of non-monetary welfare before turning to chal- Harari lenges of inequality of opportunity. The final section discusses changes in the sectoral allocations of social Note: * represents changes that are statistically significant. spending in the region from 2011 to 2016. The chapter aims to provide policy makers with information and Source: HCES, 2011; 2016; World Bank staff calculations. analysis that will be useful in improving the effectiveness of poverty reduction programs in the region. Some of the main findings include: Despite huge consumption growth, poverty in Hara- decreased from 12 percent to six percent between 2011 and ri did not decrease in a statistically significant man- 2016. Poverty did not decrease in rural Harari. In the rest of ● The poverty rate in Harari stood at seven percent in ● Non-monetary indicators including ownership of con- ner. Poverty decreased significantly only in urban areas; the the country, poverty decreased from 30 percent to 24 per- 2016, compared to 11 percent in 2011. This change sumer durables and access to drinking water improved, proportion of urban people below the national poverty line cent during the same period (Table 2). was not, however, statistically significant overall. Poverty as was the case for human capital indicators such as reduction was statistically significant only in urban areas. health and education. Harari in general has better hu- man capital indicators than the rest of the country. This ● The pattern of consumption growth in Harari was in- Table 2 DESPITE THE STRONG CONSUMPTION GROWTH, POVERTY DID NOT is also reflected in a higher score on the Human Oppor- equality-increasing, with growth in both rural and urban DECREASE IN HARARI tunity Index than the national average – implying higher areas being strongly concentrated at the top of the dis- Percentage of people below the national poverty line equality of access to key basic services for children in tribution. There was no significant consumption growth Harari compared to the rest of Ethiopia. at the bottom of the distribution in urban Harari and TOTAL URBAN RURAL growth at the bottom was negative in rural Harari. By ● Average per capita public expenditure in Harari is higher % pt. % pt. % pt. 2016 inequality in the region was higher than in the rest than in the rest of the country. Per capita expenditures 2011 2016 change 2011 2016 change 2011 2016 change of Ethiopia. on education, health and water resources are also high- Harari 11.1% 7.1% -4.0 11.7% 6.0% -5.7* 10.5% 8.5% -2.0 er in Harari. National excl. 29.6% 23.5% -6.2* 25.8% 14.6% -11.1* 30.4% 25.6% -4.8* Harari Note: * The change in poverty in Harari is significant only in urban areas. Source: HCES, 2011; 2016; World Bank staff calculations. Household Consumption Expenditure Survey (HCES) and Welfare Monitoring Survey (WMS) are the two official household sur- 85  veys administered by CSA every five year and used to monitor welfare. HCES covers mon-monetary welfare and WMS deals with non-monetary welfare like health and education. Both surveys are conducted on the same sample of households. 162 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 163 Consumption did not grow for the bottom of the pop- the distribution, reaching more than 12 percent at the top. In The rise in inequality between 2011 and 2016 is main- share of inequality – the part of inequality resulting from differ- ulation, both in urban and rural Harari. In urban areas, rural areas, consumption growth was negative for the bot- ly due to the increasing gap between urban and rural ences in average consumption within urban and rural areas consumption growth was not statistically significant for the tom 10 percent of the population, and become positive after- areas. Average consumption was already significantly high- –increased from about 47 percent in 2011 to 51 percent in bottom 15 percent of the population. But growth was posi- wards, following a similar pattern as urban areas (Figure 1). er in urban than in rural areas in 2011. The gap increased 2016. On the other hand, the “between” share of inequality – tive and strong afterwards, and increases further moving up in 2016, with both urban and rural consumption showing the part resulting from changes in distribution between urban strong growth at the top and lower growth at the bottom and rural areas – diminished from 29 percent to 22 percent of the distributions (see Figure 1). As a result, the “within” (Figure 2). Figure 1 GICS FOR HARARI URBAN AND RURAL Urban Rural Figure 2 INEQUALITY INCREASED DUE TO THE INCREASING GAP WITHIN URBAN AND RURAL AREAS Decomposition of the Gini coefficient into a between rural-urban component and a within-component Overlap Within Between Relative contribution to Gini (%) 100% 90% 23.5 27.7 80% 70% 47.2 60% 50.5 50% 40% 30% 20% 29.3 10% 21.8 0% 2011 2016 Source: HCES, 2011; 2016; World bank staff calculations. Source: HCES, 2011; 2016; World Bank staff calculations Lack of consumption growth at the bottom combined ratio of consumption at the 90th percentile compared to the Because of increased inequality, Harari did not ex- rest of Ethiopia. The redistribution effect is large and positive with relatively stronger growth at the top, for both in 10th percentile increased from 2.8 in 2011 to 4.9 in 2016, re- perience significant poverty reduction despite con- for Harari, again, even more so than the rest of the country, urban and rural areas, increased inequality. The Gini in- flecting very strong consumption growth at the top and lack sumption growth. Figure 3 decomposes the change in where it is also positive (poverty-increasing). If there were no dex increased from 27 to 37 between 2011 and 2016. The of growth at the bottom (Table 3). poverty into growth and redistribution components using the changes in the distribution between 2011 and 2016, pov- Datt-Ravallion decomposition technique. It indicates which erty in the rest of the country would have decreased by 6.1 part of the poverty change can be attributed to: (i) the change percent. In Harari, changes in inequality were large and sig- Table 3 INEQUALITY INCREASED BETWEEN 2011 AND 2016 in average consumption (growth effect), and (ii) the change nificant. Had there been no change in measured inequality, in the redistribution of consumption; that is, changes in in- poverty reduction in Harari would have been 13.7 rather than 2011 2016 equality (redistribution effect). The growth effect on poverty 3.9 percentage points. Gini index p90/p10 ratio Gini index p90/p10 ratio in Harari is negative (poverty-reducing), and is larger than the Harari 26.6 2.8 37.1 4.9 National 29.8 3.6 32.9 4.3 Source: HCES, 2011; 2016; World Bank staff calculations. 164 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 165 While poverty is generally low in Harari, households on salaried employment and pension/donation have the low- Figure 3 INCREASED INEQUALITY LED TO NO SIGNIFICANT POVERTY REDUCTION that primarily depend on wholesale and retail have the est poverty rates (Figure 5). relatively highest poverty rate. Households who depend Datt-Ravallion decomposition of poverty change between 2011 and 2016 Harari National excl. Harari POVERTY IS LOW IN GENERAL IN HARARI Contribution to change in poverty 15.0 Figure 5 10.0 Poverty rates by household’s main income source, 2016 9.8 5.0 0.0 3.1 Harari National excl. Harari -3.9 -5.0 -6.1 30% -9.2 -10.0 -13.7 25% -15.0 Poverty rate 20% Poverty reduction Consumption growth Increase in inequality 15% Source: HCES, 2011; 2016; World Bank staff calculations. 10% 5% 0% Average consumption per quintile in Harari is signifi- percent consuming twice as much as the fourth quintile, on Self - Self - Self - Salaried employment cantly higher than the rest of Ethiopia. Average annual average, reflecting strong consumption growth at the top of crop production crop and livestock wholesale and retail consumption per adult in 2016 was 7,580 Birr for the poorest the distribution. Although the distribution in Harari is always Main source of household income quintile (versus 4,853 for the rest of the country) and 22,921 significantly above that of the rest the country, the jump from Birr for the fourth quintile (versus 13,270 for the rest of the the fourth to the fifth quintile in Harari is particularly large, with Source: HCES 2016; World Bank staff calculations. country). Average consumption shows a sharp increase from consumption levels in the top quintile being nearly 20,000 Note: Only main income groups with at least 100 observations in Harari are included. the fourth to the fifth (richest) quintile, with the richest 20 Birr per year more than the rest of Ethiopia (Figure 4). Poverty is relatively lower for households headed by some secondary education is four percent, while it is zero for Figure 4 HOUSEHOLD CONSUMPTION REMAINS FAIRLY LOW EXCEPT FOR THE someone with higher education. With a poverty rate of households headed by someone with post-secondary ed- TOP QUINTILE 9 percent, poverty in Harari is highest among households ucation. Despite the significant differences in poverty rates headed by someone with either no education or incomplete across Harari and the rest of the country, Harari is similar to Average household expenditures per capita per year, 2016 primary education. The poverty rate for households headed the rest of the country in that higher education of the house- by someone who has completed primary education or have hold head is associated with lower poverty (Figure 6). Harari National excl. Harari 50,000 Average annual consumption 45,000 40,000 35,000 30,000 25,000 20,000 15,000 10,000 5,000 0 Poorest 2 3 4 Richest Consumption quintiles Source: HCES 2016; World Bank staff calculations. 166 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 167 Figure 6 EDUCATION IS A MAIN CORRELATE OF POVERTY Figure 7 OWNERSHIP OF CONSUMER DURABLES AND HOUSING CONDITIONS IMPROVED BETWEEN 2011 AND 2016 Poverty rates by educational attainment of the household head, 2016 Household durables, housing conditions and access to energy in 2011 and 2016, % of households Harari National excl. Harari with asset/characteristic 30% 28% A. Household Durables 25% 22% 2011 2016 Poverty rate 20% 80 15% 12% 10% 9% 9% 60 5% 4% 3% 40 0% 0% No education Incomplete Complete primary/ Post-secondary 20 primary Incomplete secondary Education level of household head 0 Mobile Mobile Refrigerator Refrigerator Animal Cart Motorcycle Animal Cart Motorcycle TV TV Source: HCES 2016; World Bank staff calculations. Note: The completed primary, incomplete secondary and completed secondary categories have been combined because of small sample sizes. Harari National excl. Harari NON-MONETARY WELFARE B. Housing/energy Improvements in non-monetary welfare accompanied shortage in the last 12 months before the interview de- the increase in consumption expenditure in Harari. In creased from eight percent in 2011 to zero in 2016. Food 2011 2016 Harari, the percentage of households that reported a food insecurity also decreased in the rest of the country (Table 4). 100 80 Table 4 FOOD SECURITY IMPROVED BETWEEN 2011 AND 2016 60 Incidence of food shortages and average duration of food shortages – food gap 40 2011 2016 Food shortage (%) Food gap (months) Food shortage (%) Food gap (months) 20 Harari 8.0 3.2 0 - 0 Improved wall Floor: tiles/cement Access to electricity Electricity for cooking Improved wall Floor: tiles/cement Access to electricity Electricity for cooking Improved roof Improved roof National 21.6 3.2 10.4 3.2 Source: DHS, 2011; 2016; World Bank staff calculations. Ownership of household durables and usage of elec- these durables in the region are above the rest of the country. tricity for cooking improved in Harari. Ownership of The proportion of households who used electricity for cook- household durables such as televisions, mobile phones, re- ing also improved. However, housing conditions did not im- Harari National excl. Harari frigerators, animal carts, and motorcycles increased between prove (Figure 7). 2011 and 2016, and the percentage of households owning Source: DHS, 2011; 2016; World Bank staff calculations. 168 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 169 Figure 8 ACCESS TO IMPROVED WATER SOURCES INCREASED BETWEEN 2011 Figure 9 CHILDREN’S MORTALITY DECREASED, AND THEIR HEALTH IMPROVED AND 2016 BETWEEN 2011 AND 2016 2011 2016 100 Access to water in 2011 and 2016, % of households with access/characteristics 2011 in 2011 Selected health and education variables and 2016 2016 80 100 Harari National excl. Harari A. Child and infant mortality rates (per 1000) 60 80 2011 2016 Piped water - compound Piped -outside Protected well/tube well Protected spring 40 60 100 Other improved Unimproved sources 20 40 80 100% 100% 17.9 16.8 46.4 35.2 0 20 60 Infant mortality Child mortality Under five Infant mortality Child mortality Under five 80% 5.0 4.9 80% 0 40 rate rate mortality rate rate rate mortality rate 15.2 9.3 Infant mortality Child mortality Harari Under five Infant mortality Child mortality National excl. Harari Under five 60% 23.6 60% 11.8 29.1 20 rate rate mortality rate rate rate mortality rate 9.0 17.0 40% 40% 10.1 Harari National excl. Harari 43.6 0 23.3 21.1 Infant mortality Child mortality Under five Infant mortality Child mortality Under five 32.1 20% 20% rate rate mortality rate rate rate mortality rate 11.0 14.2 Harari National excl. Harari 0% 0% Source: DHS, 2011; 2016; World Bank staff calculations 2011 2016 2011 2016 2011 2016 Source: DHS, 2011; 2016; World Bank Staff Calculations B. Immunization and health facility delivery (percent) 60 2011 2016 50 60 40 50 Access to improved drinking water sources86 changed to 20, and from 95 to 76 between 2011 and 2016. Delivery at 30 2011 2016 40 only slightly in Harari, while it improved in the rest of health facility increased from 32 to 50 percent, while the per- 60 20 30 the country. But a higher proportion of households have centage of fully immunized children increased from 34 to 42 50 10 access to improved drinking water sources in Harari than in percent. The stunting rate slightly increased from 30 percent 20 40 0 other regions combined in 2016 (83 percent versus 65 per- to 32 percent. Net primary school enrollment also improved, 10 30 Health facility Fully Stunted Health facility Fully Stunted cent) (Figure 8). while primary school completion and gross secondary enroll- 0 delivery immunized children delivery immunized children 20 children children ment decreased. Harari does better than the rest of the coun- Health facility Fully Stunted Health facility Fully Stunted Health indicators also improved, although challenges 10 delivery immunized children delivery immunized children try in all health and education indicators, except infant and Harari National excl. Harari remain. Infant, child and under-five mortality rates in Harari children children under-five mortality (Figure 9). 0 decreased respectively from 67 per thousand to 58, from 30 Health facility Harari Fully Stunted National Health facility excl. Harari Stunted Fully Source: DHS, 2011; delivery immunized 2016; World Bank children staff calculations delivery immunized children children children Harari C. Education (percent) National excl. Harari 2011 2016 80 2011 2016 70 80 60 70 50 60 40 2011 2016 50 80 30 40 70 20 30 60 10 20 50 0 10 40 Net primary Primary school Gross Net primary Primary school Gross enrolment completion secondary enrolment completion secondary 0 30 Net primary (15-24) Primary school school Gross Net primary (15-24) Primary school school Gross 20 enrolment completion enrolment secondary enrolment completion enrolment secondary 10 (15-24) Harari school (15-24) National excl. Harari school 86  Improved drinking water sources include piped water, protected well, protected spring, rainwater and bottled water. 0 enrolment enrolment Source: WMS, 2011; 2016; World Bank staff calculations. Net primary Primary school Harari Gross Primary Net primary National school excl. Harari Gross enrolment completion secondary enrolment completion secondary 170 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS (15-24) school (15-24) school 171 enrolment enrolment INEQUALITY OF OPPORTUNITY Figure 10 HARARI DOES BETTER MOST OF THE INDICATORS To assess inequality of opportunity87 in Harari and the rest of age-eligible children had completed primary school. These Coverage, HOI, and D-index for different outcomes, 2016 Ethiopia, we focus on six key outcomes: (i) primary school numbers are significantly higher than the rest of Ethiopia, enrollment, (ii) primary school completion, (iii) secondary where close to three-quarters of primary school-aged chil- Harari National excl. Harari school enrollment, (iv) access to electricity, (v) access to dren are enrolled and under one-third complete primary an improved water source, and (vi) access to a health post education. By comparison, almost one in three age-eligible 90 Coverage HOI D-index within 5 km. Circumstance variables 88 of the household in- children in Harari (aged 15 to 18) were enrolled in secondary 80 70 clude rural/urban location, gender, religion, agro-ecological school in 2016. This indicates that the education opportunity Percent/score 60 zone, consumption quintile of the child’s household, and set is quite different for the older cohort of children (15-18) 50 the household’s score on the rural accessibility index (RAI). than the younger cohort (7-14). 40 Figure 10 presents the coverage, Human Opportunity Index 89 30 The HOI scores for Harari are, like coverage scores, (HOI), and Dissimilarity index (D-index)across these key out- 20 generally much higher than the rest of Ethiopia. A high- come variables in 2016. 10 er HOI score is associated with more equitable access to a 0 In terms of coverage, Harari is ahead—sometimes far key service or outcome. For a given coverage rate, the lower alt ter alt ter alt ter st st st ec ry ool ec ry ool ec ry ool l l l m ve ess m ve ess m ve ess nd sc l co ry ool co ry ool ity oo ty oo ty oo oo po po po he wa he a he a ahead—of the rest of Ethiopia on key outcomes. Primary the HOI score, the more unequal the access to that key out- a h a h a h ric sch ici ch ici ch im ch im ch im ch 5k pro cc 5k pro cc 5k pro cc of d w of d w nd sc nd sc h h h s s pr ry s pr ry s pr ry s a a a of d co ry school completion and secondary school enrollment are low come is across different groups. In Harari, the coverage rate a a a a a a En let rim En let rim En let rim r r t t t in Ethiopia, but are significantly higher in Harari. This means of secondary school enrollment is 28 percent. However, this m m m m dp p p I I I ro ed ro ed ro ed se se se m ed m ed El El El Co rolle that the HOI scores for these key outcomes are always high- is skewed towards the urban and richer population in the Co ll Co roll d d d o in in in lle lle lle r p p p ith ith ith En En En er for Harari. They are particularly high compared to the rest region. The low HOI score of 20.5 for this outcome reflects W W W of the country when it comes to access to electricity and an strong differential access across the population in the region. improved water source (see the Coverage and HOI panels in Source: HCES, WMS 2011; 2016; World Bank staff calculations. The dissimilarity index, or D-index, shows how much Figure 14). The D-index for these variables – measuring how inequality in access exists between groups – a high- much these outcomes differ according to circumstances – is er index means less equal access across groups. The generally lower than the rest of the country. Decomposing this difference (not shown) reveals this is largely explained D-index for the three schooling outcomes in Harari is similar PUBLIC EXPENDITURE to the rest of Ethiopia, despite coverage in Harari being sig- by rural-urban differences in access rates in Ethiopia, and by nificantly better. This indicates that circumstances such as lo- In the previous sections, we discussed trends in monetary and 2016, although at the cost of spending on social sec- rural-urban and consumption quintile differences in Harari. cation, gender, and household income continue to matter in and non-monetary welfare between 2011 and 2016, and in- tors. Improvements in per capita expenditure on “pro-poor” The primary school enrollment rate in Harari is about Harari. The D-index score for electricity access in Harari is far equality of opportunities focusing on 2016, both in the Harari sectors—which includes agriculture and rural development, 8 percentage points higher than in the rest of Ethiopia. lower than the rest of the country, implying greater equality region and the rest of Ethiopia. In this section, we discuss water resources, education, and health—were stronger In the region, although 79 percent of age-eligible children in distribution of access to these key outcomes in the region trends, volume, and allocation of public expenditures to dif- in Harari compared to other regions combined, except for were enrolled in primary school in 2016, only 42 percent of compared to the rest of the country. ferent sectors to identify variation across regions and links health. Per capital expenditures in education, health, and to welfare. We use MoF data. The expenditure values are water resources in 2016 were significantly higher in Harari expressed in 2016 prices using the non-food CPI deflator. than other regions. These results are consistent with earlier We first present total per capita public expenditure and then sections showing that Harari improved in monetary and most discuss the expenditure allocation by sectors. Finally, we of non-monetary indicators. present expenditure on selected “pro-poor” sectors, which Per capita public expenditure increased between 2011 include agriculture and rural development, water resources, and 2016, and was higher in Harari than the rest of the health, and education. country for all years. Public spending in the Harari region Per capita public expenditure in Harari has been larg- increased from Birr 1,720 per person in 2011 to Birr 3,179 er than the rest of the country over the years, and a per person in 2016 (in constant 2016 Birr). Per capita public higher proportion of it is spent on the economic sector expenditure spending on average grew slightly at higher rate relative to the other regions. The share of public expen- (15 percent per year) in Harari compared to the rest of the diture on the economic sector grew in Harari between 2011 country (14 percent) (Figure 11). A full description of the methodology and variables used in the construction of this index in Ethiopia can be found in the technical 87  appendix. 88  These are generally thought of as variables over which a person does not have control, such as gender or region of birth. 89  Accessibility was also tested using a market accessibility variable, and the results were broadly consistent. 172 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 173 In Harari, the total share of spending on pro-poor sec- it decreased from 22 to 17 percent. Moreover, the share on Figure 11 PER CAPITA PUBLIC EXPENDITURE INCREASED OVER TIME tors decreased between 2011 and 2016, while spend- health significantly decreased in Harari from 12 percent in ing on health and education contracted. The share of 2011 to seven percent in 2016. Shares of spending on wa- Per capita real public expenditure, 2011 to 2016 spending on education decreased from 20 percent in 2011 ter and agriculture and rural development in the region were to 14 percent in 2016. Similarly, in other regions combined small in 2011, and remained so in 2016 (Figure 13B). Harari National excl. Harari 3,500 3,000 In 2016 Birr 2,500 Figure 13 EXPENDITURE SHARE ON HEALTH AND EDUCATION DECREASED IN 2,000 HARARI BETWEEN 2011 AND 2016 1,500 1,000 Real per capita expenditure and share of total expenditure (%) for selected sectors, 2011 and 2016 500 0 A. Real per capita expenditure in 2016 price 2011 2012 2013 2014 2015 2016 2011 2016 Source: MoF; World Bank staff calculations. 550 500 450 400 In 2016 Birr 350 A significant spending shift towards economic sectors Per capita spending91 in selected “pro-poor” activities 300 accompanied public expenditure growth in Harari. 90 increased between 2011 and 2016, except in health. In 250 200 Economic sector spending as share of the total expenditure Harari, increase in spending was particularly strong for water 150 increased from 25.2 in 2011 to 43.8 percent in 2016, while resources and agriculture and rural development; per capi- 100 50 for the rest of the country the spending on economic sector ta spending on these two activities increased from Birr 110 0 increased from 23.2 to 28.3 percent. Administration and so- to Birr 207 and from Birr 61 to Birr 133, respectively. Oth- Agriculture and Rural Development Water Resources Education Health Agriculture and Rural Development Water Resources Education Health cial-related spending decreased both in Harari and the rest of er regions also witnessed per capita spending increases in the country, and administration spending was lower in Harari pro-poor activities, but the rate of change is faster in Harari, both in 2011 and 2016 (Figure 12). except for health activity (Figure 13A). Harari National excl. Harari Figure 12 SPENDING SHARE OF ECONOMIC SECTOR INCREASED B. Expenditure share (%) Expenditure share by sectors (%), 2011 to 2016 Agriculture & Rural Dev. Water Resources Education Health Harari National excl. Harari 45 40 8.0 Administration Economic Social Other 12.0 35 7.8 2016 2016 30 22.2 6.0 2016 14.2 2016 6.5 17.4 22.5 29.7 25 20.2 11.7 20 14.3 26.7 7.1 15 31.8 37.9 10 5.5 5.6 36.3 5 5.3 5.6 25.2 6.9 6.8 2016 2016 2016 0 2.9 3.6 5.7 5.9 7.3 23.2 2016 43.8 27.8 28.3 2011 2016 2011 2016 27.7 Harari National excl. Harari Note: The sum does not add up to 100% because it includes only shares of selected “pro-poor” sectors. Note: The inner circles present the 2011 values. Source: MoF; World Bank staff calculations. Source: MoF; World Bank staff calculations. 90 Social sector includes education, health, culture and sports, and labor and social affairs while economic sector includes agriculture and rural development, water resources, trade and industry, mines and energy, transportation and communication, construction. Per capita spending refers to total spending in a particular sector divided by the population of the region. For example, per capita 91  spending on education refers to total spending on education divided by the total population of the region and not spending per student. 174 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 175 POVERTY AND INEQUALITY Food and nonfood consumption in Addis Ababa in- Consumption in Addis Ababa was higher than in other re- creased more than the rest of the country between gions combined by almost one-third in 2016 (Table 1). This 2011 and 2016. Mean consumption per adult increased is consistent with the higher consumption growth in urban by more than one-quarter in Addis Ababa alone, compared areas in general in the country. to a 13 percent increase for all other regions combined. Addis Ababa Table 1 CONSUMPTION INCREASED FASTER THAN THE REST OF THE COUNTRY IN ADDIS ABABA BETWEEN 2011 AND 2016 Mean annual household consumption per adult equivalent in December 2015 prices, 2011 and 2016 INTRODUCTION TOTAL 2011 2016 % change This chapter focuses on the evolution of socio-economic and health indicators such as poverty, inequality, Addis Ababa 12,831 16,237 26.5* access to basic services, and education in Addis Ababa between 2011 and 2016. Using data from a variety National excl. Addis Ababa 10,941 12,353 12.9* of sources, mainly the twinned household living standards surveys (HCES and WMS)92 and the Demographic and Health Survey (DHS), the chapter begins by documenting changes in monetary poverty and living stan- Note: * represents changes that are statistically significant. Source: HCES, 2011; 2016; World Bank staff calculations. dards in Addis Ababa. It then moves to an analysis of non-monetary welfare before turning to challenges of inequality of opportunity. The final section discusses changes in sectoral allocations of social spending in the region from 2011 to 2016. The chapter aims to provide policy makers with information and analysis that The increase in consumption in Addis Ababa led to a line in Addis Ababa decreased by 11 percentage points will be useful in improving the effectiveness of poverty reduction programs in the region. A fuller and more significant decrease in poverty. Poverty reduction was 93 from 28 percent in 2011 to 16 in 2016, in other regions, it detailed overview of changes in poverty in urban Ethiopia generally, and Addis Ababa in particular, can be faster in Addis Ababa compared to the rest of the country. decreased by six percentage points from 30 percent to 24 found in Chapter 5 of the World Bank’s 2019 Ethiopia Poverty Assessment. While the proportion of people below the national poverty percent (Table 2). Some of the main findings include: ● The poverty rate in Addis Ababa dropped from 28 per- improved. Addis Ababa has better human capital indi- Table 2 POVERTY DECREASED IN ADDIS ABABA FASTER THAN THE REST cent to 17 percent. In 2016, the poverty rate in Addis cators than the rest of the country. This is also reflected ETHIOPIA Ababa was significantly lower than the national poverty in a higher score on the Human Opportunity Index than Percentage of people below the national poverty line rate. the national average – implying higher equality of access to key basic services for children in Addis Ababa com- TOTAL ● Consumption growth in Addis Ababa is positive and pared to the rest of Ethiopia elsewhere in Ethiopia. 2011 2016 % pt. change strong everywhere though relatively stronger at the top Addis Ababa 28.1% 16.8% -11.3* of the consumption distribution. Inequality increased be- ● Average per capita public expenditure in Addis Ababa is tween 2011 and 2016, and by 2016 inequality in Addis higher than in the rest of the country. Per capita expen- National excl. Addis Ababa 29.6% 23.7% -5.9* Ababa was higher than in the rest of Ethiopia. ditures on education, health and water resources are Note: * represents changes that are statistically significant. also higher in Addis Ababa. Source: HCES, 2011; 2016; World Bank staff calculations. ● Non-monetary indicators including ownership of con- sumer durables, access to drinking water and health 92 Household Consumption Expenditure Survey (HCES) and Welfare Monitoring Survey (WMS) are the two official household surveys 93 The relatively slower poverty reduction in Addis Ababa could partly be due the higher rural population share (80% for Addis Ababa collected by CSA every five year and used to monitor welfare. HCES covers monetary welfare and WMS deals with non-monetary versus 75% for the rest of the country in 2016). Poverty reduction was low in rural areas in general. welfare like health and education. Both surveys are collected from the same sample of households. 176 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 177 Consumption growth in urban Addis Ababa was posi- about three percent at the bottom, and increases moving tive and strong, particularly at the top of the consump- up the consumption distribution, reaching seven percent at Box 1 Where does poverty in Addis Ababa concentrate? tion distribution. The rate of consumption growth was the top (Figure 1). Based on the geo-localization included in the 2016 HCES, it is possible to identify clusters of low and high-in- come households in Addis Ababa. Although Addis Ababa is socially and economically mixed, there are a number of distinct low-income neighborhoods. The largest cluster of poorer households (blue dots) surrounds Merkato in Figure 1 GIC FOR ADDIS ABABA Addis Ketema and Lideta (Geja Sefer neighborhood), while a second, smaller low-income cluster is found in the east, between Goro and Gurd Shola (Figure 1). Small high-income clusters (red dots) are scattered across the city. Future work will look in more detail at the interplay of location, housing, and poverty in Addis Ababa. Figure 1 LOW-  AND HIGH-INCOME CLUSTERS IN ADDIS A. Low- and high-income clusters B. Low-income only ! ! !! ! !! ! ! ! ! ! ! ! ! ! !! ! ! ! ! !! ! ! ! !! ! !! ! !! !! ! ! ! ! ! !!!!! ! !! !! ! ! ! !!! ! ! ! ! !! !! ! ! !! !! ! ! !! ! !! ! ! ! ! ! ! !! ! !! ! ! ! ! !! ! ! ! ! !! !! ! !! ! ! ! ! ! ! ! ! !! !! ! !! !! !! ! ! ! ! ! ! !! !! !!! !!! ! ! ! ! ! !! !! ! ! ! ! !! !!!! ! !! ! ! ! ! ! ! !! ! ! ! ! ! !! !! !!! ! ! ! !! ! !! ! ! !! ! !!! !! !! ! ! ! ! ! !!! ! ! ! ! ! ! ! ! !! ! ! ! !!! ! ! ! ! !! ! ! !! ! !! !!!! ! !! ! ! ! ! ! ! ! !! ! ! ! ! !!!!! !! ! ! ! ! ! ! !! ! !!! !! ! ! ! !! ! ! ! ! ! ! ! !! ! ! !! ! ! ! ! !! ! !!! ! !! ! !! ! ! ! ! ! !! !!! ! ! !! ! !!! ! ! ! ! ! !! ! ! ! !! ! !! !! ! !! ! ! ! !! ! ! ! ! ! ! ! ! !! !!! ! ! ! !!! ! ! ! ! ! ! ! ! ! ! !! ! ! ! !! !! ! ! ! !! ! !! ! !! !! ! ! !! ! ! ! !! ! ! ! ! ! ! ! ! !! !! !! ! ! ! ! !! ! !!! ! ! ! ! ! ! ! !!! ! ! ! ! !!! !!! ! ! ! !! !! ! ! ! ! !! ! ! ! ! ! ! ! ! !! ! !! ! ! ! ! !! ! ! ! ! ! !! ! !! ! ! ! ! ! !! ! !! !! ! ! ! ! !! ! ! !!! ! ! !!! ! ! ! ! !! ! !!! ! ! !! ! !! !! !!! !!!!! ! ! !! !! !!!!! !! !!! !! ! ! !! ! ! !!!! ! ! ! ! ! ! !! !! !! !! ! !!! !! ! ! !! !! ! ! !! ! !!! !!! !!!! ! ! !! ! ! ! ! ! ! !!! ! ! ! !!! ! ! !!! ! !! ! !! ! ! ! ! ! ! !! ! !! !! ! ! ! ! ! ! !! !! !!! ! !! ! ! ! ! ! ! ! !! !! !! ! ! !! ! ! ! !! ! ! ! ! !!!! ! ! !! ! ! !! !! !! ! ! ! ! ! ! !! ! ! ! ! ! !!!! !! ! !! ! ! ! ! ! !! !!! !! !! ! ! !! ! !! ! ! ! !! ! ! ! ! ! ! !! ! ! ! ! ! ! ! ! ! ! ! !! ! ! ! ! !! ! ! ! ! ! ! ! ! ! !! ! ! ! ! ! ! ! ! ! ! Source: HCES, 2011; 2016; World bank staff calculations. !! ! ! !! ! ! ! ! ! ! ! !!! ! !! ! ! ! !! ! !! ! ! ! !!! ! ! ! ! ! !! ! !! ! ! ! !! ! !! ! !! !! ! ! !! ! !! !! ! ! !! ! ! ! Relatively weak consumption growth at the bottom of consumption at the 90th percentile and consumption at the ! ! ! !! !! ! ! ! ! !! !! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! !! !! ! !! ! ! ! ! !! ! ! ! ! ! ! ! ! ! ! ! ! ! !! ! ! ! !! ! ! ! ! the distribution, combined with strong growth at the 10th percentile increased from 4.5 in 2011 to 5.5 in 2016, ! ! ! !!! ! top, led to an increase in inequality. The Gini index in- reflecting strong consumption growth at the top and weaker ! !! ! ! !! ! ! creased from 34 to 38 between 2011 and 2016. The ratio of growth at the bottom (Table 3). Source: HCES, 2016. Low-income clusters in blue. Reproduced from Chapter 5 of the World Bank’s 2019 Ethiopia Poverty Assessment. Table 3 INEQUALITY INCREASED BETWEEN 2011 AND 2016 2011 2016 Gini index p90/p10 ratio Gini index p90/p10 ratio Addis Ababa 33.6 4.5 37.8 5.5 National 29.8 3.6 32.9 4.3 Source: HCES, 2011; 2016; World Bank staff calculations. 178 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 179 Poverty reduction in Addis Ababa can be entirely at- of Ethiopia. The redistribution effect is small and positive for tributed to increases in consumption, with the increase Addis Ababa, but less so than in the rest of the country in Figure 3 HOUSEHOLD CONSUMPTION REMAINS FAIRLY LOW EXCEPT FOR THE in inequality slightly hindering poverty reduction. Figure which it is also positive (poverty-increasing). If there were no TOP QUINTILE 2 decomposes the change in poverty into growth and redis- changes in the distribution between 2011 and 2016, poverty Average household expenditures per capita per year, 2016 tribution components using the Datt-Ravallion decomposi- in the rest of the country would have decreased by 8.4 per- tion technique. It indicates which part of the poverty change centage points. In Addis Ababa, changes in inequality had a Addis Ababa National excl. Addis Ababa can be attributed to: (i) the change in average consumption relatively small effect on poverty reduction. Had there been (growth effect), and (ii) the change in the redistribution of no change in measured inequality, poverty reduction in Addis 40,000 Average annual consumption consumption; that is, changes in inequality (redistribution Ababa would have been 13.3 percentage points rather than 35,000 effect). The growth effect on poverty in Addis Ababa is neg- 11.3 percentage points. 30,000 ative (poverty-reducing), and is larger than it is in the rest 25,000 20,000 15,000 10,000 Figure 2 CONSUMPTION GROWTH DROVE POVERTY REDUCTION WHILE AN 5,000 INCREASE IN INEQUALITY HAMPERED POVERTY REDUCTION 0 Poorest 2 3 4 Richest Datt-Ravallion decomposition of poverty change between 2011 and 2016 Consumption quintiles Source: HCES 2016; World Bank staff calculations. Addis Ababa National excl. Addis Ababa Contribution to change in poverty 4.0 2.0 2.1 2.5 0.0 Figure 4 CASUAL LABORERS HAVE THE HIGHEST POVERTY RATES -2.0 -4.0 -5.9 Poverty rates by household’s main income source, 2016 -6.0 -8.4 -8.0 -10.0 -11.3 Addis Ababa National excl. Addis Ababa -12.0 -13.3 -14.0 35% -16.0 30% Poverty reduction Consumption growth Increase in inequality 25% Poverty rate Source: HCES, 2011; 2016; World Bank staff calculations. 20% 15% 10% Average consumption per quintile in Addis Ababa is jump from the fourth to the fifth quintile in Addis Ababa is 5% significantly higher than in the rest of Ethiopia, except particularly large, with consumption levels in the top Addis 0% for the very poorest. Average annual consumption per Ababa quintile nearly 10,000 Birr per year more than in the r ns il es t es en o ta lab c vic io re m an at adult in 2016 was 5,768 Birr for the poorest quintile (ver- rest of Ethiopia (Figure 3). er oy d al n itt an -s do su pl m em Ca Re le s/ sus 4,824 for the rest of the country) and 18,217 Birr for lf Se a n es io ied In Addis Ababa, households that primarily depend on ns l ho the fourth quintile (versus 13,139 for the rest of the coun- lar Pe -w Sa casual labor have the highest poverty rates. Those who try). Average consumption shows a sharp increase from the lf Se depend on salaried employment and services have the lowest fourth to the fifth (richest) quintile, with the richest 20 percent poverty rates. Addis Ababa diverges from the rest of the coun- Main source of household income consuming around twice as much as the fourth quintile, on try in that the poverty rate for households relying on remittanc- average, reflecting strong consumption growth at the top of es as their main income source is close to double (21 percent) Source: HCES 2016; World Bank staff calculations. the distribution. Although the distribution in Addis Ababa is that of other regions combined (12 percent) (Figure 4). Note: Only main income groups with at least 100 observations in Addis Ababa are included. always significantly above that of the rest the country, the 180 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 181 Educational attainment of household heads strong- education of the household head leads to decreasing pover- ly negatively correlates with poverty rates. With a rate ty rates, reaching merely four percent for households headed Figure 6 THE SERVICES SECTOR IS THE LARGEST CONTRIBUTOR TO POVERTY of 32 percent, poverty in Addis Ababa is highest among by someone with post-secondary education. Addis Ababa REDUCTION IN ADDIS ABABA households whose heads have no education. Increasing closely mirrors the trend for the rest of the country (Figure 5). Sectoral decomposition of poverty changes 2011 to 2016 Agriculture Manufacturing Construction Services Other Population shift Interaction Figure 5 EDUCATION IS A MAIN CORRELATE OF POVERTY Addis Ababa National excl. Addis Ababa 0 Change in poverty headcount Poverty rates by educational attainment of the household head, 2016 -2 Addis Ababa National excl. Addis Ababa -4 -6 35% 32% -8 28% 30% -10 25% 23% 22% Poverty rate 20% -12 20% -14 12% 13% 15% 11% Source: HCES; 2011, 2016. World Bank staff calculations. 10% 7% 6% 4% 3% 5% 0% No education Incomplete Complete Incomplete Complete Post-secondary primary primary secondary secondary NON-MONETARY WELFARE Education level of household head Improvements in non-monetary welfare accompanied for those who faced food shortage, the average number of the increase in consumption expenditure. In Addis Aba- months with a food shortage did not decrease much. The Source: HCES 2016; World Bank staff calculations. ba, the percentage of households that reported a food short- results for the rest of the country are similar, although food age in the last 12 months before the interview decreased security is significantly higher nationwide than in Addis Aba- from eight percent in 2011 to one percent in 2016. However, ba (Table 4). Services was the biggest driver of poverty reduction shift in population from agriculture to manufacturing or ser- in Addis Ababa, in line with other urban areas, but is vices) explained very little of poverty reduction in Addis Aba- Table 4 FOOD SECURITY IMPROVED BETWEEN 2011 AND 2016 very different to the rural parts of the country. Around ba over the period. Although this effect has been shown to Incidence of food shortages and average duration of food shortages – food gap 12 percent of poverty reduction in the rest of Ethiopia can be be important in other regions, it is relatively less important in explained by improvements in the services sector, while the in Addis Ababa; this is not surprising given that the share of 2011 2016 corresponding figure for Addis Ababa is far higher at almost agricultural households was already very low, falling from 2 to Food shortage (%) Food gap (months) Food shortage (%) Food gap (months) 50 percent. Gains for households working in agriculture was 1 percent of households over the period. The share of house- Addis Ababa 7.8 4.0 1.1 3.6 small, but the share of households engaged in this activity is holds primary reliant on the services sector was stable over National excl. Addis Ababa 22.1 3.2 10.7 3.3 far lower because of the urban nature of the city/region. The the period at around half, while the share engaged primari- so-called population shift effect had a negligible impact on ly in manufacturing decreased from 11 percent to around 8 Source: WMS; 2011, 2016. World Bank staff calculations. poverty reduction in Addis Ababa. This variable, often used percent (Figure 6). as an indicator of structural transformation (proxied by the 182 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 183 Ownership of household durables and use of electrici- also improved. But there was no improvement in the quality ty for cooking also improved in Addis Ababa. Ownership of other housing quality indicators such as the quality of roof. Figure 8 ACCESS TO IMPROVED WATER SOURCES IN ADDIS ABABA UNIVERSAL of household durables such as televisions, mobile phones, Improved roof and access to electricity were almost universal Access to water in 2011 and 2016, % of households with access/characteristics refrigerators, animal carts, and motorcycles increased be- even in 2011. Similar improvements were also observed in tween 2011 and 2016. The proportion of households with other regions combined (Figure 7). an improved wall and those who use electricity for cooking Addis Ababa National excl. Addis Ababa Piped water - compound Piped -outside Protected well/tube well Protected spring Other improved Unimproved sources Figure 7 OWNERSHIP OF CONSUMER DURABLES AND HOUSING CONDITIONS 0.4 2.5 IMPROVED BETWEEN 2011 AND 2016 100% 100% 31.1 14.2 48.6 36.8 Household durables, housing conditions and access to energy in 2011 and 2016, % of households 80% 80% 82.3 with asset/characteristic 68.3 0.7 A. Household Durables 60% 60% 12.3 2011 2016 9.5 17.8 2011 2016 40% 40% 120 10.7 120 22.9 21.4 100 20% 20% 100 80 8.2 11.1 0% 0% 80 60 2011 2016 2011 2016 60 40 Source: DHS, 2011; 2016; World Bank Staff Calculations 40 20 20 0 Mobile Mobile Refrigerator Refrigerator Cart Motorcycle Cart Motorcycle TV TV TV TV 0 Access to improved drinking water sources94 was al- increased from 82 to 97 percent, while the percentage of Mobile Mobile Refrigerator Refrigerator Cart Motorcycle Cart Motorcycle Animal Animal ready universal in 2011 in Addis Ababa, while it im- fully immunized children increased from 79 to 89 percent. Animal Animal proved in the rest of the country (Figure 8). The stunting rate decreased from 22 percent to 15 percent. Addis Ababa National excl. Addis Ababa However, education indicators did not improve. All health Addis Ababa National excl. Addis Ababa Improvements were also observed in health indicators. and education indicators improved for other regions com- The infant and under-five mortality rates in Addis Ababa de- bined, although they lag Addis Ababa (Figure 9). creased respectively from 35 per thousand to 32, and from B. Housing/energy 41 to 38 between 2011 and 2016. Delivery at health facility 2011 2016 2011 2016 120 120 100 100 80 80 60 60 40 40 20 20 0 wallwall tiles/cement to electricity for cooking wallwall tiles/cement to electricity for cooking roofroof roofroof 0 tiles/cement to electricity for cooking tiles/cement to electricity for cooking Improved Improved Improved Improved Improved Improved Improved Improved Electricity Electricity Access Access Floor: Floor: Electricity Electricity Access Access Floor: Floor: Addis Ababa National excl. Addis Ababa Addis Ababa National excl. Addis Ababa Source: DHS, 2011; 2016; World Bank staff calculations. 94  Improved drinking water sources include piped water, protected well, protected spring, rainwater and bottled water. 184 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 185 INEQUALITY OF OPPORTUNITY Figure 9 CHILDREN’S MORTALITY DECREASED, AND THEIR HEALTH IMPROVED 2011 2016 BETWEEN 2011 AND 2016 To assess inequality of opportunity95 in Addis Ababa and the These numbers are both much higher than the rest of Ethi- 100 rest of Ethiopia, we focus on six key outcomes: (i) primary opia, in which just over 71 percent of primary school-aged 2016 2011 in 2011 Selected health and education variables and 2016 80 school enrollment, (ii) primary school completion, (iii) second- children are enrolled and under one third complete primary 100 ary school enrollment, (iv) access to electricity (v) access to education. By comparison, more than one in three age-eli- 60 A. Child and infant mortality rates (per 1000) 80 an improved water source, and (vi) access to a health post gible children in Addis Ababa (aged 15 to 18) were enrolled 2011 2016 40 within 5 km. Circumstance variables 96 of the household in- in secondary school in 2016. This indicates that educational 60 100 clude rural/urban location, gender, religion, agro-ecological opportunity is quite different for the older cohort of children 20 40 80 zone, consumption quintile of the child’s household, and (15-18) than the younger cohort (7-14). 0 the household’s score on the rural accessibility index (RAI).97 20 60 Infant mortality Child mortality Under five Infant mortality Child mortality Under five The HOI scores for Addis Ababa are, like the coverage rate rate mortality rate rate rate mortality rate Figure 10 presents the coverage, Human Opportunity Index 0 40 scores, generally much higher than the rest of Ethio- Infant mortality Addis Child Ababa mortality Under five National Infant mortality excl. Child Addis Ababa mortality Under five (HOI), and Dissimilarity index (D-index) across these key out- rate rate mortality rate rate rate mortality rate pia. A higher HOI score is associated with more equitable 20 come variables in 2016. Addis Ababa National excl. Addis Ababa access to a key service or outcome. For a given coverage 0 In terms of coverage, Addis Ababa is ahead—some- rate, the lower the HOI score, the more unequal the access Infant mortality Child mortality Under five Infant mortality Child mortality Under five rate rate mortality rate rate rate mortality rate times far ahead—of the rest of Ethiopia on all key out- to that key outcome is across different groups. In Addis Aba- Addis Ababa National excl. Addis Ababa comes. Primary school completion and secondary school ba, the coverage rate of secondary school enrollment is 34 Source: DHS, 2011; 2016; World Bank staff calculations enrollment are low in Ethiopia, but are significantly higher in percent. The HOI score is only slightly lower at 32, indicating 2011 2016 Addis Ababa. This means that the HOI scores for these key that differences in enrollment rates across different groups is 100 B. Immunization and health facility delivery (percent) outcomes are always higher for Addis Ababa. HOI scores are not prominent. 2011 2016 particularly high compared to the rest of the country when it 80 The dissimilarity index, or D-index, shows how much 100 comes to access to electricity (see the Coverage and HOI 60 inequality in access there is between groups – a high- 80 panels in Figure 14).98 The D-index for these variables – mea- 2011 2016 er index means less equal access across groups. The suring how much these outcomes differ according to circum- 40 60 100 D-index for the three schooling outcomes in Addis Ababa is stances – is generally lower than the rest of the country. 20 far smaller than in the rest of Ethiopia, largely because cover- 40 80 The primary school enrollment rate in Addis Ababa is age in Addis Ababa is significantly better. This indicates that 0 20 60 about 16 percentage points higher than in the rest of circumstances such as gender and household income mat- Health facility Fully Stunted Health facility Fully Stunted delivery immunized children delivery immunized children Ethiopia. In the region, although 87 percent of age-eligible ter less in explaining differences in Addis Ababa than they do 0 40 children children Health facility Fully Stunted Health facility Fully Stunted children were enrolled in primary school in 2016, only 56 per- in the rest of Ethiopia. 20 delivery immunized Addis Ababa children Nationalimmunized delivery children excl. Addis Ababa cent of age-eligible children had completed primary school. children children 0 Addis Ababa National excl. Addis Ababa Health facility Fully Stunted Health facility Fully Stunted Source: DHS, 2011; delivery immunized 2016; World Bank children staff calculations delivery immunized children children children C. Education (percent) Addis Ababa National excl. Addis Ababa 2011 2016 100 2011 2016 80 100 60 80 2011 2016 40 60 100 20 40 80 0 20 60 A full description of the methodology and variables used in the construction of this index in Ethiopia can be found in the technical 95  Net primary Primary school Gross Net primary Primary school Gross enrolment completion secondary enrolment completion secondary appendix. 0 40 (15-24) school (15-24) school Net primary Primary school Gross Net primary Primary school Gross 96  These are generally thought of as variables over which a person does not have control, such as gender or region of birth. enrolment enrolment 20 enrolment completion secondary enrolment completion secondary 97  Accessibility was also tested using a market accessibility variable, and the results were broadly consistent. Addis Ababa (15-24) school National excl. Addis Ababa school (15-24) enrolment enrolment 98 Unlike in the rest of the country, access to an improved water source is not shown in this chapter as coverage of this variable is 100% 0 according to HICES 2015/2016 data. Source: WMS, 2011; 2016; World Bank Net primary staff Primary calculations. school Gross Net primary Primary school Ababa Gross Addis Ababa National excl. Addis enrolment completion secondary enrolment completion secondary 186 (15-24) school (15-24) REGIONAL ETHIOPIA school POVERTY REPORT SECTION II: REGIONAL ANALYSIS 187 enrolment enrolment Figure 10 UNSURPRISINGLY, ADDIS ABABA DOES BETTER THAN THE REST OF Figure 11 PER CAPITA PUBLIC EXPENDITURE INCREASED OVER TIME ETHIOPIA OVER ALL KEY OUTCOMES Per capita real public expenditure, 2011 to 2016 Coverage, HOI, and D-index for different outcomes, 2016 Addis Ababa National excl. Addis Ababa Addis Ababa National excl. Addis Ababa 3,600 100 3,200 Coverage HOI D-index 2,800 In 2016 Birr 90 2,400 80 2,000 Percent/score 70 1,600 60 1,200 800 50 400 40 0 30 2011 2012 2013 2014 2015 2016 20 10 Source: MoF; World Bank staff calculations. 0 ol ol ol ol ol ol ss ss ss ol ol ol ho ho ho ho ho ho ho ho ho ce ce ce sc sc sc sc sc sc sc sc sc ac ac ac A spending shift away from other sectors and towards Ababa, while it decreased in the rest of the country, but it is y y y y y y y y y y y y ar ar ar ar ar ar ar ar ar cit cit cit im im im nd nd nd im im im tri tri tri social and economic sectors accompanied public ex- still large and accounted for 32 percent in 2016 (Figure 12). pr pr pr pr pr pr co co co ec ec ec ed ed ed se se se d d d El El El lle lle lle penditure growth in Addis Ababa. 99 Social and economic et et et d d d ro ro ro lle lle lle pl pl pl Per capita spending100 in selected “pro-poor” activities En En En m m m ro ro ro sector spending as share of total expenditure increased from Co Co Co En En En increased between 2011 and 2016. In Addis Ababa, the 20 percent to 23 percent and from 49 percent to 52 per- Source: HCES, WMS 2011; 2016; World Bank staff calculations. increase was particularly strong for water resources, educa- cent, respectively. For the rest of the country, spending on tion, and health sectors; per capita spending on water, ed- economic sector increased from 19 to 25 percent, while the ucation, and health sectors increased from Birr 195 to Birr spending share on the social sector decreased from 34 to 29 370, from Birr 230 to Birr 352.3 Birr, and from Birr 101 to percent. Administration-related spending increased in Addis PUBLIC EXPENDITURE In the previous sections, we discussed the trends in mone- per capita expenditure on pro-poor sectors were stronger in tary and non-monetary welfare between 2011 and 2016, and Addis Ababa compared to the rest of the country. Per capital SPENDING SHARE ON ECONOMIC SECTOR INCREASED Figure 12 inequality of opportunities focusing on 2016 both in Addis expenditures in all sectors in 2016 were significantly higher Ababa and the rest of the country. In this section, we discuss than in other regions combined, except in agriculture and ru- Expenditure share by sectors (%), 2011 to 2016 trends, volume, and allocation of public expenditures to iden- ral development. These results are consistence with results Addis Ababa National excl. Addis Ababa tify variations across regions and links with to welfare. We presented earlier showing that Addis Ababa did better on Administration Economic Social Other use MoF data. We express all the expenditure values in 2016 monetary and most non-monetary indicators. prices using the non-food CPI deflator. We first present total 2016 2016 2016 Per capita public expenditure increased between 2011 11.7 13.7 14.5 per capita public expenditure and then discuss the expendi- 2016 and 2016, and was higher in Addis Ababa than the rest 12.0 32.1 ture allocation by sector. Finally, we present expenditure on 18.8 of the country for all years. Public spending in Addis Aba- 5.4 selected “pro-poor” sectors, which include agriculture and 2016 ba increased from Birr 2,150 per person in 2011 to Birr 3,514 22.9 19.8 33.6 41.7 rural development, water resources, health, and education. per person in 2016 (in constant 2016 Birr), with the lowest 49.4 2016 2016 19.3 2016 Per capita public expenditure in Addis Ababa is larger per capita expenditure of Birr 1,926 recorded in 2012. Al- 51.8 28.6 24.9 than in the rest of the country, and a higher propor- though per capita spending on average increased at a slight- tion is spent on economic sectors in the region. Be- ly faster rate in the rest of the country than in Addis Ababa, Note: The inner circles represent the 2011 values. tween 2011 and 2016, spending shifted in favor of social the gap in absolute terms remained significant (Figure 11). Source: MoF; World Bank staff calculations. and economic sectors from other sectors. Improvements in 99 Social sector includes education, health, culture and sports, and labor and social affairs while economic sector includes agriculture and rural development, water resources, trade and industry, mines and energy, transportation and communication, construction. Per capita spending refers to total spending in a particular sector divided by the population of the region. For example, per capita 100  spending on education refers to total spending on education divided by the total population of the region and not spending per student. 188 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 189 Birr 237, respectively. Other regions also experienced a per In Addis Ababa, the total share of spending on pro- capita spending increment on pro-poor activities, although poor activities did not change between 2011 and 2016, the rate of change is faster in Addis Ababa, except for agri- except on education. The share of spending on each pro- culture and rural development. In Addis Ababa, such growth poor sector also did not change much. The share of spend- has resulted in higher per person spending on the sectors ing on the health sector increased from five percent to seven compared to other regions combined in 2016, again except percent, while that of education decreased from 12 percent for agriculture and rural development (Figure 16A). to ten percent. The share spent on education also decreased in the rest of the country (Figure 13B). Dire Dawa Figure 13 EXPENDITURE SHARE ON HEALTH INCREASED IN ADDIS ABABA BETWEEN 2011 AND 2016 Real per capita expenditure and share of total expenditure (%) for selected sectors, 2011 and 2016 INTRODUCTION A. Real per capita expenditure in 2016 price This chapter focuses on evolution of socio-economic and health indicators such as poverty, inequality, ac- 2011 2016 cess to basic services, and education, in Dire Dawa between 2011 and 2016. Using data from a variety of 400 sources, mainly the twinned household living standards surveys (HCES and WMS)101 and the Demographic 350 and Health Survey (DHS), the chapter begins by documenting changes in monetary poverty and living stan- 300 dards in Dire Dawa. It then moves to an analysis of non-monetary welfare before turning to challenges of In 2016 Birr 250 inequality of opportunity. The final section discusses changes in the sectoral allocations of social spending in 200 the region from 2011 to 2016. The chapter aims to provide policy makers with information and analysis that 150 100 will be useful in improving the effectiveness of poverty reduction programs in the region. A fuller and more 50 detailed overview of changes in poverty in urban Ethiopia can be found Chapter 5 of the World Bank’s 2019 0 Ethiopia Poverty Assessment. Agriculture and Rural Development Water Resources Education Health Agriculture and Rural Development Water Resources Education Health Some of the main findings include: ● The poverty rate in Dire Dawa stood at seven percent ● Non-monetary indicators including ownership of con- in 2016, compared to 11 percent in 2011. This change sumer durables and access to drinking water improved, was not, however, statistically significant. Poverty reduc- as was the case for human capital indicators such as Addis Ababa National excl. Addis Ababa tion was statistically significant only in urban areas. health and education. Dire Dawa in general has better human capital indicators than the rest of the country. B. Expenditure share (%) ● The pattern of consumption growth in Dire Dawa was in- This is also reflected in a higher score on the Human Agriculture & Rural Dev. Water Resources Education Health equality-increasing, with growth in both rural and urban Opportunity Index than the national average – implying 50 areas being strongly concentrated at the top of the dis- higher equality of access to key basic services for chil- 45 tribution. There was no significant consumption growth 8.5 dren in Dire Dawa compared to the rest of Ethiopia. 40 at the bottom of the distribution in urban Dire Dawa and 35 8.0 30 23.7 growth at the bottom was negative in rural Dire Dawa. ● Average per capita public expenditure in Dire Dawa is 25 18.5 By 2016 inequality in the region was higher than in the higher than in the rest of the country. Per capita ex- 5.2 6.7 20 11.8 rest of Ethiopia. penditures on education and health are also higher in 9.9 15 10 4.8 4.9 Dire Dawa. 10.0 10.4 5 7.8 7.7 0.4 0.7 0 2011 2016 2011 2016 Addis Ababa National excl. Addis Ababa Note: The sum does not add up to 100% because it includes only shares of selected “pro-poor” sectors. Source: MoF; World Bank staff calculations. 101 Household Consumption Expenditure Survey (HCES) and Welfare Monitoring Survey (WMS) are the two official household surveys collected by CSA every five year and used to monitor welfare. HCES covers monetary welfare and WMS deals with non-monetary welfare like health and education. Both surveys are collected from the same sample of households. 190 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 191 POVERTY, AND INEQUALITY Consumption growth in urban Dire Dawa was pos- increases moving up the distribution. In rural areas, on the itive for the entire consumption distribution, while other hand, there was no consumption growth for the first Food and nonfood consumption in Dire Dawa increased attributed to urban areas, where consumption increased by consumption growth in rural areas varied but was on and last quarters of the consumption distribution, while it significantly between 2011 and 2016, although only ur- more than three-quarters, higher than the increase of 34 average lower. In urban areas, consumption growth was was positive in the middle of the distribution (Figure 1). ban areas experienced the full benefit of this increase. percent at the national level. Rural areas, however, were left positive and strong across the consumption distribution, and Mean consumption per adult increased by more than half behind. Consumption growth in rural Dire Dawa was not sta- in Dire Dawa alone compared to 13 percent for all other re- tistically significant (Table 1). gions combined. The increase in consumption can be solely Figure 1 GIC FOR DIRE DAWA: URBAN AND RURAL Table 1 CONSUMPTION INCREASED IN DIRE DAWA FASTER THAT THE REST OF Urban Rural THE COUNTRY BETWEEN 2011 AND 2016 Mean annual household consumption per adult equivalent in December 2015 prices, 2011 and 2016 TOTAL URBAN RURAL 2011 2016 % change 2011 2016 % change 2011 2016 % change Dire Dawa 11,268 17,428 54.7* 11,617 20,718 78.3* 10,532 11,393 8.2 National excl. 11,007 12,475 13.3* 13,947 18,614 33.5* 10,433 11,014 5.6* Dire Dawa Note: * represents changes that are statistically significant. Source: HCES, 2011; 2016; World Bank staff calculations. Poverty in Dire Dawa also decreased faster than the than two-thirds (falling from 35 to 11 percent), while poverty rest of the country between 2011 and 2016. The propor- reduction in rural areas was not statistically significant. For tion of people below the national poverty line in Dire Dawa the rest of the country, urban poverty and rural poverty de- Source: HCES, 2011; 2016; World bank staff calculations. decreased by 13 percentage points from 28 percent in 2011 crease by 11 percentage points and 5 percentage points, to 15 in 2016. In urban areas, poverty decreased by more respectively (Table 2). Lack of consumption growth at the bottom of the ru- ratio of consumption at the 90th percentile compared to the ral consumption distribution, combined with strong 10th percentile increased from 3.4 in 2011 to 5.7 in 2016, Table 2 POVERTY DECREASED IN DIRE DAWA BUT ONLY IN URBAN AREAS growth in urban areas—particularly at the top of the reflecting strong consumption growth at the top and lack of Percentage of people below the national poverty line distribution—led to an increase in inequality. The Gini growth at the bottom (Table 3). index increased from 29 to 39 between 2011 and 2016. The TOTAL URBAN RURAL % pt. % pt. % pt. 2011 2016 change 2011 2016 change 2011 2016 change Table 3 INEQUALITY INCREASED BETWEEN 2011 AND 2016 Dire Dawa 28.3% 15.4% -12.8* 34.9% 11.1% -23.8* 14.2% 23.3% 9.1 National excl. 29.6% 23.5% -6.1* 25.5% 14.6% -10.8* 30.4% 25.6% -4.8* Dire Dawa 2011 2016 Note: * represents changes that are statistically significant. Gini index p90/p10 ratio Gini index p90/p10 ratio Source: HCES, 2011; 2016; World Bank staff calculations. Dire Dawa 29.3 3.4 39.4 5.7 National 29.8 3.6 32.9 4.3 Source: HCES, 2011; 2016; World Bank staff calculations. 192 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 193 The rise in inequality between 2011 and 2016 is mainly from differences in average consumption between urban due to the increasing gap between urban and rural ar- and rural areas – rose very sharply from 7 percent in 2011 Figure 3 CONSUMPTION GROWTH DROVE POVERTY REDUCTION WHILE AN eas. The difference in average consumption between urban to 33 percent in 2016. Rising inequality within both urban INCREASE IN INEQUALITY HAMPERED POVERTY REDUCTION and rural areas was modest in 2011, but the gap became and rural areas was not as evident, as the “within” share of Datt-Ravallion decomposition of poverty change between 2011 and 2016 much larger by 2016. Urban consumption showed strong inequality – the part resulting from changes in distribution growth, and while average rural consumption also grew, it within urban and rural areas – falling from 60 percent to 56 Dire Dawa National excl. Dire Dawa was not nearly as strong (see Table 1). As a result, the “be- percent (Figure 2). Contribution to change in poverty 20.0 tween” share of inequality – the part of inequality resulting 15.0 10.0 13.5 5.0 2.6 0.0 -6.0 Figure 2 INEQUALITY INCREASED DUE TO THE INCREASING GAP BETWEEN -5.0 -8.6 URBAN AND RURAL AREAS -10.0 -12.8 -15.0 Decomposition of the Gini coefficient into a between rural-urban component and a within-component -20.0 -25.0 -26.4 -30.0 Overlap Within Between Poverty reduction Consumption growth Increase in inequality Relative contribution to Gini (%) 100% 90% 33.4 11.1 Source: HCES, 2011; 2016; World Bank staff calculations. 80% 56.2 70% 60% 60.1 50% Average consumption per quintile in Dire Dawa is sig- than twice as much as the fourth quintile, on average; reflect- 40% nificantly higher than the rest of Ethiopia, except for ing strong consumption growth at the top of the distribution. 30% 32.7 20% the very poorest. Average annual consumption per adult n Although the distribution in Dire Dawa is always significantly 10% 2016 was 6,237 Birr for the poorest quintile (versus 4,850 for above that of the rest the country, the jump from the fourth 0% 6.5 the rest of the country) and 17,916 Birr for the fourth quin- to the fifth quintile in Dire Dawa is particularly large, with con- 2011 2016 tile (versus 13,276 for the rest of the country). Average con- sumption levels in the top Dire Dawa quintile nearly 15,000 Source: HCES, 2011; 2016; World Bank staff calculations sumption shows a sharp increase from the fourth to the fifth Birr per year more than in the rest of Ethiopia (Figure 4). (richest) quintile, with the richest 20 percent consuming more Poverty reduction in Dire Dawa can be entirely at- rest of Ethiopia. The redistribution effect is large and positive tributed to increases in consumption, with the in- (poverty-increasing) for Dire Dawa, reflecting the big jump in HOUSEHOLD CONSUMPTION REMAINS FAIRLY LOW EXCEPT FOR THE Figure 4 crease in inequality significantly hindering poverty the Gini coefficient discussed earlier. If there were no chang- TOP QUINTILE reduction. Figure 3 decomposes the change in poverty into es in the distribution between 2011 and 2016, poverty in the growth and redistribution components using the Datt-Raval- rest of the country would have decreased by 6 percentage Average household expenditures per capita per year, 2016 lion decomposition technique. It indicates which part of the points. In Dire Dawa, changes in inequality had a very large Dire Dawa National excl. Dire Dawa poverty change can be attributed to: (i) the change in aver- impact on poverty reduction. Had there been no change in 45,000 age consumption (growth effect), and (ii) the change in the measured inequality, poverty reduction in Dire Dawa would Average annual consumption 40,000 redistribution of consumption; that is, changes in inequality have been about twice as large as when ended up being 35,000 (redistribution effect). The growth effect on poverty in Dire the case. 30,000 Dawa is negative (poverty-reducing), and is larger than in the 25,000 20,000 15,000 10,000 5,000 0 Poorest 2 3 4 Richest Consumption quintiles Source: HCES 2016; World Bank staff calculations. 194 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 195 Figure 5 MIXED FARMERS HAVE THE HIGHEST POVERTY RATE Figure 6 EDUCATION IS A MAIN CORRELATE OF POVERTY Poverty rates by household’s main income source, 2016 Poverty rates by educational attainment of the household head, 2016 Dire Dawa National excl. Dire Dawa Dire Dawa National excl. Dire Dawa 30% 28% 40% 35% 25% 22% 22% 30% Poverty rate 20% Poverty rate 17% 25% 15% 13% 20% 15% 10% 7% 10% 4% 5% 3% 5% 0% 0% Self - crop and livestock Self - wholesale and retail Salaried employment No education Incomplete Complete primary/ Secondary and primary Incomplete secondary above Main source of household income Education level of household head Source: HCES 2016; World Bank staff calculations. Source: HCES 2016; World Bank staff calculations. Note: Only main income groups with at least 100 observations in Dire Dawa are included. Note: The completed primary and incomplete secondary, and completed secondary and post-secondary categories have been com- bined because of small sample sizes. In Dire Dawa, households that primarily depend on rate of 22 percent, poverty in Dire Dawa is highest among The pattern of poverty reduction in Dire Dawa was urban areas. Changes in rural areas were poverty-increas- crop and livestock production have the highest poverty households headed by someone with no education. Increas- very different than in the rest of Ethiopia, with pov- ing, increasing the overall poverty rate by about 3 percentage rates.102 Those who depend on trade (wholesale and retail) ing education of the household head leads to decreasing erty increasing in the rural woreda of the region be- points – in line with the negative growth at the bottom of the and salaried employment have the lowest poverty rates. For poverty rates, reaching merely three percent for households tween 2011 and 2016. Figure 7 shows that practically all rural distribution (as shown in Figure 2). the rest of the country, poverty is highest for those who de- headed by someone who has at least completed secondary poverty reduction in Dire Dawa was driven by changes in pend on casual labor, and lowest for those who depend on education. Dire Dawa closely mirrors the trend for the rest of salaried employment (Figure 5). the country, although poverty is in general lower in Dire Dawa (Figure 6). Educational attainment of household heads strongly Figure 7 POVERTY REDUCTION IN DIRE DAWA LOOKED VERY DIFFERENT TO negatively correlates with poverty rates. With a poverty POVERTY REDUCTION IN THE REST OF ETHIOPIA Rural-urban decomposition of poverty changes 2011 to 2016 Rural Urban Population shift Interaction Dire Dawa National excl. Dire Dawa 5 Change in poverty headcount 0 -5 -10 -15 -20 Source: HCES, 2011; 2016; World Bank staff calculations. 102 Results are only shown for main income sources that have at least 100 households in the relevant category. 196 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 197 Services was the biggest driver of poverty reduction construction sector was far smaller. The so-called “popula- NON-MONETARY WELFARE in Dire Dawa, which in line with other urban areas, but tion shift” effect had negligible impact on poverty reduction is very different to the rural parts of the country. Im- in Dire Dawa. This variable, often used as an indicator of Improvements in non-monetary welfare accompanied who faced food shortage, the average number of months provements in the services sector account for around 12 structural transformation (proxied by the shift in population the increase in consumption expenditure. In Dire Dawa, with a food shortage did not decrease. The results for the percent of poverty reduction in the rest of Ethiopia, while the from agriculture to manufacturing or services) explained very the percentage of households that reported a food shortage whole country are similar, although food security is lower in corresponding figure for Dire Dawa is far higher at around 80 little of poverty reduction in Dire Dawa over the period. Al- in the last 12 months before the interview decreased from 14 Dire Dawa (Table 4). percent. Progress of households working in agriculture re- though this effect has been shown to be important in oth- percent in 2011 to eight percent in 2016. However, for those gressed, causing increase in regional poverty rate by about er regions, it is relatively less important in Dire Dawa. The 3 percentage points. This is also reflected in the rural/ur- region had a stable sectoral composition over the period. ban data shown in Figure 7. The construction sector also The share of households engaged primarily in agriculture fell Table 4 FOOD SECURITY IMPROVED BETWEEN 2011 AND 2016 drove poverty reduction in the region – accounting for about slightly over the period (30 percent to 29 percent), while the Incidence of food shortages and average duration of food shortages – food gap 1.8 percentage points of the overall drop in poverty. This is share in services increased by one percentage point to 41 different to the rest of the country where the effect of the percent (Figure 8). 2011 2016 Food shortage (%) Food gap (months) Food shortage (%) Food gap (months) Dire Dawa 13.5 1.6 7.7 2 National excl. Dire Dawa 21.6 3.3 10.4 3.2 Figure 8 THE SERVICES SECTOR IS THE LARGEST CONTRIBUTOR TO POVERTY REDUCTION IN DIRE DAWA Source: WMS; 2011, 2016; World Bank staff calculations. Sectoral decomposition of poverty changes 2011 to 2016 Agriculture Manufacturing Construction Services Ownership of household durables and housing quality households with an improved roof, improved wall, improved Other Population shift Interaction also improved in Dire Dawa, although it remains low. floor, access to electricity, and who use electricity for cook- Dire Dawa National excl. Dire Dawa Ownership of household durables such as televisions, mo- ing also improved. But there was no significant improvement 4 Change in poverty headcount bile phones, and refrigerators increased between 2011 and in other housing quality indicators such as the quality of wall 2 0 2016, and the percentage of households owning these du- and floor. The rest of the country also experienced similar -2 rables are above other regions combined. The proportion of improvements (Figure 9). -4 -6 -8 -10 -12 -14 -16 -18 Source: HCES; 2011, 2016. World Bank staff calculations. Photo: World Bank/Chris Terry 198 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 199 Figure 9 OWNERSHIP OF CONSUMER DURABLES AND HOUSING CONDITIONS Figure 10 ACCESS TO IMPROVED WATER SOURCES INCREASED BETWEEN 2011 IMPROVED BETWEEN 2011 AND 2016 AND 2016 Household durables, housing conditions and access to energy in 2011 and 2016, % of households Access to water in 2011 and 2016, % of households with access/characteristics with asset/characteristic Dire Dawa National excl. Dire Dawa A. Household Durables Piped water - compound Piped -outside Protected well/tube well Protected spring 2011 2016 Other improved Unimproved sources 80 100% 100% 70 12.5 10.5 46.5 35.3 4.9 1.1 60 80% 1.4 2.2 80% 50 60% 44.8 40.3 60% 11.8 40 9.1 17.1 30 40% 40% 10.2 41.2 20 35.7 23.2 21.0 20% 20% 10 14.1 0 0% 0% 10.9 Mobile Mobile Refrigerator Refrigerator Animal Cart Motorcycle Animal Cart Motorcycle TV TV 2011 2016 2011 2016 Source: DHS, 2011; 2016; World Bank Staff Calculations Dire Dawa National excl. Dire Dawa Access to improved drinking water sources in Dire 40 to 56 percent, while the percentage of fully immunized Dawa was already high in 2011, and slightly improved children increased from 59 to 76 percent. The stunting rate in 2016. The proportion of household who have access to decreased from 52 percent to 46 percent. Mortality rates did B. Housing/energy improved source of drinking water103 increased from 87 per- not decrease, except for the child mortality rate that dropped 2011 2016 cent to 89 percent. For the rest of the country, the proportion from 29 per thousand children to 21. Education indicators of households with access to improved water sources in- did not improve and primary school completion deteriorated. 100 creased from 53 percent to 65 (Figure 10). Health and education indicators improved in the rest of the 80 country (Figure 11). Children’s immunization and health facility delivery 60 also improved. Delivery at health facility increased from 40 20 0 Improved wall Floor: tiles/cement Access to electricity Electricity for cooking Improved wall Floor: tiles/cement Access to electricity Electricity for cooking Improved roof Improved roof Dire Dawa National excl. Dire Dawa Source: DHS, 2011; 2016; World Bank staff calculations. 103 Improved drinking water sources include piped water, protected well, protected spring, rainwater and bottled water. 200 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 201 INEQUALITY OF OPPORTUNITY Figure 11 CHILDREN’S IMMUNIZATION AND HEALTH FACILITY DELIVERY 2011 2016 To assess inequality of opportunity104 in Dire Dawa and the are low in Ethiopia, but are significantly higher in Dire Dawa. IMPROVED BETWEEN 2011 AND 2016 100 rest of Ethiopia, we focus on five key outcomes: (i) primary This means that the HOI scores for these key outcomes are 2016 2011 in 2011 Selected health and education variables and 2016 80 school enrollment, (ii) primary school completion, (iii) second- always higher for Dire Dawa. They are particularly high com- 100 A. Child and infant mortality rates (per 1000) ary school enrollment, (iv) access to electricity, and (v) access pared to the rest of the country when it comes to access 60 80 2011 2016 to an improved water source. Circumstance variables105 of to electricity and access to an improved water source (see 40 60 100 the household include rural/urban location, gender, religion, the Coverage and HOI panels in Figure 14).107 The D-index agro-ecological zone, consumption quintile of the child’s for these variables – measuring how much these outcomes 20 40 80 household, and the household’s score on the rural accessi- differ according to circumstances – is generally lower than in 0 20 60 bility index (RAI).106 Figure 12 presents the coverage Human the rest of the country. Infant mortality Child mortality Under five Infant mortality Child mortality Under five rate rate mortality rate rate rate mortality rate Opportunity Index (HOI), and Dissimilarity index (D-index) 0 40 The primary school enrollment rate in Dire Dawa is Infant mortality Child across these key outcome variables in 2016. Diremortality Dawa Under five Infant mortality Child National mortality excl. Under five Dire Dawa about 13 percentage points higher than in the rest of 20 rate rate mortality rate rate rate mortality rate In terms of coverage, Dire Dawa is ahead—sometimes Ethiopia. In the region, although 84 percent of age-eligible Dire Dawa National excl. Dire Dawa 0 far ahead—of the rest of Ethiopia on all key outcomes. children were enrolled in primary school in 2016, only 56 per- Infant mortality Child mortality Under five Infant mortality Child mortality Under five rate rate mortality rate rate rate mortality rate Primary school completion and secondary school enrollment cent of age-eligible children had completed primary school. Dire Dawa National excl. Dire Dawa Source: DHS, 2011; 2016; World Bank staff calculations 2011 2016 B. Immunization and health facility delivery (percent) Figure 12 DIRE DAWA DOES BETTER THAN THE REST OF ETHIOPIA OVER ALL KEY 80 70 2011 2016 OUTCOMES 80 60 Coverage, HOI, and D-index for different outcomes, 2016 70 50 60 40 2011 2016 Dire Dawa National excl. Dire Dawa 50 30 80 40 20 70 100 Coverage HOI D-index 30 10 60 90 20 50 80 0 Percent/score 10 40 Health facility Fully Stunted Health facility Fully Stunted 70 0 30 delivery immunized children delivery immunized children 60 Health facility children Fully Stunted Health facility children Fully Stunted 50 20 delivery immunized children delivery immunized children 40 10 Dire Dawa excl. Dire Dawa Nationalchildren children 30 0 20 Health facility Dire Dawa Fully Stunted Fully Dire Dawa National excl. Health facility Stunted immunized children delivery immunized 10 Source: DHS, 2011; delivery 2016; World Bank staff calculations children children children 0 er er er ec ry s ol ec ry s ol ec ry s ol Im y a ol Im y a ol Im y a ol ed s ed s ed s co ry s ol co ry s ol co ry s ol ov ces ov ces ov ces C. Education (percent) o o o at at at o o o o o o Dire Dawa National excl. Dire Dawa nd ch nd ch nd ch ch ch ch im ch im ch im ch w w w c c c pr y s pr y s pr y s 2011 2016 r r r cit cit cit a a a a a a En lete ima En lete ima En lete ima pr pr pr tri tri tri 80 r r r m dp m dp m dp d d d se se se 2011 2016 El El El Co olle Co lle Co lle 70 d d d ro ro lle lle lle r p p p 80 En En En 60 ro ro ro 70 50 60 40 2011 2016 Source: HCES, WMS 2011; 2016; World Bank staff calculations. 50 80 30 40 70 20 30 60 10 20 50 0 Net primary Primary school Gross Net primary Primary school Gross 104 A full description of the methodology and variables used in the construction of this index in Ethiopia can be found in the technical 10 40 enrolment completion secondary enrolment completion secondary appendix. 0 30 Net primary (15-24) Primary school school Gross Net primary (15-24) Primary school school Gross 105  These are generally thought of as variables over which a person does not have control, such as gender or region of birth. 20 enrolment completion enrolment secondary enrolment completion enrolment secondary 106  Accessibility was also tested using a market accessibility variable, and the results were broadly consistent. 10 (15-24) Dire Dawa school Dire Dawa school (15-24) National excl. 107 Unlike in other regions, access to an improved water source is not shown in this chapter as coverage of this variable is 100% ac- 0 enrolment enrolment cording to HICES 2015/2016 data. Source: WMS, 2011; Net2016; World Bank primary staff Primary calculations. school Gross Net primary Dire Dawa Primary school National excl. Dire Dawa Gross enrolment completion secondary enrolment completion secondary (15-24) school (15-24) school 202 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 203 enrolment enrolment These numbers are both much higher than the rest of Ethi- the coverage rate of secondary school enrollment is 37 per- opia, in which just over 71 percent of primary school-aged cent. The HOI score is only slightly lower at 34, indicating Figure 14 SPENDING SHARE ON ECONOMIC SECTOR INCREASED children are enrolled and under one-third complete primary that differences in enrollment rates across different groups is Expenditure share by sectors (%), 2011 to 2016 education. By comparison, more than one in three age-el- not prominent. igible children in Dire Dawa (aged 15 to 18) were enrolled Dire Dawa National excl. Dire Dawa The dissimilarity index, or D-index, shows how much in secondary school in 2016. This indicates that educational inequality in access there is between groups – a high- Administration Economic Social Other opportunity is quite different for the older cohort of children er index means less equal access across groups. The (15-18) than the younger cohort (7-14). 2016 2016 2016 D-index for the three schooling outcomes in Dire Dawa is 21.0 13.9 29.7 The HOI scores for Dire Dawa are, like the coverage far smaller than in the rest of Ethiopia, largely because of 23.3 24.9 scores, generally much higher than the rest of Ethio- coverage in Dire Dawa is significantly better. This indicates 3.4 2016 2016 32.0 14.1 31.7 37.9 pia. A higher HOI score is associated with more equitable that circumstances such as gender and household income 9.6 access to a key service or outcome. For a given coverage matter less in explaining differences in Dire Dawa than they 42.1 2016 23.3 2016 rate, the lower the HOI score, the more unequal the access do in the rest of Ethiopia. 27.8 28.5 2016 to that key outcome is across different groups. In Dire Dawa, 32.8 Note: The inner circles present the 2011 values. Source: MoF; World Bank staff calculations. PUBLIC EXPENDITURE In the previous sections, we discussed the trends in mon- Although average per capita public expenditure in Dire combined, which is consistent with results presented earlier At 23 percent, spending share of other sectors was already etary and non-monetary welfare between 2011 and 2016, Dawa is larger than in the rest of the country, the re- showing that Dire Dawa is better off on most education and large in Dire Dawa in 2011, and further increased to 32 per- and inequality of opportunities focusing on 2016 both in Dire gion spends a significantly lower proportion on eco- health indicators. cent in 2011 (Figure 14). Dawa and the rest of the country. In this section, we discuss nomic sectors. The share of “other” expenditure remained the trends, volume, and allocation of public expenditures to high in Dire Dawa between 2011 and 2016, and spending Per capita public expenditure increased between 2011 For pro-poor sectors, per capita spending109 in Dire identify variation across regions and links to welfare. We use shifted away from social sectors. There was improvement and 2016 and is higher in Dire Dawa than the rest of the Dawa increased for agriculture and rural development MoF data. We express all the expenditure values in 2016 in per capita expenditure on agriculture and rural develop- country. Public spending in Dire Dawa increased from Birr and health sectors. Per capita spending on agriculture and prices using non-food CPI deflator. We first present total per ment and health in Dire Dawa, although spending in these 1,559 per person in 2011 to Birr 2,682 per person in 2016 (in rural development and health sectors increased from 62 Birr capita public expenditure and then discuss the expenditure areas is weak compared to the rest of the country. Per capita constant 2016 Birr), with the lowest per capita expenditure to 83 Birr and from 245 Birr to 276 Birr, respectively, be- allocation by sector. Finally, we present expenditure on se- spending on water and education sectors did not improve of Birr 1,559 recorded in 2011. On average, per capita public tween 2011 and 2016. However, per capita spending on wa- lected “pro-poor” sectors, which include agriculture and rural in Dire Dawa, while they increased in other regions com- expenditure increased slightly faster in the rest of the country ter resources and health showed modest decline. Per capita development, water resources, health, and education. bined. Per capital expenditures in health and education in compared to Dire Dawa, but the gap in absolute terms re- spending on all pro-poor sectors increased in the rest of the 2011 and 2016 were significantly higher than in other regions mains high (Figure 14). country. Dire Dawa spends significantly higher per capita on health and education compared to other regions combined A spending shift towards economic sectors accom- (Figure 16A). panied public expenditure growth in Dire Dawa.108 Figure 13 PER CAPITA PUBLIC EXPENDITURE INCREASED OVER TIME Economic sector spending as a share of total expenditure The total share of spending on pro-poor activities con- increased from ten percent to 14 percent, while spending tracted between 2011 and 2016 in Dire Dawa and the Per capita real public expenditure, 2011 to 2016 on the social sector decreased from 42 to 33 percent. For rest of the country. In Dire Dawa, spending shares on all the rest of the country, the spending share on the economic pro-poor sectors, except agriculture and rural development, Dire Dawa National excl. Dire Dawa sector increased from 23 percent to 29 percent, while the decreased. The decrease in the share of spending was par- 2,800 spending share of the social sector decreased from 32 per- ticularly strong in education, where it dropped from 26 per- 2,400 cent to 28 percent. Administration-related spending share cent to 18 percent. The spending share on education also In 2016 Birr 2,000 decreased both in Dire Dawa and the rest of the country. dropped in the rest of the country (Figure 16B). 1,600 1,200 800 400 0 108 Economic sector includes agriculture and rural development, water resources, trade and industry, mines and energy, transportation 2011 2012 2013 2014 2015 2016 and communication, construction. 109 Per capita spending refers to total spending in a particular sector divided by the population of the region. For example, per capita Source: MoF; World Bank staff calculations. spending on education refers to total spending on education divided by the total population of the region and not spending per student. 204 ETHIOPIA REGIONAL POVERTY REPORT SECTION II: REGIONAL ANALYSIS 205 Figure 15 EXPENDITURE SHARE ON EDUCATION DECREASED IN DIRE DAWA References BETWEEN 2011 AND 2016 Bachewe, F.N. et al. (2015). Agricultural Growth in Ethiopia (2004-2014): Evidence and Drivers. EDRI/IFPRI Ethiopia Real per capita expenditure and share of total expenditure (%) for selected sectors, 2011 and 2016 Strategy Support Program Working Paper 81. A. Real per capita expenditure in 2016 price Central Statistical Agency of Ethiopia (2018). The 2015/16 Ethiopian Household Consumption Expenditure (HCE) Survey. 2011 2016 Results for: Country Level. Statistical Report. Addis Ababa: Central Statistical Agency of Ethiopia. 450 de Barros, R. P., R., Ferreira, R. F. H. G., Molinas Vega, J. R. and Chanduvi, J.S. (2009). Measuring Inequality of 400 Opportunities in Latin America and the Caribbean. Washington, DC: World Bank. 350 In 2016 Birr 300 FDRE. (2012). Ethiopia’s Progress Towards Eradicating Poverty: An Interim Report on Poverty Analysis Study (2010/2011). 250 200 Federal Democratic Republic of Ethiopia: National Planning Commission. 150 100 FDRE. (2017). Ethiopia’s Progress Towards Eradicating Poverty: An Interim Report on 2015/16 Poverty Analysis Study. 50 Ethiopia: National Planning Commission. 0 Agriculture and Rural Development Water Resources Education Health Agriculture and Rural Development Water Resources Education Health Ferreira, F. and Walton, M. (2006). Inequality of opportunity and economic development, in Kochendörfer-Lucius, G. and Pleskovic, B. (eds.) Equity and Development. Washington DC: World Bank. Ravallion, M. and 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. Dire Dawa National excl. Dire Dawa Valderrama, D. and Viveros, M. (2014). Sectoral gains and population shifts. Poverty and inequality measures in practice: A basic reference guide with Stata examples, World Bank Group, Poverty and Equity Global Practice. B. Expenditure share (%) World Bank and DFID (2019). Poverty and Vulnerability in the Ethiopian Lowlands: Building a More Resilient Future. Agriculture & Rural Dev. Water Resources Education Health Washington, D.C.: World Bank. 50 45 13.9 World Bank. (2020). Ethiopia Poverty Assessment: Harnessing Continued Growth for Accelerated Poverty Reduction. 40 8.0 35 Washington, D.C.: World Bank. 7.8 30 25.8 11.5 22.1 25 17.4 20 18.3 15 10 5.5 5.6 5 3.8 2.1 3.5 3.4 6.9 6.8 0 2011 2016 2011 2016 Dire Dawa National excl. Dire Dawa Note: The sum does not add up to 100% because it includes only shares of selected “pro-poor” sectors. Source: MoF; World Bank staff calculations. 206 ETHIOPIA REGIONAL POVERTY REPORT REFERENCES 207 ANNEX A Sectoral decompositions of changes in poverty We use the Ravallion and Huppi (1991) method to at- decompose changes in poverty into an intra-sectoral effect tribute changes in poverty to shifts in intra-sectoral (changes in poverty levels within each group), and a popula- household composition compared to shifts in popu- tion shift effect (changes in the characteristics of the popula- lation. This approach exploits the additive decomposability tion). Decomposing poverty changes in this way allows us to of the standard Foster-Greer-Thorbecke (FGT) measures of determine which of the two factors was more responsible for poverty in order to generate the decomposition. The aim is to driving poverty changes over a given time period. Box A1 Decomposing poverty changes over time The poverty decompositions used in this chapter are based on Ravallion and Huppi (1991).110 The decompositions reflect changes in the poverty headcount rate, but extend naturally to the poverty gap and poverty gap squared mea- sures. Poverty at time t is given as Pt . The change in poverty between t and t+1 is composed of the following effects: n sit ( Pi,t+1 – Pi,t ) Pt+1 – Pt = ∑ i=1 Intra-sectoral effect n Pit (si,t+1 – si,t ) +∑ i=1 Population shift effect +∑ i=1 n (Pi,t+1 – Pi,t )(si,t+1 – si,t ) Interaction effect For the purposes of this chapter, i represents the specific sector, and n is the number of sectors. These include: urban versus rural; the five main sectors of occupation; self-employed versus not self-employed. Pi,t is the poverty rate of group i in period t. Finally, si,t is the population share of group i in period t. An illustrative example, adapted from Valderrama and Viveros (2014) may be useful in explaining the counterfactual assumptions underlying this decomposition. Consider that in time period 1 there are three sectors – agriculture, manufacturing and services – in which a household head could be employed. Each of these sectors has a different poverty rate associated with it, with the highest poverty rate in the agriculture sector and the lowest poverty rate in the services sector. Consider a simple change in which a group of households shifts from the agriculture sector to the services sector. If, after the shift, the within-sector poverty rates remain the same, then the national poverty rate must have decreased. This is a pure population shift effect – the drop in national poverty was driven entirely by pop- ulation shift from agriculture to services. Consider now a situation in which the poverty rate in the agriculture sector decreased but no households change sectors.111 Again, the national poverty rate would decrease. In this situation, however, the decrease would have been driven entirely by the intra-sectoral effect. In practice, as shown in this chapter, the overall poverty change will be a combination of both the population shift effect and the intra-sectoral effect, along with an interaction effect to balance out the accounting exercise. Photo: World Bank/Chris Terry Poverty is defined at the household level, and for this reason the chapter uses the occupational characteristics of the household 110  head as the characteristic that defines the entire household. Finally, the caveat should be added that the decomposition is purely a statistical exercise that should be used to understand past changes, rather than a tool that should be used to estimate future trends in poverty. 111 This could also be the case if the poverty rate decreases in one sector but there is no net mobility across sectors. 208 ETHIOPIA REGIONAL POVERTY REPORT ANNEX A: SECTORAL DECOMPOSITIONS OF CHANGES IN POVERTY 209 ANNEX B Constructing the Human Opportunity Index for Ethiopia The Human Opportunity Index (HOI) is widely used to The extent of inequality of opportunity is measured us- measure inequality of opportunity. The HOI captures ing the D-index. This index calculates how much access to both (i) the overall access to basic services, such as edu- services varies by birth characteristics, such as socio-eco- cation, water and electricity, (the coverage rate); and (ii) in- nomic status and the location of a household. A D-index of equality in access (de Barros, Ferreira et al. (2009)). If access zero indicates perfect equality (no gaps in access to services to a basic service is perfectly equal, then the HOI is the same across circumstance groups), whereas a D-index of “one” in- as the coverage rate. As access becomes more unequal, dicates perfect inequality. More information on the construc- the HOI becomes lower. tion of the HOI and the D-index is provided in Box 2. Box B1 Constructing the HOI and the D-Index  The central question behind the HOI is to what extent circumstances beyond one’s control influence access to a set of important basic services. Simply put, the HOI takes the coverage level of a basic service or “opportunity” (for example whether a child is enrolled in primary education) and combines this with the extent to which that opportunity is determined to be beyond the control of the child (for example being born in a rural rather than urban area, or being a girl rather than a boy). Ideally, random circumstances should play no role in determining access to opportunities. The D-index measures dissimilar access rates to a given basic opportunity for groups of children where groups are defined by circumstance characteristics (for example, area of residence, or gender) compared to the average access rate to the same service for the population as a whole. The sample is stratified into groups, or “cells,” so that all individuals in any given cell have the same combination of circumstances. The resulting subgroups are known as “types” (de Barros, Ferreira et al. (2009)). These cells are then compared to one another. The difference in outcomes between cells can be attributed to inequality of opportunity, while the differences within cells can be considered the result of effort or luck. The D-index summarizes all gaps into a single measure by weighting them according to the population share in each circumstance group. The D-index generates a value between 0 and 1. In a society in which there is no inequality of access, the D-index is be zero. If average access is denoted by p, the specific access rate of group i is pi, and the share of group i in the population is given by ßi then the D-index is: n ∑ 1 D= ßi | pi – p| 2p i =1 The HOI can then be calculated as: HOI = p (1 – D) Photo: World Bank/Chris Terry The measure is also decomposable so that the extent to which specific opportunity sets contribute to the dis- similarity can be assessed. This means that the contribution of different circumstances to overall inequality of opportunity can be determined. 210 ETHIOPIA REGIONAL POVERTY REPORT ANNEX B: CONSTRUCTING THE HUMAN OPPORTUNITY INDEX FOR ETHIOPIA 211