The Uganda Poverty Assessment Report 2016 Farms, cities and good fortune: Assessing poverty reduction in Uganda from 2006 to 2013 Report No. ACS18391 a The Uganda Poverty Assessment Report 2016 Farms, cities and good fortune : assessing poverty reduction in Uganda from 2006 to 2013 SEPTEMBER 2016 i 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 Group concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this work is subject to copyright. Because The World Bank Group encourages dissemination of its knowledge, this work may be reproduced, in whole or in part, for noncommercial purposes as long as full attribution to this work is given. Any queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2422; e-mail: pubrights@worldbank.org. Photography: Clarence Tsimpo, Great Lakes Film Production Ltd and Digital Media Network Ltd Cover design : Artfield Graphics Ltd, www.info@artifield.com Design and Layout: Artfield Graphics Ltd. Printed in Uganda by Artfield Graphics Ltd Additional material relating to this report can be found on the World Bank Uganda website (www.worldbank.org/ uganda). The material includes a fact sheet. © 2016 International Bank for Reconstruction and Development / International Development Association or The World Bank Group 1818 H Street NW Washington DC 20433 Telephone: 202-473-1000 www.worldbankgroup.org CONTENTS ACKNOWLEDGMENTS vii | ABBREVIATIONS AND ACRONYMS ix | EXECUTIVE SUMMARY x OVERVIEW xii | INTRODUCTION xxx CHAPTER 1: UGANDA’S PROGRESS IN REDUCING POVERTY 1.1. Recent progress in poverty reduction 3 | 1.2 Is the national poverty line a good measure of poverty in Uganda? 7 | 1.3 The incidence of progress and shared prosperity 11 | 1.4 Who are the poor in Uganda? 18 1.5. Conclusion and outlook: Ending extreme poverty in Uganda 21 CHAPTER 2: NON-MONETARY DIMENSIONS OF POVERTY IN UGANDA 2.1 Housing conditions 27 | 2.2 Infrastructure services 28 | 2.3 Physical capital 32 | 2.4 Human capital 33 2.5 Conclusion 42 CHAPTER 3: HOW DID UGANDA REDUCE POVERTY? 3.1 Growth, not redistribution, drives poverty reduction in Uganda 45 | 3.2 Demographic change, urbanization, and poverty reduction 49 | 3.3 Agricultural growth has been particularly important for poverty reduction 52 | 3.4 Human capital, access to infrastructure, and poverty reduction 58 | 3.5 Conclusion 61 CHAPTER 4: AGRICULTURAL GROWTH AND POVERTY REDUCTION IN UGANDA 4.1 Agriculture and poverty in Uganda 63 | 4.2 Weather, prices, peace, and consumption growth 75 4.3 Increasing the resilience of Ugandan households 80 | 4.4 Conclusion 81 CHAPTER 5: NON-AGRICULTURAL GROWTH IN UGANDA 5.1 Characteristics of households that have experienced non-agricultural income growth 87 5.2 Identifying constraints to non-agricultural income growth 89 | 5.3 Conclusion 95 CHAPTER 6: MOVING OUT AND UP: MIGRATION AND POVERTY IN UGANDA 6.1 The impact of migration on poverty reduction 98 | 6.2 Who migrates? 101 | 6.3 What aids and constrains migration? 104 | 6.4 Conclusion 109 CHAPTER 7: EDUCATION AND HEALTH SERVICES: QUALITY OF INPUTS, USER SATISFACTION, AND COMMUNITY WELFARE LEVELS 7.1 Quality of inputs at the school level 111 | 7.2 Quality of inputs at the health center level 113 7.3 Knowledge and behavior of teachers 116 |7.4 Knowledge and behavior of health workers 117 7.5 Outcomes at the school level118 | 7.6 User satisfaction with facilities 120 | 7.7 Conclusion 122 iii REFERENCES 123 ANNEXES Annex 1: Exploring Patterns of Food and Non-Food Consumption Over Time, Methodology 130 Annex 2: Additional Tables on Service Delivery and Welfare 132 LIST OF FIGURES Figure 1: Headcount poverty rate, national and international poverty line, 1993 to 2013 xiii Figure 2: Annual reduction of poverty headcount at international poverty line, selected Sub-Saharan Africa countries (2003–2013) xiv Figure 3: Rising inequality: the Gini coefficient from 1993–2013 xv Figure 4: Inequality is increasing, but remains moderate compared to the region (percent, latest survey year) xv Figure 5: Share of poor population in each region, 2006–2013 xvi Figure 6: Sectoral contribution to poverty reduction, 2006 to 2013, main source of income xviii Figure 7: Real income per capita by source of income, bottom 40 percent xviii Figure 8: Share of crop income derived from crop sales, bottom 40 percent, 2006–2012 xx Figure 9: Locational contribution to poverty reduction xx Figure 10: Contribution of education to consumption growth xxii Figure 11: Higher dependency ratios held back consumption growth, especially for the poorest xxii Figure 12: Self-reported coping mechanism xxiv Figure 13: Primary completion rate is among the lowest in the world xxv Figure 14: Inputs and user satisfaction by welfare quintiles in education sector xxvii Figure 1.1: Two decades of progress in reducing poverty 5 Figure 1.2: Structure of non-food spending over time 9 Figure 1.3: Prices of food items, 1993–2013 10 Figure 1.4: Structure of food spending over time 10 Figure 1.5: The incidence of consumption growth, 1993 to 2013 13 Figure 1.6: GDP per capita growth, 1993 to 2013 14 Figure 1.7: Coffee and maize prices, 1993 to 2013 14 Figure 1.8: Decomposing poverty reduction into growth and redistribution 15 Figure 1.9: Inequality in Uganda 17 Figure 1.10: Where do the poor live? 18 Figure 1.11: Trends in poverty incidence 23 Figure 1.12: Trends in poverty incidence for different regional growth rates 23 Figure 2.1: Distribution of households by main type of construction materials (%), 2006–2013 28 Figure 2.2: Access to improved water source vs. GNI per capita 29 Figure 2.3: Percentage of households using an improved latrine 30 Figure 2.4: Access to electricity (% of population) 31 Figure 2.5: Changes in asset ownership, by consumption quintile, 2006–2013 (absolute numbers) 33 Figure 2.6A: Trends in adult literacy rates (%) 34 Figure 2.6B: Literacy gap across cohorts (%) 34 iv Figure 2.7: Net enrollment and primary completion rates 36 Figure 2.8: Out-of-school rate for lower secondary and reason for dropping out of school 37 Figure 2.9: Trends in childhood mortality, 2001–2011 38 Figure 2.10: Under-five mortality by region and maternal mortality rates 39 Figure 2.11: Maternal and under-five mortality rates in Uganda and international comparison 39 Figure 2.12: Malnutrition prevalence, underweight (% of children under five years) versus income per capita 41 Figure 2.13: Nutritional status of children under 5 years, by region and consumption quintile, 2011 (in percent) 41 Figure 3.1: Limited public transfers for Ugandan households 45 Figure 3.2: Informal transfers are a prevalent, but not important, source of income 46 Figure 3.3: Using counterfactuals to quantify changes that have been important to poverty reduction 48 Figure 3.4: Demographic change and poverty reduction, 2006–2013 51 Figure 3.5: Household labor income and poverty reduction, 2006–2013 53 Figure 3.6: Sectoral growth rates 56 Figure 3.7: Educational attainment is associated with consumption growth, except for the wealthiest households 59 Figure 3.8: Increased access to electricity and piped water is associated with consumption growth 60 Figure 4.1: Structure of agricultural income by region, 2012 64 Figure 4.2: Share of crop income derived from crop sales, bottom 40 percent, 2006–2012 65 Figure 4.3: Price, conflict, and weather trends from 2005/06 to 2011/12 71 Figure 4.4: Regional differences in per capita crop income growth, 2005/6 to 2011/12 74 Figure 4.5: Self-reported coping mechanisms 81 Figure 4.6: Education mitigates the impact of climate shocks 81 Figure 5.1: Sectoral growth in comparison to the region 86 Figure 5.2: Type of employment and education 89 Figure 5.3: Average monthly and hourly wages by the level of education 90 Figure 5.4: Access to finance 93 Figure 5.5: Distance to nearest population center with +20,000 (km) 94 Figure 5.6: Nonfarm self-employment productivity and distance 95 Figure 6.1: Consumption of migrants and non-migrants, before (2006) and after (2010) migration 99 Figure 6.2: Consumption of rural and urban migrants, before (2006) and after (2010) migration 99 Figure 6.3: Age distribution of migrants relative to stayers 103 Figure 6.4: Education of household head, households that send migrants relative to those that do not 104 Figure 7.1: Inputs for primary schools by welfare and subregion 113 Figure 7.2: Inputs for health facilities by welfare and subregion 115 Figure 7.3: Primary school teachers’ assessment by welfare quintiles 116 v Figure 7.4: Share of health workers giving the correct diagnostic (5 tracer conditions) 118 Figure 7.5: Share of health workers giving the correct diagnostic for post-partum hemorrhage and neonatal asphyxia 118 Figure 7.6: Pupil assessment (score) 119 Figure 7.7: Inputs and user satisfaction by welfare quintiles in education sector 121 LIST OF TABLES Table 1: Impact of weather on diversification xxi Table 1.1: Poverty from 1993 to 2013, national and international line 4 Table 1.2: Reductions in the depth and severity of poverty at the national line, 1993 to 2013 5 Table 1.3: National poverty rates by region 5 Table 1.4: Spending on non-food items among poor households, 1993 and 2013 9 Table 1.5: Shared prosperity, 1993–2013 14 Table 1.6: Human capital, asset ownership, and access to infrastructure across regions 19 Table 1.7: Fertility rates and dependency ratios, 2006–13 20 Table 1.8: Human and physical capital and livelihoods among the bottom 40 percent, 2013 21 Table 1.9: Poverty statistics in 2030 24 Table 2.1: Trends in net enrollment rates in primary schools (%) 35 Table 2.2: Perceptions about poverty (%) 42 Table 3.1: Structure of household income, 1993 to 2013 54 Table 3.2: Agricultural GDP growth rates for selected Eastern African countries, 2000–2012 55 Table 3.3: Real per capita Income growth by source of income, 2006 to 2012 57 Table 3.4: Relationship between income and consumption, 2006-2010 57 Table 3.5: Maximum level of education attended by a household member, 2006–2013 59 Table 4.1: Agricultural income, 2012 64 Table 4.2: The nature of crop income, 2012 65 Table 4.3: Household characteristics, by wave 66 Table 4.4: Drivers of agricultural income growth 73 Table 4.5: Changes in agricultural income: a regional story 75 Table 4.6: Impact of weather, prices, and peace on income and consumption 77 Table 4.7: Impact of weather, prices, and peace on income and consumption: Bottom 40 percent 78 vi Table 4.8: Welfare changes: A regional story 78 Table 4.9: Impact of weather, prices, and peace on weight for age and weight for height 79 Table 5.1: Source of household income by sector 86 Table 5.2: Moving in and out of non-agricultural employment 87 Table 5.3: Characteristics of households in 2011 that experienced non-agricultural growth from 2011 to 2012 88 Table 5.4: Determinants of non-agricultural household income 91 Table 6.1: Share of households which sent a work migrant, by region, location, and year 97 Table 6.2: Impact of migration 100 Table 6.3: Characteristics of households that send working migrants 102 Table 6.4: Characteristics of individuals who migrate before migration 103 Table 6.5: Correlates of a household’s decision to send a migrant 108 Table 6.6: Correlates of migration at the individual level 108 Table 6.7: Shocks, distance, and the probability of migration of 15–24-year-olds 108 Table A1.1: Regression of food share 131 Table A2.1: Primary schools—resources and management 132 Table A2.2: Health facilities—caseload, workers, management, and drug availability 134 Table A2.3: Health facilities—infrastructure and equipment 136 Table A2.4: Assessment of teacher knowledge and teaching quality 138 Table A2.5: Assessment of health worker knowledge 139 Table A2.6: Assessment of student performance 140 Table A2.7: Correlates of pupil achievement (Probit model) 141 LIST OF BOXES Box 1.1: How poverty is measured in Uganda 11 Box 1.2: Inequality Measures 16 Box 1.3: Spatial Dimensions of Poverty 19 Box 3.1: What does decomposing changes in poverty entail? 47 Box 3.2: Expanding fiscal policy: How can oil revenues accelerate poverty reduction in Uganda? 48 Box 3.3: Agricultural growth in national accounts and survey data 55 vii ACKNOWLEDGMENTS The World Bank greatly appreciates the close collaboration with the Government of Uganda (the Ministry of Finance, Planning and Economic Development [MoFPED] and the Uganda Bureau of Statistics [UBOS], in particular) in the preparation of this report. The team preparing this report was led by Ruth Hill (Senior Economist, GPV01) and Clarence Tsimpo Nkengne (Senior Economist, GPV01) and consisted of Habtamu Fuje (Chapter 1), Alvin Etang (Chapters 2 and 7), Carolina Mejia-Mantilla (Chapter 4), Siobhan Murray (Chapter 4), Johanna Fajarado Gonzalez (Chapters 4 and 5), Tomomi Tanaka (Chapter 5), Michael O’Sullivan (Chapter 6), and Edouard Mensah (Chapter 6). The list of background papers, authors, and reviewers that formed the basis of Chapters 2 to 7 is provided below. This report was prepared under the guidance of Diarietou Gaye (Country Director, AFCE2), Christina Malmberg-Calvo (Country Manager, AFMUG), Thomas O’Brien (Country Program Coordinator, AFCE2), Apurva Sanghi (Program Leader, AFCE2), Johan A. Mistiaen (Program Leader, AFCE2), and Pablo Fajnzylber (Practice Manager, GPV01). Markus Goldstein (Lead Economist, AFRCE), Kathleen Beegle (Program Leaders, AFCW1) and Obert Pimhidzai (Economist, GPV02) were peer reviewers for the report. In addition the team is grateful for comments and input from Maura Leary, Jeeyeon Seo, Jean-Pascal Nganou, Rachel Sebudde, Sheila Kulubya, Franklin Mutahakana, Dr. James Muwonge (UBOS), Vincent Ssennono (UBOS), Albert Musisi (MoFED), Joseph Enyimu (MoFED), Dr. Jean-Louis Kasekende (BoU), Dr. Sarah Ssewanyana (EPRC), Dr. Ibrahim Kasirye (EPRC) and seminar participants at seminars held at the Economic Policy Research Center (EPRC), Makerere University, and United Nations Children’s Emergency Fund (UNICEF), Kampala. The team also thanks Martin Buchara, Senait K. Yifru, Damalie Nyanja, Gladys Alupo, Barbara Katusabe for assistance provided during the preparation of this report. In addition, the report benefited from proofreading by Shlamraj Ramraj and Akashee Medhi. Financial support from the Global Facility for Disaster Reduction and Recovery, and the Belgium Trust Fund is gratefully acknowledged. Background papers “Beyond Income Poverty: Non-Monetary Dimensions of Poverty in Uganda” by Alvin Etang and Clarence Tsimpo and reviewed by David Newhouse and Carolina Mejia-Mantilla. “Welfare, Income Growth and Shocks in Uganda” by Ruth Hill and Carolina Mejia-Mantilla and reviewed by Johannes Hoogeveen and Luc Christiaensen. Funding was also provided by GFDRR for the analysis undertaken in this paper. “Moving Out and Up: Panel Data Evidence on Migration and Poverty in Uganda” by Edouard Mensah and Michael O’Sullivan and reviewed by Joao Montalvao and Francis Mwesigye (Makerere University and EPRC, Uganda). “Education and Health Services in Uganda: Quality of Inputs, User Satisfaction, and Community Welfare Levels” by Clarence Tsimpo, Alvin Etang, and Quentin Wodon and reviewed by Andrew Dabalen and Christophe Rockmore. viii ABBREVIATIONS AND ACRONYMS ACLED Armed Conflict Location and Event Data CAPI Computer Assisted Personal Interviews CEM Country Economic Memorandum CPI Consumer Price Index EPRC Economic Policy Research Center GDP Gross Domestic Product GNI Gross National Income GoU Government of Uganda ICT Information and Communication Technology MoFPED Ministry of Finance, Planning, and Economic Development NMS National Medical Store PPP Purchasing Power Parity PTA Parents and Teachers Association PTB Pulmonary tuberculosis RIF Recentered Influence Functions RIGA Rural Income-Generating Activities SDI Service Delivery Indicators SMC School Management Committee UBOS Uganda Bureau of Statistics UDHS Uganda Demographic and Health Survey UNICEF United Nations Children’s Emergency Fund UNHS Uganda National Household Survey UNPS Uganda National Panel Survey UPE Universal Primary Education USE Universal Secondary Education WDI World Development Indicators WRSI Water Requirement Satisfaction Index Vice President : Makhtar Diop Senior Director : Ana Revenga Country Director : Diarietou Gaye Country Manager : Christina Malmberg-Calvo Practice Manager : Pablo Fajnzylber Task Team Leaders : Ruth Hill and Clarence Tsimpo Nkengne ix Uganda has set out an ambitious agenda for its future; its 2040 Vision foresees a middle-income EXECUTIVE country with the majority of its citizens living in SUMMARY urban areas, having smaller families, and earning income in non-agricultural sectors. Uganda’s progress in reducing poverty over the last And, of more concern, it is not clear that the processes two decades is a remarkable story of success. From that brought about gains in the past will be enough 1993 to 2006, annual reduction in the national poverty to address the future poverty challenge in Uganda, rate of 1.9 percentage points a year resulted from the particularly in the impoverished Northern and Eastern restoration of peace and stability to much of the country regions. after Yoweri Museveni came into power, the series of Acknowledging both the impressive progress and economic liberalization reforms that were implemented, its limitations, however, it is helpful to look at the and the investments of households and firms that these factors that contributed to Uganda’s poverty reduction encouraged. since 2006 and to examine policies that have worked Poverty reduction has remained impressive since 2006— alongside possible improvements to make progress the period of focus for this report—even though it has more sustainable into the future. fallen more slowly. The national poverty rate fell by 1.6 Much of Uganda’s poverty reduction was built on percentage points per year and the international extreme agricultural income growth that particularly benefited poverty rate fell by 2.7 percentage points per year, the poor households. Peace in northern Uganda, improved second fastest reduction in extreme poverty per year in regional crop markets, and good weather drove growth Sub-Saharan Africa during this time. in agricultural incomes. Modest gains in education also Uganda’s poverty reduction since 2006 has coincided contributed to growth, as did urbanization. with a period of slower economic growth. Despite Uganda’s formula for success is one that works especially this, poor households still experienced consumption well when conditions are favorable, particularly in growth and poverty fell. Understanding why this was agriculture. And luck has been on Uganda’s side in and whether it is sustainable offers lessons for other the last decade. Good rainfall and prices can account countries grappling with how to ensure that the poor can for two-thirds of the growth in crop income of the still see improvements in their lives, even in the face of a bottom 40 percent from 2006 to 2012. Prices reflect not slowing global economy. just improvements in marketing efficiency resulting However, Uganda’s success is not without caveats. from market liberalization, but also many factors In 2013, more than a third of its citizens lived below beyond domestic policy control: positive price trends the international extreme poverty line of US$1.90 a in international markets and increased demand for day. What’s more, the low national poverty rate of Ugandan crop exports in regional markets as a result of 19.7 percent is based on a poverty line that was set peace in South Sudan and the Democratic Republic of over twenty years ago and is now too low, and not Congo. reflective of a reality in which too many Ugandans live There was little fundamental change in how the today. Vulnerability has also remained high. For every households earned their income that benefited poverty three Ugandans that moved out of poverty, two fell reduction—either in agriculture or in other sectors. into poverty. Poverty has also become increasingly Most households continue to earn income in informal, concentrated in the Northern and Eastern regions of the low-investment, low-productivity activities such as country. x traditional crop farming and small-scale retail trading, delivery continue. Teacher absenteeism keeps students and there has been little change in the proportion of from learning and achieving, and teachers and health households that count agriculture as their main sector workers often lack the minimum knowledge to properly of employment since 2006. In addition, persistently high teach pupils or treat patients. As a consequence, children fertility rates held back poverty reduction. A quarter may go to school, but not master the knowledge that of Uganda’s households are female headed and these they need to be successful in the labor market. Similarly, households experienced lower productivity largely public and private spending on health access does not because of the higher time-burden of childcare that they guarantee that people are receiving quality service. face. Limited spending on safety nets also resulted in All these can have a negative impact on people’s skill fiscal redistribution having little direct impact on poverty attainment and health, even more so for the poor as reduction. they experience the lowest quality of services. Improving community-based monitoring and demand-side Uganda has set out an ambitious agenda for its future; accountability is an important part of the solution, but its 2040 Vision foresees a middle-income country with more than this will be needed. Poorer communities are the majority of its citizens living in urban areas, having more likely to express satisfaction with any services that smaller families, and earning income in non-agricultural they are receiving, even though their quality is worse sectors. Sustained gains in poverty reduction and the than in better-off communities. achievement of this vision for Uganda will require a fundamental shift in the nature of production—from low- Liberalization of markets has been important to investment, informal activities to higher-capital, more Uganda’s success in the past, but some markets are productive employment and a more rapid reduction in currently failing to work. The low quality of agricultural fertility rates. inputs in domestic markets results in poorer quality outputs and lower earnings. If authentic technologies To make this happen, effective public investment replaced these low-quality products, average returns in services such as education, health, agricultural for smallholder farmers would be over 50 percent. extension, and safety nets will be crucial. Structural Increasing the adoption of more modern technologies change undoubtedly also requires a focus on firm will entail improving the quality of inputs in local markets growth and job creation, but for this growth to be through certification (public or private). Improvements inclusive of the poorest households, it must be paired in rural financial markets are also needed to increase the with investments in education, skills, and finance, access to financial capital that is required for agricultural especially for vulnerable groups such as adolescent input purchases, nonfarm employment growth, and rural girls. The significant increase in primary enrollment to urban migration. This will be imperative to ensure rates brought about by the benefits of the Universal consumption growth for poor households regardless of Primary Education (UPE) program has yet to translate weather variations and regional and international prices. into substantial improvements in educational outcomes. Primary completion rates were merely 53 percent in Although there is an important role for the state in 2011, much lower than countries with similar income bringing about the change Uganda needs to see, the levels. Pregnancy is the fourth most common reason for continued importance of security and liberalized markets dropping out of secondary school: in 2013, 1 in 10 girls cannot be underestimated. Ensuring continued stability report dropping out of secondary school as a result of in the region and further promoting efficient crop pregnancy. Public transfers to households are negligible markets and regional exports will be important for future in Uganda—total spending on direct income support income growth in Uganda. This growth, when paired with to poor households was 0.4 percent of gross domestic an inclusive policy framework and stronger investments product (GDP) compared with 1.1 percent in other low- in basic services, can lead to more sustainable poverty income countries in Africa. reduction and improvements in the quality of life of millions of Ugandans. But for these public investments to be effective, Uganda cannot let implementation gaps and poor service xi 1. Uganda’s progress in reducing poverty from 1993 to 2006 is a remarkable story of success that has been well told. Annual reductions in the national poverty rate of 1.9 percentage points a year resulted from the restoration of peace and stability to much of the country after Yoweri Museveni came into power in 1986, the series of economic liberalization reforms that were implemented in the 1990s, and OVERVIEW the investments of households and firms that these encouraged (see for example, Collier and Reinikka 2003; World Bank 2007). 2. The narrative of Uganda’s continued, albeit slightly slower, progress in reducing poverty since 2006 is less familiar. Uganda reduced the proportion of people living on less than US$1.90 per person per day by 2.7 percentage points per year, the second fastest percentage point reduction in extreme poverty per year in Sub-Saharan Africa during this period.1 The national poverty rate continued to fall by 1.6 percentage points per year. However, during this time the national poverty line, set using data from 1993, became an increasingly poor standard against which to measure who was poor. 3. This was a period in which growth slowed as the gains from reforms years earlier had been fully realized, and weak infrastructure and increasing corruption increasingly constrained private sector competitiveness (World Bank 2015). How, in this context, was Uganda still able to secure inclusive consumption growth for many of its citizens? Understanding the drivers of recent poverty reduction is important for offering lessons on how to reduce poverty further in the future not only in Uganda, but also for other countries in the region that have not experienced such progress. 4. This report examines Uganda’s progress in reducing poverty, with a specific focus on the period 2006 to 2013. The report shows that high growth from 2006 to 2010 benefited poverty reduction. Although growth slowed for all households from 2010, poor households were able to maintain above average consumption growth and poverty reduction did not falter. Agricultural income growth particularly benefited poor households aided by peace in northern Uganda, improved regional markets, and good weather. Modest gains in education also appear to have contributed to the growth for poor households, as did urbanization. However there was little fundamental change in the nature of production that benefited poverty reduction—either in agriculture or in other sectors. In addition, persistently high dependency ratios held back poverty reduction, and limited spending on safety nets resulted in fiscal policy 1. Uganda reduced the extreme poverty rate by 2.7 percentage points a year, second only to Chad, which reduced the extreme poverty rate by 3.1 percentage points per year. This is using poverty numbers reported in Povcalnet as of January 2016 and using the surveys deemed comparable by World Bank 2016. xii contributing to neither poverty reduction nor to progress in well-being since 1993. The UBOS has improving vulnerability. conducted high-quality household surveys every three to four years that have provided a comparable 5. Is Uganda on a path to end extreme poverty? series of data on poverty and other household The benefits of security and liberalized markets characteristics for the last twenty years. Uganda for poverty reduction cannot be underestimated is one of the few countries in the region to have and will likely aid future poverty reduction as they achieved this level of comparable, frequent poverty have done in the past. However, sustained gains monitoring over time. Without this, it would not be also require a fundamental shift in the nature of possible to document the lessons Uganda provides. production from low-investment, informal activities to higher-capital, more productive employment. A RECORD OF PROGRESS This in turn requires effective public investment in services (such as education, health, rural finance, 7. Uganda recorded impressive rates of poverty quality of agricultural inputs and extension reduction in the last two decades. The proportion services), infrastructure (such as regional corridors of the Ugandan population living in poverty— and electricity), and safety nets. Addressing whether measured using the national poverty line this requires addressing public investment or the international poverty line—more than halved implementation gaps and improving service from 1993 to 2013 (Figure 1). The proportion of the delivery. population living below the national poverty line declined from 56.4 percent in 1993 to 19.7 percent 6. Before turning in further detail to the key in 2013.2 The proportion of households living below findings of the report, it is important to note the international extreme poverty line of US$1.90 that the analysis undertaken in this report a day (2011 prices) fell from 68.1 percent in 1993 is only possible because the Government of to 34.6 percent in 2013. The depth and severity of Uganda (GoU) has invested in a high quality poverty have also fallen consistently. series of household surveys to document FIGURE 1: Headcount poverty rate, national and international poverty line, 1993 to 2013 Source: Uganda National Household Survey (UNHS) 2006–2013. Most districts have difficulties in accessing basic services such as safe water 2. The national poverty line ranges from US$0.88 to US$1.04 2005 PPP per capita depending on the region. Poverty in Uganda is calculated using a cost-of-basic-needs approach. Consumption expenditure data is collected on food and non-food items through the UNHS conducted every three to four years. The poverty line was set in 1993 by calculating the cost of consuming 3,000 calories per adult equivalent and then adding an amount (the amount depending on the region) to capture non-food expenditures. The poverty line has only been updated for the cost of inflation since then and is low by international standards. xiii 8. Progress in the period of focus for this report, due to the fact Uganda started with a high poverty 2006 to 2013, has been a little slower but still rate. However, even considering the percentage very fast by regional standards. The international reduction in poverty, Uganda’s performance has extreme poverty rate—the proportion of been impressive—the fifth fastest in the region households living on less than US$1.90 purchasing during this time. The national poverty rate fell by power parity (PPP) per day—fell by 2.7 percentage 1.6 percentage points a year during this period, points per year since 2003. Although this was slower only slightly slower than the 1.9 percentage point than the rate of progress in earlier years, it was reduction recorded from 1993 to 2006. However, still the second fastest percentage point reduction the national poverty line has not been updated in poverty per year in Sub-Saharan Africa (Figure since 1993, causing this to become an increasingly 2). The high percentage point reduction is in part poor measure of who is poor in Uganda today. FIGURE 2: Annual reduction of poverty headcount at international poverty line, selected Sub-Saharan Africa countries (2003-2013) Source: Staff calculations using Povcalnet. 9. Recent gains in poverty reduction have occurred has been marginal and inequality fell from during a period in which growth started to slow. 2010 to 2013. Inequality increased in rural and Although growth slowed for all households, poor urban Uganda from 1993 to 2010, by any measure. households still experienced consumption growth National inequality, as measured by the Gini and poverty fell. Peace in northern Uganda and index, increased from 0.36 in 1993 to 0.42 in 2010 agricultural income growth aided consumption (Figure 3). This finding holds when looking at other growth for poorer rural households, even though measures of inequality such as the Theil index with better-off urban households did not fare as well. the parameter α=−1 which emphasizes inequality As a result, the period from 2010 to 2013 was for lower incomes, and the absolute and relative the only period in the last twenty years in which difference between the bottom 10 percent and the consumption growth was higher among the bottom top 90 percent. However, the increase has been 40 percent (2.3 percent per year) than among the marginal and Uganda has a moderately low rate top 60 percent (1.6 percent per year). of inequality compared to other countries in the region (Figure 4). Inequality fell from 41.5 percent 10. In general, growth brought rising inequality in 2010 to 38.5 percent in 2013, a reduction of 1 as well as rising consumption but the increase percentage point in the Gini per year. xiv 11. Trends in non-monetary well-being also point For example, between 2001 and 2011, under-five to improvements in the well-being of Ugandan mortality dropped by 5.2 annually for Sub-Saharan households. Infant mortality dropped from 88 African countries and by only 2.4 for the world. in 2001 to 76 in 2006 and 54 in 2011. Under-five Education outcomes have also improved over mortality stood at 90 in 2011, having declined from time, for example the primary net enrollment rate 152 in 2001 to 137 in 2006.3 Between 2001 and increased from 84 percent in 2006 to 86 percent in 2011, under-five mortality dropped by 5.6 annually 2013. In addition, ownership of modern assets such in Uganda. This was a considerable improvement as telephones and motorcycles increased, while in comparison to regional and global averages. ownership of traditional assets, such as bicycles, fell. FIGURE 3: Rising inequality: the Gini coefficient from 1993–2013 Source: UNHS 1993–2013. FIGURE 4: Inequality is increasing, but remains moderate compared to the region (percent, latest survey year) Source: World Development Indicators (WDI). BUT MANY CHALLENGES REMAIN 12. However, despite the substantial progress that 13. The low national poverty rate of 19.7 percent has been sustained over two decades, Uganda reflects a poverty line that is too low and not remains a very poor country. In 1993, Uganda a reality in which only a fifth of Ugandans are was one of the poorest countries in the world, so, unable to meet their basic needs. When the even after two decades of progress, poverty is still national poverty lines are converted into 2011 widespread. In 2013, more than a third of its citizens PPP they vary from 72 percent to 82 percent of the live below the international extreme poverty line of international extreme poverty line of US$1.90. The US$1.90 a day. international extreme poverty line is designed to 3. Infant mortality and under-five mortality are per 1,000 children xv capture the average national poverty line among percent of households in Uganda use electricity for the world’s poorest countries, so the fact that lighting. Uganda’s poverty lines are much lower, suggests that the poverty lines used in Uganda are too low. 16. In addition, vulnerability to poverty in Uganda is high. Between 2005 and 2009, for every three 14. An updated national poverty line that reflects Ugandans who were lifted out of poverty, two the changes in consumption patterns of fell back into poverty, illustrating the fragility of Ugandan households since 1993 suggests a the gains realized by the poorest households poverty rate in the range of 33 to 35 percent. (Ssewanyana and Kasirye 2012). Uganda’s The national poverty lines were set using data from success in reducing poverty has resulted in many 1993 and have not been updated to reflect the real households that are living just out of poverty who price increases of some foods that poor households remain vulnerable to falling back in to poverty in consume and the changing nature of food and non- the face of a negative shock. food consumption in Uganda. Poverty lines that are 25 percent to 30 percent higher would reflect the 17. Poverty has become increasingly concentrated changes in consumption over the last 15 years and in the Northern and Eastern regions of Uganda would bring the lines closer to the standard used by as the Central and Western regions have other low-income countries. experienced more rapid poverty reduction. There are large and increasing regional variations 15. Although there was improvement in the non- in poverty with most of the poor concentrated monetary dimension of well-being, the country in the north and the east. In 2006, approximately still faces widespread deprivation. Despite 68 percent of the poor lived in the northern and improvement over the last decade, access to eastern parts of the country. Seven years later, this basic infrastructure services remains abysmally proportion increased to 84 percent. Poverty has low, particularly for the poor. Access to improved fallen in all regions, but gains have been slower in sanitation facilities remains very low by regional the poorer Northern and Eastern regions (Figure 5). and international standards. Less than a third The annual percent reduction in poverty has been of households (31.3 percent) have adequate almost twice as high in the Central and Western sanitation and a quarter of poor households regions (7.4 and 7.9 percent respectively) than have no toilet facility at all. Access to electricity in in the Northern and Eastern regions (3.1 and 4.7 Uganda is one of the lowest in the world. Only 14 percent respectively). FIGURE 5: Share of poor population in each region, 2006–2013 Source: UNHS 2006–2013. xvi 18. High fertility rates and widespread acceptance among teenage girls, also jeopardize educational of discriminatory attitudes to women hold attainment. Pregnancy is the fourth most common back the participation of women in Uganda’s reason for dropping out of secondary school: in development, despite impressive gains in 2013, 1 in 10 girls report dropping out of secondary primary female enrollment, maternal mortality, school as a result of pregnancy. Lower rates of and poverty reduction among female-headed agricultural productivity among female-headed households. Although, on average, female-headed households can largely be accounted for by the households are no poorer than male-headed higher childcare demands they face (Ali et al. households, some groups of female-headed 2015). Perceptions also limit Uganda’s progress in households are particularly vulnerable to poverty. reducing gender inequalities: perceptions of gender Female widows are almost twice as likely to be appropriate economic roles have been found to poor compared to male widowers. Maternal account for lower female earnings (Campo et al. mortality rates have been falling but are still high 2015), and worryingly, nearly four in every five and given each woman goes through six births on Ugandan women accept domestic violence—the average, having children still poses a significant second highest acceptance of domestic violence in risk to women. High pregnancy rates, particularly Sub-Saharan Africa (World Bank 2016a). Long queue at a Health center in Kabong District. xvii THE DRIVERS OF PROGRESS: AGRICULTURE, URBANIZATION, AND EDUCATION AGRICULTURE 19. Poverty reduction among households in main sector of employment, half of those engaged agriculture accounts for 79 percent of national in agriculture have additional sources of income poverty reduction from 2006 to 2013 (Figure from non-agricultural activities. However, poverty 6). To some extent this is to be expected as the fell just as fast for agricultural households that agricultural sector is the main sector of employment were solely engaged in agriculture as for those with for households in Uganda, particularly so for poorer diversified income sources, suggesting that growth households. Although the agricultural sector is the in agricultural incomes drove poverty reduction. FIGURE 6: Sectoral contribution to poverty reduction, 2006 to 2013, main source of income 20. High rates of agricultural income growth were observed from 2006 to 2012, particularly for the poorest. Figure 7 shows how different sources of income have grown for households in the bottom 40 percent from 2006 to 2012. Agricultural income grew at 6 percent per capita per year. Agricultural income growth is also found to be more strongly correlated with consumption growth than other sources of income growth, particularly for the bottom 40 percent. Source: UNHS 2006–2013. FIGURE 7: Real income per capita by source of income, bottom 40 percent Source: Uganda National Panel Survey (UNPS) 2006–2012 xviii 21. Agricultural incomes grew because the household income coming from crop sales government got right some key fundamentals increased from 2006 to 2012. The share of that provided the incentives to invest time households in the bottom 40 percent that are in agricultural production and engage in selling crops increased from 60 percent in 2006 to agricultural markets. Conflict with the Lord’s 72 percent in 2012 (Figure 8). It is crops that are Resistance Army in the Northern region of Uganda produced for domestic and regional consumption was stabilized in 2008 and this had a positive that dominate crop income. Coffee is important impact on crop income. Establishing peace for some households, but does not comprise more was associated with a doubling (a 112 percent than 10 percent of crop income in any region. growth) in crop income in affected areas. In This is consistent with the export data that shows addition, markets, particularly in the north and that coffee fell from comprising three-quarters of east, have been improving since 2006 because of exports at the beginning of the 1990s to a third infrastructure investments, new export markets of exports by 2005 (World Bank 2007) and that 41 opening up in South Sudan, DRC and in Kenya, percent of exports now go to Uganda’s four regional better market information for farmers and traders neighbors (in order of importance): South Sudan, (because of the development of a well-functioning the Democratic Republic of Congo, Kenya, and information and communication technology Rwanda (World Bank 2016b). [ICT] sector), and growth in trade services, which improved marketing efficiency. This has contributed to real relative price increases for 24. Agricultural growth was not driven by agricultural commodities that poor farmers grow technology adoption or change in the nature and sell. of production. When extension services were provided crop income was 20 percent higher, but few households received extension services. 22. Luck was also on Uganda’s side: good weather Extension services expanded from 8 percent of benefited many households and positive price households in 2006 to 12 percent of households trends in international and regional markets in 2013. There was very little growth in the use of aided real crop price increases. Prices reflect improved inputs and as a result modernization of not just improvements in marketing efficiency, agricultural practices contributed very little to crop but also favorable changes in supply and demand income growth. Understanding why farmers did not conditions within and outside of Uganda. Peace adopt agricultural technologies during this time of in South Sudan and the Democratic Republic high prices and designing policies that help farmers of Congo provided new sources of demand for overcome these constraints needs to be a key area Ugandan food production. Good rainfall and prices of action going forward. Recent research suggests account for 51 percent of the improvement in crop that poor quality of inputs, limited access to credit, income for all households and 66 percent of the and lack of knowledge are binding constraints. The improvement in crop income for the bottom 40 high prevalence of low-quality inputs in domestic percent. A 10 percent increase in water sufficiency input markets results in negative returns on increases crop income by 9.9 percent. A 10 percent average, even though prices are high. If authentic increase in the price of maize or beans increases technologies replaced these low-quality products, crop income by 4.5 and 9.2 percent, respectively. average returns for smallholder farmers would be over 50 percent. 23. The importance of regional and domestic markets in contributing to agricultural growth is confirmed by the fact that the share of xix FIGURE 8: Share of crop income derived from crop sales, bottom 40 percent, 2006–2012 25. Urbanization can account for one-tenth of the poverty reduction that took place from 2006 to 2013, accounting for the movement of 180,000 people out of poverty. While the bulk of Uganda’s 35 million inhabitants live in rural areas, the country is urbanizing at a considerable pace. Between 2002 and 2014 the share of Uganda’s population living in urban areas increased by more than 50 percent, from 12.1 percent to 18.4 percent (UBOS 2014b). Urbanization has been an important driver of poverty reduction from 2006 to 2013 (Figure 9). Migration, in addition to demographics and redistricting, contributes to urbanization. Source: Staff calculations using rural income-generating activities (RIGA) income aggregates calculated from UNPS 2006–2012 FIGURE 9: Locational contribution to poverty 26. Careful analysis on the impact of migration reduction suggests it results in consumption growth that is 14.6 percent higher per year for migrants compared to those who do not migrate. Migration has a large and positive impact, both for those who move to rural destinations and those who move to urban destinations, but the impact of migration is larger when it entails moving to an urban area. Annual consumption growth is 16.3 percent higher for those who migrate to urban destinations and 14 percent higher for those who migrate to rural destinations. Migration can bring about welfare gains if individuals are able to move from areas where the return to labor is low to areas where the return to labor is higher because of Source: Staff calculations using UNHS 2006–2013 better market opportunities. This appears to have been the case for both rural-urban migration and rural-rural migration in which migrants often came from remote, conflict affected rural areas. Migration can also aid poverty reduction through the remittances that it allows. Currently little is known about the role of remittances in bringing about poverty reduction in Uganda. xx 27. Urban migration is facilitated by education and EDUCATION access to finance and hindered by remoteness 29. Although progress on education has been slow, and lack of access to social networks in urban progress has aided poverty reduction, accounting areas. Those who are more educated are more for half of the consumption growth experienced likely to migrate and more likely to send household by poor households. Households with higher levels members to migrate. Even once controlling for of education have higher agricultural incomes and other factors, a one-year increase in schooling more productive nonfarm enterprises. Education leads to 0.1 percent increase in the incidence of also enables migration and helps households out-migration. Having a formal loan and a savings gain more productive wage employment. The account increases the likelihood of becoming a estimated returns to education in Uganda range migrant-sending household by 3 and 6 percentage from 4.5–8.3 percent (Lekfuangfu et al. 2012). Over points, respectively, controlling for other factors. the last decade, there was slow improvement in Access to finance can also help overcome the costs human capital outcomes but the slight increase of associated with migrating from a remote area to a the share of households with secondary education distant urban center. There is also some evidence aided consumption growth. Decomposition analysis that access to mobile phones helps overcome suggests this improvement can account for half barriers associated with limited social networks in of the consumption growth of households at the urban areas. bottom of the consumption distribution (Figure 10). The strong positive correlation of secondary 28. Some migration—both rural and urban—is the education and consumption growth is particularly result of experiencing loss of income, assets, important for poor households. or security. Young, working age individuals from 30. Higher educational outcomes contribute to areas with higher levels of conflict-related fatalities growth in wage employment income and were more likely to migrate and migrated to rural migration and enables households to diversify in areas. Young, working age individuals from areas the face of shocks. Panel data analysis shows that with high levels of rainfall-induced harvest losses as households have increased the level of education were more likely to migrate to urban areas. Losing of household members they are more likely to see assets and having no network to rely on in a time of growth in wage income and in migration, particularly need also encouraged migration. While migration to urban areas. Having some secondary education helped increase the welfare of these individuals in implies a 1.4 percent reduction in the intensity of the face of shocks, it is not clear whether migration a weather shock for households in the bottom 40 is the optimal instrument to manage risk. Reducing percent. More education facilitates diversification exposure to risk and increasing access to other by enabling increased participation in the labor tools with which to manage shocks when they do market. Productivity in agriculture is also higher for occur may prove more beneficial in the long term. those with higher levels of education. Pupils in class in Alidi Primary School, Oyam District xxi FIGURE 10: Contribution of education to consumption growth Source: Staff calculations using UNHS 2006–2013. WHAT DID NOT CONTRIBUTE? DEMOGRAPHICS, STRUCTURAL CHANGE, AND REDISTRIBUTIVE FISCAL POLICY DEMOGRAPHICS 31. Uganda has one of the youngest and FIGURE 11: Higher dependency ratios held back most rapidly growing populations in the consumption growth, especially for the poorest world. About half (48.7 percent) of Uganda’s population is younger than 15, well above Sub-Saharan Africa’s average of 43.2 percent and world average of 26.8 percent. The country’s population growth rate, currently at 3.3 percent, is also above Africa’s average. 32. An increasing dependency ratio held back consumption growth from 2006 to 2013, reducing the consumption growth of the poorest households by 15 percent to 20 percent. Although the fertility rate is high, it has been slowly falling over the last two Source: Staff calculations using UNHS 2006–2013 decades. However, the drop in fertility rates in recent years has yet to substantially change the demographic composition of Ugandan The drop in fertility rates in recent years has households. The dependency ratio has been yet to substantially change the demographic increasing, particularly for poorer households. composition of Ugandan households. The This increase held back consumption growth dependency ratio has been increasing, from 2006 to 2013 (Figure 11). Reducing the particularly for poorer households. dependency ratio will benefit consumption growth, particularly for poorer households. xxii STRUCTURAL CHANGE 33. There has been little change in the proportion of diversification was rapidly increasing, it was not households that count agriculture as their main a major driver of progress from 2006 to 2013. sector of employment since 2006. This is despite Poverty reduction was just as fast for those solely high growth rates in services and manufacturing in agriculture as it was for those with diversified during this period. Additionally few households income sources. However, some households did have diversified into nonfarm activities. From 1993 experience growth in non-agricultural incomes to 2006, many households stayed in agriculture, and this aided improvements in consumption and but diversified their sources of income by taking reductions in poverty. additional income activities in non-agricultural sectors (Fox and Pimhidzai 2011). This trend has 35. Diversification has increased the resilience not been observed since 2006. The high rates of of households to shocks by making them growth in non-agricultural sectors resulted in job less vulnerable to the impact of bad weather. creation keeping pace with growth in the working Weather has a smaller impact on consumption age population, but not outpacing growth. than it does on crop income because households are able to increase income from non-agricultural 34. Structural change and diversification was not activities (Table 1). If agricultural income is affected a major driver of poverty reduction since 2006, by climate shocks, households can offset this although growth in nonfarm incomes helped with increased nonfarm income. As a result, a lot some households. Although diversification may of movement in and out of nonfarm activities by have driven poverty reduction before 2006, when agricultural households is observed. TABLE 1: Impact of weather on diversification Impact of 10 per- Crop income Non-agricultural Nonfarm self-em- Consumption cent reduction in wage income ployment income rainfall on… All households 18.9*** −36.3*** −28.0*** 4.8*** Bottom 40 percent 24.2*** −43.7*** −33.3*** 4.0** Source: Staff calculations using UNPS 2006–2012. Note: Significance levels are reported as follows: *p < 0.10, ** p < 0.05, *** p < 0.01. 36. Unhelpful gender norms, low levels of education, engage in, causing them to go into lower productive and lack of access to infrastructure and finance sectors (Campo et al. 2015). has limited the degree to which households move out of agriculture. Low education, lack of 37. Limited firm growth and job creation has also access to financial instruments (both savings and resulted in structural change contributing little credit), and lack of access to requisite infrastructure to poverty reduction. While an analysis of the (such as electricity) has constrained non-agricultural constraints to firm growth is beyond the scope of income growth for many households. In addition, this report, and have been discussed elsewhere strong gender norms have constrained non- (for example, World Bank 2015), the results of agricultural income growth for many women during the analysis undertaken show that the limited this period. Female adolescents are likely to give growth of non-agricultural jobs for the bottom 40 birth and get married young, limiting their income percent has been a missed opportunity for Uganda. earning potential (Bandiera et al. 2015). Gender Structural change could have contributed to norms influence the type of activities women poverty reduction more had this been present. xxiii REDISTRIBUTIVE FISCAL POLICY 38. Growth, not redistribution, drives poverty highlighting the absence of reliable official safety net reduction in Uganda reflecting a limited use programs. Safety nets provided by savings, family, of fiscal policy to redistribute incomes in and friends are of paramount importance in the comparison to other countries in the region. absence of official safety net programs. In a context Public transfers to households are negligible in which income volatility is high, limited formal in Uganda. The proportion of poor households safety nets result in considerable vulnerability to receiving any kind of transfer is 5 percent. Uganda’s poverty. Savings cannot help mitigate large shocks total spending on social security in 2013 was 1 and reliance on families and friends in the absence percent of GDP compared to an average of 2.8 of formal safety nets is not always ideal. If all are percent for other countries in Africa. Of that 1 affected by the same bad event (for example, poor percent, only 0.4 percent was spent on direct rains or low cash crop prices), they are unable income support to poor households, compared with to provide help. Not only does the lack of formal 1.1 percent in other low-income countries in Africa safety nets result in households falling into poverty (World Bank 2015). when setbacks occur, it also limits consumption growth for poor and vulnerable households. These 39. There is also limited government support households avoid investing in risky production available to households to manage shocks to activities even when returns are high. In addition, welfare. Figure 12 indicates that households rely excessive reliance on informal networks can result on savings (35 percent) and help from family (25 in individuals hiding or foregoing income to avoid percent) to mitigate the impact of shocks. Very few the risk of this type of informal taxation in the future report receiving support from the government, (Fafchamps and Hill 2015; Jakiela and Ozier 2015). Uganda’s total spending on social security in 2013 was 1 percent of GDP compared to an average of 2.8 percent for other countries in Africa. FIGURE 12: Self-reported coping mechanism Source: Nikolaski et al. (2015) using UNPS 2011. xxiv IMPROVING HEALTH AND EDUCATION OUTCOMES FOR POVERTY REDUCTION IN UGANDA 40. Although fiscal policy does not play an important low, averaging about 3.2 percent of GDP annually. role in directly redistributing income to reduce It is also here that the implementation gap that poverty, public spending can provide an has been increasing in recent years has limited the important role in facilitating poverty reduction effectiveness of government. through the provision of basic services. However, the share of public spending on education and 41. The significant increase in primary enrollment health services is low in Uganda, in comparison rates has yet to translate into substantial to regional peers. In 2013, public spending on improvements in educational outcomes. The health accounted for only 24 percent of the total high primary school enrollment rates among both expenditure on health. In contrast, this share was poor and rich children reflect the benefits of the UPE 37 percent among low-income countries and program that was introduced by the GoU in 1997. almost 44 percent among developing economies However, primary completion rates are lower than in Sub-Saharan Africa (World Bank 2015). This expected, and the trends show that the completion is compounded by the fact that overall public rate fell as more children were enrolled in school. spending is low because of limited tax revenue Uganda’s gross primary completion rate was 53 generation. Because of low levels of spending, percent in 2011. When compared with its peers, out-of-pocket payments are generally higher in Uganda’s primary completion rate is low (Figure 13). Uganda than those in other countries in the region As a consequence, the out-of-school rate for lower and in countries with similar levels of GDP per secondary is much higher than its income peers. capita. Public investment in education also remains FIGURE 13: Primary completion rate is among the lowest in the world Source: WDI. xxv 42. More and better health and education inputs analysis that the level of inputs and their seem to be available in better-off locations, as quality is higher in better-off communities. The expected. Consider, for example, the number perceived quality of service is negatively correlated of pupils per classroom. These ratios are much with community welfare. The likely explanation higher for the poorest quintile of communities is that poor communities are so deprived that than the richest. A typical classroom in the poorest their expectations are low. This leads them to be quintile has 116 pupils, while the corresponding more rapidly satisfied with the services they get. figure for the richest quintile is 58 pupils. Teacher By contrast, better-off communities have higher absenteeism rates at the level of schools or expectations and, therefore, are more demanding classrooms are also negatively correlated with about quality and less satisfied even if objectively welfare.4 For communities in the poorest quintile, they are getting comparatively better services. This about four out of ten teachers are absent from has a series of implications on how to deal with school. Teachers are more likely to be absent community feedback, including importance of in poorer areas. Unlike in the education sector, access to information and education of beneficiaries there is no apparent correlation between health on what quality to expect. This result also implies workers’ absenteeism and the welfare level of that community feedback as such is useful, but communities. However, there is a clear correlation should not be the sole source of monitoring between patient caseload and community welfare.5 information. A health worker in the poorest quintile provides consultations to six outpatients per day (median 45. The contrast between satisfaction and value) versus three for staff in facilities in the richest quality of service provision raises questions quintile of communities. Sick people in poor areas for the effectiveness of community based are more likely to face overcrowding and long monitoring and the demand for accountability. queues while visiting their health centers. If the population in poor communities has low expectations or is not exposed enough to what 43. The low quality of inputs negatively affects services of good quality should look like, to be service delivery outcomes, especially in poor able to indeed assess quality, it is not clear that areas. Teacher’s absenteeism constitutes a barrier it can effectively lobby for quality services. For to pupil’s achievement. Similarly, teachers and social accountability mechanisms to be effective, health workers often lack the minimum knowledge additional measures may be needed to enable to properly teach or treat patients. Evidence disadvantaged communities to properly monitor suggests that workers knowledge is lower in poor the services they receive. The issue is not specific communities. As a consequence, children may go to Uganda, and there are examples of social to school, but not master the necessary knowledge accountability initiatives with mixed results (Fox that they need to be successful on the labor market. 2015). Issues of political economy may also have to Similarly, public and private spending on health be considered for social accountability measures to will not guarantee value for money. All these have work (Joshia and Houtzagerb 2012). The importance the potential to have a negative impact on human of information for a positive impact of community capital accumulation, even more so for the poor, as monitoring has been documented for the case they experienced the lowest quality of services. of Uganda by Reinikka and Svensson (2005) and Svensson et al. (2015) among others. Reinikka 44. Poorer communities are more likely to be and Svensson (2005) conducted an experiment satisfied with the services that they are that shows that making information on budget receiving, even though it is clear from the allocation available to beneficiaries reduces 4. That is, whether teachers are in the classroom even if they may be in the school. 5. Patient caseload is defined as the average number of outpatient visits a health worker attends to per working day. xxvi corruption and elite capture and has a positive monitoring and there are financial incentives for impact on enrollment and educational outcomes. teachers at stake. Moreover, they also found that Svensson et al. (2015) conducted an experiment parents generate significantly less reliable reports on community-based monitoring of absenteeism than head teachers do. Overall, in a context where versus head teachers monitoring. They found that poverty and expectations are a problem, more local monitoring improves teacher attendance needs to be done for social accountability to be but only when the head teacher is responsible for effective. FIGURE 14: Inputs and user satisfaction by welfare quintiles in education sector Absenteeism, Pupil per classroom per teacher Teacher and pupil knowledge Satisfaction and health workers absenteeism Satisfaction and child mortality Source: Staff calculations using the 2013 Service Delivery Indicators (SDI) survey, the UNHS 2012/13, and the Uganda Demographic and Health Survey (UDHS) 2011. If the population in poor communities has low expectations or is not exposed enough to what services of good quality should look like, to be able to indeed assess quality, it is not clear that it can effectively lobby for quality services. xxvii ENDING EXTREME POVERTY IN UGANDA 46. This report has documented that Uganda has 50. Modernizing agricultural production will continued to reduce poverty from 2006 to require a focus both on fostering demand 2013, even as growth faltered. Although growth for agricultural products and on addressing slowed for all households, poor households still the constraints households face in making experienced consumption growth and poverty investments. Continued efforts in increasing fell. Agricultural growth drove much of this demand for agricultural production through poverty reduction aided by peace in the north, regional trade, growth in urban demand, and improvements in domestic and regional food investments in agro-processing industries are markets, favorable international prices, and good needed to keep prices of agricultural commodities weather. Urbanization and modest improvements high. Addressing constraints to modern input in education outcomes also contributed to poverty adoption will entail improving the quality of inputs reduction. in local markets through certification (public or private), and complementary investments in 47. However, it is not clear that the processes that extension and credit to address the knowledge brought about gains in the past can be relied and financial constraints farmers face. This is upon to address the continuing challenge of particularly important in the Northern and Eastern extreme poverty in Uganda, particularly in the regions where agricultural income growth is impoverished Northern and Eastern regions. particularly vulnerable. Addressing the volatility Uganda’s formula for success is one that works of returns to investing in agriculture in this when conditions are favorable, particularly in region—through safety nets or other insurance agriculture. Moreover, luck was on Uganda’s side. mechanisms—may also be needed. There was little fundamental change in the nature of production that benefited poverty reduction— 51. Increasing the contribution of non-agricultural either in agriculture or in other sectors. income growth to poverty reduction requires a focus on firm growth and job creation, but 48. The benefits of security and liberalized markets also investments in education and increased cannot be underestimated and will likely aid financial inclusion. An assessment of the future poverty reduction, as they have done in constraints to firm growth are beyond the focus of the past. Ensuring continued stability in the region, this report, but this report has shown that for non- and further promoting efficient crop markets and agricultural growth to be inclusive of the poorest regional exports such as through investments in households, investments in education and skills regional corridors and improving export efficiency training for the poorest are needed (especially for will be important for future agricultural growth in vulnerable groups such as adolescent girls), as well Uganda, and this benefits poor households. as stronger financial markets for savings and credit. When urbanization occurs this brings direct gains 49. However, sustained gains also require a to those who move, and evidence suggests that fundamental shift in the nature of production investments in education and financial markets will from low-investment, informal activities to aid migration. higher-capital, more productive employment. This in turn requires effective public investment in 52. Improving educational outcomes and services (such as education, health, and agricultural addressing knowledge gaps through extension extension) and safety nets. Without this, it is hard and vocational training will require improving to ensure sustained progress in poverty reduction, service delivery. Although the analysis highlights reduce vulnerability, and address regional many benefits to higher education, progress in inequality. improving educational outcomes has been slow. xxviii The quality of service delivery is lower for poorer households and poorer households are also less vocal about the poor quality of service delivery they receive, limiting the effectiveness of local accountability mechanisms to improve service delivery in poor communities. 53. Concerted action to reduce fertility rates is also needed to reduce the strain that high dependency ratios puts on poverty reduction and to improve the socioeconomic status of women. Investing in education and economic opportunities for adolescent girls helps to reduce fertility rates. Vocational training center in Moyo District xxix This report examines Uganda’s progress in reducing poverty over the last two decades, with a specific focus on the period 2006 to 2013. Uganda’s progress in reducing poverty from 1993 to 2006 is a remarkable story of success that has been well told. Annual reductions in the national poverty rate of 1.9 percentage points a year resulted from the restoration of peace and stability to much of the country after Yoweri Museveni came into power in 1986, the INTRODUCTION series of economic liberalization reforms that were implemented in the 1990s, and the investments of households and firms that these encouraged (see for example, Collier and Reinikka 2003, World Bank 2007). The narrative of Uganda’s continued, albeit slightly slower, progress in reducing poverty since 2006 is less familiar. This was a period in which growth started to slow, as the gains from reforms years earlier had been fully realized, and weak infrastructure and increasing corruption increasingly constrained private sector competitiveness (World Bank 2015). During this period, the national poverty rate still fell by 1.6 percentage points per year and Uganda still recorded the second fastest percentage point reduction in extreme poverty per year in Sub-Saharan Africa since 2000, an African success story.6 This report examines how, in this context, Uganda was still able to secure consumption growth for many of its citizens. Uganda has a wealth of household survey data that has been used in this work. The quality, regularity, and comparability of available household surveys set Uganda apart from many other countries in the region. The core of the analysis undertaken in the report uses two series of surveys: (a) the UNHS undertaken in 1992/93, 1999/2000, 2002/03, 2005/06, 2009/10, and 2012/13 (henceforth referred to as 1993, 2000, 2003, 2006, 2010, and 2013), and (b) the UNPS undertaken in 2005/06, 2009/10, 2010/11, and 2011/12 (henceforth referred to as 2006, 2010, 2011, and 2012). The UNHS is a nationally representative cross-section and it is from this series that the official consumption aggregates and monetary poverty estimates are derived. This series also provides official statistics on many non-monetary dimensions of well-being. As its name suggests, the UNPS is a panel survey in which households surveyed in UNHS 2006 were revisited in subsequent survey rounds. The sample was nationally representative in 2006 and a random sample of split-offs from sample households have also been followed with the aim of keeping the survey national representative. The survey collects much of the same data as in the UNHS but in addition has detailed information on agriculture, income earned from other sources, and anthropometric data. The advantages of using a panel survey for the analysis of poverty trends are described in Box 1. 6. Uganda reduced the extreme poverty rate by 2.7 percentage points a year, second only to Chad who reduced the extreme poverty rate by 3.1 percentage points per year. This is using poverty numbers reported in Povcalnet as of January 2016, and using the surveys deemed comparable by World Bank 2016. xxx The UDHS undertaken in 2001, 2006, and 2011 and the neither poverty reduction nor to improving vulnerability. SDI survey undertaken in 2013 complement this analysis. The UDHS is a nationally representative cross section Chapter 4 explores the nature of agricultural growth that designed to provide population and health indicator has reduced poverty in further detail and examines what estimates for the country as a whole and for urban and drove progress for poor households during this period. rural areas separately. Estimates can also be reported for Chapter 5 explores why structural change contributes the ten subregions of Uganda. The SDI survey collects so little to progress by examining the constraints facility-based data from primary schools and health households in the bottom 40 percent face in moving out facilities. The sample frame is the list of all facilities in of agriculture. Chapter 6 uses panel analysis to quantify the country. The survey instruments incorporate recent the welfare gains from migration and to explore who has innovations in measuring provider competence and benefited from migration and what constrains migration effort (World Bank 2013). The sample design is national, of others during 2006 to 2012. with the possibility of disaggregating results by rural/ urban locations as well as regions and by type of provider In looking back to explain drivers and constraints of (public or private) for both education and health. progress, these chapters point to a number of priorities for ending extreme poverty in Uganda. Peace in northern Chapters 1 and 2 synthesize progress since 1993, but Uganda, improved regional markets and good weather with a focus on 2006 to 2013. Chapter 1 starts with a drove growth in agricultural incomes. The benefits of focus on monetary poverty. In addition to documenting security and liberalized markets will likely aid future trends in national and international poverty and poverty reduction as they have done in the past. inequality it examines the incidence of consumption However, there was little fundamental change in the growth; assesses whether the poverty line is too low nature of production that benefited poverty reduction— given the changes in the consumption patterns of the either in agriculture or in other sectors. Sustained welfare poor since the line was set in 1993; and simulates future gains also require a fundamental shift in the nature of poverty trends. Chapter 2 takes as its focus progress in production from low-investment, informal activities non-monetary dimensions of well-being and in particular to higher-capital, more productive employment. The assesses the degree to which households in Uganda analysis highlights that transitions require effective have experienced change in non-monetary dimensions public investment in services (such as education, health, of well-being that are commensurate with the country’s and agricultural extension), infrastructure (such as economic development. regional corridors and electricity) and safety nets. The overwhelming conclusion of Chapters 1 and 2 is Increasing the effectiveness of public investment that there has been substantial progress in well-being in Uganda for poverty reduction, in turn requires in Uganda since 2006. In the chapters that follow, addressing improving service delivery. For example, the the factors that have contributed to this progress are analysis shows that education increases agricultural explored. Chapter 3 examines the drivers of poverty income, aids migration and transitions out of agriculture, reduction through decomposition analysis using the and reduces vulnerability. Yet progress in improving UNHS and through analysis of the UNPS that has educational outcome has been slow. Chapter 7 takes as followed the same households through this period. its focus the relationship between service delivery and It highlights the importance of agriculture, urban poverty reduction, highlighting that the quality of service migration, and modest gains in education. It also delivery is lower for poorer households and that poorer highlights the limited role of structural change since households are also less vocal about the poor quality of 2006, the persistently high dependency ratios which held service delivery they receive, limiting the effectiveness back poverty reduction and limited spending on safety of local accountability mechanisms to improve service nets, which have resulted in fiscal policy contributing to delivery in poor communities. xxxi BOX 1: Use of Panel Data for Poverty Analysis This report draws on the nationally representative UNPS to analyze the drivers of welfare changes in Uganda over time. Panel data provides a number of advantages for the analysis of welfare outcomes. It allows the same household to be followed over time, making it possible to calculate the income and consumption growth of a given household over time. Panel data also allows for regression analysis to look at how changes in the characteristics or behavior of the household or individual over time have contributed to changes in welfare. This is arguably a stronger basis for identifying what has caused welfare improvements than just looking at the characteristics of those that are poor or non-poor. However, caution is still warranted in drawing causal conclusions from panel analysis, as it is possible that a characteristic of the household not captured in the analysis allowed the household to both change behavior and experience welfare gains. Inferring that the behavior change caused this improvement would be erroneous. The core pieces of analysis in Chapters 4 and 6 thus rely on changes that can be considered exogenous. In addition, the attrition present in panel surveys—and in the UNPS in particular—makes it less representative of Ugandan households over time. Households that stay in their original location are more likely to be found in a successive visit, but households that have moved or new households that have formed from old households are less likely to be found. Controlling for attrition in the analysis is difficult. This has been addressed in this report by: (a) not using the UNPS to develop descriptive statistics if the same variable is available in the UNHS, (b) focusing analysis on households that have not moved (Chapter 4) or specifically analyzing the splits and moves (as is done in Chapter 6). Pupils in class in Alidi Primary School, Oyam District xxxii CHAPTERS 1 UGANDA’S PROGRESS IN CHAPTER: 1 REDUCING POVERTY Uganda has recorded impressive rates of poverty reduction in the last two decades. The proportion of the Ugandan population living in poverty—whether measured using the national poverty line or the international poverty line—more than halved from 1993 to 2013. 1. The Ugandan economy has experienced high growth through much of the last two decades. Peace and stability were restored in much of the country in 1986 when Yoweri Museveni came into power and then in the north of Uganda in 2008. Stability and the series of economic liberalization reforms that were implemented in the 1990s contributed to high growth (see for example, Collier and Reinikka 2003, World Bank 2007). Growth started to slow in 2010 as the gains from peace and the reforms years earlier had been fully realized, and weak infrastructure and increasing corruption increasingly constrained private sector competitiveness (World Bank 2015). 2. This chapter documents that Ugandan households have also experienced progress in monetary well-being during the last two decades, including during the period of focus for this report, 2006 to 2013. Although consumption growth has been lower, on average, in recent years, it has become increasingly pro-poor. The period from 2010 to 2013 was the only period in the last twenty years in which consumption growth benefited the poor more than the rich and inequality fell. The national poverty rate fell by 1.6 percentage points a year since 2006 (compared to 1.9 percentage points a year before then) and the international poverty rate fell by 2.7 percentage points per year (much higher than the regional average of 0.74 during this period). 2 During 2006 to 2013, Uganda had the second fastest poverty is now concentrated in the Northern and percentage point reduction in poverty per year in Eastern regions of the country where progress is Sub-Saharan Africa. slower. 3. However, this progress is not without its 4. This chapter documents trends in national challenges. Uganda remains a very poor country. and international poverty rates incorporating In 2013, more than a third of Ugandans lived findings from World Bank and other studies that below the international extreme poverty line of have also documented progress in well-being US$1.90 a day. The low national poverty rate of over this period (for example, Ssewanyana and 19.7 percent reflects a poverty line that is too low Kasirye 2013, MoFPED 2014, UBOS 2014a, World and not a reality in which only a fifth of Ugandans Bank 2015). It assesses whether the poverty line are unable to meet their basic needs. An updated is too low given the changes in the consumption national poverty line that reflects the changes patterns of the poor since the line was set in 1993 in consumption patterns of Uganda households and examines what the implications of a higher since 1993 suggests a higher national poverty poverty line would be for poverty incidence in line is needed. Even with a higher line, progress Uganda. The chapter then turns to examining in reducing poverty has been impressive over the incidence of consumption growth and how the last two decades. Yet progress has pushed the distribution of consumption of Ugandan many households just out of poverty and they are households has changed over time. The chapter vulnerable to falling back in to poverty. In addition, concludes by providing a profile of characteristics regional disparities are increasing over time and of the poor and simulating future poverty trends. 1.1. Recent progress in poverty reduction 5. Uganda has recorded impressive rates of poverty fast in the last decade with international extreme reduction in the last two decades. The proportion poverty falling from 62.2 percent in 2003. of the Ugandan population living in poverty— whether measured using the national poverty line 6. The depth and severity of poverty have also or the international poverty line—more than halved fallen consistently. Measured at the national from 1993 to 2013 (Table 1.1 and Figure 1.1.1). poverty line, the poverty gap—the average amount The proportion of the population living under the that each household lives beneath the poverty line national poverty line declined from 56.4 percent (expressed as a percentage of the poverty line)—fell in 1993 to 19.7 percent in 2013.7 The proportion from 11.9 percent in 2003 to 5.2 percent in 2013 of households living beneath the international (Table 1.2)8. The severity of poverty, an index that extreme poverty line of US$1.90 a day (2011 prices) gives more weight to those households who fall fell from 68.1 percent in 1993 to 34.6 percent in substantially below the poverty line, fell from 5.1 2013. The rate of progress has been particularly percent to 2 percent. 7. The national poverty line ranges from US$0.88 to US$1.04 2005 PPP per capita depending on the region. Poverty in Uganda is calculated using a cost-of-basic-needs approach. Consumption expenditure data is collected on food and non-food items through the UNHS conducted every three to four years. The poverty line was set in 1993 by calculating the cost of consuming 3,000 calories per adult equivalent and then adding an amount (the amount depending on the region) to capture non-food expenditures. The poverty line has only been updated for the cost of inflation since then and is low by international standards. 8. This measure reflects the depth of poverty as well as its incidence. The indicator is often described as the per capita amount of resources needed to eliminate poverty or reduce the poor’s shortfall from the poverty line to zero, through perfectly targeted cash transfers. 3 7. Uganda has experienced one of the fastest mortality and child nutrition. However, Uganda reductions in extreme poverty seen in Sub- is still lagging behind in many dimensions. For Saharan Africa. Uganda’s reduction in poverty has instance, access to electricity is one of the lowest kept pace with the strong growth in gross national in the world. Education outcomes have improved income (GNI) per capita that it experienced from as well, but the significant increase in primary 1999 to 2013 (Figure 1.1.2). Uganda had the second enrollment rates has yet to translate at higher fastest percentage point reduction in poverty levels. Chapter 2 looks at non-monetary well-being per year in Sub-Saharan Africa during the period in detail. of focus for this study (2006 to 2013), an African success story (Figure 1.1.3). 9. However, Uganda is still a poor country; more than a third of the country still lives in extreme 8. Trends in non-monetary well-being also tell poverty as measured by the international the same story of rapid improvements in the poverty line of US$1.90 a day. Figure 1.1.4 well-being of Ugandan households, but there indicates that in comparison to other countries in is still much left to be achieved. The share of Sub-Saharan Africa, Uganda experiences moderate households with improved roof material went up poverty rates. The poverty gap (at US$1.90 2011 by 7 percentage points, from 61 percent in 2006 PPP per capita per day) indicates that it will take to 68 percent in 2013. Nearly three-quarters of an average payment of US$70 per capita per households in Uganda had access to improved year to eliminate extreme poverty in Uganda. water sources in 2013. Ownership of modern Understanding the drivers of recent poverty assets such as mobile phones and motorcycles reduction is important both for offering lessons on has increased. Performance on adult literacy is how to reduce poverty further in the future not only way above expected, given the GNI level, and is on in Uganda, but also for other countries in the region the rise. Cross-country regressions also suggest that have not experienced such a remarkable that Uganda performs well on child and maternal reduction in poverty. TABLE 1.1: Poverty from 1993 to 2013, national and international line Proportion of the Population Living Beneath National Poverty Line* International Poverty Line** 1993 56.4 68.1 2000 33.8 52.1 2003 38.8 62.2 2006 31.1 53.2 2010 24.5 41.5 2013 19.7 34.6 Source: UNHS 1993–2013. Note: * Ranges from US$0.94 to US$1.07 PPP per capita per day depending on the region of the country.10 ** US$1.90 2011 PPP per capita per day. 9. Uganda reduced the extreme poverty rate by 2.7 percentage points a year, second only to Chad which reduced the extreme poverty rate by 3.1 percentage points per year. This is using poverty numbers reported in Povcalnet as of January 2016 and using the surveys deemed comparable by World Bank 2016. 10. This is calculated by converting the region-specific national poverty lines to U.S. dollar 2005 PPP and dividing by the average ratio of adult equivalents to individuals (given the national poverty line is a per adult equivalent line). 4 TABLE 1.2: Reductions in the depth and severity of poverty at the national line, 1993 to 2013 Poverty Depth Poverty Severity National Rural Urban National Rural Urban 1993 20.3 22 8.3 9.9 10.81 3.48 2000 10 11.2 2.1 4.25 4.79 0.68 2003 11.9 13.1 3.9 5.1 5.7 1.6 2006 8.7 9.7 3.5 3.5 3.9 1.4 2010 6.7 7.6 1.8 2.8 3.1 0.6 2013 5.2 6 2.5 2 2.4 0.9 Source: UNHS 1993–2013. Note: 2000 excludes Kitgum, Gulu, Bundibugyo, Kasese, and Pader districts. TABLE 1.3: National poverty rates by region Year Region Central Eastern Northern Western 1993 45.6 58.8 73.5 52.7 2000 19.7 34.9 63.7 26.2 2003 22.3 46.0 63.0 32.9 2006 16.4 35.9 60.7 20.5 2010 10.7 24.3 46.2 21.8 2013 4.7 24.5 43.7 8.7 Percentage point reduction, 17.6 21.5 19.3 24.2 2003–2013 Annual percent reduction, 7.9% 4.7% 3.1% 7.4% 2003–2013 Source: UNHS 1993–2013 FIGURE 1.1: Two decades of progress in reducing poverty 1. Headcount poverty rate, national and 2. Poverty reduction and growth in GNI per capita, international poverty line, 1993 to 2013 Uganda (marked) and all other countries 5 3. Annual reduction of poverty headcount at international 4. Extreme poverty, Uganda and other African poverty line for selected countries (2003-2013) countries (latest survey year) 5. National poverty rates by region 6. Poor, vulnerable, and middle class 100 10.2 90 22.4 21.2 28.7 32.6 80 37.0 33.4 70 Population share (%) 60 39.9 43.9 50 40.2 42.9 40 43.3 30 56.4 20 38.8 33.8 31.1 24.5 10 19.7 0 1992/3 1999/2000 2002/3 2005/6 2009/10 2012/13 Poor Insecure non-poor Middle class Sources: 1, 5: UNHS 1993–2013; 2 Gable, Lofgren, and Osorio-Rodarte (2015); 3, 4: WDI; 6: MoFPED 2015. Notes: 2: Each point represents a country, the years denote Uganda’s values for 1999, 2013, and 2013 (projected). The regression line and confidence interval is also shown. The graph uses 2005 PPP and the poverty rate of US$1.25 2005 PPP per day. 6: Poverty Status Report: Absolute poor: living below the national poverty line; Insecure: living below twice the poverty line; Middle class: living above twice the poverty line. 10. Poverty has fallen in all regions, but gains have rate, vulnerability to poverty in Uganda is high. been slower in the poorer Northern and Eastern Nearly 43 percent of Ugandans were insecure regions (Figure 1.1.5). As Table 1.3 shows, the non-poor in 2013, defined as those living above annual percent reduction in poverty has been the poverty line but living on less than twice the almost twice as high in the Central and Western poverty line (Figure 1.1.6)11. Between 2005 and regions (7.4 and 7.9 percent, respectively) than 2009, for every three Ugandans who were lifted out in the Northern and Eastern regions (3.1 and 4.7 of poverty, two fell back into poverty, illustrating percent, respectively). However, the percentage the fragility of the gains realized by the poorest point reduction in poverty has been similar across households (Ssewanyana and Kasirye 2013). regions. Spatial concentration of poverty in the Uganda’s success in reducing poverty has resulted Northern and Eastern regions is occurring as a in many households that are living just above the result. poverty line who remain vulnerable to falling under the poverty line in the face of a negative shock. 11. In spite of the significant decline in the poverty 11. As per the Poverty Status Report 2014 produced by the MoFPED. 6 1.2 Is the national poverty line a good measure of poverty in Uganda? 12. The national poverty line used to define an food items have changed substantially since 1993 individual as poor or non-poor in Uganda is and households may have adjusted their food low—about three-quarters of the international consumption patterns in response. In addition, if extreme poverty line of US$1.90—and results in the goods that make up the basket of consumption a low national poverty rate. Poverty in Uganda that sets the poverty line experienced inflation is measured by assessing whether a household higher than the CPI, using the CPI may not have consumes enough to meet their basic food needs allowed the poverty line to keep up with the cost of and other necessary expenditures. The amount living. needed for basic food needs and other necessary expenditures is captured in the national poverty 15. This section examines how consumption lines. Uganda has different poverty lines for patterns have changed over time and what this different regions to allow for the fact that the cost means for how poverty is measured in Uganda of living varies across different parts of the country and the trends in poverty reduction over time. (see Box 1.1 for more details on how poverty The amount and structure of non-food spending is is measured in Uganda). When these poverty examined first. Then the structure of food spending lines are converted into 2011 PPP they vary from and the degree to which the value of the food US$1.36 to US$1.55, 72 percent to 82 percent of the basket has been properly updated by using the CPI international extreme poverty line of US$1.90. The since 1993. international extreme poverty line is designed to capture the average national poverty line among 16. The share of consumption that the poor spend the world’s poorest countries, so the fact that on non-food items is 6 to 26 percent higher in Uganda’s poverty lines are much lower suggests 2013 than in 1993 when the poverty line was that the poverty line in Uganda is perhaps too low.12 set. Table 1.4 presents results on how the share of non-food items in total consumption of poor 13. The national poverty line in Uganda was households has changed over time in Uganda. established using data from 1993 and has been In column 1, the results reported in Appleton updated using the Consumer Price Index (CPI) et al. (1999) are presented. In columns 2 and 3, since then. The poverty line was set based on an the same method used by Appleton et al. (see in-depth analysis of the pattern of food and non- Annex 1 for details) is used to estimate the share food consumption among Uganda’s poor (Appleton of non-food items in total consumption in 2013. et al. 1999). The share of non-food consumption for food poor households is presented in column 2 and for the 14. However, much has changed in Uganda since bottom 50 percent in column 3. In 1993, these two 1993 and the amount poor households need to groups were identical, as food poor households cover the basic food and non-food needs may comprised the bottom 50 percent of households, be quite different. Consumption patterns are but this is no longer the case. The share of non-food likely to have changed since 1993, reflecting the expenditure is higher in 2013 in all regions, with different realities of living in Uganda today. For particularly large changes in the rural parts of the example, in 1993, no household owned a mobile Central and Northern regions. Without adjusting phone, yet today most households in Uganda own the food basket, this increase in share of non-food mobile phones and purchase credit on a regular expenditure would entail a 5 percent increase in the basis to make and receive calls. Relative prices of poverty line. 12. This conversion takes into account the fact that the national poverty line uses consumption per adult equivalent and the international poverty line uses consumption per capita. 7 17. Although the overall amount of spending on reflecting that some foods had become much non-food items has increased since 1993, the more expensive. Figure 1.4 presents data on the structure of non-food expenditure has not share of consumption spending on the seven most changed much. Figure 1.2 presents the share of important food items that together comprised expenditure on the major groups of non-food items half of food expenditure in 1993. In 2013, these in total non-food expenditure in 2000 and 2013 and items also comprised almost half of consumption shows little change over time. In addition, when expenditure (47 percent), but sorghum and maize expenditure on selected items is tracked from had become significantly more important and 2000 to 2013 there is little change in the relative sweet potato and matooke less so. The price of share of these items in total non-food expenditure, matooke and sweet potatoes increased during this even though they do fluctuate. However, one big time, perhaps providing part of the explanation as change is expenditure on telephone services. This to their declining share, but not fully, as the real was nonexistent in 2000, but by 2013 comprised 2 price of sorghum and maize also increased during percent of non-food expenditure. this time. Changes in the relative prices of food items and changing consumption patterns require 18. Household survey data indicates that the prices the items in the food basket to be updated. of food items in the food poverty line basket may have risen faster than the CPI on average. 20. This analysis suggests that the national poverty In 1993, a food basket that provides 3,000 calories lines are too low to reflect the cost of basic per adult equivalent was defined. The cost of this needs of Ugandan households in 2013. This basket was USh 11,463 per month in 1993 prices. analysis also suggests that the national poverty If the CPI is used to adjust this basket, the cost is lines in 2013 should be higher than the lines USh 46,263 (2013 prices). However, when the cost currently used. The items in the basket of food of purchasing this same basket is recomputed consumption need to be updated, as does the using the unit food prices recorded in UNHS amount by which the food consumption basket 2013, the value is 43 percent higher: USh 66,067 is scaled to account for non-food consumption. (2013 prices).13 14 This could, in part, be driven A fuller analysis of consumption needs of poor by methodological differences (although to the households is needed to determine what the new extent possible, the same assumptions as used in basket and line should be, but the existing analysis Appleton et al. 1999 were adopted), but it could suggests the current line is too low. also reflect that the prices of some items in the consumption basket have risen faster than the 21. A higher poverty line would raise the national CPI. In particular, Figure 1.3 shows that the prices poverty rate—perhaps to 33–35 percent—but of sweet potatoes, meat, fish, matooke, sorghum, this higher rate still represents significant millet, and sim-sim increased much faster than the progress in reducing poverty over the last two prices of other goods. decades. Without re-estimating what should be in the food consumption basket of the poorest 19. The structure of food consumption has also households, it is not possible to know how much changed substantially across time, in part the national poverty line should be increased by. 13. Households were asked the values and quantities of items they consumed, and dividing the value by the quantity provides the unit price. Quantities were often reported in nonstandard units and the quantities measured in nonstandard units were converted into kilograms using conversion factors reported in the survey as well as the conversion factors used in the consumption module of the Living Standards Measurement Study–Integrated Surveys on Agriculture (LSMS-ISA). Not all nonstandard units were converted, but enough to provide unit values. 14. The two most common types of consumption recorded are ‘consumption of own produce’ and ‘consumption in the household’ of produce that is purchased. Prices imputed from own consumption are consistently lower across almost all items, and there is a valid concern that households might systematically undervalue consumption from own production. Therefore, the price from purchased ‘consumption in the household’ is used. This is also done when calculating the official household food consumption aggregate. 8 If the basket did not change, the analysis suggests an overestimate. An increase of 25 percent to 30 the poverty line may need to increase by 50 percent could be enough. Increasing the national percent—a 44 percent increase in the basket and a poverty lines by this amount would also bring them 5 percent increase in non-food consumption. This closer to the international extreme poverty line. is very close to the poverty line re-estimation done This would increase the national poverty rate in by Appleton (2003). However, given households 2013 to 32.7 to 35.2 percent. Although higher, this do substitute away from foods that become poverty rate still represents significant progress in relatively more expensive, this would likely be reducing poverty over the last two decades. TABLE 1.4: Spending on non-food items among poor households, 1993 and 2013 Share of Total Expenditure on Non-food Items Percentage Change Food Poor in Share of Non-food, Food Poor Bottom 50 Region Households, 1993–2013 (Percent) Households, 1993 Percent, 2013 2013 Central Rural 0.39 0.49 0.50 26 Central Urban 0.51 0.58 0.59 14 East Rural 0.35 0.40 0.40 14 East Urban 0.44 0.49 0.49 11 North Rural 0.32 0.39 0.40 22 North Urban 0.41 0.48 0.48 17 West Rural 0.36 0.38 0.38 6 West Urban 0.42 0.47 0.47 12 Source: Column 1 is results from Appleton et al. (1999). Columns 2 and 3 are results of staff calculations.15 FIGURE 1.2: Structure of non-food spending over time Source: UNHS 2000 and 2013. Note: This excludes imputed value of freely acquired water, charcoal, and firewood. 15. Note that the results in column 2 and 3 are not much different from each other. This is because the two reference groups do not differ significantly in terms of demographic characteristics and the coefficients of these demographic characteristics (in a regression of non-food share) are not large compared to the constant term and region dummies. The weighted demographic characteristics, weighted by the corresponding coefficient, of the food poor and the bottom 50 percent of households is 0.077 and 0.081, respectively. The weighted difference in demographic characteristics of the two groups is 0.004 only, and this minor difference results in 0.01 difference in non-food share in Central Rural, Central Urban, and Northern Rural categories. In other locations, the use of a different reference group does not affect non-food share. 9 FIGURE 1.3: Prices of food items, 1993–2013 Source: Appleton et al. 1999 and UNHS 2013. Note: Size of bubble reflects the share of the food consumption basket comprised by the item. FIGURE 1.4: Structure of food spending over time Source: UNHS 2000 and 2013. Shop keeper in Ttula - Kawmpe, Kampala 10 BOX 1.1: How poverty is measured in Uganda The poverty line was set in 1998 using 1993 data by estimating the amount of expenditure needed to satisfy the minimum daily calorie requirements and basic non-food needs. Appleton et al. (1999) identified the 28 commonly consumed food items and the corresponding amount consumed to meet 3,000 calories per adult equivalent. Calorie requirement varies by age and gender, and hence the 3,000 calories is per adult equivalence. Based on the population structure then, the average per capita calorie need was 2,283 calories. The minimum expenditure on basic non-food needs was estimated using the classic approach of Ravallion and Bidani (1994) by identifying the non-food expenditure of households that are just on the food poverty line. The justification for using these households’ non-food expenditure as a reference is that the poor have sacrificed some of their need for calories to buy the non-food items. Therefore, these non-food expenditures should also be regarded as meeting essential needs. The non-food expenditure was allowed to vary by region and rural/urban areas to account for spatial differences prices (Appleton et al. 1999). The poverty line is the sum of expenditure on basic food and non-food items. Since 1993, the CPI has been used to update this poverty line. Source: Appleton et al. (1999). 1.3 The incidence of progress and shared prosperity 22. Reducing the number of people living below the the bottom 40 percent. This group is the focus of national poverty line is a significant measure the World Bank Group’s goal of shared prosperity. of progress. However, this is just one measure of In Uganda, this group comprises all of those living how Ugandan households have fared. This section below the national poverty line as well as some takes a closer look at changes in the distribution living above the national poverty line who are of consumption in Uganda from 1993 to 2013, vulnerable to falling back into poverty. The bottom focusing on 2006 to 2013, and sheds light on the 40 percent is a group referred to in much of the role of growth and redistribution in bringing about analysis in subsequent chapters also. changes in poverty. Much of the analysis refers to INCIDENCE OF GROWTH AND SHARED PROSPERITY 23. The period from 1993 to 2000 was a period Uganda has remained 84 percent rural throughout of recovery and stabilization and yielded this time) the bottom 40 percent of the population high consumption growth for all households benefited from growth of 5.3 percent annually (an average of 5.3 percent per annum) and and the top 60 percent benefited from growth of substantial poverty reduction. Internal peace, 4.6 percent (Figure 1.5.1 and Table 1.5). In urban fiscal discipline, and the removal of implicit areas, the pattern of progress was even more rapid, taxation through liberalization of the exchange rate particularly for wealthier households. The bottom and coffee marketing provided the environment 40 percent in urban areas saw incomes increase by needed for growth in household consumption 6.9 percent per annum and the top 60 percent had (Collier and Reinikka 2003). In rural areas (which consumption growth of 8.6 percent.16 dominate the national distribution, given that 16. It is possible that the growth rates are somewhat inflated, given the 2000 survey could not be carried out in some districts where fighting was ongoing. Even taking this into account, consumption growth and poverty reduction during this period was high and impressive. 11 24. From 2000 to 2006, GDP per capita growth rates strong, pro-poor consumption growth in rural dropped and poverty fell marginally (Figure areas (Figure 1.5.4). Just as from 2000 to 2006, 1.6). Rural households experienced low levels GDP per capita growth was less than 3 percent of consumption growth, particularly the bottom (Figure 1.6), yet poverty fell by 5 percentage points. 40 percent for whom growth was 0.9 percent per Higher average household consumption growth annum. In urban areas, household consumption was observed from 2010 to 2013 (1.9 percent), growth was negative (Figure 1.5.2). The national than during 2000 to 2006 (1.3 percent), as a result poverty rate only fell by a couple of percentage of strong consumption growth in rural areas. points as a result, from 33.8 to 31.1 percent. Rural consumption growth was also pro-poor: the 25. High levels of broad-based consumption growth consumption growth rate of the bottom 40 percent were again realized from 2006 to 2010, reflecting was 2.0 percent compared to a consumption high GDP growth, the cessation of conflict in growth rate of 1.0 percent among the top 60 the north of Uganda, and improving terms percent. In contrast urban growth rates were of trade for many farmers (Figure 1.5.3). The negative, although more so for the urban middle establishment of peace in the north of Uganda class (−2.6 percent). benefited households in the Northern, Eastern, and 28. The pattern of pro-poor growth from 2010 Central regions (Figure 1.5.5). Prices for food goods to 2013 is again consistent with price trends were also high during this period, benefiting rural during this period. Subsequent chapters further households. Consumption growth of the bottom 40 percent in rural areas averaged 3.2 percent and for examine the factors underpinning these high rates the top 60 percent it was 3.0 percent. Urban areas of pro-poor consumption growth, but Figure 1.7 also saw high levels of growth, although this growth helps point to some of the external factors that may was less pro-poor. On average, consumption contribute to the pattern of consumption growth growth in urban areas was 5.1 percent for the observed. International coffee prices increased, bottom 40 percent and 5.7 percent for the top 60 thereby increasing the terms of trade for coffee- percent. Given that households in urban areas tend producing households and resulting in very high to be wealthier, nationally growth was marginally rates of consumption growth for households in higher for the top 60 percent (3.5 percent) than for the Western region (Figure 1.5.6).17 The bottom the bottom 40 percent (3.4 percent). 40 percent of households in the Western region experienced annual income growth of 7.5 percent in 26. Although consumption growth was on average these three years. Prices of maize and other staples very strong during this period, households in the were also high, increasing the terms of trade for Western region fared badly from 2006 to 2010. many rural households in other regions. Domestic Figure 1.5.5 shows that consumption growth was markets in Uganda are characterized by low entry negative for most households in the Western region. costs and high levels of competition, which allows Figure 1.7 helps partially explain why: coffee prices in 2010 were almost identical to prices in 2006, but changes in market prices to be transmitted quickly the higher price of food (indicated in the graph to farmers (Fafchamps and Hill 2008). However, with maize prices, but present for other staples too) although higher food prices may have aided rural resulted in the terms of trade worsening for coffee- households, the excessively high rates of inflation producing households, which are predominantly in observed during the election spending in 2011 hurt the Western region. urban households. 27. Strong poverty reduction was recorded from 29. The period from 2010 to 2013 was the only 2010 to 2013, even though this was a period period in the last twenty years in which of lower GDP per capita growth, because of consumption growth benefited the poor more 17. Even though coffee is a perennial crop, high prices translate into immediate welfare gains as farmers exert more labor on maintaining and pruning the tree and on harvesting coffee when coffee prices are high (Hill 2010). 12 than the rich and inequality fell. The average period in which the growth rate of the bottom 40 annual consumption growth rate of the bottom 40 percent was higher than the growth rate of the top percent is used to assess shared prosperity. This 60 percent (2.3 percent compared to 1.8 percent). growth rate can be compared to a relative target— However, from 1993 to 2000 and 2006 to 2010 very the growth rate of the top 60 percent—to determine high growth rates were observed for the bottom 40 whether progress has been shared; or to an percent (5.4 and 3.4 percent, respectively). Shared absolute target, when 3 percent per annum is often prosperity was not met by any measure from 2000 used. The period from 2010 to 2013 was the only to 2006 (Table 1.5). FIGURE 1.5: The incidence of consumption growth, 1993 to 2013 1. 1993–2000: High, broad based growth 2. 2000–2006: Stagnation and worsening inequality 3. 2006–2010: High, inequality increasing growth 4. 2010–2013: Low, pro-poor consumption growth 5. Regional consumption growth from 2006 to 2010 6. Regional consumption growth from 2010 to 2013 Source: Staff calculations using UNHS 1993–2013. 13 TABLE 1.5: Shared prosperity, 1993–2013 1993–2000 2000–2006 2006–2010 2010–2013 Bottom Top Bottom Top Bottom Top Bottom Top 40% 60% 40% 60% 40% 60% 40% 60% National 5.4 5.2 0.8 1.6 3.4 3.5 2.3 1.6 Rural 5.3 4.6 0.9 1.8 3.2 3.0 2.0 1.0 Urban 6.9 8.1 −1.3 −0.4 5.1 5.7 −0.6 −2.6 Regions: Central 7.0 6.3 1.1 2.6 4.5 7.2 3.8 0.0 Eastern 6.2 4.8 0.1 0.2 5.2 4.9 0.2 –1.5 Northern 1.9 1.9 2.2 0.9 3.1 7.7 1.6 2.4 Western 7.0 5.5 1.2 2.2 –0.9 –0.6 7.5 5.7 Source: Staff calculations using UNHS 1993–2013. FIGURE 1.6: GDP per capita growth, 1993 to 2013 Thatching a hut in Kotido District FIGURE 1.7: Coffee and maize prices, 1993 to 2013 Source: Staff calculations using UNHS 1993–2013. 14 30. As a result, 2010 to 2013 was the only period Results of this decomposition are presented in in which redistribution contributed to poverty Figure 1.8. Until 2010, all poverty reduction in reduction. Poverty reduction can be decomposed Uganda resulted from growth. Changes in the shape into a part that comes from an average increase of consumption distribution—redistribution— in consumption across the population (‘growth,’ undermined progress in poverty reduction, as richer that is, the consumption levels of all households households were consistently gaining more than increasing) and that which comes from a change poor households. However, from 2010 to 2013, both in the shape of the consumption distribution growth and redistribution contributed to poverty (‘redistribution,’ that is, consumption of the poorest reduction, as poorer households gained more than growing faster than consumption of the richest). richer households. FIGURE 1.8: Decomposing poverty reduction into growth and redistribution Source: Staff calculations using UNHS 1993–2013. INEQUALITY 31. The growth incidence analysis also provides the absolute and relative difference between the some indication as to how inequality has bottom 10 percent and the top 90 percent (Figures changed over time and the next paragraphs 1.9.3 to 1.9.5). present information on summary measures 33. However, the increase has been marginal and of inequality. Box 1.2 outlines the inequality Uganda has a moderately low rate of inequality measures used. compared to other countries in the region. The 32. Inequality has been steadily increasing in rural change in the Gini from 1993 to 2010 has been an and urban Uganda from 1993 to 2010, by any annual increase of 0.4 percentage points per year. measure. Inequality, as measured by the Gini Figure 1.9.1 shows that Uganda faces moderately index, increased from 35.7 percent in 1993 to 41.5 low inequality in comparison to other countries in percent in 2010 (Figure 1.9.2). This finding holds the region. Inequality is higher in urban areas than when looking at other measures of inequality such in rural areas, as is often the case, but the increase in as the Theil index with the parameter α=−1 which inequality in urban areas has occurred at the same emphasizes inequality for lower incomes and speed as the increase in inequality in rural areas. 15 34. Inequality fell from 2010 to 2013, consistent with the finding that changes in the consumption distribution favored the poor during this period. Inequality fell from 41.5 percent in 2010 to 38.5 percent in 2013, a reduction of 1 percentage point in the Gini per year. BOX 1.2: Inequality Measures While poverty measures absolute deprivation with respect to a given threshold, inequality is a relative measure of poverty indicating how little some parts of a population have relative to the whole population. In the context of monetary poverty, equality can be defined as an equal distribution of consumption/ income across the population. This means that each share of the population owns the same share of consumption/income. The Lorenz Curve compares graphically the cumulative share of the population with their cumulative share of consumption/income. A perfectly equal consumption/income distribution is indicated by a diagonal. The other extreme is complete inequality where one individual owns all the consumption/income. These two (theoretical) extremes define the boundaries for observed inequality. The Gini coefficient is the most commonly used measure for inequality. A Gini coefficient of 0 indicates perfect equality while 1 signifies complete inequality. In relation to the Lorenz Curve, the Gini coefficient measures the area between the Lorenz Curve and the diagonal. The Theil index measures inequality based on an entropy measure. A parameter α controls emphasis to measure inequality for higher incomes (larger α) or lower incomes (smaller α). The Theil index with parameter α = 1 is usually called Theil T while using α = 0 is called Theil L or log deviation measure. Relative and absolute income differences can be used to compare inequality dynamics over time. Usually, percentiles are used to compare incomes of different groups. For example, p90/p10 is the ratio (for relative incomes) or difference (for absolute incomes) of the average income in the 90th and 10th percentile. Source: World Bank’s Poverty Handbook. Firewood collection - Moyo District 16 FIGURE 1.9: Inequality in Uganda 1. Gini in comparison to other countries in the region (percent, latest survey year) 2. Gini over time 3. Theil over time 4. Absolute difference (Ugandan shillings) 5. Relative difference (percent) Source: 1: WDI; 2–5: Staff calculations using UNHS 1993–2013. 17 1.4 Who are the poor in Uganda? 35. Most of the poor in Uganda live in rural areas. well as to reduce social and political tensions that Nearly 84 percent of the population and 90 percent can emerge from stark differences across regions of the poor lived in rural areas in 2013. One in four (Box 1.3). rural Ugandans lives in poverty compared to just 37. In particular, two subregions in the north, the one in ten urban Ugandans. North East and West Nile subregions, have a 36. There are large and increasing regional very high poverty headcount. Almost three in variations in poverty with most of the poor four residents (74 percent) in North East subregion concentrated in the north and the east. In 2006, live below the national poverty line (Figure approximately 68 percent of the poor lived in the 1.10.2). The North East subregion is also the least Northern and Eastern regions of the country. Seven populous. Poverty is also much higher than the years later, this proportion increased to 84 percent national average in the West Nile and Mid-Northern (Figure 1.10.1). About 47 percent of the poor live subregions where 43 percent and 35 percent of in the Northern region and another 37 percent live the population live in poverty, respectively. On in the Eastern region. A focus on the Northern and the contrary, Kampala has a poverty rate of only 1 Eastern regions will be needed for Uganda to end percent and poverty is in single digits in the Central extreme poverty and boost shared prosperity as 1 and Central 2 subregions. In 2013, approximately 84 percent of the poor lived in the Northern and Eastern regions of the country. FIGURE 1.10: Where do the poor live? 1. Population in each region by welfare ventile 2. Poverty rates by subregion Source: UNHS 2013. Note: In Figure 1.10.2, the size of the circle is proportional to population size of the subregion. 18 BOX 1.3: Spatial Dimensions of Poverty Households in Uganda’s Northern, Eastern, and Western regions have much lower levels of human capital, fewer assets, and more limited access to infrastructure than households in the Central region. The Northern region is the worst, largely because the conflict took lives, damaged communities, destroyed assets, and had lasting effects on the aspirations of many individuals. Households in the north are larger and more likely to be headed by a woman and are more likely to have a household head with no education (Table 1.6). Most households own land but they are less likely to own other assets and have lower access to infrastructure services. The Eastern region also lags behind the Central and Western region in nearly all of these measures. TABLE 1.6: Human capital, asset ownership, and access to infrastructure across regions Central Eastern Northern Western Household size 4.2 5.4 5.0 4.8 Dependency ratio 101 130 134 116 Household is headed by a female (%) 30 30 35 31 Head has no education (%) 14 19 27 25 Head has some primary education (%) 43 50 41 41 Head has completed primary education (%) 9 7 8 11 Head has some secondary education (%) 19 15 12 11 Head has completed secondary education (%) 7 5 3 5 Head has tertiary education (%) 6 3 5 5 Literacy rate among 18+ year-olds (% literate) 79 60 56 72 Owns a mobile phone (%) 82 52 35 63 Has electricity (%) 40 6 3 8 Has piped water (%) 20 5 1 6 Availability of tarmac roads (%) 53 21 19 27 No toilet (%) 5 8 29 2 Owns land (%) 59 78 80 86 Source: UBOS 2013. Report on UNHS 2013. Households in the Northern region also have more limited access to markets and services. For households in these regions, distances to schools and health services are much larger as are distances to markets. The provision of agricultural extension and veterinary services is much lower and this is of concern given the reliance of these households on agriculture and livestock income. Rural financial institutions are almost entirely absent in the north. These constraints have limited the accumulation of human capital and the extent to which households can use their assets to earn a return in these regions. Household income among the bottom 40 percent is low in the Eastern and Northern regions and heavily reliant on food crops and livestock farming. Livestock income comprises 39 percent of the agricultural income of the bottom 40 percent who live in the north. In addition, rainfall is lower and more volatile in the north increasing the vulnerability of households in this region, while households in the east are particularly vulnerable to the collapse of maize prices (Chapter 4). 19 38. Those in the bottom 40 percent live in larger income groups. Households in the bottom 40 families and have more dependents than the top percent are just as likely to be headed by a woman 60 percent. Households in the bottom 40 percent as households in the top 60 percent. This means have 6 members on average compared to 4.6 in the that on average female-headed households are top 60 percent. As a result, the dependency ratio is no less likely to be poor. This is true in both rural 13 percentage points higher for those living in the and urban Uganda. However, households that are bottom 40 percent. This gap between the bottom headed by female widowers are more likely to be 40 percent and top 60 percent has remained poor than households headed by male widowers constant between 2006 and 2013 (Table 1.7). In (18 percent compared to 11 percent, significant at addition, the proportion of households headed by 10 percent). This is consistent with findings on the women has increased slightly during this period poverty of female widows in Uganda in the 1990s but this has happened for households across all (Appleton 1996). TABLE 1.7: Fertility rates and dependency ratios, 2006–13 2006 2010 2013 Bottom Bottom Top Bottom Top Top 60 40 40 60 40 60 Household composition Children ages 0 to 5 1.5 1.0 *** 1.4 1.0 *** 1.4 1.0 *** Children ages 6 to 14 1.2 0.9 *** 1.3 0.7 *** 1.2 0.8 *** Male adults ages 15 to 59 1.1 0.9 *** 1.2 0.8 *** 1.2 0.8 *** Female adults ages 15 to 59 2.1 2.1 2.0 2.0 * 1.9 1.9 Seniors v 60 0.2 0.2 ** 0.2 0.2 *** 0.3 0.2 *** Household size 6.1 5.1 *** 6.1 4.6 *** 6.0 4.6 *** Dependency ratio 136.1 98.3 *** 142.9 97.7 *** 141.8 99.6 *** Head is female 27.4% 26.6% 31.4% 29.5% 31.4% 30.7% Source: UNHS 2013. Note: Stars indicate whether bottom 40 percent and top 60 percent are significantly different using a Wald test. *** indicates significantly different at 99% confidence, ** at 95% confidence, and * at 90% confidence. 39. Ugandan households have a higher level of 2006 (Chapter 2). However, these levels of access education than in the past, but it remains low, remain relatively low by international and even particularly among poorer households. Although regional standards, with only 12.4 percent and 6.8 there has been much progress in educational percent of households having access to electricity attainment in recent years (see Chapter 2), and piped water, respectively, in 2013. In addition, many working-age adults still have low levels of there are large variations in asset ownership and education—only 23.8 percent of household heads access to infrastructure services between the rich had higher than primary education. Within the and the poor. Mobile phone ownership is only 37 bottom 40 percent of the population, this is only 11 percent among the bottom 40 percent compared percent. with 70 percent among the top 60. Almost no households in the bottom 40 percent have access to 40. Access to infrastructure services, particularly electricity or piped water, compared with 20 percent for the poor, remains low even by regional and 10 percent, respectively, in the top 60 (Table standards. By 2013, more households owned 1.8). Interestingly, more poor households report land, mobile phones, and motorcycles, and also to owning land, reflecting the predominance of accessed electricity and piped water, compared with farming as their prime occupation. 20 TABLE 1.8: Human and physical capital and livelihoods among the bottom 40 percent, 2013 Proportion of Individuals That Live in a Household in Which… Bottom 40 Top 60 Education level of the head of the household is: None 29.4 16.3 Primary 58.5 49.0 Secondary 11.4 27.3 Tertiary 0.7 7.5 The household owns the following assets: Bicycle 30.5 30.9 Mobile phone 36.7 70.4 Electricity 1.7 19.6 Piped water 0.4 10.2 Land 83.2 74.0 Main income source of the household is: Farming 52.6 38.8 Wage employment 20.4 25.4 Other source 27.0 35.7 The household owns a nonfarm business 31.6 40.0 Source: UNHS 2013. 41. Poorer households are more likely to report 40 percent (Table 1.8). In addition, although crop farming as their primary occupation. More than income is becoming less important over time it is half of the households in the bottom 40 percent still the main source of income for most households (53 percent) depend on agricultural production at the bottom of the consumption distribution, with as their main source of income compared with 39 richer households reporting higher levels of wage percent of those in the top 60. Wage employment employment income and income from nonfarm and ownership of a nonfarm business is higher household enterprises (Table 1.8). among the top 60 percent than among the bottom 1.5. Conclusion and outlook: Ending extreme poverty in Uganda 42. This chapter has documented Uganda’s year in Sub-Saharan Africa. Consumption growth impressive rate of poverty reduction in the last has slowed in recent years, but it has become two decades. Uganda’s progress, during the period increasingly pro-poor which has allowed poverty of focus of this report, was slower than from 1993 to rates to continue to decline. 2006, but still very fast. The poverty rate measured 43. However, Uganda’s progress in reducing poverty against the international line of US$1.90 PPP per day is not an unqualified success and Uganda fell by 2.7 percentage points per annum, the second fastest percentage point reduction in poverty per remains a very poor country. The low national 21 poverty rate of 19.7 percent reflects a poverty line violated. Therefore, for each scenario household that is too low. An updated poverty line would consumption is also simulated assuming for a suggest a third of Ugandans remain unable to pro-poor growth scenario in which growth rates are meet their basic needs. In addition, vulnerability to higher for the bottom 40 percent than for the top 60 poverty is high which makes it hard for individuals percent, as was the case from 2010 to 2013. to sustain gains in welfare. Moreover, poverty is 46. In the most optimistic scenario, extreme poverty increasingly concentrated in the Northern and will be almost eradicated, reduced to 4 percent, Eastern regions. by 2030. Figure 1.11 and Table 1.9 present results 44. Is Uganda on a path to end extreme poverty by from the simulation analysis detailing the trend 2030? In Chapters 3 to 7 of this report we examine in poverty rates over time under the scenarios in further detail what has driven progress in Uganda, considered. Poverty rates in 2030 range between 4 and this provides some insight into whether or not and 21 percent. The most optimistic scenario entails Uganda is on a path of sustained poverty reduction reducing extreme poverty to 4 percent by 2030, that would allow it to end extreme poverty. This which would be a remarkable achievement, given section reports simulation results to examine what that 34.6 percent of the population is in poverty in poverty rates may be in Uganda in the next 5, 10, 2013. and 15 years, if recent patterns of consumption 47. Achieving this low level of extreme poverty growth continue. As the rest of the report highlights requires both high and pro-poor growth, though, this is not guaranteed. Three scenarios something that Uganda has not been able to are identified in which the average growth rate is achieve concurrently in the last two decades. estimated based on recent history:18 The scenarios point to a number of reasons why 4 • Pessimistic scenario assumes annual average percent extreme poverty in 2030 may be an overly consumption growth of 1 percent, which is about optimistic projection. First, this scenario assumes the growth rate observed in the low growth period consumption growth rates averaging 4 percent, from 2000 to 2006. which is a growth rate for consumption that has not been observed since the high growth period of 1993 • Intermediate scenario assumes annual average to 2000. Secondly, this assumes higher growth rates consumption growth of 2.5 percent, which is about for the bottom 40 percent, something only seen the average growth rate observed over the period from 2010 to 2013. 2006 to 2013. 48. A more realistic scenario predicts an extreme • Optimistic scenario assumes annual average poverty rate of 12 percent by 2030. This still consumption growth of 4 percent, higher than the represents an impressive reduction in poverty. consumption growth rates observed since 2000, but A more realistic scenario is a growth rate of 2.5 lower than the very high rates observed from 1993 percent, the average of the growth rate observed to 2000. from 2006 to 2013, and growth that is not pro-poor. 45. Assuming the same growth rate for all 49. However, although historical trends suggest this households in the population, household scenario may be more realistic, caution should consumption is multiplied by 1 plus the growth be noted given the increasing concentration of rate for each year in the simulation. However, as intransigent poverty in the Northern and Eastern growth incidence curves indicate, the assumption regions. Regional inequality has been worsening of average growth across the population is usually 18.. The label of the scenarios (pessimistic to optimistic) refers to the average assumed growth rate. It does not imply that growth distribution across the population is ‘better’ in the optimistic scenario than in the pessimistic scenario. 22 in recent years and the majority of Uganda’s poor 50. In a scenario in which policies and investments are now concentrated in the Northern and Eastern are unable to bring about faster growth in the regions. Consumption growth rates have, on Northern region, extreme poverty in 2030 will be average, been lower in the Northern and Eastern 13 percent. A series of scenarios are conducted in regions than in the rest of the country and unless which household consumption growth rates remain this trend is reversed, assuming a growth rate of 2.5 lower in the Northern and Eastern regions. Results percent for the poorest households in Uganda is are presented in Figure 1.12 and Table 1.9. overly optimistic. FIGURE 1.11: Trends in poverty incidence FIGURE 1.12: Trends in poverty incidence for different regional growth rates Source: Staff calculations using UNHS 2013. 23 TABLE 1.9: Poverty statistics in 2030 Headcount Depth Severity 2012 34.6 10.3 4.4 Neutral Growth Pessimistic 24.1 6.6 2.7 Intermediate 12.3 3.1 1.2 Optimistic 5.9 1.3 0.4 Bottom 40% Grow Faster Pessimistic 22.3 2.6 0.9 Intermediate 10.6 2.6 0.9 Optimistic 4.2 0.9 0.3 Region-specific Growth Rates Pessimistic 24.2 6.8 2.8 Intermediate 13.4 3.5 1.3 Optimistic 7.4 1.7 0.6 Source: Staff calculations using UNHS 2013. 24 25 NON-MONETARY DIMENSIONS CHAPTER: 2 OF POVERTY IN UGANDA Uganda’s progress in reducing income poverty is strongly reflected in some non-monetary indicators of welfare, although the country still has a long way to go on some dimensions. 51. Chapter 1 highlighted the impressive performance in reducing monetary poverty over the last decade. The proportion of the population living under the national monetary poverty line declined from 56.4 percent in 1993 to 19.7 percent in 2013. 52. Poverty is multidimensional in nature and there are some limitations to relying solely on the monetary poverty measures. It has been well documented in literature that well-being is a broad description of the state of people’s living conditions (for example, McGillivray and Clarke 2006; Saunders 2005). Beyond monetary poverty, it is important to have a more comprehensive understanding of how the country has performed on other dimensions of well-being. Socioeconomic indicators of well- being can provide a valuable complement to existing monetary measures of poverty, and this would allow to better target programs and policies to reach those who need them the most. Non-monetary aspects of well- being can complement the monetary measure. 53. This chapter analyzes levels and trends of non-monetary poverty indicators in Uganda focusing on selected dimensions of housing conditions, infrastructure services, physical capital, and human capital. The selection of non-monetary indicators was guided by 26 literature on multidimensional poverty (See Etang although the country still has a long way to and Tsimpo 2016 for more details). Although go on some dimensions. Ownership of modern very comprehensive, the list of non-monetary assets and the share of households using improved indicators analyzed in this chapter is not exhaustive. roofs increased over the last decade. Education The indicators used are categorized into four outcomes have improved as well, but the significant broad dimensions: (a) housing conditions, (b) increase in primary enrollment rates has yet to infrastructure services, (c) physical capital, and (d) translate at higher levels. There was a substantial human capital. decline in all components of child mortality, but malnutrition continues to be widespread. The 54. The analysis shows that Uganda’s progress in evidence presented in this chapter points to two reducing income poverty is strongly reflected areas that require special attention: infrastructure in some non-monetary indicators of welfare, and educational outcomes beyond enrollment. 2.1 Housing conditions 55. The share of households using improved roof measure of poverty) regarding housing materials has expanded, but improvements in construction materials in 2006, 2010, and 2013. wall and floor materials have stalled. Figure 2.1 The most visible distinction between the poor shows that usage of improved roof materials has and non-poor was the materials used to roof the slightly increase between 2006 and 2013, providing house (Figure 2.1.2). The share of households with evidence for rising living standards, including for improved roof material was substantially (at least rural households (Figure 2.1.1). At the national 35 percentage points) higher among the non-poor level, the share of households with improved roof for each of the three years. The materials used material went up by 7 percentage points, from 61 for the walls and floor show significant variations percent in 2006 to 68 percent in 2013. Improved between poor and non-poor households. The wall material went up by 4 percentage points at share of households with improved wall and floor the national level and improved floor material by materials was 28–30 percentage points higher only 2 percentage points. Interestingly, the slight among the non-poor across 2006 and 2013. An rise in improved housing conditions between 2006 important point, worthy of note, is that the gap and 2013 seems to have occurred mainly for the between poor and non-poor households increased roof of the house,19 a bit more so for households in slightly between 2006 and 2013 with regard to the rural areas (by 5 percentage points) than in the improved roof and wall materials, although it was urban areas (3 percentage points). The majority of stable for improved floor materials. Increases in urban households have cement floors, while less poverty rate are associated with decreases in the than 20 percent of rural households do so. The use of improved roof materials, and vice versa. The fact that the majority of rural households continue Northern region with the highest poverty rates in to live in dwellings with earth (mud) floors is a 2006, 2010, and 2013 was also the region with low concern, as this can pose health risks. use of improved roof materials during the same periods. The opposite is true for the Central region 56. Stark differences persist between poor and with high levels of improved roof materials and low non-poor households (based on the monetary poverty rates in all three years. 19. The type of roof is often used in developing countries as a proxy-indicator for poverty, among others for targeting purposes of unconditional cash transfer program. 27 FIGURE 2.1: Distribution of households by main type of construction materials (%), 2006–2013 1. By location 2. By poverty status Source: UNHS 2006, 2010, and 2013. Notes: The definition of improved roof material includes iron sheets and tiles. Improved wall material includes burnt bricks with mud, burnt bricks with cement, cement blocks, and stones. Improved floor includes cement and mosaic or tiles. 2.2 Infrastructure services 57. Access to improved water has expanded overall Eastern regions recorded the most improvement during the past decade, but regional and over this period. The same is true for the second, socioeconomic inequities in access persist. third, and fifth consumption quintiles. Improved water sources are broadly available, with access having increased modestly over the 58. Uganda’s access to an improved source of last decade.20 At the national level, the share of drinking water was slightly above expected households with access to improved sources of levels and progress over time was faster than drinking water increased by 4 percentage points the expected level. Access to improved sources of between 2006 and 2013. While nearly three drinking water was relatively high by international quarters of households in Uganda had access to standards. Also, Uganda performs better than the improved water sources of drinking water in 2013, a average country in Sub-Saharan Africa and better substantial share of households still lacked access than its East African Community counterparts in to this basic need. Access among residents of 2012. With respect to the pace of progress over Kampala is almost universal (95 percent). In other time, cross-country correlations with GNI per urban areas, 84 percent of households have access capita indicate that progress in access to improved to improved water sources in 2013, compared to water sources was faster than could be expected, 67 percent in rural areas. Access to improved water given the change in GNI during 2000–2012. The increased between 2006 and 2013 across all regions performance may be related to a focus on low-cost and consumption quintiles. The Western and the type of supply in rural areas (borehole), under a 20. The World Health Organization (WHO/UNICEF Joint Monitoring Programme) defines ‘improved’ sources of drinking water as including piped water into the dwelling, piped water into a yard/plot, a public tap or standpipe, a tube well or borehole, a protected dug well, a protected spring, bottled water, and rain water. ‘Unimproved’ sources of drinking water include an unprotected spring, an unprotected dug well, a cart with small tank/drum, a tanker-truck, and surface water (WHO and UNICEF 2006). 28 pro-poor strategy. The fact that access to improved of drinking water is associated with increases in water sources increased as poverty declined during income (GNI per capita). There does not seem to be the past decade is probably not surprising given a significant gender difference with regard to access a high correlation between the two, according to to piped water, with about 8 percent of female- cross-country data for low- and middle-income headed households having piped water compared countries (Figure 2.2). Access to improved sources to 7 percent for male-headed households. FIGURE 2.2: Access to improved water source vs. GNI per capita Source: Gable, Lofgren, and Osorio-Rodarte (2015) 59. Sanitation remains a serious issue as only a small are more likely to have access to improved minority of households has adequate sanitation. sanitation compared to households in rural areas. Furthermore, there is a strong link between poverty The data show that 19.7 percent of households in and the presence of improved toilet facilities. Kampala and 18.6 percent in other urban areas had Figure 2.3.1 provides estimates of the share of the access to improved sanitation against 12.3 percent population with access to improved sanitation in rural areas. This is also the case for shared but based on UNHS 2013 data (due to changes in improved sanitation (50.5 percent of households questionnaire categories, it is difficult to provide a in Kampala and 36.1 percent in other urban areas trend over time). The data suggest that only 14.0 compared to 9.4 percent in rural areas). Looking at percent of households have access to improved sanitation from a gender dimension, UNHS 2013 sanitation. If unimproved facilities are split between data suggest that the share of female-headed shared but improved facilities and unimproved households that have no toilet is slightly higher facilities, the proportion of households with a than the corresponding number for male-headed shared improved facility is 17.3 percent. Clearly, households (12 percent and 9 percent, respectively). most households do not have access to adequate This finding is consistent with evidence of a strong sanitation, and when they do have access, in most correlation between poverty and lack of toilet cases the facilities used are shared, often by too facilities—poor households are mostly those many households. A rural/urban breakdown of without a toilet facility, and it is known that female- access to sanitation shows that urban households headed households are more likely to be poor. 29 60. In 2011, Uganda’s access to improved sanitation for those in urban areas. A big challenge remains was slightly above expected levels. Overall, in terms of access to improved sanitation facilities Uganda performs slightly better given the level in urban areas, where Uganda is performing below of GNI. However, access to improved sanitation expectation compared to other countries (as shown facilities remains low by international standards in Figure 2.3.2). FIGURE 2.3: Percentage of households using an improved latrine 1. By welfare quintiles 2. Improved latrine vs. GNI per capita, urban population only Source: UNHS 2013. Source: Gable, Lofgren, and Osorio-Rodarte (2015). 61. Residential coverage of electricity remains Wodon 2014b). There has been a recent increase in very low. Only one out of seven households used alternative forms of electricity coverage, especially electricity for lighting in 2013. At the national through solar generation, but overall coverage level, 14 percent21 of households in Uganda use rates still remain very low. Tsimpo and Wodon electricity for lighting.22 Figure 2.4.1 indicates that (2014b) argue that the slight increase in electricity there was a slight increase in the percentage of coverage, despite increases in connections, is households across Uganda that used electricity as because of population growth and a reduction in the main source of fuel for lighting over the survey household size as well. periods from 10 percent in 2006 to 12 percent in 2010 and then to 14 percent in 2013, resulting 62. There is a strong correlation between poverty in 4 percentage points increase in electricity and use of electricity, and connection rates use between 2006 and 2013. While UMEME’s are virtually nonexistent in the bottom 40 distribution network has grown over the last few percent. As Figure 2.4.1 shows, access to electricity years, residential coverage rates remain very low decreases with poverty. Not surprisingly, electricity due to limited access rates at the neighborhood coverage rates are much higher among households or village level and limited take-up by households in the top 60 percent of the distribution than of the service when access is (at least in principle) among those in the bottom 40 percent. About available in the area where they live (Tsimpo and 17 percent households, on average, for the top 21.. This number is based on the UNHS 2013 survey and is consistent with findings of the Energy for Rural Transformation Survey 2012 and the UDHS 2012 which found that electricity is used for lighting by about 15 percent of households (UBOS 2014, UNHS 2013 Report) 22. Electricity sources include national grid, solar, personal generator or community/thermal plant. 30 60 percent of the distribution use electricity for and children spend a considerable amount of time lighting, whereas connection rates are virtually on households chores, including collecting water nonexistent among the bottom 40 percent. and fuel, cooking, and taking care of children and the elderly (Blackden and Wodon 2006; WHO and 63. There exist stark differences in electricity UNICEF 2006). A connection to the electricity or usage across rural and urban households. piped water network eases access to timesaving During the last decade, more than 40 percent of technology and therefore reduces domestic urban households used electricity for lighting work, especially for women. Tsimpo and Wodon compared to a mere 4 percent in rural areas. For (2014b) use UNHS 2013 data to demonstrate that the rural households, the number has remained if electricity or piped water were provided to all fairly stable over the last decade. The share of households living in areas where the network is urban households that used electricity for lighting available at the neighborhood level, connections increased from 41 percent in 2006 to 48 percent for households not yet connected would enable in 2010 and then fell by 8 percentage points to 40 women to increase market work by up to two hours percent in 2013. It is surprising that the urban usage per week. This has additional impact on household rate fell substantially between 2010 and 2013. The income and poverty. data show that the gain in access to electricity from 2006 to 2013 happened in rural areas. According 64. Although the share of Uganda’s population to UNHS 2013 data, there is only a slight gender with electricity access has improved slightly difference with respect to access to electricity, during the last decade, it is still far below what with 11 percent of female-headed household is expected.23 Access to electricity in Uganda is having access compared to 12 percent for male- one of the lowest in the world (Figure 2.4.2). Access headed households. Availability of electricity to electricity remains very low even by regional and network (piped) water may help in reducing standards, with only 18 percent of the population time spent on domestic chores and increase having access in 2012. This is half the average for economic opportunities and earnings, especially Sub-Saharan Africa and almost a fifth of the world for women, ultimately reducing poverty. Women average. FIGURE 2.4: Access to electricity (% of population) By location and poverty status 2. Uganda compared to other countries Source: UNHS 2013. Source: Gable, Lofgren, and Osorio-Rodarte (2015). 23. Data on electricity access are provided by the International Energy Association. The access indicator refers to the population share with access to electricity in their homes. While this definition leaves out access to production sectors, an indicator based on a broader definition would paint a similar picture. 31 2.3 Physical capital 65. Ownership of modern assets increased while more households have a mobile phone in 2013 ownership of traditional assets deteriorated. compared to 2010. This is probably not surprising Figure 2.5 presents the distribution of households given that mobile phone ownership has increased by ownership of some of the key assets. More substantially across Africa. As with land ownership, households own land, mobile phones, and a gender gap appears in terms of mobile phone motorcycles, at the expense of pedal cycles. Land ownership. The share of households that own a ownership information was not collected in 2006. mobile phone is substantially higher when the household head is a male (66 percent) than if it is a 66. The proportion of households who owned a female (50 percent). As male-headed households piece of land appears to remain stable between are generally richer, they can more easily afford the 2000 and 2013. However, land ownership cost of purchasing and maintaining a mobile phone increased for the poor. About three-quarters than poorer households can. of households own a piece of land. This share increased between 2010 and 2013, particularly for 68. Ownership of motorcycles is low, but on the rise. the poor who are mostly involved in agriculture. The share of households owning a motor cycle This can be considered as a positive sign as it remains low at 6.7 percent in 2013. However, this can potentially contribute to improvement of represents a major improvement from the mere 2.6 their productivity and living standards. Even if the percent in 2006. land was not directly used for agriculture, land ownership, if accompanied by formal titles, can 69. The ownership of mobile phones and help households to access credit that could be motorcycles appear to have improved used to improve their welfare. A gender breakdown substantially more among the well-off (top 60 of the data shows that the share of male-headed percent). As illustrated in Figure 2.5, the proportion households that own land in 2013 is higher than of bottom 40 percent households having a mobile the corresponding number for female-headed phone has grown substantially, by 35 percentage households (about 86 percent and 76 percent, points, on average, compared to 46 percentage respectively). Given that land is a productive asset points for the top 60 percent households.24 With and agricultural productivity is lower among regard to motorcycles, increase in ownership female-headed households (see Chapter 4), finding between 2006 and 2013 remained fairly stable ways to improve land ownership among female- among the bottom 40 percent households while it headed households would benefit efforts to close increased by 5 percentage points among the top 60 this gap and reduce poverty among households percent. headed by females in Uganda. 70. Conversely, ownership of more traditional 67. There was a notable increase in the proportion assets such as bicycles has declined. It seems of households who own a mobile phone. About that households have replaced bicycles by a more 170 percent more households owned a mobile modern transport mode, as can be seen from the phone in 2010 than in 2006, and 30 percent decline of bicycles and increase of motorcycles. 24. Asset ownership growth could also be shown in terms of percentage change. However, it would be more informative to show in percentage point terms rather than percentage change, because a large relative increase from a very small base may not be very meaningful in an absolute sense. 32 This is consistent with Seff et al. (2014), who, using are not necessarily an indicator of declining Tanzania National Panel Survey data, show that levels of wealth. Rather, the rise in motorcycle households tend to replace traditional devices ownership, coupled with the decline in bicycle such as radios and bicycles by more upgraded ownership, supports the notion that these goods goods, such as TVs or motorcycles. Thus, the are substitutes of each other. declining levels of bicycle ownership observed FIGURE 2.5: Changes in asset ownership, by consumption quintile, 2006–2013 (absolute numbers) Source: UNHS 2006 and 2013. Note: Changes are calculated between 2006 and 2013, except for land, which is between 2010 and 2013 because land ownership data was not collected in 2006. 2.4 Human capital EDUCATION25 71. Adult literacy rates are high in Uganda, given given that this is a stock variable. Adult literacy rates its income level, but have not changed much are substantially higher among males than females. over time although progress in youth literacy One might expect to see more rapid change in rates has been faster. Adult literacy is substantially literacy rates among young adults. The youth higher when compared to countries with a similar literacy rate has improved over time, and this is the GNI level (Figure 2.6A.2). The national adult literacy case for both males and females between 15 and 24 rate (for those ages 18 years and above) stands at years old. It is found that male and female youths 68 percent (Figure 2.6A.1), and has been fairly 26 have similar levels of literacy (Gable, Lofgren, and stable between 2006 and 2013 as one might expect Osorio-Rodarte 2015). 25. This report looks at both stock variables such as adult literacy rates and flow variables such as school enrollment. Stock variables should not be expected to change much in the short run, while flow variables should. 26. Adult literacy rate: the percentage of the population, ages 18 and above, who can, with understanding, read and write a short, simple statement on their everyday life. 33 FIGURE 2.6A: Trends in adult literacy rates (%) 1. Adult literacy rate (%) 2. Adult literacy rate vs. GNI per capita Source: UNHS 2006, 2010, and 2013. Source: Gable, Lofgren, and Osorio-Rodarte (2015). 72. Gender gap in literacy rate has closed. Given than women for the older cohorts. This is probably the averages in Figure 2.6B and the large gender partly because schools are more accessible now disparity, Figure 2.7 graphs the literacy rate across than in the past decades, enabling more young cohorts, suggesting that younger males and females women to study now while they could not do so in are equally literate while there are more literate men the past due to lack of nearby school facilities. FIGURE 2.6B: Literacy gap across cohorts (%) Source: UNHS 2013. 34 73. Net enrollment in primary schools are high negative impact on schooling. For instance, Nyqvist and have increased over time. Primary school (2012) shows that in Uganda, a decrease in rainfall enrollments (6–12 years) increased slightly between is associated with a reduction in female enrollment 2006 and 2013. According to UNHS data the primary in grade 7 (primary school). However, this effect is net enrollment rate increased from 84 percent in significant for older girls only. There is no significant 2006 to 86 percent in 2013 (Table 2.1). This is up effect of rainfall variation on the enrollment of boys from 67 percent in 1995 and 79 percent in 2000. and young girls. Table 2.1 indicates that there is Interestingly, primary net enrollment deteriorated no marked difference in male versus female net in 2010 before recovering in 2013. The same holds enrollment. It should be noted that, unlike Table 2.1, for a number of indicators on education, and there which covers all grades and ages, Nyqvist focused might be a common explanation for the oscillation. on grade 7 only and split boys and girls by age. Exogenous shocks affecting incomes often have TABLE 2.1: Trends in net enrollment rates in primary schools (%) Boys Girls Total 2006 84 85 84 2010 82 83 83 2013 85 87 86 Source: UBOS reports based on UNHS 2006, 2010, and 2013. 74. According to cross-country regression analysis, everywhere, boys and girls alike, will be able to Uganda’s net primary enrollment rates are complete a full course of primary schooling. above the expected level when compared to other countries with similar incomes. Primary 75. Uganda has been successful in enrolling children school enrollment rates are on the rise and higher in primary school but completion rates are lower than expectations, given the GNI level (Figure than expected, and the trends show that the 2.7.1). The expansion of enrollment in primary situation deteriorated over the last decade. The schools was observed for both males and females primary completion rate has generally fallen since (Gable et al. 2015). The magnitude of increase the beginning of the 2000s (Figure 2.7.2). Ideally, in net primary enrollments for boys and girls completion should be timely. This means that most was similar. The high primary school enrollment of the population in the targeted age group (12 years) rates among both poor and rich children reflect should complete the last grade at the age of 12 the benefits of the UPE program that was years. Uganda’s gross primary completion rate was introduced by the GoU in 1997. Under the UPE 53 percent in 2011. This is mainly due to a very high program, all tuition fees and Parents and Teachers primary school dropout rate of 75.2 percent. When Association (PTA) charges for primary education compared with its income peers (GNI per capita), were abolished to ensure that by 2015 children Uganda’s primary completion rate is very low. 35 FIGURE 2.7: Net enrollment and primary completion rates 1. Net enrollment in primary vs. GNI per capita 2. Primary completion rate vs. GNI per capita Source: Gable, Lofgren, and Osorio-Rodarte (2015). Source: Gable, Lofgren, and Osorio-Rodarte (2014). 76. Enrollment in secondary schools remains very completion rates are very low in Uganda. Perhaps low, meaning that the increase in primary the low completion rates are because parents school enrollment has yet to translate at higher cannot continue investment (for example, when a levels. Here, the analysis focuses on the out-of- shock occurs), or they do not see the investment school rate (that is, the inverse of net enrollment).27 being worthwhile (due to perceived low returns, Out-of-school rate stands at 23 percent in 2011. child’s poor performance, and so on). Second, cost This is within the expected level when compared seems to be a very important factor preventing to other countries with similar incomes (Figure many children from attending secondary school, 2.8.1). Secondary enrollment rates remain low, and it is the main reason for dropping out (Figure regardless of the Universal Secondary Education 2.8.2). On the other hand, almost no child stated (USE) program introduced by the GoU in 2007. physical accessibility (that is, distance) or that Although secondary schools tuition fees were further schooling was not available. Third, the abolished, students still have to pay boarding fees, other major reasons are related to attitude toward uniform costs, and for school materials, among education. These include children not willing to others costs. This is reflected in the estimated attend, pregnancy and poor academic progress, share of monthly expenditure on education, which and parents not wanting the child to continue decreased from 9.6 percent in 2006 through 8.5 school. This negative attitude toward education percent in 2010 to 7.5 percent in 2013. is mostly seen among children in the bottom 40 percent. Finally, an economic shock is the other 77. The low secondary enrollment rates are due to main reason for dropping out of secondary school, several factors, including the poor performance with about 11 percent of dropouts citing sickness/ at primary level, affordability, and attitude/ calamity in family as the most important factor tradeoffs. First, not enough children complete preventing them from attending school. primary school. As shown above, primary 27. The out-of-school rate for children of lower secondary school age is defined as the number of children of official lower secondary school age who are not enrolled in lower secondary school expressed as a percentage of the population of official lower secondary school age (Gable et al. 2014). 36 78. Child marriage and early pregnancy have a marriage and support girls who marry early.28 large negative impact on education attainment, Curbing early marriage and pregnancy will also especially for girls. While primary completion help reduce the fertility rate and, subsequently, the rates are low in Uganda, they have converged for dependency ratio thus affecting welfare positively. boys and girls. This is mainly because the male completion rate is declining over time (Gable et 80. There is a strong correlation between poverty al. 2014). As documented by Wodon et al. (2016a), and education. The Central region has the lowest child marriage and early pregnancy appeared to percentage of persons with no formal schooling be one of the main reasons why girls drop out of together with the lowest poverty rate. On the other school prematurely. The issue of early pregnancy is hand, the high share of people with no formal mentioned by 16.2 percent of parents as the main education in the Northern region is associated reason for girls dropping out. strongly with high poverty rates in the region. Finding that poverty is strongly correlated with 79. These results have policy implications. education has policy implications. Promoting Obviously, free tuition alone is not enough policies and programs to achieve UPE as well as for primary completion rates and secondary promoting transition from primary to secondary, enrollment rates. Efforts to improve secondary and, subsequently, tertiary education will be school enrollments must start with programs important for poverty reduction. Education that would boost primary school completions. equips people with the needed skills to transition In addition, social protection programs that can from subsistence agriculture to more productive enable households to cope with negative shocks activities. Furthermore, a better-educated would enable their children to stay in secondary population would likely be more productive, school when a shock hits. Various types of participating more efficiently in promoting interventions can be also considered to delay economic growth and poverty reduction. FIGURE 2.8: Out-of-school rate for lower secondary and reason for dropping out of school 1. Out-of-school rate for lower secondary vs. 2. Main reasons for dropping out of school, GNI per capita children ages 13–18, 2013 Source: Gable, Lofgren, and Osorio-Rodarte (2015). Source: Authors’ calculation, based on UNHS 2013. 28. Such interventions will include: (a) empowering girls with information, skills, and support networks; (b) educating and mobilizing parents and community members; (c) enhancing the accessibility and quality of formal schooling for girls; (d) offering economic support and incentives for girls and their families; and (e) fostering an enabling legal and policy framework. See Wodon et al. (2016) for more details. 37 HEALTH AND NUTRITION 81. According to cross-country regression analysis, find that all these indicators are on a declining Uganda’s under-five mortality rates seem to trend since 2000. Uganda achieved the Millennium be exactly at the expected level. There has Development Goals target of reducing child been a remarkable decline in all components mortality by two-thirds by 2015 before the target of early childhood mortality over the 15-year date (compared with 1990). period preceding UDHS 2011. There have been substantial decreases in early childhood mortality 82. Under-five mortality is significantly higher in rates (Figures 2.9, 2.10.1, 2.10.2, and 2.11.1). Infant rural areas than in urban areas. The mortality mortality (which measures the probability of rates were lowest in Kampala and highest in the infants dying before their first birthday per 1,000 Mid-Northern. This shows that there is a relation live births) dropped from 88 in 2001 to 76 in 2006 between child mortality and poverty, with Kampala and 54 in 2011. For the five years preceding UDHS having the lowest poverty rates and the Mid- 2011, the child mortality rate was 38 per 1,000 live Northern one of the subregions with high poverty births. This implies that one in every twenty-six levels. Indeed, under-five mortality rates were children, who survived the first birthday, does lowest among the top 60 percent. not live to the fifth birthday. Under-five mortality, which measures the probability of children dying 83. Uganda has also made considerable progress between birth and the fifth birthday, stood at 90 to reduce maternal mortality over the past two in 2011, having declined from 152 in 2001 to 137 decades. Uganda’s maternal mortality rate declined in 2006. Declining trends were also observed for from 600 to 440 deaths per 100,000 live births neonatal and post-neonatal rates. It is positive to between 1990 and 2011 (Figures 2.10.2 and 2.11.2). FIGURE 2.9: Trends in childhood mortality, 2001–2011 Source: UBOS UDHS Reports 2001, 2006, and 2011. Notes: According to UBOS and ICF (2012) age-specific mortality rates are categorized and defined as follows: (a) neonatal mortality: the probability of dying within the first month of life; (b) post-neonatal mortality: the arithmetic difference between neonatal and infant mortality; (c) infant mortality: the probability of dying before the first birthday; (d) child mortality: the probability of dying between the first and the fifth birthday; and (e) under-five mortality: the probability of dying between birth and the fifth birthday. All rates are expressed per 1,000 live births except for child mortality, which is expressed per 1,000 children surviving to 12 months of age. 38 FIGURE 2.10: Under-five mortality by region and maternal mortality rates 1. Under-five mortality by region and 2. Maternal and under-five mortality rates, consumption quintile, 2011 1990–2012 Source: UDHS 2011. Source: Gable et al. (2014). FIGURE 2.11: Maternal and under-five mortality rates in Uganda and international comparison 1. Under-five mortality rate vs. GNI per capita 2. Maternal mortality rate vs. GNI per capita Source: Gable et al. (2015). Source: Gable et al. (2015). 84. Anthropometric indicators for young children (at 45 percent). It came down to 38 percent in show some improvement since 1995, but the 2006 and dropped further in 2011 (to 33 percent). trends are uneven and malnutrition continues Nevertheless, the level of stunting remains quite to be widespread. Stunting, defined as low height high (Figure 2.12.1). Childhood stunting has for age and an indicator of chronic malnutrition, long-term effects that are often irreversible. It can was consistently high between 1995 and 2001 cause delayed motor function and diminished 39 cognitive ability; and children with low height-for- 86. However, the puzzle revealed by this analysis age in their early years may exhibit poor academic is that the patterns of the nutritional outcomes performance later in life (Seff et al. 2014; UNICEF are not as expected across regions and welfare. 2007). In Uganda, wasting decreased slightly Stunting levels are higher in rural areas. Stunting from 7 percent in 1995 to 5 percent in 2001, but incidences are lowest in Kampala, followed by has remained fairly unchanged since then. The the North East and Eastern subregions (Figure incidence of underweight in Uganda stands at 14 2.13). Finding that the North East and Eastern percent in 2011, decreasing by 8 percentage points subregions outperform other subregions in terms since 1995, and has been declining gradually over of stunting levels (with the exception of Kampala) the periods from 1995 through 2001 and 2006 to is surprising and not as expected. The North East is 2011. The outlook seems positive, particularly the poorest subregion and the Eastern subregion for stunting and underweight. The results show is one of the poorest subregions of Uganda. The a downward trend in the percentage of children findings in Chapter 4 do point to the fact that there stunted and underweight over the last two UDHSs, are improvements in reducing malnutrition when but the percentage of children who are wasted has income increases, but the findings here show that remained stable. the level of malnutrition is not correlated with the level of poverty, as would be initially expected. This 85. The incidence of being underweight is lower may in part be due to the diverse possible causes in Uganda than in other low-income countries, of malnutrition, including not enough nutrients and progress was recorded over the last in available staple foods, lack of knowledge of decade. According to cross-country data for low- adequate feeding, and lack of safe water and and middle-income countries, there is a strong sanitation. However, the low correlation between correlation between poverty and malnutrition poverty and nutrition outcomes has been observed (Figure 2.12.2).29 Thus, it is not surprising to find that in other contexts and cannot be fully explained. both poverty and malnutrition have declined during Speaking in the Lionel Robbins Memorial lectures in the recent decades of strong growth in Uganda. The December 2015, Deaton talked about poverty and expected number is 19.6 percent for a country with inequality in India.31 He noted that malnutrition is Uganda’s income per capita (Gable et al. 2014).30 present at all levels of consumption and more so Uganda’s current underweight rate of 14.1 percent among the rich than the poor. This is consistent of children under five years of age is slightly below with the finding in Uganda. Deaton noted that it the expected value. This means that incidences is not clear why this is the case in India, and this of being underweight in Uganda are fewer than in is true for Uganda also. Further research work is comparable countries. Perhaps this progress was needed to better understand and explain the puzzle partly due to the benefits of the Uganda Nutrition of why rates of malnutrition are not correlated with Action Plan that was launched in 2011. poverty. 29. The correlation coefficient between the two variables is 0.60 in non-log form and 0.72 in log form. 30. The under-five underweight rate is defined as the percentage of children under the age of five years whose weight for age is more than two standard deviations below the median for the international reference population ages 0–59 months (WDI). 31. http://www.livemint.com/Politics/jYcyQ0VZ6JZNhejdOpdywL/Angus-Deaton-on-India-at-the-LSE.html 40 FIGURE 2.12: Malnutrition prevalence, underweight (% of children under five years) versus income per capita 1. Trends in nutritional status of children 2. Malnutrition prevalence, underweight (% of under five years (%) children under five years) versus income per capita Source: UDHS 1995, 2001, 2006, and 2011. Source: Gable et al. (2014). FIGURE 2.13: Nutritional status of children under 5 years, by region and consumption quintile, 2011 (in percent) Source: UDHS 2011. 87. Objective and subjective indicators of poverty classify the household into very poor, poor, neither are similar. The UNHS 2013 contains information poor nor rich, rich, where would you put your own on people’s perceptions of poverty. Evidence household?” The results in Table 2.2 suggest that shows that Uganda has been successful in reducing the level of subjective poverty (16.1 percent) is not poverty in the last decade. However, do people much different from the income poverty rate of 19.7 necessarily feel better-off? It would be interesting percent. The majority of households (92.6 percent) to check whether people classified as poor based that are classified as income poor indeed are either on income actually consider themselves as poor poor or very poor. when asked the question: “If you were asked to 41 TABLE 2.2: Perceptions about poverty (%) Self-assessed Poverty Income Poverty Status Total Status Poor Non-poor Insecure Middle Class Very poor 36.3 17.3 8.2 16.1 Poor 56.5 61.2 47.3 54.1 Neither poor nor rich 7.1 20.9 42.8 28.8 Rich 0.1 0.5 1.7 1.0 Source: UNHS 2013. 2.5 Conclusion 88. Uganda’s progress in reducing income poverty in power generation to improve access to electricity. is strongly reflected in some non-monetary Usage of improved sanitation is very low, and indicators of welfare. Cross-country regressions improving access to this facility will be important suggest that Uganda performs well on improved for the population well-being. There is also a need water, adult literacy, child and maternal mortality, to increase primary education completion rates, and child nutrition. as well as secondary education enrollment and completion rates, especially for girls, by addressing 89. The evidence presented in this chapter points to issues related to early marriage/pregnancy. As it areas where the country is performing less than will be illustrated in the next chapter, improved expected and which require special attention: educational outcomes are important for improving access to electricity and improved sanitation and people’s income generation capacity, which can lift education. The GoU needs to improve investment many out of poverty. 42 43 HOW DID UGANDA REDUCE CHAPTER: 3 POVERTY? Agriculture was the main driver of poverty reduction. Other important factors that contributed to poverty reduction include increased peace and stability in the North, urbanization and education. 90. This chapter examines the factors behind Uganda’s success in reducing poverty from 2006 to 2013. It relies on analysis of the panel survey that has followed the same households through this period (UNPS) and decomposition analysis using the UNHS. The advantages of panel analysis for assessing drivers of poverty reduction are outlined in Box 1 and the decomposition methods used in this chapter are summarized in Box 3.1. 91. The findings highlight the importance of agriculture, urban migration, and modest gains in education. It also highlights the limited role of structural change since 2006, the persistently high dependency ratios which held back poverty reduction, and limited spending on safety nets, which have resulted in change in the distribution of household consumption having little direct impact on poverty reduction or on improving vulnerability. 44 3.1 Growth, not redistribution, drives poverty reduction in Uganda 92. Public transfers to households are negligible households in Uganda, in comparison to other in Uganda. Figure 3.1 shows the proportion of countries in the region. Uganda’s total spending households reporting receiving any pensions, on social security in 2013 was 1 percent of GDP insurance, social security benefits, and other compared to an average of 2.8 percent for other transfers across the income distribution. Less countries in Africa. Of that 1 percent, only 0.4 than 10 percent of households at any point in the percent was spent on direct income support to income distribution receive these transfers. The poor households, compared with 1.1 percent in proportion of poor households receiving transfers other low-income countries in Africa (Uganda is only 5 percent. All incomes from pensions, Systematic Country Diagnostic). In addition to insurance, scholarships, and alimony are included, low spending on public transfers, there is, more and this may include private as well as public broadly, a limited use of fiscal policy to redistribute sources and as such overestimate the proportion incomes. De facto tax rates are very low in Uganda. of households receiving state transfers. Only 4.5 The International Monetary Fund 2013 Article IV percent of the total population received any kind report documents that Uganda faces one of the of direct income support and only 5 percent of the highest revenue gaps among Sub-Saharan African working population is part of a pension scheme. countries. Redistributive fiscal policy thus plays a limited role in directly reducing inequality and 93. This reflects a limited use of fiscal policy addressing poverty. Box 3.2 discusses how, as oil to directly improve the incomes of poor revenues increase fiscal space, this could change. FIGURE 3.1: Limited public transfers for Ugandan households Source: Staff calculations using UNHS 2013. 94. Rates of informal redistribution are much income. The data available indicates that only 4 higher, but remittances and transfers comprise percent to 8 percent of household income (or 3 a small share of income. Many Ugandans—32 percent to 6 percent when compared to reported percent to 53 percent of all households—report consumption) comprises transfers received from receiving transfers or remittances from family others. Transfers and remittances are a more and friends. However, as Figure 3.2 indicates, important share for the top 60 percent than for the these transfers comprise a small share of bottom 40 percent. 45 FIGURE 3.2: Informal transfers are a prevalent, but not important, source of income Source: Staff calculations using UNHS 2013. 95. Given the limited role of public and private growth, and migration have brought about poverty transfers as a source of income for poor reduction in Uganda, and how they can continue to households, growth in labor income is what drive gains in the future. drives poverty reduction in Uganda. Section 3.2 examines Uganda’s demographic change in 96. Although fiscal policy is not redistributing the recent past and how the share and location income to directly reduce poverty, public of the working age population has changed and spending does play a role in facilitating poverty contributed to poverty reduction. Section 3.3 reduction through the provision of basic examines the type of labor income growth that services. The contribution of education and public Uganda has experienced and why this has been utilities is considered in section 3.4 and analysis on good for poverty reduction. Chapters 4 to 6 examine how to improve the quality of service delivery for labor income growth in further depth, examining poverty reduction is discussed in Chapter 7. how agricultural growth, rural non-agricultural 46 BOX 3.1: What does decomposing changes in poverty entail? In this chapter, the results of two decomposition methods are presented. The first method is the Recentered Influence Functions (RIF, Firpo et al. 2009) in which traditional Oaxaca-Blinder decompositions are applied to different percentiles of the consumption distribution. This allows an assessment of the amount of poverty reduction that can be accounted for by changes in the characteristics of households and individuals (‘endowments’) compared to the changing nature of the Ugandan economy and poverty. The second method, the Ravallion and Huppi (1991) inter-sectoral decomposition method quantifies how much poverty reduction among different groups or movement between different groups accounts for national poverty reduction. Both decomposition methods rely on defining a counterfactual scenario and estimating what would have happened to poverty had the counterfactual scenario occurred. By defining a counterfactual scenario, the changes that have been important to overall poverty reduction can be quantified. Figure 3.3 depicts how this can work for two different counterfactual scenarios. In the Ravallion and Huppi method, the focus is on a counterfactual of no change in the proportion of population in different sectors; and a counterfactual of no change in poverty among people in a given sector. These counterfactuals are used to examine the amount of poverty reduction that took place within sectors (as if sectors had not changed) and the amount of poverty reduction that took place because of people moving between sectors. In the RIF analysis, the focus is on a counterfactual of a constant relationship between endowments and poverty in Uganda over 2006 to 2013. This counterfactual is used to determine which changes in endowments could have contributed to poverty reduction, and how much poverty reduction could have changed because of a changing relationship between poverty and endowments. The latter is sometimes referred to as changes in the returns to endowments, but really it represents how the conditional correlation between a given endowment and consumption has changed. The RIF decomposition is carried out at five points of the distribution, representing five different welfare groups: the 10th percentile, the 25th percentile, the 50th percentile (median), the 75th percentile, and the 90th percentile. This exercise can be done robustly only at the national level because of the small sample size in urban areas. In all decomposition approaches, there is an interaction effect which can be interpreted as a measure of the correlation between population shifts and inter-sectoral changes in poverty in the Ravallion and Huppi method and changes in endowments and returns in the RIF analysis. In the decompositions shown here it is quite small. Source: World Bank’s Poverty Handbook. 47 BOX 3.2: Expanding fiscal policy: How can oil revenues accelerate poverty reduction in Uganda? The Country Economic Memorandum (CEM) recently produced by the World Bank, highlighted the importance of oil as a source of fiscal revenue when production starts. It argued that to maximize the socioeconomic impact of its new revenue, Uganda should increase public investment gradually and save some of its oil revenue in the early years of production to finance countercyclical policies (given the volatile nature of oil prices) and to mitigate Dutch Disease effects. However, how should public investment best be allocated to facilitate sustainable, inclusive growth and aid poverty reduction? Economic simulations undertaken for the CEM indicate that, initially, investment in transport and energy infrastructure would aid private sector development and have a stronger impact on growth. In the long-term, however, education and health spending will be more effective. Manufacturing and modern services—and the success of the government’s diversification strategy—depend on a healthy and well- educated labor force. The CEM also highlighted that future infrastructure development programs should give priority to the poorest/underdeveloped regions of the country to promote economic and political stability. In addition, social programs focused on the poor should be designed and implement to reduce poverty. Specifically, the CEM suggests that direct cash transfers to poor households, linked with changes in health and education practices should be considered and tested. Source: World Bank 2016. “Economic Diversification and Growth in the Era of Oil and Volatility” Uganda CEM. FIGURE 3.3: Using counterfactuals to quantify changes that have been important to poverty reduction 48 3.2 Demographic change, urbanization, and poverty reduction SLOWLY DECLINING FERTILITY HAS NOT YET CONTRIBUTED TO DEMOGRAPHIC CHANGE OR POVERTY REDUCTION 97. Uganda has one of the youngest and most term, drops in fertility tend to lead to increased rapidly growing populations in the world. female labor market participation and better About half (48.7 percent) of Uganda’s population is human capital outcomes for younger generations younger than 15, well above Sub-Saharan Africa’s as more resources (public and private) can be average of 43.2 percent and the world average of invested in the education and health of each child. 26.8 percent. The country’s population growth rate, currently at 3.3 percent, has also been steadily 100. The slight drop in fertility rates in recent years above Africa’s average, except for the period of peak has not changed the demographic composition prevalence in HIV/AIDS in early 2000s (World Bank of Ugandan households. Dependency ratios have 2011). been increasing, particularly for poorer households. As illustrated in Chapter 1, the dependency ratio 98. The fertility rate has been slowly falling over remains high and increased slightly from 1.11 to the last two decades but it remains high. The 1.14 between 2006 and 2013. The increase of the total fertility rate remained stable at a high level dependency ratio was more pronounced for poor (around seven children per woman) between the households than for the non-poor households: 1960s and the mid-1900s (Figure 3.4.1), resulting in the average dependency ratio in poor households high population growth.32 This is in sharp contrast increased from 1.38 in 2006 to 1.47 in 2013 (increase to neighboring Kenya and other countries in the of 6.6 percent), while in non-poor households it region, such as Ghana and Ethiopia. Since 1995, the increased by 5.8 percent (from 1.02 in 2006 to 1.08 country has started a slow demographic transition. in 2013). Total fertility rates started dropping steadily, from 7 in 1995 to 6.6 in 2005 and 5.9 in 2013. However, 101. Increasing dependency ratios held back both the high fertility rates and the youthfulness of consumption growth from 2006 to 2013. Changes the population bring a very high youth dependency in the dependency ratio between 2006 and 2013 ratio. have not been favorable for consumption growth (Figure 3.4.2). The increase of the dependency ratio 99. Lower fertility can have positive effects on was more costly in reducing consumption growth household living standards in both the short for poor households. The demographic transition and longer term, and Ugandan households process has yet to effectively materialize in poor are missing out on these benefits. In the short households. term, lower fertility rates translate into smaller households and lower child dependency rates. 102. However, if the dependency ratio can be Fewer dependent children in a household mean reduced, consumption growth will benefit. less strain on household resources and an increase Reducing the dependency ratio, particularly in per equivalent adult consumption. In the long- for poorer households, is important for poverty 32. The total fertility rate is defined as the average number of children a hypothetical cohort of women could be expected to have at the end of the reproductive period. 49 reduction. As the demographic transition is an increasing body of evidence that points to progresses in Uganda, the working-age population high fertility rates reducing the economic capacity in Uganda will grow quickly (faster than the of women, thereby limiting the extent to which economically dependent population), causing women can contribute to and participate in dependency ratios to progressively decrease. The economic growth. A major factor contributing analysis suggests this would be associated with to lower rates of agricultural productivity found improvements in household living standards and among women is the childcare demands they poverty reduction. A recent impact evaluation face which reduces the time they can allocate to shows that targeting adolescent girls as they agricultural production (Ali et al. 2015). Bandiera et transition from school to work and providing al. (2015) found that a program supporting life skills them with vocational training and information and livelihood training for teenage girls ages 14 to on sex, reproduction, and marriage reduces teen 20, simultaneously reduced the fertility rate by 26 pregnancy and early marriage, contributing to percent and increased employment by 72 percent. reduced fertility rates (Bandiera et al. 2015). 104. Faster progress in reducing fertility will also 103. Reducing fertility rates is also imperative to reduce the pressure on education and health improving the socioeconomic status of women. services, allowing for better service delivery and The total fertility rate of 5.9 is an average of 6 better investments in human capital outcomes. child births per women. Maternal mortality rates Currently Uganda has 5.7 million primary school have been falling (Chapter 2) but are still high, age children (children ages 5 to 14), but this will and multiple births pose a significant health risk increase to 6.6 million in five years’ time and 7.5 to women. High pregnancy rates, particularly million in ten years’ time (calculations using data among teenage girls, also jeopardize educational from United Nations: http://data.un.org). The attainment. Pregnancy is the fourth most common challenges faced in delivering high quality services reason for dropping out of secondary school: in that are outlined in Chapter 7, will become even 2013, 1 in 10 girls report dropping out of secondary more severe if fertility rates are not reduced. school as a result of pregnancy. Additionally, there URBANIZATION HAS BEEN IMPORTANT FOR POVERTY REDUCTION 105. Uganda is predominantly a rural country and accounting for the movement of 180,000 poverty reduction has thus been concentrated people out of poverty. Census data shows that in rural areas. In Uganda, 82 percent of the Uganda’s urban population increased by half a population lives in rural areas (2014 census). A percentage point per year from 2002 to 2014. This higher share of poor Ugandans live in rural areas is an estimated increase of 3.5 percentage points given the higher rates of poverty in rural Uganda in the urban population from 2006 to 2013. This compared to urban Uganda. Figure 3.4.3 shows small increase accounted for 10 percent of poverty that 80 percent of poverty reduction took place in reduction given the substantially higher welfare of rural Uganda. Reductions in poverty in urban areas households in urban areas. contributed to poverty reduction from 2006 to 2010, but not after then, partly because the urban poverty 107. Welfare gains from rural to urban migration rate was so low by 2010. contribute to the role urbanization plays in reducing poverty. Some urbanization is likely a 106. However, urbanization accounted for one- result of higher rates of natural population growth tenth of poverty reduction from 2006 to 2013, in urban areas than rural areas because of lower 50 mortality rates, but migration of individuals from where the return to labor is low to areas where the rural to urban areas also helps. Figure 3.4.4 uses return to labor is higher because of better market panel data in which individuals who migrated were opportunities (Harris and Todaro 1970; Lewis 1954). tracked over time and shows how consumption Migration can also help bring welfare gains for increases much more for an individual when he a household by helping the household diversify or she moves from living in rural Uganda to living income sources and minimize risk (Rosenzweig in an urban center, than for an individual who and Stark 1989; Stark and Bloom 1985). Chapter 6 does not move. Migration can bring about welfare looks at the impact of migration on welfare and the gains if individuals are able to move from areas drivers of migration in more detail. FIGURE 3.4: Demographic change and poverty reduction, 2006–2013 2. Higher dependency ratios held back 1. Uganda’s demographic transition has consumption growth, especially for the been slow poorest, 2006–2013 3. Rural areas and urbanization are 4. Migration to urban areas increases important for poverty reduction consumption Source: 1: Canning et al. (2015); 2 and 3: Staff calculations using UNHS 2006–2013; 4: Mensah and O’Sullivan (2016) using UNPS 2006 and 2010. 51 3.3 Agricultural growth has been particularly important for poverty reduction 108. Cross-country analysis finds that growth in the importance of growth in coffee incomes (as a result sectors in which the poor are employed is more of coffee marketing liberalization and favorable poverty reducing than growth in other sectors international prices), growth in agricultural (Loayza and Raddatz 2010; and Christiaensen et productivity for goods produced for self- al. 2013). In this section, we characterize the nature consumption, and growth in nonfarm enterprises of employment and income for poor households for poverty reduction (Deininger and Okidi 2003; and assess what type of income growth was most Fox and Pimhidzai 2011). This analysis in this important for poverty reduction in Uganda from section highlights the importance of continued 2006 to 2013. Analysis conducted on poverty trends of agricultural growth post-2006 in bringing reduction from 1993 to 2006 highlighted the about poverty reduction. JOBS AND INCOME OF UGANDAN HOUSEHOLDS: DIVERSIFIED BUT NOT INCREASINGLY SO 109. The agricultural sector is the main sector 111. Households diversified their sources of income of employment for households in Uganda, from 1993 to 2006. Since 2006 little additional particularly so for poorer households. Agriculture diversification has been observed. Fox and is cited as the main sector of employment for 72 Pimhidzai (2011) document dramatic growth percent of the workforce in 2013 and 81 percent of in the number of sources of income Ugandan households report engaging in some agricultural households reported from 1993 to 2006 (Table production. The poorest and the bottom 40 percent 3.1). The proportion of households that reported are even more concentrated in agriculture: 89 income from nonfarm self-employment increased percent of poor households and 90 percent of the by 1 percentage point a year from 28 percent in bottom 40 percent receive income from agricultural 1993 to 41 percent in 2006. Structural change was production. occurring during this time. Not by households moving out of agriculture, but by households 110. However, half of those engaged in agriculture staying in agriculture and taking on informal sector have additional sources of income from activities in agriculture and services. However, non-agricultural activities. Only 41 percent of there has been very little increase in diversification households derive income only from agricultural and very little structural change since 2006 with activities, 40 percent of households are engaged diversification of household income sources similar in some form of employment in both agriculture in 2013 to 2006. and non-agricultural sectors. The majority of non-agricultural income is also earned in self- 112. Poorer households are less diversified. On employment rather than wage employment. average, half of household income comes from In 2013, 42 percent of households earned non- agricultural production, but for the bottom 40 agricultural income from self-employment and percent, three-quarters comes from agriculture. 24 percent of households earned non-agricultural Information on real income per capita for income from wage employment. households in Uganda across time is presented 52 in Figure 3.5.1 for all households and for the 40 percent. Income from nonfarm self-employment bottom 40 percent in Figure 3.5.2. Together, crop, 33 is the second most important source of income livestock, and agricultural wage income comprised followed by non-agricultural wage income (for all 50 percent of the income of Ugandan households in households and for the bottom 40 percent).34 2012 and 73.8 percent of the income of the bottom FIGURE 3.5: Household labor income and poverty reduction, 2006–2013 1. Real income per capita by source of income, all 2. Real income per capita by source of income, bottom 40 3. Sectoral contribution to poverty reduction, 4. Sectoral contribution to poverty reduction, 2006 to 2013, main source of income 2006 to 2013, all sources of income Source: 1 and 2: Staff calculations using UNPS 2006–2012; 3 and 4: Staff calculations using UNHS 2006–2013. 33. The data represents weighted averages of income from crop farming, livestock production, wage employment (in agriculture and non- agriculture sectors) and nonfarm self-employment. All values are in 2011 prices. 34. Finding a measure of non-agricultural self-employment income that compares well to the measures of gross agricultural income used is not straightforward. Much self-employment income comes from trade and taking only gross sales does not give an idea of how much was earned. To account for this net self-employment income in the analysis, which is gross self-employment income net of raw materials, operating expenses, and wages paid to others. Raw materials account for 81 percent of these expenditures. Operating expenses and wages paid to others account for 12–13 percent of gross income, suggesting that self-employment income would be a marginally more important source of income were these expenses not netted out. 53 TABLE 3.1: Structure of household income, 1993 to 2013 Proportion of Households Reporting Receiving Income from: 1993 2006 2013 Wage employment in agriculture 10.7 20.9 22.7 Wage employment out of agriculture (private and public) 21.2 27.2 24.0 Nonfarm self-employment 27.7 41.4 42.5 Agricultural self-employment 82.0 77.3 75.8 Source: Fox and Pimhidzai (2011) using UNHS 1993 and 2006. Authors’ calculations for 2013 using UNHS 2013. Note: 2013 data estimated from a labor module, not income. Using income data suggests a higher share earning income from self-employment in agriculture (86 percent), a higher share earning income from self-employment in non- agriculture (45 percent), and 41 percent of households earning wage income (agriculture/non-agriculture not specified). AGRICULTURAL GROWTH (NOT DIVERSIFICATION) ACCOUNTS FOR POVERTY REDUCTION 113. Poverty reduction among households in 115. Poverty reduction was just as fast for those agriculture accounts for 79 percent of national solely in agriculture, as for those with poverty reduction from 2006 to 2013 (Figure diversified income sources. The share of poverty 3.5.3). This is when households in agriculture reduction accounted for by households solely are defined as all those households that report in agriculture is high, commensurate with the agriculture as their main sector of employment. share of this type of household among those The large contribution of this group to national who were poor in 2006 (Figure 3.5.4). The share poverty reduction is perhaps not surprising of poverty reduction accounted for by diversified given that 72 percent of Uganda’s population households was also equivalent to their share in cite agriculture as their main income source the poor population in the beginning of the period, (UBOS 2014a). Kaminski and Christiaensen (2014) indicating that it was not only for these households undertake decomposition analysis using the UNPS for whom poverty reduction was faster. and find that agricultural growth contributed to 70 percent of the poverty reduction observed from 116. These findings suggest that although 2006 to 2010. They also estimate that agricultural diversification may have driven poverty income growth accounted for 18 percent of reduction before 2006 when diversification consumption growth from 2006 to 2010, because was rapidly increasing, it was not the main of the lower importance of agricultural income for driver of progress from 2006 to 2013. There is non-poor households. a commonly held view that diversification has enabled predominantly agricultural households to 114. However, as Table 3.1 suggests, agricultural become less poor. This may have been true before households have diverse sources of income. 2006 when many households were acquiring an Was it agricultural growth or growth in incomes additional non-agricultural income source, but from other sources that contributed to poverty this was no longer true after 2006. Instead, the reduction? To answer this question, agricultural findings are consistent with literature that points households are categorized into those that derive to agricultural income growth as a major source income solely from agriculture and those with of poverty reduction in the country (Dorosh and agricultural and non-agricultural income sources. Thurlow 2012; MoFPED 2014; Kassie, Shiferaw, and Results are presented in Figure 3.5.4. Muricho 2011). 54 117. Agriculture’s seemingly significant contribution The growth in agricultural income recorded in to poverty reduction is consistent with the high survey data is not consistent with national account rates of agricultural income growth observed estimates (Box 3.3). For poor households, growth from 2006 to 2013. On average, real per capita in non-agricultural per capita income was equal crop income grew by 9 percent per year, and by to growth in agricultural income (6 percent), but it 11 percent for the poorest 40 percent (Table 3.3 accounts for a much smaller share of total income. and Figures 3.5.1 and 3.5.2). Overall, agricultural Growth in non-agricultural income was lower when income per capita grew by 5 percent annually on considering all households. average and 6 percent for the bottom 40 percent.35 BOX 3.3: Agricultural growth in national accounts and survey data Table 3.3 indicates substantial growth in real per capita agricultural incomes from 2006 to 2012 based on household survey data. In contrast, limited agricultural growth was recorded in the national accounts from 2006 to 2012. The national accounts suggest agricultural growth in Uganda has been consistently low, averaging only 2 percent over the past five years (see Figure 3.6) and below the performance achieved by other regional economies (see Table 3.2). It is difficult to explain why there is this divergence between the national accounts estimates of agricultural growth and those found in the survey data. The national accounts estimates are not based on any additional sources of survey information (such as agricultural sample surveys which are often used in other countries to underpin estimates of agricultural value added) and it has been a number of years since an agricultural census was undertaken so it is difficult to assess what underpins the national accounts estimates and thus what might cause the divergence. The UNPS may be biased to households that have stayed in agriculture, as households that attritted over time are probably more likely to be those that have moved out of agriculture. However, given that the nationally representative cross sections suggest that many households have stayed in agriculture during this time, this is unlikely to be a large source of bias. TABLE 3.2: Agricultural GDP growth rates for selected Eastern African countries, 2000–2012 Country 2000–2009 2010–2012 Ethiopia 6.6 6.3 Tanzania 4.6 3.9 Kenya 2.3 4.6 Uganda 2.6 1.5 Source: World Bank, WDI. 35. It is worth noting that the panel analysis may overestimate national average per capita agricultural growth (and underestimate national average per capita non-agricultural growth) as households that attritted over time are probably more likely to be those that have moved out of agriculture. However, the nationally representative cross sections undertaken during this time show that many households have stayed in agriculture, so this is unlikely to be a large source of bias. 55 118. Agricultural income growth is also found to be consumption growth than other sources of income more strongly correlated with consumption growth. This correlation is larger for the bottom growth than other sources of income growth, 40 percent (column 2) indicating that agricultural particularly for the bottom 40 percent. Growth income growth has been more important for in real per capita income from different sources poverty reduction during this period than other is correlated with household consumption to types of income growth. Chapters 4 and 5 look ascertain whether growth in some sources of further at agricultural and non-agricultural income have been more important for increasing income growth and poverty reduction to examine consumption than others. The results are 36 how agricultural growth contributed to poverty presented in Table 3.4 and indicate that agricultural reduction in this period and what holds back income growth is more strongly correlated with diversification and growth in nonfarm income. FIGURE 3.6: Sectoral growth rates Source: Uganda Fourth Economic Update. Dry food grains on a market stall 36. Specifically, a fixed effects model was estimated using the log of per capita consumption and the log of per capita income, allowing an analysis of the relationship between changes in income and changes in consumption. Interview year and month fixed effects were also included. The analysis was conducted only for 2005/06 and 2009/10 as there is a marked reduction in the consumption aggregate after 2009/10 that is hard to explain and is inconsistent with the national poverty trend. It may result from methodological differences in the collection of consumption data in the 2010/11 and 2011/12 survey rounds. Computer Assisted Personal Interviews (CAPI) was introduced in the UNPS for the 2010/11 and 2011/12 rounds and this may have resulted in a reduction in reported consumption. CAPI was not introduced in the nationally representative cross-sectional survey, the UNHS. 56 TABLE 3.3: Real per capita Income growth by source of income, 2006 to 2012 Agricultural Income Non-agricultural Income Self Crop Livestock Wage Total Wage Total employment All households 2006 115,320 49,322 50,147 214,788 118,582 138,486 257,068 2010 165,735 82,590 23,056 271,380 147,255 113,697 260,951 2011 145,938 78,789 21,651 246,378 175,932 103,013 278,945 2012 195,194 75,527 18,924 289,645 168,204 121,480 289,684 Annual growth 9% 8% –10% 5% 6% –1% 2% Bottom 40 percent 2006 99,423 39,696 62,849 201,968 29,566 42,219 71,785 2010 140,172 72,179 30,777 243,128 40,385 48,015 88,400 2011 133,862 63,163 29,428 226,454 47,188 50,536 97,723 2012 191,205 71,004 28,637 290,847 49,953 53,056 103,009 Annual growth 11% 12% –8% 6% 8% 3% 6% Source: Staff calculations using UNPS 2006–2012. Note: Value of non-agricultural self-employment income for 2011 is interpolated between 2010 and 2012. TABLE 3.4: Relationship between income and consumption, 2006-2010 1 2 Log of Per Capita Consumption All Households Bottom 40 Percent Log of per capita real crop gross income 0.0324*** 0.0416*** (0.00805) (0.0103) Log of per capita real livestock gross income 0.00573** 0.00479 (0.00283) (0.00347) Log of per capita real agricultural wage 0.00127 0.00186 (0.00239) (0.00278) Log of per capita real non-agricultural wage (0.00271) (0.00359) Log of per capita real self-employment income 0.00934*** 0.0106*** (0.00246) (0.00302) Constant 10.28*** 9.942*** (0.140) (0.189) Observations 4,171 3,017 R-squared 0.086 0.095 Number of households 2,644 1,853 Source: Authors’ calculations using UNPS 2005/06 and 2009/10. Notes: The dependent variable is log of real per capita consumption. Household, year, and month of interview fixed effects are included but not shown. Robust standard errors are in parentheses. Coefficient statistically significant at: ***1%, ** 5%, *10%. 57 3.4 Human capital, access to infrastructure, and poverty reduction SMALL IMPROVEMENTS IN HUMAN CAPITAL HAVE BEEN ASSOCIATED WITH POVERTY REDUCTION 119. Over the last decade, there was slow percent and in rural self-employment, it ranges improvement in human capital outcomes. For from 6.8 percent in agriculture to 6.1–8.3 percent in example, as illustrated in Chapter 2, adult literacy non-agriculture (Lekfuangfu et al. 2012). Assessing rates remained almost flat between 2006 and 2013. returns to education is challenging as it is difficult to Little progress is also observed when considering disentangle the effect of education on income from all household members: between 2006 and 2013, the effect of other characteristics associated with there was a 2 percentage point increase in the high education that also result in higher incomes, proportion of households with at least one member for example parental education or self-discipline, with secondary education and a corresponding 2 but estimates using the introduction of UPE to try percentage point reduction in the proportion of and identify the causal impact of education suggest households in which the highest level of education that if anything the returns to education are higher achieved was primary (Table 3.5). However, for in Uganda than the estimates suggest (Lekfuangfu the poorest 40 percent the progress was twice as et al. 2012). fast. These households experienced a 4 percent increase in the share of households with a member 121. To assess the correlation between human capital with secondary education or higher. The share of and consumption growth, the relationship individuals in a household that achieves higher between the household’s endowment in levels of education follows a similar trend. Very few education and consumption is examined. The are able to make it up to the tertiary level. proxy that is used here is the maximum level of education achieved by any member of the 120. Education and skills allow households to household. There is a difference between attending improve their living standards by accessing school and effectively acquiring the relevant skill more productive jobs and by increasing their that matters for the labor market and poverty productivity in self-employment activities. reduction, but in the absence of a measure of skills, Estimates suggest that the rates of return to the level of education achieved is used. Chapter 7 education in Uganda are high, both in wage looks more in-depth into these issues of quality of employment and in self-employment in and out service delivery and why educational attainment of agriculture. The return to an additional year of and skills acquired have not increased as would be schooling in urban wage labor markets is 4.5–6.5 expected in Uganda. 58 TABLE 3.5: Maximum level of education attended by a household member, 2006–2013 Welfare Quintile Quintile 1 Quintile 2 Quintile 3 Quintile 4 Quintile 5 Total 2006 No education 4.0 1.2 1.3 1.4 1.0 1.8 Primary 73.1 62.3 55.5 46.4 22.4 51.9 Secondary 21.5 35.0 39.5 45.9 49.5 38.3 Tertiary 1.3 1.5 3.7 6.3 27.1 8.0 Total 100.0 100.0 100.0 100.0 100.0 100.0 2013 No education 3.3 1.9 1.1 1.4 1.6 1.9 Primary 70.0 57.6 53.6 42.1 27.4 50.1 Secondary 26.1 37.4 40.2 48.1 49.3 40.2 Tertiary 0.6 3.2 5.1 8.4 21.7 7.8 Total 100.0 100.0 100.0 100.0 100.0 100.0 Source: Staff calculations using the UNHS 2006 and 2013. 122. Although progress on education has been slow, household characteristics, changes in education are it has been associated with income growth, picking up other characteristics of households that accounting for almost half of the consumption have changed over time and that are also associated growth experienced by the poorest households with (or driving) consumption growth. In subsequent in decomposition analysis (Figure 3.7). As one chapters we use panel data to further examine moves up the consumption distribution, education the causal role that education plays in increasing accounts for less and less consumption growth until agricultural and non-agricultural incomes (Chapters it accounts for nothing for those at the top of the 4 and 5) and in encouraging migration (Chapter 6), distribution. It is possible that in decomposition to understand whether it did indeed have a large analysis undertaken with a limited set of observed impact on consumption growth, and if so why. FIGURE 3.7: Educational attainment is associated with consumption growth, except for the wealthiest households Source: Staff calculations using the UNHS 2006 and 2013. Note: Contribution of change in education level of household members to growth in per equivalent adult consumption (percent). 59 IMPROVEMENT IN ACCESS TO UTILITIES OVER THE LAST DECADE IS ASSOCIATED WITH CONSUMPTION GROWTH 123. Access to electricity and piped water is low in more likely to benefit from improved access. Uganda, but has improved in the last decade. As illustrated on Figure 3.8, increased access to As Chapter 2 discusses, the share of households electricity and residential piped water between connected to the electricity grid increased from 2006 and 2013 is associated with an increase of 10 percent in 2006 to 14 percent in 2013. In 2013, consumption, particularly for the non-poor. It is 7 percent of households were connected to the not possible to know from this analysis whether residential piped water network, a slight increase or not this association is causal. Given the limited from 5.1 percent in 2006. impact of electricity on nonfarm income growth (Chapter 5) and the limited nonfarm income growth 124. Increased access to electricity and piped water experienced by poor households during this period were associated with consumption growth, (Table 3.3), it may not be. particularly for wealthier households that were FIGURE 3.8: Increased access to electricity and piped water is associated with consumption growth Source: Staff calculations using the UNHS 2006 and 2013. Note: Contribution of change in education level of household members to growth in per equivalent adult consumption (percent). 60 3.5 Conclusion 125. This chapter has documented the importance 126. This chapter also highlights three factors of agricultural income growth, urbanization that did not contribute to poverty reduction: and improvements in human capital—albeit demographic transition, structural change, limited—in accounting for the poverty reduction and public safety nets. Uganda has one of the that Uganda has experienced. High rates of youngest and most rapidly growing populations in the world. An increasing dependency ratio held agricultural income growth per capita, 6 percent, back consumption growth from 2006 to 2013, were observed from 2006 to 2012, particularly for reducing the consumption growth of the poorest the poorest, and this growth is strongly correlated households by 15 percent to 20 percent. Securing with growth in consumption for the bottom 40 more rapid reductions in the fertility rate in Uganda percent. The share of the population living in is essential both for poverty reduction and for urban areas in Uganda increased by 6.3 percentage improving the socioeconomic status of women. points from 2006 to 2014 and this accounted for 10 Another area that saw little change in this period percent of Uganda’s poverty reduction given the was the structure of household incomes. Although better economic opportunities available in urban rapid diversification out of agriculture was observed areas. Although education outcomes improved prior to 2006, little movement has been observed slowly from 2006 to 2013, this improvement can since then, suggesting that high rates of growth in account for substantial income growth among the services and industry has not resulted in high rates poorest. Returns to education still appear to be of job creation. Job creation in services and industry just kept up with population growth. Finally, public high in Uganda, both for rural households engaged transfers to poor households in Uganda are minimal in agriculture and nonfarm activities and for urban and contribute little to poverty reduction reflecting households with members in wage employment. a limited use of fiscal policy to directly improve Chapters 4 to 6 explore the role of agriculture, the incomes of poor households in Uganda, in migration, and education in more detail. comparison to other countries in the region. 61 AGRICULTURAL GROWTH CHAPTER: AND POVERTY REDUCTION 4 IN UGANDA 37 Agricultural incomes grew because the government got some key fundamentals right that provided the incentives to invest in agriculture. Luck was also on Uganda’s side: good weather benefited many households and positive price trends in international and regional markets aided real crop price increases. 127. Chapter 3 highlighted the important role that agricultural income growth has played in reducing poverty in Uganda from 2006 to 2013. Half of all poverty reduction occurred within households whose only income source was in agriculture. This increased to nearly 80 percent when considering households with other income sources, but the main occupation is in agriculture. 128. This chapter assesses the factors that have contributed (and those that have not) to growth in agricultural income of households in recent years. The focus of the analysis is on crop income earned through self-employment, as this constitutes two-thirds of household agricultural income. Changes in production practices of households as well as changes in the external environment that may have had a direct impact on crop income or affected how households decided to produce. 129. The analysis shows that Uganda was able to get many of the fundamentals right. The government secured stability in the north and enabled private markets for agricultural produce to develop across the country, resulting in real relative price increases for agricultural 37. This chapter draws on the background paper: “Welfare, income growth, and shocks in Uganda” by Ruth Hill and Carolina Mejia-Mantilla. 62 commodities that poor farmers grow and sell. similar plot sizes in the same region. Gender Ensuring continued stability in the region and differences in household labor, childcare further promoting efficient crop markets and responsibilities, education, and extension regional exports will be important for future crop contribute to this large gender gap between female income growth in Uganda. and male farm managers. 130. Also, luck was on Uganda’s side: good 133. The reliance on weather and prices also offers weather benefited many households and the some cause for concern. When prices are poor positive price trends in international food and or when the rains do fail, crop income growth commodity markets during this period aided falters and consumption falls, reversing real crop price increases. As a result, a favorable gains in poverty reduction. This is indeed what external environment (some of it policy induced happened in the Northern and Eastern regions in and some of it not), accounted for two-thirds of 2011. Households need to be able to both benefit the change in agricultural income among poor from good prices and weather and have access households, contributing to higher household to coping mechanisms to be protected from low consumption and lower poverty. prices and poor weather. Sustained growth in incomes and welfare will also require productivity 131. However, there are also areas where less growth in agriculture—possibly through the use progress was made: extension services of improved seeds, fertilizer, pesticides, and remain limited and production practices did irrigation—and diversification to other more not change much. There was very little growth remunerative forms of employment. in the use of improved inputs and as a result modernization of agricultural practices contributed 134. Diversification of income offers households very little to crop income growth. Understanding the ability to protect consumption from why farmers did not adopt agricultural technologies weather shocks, but it is not enough to fully during this time of high prices and designing protect consumption, and better safety nets policies that helps farmers overcome these are needed. Education is essential to enabling constraints needs to be a key area of action going households to diversify and better-educated forward. Recent research suggests that poor quality households had consumption that was better of inputs, limited access to credit, and lack of insured from weather shocks as a result. However, knowledge are binding constraints. the ability to diversify does not fully insure consumption. The inability of Uganda to implement 132. In addition, large gender differences in a functioning public safety net system has resulted agricultural productivity limit the equity of in households relying on informal networks and agricultural growth. Female farm managers own savings to manage shocks. These are imperfect are 13 percent less productive than male farm insurance mechanisms and as a consequence high managers are. The gap increases to 33 percent levels of vulnerability are observed. when comparing male and female farmers with 4.1 Agriculture and poverty in Uganda 135. For households in Uganda—both rich quarter of agricultural income, and the remaining and poor—agricultural income is largely comes from agricultural wage employment (Table comprised of crop income earned through 4.1). Livestock income and crop income have self-employment. Self-employment crop income grown at an equal pace from 2006 to 2012 for comprises two-thirds of agricultural income, all households and the bottom 40 percent alike livestock self-employment income comprises a (Chapter 3). Agricultural wage income has fallen. 63 Table 4.1: Agricultural income, 2012 Proportion of Agricultural Income from: All Households Bottom 40 Percent Self-employment in crop production 0.67 0.66 Self-employment in livestock 0.26 0.24 Agricultural wages 0.07 0.10 Source: UNHS 2012. Figure 4.1: Structure of agricultural income by region, 2012 Source: Staff calculations using RIGA income aggregates calculated from UNPS 2012. 136. For households in the poorer Northern regions except the Northern region, and cassava is region, livestock income is more important, important in all regions except the Western region. comprising 35 percent of agricultural income, The crops produced are very similar among the but crop income still dominates (Figure 4.1). bottom 40 percent. This is true for all households, on average, and households in the bottom 40 percent. This is 138. The share of household income coming from in contrast to the Central and Western regions crop sales has increased from 2006 to 2012. where livestock comprises about 20 percent of Figure 4.2 shows that the share of crop income agricultural income. Income from livestock in the marketed has increased over time for the bottom 40 north is dominated by income from crop sales, percent. The share of households in the bottom 40 whereas sales of by-products such as milk are a percent selling crops has increased from 60 percent more important share of livestock income in other in 2006 to 72 percent in 2012. regions (18 percent to 25 percent compared to 9 percent in the north). 139. Crops that are produced for domestic and regional consumption dominate crop income. 137. Maize, beans, matooke, and cassava are the Coffee is important for some households, but does four most important crops grown in Uganda, not comprise more than 10 percent of crop income as a share of total crop income. Table 4.2 in any region. Sunflower produced for commercial indicates that maize and beans are universally production has increased in importance in recent important—comprising 10 percent or more of crop years, particularly in the north, but it is still a incomes in all regions. Matooke is important in all relatively small share of crop income. The growth 64 of sugarcane, particularly in the Eastern region has of exports by 2005 (World Bank 2007) and that 41 been reported, but by 2012 it was not comprising percent of exports now go to Uganda’s four regional more than 1 percent of crop income in that region. neighbors (in order of importance): South Sudan, This is consistent with the export data that shows the Democratic Republic of Congo, Kenya, and that coffee fell from comprising three-quarters of Rwanda (World Bank 2015). exports at the beginning of the 1990s to a third Figure 4.2: Share of crop income derived from crop sales, bottom 40 percent, 2006–2012 Source: Staff calculations using RIGA income aggregates calculated from UNPS 2006–2012. Table 4.2: The nature of crop income, 2012 Proportion of All Households Bottom 40 Percent Crop Income from (Average) National Central Eastern Northern Western National Central Eastern Northern Western Sales of crops 0.30 0.30 0.27 0.28 0.35 0.26 0.26 0.24 0.26 0.31 Beans 0.16 0.18 0.11 0.13 0.21 0.15 0.19 0.13 0.14 0.21 Maize 0.17 0.15 0.25 0.16 0.10 0.17 0.18 0.25 0.15 0.12 Matooke 0.16 0.25 0.08 0.02 0.34 0.10 0.20 0.08 0.01 0.30 Cassava 0.11 0.09 0.15 0.14 0.04 0.12 0.11 0.16 0.15 0.05 Sweet Pota- 0.09 0.15 0.11 0.06 0.06 0.08 0.17 0.12 0.06 0.07 toes Groundnuts 0.06 0.02 0.08 0.06 0.05 0.04 0.02 0.06 0.05 0.04 Coffee 0.04 0.08 0.03 0.01 0.05 0.03 0.06 0.03 0 0.06 Sorghum 0.04 0.00 0.03 0.09 0.02 0.05 0 0.03 0.11 0.02 Finger Millet 0.03 0.01 0.06 0.04 0.02 0.03 0 0.05 0.04 0.02 Simsim 0.02 0.00 0.01 0.06 0.00 0.02 0 0 0.07 0 Sunflower 0.02 0.00 0.00 0.05 0.00 0.02 0 0 0.06 0 Source: Staff calculations using UNPS 2012. Note: Red indicates a share 10 percent and higher in a given region; green indicates a share between 4 percent and 10 percent. 65 Table 4.3: Household characteristics, by wave   2006 2010 2011 2012 Mean s.d. Median Mean s.d. Median Mean s.d. Median Mean s.d. Median Age of household 43.75 15.10 41.00 47.62 14.90 45.00 48.17 14.90 46.00 48.73 14.61 46.00 head Household head is male 0.74 0.44 0.72 0.45 0.69 0.46 0.68 0.47 Education of household 2.49 1.29 2.00 2.43 1.28 2.00 2.54 1.31 2.00 2.45 1.27 2.00 head Distance to market selling agricultural 10.05 10.92 7.33 6.99 8.43 4.00 6.92 9.19 4.00 5.15 5.09 4.00 inputs in Km Received any visits by extension 0.09 0.28 0.00 0.16 0.37 0.00 0.08 0.27 0.00 0.12 0.33 0.00 services in past 12 months Total area planted self- 2.79 3.22 1.82 3.69 3.56 2.43 3.10 3.18 2.02 2.90 3.07 1.78 reported, in Ha Renter (land) 0.23 0.14 0.19 0.19 Use of fertilizer (1=yes) during 0.17 0.22 0.22 0.24 the year Use of pesticides (1=yes) during 0.13 0.16 0.14 0.12 the year Use of seeds and seedlings (1=yes) during 0.64 0.80 0.69 0.71 the year Any hired labor used (1=yes) 0.56 0.57 0.52 0.44 during the year Number of fatalities in a 25 4.78 21.3 0 1.64 6.07 0 2.37 10.63 0 0.28 1.39 0 km radius Source: Staff calculations using UNPS 2005/06–2011/12 Examining crop income growth. Note: s.d. = Standard Deviation. 140. Given the importance of crop income, this and available for sale are considered separately section examines what factors contributed from the role of the external environment: the to its growth from 2006 to 2012. Some of the introduction of peace, the nature of the weather, change is likely to have come from the substantial changes in prices, and access to markets. Of increase in crop marketing during this period. course, the external environment influences how The role of changing household farming practices households decide to farm and this relationship is that can increase the amount of crops produced considered in the discussion. 66 PRODUCTION PRACTICES 141. Production practices are significantly associated with a 2 percent rise in agricultural correlated with crop incomes in Uganda, but income; 1.98 percent for the bottom 40 percent. The production practices did not change much estimates include individual fixed effects to account between 2006 and 2012, so they contributed for time-invariant unobserved characteristics that little to crop income growth. To capture the simultaneously affect crop income growth and impact of changes in production practices on crop production practices, but it is still possible that income growth, data on the area and ownership time-varying characteristics are in part responsible of the plot being harvested, the use of fertilizer, for the observed relationships. improved seeds and pesticides, household labor inputs (both hired labor and family labor), access to 143. However, although there was some increase in extension, and household demographics are used the proportion of households using fertilizer in a fixed effects panel regression analysis. 38 For and pesticides during this period, the increase households that did change production practices, was relatively marginal. The proportion of large changes in income were observed but few households using fertilizer increased from 17 households changed production practices during percent in 2006 to 24 percent in 2012 while this time. pesticide use hovered around 12–13 percent (Table 4.3). As a result, technology adoption did not 142. A household in the bottom 40 percent in 2006 contribute to large increases in crop incomes on that adopted both fertilizer and pesticides average. has a crop income that is 36 percent higher than the crop income for those that adopt 144. Households that farmed more land received neither. Table 4.4 presents regression results higher per capita crop income, but not by using four rounds of the UNPS panel. These results much, and there was little increase in the area show that per capita crop income is significantly cultivated during this period. The coefficient higher among those who farm more land and estimates suggest that an increase of 1 ha in the apply more labor, fertilizer, and pesticides. Using area of land farmed is associated with an increase improved seeds does not have a significant effect in crop income of only 2 percent. In addition, very on crop income. Households that use fertilizer and little change in the area of land cultivated was pesticides have crop incomes that were 12 percent recorded during this time. Detailed analysis on and 19 percent higher, respectively, than those area of land cultivated in Uganda and other Sub- households which did not. The increase is even Saharan Africa countries shows that relying on self- higher for households that were in the bottom 40 reported land areas results in considerable (and percent at the start of the period: crop incomes systematic) measurement error (Kilic et al. 2014 are 22 percent higher for those using fertilizer and Carletto et al. 2015). Indeed the self-reported and 14 percent higher for those using pesticides. area of land cultivated fluctuated over the four A 1 percent increase in the value of pesticide is rounds perhaps more than the true area of land 38 The regression run is where is the log of the real value of per capita crop income of household i at time t. is a set of variables representing production practices, containing the average plot area harvested by household i at time t, and an indicator variable if the household owns or owns and rents plots (only renter is the excluded category), dummy variables for inputs such as fertilizer, pesticide, seeds/seedlings, and hired labor, and the amount of family labor spent on the farm. is a set of variables capturing the external environment. It includes the distance in kilometers of household to the nearest market selling agricultural inputs at time t, whether extension services were provided to any household in the community, prices of maize and beans at the nearest major urban market to household at time , the WRSI experienced by household at time , and the number of fatalities in proximity to household at time . The regression is run with household fixed effects to control for time-invariant household characteristics. For more details see Hill and Mejia-Mantilla (2016). 67 cultivated. However, there is very little growth in the 146. Human capital influences the type of labor land cultivated over the period and, as a result, it available for crop income and real per capita does not appear that expansion of land cultivated crop income is higher for those who are by these households contributed much to the educated. Specifically, compared to those with no increases in average per capita income growth education, agricultural income is 26 percent higher observed. in households whose head had some primary education, 34 percent higher in households whose 145. Increased household labor on crop production head completed primary school, 25 percent higher accounts for 10 percent of the growth in crop for those with some secondary education, and income. Households that apply more labor—both 42 percent higher for those with post-secondary family labor and hired labor—have higher crop education. This is consistent with rates of return to incomes, as expected. A 10 percent increase in the education in agriculture estimated by Lekfuangfu et number of days of family labor provided by the al. (2012). This suggests that increasing educational household is associated with an increase in crop attainment can contribute to crop income growth, income by 2 percent. The amount of household but the causality of the impact of education on labor reportedly spent on agricultural production crop income growth is hard to estimate with data increased substantially between 2006 and 2011, available. It is also worth noting that improving falling again in 2012. This may not reflect a true human capital of existing farmers requires change in household labor applied during this education of adults, such as is possible through time. However, even if this does represent a real extension. increase and if the return to this was as estimated in Table 4.4, the increase of 50 percent reported 147. Crop income was 20 percent higher in villages would only account for 10 percent of the increase where extension services were provided, but in crop income. Regression results indicate that few households received extension services. households that hire labor have agricultural Extension services expanded by 50 percent from production that is higher by 15–25 percent, but the 2006 to 2012. However, extension was expanding use of hired labor actually fell during this time. from a low base. Eight percent of households received extension services in 2006 and 12 percent 68 of households received extension services in 2012 labor showed marked increases during this time. (Table 4.3). The relationship between extension In general, input use is very low in Uganda in and crop income growth appears to come from the comparison to other countries in the region with increased use of inputs that extension services may data collected using a similar survey instrument encourage. When use of inputs is controlled for, (Binswanger and Savastano 2014; Sheehan and extension has no additional effect. However, even Barrett 2014). though household fixed effects are included in all regression estimates and the measure of extension 150. Recent research highlights that low quality used is availability of extension services in a village, inputs are prevalent in local markets limiting it is hard to estimate the causal impact of extension the returns to adoption. Bold et al. (2015) tested from panel data. Experimental evidence suggests it the quality of agricultural inputs purchased in local can increase income when combined with access to markets. They found that, on average, 30 percent credit (Bandiera et al. 2015). of nutrients are missing in fertilizer, and that more than 50 percent of hybrid maize seeds are 148. Gender differences in access to labor, not authentic. This low quality results in negative education, and extension account for a large returns on average, even though prices are high. gender gap between female and male farm If authentic technologies replaced these low- managers: 33 percent when comparing men quality products, average returns for smallholder and women with similar farm sizes in the same farmers would be over 50 percent. Public regulation region. Closing the gap of a third of production and certification has not proven effective in is key to increasing agricultural productivity in guaranteeing quality products in this market. An Uganda equitably. Although, all else equal, male- ongoing impact evaluation by the International headed households have lower levels of crop Food Policy Research Institute (IFPRI) is assessing income than female-headed households, all else whether privately provided e-verification can is usually not equal and a comparison of male and provide farmers with a guarantee that the product female farm managers shows that women have they are purchasing is of high quality. Lessons can 13 percent lower productivity than men. This gap also be learned from an evaluation of strategies to increases to 33 percent once the plot size and the improve the quality of malaria medicine available region of residence is controlled. Female farm in Uganda. Flooding the market with high-quality managers have fewer household members to malaria drugs certified by a locally respected provide labor, have a larger share of children in the nongovernmental organization brought about an household, which carries a significant childcare increase in the quality of Malaria medicine found in burden, and have lower levels of education, all of retail pharmacies (Björkman Nyqvist et al. 2012). which contributed to the gap. In addition, women use less fertilizer and appear to benefit less from 151. Farmer behavior suggests that farmers are extension when they receive it. See Ali et al. (2015) aware of the returns to using inputs that they for a fuller discussion of differences in productivity face and that this can explain the low rates of between male and female farmers in Uganda. input use. The average return to using inputs is estimated to be negative, but the actual returns 149. The fact that production practices did not each farmer faces depends on the input and crop change from 2006 to 2012 is a puzzle, as this prices he or she faces, and his or her ability to was a period during which the returns to secure good inputs. Low adoption rates indicate investing in crop production were increasing. that many farmers know the returns they face are The return to investing more in inputs was negative. However, for some farmers who face increasing considerably—food prices were high and particularly good prices (that is, low input or high the weather was favorable—but only household output prices) or who have good networks that 69 allow them to ensure they are getting inputs of income, and that technology adoption requires good quality, returns to using inputs are positive complementary investments to be made. Bandiera and it will be these farmers who use inputs and et al. (2015) show that when credit and extension is experience higher returns. This is what is observed offered together they increase crop income by 50 in Tables 4.3 and 4.4. A small proportion of farmers percent. This effect is not observed when extension use inputs and, on average, farmers who use inputs or credit is provided alone. They also find that face high returns. extension has the largest impact on crop income for households that are between 30 and 60 minutes to 152. Other recent research highlights that the nearest trading center. This could reflect the fact complementary investments in credit, that it is easier for these households to purchase extension, and markets are needed to inputs or that it is easier for these households to sell encourage crop income growth. An ongoing their output, increasing the economic return from impact evaluation provides evidence that farmers increased production. face multiple constraints in improving crop THE EXTERNAL ENVIRONMENT 153. Changes in the external environment can an increase in the number of fatalities reported in have an impact on crop income directly and/ 2011 but this fell again by 2012. Weather conditions or indirectly through the way that households were in general good, with rainfall deficits less produce. For example, good weather has a than 20 percent in most cases. However, 2010 direct impact on crop income by determining was a challenging year for households and higher production quantities but it can also affect crop losses were observed (although no higher than 30 income indirectly through the household’s decision percent). Maize and beans prices increased from to apply inputs as a response to weather. Good 2006 to 2010. The real price of beans continued prices for crops increase crop income but they to rise in most markets in 2011, but maize prices also increase the incentives to produce and may crashed in that year, recovering in the subsequent encourage increased input use or labor as a result. season. Changes in the external environment that may have affected crop income are analyzed by looking at the 155. Changes in wholesale market prices may impact of wholesale market prices, weather shocks, reflect the beneficial effects of improved conflict fatalities, and changes in market access. infrastructure investments, increased efficiency in domestic markets, and 154. There were marked changes in the external development of new export markets. Markets environment from 2006 to 2012: conflict in in the north and east have been improving since the north ceased, prices increased but were 2006 because of infrastructure investments, new volatile, and, in general, the weather was good. export markets opening up in South Sudan and in Figure 4.3 presents data on weather, price, and Kenya, and improved access to market information conflict by region across the years considered in (because of the growth of the ICT sector) and the study. Conflict with the Lord’s Resistance Army growth in trade services which improved efficiency in the Northern region of Uganda (also affecting in markets. Svensson and Yanagizawa (2009) shows households in the northern parts of the Central that improved access to market information helped region) was stabilized in 2008 and the impact of this farmers who were better informed to bargain for is seen clearly in the reduction of conflict related (and receive) higher prices. However, changes in fatalities reported in Armed Conflict Location and supply and demand conditions within and outside Event Data (ACLED) from 2006 to 2010. There was of Uganda also have a large impact on price trends. 70 Figure 4.3: Price, conflict, and weather trends from 2005/06 to 2011/12 Source: Rainfall: Staff calculations using geoWRSI v 3.0, with global Potential Evapo-Transpiration (PET) and Rainfall Estimate (RFE) v2 (2001–2014) time series. Fatalities: ACLED. Prices: UBOS market price data collected for the CPI. Note: WRSI = Water Requirement Satisfaction Index. 156. Good rainfall and price changes account for 51 increase in the price of maize or beans increases percent of the improvement in crop income crop income by 4.5 and 9.2 percent, respectively. for all households and 66 percent of the Incomes of poorer households (those in the bottom improvement in crop income for the bottom 40 40 percent in 2005) are even more dependent percent. The strongest drivers of changes in crop on climate and prices. This is likely because the incomes are changes in rainfall and prices (Table majority of poorer households are located in the 4.4).39 40 A 10 percent increase in water sufficiency Northern and Eastern regions and farming in these increases crop income by 9.9 percent. A 10 percent areas is more likely to be unimodal and experience 39. Only those variables that can be considered to represent the external environment are included in these regressions. This is done for two reasons. First, given these variables have an impact on production practices, a regression that includes production practices as independent variables does not allow the full impact of changes in the external environment to be captured. Secondly, given these variables are exogenous to household production decisions, they provide more robust estimates of drivers of changes in income. It is possible that changes in distance to market and provision of extension services in the community are not fully exogenous, with investments in infrastructure and services being targeted to communities that are more (or less) agriculturally productive. Regressions are also run in which distance to market and provision of extension are excluded, leaving only prices, weather, and conflict. The regression results presented do not include year fixed effects given the objective of the analysis is to explain changes in crop income across years. However, it is possible that other differences across years, correlated with changes in the external environment, are driving the results. To test this, a regression model including year fixed effects is also estimated. The results show the continued significance of weather, prices, peace, and extension provision. 40. As a final robustness check, a specification was run in which prices of regional crops—matooke in the center and west, and cassava in the north and east—were included instead of beans prices (results not shown). These results also showed the same findings: production practices played a role, but changes in the external environment were the main drivers of changes in crop income in Uganda. 71 larger yield variation because of rainfall. Regional 157. Peace is strongly associated with increased variations are explored further below (Table agricultural income growth. Every 1 percent 4.5). For households in the bottom 40 percent, a reduction in the number of fatalities in a 25 km 10 percent increase in rainfall and a 10 percent radius of the village was associated with crop increase in maize and beans prices, results in a 13.4 income growth of 1.3 percent. The establishment percent and 13.0 percent increase in crop income, of peace observed between 2006 and 2010 was respectively. Changes in distance to local market associated with a doubling (a 112 percent growth) had no effect on crop income growth. in crop income. 72 Table 4.4: Drivers of agricultural income growth Bottom 40 Bottom 40 All Households All Households Percent Percent Farming practices Total area planted self-reported, in Ha 0.00734** 0.00846   (0.00313) (0.00674) Renter (land) 0.0682 −0.0343   (0.126) (0.189) Use of fertilizer 0.0846 0.217**   (0.0523) (0.0904) Use of pesticides 0.149*** 0.147**   (0.0479) (0.0695) Used improved seeds/seedlings 0.0238 0.0407   (0.0549) (0.0760) Hired labor used 0.148*** 0.209*** (0.0475) (0.0653) Log of number of days of family labor 0.173*** 0.231*** (0.0343) (0.0476) External environment Distance to output market (zkm) −0.00613 −0.00747 −0.0260 −0.0135   (0.0194) (0.0304) (0.0259) (0.0382) Any extension in village in past 12 months 0.0600 –0.00359 0.200*** 0.222***   (0.0457) (0.0726) (0.0554) (0.0839) Log of rainfall (percent of needs measured 0.986*** 1.356*** 2.064*** 2.683*** by WRSI) (0.196) (0.280) (0.362) (0.541) Log of maize price 0.446*** 0.544*** 0.439*** 0.609*** (0.0674) (0.0970) (0.0879) (0.118) Log of beans price 0.922*** 1.295*** 1.046*** 1.191*** (0.143) (0.213) (0.166) (0.232) Log of number of fatalities 0.00849 0.0406 −0.132** −0.152* (0.0413) (0.0577) (0.0606) (0.0787) Constant −2.048 −7.283*** −6.788*** −11.71*** (1.521) (2.252) (2.143) (3.007) Observations 5,145 2,501 6,184 2,991 Number of HHID 1,806 871 1,962 934 Source: Staff calculations using UNPS 2006–2012. Note: Dependent variable is log of real per capita crop income. Household fixed effects are included. Robust standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1. REGIONAL VARIATIONS 158. The level of agricultural income varies high, this reflects the fact that a much lower share substantially across regions. High levels of of total income in the Central region comes from crop income are recorded in the Western region agriculture. In the Northern and Eastern regions, and the lowest levels of crop income are seen in agricultural income is the dominant source of the Northern region (Figure 4.4). Although crop income, as in the Western region, but overall levels income in the Central region is not particularly of income are much lower. 73 159. Agricultural income growth also varies across negative growth rate in the north and east resulted regions, and was negative between 2010 and in both regions falling behind the center and west 2011 in the east and north. Although growth (Figure 4.4). recovered between 2010/11 and 2011/12, the Figure 4.4: Regional differences in per capita crop income growth, 2005/6 to 2011/12 Source: Staff calculations using UNPS 2005/06–2011/12. 160. The external environment was changing in Western and Central regions are more economically different ways across the four regions during developed. They have had many more years of this period. The Northern region in Uganda is the stability than the Northern region and these regions most drought prone and although rainfall was, have seen substantial development during this in general, good during 2005/06 to 2011/12, the time. More stable climatic conditions and rapid rainfall shortfall in 2009/10 was much larger in the urban growth in and around Kampala has also north than elsewhere in the country (Figure 4.3). helped. The role of the external environment on The Eastern region also experienced quite variable crop income growth is analyzed separately for the rainfall. 41 The north is also the part of the country four regions (Table 4.5). that experienced conflict until the cessation of hostilities in the late 2000s, and, thus, it saw the 161. Weather is a strong driver of crop income largest change in the number of fatalities due to growth in the north and east, but not in other conflict related violence. Maize prices are expected regions. Weather is particularly important in the to be particularly important in the north and east, north: a 10 percent rainfall shortfall results in a both because of its predominance in production reduction in crop income of 38.3 percent in the in the east, but also because a lot of maize trade north (compared to 8.7 percent in the east). with Kenya and South Sudan goes through these regions. There are also large and increasing 162. Prices have been important in all regions, but regional variations in welfare across Uganda. The maize prices have only been important in the 41. Data also suggest larger losses on average in the west across the four years, but this may be because a maize model has been used to calculate the losses while this is not a crop grown in the west. The inclusion of regional dummies or household fixed effects controls for this persistent difference in the analysis. 74 north and east. A 10 percent reduction in the 163. The importance of the external environment in maize price results in a 6.6 percent and 11.1 percent bringing about crop income growth is strongest reduction of agricultural income in the east and in the north, followed by the east, making north, respectively, while it had no impact in the growth in these regions particularly vulnerable center and west. Beans prices are important in all to shocks. These are also the regions that regions, with a 10 percent increase in the beans experienced negative income growth from 2010 to prices increasing income by 6.3 percent to 13.5 2011, highlighting that while the dependence on percent across regions. The results also indicate the external environment benefited households in that the cessation of violence in the late 2000s only these regions, when peace was being established, affected crop income growth in the north. rainfall was good, and prices were rising, it hurt them when rainfall fell and when maize crop prices collapsed in 2011. Table 4.5: Changes in agricultural income: a regional story Centre East North West Log of rainfall (percent of needs measured by WRSI) −0.335 0.868** 3.826*** 0.283 (0.825) (0.370) (0.578) (0.524) Log of maize price 0.243 0.657*** 1.112*** 0.00646 (0.219) (0.114) (0.166) (0.132) Log of beans price 0.627* 0.936*** 1.348*** 1.074*** (0.340) (0.318) (0.350) (0.203) Log of number of fatalities 0.129 −0.0721 −0.131** −0.0521 (0.167) (0.149) (0.0651) (0.221) Constant 7.595 −1.921 −21.00*** 4.008 (5.108) (2.906) (3.686) (3.361) Observations 1,585 2,114 2,253 1,856 Number of HHID 504 674 735 626 Source: Staff calculations using UNPS 2005/06–2011/12. Notes: Dependent variable is log of real per capita crop income. Household fixed effects analysis with robust standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1. 4.2 Weather, prices, peace, and consumption growth 164. Can prices and weather explain the growth The impact of positive trends in prices, weather, in consumption observed from 2006 to 2012, and peace on household consumption growth is given their importance in driving agricultural explored in Table 4.6. Column 1 reports the results income growth, and is peace as positively for crop income that were discussed in section 4. associated with consumption growth as Columns 2 to 5 detail results for livestock income, it is with agricultural growth? Agricultural agricultural wage income, non-agricultural income, income is the most important source of income and nonfarm self-employment income. Column 6 for households, particularly those in the bottom examines the impact on household consumption 40 percent, but it is only one of many sources using consumption data for 2006 and 2010, the of income (Chapter 3). Can these drivers of crop years for which comparable consumption data was income growth explain consumption growth, collected in the UNPS. Table 4.7 presents the same particularly among those that were poor in 2006? results for the bottom 40 percent. 75 165. Good weather and higher prices were households to protect consumption from the important drivers of consumption growth but downside of risk is essential to securing gains in the impact is more muted than the impact on welfare for these households. income. A 10 percent increase in water sufficiency results in consumption growth of 4.8 percent in per 168. Weather has a smaller impact on consumption capita consumption (4.1 percent when considering than income because households have households in the bottom 40 percent in 2005/06) diversified sources of income and bad weather compared to its impact of 9.9 percent crop income is compensated by higher non-agricultural growth. A 10 percent increase in the price of maize income. Rainfall shocks do not affect income and beans results in consumption growth of 5.1 from livestock. However, wage employment and percent. The impact is almost double for the self-employment out of agriculture is significantly bottom 40 percent—a 10 percent price increase negatively affected by poor weather. The results results in 10.5 percent consumption growth. suggest that diversification of productive activities can be an important risk hedging strategy for 166. The consumption of households in the north households in Uganda, particularly the poorest. If and east is more reliant on prices and weather agricultural income is affected by climate shocks, than the consumption in wealthier households households can offset this with increased nonfarm in the center and west. Given the limited income. It is not clear whether household labor sample size, households in the north and east are is pulled into own-farm agricultural production pooled together in the regression analysis, as are because of the increased demand for agricultural households in the center and west. Also just beans labor when the rainfall is good or whether prices were considered (Table 4.8). The difference is household labor is pushed out of agriculture a largest when considering prices where a 10 percent result of a desperate need to smooth consumption increase in the beans price is associated with a 6.7 when rainfall is bad. However, although some percent increase in consumption in the north and of the weather shock can be insured through east and a 2.5 percent increase in consumption in diversification, the fact that weather still affects the center and west. consumption shows that households are not able to fully insure their consumption from the impact 167. The dependence of consumption on weather of weather. and prices can be a source of welfare improvements when the weather is good 169. In contrast, price decreases affect all sources and prices are rising, but it also puts welfare of income negatively. This means that when gains at risk of being reversed if the weather prices are good, total income is positively affected, fails or prices fall. This reliance on the external but conversely when prices are bad, households environment contributes to the high levels of are not able to mitigate crop income shortfalls vulnerability to poverty that are observed in by increasing income from other sources. The Uganda. Indeed this was observed for many exception to this is agricultural wage income, which households in the north and east in 2011. Poor is surprising, given findings in other countries, prices resulted in lower incomes and consumption that agricultural wage labor is positively affected and this decline in welfare had not fully been by crop price increases and the expectation that reversed by 2012. In the North and East, the greater higher prices would result in increased demand reliance of households on weather and prices has for agricultural wage labor. It is not clear why a both been a source of welfare improvements and negative relationship is observed in this context. vulnerability for Northern and Eastern households. The impact of prices on consumption is, however, Ultimately increasing the resilience of these smaller than the impact of prices on crop income, 76 indicating that even though households are not significant relationship with consumption is able to diversify to manage price risk, they are able not observed. The results suggest that this may to reduce the impact of prices on consumption by be because households switched out of wage other means. labor activities into self-employment activities in agriculture as peace was restored. Further analysis 170. Although the cessation of violence is positively is needed to confirm this finding. associated with crop income growth, a Table 4.6: Impact of weather, prices, and peace on income and consumption   (1) (2) (3) (4) (5) (6)   Non- Nonfarm Self- Consumption Crop Livestock Agricultural agricultural employment (2005/06, Income Income Wage Income Wage Income 2009/10) Income Log of rainfall 1.886*** −0.198 −4.853*** −3.627*** −2.796*** 0.478*** (percent of needs (0.343) (0.833) (0.706) (0.701) (0.750) (0.147) measured by WRSI) Log of maize price 0.492*** −0.0671 −1.130*** −0.0973 −0.401 −0.218** (0.0840) (0.264) (0.338) (0.339) (0.371) (0.0975) Log of beans price 1.091*** 1.213** −1.453*** 4.263*** 1.175** 0.729*** (0.155) (0.516) (0.506) (0.422) (0.506) (0.125) Log of number of −0.146*** −0.227 0.451*** 0.323** 0.177 −0.00909 fatalities   (0.0515) (0.142) (0.135) (0.134) (0.145) (0.0143) Constant −6.619*** −0.196 36.68*** −12.84** 9.026 5.127***   (2.010) (5.792) (5.804) (5.190) (6.101) (1.172) Observations 6,852 6,986 6,497 6,497 6,497 3,154 Number of HHID 2,044 2,046 2,045 2,045 2,045 1,946 Source: Staff calculations using UNPS. Notes: Household fixed effects estimation with robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. 77 Table 4.7: Impact of weather, prices, and peace on income and consumption: Bottom 40 percent   (1) (2) (3) (4) (5) (6)   Non-agricul- Nonfarm Livestock Agricultural Consumption Crop Income tural Wage Self-employ- Income Wage Income (2005/6, 2009/10) Income ment Income Log of rainfall 2.417*** 1.533 −6.419*** −4.366*** −3.335*** 0.405** (percent of (0.506) (1.177) (1.017) (0.946) (1.015) (0.190) needs measured by WRSI) Log of maize 0.715*** 0.437 −1.418*** −0.281 –0.435 −0.00504 price (0.113) (0.377) (0.519) (0.466) (0.504) (0.122) Log of beans 1.247*** 2.019*** 0.0956 4.078*** 1.678** 1.049*** price (0.214) (0.722) (0.752) (0.542) (0.701) (0.140) Log of number of −0.187*** –0.287 0.690*** 0.455*** 0.103 −0.00918 fatalities   (0.0662) (0.177) (0.185) (0.170) (0.184) (0.0163) Constant −11.48*** −16.37** 35.56*** −8.024 7.701 1.782   (2.818) (8.120) (8.563) (6.843) (8.198) (1.463) Observations 3,334 3,359 3,102 3,102 3,102 1,502 Number of HHID 966 966 964 964 964 927 Source: Staff calculations using UNPS 2005/06–2011/12. Notes: Household fixed effects estimation with robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. Table 4.8: Welfare changes: A regional story Centre and West North and East Log Real Consumption Per Capita Log of rainfall (percent of needs measured by WRSI) 0.444** 0.488** (0.825) (0.219) Log of beans price 0.245* 0.674*** (0.130) (0.122) Log of number of fatalities 0.057 0.004 (0.053) (0.016) Constant 7.239 4.050** (1.603) (1.701) Observations 1,585 1618 Number of HHID 504 1022 Source: Staff calculations using UNPS 2005/06–2009/10. Notes: Dependent variable is real per capita consumption. Month of interview dummies included but not shown. Household fixed effects estimation with robust standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1. 78 171. Weather and prices also affect nutritional these regressions much smaller. For this reason outcomes. A 10 percent reduction in rainfall only one price—the prices of beans—is considered. reduced the weight for age of children under 5 Although the results are not consistently significant years in the bottom 40 percent by 5.9 percent. across specifications, they do show that weight Thus far, the results presented have relied on a for height and weight for age is positively affected monetary dependent variable and thus prices by rainfall and by higher prices, as suggested by have both been part of the construction of the the regressions on income and consumption. A 10 dependent variable as well as an explanatory percent reduction in rainfall reduced the average variable included in the analysis. As a robustness weight for age z-score of children under 5 years by check on the findings of the analysis, a non- 3.5 percent. This impact increases to 5.9 percent monetary measure of welfare that is correlated with for children in the bottom 40 percent. These results consumption is used: z-scores (standard scores) indicate that changes in crop income do improve of weight for age and weight for height among nutritional outcomes, even though evidence in children less than 5 years of age in the household. Chapter 2 shows that this is not all that matters and Results are presented in Table 4.9. This data was children in wealthier regions are not necessarily less only collected from 2010 onward and only collected likely to suffer from malnutrition. for children, making the sample size available for Table 4.9: Impact of weather, prices, and peace on weight for age and weight for height   (1) (2) (3) (4) Weight for Weight for Weight for Weight for Age Z-score Height Z-score Age Z-score Height Z-score Log of rainfall (percent of needs 0.364** 0.397 0.586*** 0.512 measured by WRSI) (0.158) (0.381) (0.223) (0.669) Log of beans price 0.194 0.704 0.364 1.213** (0.284) (0.434) (0.404) (0.609) Log of number of fatalities 0.0259 −0.0230 0.0536 0.0330 (0.0503) (0.0738) (0.0638) (0.0885) Constant −3.798* −6.413* −6.009** −10.35* (2.059) (3.801) (2.914) (6.139) Observations 1,658 1,643 803 801 Number of HHID 957 953 465 465 Bottom 40 percent  No No Yes Yes Source: Staff calculations using UNPS 2010–2012. Notes: Household fixed effects estimation with robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. In each case the dependent variables is averaged across all children below 5 years in the household. 79 4.3 Increasing the resilience of Ugandan households 172. Formal safety nets are available to very few 173. Are households with a higher level of human households in Uganda. The results presented capital and access to financial instruments, in Section 4.3 suggest that it is desirable for such as having a savings account and having households to be more fully insured against a loan, better able to smooth the impact of shocks than they currently are. UNPS households climate shocks and price declines? The only were asked to report the most important types factor that helped households to mitigate the of coping mechanisms used if they faced an adverse effect of shocks was the level of education adverse shock in the last year (the answers were of the household head. Households that have not mutually exclusive). As seen in Figure 4.5, a savings account or a loan from a financial households rely on savings (35 percent) and help institution are not more resilient to these shocks. from family (25 percent) to mitigate the impact of Similarly, enhanced access to markets where shocks. Very few report receiving support from the agricultural inputs are sold and where agricultural government, highlighting the absence of reliable products are sold as well as technical assistance, official safety net programs. Safety nets provided do not make a difference in the way households by savings, family, and friends are of paramount are affected by climate shocks and crop price importance in the absence of official safety net declines.42 programs. However, reliance on informal insurance mechanisms has been shown to reduce incentives 174. Higher levels of education of the household for productive investments among rural households head reduce the negative effect of rainfall in Uganda (Fafchamps and Hill 2015). shocks on consumption, compared to households where the head has no education at all. Having primary education reduces the effect of a weather shock by 2.8 percent compared to those with no education, while for those with complete secondary education, the reduction increases to 4.9 percent (Figure 4.6). Something similar occurs if we look at the effect of climate shocks on per capita consumption, albeit the magnitude is smaller: having some secondary education implies a 1.4 percent reduction in the intensity of the shock for households in the bottom 40 percent. More education facilitates diversification by enabling increased participation in the labor market, particularly in the non- agricultural sector. In addition, individuals who are more educated may assess and respond to risk more successfully. In both cases, crop income and per capita consumption, the higher the education level, the larger the impact for the households that belong to the bottom 40 of the distribution. 42. Instead of using the subjective responses of households, objective measures were used. For example, instead of using the response that the household used savings as a coping mechanism, an indicator that the household has a savings account was used. 80 Figure 4.5: Self-reported coping mechanisms Source: Nikoloski et al. (2015). Figure 4.6: Education mitigates the impact of climate shocks Source: Staff estimation using UNPS 2006–2012. Note: Results statistically significant at the 10 percent level for crop income. For consumption, only ‘some secondary’ education for the bottom 40 percent is statistically significant at the 5 percent level. 4.4 Conclusion 175. Agricultural incomes grew because the improving since 2006 because of infrastructure government got some key fundamentals investments, new export markets opening up right that provided the incentives to invest in South Sudan and in Kenya, better market time in agricultural production and engage information for farmers and traders (because of the in agricultural markets. Conflict with the Lord’s development of a well-functioning ICT sector), and Resistance Army in the Northern region of Uganda growth in trade services, which improved marketing was stabilized in 2008 and this had a positive efficiency. This has contributed to real relative price impact on crop income. In addition, markets, increases for agricultural commodities that poor particularly in the north and east, have been farmers grow and sell. 81 176. Luck was also on Uganda’s side: good weather gains in poverty reduction. This is indeed what benefited many households and positive price happened in the Northern and Eastern regions in trends in international and regional markets 2011. Households need to be able to both benefit aided real crop price increases. Prices reflect from good prices and weather and have access to not just improvements in marketing efficiency, coping mechanisms, such as public safety nets, to but also favorable changes in supply and demand be protected from low prices and poor weather. conditions within and outside of Uganda. Peace Productivity growth in agriculture—possibly in South Sudan and the Democratic Republic through the use of improved seeds, fertilizer, of Congo provided new sources of demand for pesticides, and irrigation—and diversification to Ugandan food production. Good rainfall and prices other more remunerative forms of employment can account for 51 percent of the improvement in crop also improve resilience. This requires addressing income for all households and 66 percent of the the challenge of low quality agricultural inputs and improvement in crop income for the bottom 40 constraints (such as credit, extension, and access to percent. markets) that some farmers face. 177. Although households increased the volume 179. For agricultural growth to be truly inclusive, that they marketed during this time, there was it needs to address the gender productivity very little change in the nature of agricultural gap that still persists in agriculture. One of production. In the bottom 40 percent, the share the biggest constraints female farmers face in of households selling crops increased from 60 comparison to male farmers is their limited access percent in 2006 to 72 percent in 2012. When to labor and high childcare demands. Lowering extension services were provided crop income was the fertility rate will help address this constraint in 20 percent higher, but few households received the long term, but exploring community childcare extension services. Extension services expanded may provide some gains in the immediate term. In but from 8 percent of households in 2006 to 12 addition improving access to inputs and tailoring percent of households in 2012. There was little extension services toward women may help growth in the use of improved inputs and as a result address the fact that women currently use fewer modernization of agricultural practices contributed inputs and gain less from extension when they do very little to crop income growth. receive it. 178. The reliance on weather and prices also offers some cause for concern. When prices are poor or when the rains do fail, crop income growth falters and consumption falls, reversing 82 83 NON-AGRICULTURAL CHAPTER: GROWTH IN UGANDA 5 High growth in value-addition in industry and services has not been accompanied by a higher share of the workforce being employed in these sectors, limiting the degree to which these sectors contributed to poverty reduction. 180. Uganda has experienced high growth in industry and services when compared to the regional average. Between 2003 and 2014, the mean annual growth rates of industry and services were 12.2 percent and 8.2 percent, respectively, which were higher than the average of developing Sub-Saharan Africa (3.5 percent for industry and 7.5 percent for the service sector, Figure 5.1). On the other hand, Uganda’s mean annual growth rate of agriculture value added was 2.0 percent during the same period, which is lower than the average of developing Sub-Saharan Africa (5.4 percent). 181. The growth of services was largely driven by the expansion of posts and telecommunication services, which reflects the rapid growth of the ICT sector. As shown in Figure 5.1.4, the fastest growth within the services sector came from post and telecommunication services. 182. High growth in value-addition in industry and services has not been accompanied by a larger proportion of the workforce employed in these sectors, suggesting that the job creation brought about by non-agricultural growth has only just kept up with population growth. Table 5.1 summarizes the source of household income in 2006 and 2013. In 2006, 60 percent of households had income from non-agricultural sectors, while only 49 percent of the 84 bottom 40 percent households had income from associated with higher consumption growth (Table non-agricultural sectors. In 2013, the proportion 3.3). In addition, Chapter 4 provides evidence that of households with non-agricultural income was the ability to diversify into non-agricultural income very similar, decreasing slightly to 59 percent for sources when agricultural conditions are less all households and 47 percent for the bottom 40 favorable has helped households be more resilient percent households. However, specialization in to shocks in agricultural income (Table 4.6). these sectors did increase over this period: in 2006, 12 percent of all households were specialized 184. This chapter examines which households have in industry and services, and by 2013, this had experienced non-agricultural income growth increased to 17 percent. This is consistent with and what constrains further non-agricultural the discussion in Chapter 3 that since 2006 little income growth. It examines income from both additional income diversification has been self-employment and wage employment in non- observed. agricultural activities. The chapter focuses on constraints faced by households, and does not 183. However, the majority of Ugandan households examine what has constrained firm creation of jobs derive some form of income from industry in non-agricultural sectors in Uganda. Addressing and services, and growth in this income the constraints households face in increasing has increased consumption and reduced non-agricultural income will help increase the vulnerability. Chapter 3 highlights that, although inclusivity of non-agricultural growth, but more agricultural income growth is more strongly fundamentally, stronger job-creating firm growth is associated with consumption growth for the needed to drive poverty reduction in this area and bottom 40 percent, non-agricultural income this requires addressing the constraints firms face in growth—particularly from self-employment—is also growing and in creating new jobs. 85 FIGURE 5.1: Sectoral growth in comparison to the region 1: Growth in value addition in 2: Growth in value addition in industry agriculture (Index 2003 = 100) (Index 2003 = 100) 3: Growth in value addition in services 4: Real growth in value addition across (Index 2003 = 100) selected sectors (Index 2003/04 = 100) Source: 1–3: WDI; 4: Uganda Systematic Country Diagnostic Table 5.1: Source of household income by sector All Households Bottom 40% 2006 2013 2006 2013 Agriculture only 40% 41% 51% 53% Industry only 2% 3% 1% 2% Services only 8% 11% 2% 5% Agriculture and industry 14% 13% 18% 16% Agriculture and services 27% 22% 21% 18% Industry and services 2% 3% 1% 1% All sectors 6% 5% 6% 3% Source: Authors’ calculations using UNHS 2006 and 2013 86 5.1 Characteristics of households that have experienced non-agricultural income growth 185. There are significant movements both in and agricultural sectors. The net changes in the out of non-agricultural sectors as households percentages of the bottom 40 percent households, adjust their time spent in agricultural and non- which engage in both non-agricultural wage agricultural activities depending on the returns employment and self-employment, are negative to these activities in a given year. The UNPS in all periods, because more households in the was analyzed to examine how many households bottom 40 percent exited non-agricultural wage move in and out of wage employment and self- employment and self-employment than went employment in non-agricultural sectors over time. into non-agricultural wage employment and self- Table 5.2 shows this change for three periods: employment. Table 5.2 confirms that in net terms, a 2006 to 2010, 2010 to 2011, and 2011 to 2012. The higher proportion of the bottom 40 percent moved net changes in engagement in non-agricultural into agriculture (out-of-wage and self-employment) wage employment and self-employment were than wealthier households. This is consistent close to zero in most periods, confirming the trend with findings in Chapter 4 that households reduce reported in Table 5.1. This is consistent with the nonfarm income when external conditions for finding in Chapter 4 that households increase and agricultural production are favorable (prices are reduce their income in nonfarm activities based on high and weather is good), as was the case during whether conditions—namely weather and prices— this period. This is also consistent with the findings are favorable to agricultural production in a given by Nagler and Naude (2014) that higher income is year. associated with the probability of having non- 186. More households in the bottom 40 percent agricultural enterprises. exit non-agricultural sectors than enter non- Table 5.2: Moving in and out of non-agricultural employment   Moving into… Moving out of… Net Change in… Wage Self- Wage Self- Wage Self- Employment employment Employment employment Employment employment All households 2006 to 2010 10.5% 19.0% 12.4% 17.4% −1.9% 1.6% 2010 to 2011 17.3% 20.1% 18.6% 21.6% −1.3% −1.5% 2011 to 2012 17.0% 18.9% 17.3% 21.3% −0.3% −2.4% Bottom 40 percent 2006 to 2010 5.3% 13.1% 11.4% 17.8% −6.1% −4.8% 2010 to 2011 14.1% 15.6% 18.0% 20.2% −3.9% −4.5% 2011 to 2012 14.2% 18.6% 18.4% 21.5% −4.2% −2.9% Source: Authors’ calculations using UNPS 2006–2012 87 187. Those that were able to increase their self- growth in nonfarm self-employment income more employment income were more likely to live likely, but for the bottom 40 percent, it is access to in households that were headed by young, own savings that is most important, not access to educated men with better access to finance. credit. Table 5.3 presents the characteristics of households 188. Those that saw growth in wage income were that saw income growth in non-agricultural income. also more likely to be in households headed Data is presented for 2011 to 2012, but similar by young, educated males, but education results hold for different periods. The first columns appears to be more important for the bottom describe the characteristics of those that increased 40 percent. In addition, households who increased (or did not increase) self-employment income and their non-agricultural wage income had lower the latter columns describe the characteristics of levels of agricultural income. This suggests that those that increased non-agricultural wage income. non-agricultural wage income is a substitute, rather Households headed by young men are most likely than complement of agricultural income (Table 5.3). to increase self-employment income. Education is Figure 5.2 shows educated individuals are more important, on average, but less so when focusing likely to be engaged in wage employment and less on the bottom 40 percent. Access to finance makes likely to be self-employed. Table 5.3: Characteristics of households in 2011 that experienced non-agricultural growth from 2011 to 2012 All Households Self-employment Income Non-agricultural Wage Income Increased No Increase Increased No Increase Male headed household 0.76 0.69 *** 0.75 0.7 Age of head 44.4 47.4 *** 44.37 46.95 *** Education of head 2.72 2.54 *** 3.07 2.52 *** Has a mobile phone 0.57 0.47 *** 0.61 0.48 *** Distance to market (km) 4.67 5.12 4.44 5.09 Has savings 0.36 0.3 ** 0.36 0.31 Has a loan 0.46 0.4 ** 0.44 0.41 Real crop income (shillings, 540 530 460 540 ** thousands) Land owned 2.31 2.37 1.34 2.5 Bottom 40 Percent Self-employment Income Non-agricultural Wage Income Increased No Increase Increased No Increase Male headed household 0.76 0.68 ** 0.74 0.69 Age of head 44.76 47.94 *** 46.3 47.33 Education of head 2.18 2.24 2.64 2.16 *** Has a mobile phone 0.33 0.3 0.43 0.29 *** Distance to market (km) 7.58 8.22 6.72 8.28 Has savings 0.31 0.24 * 0.31 0.25 Has a loan 0.36 0.35 0.35 0.35 Real crop income (shillings, 410 460 410 450 thousands) Land owned 1.22 1.22 1.31 1.2 Extension visits 0.04 0.05 0.08 0.04 Source: Authors’ calculations using UNPS 2011–12. Note: *** indicates significantly different at 1%, ** indicates significantly different at 5%, and * indicates significantly different at 10%. 88 Figure 5.2: Type of employment and education FIGURE 5.2: Type of employment and education     Source: Staff calculations using UNHS 2013 5.2 Identifying constraints to non-agricultural income growth 189. This section presents findings from panel The study points out that women face limited analysis and recent impact evaluations to networks and information about entering into examine whether gender, education, and male-dominated sectors and that this can constrain access to finance constrains growth in non- their non-agricultural earnings potential. The agricultural incomes. It also looks at the role of findings of this study are consistent with empirical access to infrastructure and markets. evidence from various countries that shows that female entrepreneurs earn lower incomes than 190. Women are generally engaged in lower-earning men (Berge et al. 2014; De Mel et al. 2008). self-employment activities and are less likely to experience self-employment income growth, 191. Poor households have lower educational but women who are running businesses in attainment and face lower gains from moving male-dominated sectors make profits as into non-agricultural wage employment. much as men do. Campo et al. (2015) find that, Figure 5.3 shows that non-agricultural wages controlling for the sector in which a woman works, increase quite rapidly with education in Uganda. women make just as much as men. However, The monthly wage of individuals with some upper women tend to choose less profitable sectors. secondary school education is more than twice Women are more likely to work in sectors that are as large as the monthly wage of individuals with considered female, such as hairdressing and retail some primary school education. Nagler and trade. Women who cross over into male-dominated Naude (2014) report that lower levels of education sectors make as much as men and three times are also associated with lower returns to self- more than women who stay in female-dominated employment. Estimating the returns to education sectors. This study suggests women are self- is challenging as unobservable characteristics selecting themselves into less productive sectors. of individuals—such as work discipline—often 89 determine both an individual’s educational in agriculture, but they are lower. This means attainment and the income they are able to secure. that the gain from moving from agriculture to However, available evidence for Uganda suggests non-agricultural wage employment is lower for that there is a considerable return to education in someone with primary education than for someone the non-agricultural sector. Efforts to control for with secondary or post-secondary education. As endogeneity suggest that the estimated returns discussed in Chapter 1, educational attainment are if anything, underestimated (Lekfuangfu, et is lower among the bottom 40 percent, posing a al. 2012). There are also returns to education constraint to these households. Figure 5.3: Average monthly and hourly wages by the level of education Source: World Bank (2014) “Workforce Development and Returns to Education in Uganda.” 192. Panel analysis suggests that educational regression results in Table 5.4 show that maximum attainment is a determining factor of non- years of education among household members agricultural wage income for the poor. do not influence household income in non- Table 5.4 summarizes the results of fixed-effects agricultural sectors, in either wage employment regression using UNPS data from 2010 and 2011. or self-employment, when we use data for all Dependent variables are logged wage income in households. However, it determines wage income non-agricultural sectors, logged non-agricultural among the bottom 40 percent of households. This self-employment income, and logged total income implies educational attainment is an important in non-agricultural sectors. Because agricultural determinant factor of non-agricultural wage income is endogenous, it is instrumented with employment only for the poor, and there is a the WRSI calculated from satellite rainfall data for potential gain from investment in school education each pixel using a cassava crop model calibrated among the poor. to the growing seasons across Uganda. The 90 Table 5.4: Determinants of non-agricultural household income All Households Bottom 40 Percent Self-em- Self-em- Wage Total Wage Total ployment ployment Log of real gross agricultural −0.773 −0.950 −0.271 −0.698 −0.109 −0.070 income per capita (0.698) (0.675) (0.651) (0.666) (0.654) (0.693) Maximum number of years of 0.055 0.053 0.129 0.424** −0.125 0.362* education in the household (0.112) (0.112) (0.106) (0.196) (0.197) (0.208) Has a savings accounts with 1.813*** 1.641** 1.325** 2.396* 2.511 4.258*** formal institutions (0.670) (0.699) (0.661) (1.397) (1.562) (1.651) Obtained loan in past 12 −0.397 0.531 0.280 0.343 1.107* 1.244** months (0.397) (0.402) (0.383) (0.586) (0.594) (0.628) Distance to nearest population −0.061 −0.074 −0.026 –0.340 0.0450 −0.365 center with +20,000 (km) (0.089) (0.087) (0.083) (0.411) (0.416) (0.440) 0.201 −2.840** −0.694 1.807 −0.532 −0.230 Connection to electricity (1.391) (1.413) (1.359) (3.058) (3.107) (3.283) Number of working age adults 0.857*** 0.531** 0.860*** 0.836*** 0.285 0.639* in the household (0.211) (0.218) (0.208) (0.320) (0.333) (0.351) Observations 3,300 3,106 3,140 1,422 1,360 1,366 Male-headed households only   All Households Bottom 40 Percent Self-em- Self-em- Wage Total Wage Total ployment ployment Log of real gross agricultural −0.678 −0.793 −0.342 −0.571 −0.144 −0.136 income per capita (0.635) (0.610) (0.603) (0.630) (0.635) (0.672) Maximum number of years of 0.0348 0.167 0.143 0.386 −0.156 0.292 education in the household (0.136) (0.132) (0.128) (0.258) (0.269) (0.284) Has a savings accounts with −0.144 −0.094 −0.085 −0.543 0.512 −0.174 formal institutions (0.107) (0.105) (0.101) (0.603) (0.622) (0.656) Obtained loan in past 12 1.592** 1.650** 0.969 2.155 3.350* 4.482** months (0.759) (0.789) (0.768) (1.733) (2.028) (2.140) Distance to nearest popula- −0.286 0.294 0.328 0.000 0.704 0.702 tion center with +20,000 (km) (0.449) (0.446) (0.434) (0.662) (0.696) (0.734) 0.519 −2.056 −0.106 1.796 −0.601 −0.337 Connection to electricity (1.462) (1.462) (1.440) (3.076) (3.183) (3.355) Number of working age adults 0.546** 0.249 0.435* 0.407 0.220 0.205 in the household (0.260) (0.258) (0.250) (0.390) (0.416) (0.438) Observations 2,516 2,352 2,380 1,046 992 998 Female-headed households only   All Households Bottom 40 Percent Self-em- Self-em- Wage Total Wage Total ployment ployment Log of real gross agricultural 0.087 −0.578 0.004 −0.089 −0.366 −0.319 income per capita (0.545) (0.536) (0.479) (0.557) (0.543) (0.536) Maximum number of years of −0.0577 0.094 0.126 0.146 0.207 0.478* education in the household (0.144) (0.144) (0.127) (0.267) (0.261) (0.258) Has a savings accounts with 0.142 −0.086 0.0518 −0.324 −0.005 −0.320 formal institutions (0.130) (0.126) (0.113) (0.465) (0.464) (0.459) Obtained loan in past 12 2.253** 2.684** 2.149** 3.686 5.850** 7.114*** months (1.014) (1.048) (0.908) (2.573) (2.718) (2.684) Distance to nearest popula- −1.267* 0.521 −0.535 −0.846 1.728* 1.759* tion center with +20,000 (km) (0.668) (0.677) (0.603) (1.052) (1.046) (1.032) 1.909 −2.655 0.546 8.624 −1.241 0.138 Connection to electricity (1.867) (1.962) (1.748) (6.260) (6.212) (6.133) Number of working age adults 1.317*** 0.988*** 1.411*** 1.469*** 0.116 1.266** in the household (0.309) (0.307) (0.272) (0.540) (0.540) (0.533) Observations 1,354 1,282 1,298 492 478 480 Note: Instrumental-variables regressions (fixed effects). Log real gross agricultural income per capita (crop and livestock) is instrumented with WRSI calculated from satellite rainfall data for each pixel using a cassava crop model calibrated to the growing seasons across Uganda. 91 193. Poor households have limited access to credit, agricultural self-employment for the bottom 40 but access to credit has improved for the percent of households, even though it did not poor. Better access to loans increased self- increase income from self-employment for all employment income among the poor. Access households. to credit is also a very critical factor for developing 194. Access to savings also is strongly correlated non-agricultural self-employment. However, poor with increased non-agricultural income. There households have had limited access to credits. In is a large gap in access to savings accounts between 2006, 20 percent of all households had household the bottom 40 percent and other households. In members who obtained loans in the past 12 2011, 12 percent of all households had at least months, while only 14 percent of the bottom 40 one member with a savings account with a formal percent had household members who obtained financial institution, while it was only 4 percent loans in the past 12 months (Figure 5.4). In 2012, for the bottom 40 percent of households. The the gap between the bottom 40 percent and other regression results in Table 5.4 suggest that non- households narrowed. Among all households, agricultural income is higher for those with savings 42 percent had members who obtained loans, account with formal institutions, more so than for while 39 percent of the bottom 40 percent had those with credit. Access to savings is significantly household members who obtained loans in the correlated with income even when regressions past 12 months. The gap in the proportion of are run separately for male- and female-headed households, which obtained loans with formal households. This result is consistent with empirical sources, also narrowed between the bottom 40 findings from many countries that savings has percent and other households. In 2006, 28 percent relatively positive welfare impacts than credit (Van of households had members who obtained loans Rooyen et al. 2012). It may be because investment from formal sources, while only 9 percent of the in non-agricultural businesses is often made out of bottom 40 percent had household members who savings. Thus, improving access to savings accounts obtained loans from formal institutions. In 2012, has a great potential to increase non-agricultural the gap between the bottom 40 percent and other income. Mobile money is a promising way to households shrank. Among all households, 44 promote financial inclusion in Uganda. Gutierrez percent had household members who obtained and Choi (2014) report that Uganda has the largest loans from formal sources, while 42 percent of share of the population using mobile phones to bottom 40 percent households had household make monetary transactions, even though half of members who obtained loans from formal sources. the users of mobile money services are unbanked. The regression results in Table 5.4 suggest that access to loans increased income from non- 92 Figure 5.4: Access to finance Source: UNPS 2006–2011. 195. Results of impact evaluations suggest poor to USh 31,300, cash savings tripled, and short-term women benefit from cash grants and business expenditures and durable assets increased 30 training, as they are the most financially percent to 50 percent relative to the control group constrained. An earlier analysis suggests female- which did not receive cash grants or training. The head households are less likely to be able to program had the strongest impacts on the people increase non-agricultural income. Blattman et al. with the lowest levels of capital and access to (2016) provided women in poor households with credit. Their finding is consistent with the meta- cash grants of approximately US$150 and basic analysis that financing support is more effective for business skills training in a war-affected region in women compared to other interventions, because northern Uganda. The women were encouraged to poor women are the most credit-constrained start retail businesses. Most started and sustained group of people in the society (Cho and Honorati small retail businesses with the cash grant, while 2014). However, Fiala (2015) offered either capital they continued farming. A year after the program, with repayment (subsidized loans) or without monthly cash earnings doubled from USh 16,500 (grants) to both male and female microenterprise 93 owners in poor households and randomly offers electricity at home, while 19.6 percent of the top 60 business skills training. He found no effect for households have electricity at home. Golumbeanu female enterprises from either form of capital or the and Barnes (2013) report that a very simple training, but found large effects for men with access home wiring costs about US$108 in Uganda and to loans combined with training. a security deposit of US$43 is required to obtain electricity at home. The total connection charge is 196. Impact evaluation studies provide evidence 61.6 percent of the average monthly income. This that there is strong demand for financial implies it is hard for the poor to afford electricity. and skill training programs among youth, The regression results in Table 5.4 indicate that especially among women, and such programs access to electricity is not a determining factor of can increase their earnings. Blattman et al. non-agricultural income. However, because the (2014) conducted an unconditional cash transfer percentage of households with connection to program for youth, and followed young adults for electricity is so low among the bottom 40 percent, two and four years after receiving grants equal it is difficult to conclude that there is no impact of to annual incomes. Most started new skilled access to electricity on non-agricultural income trades and labor supply increased by 17 percent. for the poor. Poor households also tend to live far Earnings rose nearly 50 percent, especially among from cities. Figure 5.5 shows that the bottom 40 women. This suggests that young women face percent of households live around 25 km away larger financial constraints than young men do. from cities with a population of at least 20,000 Bandiera, Goldstein et al. (2010) analyzed the people. The regression results in Table 5.4 do not intention to participate in training programs of indicate that the distance to cities affects non- adolescent girls (Bangladesh Rural Advancement agricultural income. However, Nagler and Naude Committee’s Adolescent Development Program). (2014) demonstrate non-agricultural household The program emphasizes the provision of life skills, enterprises located up to 10 km from a population entrepreneurship training, and microfinance. They center are the most productive, followed by found that the program attracts girls who are likely household enterprises residing up to 25 km and to place a high value on financial independence: 50 km away, respectively, as shown in Figure 5.6. single mothers and girls who are alienated from Their results suggest that the poor engaged in their families. non-agricultural self-employment may benefit from 197. Access to electricity and markets does not living near towns. Land size was also included as seem to influence non-agricultural household an independent variable in all regressions but it income. As discussed in Chapter 1, only 1.7 percent was not significantly correlated with income. This is of the bottom 40 percent of households have consistent with Table 5.3. Figure 5.5: Distance to nearest population center with +20,000 (km) Source: UNPS 2006–2011. 94 FIGURE 5.6: Nonfarm self-employment productivity and distance Source: Nagler and Naude (2014). 5.3 Conclusion 198. This chapter examined which households 199. The chapter also examined what constrains have experienced non-agricultural income further non-agricultural income growth, and, growth, both in self-employment and wage in particular, examined the findings from employment. Uganda has experienced high randomized controlled trials undertaken in growth in industry and services when compared Uganda to identify what interventions would to the regional average. However, high growth help increase non-agricultural income growth. in value-addition in industry and services has Those that were able to increase their self- not been accompanied by a higher share of the employment and wage income were more likely workforce being employed in these sectors, limiting to live in households that were headed by young, the degree to which these sectors contributed to educated men with better access to finance. Results poverty reduction. The growth in these sectors did of impact evaluations suggest that poor women not result in job creation faster than population can benefit from cash grants and business training, growth. The net changes in the percentages of the as they are the most financially constrained. bottom 40 percent of households, which engage Randomized controlled trials (RCTs) provide a clear in both non-agricultural wage employment and indication of the types of interventions that work; self-employment, are negative, because more however, they are often implemented on a small households in the bottom 40 percent exited scale. It is not clear whether these interventions non-agricultural wage employment and self- will also work at scale for growing self-employment employment than went into non-agricultural wage and encouraging income diversification among employment and self-employment. the poor. More empirical evidence is needed on programs implemented at scale. 95 MOVING OUT AND UP : CHAPTER: MIGRATION AND POVERTY 6 IN UGANDA 43 Most of Uganda’s rural migrants tend to move within their own region or to another rural area. Migration generates substantial welfare gains—with even larger gains accruing to those who migrate to urban areas. 200. While the bulk of Uganda’s 35 million inhabitants live in rural areas, the country is urbanizing at a considerable pace. According to recent census data, the country’s overall population density grew by 41 percent between 2002 and 2014 and the share of Uganda’s population living in urban areas increased by more than 50 percent (from 12.1 percent to 18.4 percent) over the same period (UBOS 2014b). An alternative measure of urbanization that is comparable across countries, the agglomeration index, suggests that Uganda’s urban share is actually higher than these rates would suggest, at 25 percent (World Bank 2012). 201. Urbanization has been an important driver of poverty reduction from 2006 to 2013, because of the much lower rates of poverty present in urban areas. Chapter 3 highlighted that urbanization accounts for 10 percent of the poverty reduction achieved from 2006 to 2013. 202. Migration, in addition to demographics and redistricting, contributes to urbanization. In Sub-Saharan Africa, lower mortality rates in urban areas result in higher natural population growth rates in urban areas, even in the presence of lower fertility rates (Jedwab 43. his chapter draws on the background paper: “Moving Out and Up: Panel Data Evidence on Migration and Poverty in Uganda” by Edouard Mensah and Michael O’Sullivan. 96 et al. 2015). Some of the expansion is due to a to households.44 The UNPS data used for this redefinition of administrative boundaries for urban analysis indicated that 3 to 5 percent of households areas. However, some is likely because of rural to reported sending out a work migrant during the first urban migration. two survey waves with an increase to 13 percent in later rounds (Table 6.1). This jump may be tied to 203. This chapter considers the role of rural to a change in the way the household roster module urban migration and internal migration, more was administered, because 2011 was the first year broadly, in bringing about poverty reduction in in which the UNPS employed computer-assisted Uganda. It uses panel data regression analysis to personal interviewing methods for data collection. quantify the causal impact of migration on welfare. Year dummies are included in all regressions that It uses the same panel to explore who has benefited use all years of the UNPS. from migration and what constrains migration of others. The role of international migration is not 205. Despite the mobility of its population, most of considered, given the lack of data on this. Uganda’s rural migrants tend to move within their own region or to another rural area. An 204. Uganda is a country characterized by a analysis of 2002 census data, found that—though relatively high degree of spatial mobility. In the rural and urban populations are mobile—most period of four years from 2005 to 2009, 22.9 percent migration events in Uganda occur within the same of individuals moved to other districts. Migration region and the majority of migrants into Kampala patterns are likely tied to the country’s substantial come from the adjoining Central region (Mukwaya regional and rural-urban wealth disparities, which et al. 2012). shape the sets of economic opportunities available Table 6.1: Share of households which sent a work migrant, by region, location, and year   2006 2010 2011 2012 All households 0.03 0.05 0.13 0.13 (0.20) (0.22) (0.31) (0.33) Regions Kampala 0.06 0.03 0.17 0.18 Central 0.04 0.06 0.14 0.17 Eastern 0.02 0.03 0.11 0.11 Northern 0.02 0.02 0.09 0.09 Western 0.04 0.07 0.16 0.16 Source: Authors’ calculations with UNPS. Standard deviations reported in parentheses. 206. The findings of this chapter suggest that are associated with migration to rural areas. The migration generates substantial welfare gains— findings suggest that policies to capture the welfare with even larger gains accruing to those who gains from migration to cities should focus on migrate to urban areas. Rainfall shocks serve as a investments in education for men and women in push factor for urban migration, while remoteness, rural areas as well as ICT and financial inclusion for violent conflict, and weak urban migrant networks rural households. 44. While economic considerations lead many of Uganda’s migrants to move, other factors also drive migration decisions. For example, insecurity and conflict, particularly in the North of the country during the 2000s, prompted the displacement and forced migration of large segments of the rural population (Mulumba and Olema 2009). A period of reverse migration then followed, with an influx of displaced residents returning to the North (World Bank 2012). 97 6.1 The impact of migration on poverty reduction 207. Migration can bring about welfare gains if continuing to be wealthier than non-migrants. individuals are able to move from areas where Figure 6.2 presents the same data as Figure 6.1, the return to labor is low to areas where the but disaggregating migrants to rural and urban return to labor is higher because of better areas. Migrants to rural areas were poorer than market opportunities. For example, an individual non-migrants in 2006, before moving. After may be able to earn a higher income if she moves migrating this difference between rural migrants from being engaged in agriculture in a rural village and non-migrants was almost closed. Migrants to to a job in Kampala (Harris and Todaro 1970; urban areas were better-off than non-migrants both Lewis 1954). Migration can also help bring welfare before and after migration. gains for a household by helping the household 210. Identifying the true impact of migration on minimize risk, diversify income sources, and relax welfare is challenging. Those who migrate often the constraints existing in the markets for factors of differ in unobservable ways from those who do production (capital, credit, land, and labor) through not. For example, migrants may have more drive, remittances (Azam and Gubert 2006; Rosenzweig and tolerance for risk and uncertainty than non- and Stark 1989; Stark and Bloom 1985). migrants. This makes it difficult to disentangle 208. A simple comparison of the welfare distribution what contributes to welfare differences between of those who migrate and those who do not, migrants and non-migrants: the fact they migrated suggests that migration in Uganda provides or their difference in attitude and outlook? These welfare benefits for those who migrate. The unobservable differences may have resulted in distributions of consumption for those who migrate welfare differences for migrants even if they had not and those who do not are presented in Figure migrated. Recent studies on the welfare impacts 6.1. Consumption is presented for 2006, before of migration have used experimental methods anyone migrates, and for 2010, after migration (McKenzie and Sasin 2007, Bryan, Chowdhury, and has occurred for those who migrate. The graph Mobarak 2014) or panel regression analysis with for 2006 shows that migrants and non-migrants instrumental variables (Beegle, De Weerdt, and had very similar levels of consumption before Dercon 2011; de Brauw, Mueller, and Woldehanna migrating—the two lines reflecting that the two 2013) to try and identify the impact of migration distributions lie almost on top of each other. In on welfare. The analysis in this section follows 2010, the consumption distribution of migrants Beegle et al. (2011) and uses panel regression is to the right of the consumption distribution of analysis of individual household members with non-migrants, particularly for the top two-thirds household fixed effects and instrumental variables of the distribution, indicating that migration was to instrument for the decision of an individual to beneficial. This is consistent with findings reported migrate.45 Further details on the analytical method in earlier World Bank reports: an unpublished are provided in the background paper from which analysis of the UNHS 2006 found a positive this chapter is drawn (O’Sullivan and Mensah 2016). correlation between labor mobility and per capita expenditure (World Bank 2008). 211. Analysis finds that migration results in consumption growth that is on average 14.6 209. This beneficial effect is the result of migrants percent higher per year than for those who do to rural areas ‘catching up’ with the welfare not migrate. The results of the panel regression of non-migrants and migrants to urban areas analysis with instrumental variables are presented 45. The instruments used are a WRSI reflecting rainfall shocks experienced by households, number of conflict fatalities, distance to regional capital, share of one’s ethnicity living in urban areas, and an individual’s position in the household. For more details, see O’Sullivan and Mensah (2016). 98 in the first column of Table 6.2 and show a sizeable migration, the bulk of Uganda’s migration flows still welfare impact—58.2 percent additional growth in occur within rural areas. consumption compared to non-migrants—that is 213. The gains from rural-to-rural migration may, strongly significant. at first, seem surprising. While it is expected 212. Migration has a large and positive impact, both that opportunities for employment in urban areas for those who move to rural destinations and are likely to yield higher returns it is not clear that for those who move to urban destinations. moving to another rural area would result in better Columns 2 and 3 of Table 6.2 present the impact employment opportunities. However, as the next of migration on consumption for those who move section explores in greater detail, rural-to-rural to rural areas and those who move to urban areas. migrants are often those moving from conflict- Annual consumption growth is 14 percent higher affected or remote rural areas to rural areas that for those who migrate to rural destinations and offer stability and better access to markets. It is thus 16.25 percent higher for those who migrate to urban plausible that strong welfare gains result from rural- destinations. Despite the larger gains from urban to-rural migration also. Figure 6.1: Consumption of migrants and non-migrants, before (2006) and after (2010) migration Source: UNPS 2006, 2010. Figure 6.2: Consumption of rural and urban migrants, before (2006) and after (2010) migration Source: UNPS 2006, 2010. 99 Table 6.2: Impact of migration (1) (2) (3) Rural Migrants Urban Migrants All Households and Non-migrants and Non-migrants Migrated across survey waves (1=mover, 0=stayer) 0.582*** (0.142) Migrated to rural areas (1=mover, 0=stayer) 0.560*** (0.124) Migrated to urban areas (1=mover, 0=stayer) 0.651*** (0.233) Male 0.003 (0.004) Unmarried 0.001 0.002 −0.002 (0.007) (0.003) (0.007) Unmarried male −0.000 −0.004 0.004 (0.006) (0.003) (0.006) Age category (reference: ages 10–14) Ages 15–24 −0.004 0.001 −0.006 (0.006) (0.002) (0.006) Ages 25–34 −0.006 −0.001 −0.006 (0.007) (0.003) (0.008) Ages 35–49 −0.001 0.000 −0.003 (0.006) (0.003) (0.006) Ages 50–65 0.013* 0.001 0.012* (0.007) (0.003) (0.006) Ages 66 plus 0.015* 0.005 0.012 (0.009) (0.004) (0.009) Number of effective years of schooling completed 0.001* 0.000 0.001 (0.001) (0.000) (0.001) Observations 11,338 10,783 10,824 Number of households 2,400 2,319 2,290 Source: Authors’ calculations with UNPS 2006 and 2010. Note: Initial Household Fixed Effects. Robust standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1. 214. Estimates of the impact of migration for other for movers to rural areas, and a 66 percentage point countries also find large gains to migration. consumption growth for movers to urban areas. Beegle et al. (2011) in Tanzania estimate a 36 Ignoring the direction of the move, de Brauw et al. percentage point growth in consumption over a (2013), find migrants achieve 110 percent higher period of 14 years, relative to staying, 18 percentage consumption than non-migrants, in Ethiopia, over a point to 27 percentage point consumption growth period of four years. 100 215. These estimated impacts do not take into attendance drops among schoolchildren whose account the impact of remittances on households have lost an adult due to migration. sending households. Chapter 3 documented However, school attendance is found to increase that remittances are not a large share of overall when the child migrates either solo or with his or income, but they are received by many households. her parents (Ferrone and Giannelli 2015). Strobl Recent unpublished work using the UNPS sample and Valfort (2015) combine 2002 census data suggests that remittances can be a vehicle for with weather information to examine the impact financial inclusion. The authors rely on household of weather-induced migration on employment fixed effects estimations and uncover a positive outcomes for non-migrants in Uganda. They relationship between internal remittances and uncover an adverse effect of migration on formal credit (Gross and Ntim 2014). employment outcomes for residents in receiving communities—particularly in areas with fewer 216. However, migration is not universally roads (a proxy for low capital mobility). Mwesigye beneficial, as it can have negative impacts and Matsumoto (2013) also find that communities on those who do not migrate, within the with a higher relative share of migrants are more household or within the community. An likely to experience land conflicts. These negative analysis on the links between migration and side effects need to be managed. schooling, which uses the UNPS datasets, finds that 6.2 Who migrates? 217. Households located in poorer regions are less supply of adult household labor, male or female, likely to send migrants, even though there is associated with a higher probability of sending are more gains from migration for these out a work migrant, presumably because these households. Households in the poorer regions of households are more likely to have underemployed Uganda (Eastern and Northern regions) are 3 to 5 adult labor, which reduces income that might percentage points less likely to send work migrants be lost from a member migrating. On average, when compared with households in the Central there is a difference of one member, between the and Western regions (Table 6.1). Households in ages 15 and 59, between those who migrate and the poorest regions of Uganda have the most to those who do not. Both de facto female heads of gain from migration given the average levels of household, who report being married, and de jure welfare are higher for households living outside of female heads who report being single, divorced, the north and east. On average, households that or widowed are more likely to send a migrant than send migrants live 24 km closer to Kampala than male-headed households (Table 6.3). De facto households that do not send migrants. female-headed households may be more likely to send out a migrant because the male head has 218. Migrant-sending households have a larger previously migrated making it easier for other number of adults and are more likely to be family members to migrate. headed by a woman. Having a larger relative It is expected that opportunities for employment in urban areas are likely to yield higher returns but, it is not clear that moving to another rural area would result in better employment opportunities. 101 Table 6.3: Characteristics of households that send working migrants (1) (2) (2) – (1) Coefficient in   No Migrant Sent Migrant Difference Regression+ De facto female-headed household 0.09 0.09 0.01 0.06*** De jure female-headed household 0.2 0.23 0.03* 0.03** Age of household headv 42.33 50.4 6.84*** 0.00*** Number of adult males (15–59) in household 1.03 1.49 0.38*** 0.04*** Number of adult females (15–59) in household 1.14 1.66 0.4*** 0.04*** Number of adults ages 60+ in household 0.21 0.41 0.18*** 0.03** Source: UNPS 2006–2011 with classification of migrant status from UNPS 2010–2012, respectively. Note: + Random effects regression controlling for demographics, education, and regional fixed effects. Significance levels are reported as follows: *p < 0.10, ** p < 0.05, *** p < 0.01. 219. Young adults are most likely to migrate. Figure countries, including Uganda, associate rural-urban 6.3 shows that movers are more likely to be young migration with females in their twenties who reach adults (15–24 age category) and least likely to cities for marriage purposes. However, females may be above 50. Migrants selected for the move are also leave their communities for reasons other than selected to be of an economically active age. This is marriage, such as independence from social and even more pronounced for movers to urban areas. family constraints, employment, and education (Chant 1992; Tacoli 1998). 220. Women and men are just as likely to migrate to rural and urban areas, but when women 221. The ranking of the individual in the household migrate to urban areas they are more likely to also plays a role in determining migration. be single than those who do not migrate. Table Those who migrate are less likely to be a head or 6.4 shows that those who migrate (‘movers’) and spouse, or male or female child of the head, when those who do not (‘stayers’) have similar shares of compared to stayers (Table 6.4). However, once males and females in their subsamples. Movers other individual characteristics such as gender, are more likely than stayers to be unmarried. age, and education are controlled for, being a This is driven by unmarried women being more child of the head increases the probability that an likely to migrate to urban areas. Those who are individual migrates. It increases the probability married are just as likely as those who are not to of migration to rural areas by 1 percent and the move to rural areas, and married men are just as probability of migration to urban areas by 2 likely as unmarried men to move to urban areas. percent. Those who migrate are more likely to be Brockerhoff and Eu (1993), in their demographic the oldest children. and health studies of eight Sub-Saharan African 102 Figure 6.3: Age distribution of migrants relative to stayers Source: UNPS 2006 with classification of migrant status from UNPS 2010. Note: These differences are statistically significant at 1% with the exception of the age category 25 to 34 where there is no statistically significant difference between migrants and stayers, and the rural and urban results for those ages 10 to 14, and those ages 50 to 65, which are significant at the 5% level. Table 6.4: Characteristics of individuals who migrate before migration Non-mi- Migrants Significance of Difference grants To Urban To Rural (1)– (1) All (2) (1)–(2) (1)–(4) (3) (4) (3) Male 0.47 0.48 0.47 0.49 Head or spouse 0.37 0.20 0.16 0.26 *** *** *** Child of head 0.29 0.19 0.18 0.22 *** *** *** Male child of head 0.15 0.10 0.09 0.12 *** *** *** Female child of head 0.14 0.09 0.09 0.10 *** *** *** Age rank (highest value for 3.72 4.42 5.10 3.52 *** *** * oldest) Unmarried 0.62 0.65 0.68 0.61 ** *** Unmarried male 0.29 0.31 0.31 0.30 Unmarried female 0.33 0.35 0.37 0.32 *** Years of schooling completed 4.76 5.01 5.36 4.55 *** *** * Log of real consumption per 10.88 10.95 11.15 10.68 *** *** *** adult equivalent Is poor 0.22 0.23 0.18 0.30 *** *** Number of observations 10,850 1,750 1,002 748 Source: UNPS 2006 with classification of migrant status from UNPS 2010. Note: Significance levels are reported as follows: *p < 0.10, ** p < 0.05, *** p < 0.01. 103 222. Those who migrate to urban areas are more is not strongly significant (Table 6.4). Even once educated than those who do not migrate. controlling for other factors, a 1-year increase in Migrants to rural areas are less educated than schooling leads to a 0.1 percent increase in the non-migrants are. On average, those who migrate incidence of out-migration.46 Fathers of movers have completed 0.25 more years of schooling than tend to be better educated than those of stayers, stayers. However this gap more than doubles to while mothers of movers have received less 0.6 years for those who migrate to urban areas. In education than those of stayers. However, the contrast, those who migrate to rural areas are less heads of households that send work migrants are educated than non-migrants are, although this more educated than those that do not (Figure 6.4). Figure 6.4: Education of household head, households that send migrants relative to those that do not Source: UNPS 2006–2011. 223. On average, migrants are no poorer than non- stayers at the baseline (23 percent and 22 percent, migrants before migrating. However, migrants respectively), movers to urban areas are less poor to rural areas are poorer than non-migrants than stayers (18 percent versus 22 percent), while are and migrants to urban areas are richer than movers to rural areas are poorer than stayers (30 non-migrants are. Although movers come from percent versus 22 percent) (Table 6.4). households that are just as likely to be poor as 6.3 What aids and constrains migration? 224. Given the welfare gains associated with rural to This section further examines the characteristics urban migration, it is important to understand of those who were able to migrate and uses panel what drives and constrains migration and how regression analysis to identify some of the drivers constraints to migration can be overcome. and constraints of migration in Uganda. 46. This includes household fixed effects. 104 INVESTMENTS IN HUMAN, FINANCIAL, AND SOCIAL CAPITAL 225. The finding that those who are more educated account increases the likelihood of being a migrant- are more likely to migrate and more likely to sending household by 3 and 6 percentage points, send household members to migrate suggests respectively. Facilitating households’ access to that educational investments may facilitate savings and credit products could help overcome out-migration. Other studies for countries in liquidity constraints to migration. Sub-Saharan Africa also highlight the importance 228. However, it is access to finance for the sending of education for out-migration. Brockerhoff and Eu household, not the individual migrant, that is (1993) highlight evidence that educated females associated with migration. At the individual level, most likely migrate to cities. In their analysis, Beegle those who migrated were 6 percentage points less et al. (2011) also highlight the positive influence likely to have received a loan at the baseline than (with a convex effect) of individual education on those who did not (Table 6.6). As such, the focus migration and consumption. of financial inclusion programs should be on the 226. In addition to human capital, financial capital sending household, helping them access loans on can drive household migration decisions. Panel the migrant’s behalf, rather than on the migrant. evidence from rural South Africa, for example, 229. International evidence points to the suggests that relaxing the credit constraints importance of social networks, in addition of households through transfer schemes can to human and financial capital, in facilitating boost employment through labor migration migration. Network relationships build upon (Ardington, Case, and Hosegood 2009). One of social connections of kinship, friendship, and the few randomized experiments that examines shared community origin to reduce costs and the gains and constraints to domestic migration risks associated with the movement and increase finds that a small monetary provision for migration the net expected gains from migration (Massey transportation costs has a large impact on domestic et al. 1993). Evidence of reliance on networks for migration in Bangladesh, driving substantial welfare lowering migration costs and risks largely exist in gains (Bryan et al. 2014). However, Beegle et al. international migration literature (Ilahi and Jafarey (2011) find that migration is not associated with 1999; Massey et al. 1993; McKenzie and Rapoport financial constraints. 2007). 227. Household access to finance is associated 230. Social networks appear to be important in with higher levels of migration in Uganda. At Uganda too with migrant-sending households the household level, those households that sent having stronger migrant networks, particularly migrants are 13 percentage points more likely in urban areas. Households that send migrants to have a formal loan than households that did are more likely to have a household head from an not (28 percent compared to 17 percent), and 15 ethnicity with a higher share of migrants (Table percentage points more likely to have a savings 6.5).47 Migrants, particularly urban migrants, are account with a formal institution (29 percent compared to 15 percent). These differences more likely to be from an ethnicity that has a persist even when controlling for other household larger share of its people living in cities. The share characteristics in a regression framework (Table of urban residents within one’s ethnicity is 14.1 6.5). Having a formal loan and a formal savings percent among migrants and 16.1 percent among 47. The share of migrants in a given ethnicity is calculated using 2002 census data. 105 migrants to urban areas, compared to 12.6 percent to closer rural destinations. On average, movers among stayers (Table 6.6).48 Those lacking urban live closer to Kampala and to their regional capital ethnic-based networks migrate to rural rather than at the baseline when compared with stayers (Table to urban areas. Within a regression framework, a 1 6.6). However, rural movers are located further percent decrease in one’s shared ethnicity in urban away from Kampala and their regional capital than areas is associated with a 6.7 percent increase in stayers. At a 5 percent level of significance, for the propensity to migrate to rural areas (at a 10 individuals of prime age to migrate, a one log-unit percent level of significance). increase in the distance to the regional capital is associated with a 0.6 percent increase in the 231. ICT can help individuals overcome limited incidence of rural migration (Table 6.7). The results ethnic networks to facilitate migration. Muto suggest that, for those living far from regional (2012) uses panel data from 94 rural villages across capitals, a less costly move to a rural area is the Uganda to explore the relationship between only viable option. Finding ways to remove these information and ethnic migration networks. Using constraints—such as through improved access cellular network coverage as an instrument, she to credit and ICTs—will enhance urban migration finds that households with a mobile phone are opportunities for rural households, especially for more likely to send out a migrant for employment those individuals of an economically active age. and that this effect is larger for households with smaller ethnic networks in Kampala. This result 233. There is little evidence that service availability suggests that information received through mobile influences migration decisions. Previous analysis technologies can facilitate spatial mobility. In line for Uganda found that a lack of service amenities with Muto (2012), mobile phone ownership is found in rural areas was associated with greater out- to increase the probability that a household sends migration (World Bank 2012). However, although a migrant by 3 percent, even when controlling for households that send migrants are more likely to household wealth (Table 6.5).49 live in closer proximity to an elementary school and health clinic, controlling for other factors, there 232. Living in a remote area constrains individuals is little significant difference in access to services from affording the long and costly move to between those that send migrants and those that urban areas and is associated with migration do not (Table 6.5). MIGRATION AS INSURANCE 234. Studies on migration in other settings have 235. There is some evidence that shocks do underscored the idea that shocks can be a main influence migration patterns in Uganda. driver of migration, as much as investments Rainfall shocks were found to spur an exit from in human, social, and financial capital. For agriculture in favor of urban areas. Rainfall example, Kleemans (2015) discusses that migration shocks are measured by a WRSI, which measures may evolve as an ex post risk coping strategy to the amount of rainfall against the ideal that is survive, in the face of negative income shocks, or as required for optimal maize production. Those an investment strategy to increase future expected who migrated experienced a lower WRSI in 2005 income. Beegle et al. (2011) find that young people than stayers, indicating a higher rainfall deficit experiencing rainfall shocks in Tanzania are more experienced among those who moved. The rainfall likely to migrate. deficit faced by urban movers relative to stayers 48. The share of urban residents in a given ethnicity is calculated using 2002 census data. 49. One may suspect that some of these household-level variables, such as access to formal savings and mobile phones, are merely correlates of having a higher level of welfare (which is also associated with out-migration). However, the results reported are robust to the inclusion of lagged household welfare levels, suggesting that these point estimates are not merely artifacts of higher pre-migration consumption levels. 106 is twice as large as the one for rural movers (Table migration was to rural areas. 6.6). As a result, in a regression framework, rainfall 237. Asset losses are associated with migration. shocks are strongly predictive of urban migration. There is a positive and significant 4 percentage A 25 percent reduction in the WRSI (that is, an point relationship between a household increased rainfall deficit) leads to a 2.6 percent experiencing a theft or fire and future out-migration increase in the incidence of out-migration for young once household fixed effects are controlled for. adults (Table 6.7). Migration to rural areas is no Two factors could be behind these results. It could higher for those who experience rainfall shocks. be the case that with fewer assets, households In rain-fed agricultural areas and in the absence are less committed to stay in their home village, of crop insurance, rainfall deficits lead some encouraging the household to migrate. It could individuals to escape from rural areas and settle in also be that migration is an economic coping urban areas. mechanism for households that have experienced 236. Violent conflict is also associated with asset shocks. migration, but to rural areas. Movers come from 238. In further support of the idea that migration is areas that are more prone to violent conflict than in part a household’s attempt to insure itself stayers come from, with an even larger incidence against shocks, households with stronger of conflicts for urban movers than for rural movers networks to rely on in the face of shocks are (Table 6.6). However, in a regression framework, less likely to send migrants. A household’s conflict is only significantly correlated with rural reliance on networks for insuring against shocks migration, not urban migration. A doubling of the is associated with a 3 percentage point lower number of conflict-related fatalities is associated likelihood of sending out a migrant in the next with a 0.8 percent increase in the incidence of survey round, once other household characteristics out-migration of young individuals of prime age have been controlled for (Table 6.5). This finding to migrate. In 2005, the Northern region in Uganda suggests that households with less robust local faced conflict with four times the number of support networks could instead rely on migration to fatalities (22) than were recorded in the Central deal with risk. region (6), which was the region with the next highest fatality rate. Young individuals were 3.6 239. To some degree, migration in Uganda has aided percent more likely to migrate from Northern poverty reduction by allowing households to Uganda than from Central Uganda (Table 6.7). This manage shocks. However, it is not clear from this 107 analysis whether migration of this type should be risk. Reducing households’ exposure to risk or encouraged. Although this type of migration aided increasing their access to other ways to manage risk poverty reduction when it occurred, migration may be a more sustainable approach to increase may or may not be the optimal strategy to manage resilience. Table 6.5: Correlates of a household’s decision to send a migrant (1) (2) (2) – (1) Coefficient   No Sent Difference in Regres- Migrant Migrant sion+ Household has a formal loan of any type 0.17 0.28 0.11*** 0.03*** Household member has a savings account with a 0.14 0.29 0.11*** 0.06*** formal institution Share of migrants within head’s ethnicity 0.17 0.19 0.01*** 0.07 Reliance on networks for insuring shocks 0.32 0.20 −0.08*** −0.03*** Household owns mobile phone 0.41 0.64 0.23*** 0.03*** Elementary school within one hour of household 0.81 0.87 0.09*** −0.02* Health center/clinic within one hour of 0.69 0.82 0.13*** 0.01 household Source: Staff calculations using UNPS 2006–2011. Table 6.6: Correlates of migration at the individual level Migrants Significance of Difference Non- All To Urban To Rural migrants (1)–(2) (1)–(3) (1)–(4) (2) (3) (4) (1) Individual received loan from 0.117 0.147 0.076 *** *** *** any source Log kilometers from Kampala 4.820 4.698 4.220 5.209 *** *** *** Log kilometers from regional 4.195 3.935 3.443 4.461 *** *** *** capital Log WRSI maize 4.388 4.372 4.365 4.380 *** *** * Log number of fatalities 0.803 1.735 1.973 1.417 *** *** *** Share of one’s ethnicity living 0.126 0.141 0.161 0.114 *** *** *** in urban areas Source: Staff calculations using UNPS 2006–2011. Table 6.7: Shocks, distance, and the probability of migration of 15–24-year-olds Effect on Probability of Migration of 15–24-Year-Olds, Percent All To Urban To Rural Decrease in WRSI from 100 to 75 percent 1.6 2.6* 1.0 Increase in number of fatalities from 6 to 24 3.6** 0.8 3.6*** Log kilometers from regional capital 0.008** 0.006** 0.002 Source: Staff calculations using UNPS 2006–2011. 108 6.4 Conclusion 240. This chapter has highlighted the strong welfare potential in urban areas. However, it is not clear impact of migration—both to rural and urban that migration is the optimal response to a shock. areas—but particularly to urban areas. The Policies are also needed to reduce exposure to welfare impact of migration strongly supports risk and increase a household’s access to markets urbanization and pro-rural-urban migration policies and public programs that help it manage risk. The for their linkage to poverty reduction in Uganda and restoration of peace in northern Uganda was a similar developing countries. major step in reducing exposure to risk. 241. The evidence is consistent with low levels 243. Improving education outcomes for women may of education, lack of access to finance, long also require programs that encourage delaying distances to urban centers, and limited young women’s age at marriage. To ensure migrant networks in urban areas constraining females take full advantage of urban migration migration for some households. Improving opportunities for their own welfare and to facilitate education, access to finance, and access to ICT remittance transfers to their parent households, would help these households migrate. programs that delay young women’s age at marriage—such as adolescent empowerment 242. Migration is often undertaken to help mitigate interventions (Bandiera et al. 2014)—should be the impact of negative shocks. Policies that considered. The results highlight the importance of allow free movement can transform the lives of investments in the education of rural populations, rural individuals prone to shocks by offering them which would increase human capital and enhance migration opportunities to boost their earning the migration potential for future generations. After work Traffic, down town -Kampala 109 EDUCATION AND HEALTH CHAPTER: SERVICES: QUALITY OF 7 INPUTS, USER SATISFACTION, AND COMMUNITY WELFARE LEVELS 50 Poorer communities tend to have services of lower quality, but are more satisfied with the services that they are receiving. 244. A better-educated and healthy population is more likely to transition from subsistence agriculture to more productive jobs. Chapters 2 and 5 highlighted the importance of human capital for poverty reduction. In Uganda, education is a key predictor if earnings as well as household consumption (see, for example, Fox and Pimhidzai 2011; and Tsimpo and Wodon 2014a). Apart from its impact on livelihoods, the case for investments in education can also be made on the basis of its impact on health outcomes, among others. 245. Since 1997, the GoU has implemented a series of policies as well as made substantial budget investments to improve education and health services as well as the demand for those services. On the supply side, key policies include building and renovating schools and health centers; purchasing adequate instructional materials; training, hiring, and retaining teachers and health workers; improving the drugs policy under the national medical store (NMS); reducing teacher and health worker absenteeism; and serving areas that are hard to reach and hard to stay in. On the demand side, important policy reforms have been adopted as well, including for UPE, USE, school feeding programs, mama kits, and national immunization days, among others. 50. This chapter draws on the background paper: “Education and Health Services in Uganda: Quality of Inputs, User Satisfaction, and Community Welfare Levels,” by Clarence Tsimpo, Alvin Etang, and Quentin Wodon. 110 246. This has led to improvements in access education and health, the level of welfare of to education and health, but quality has communities, and the satisfaction of users deteriorated. For example, while access to with facilities. The basic idea is to combine data education has improved, quality remains an issue from two different surveys to provide a profile of and most students do not learn nearly enough. the quality of services available in communities Arguments have actually been suggested that in relationship to their level of welfare, while also access has increased at the cost of quality, a assessing rates of user satisfaction with the services problematic outcome because quality is essential provided. This chapter draws heavily from two for economic growth (Hanushek and Woessman datasets: The SDI survey of 2013 and the UNHS of 2012). 2012/13. The SDI is used to compute the indicators on the supply and quality of services. The UNHS 247. This chapter aims to assess the relationships is used to rank facilities by welfare and to derive between the quality of services in users’ satisfaction.51 7.1 Quality of inputs at the school level 248. In general, more and better inputs seem to be The two poorest regions (Northern and Eastern) available in better-off locations, as expected. are the regions with the highest rates of teacher Consider, for example, the pupil per classroom absenteeism. Absenteeism may be driven in part and pupil to teacher ratios. These ratios are much by the fact that some locations in these regions higher for the poorest quintile of communities are hard to reach (due to poor roads) and hard than the richest (Figure 7.1 and Annex 2, Table to live in (specific areas in Uganda are classified A2.1). A typical classroom in the poorest quintile administratively as ‘hard to reach/hard to stay’). has 116 pupils, while the corresponding figure for Teacher absenteeism leads to inefficiency in the richest quintile is 58 pupils. A teacher in the public spending because teachers are paid with poorest quintile has to attend to 58 pupils, while little benefits for students. While some level of a corresponding teacher in the richest quintile absenteeism may be warranted, prevailing rates are attends to 31 pupils, on average.52 Overcrowding clearly much too high, with likely consequences in of pupils in classrooms in poorer areas is likely to terms of student learning (Finlayson 2009). Notably, have negative consequences on learning outcomes. absenteeism is higher among head teachers. The Northern region, which also happens to be the Close to two out of five head teachers were not poorest region in Uganda, has the worst ratios. present. This certainly contributes to weakening the accountability mechanism at the school level. 249. Teacher absenteeism rates53 at the level of schools or classrooms are also negatively 250. Absenteeism rate is lower for female teachers. correlated with welfare. Teachers are more likely Female teachers are more likely to be present at to be absent in poorer areas. For communities in school and in the classroom. School absenteeism the poorest quintile, about four out of ten teachers rate for female teachers is 20 percent, which is 6 are absent from school. The corresponding figure percentage points lower than male. This difference for the richest quintile is two out of ten teachers. is statistically significant. Similarly, classroom 51. The UNHS provides information on household welfare. Each district of the country is split into two parts: urban and rural. The average household welfare is then attributed to the facilities in the SDI survey. Subsequently, this allows ranking of the facilities by quintiles of welfare. 52. Results from a Wald test confirm that the differences between the poorest and the richest quintiles are statistically significant 53. The SDI survey uses a standardized, internationally benchmarked methodology to measure absenteeism through unannounced visits. SDI teams conduct two visits to each facility. The first is announced in advance so as to increase the likelihood of being able to collect data on key indicators. The second visit, which happens during a seven-day period following the first visit, is unannounced and its purpose is to ascertain the whereabouts of staff. Staff who are not in the facility because it is not their shift are not considered absent. Health workers who are not in the facility because they are carrying out outreach activities are likewise not considered absent. 111 absenteeism for females is 44 percent, which schools have a functioning PTA, even though, in is 14 percentage points lower than their male principle, PTAs have been abolished in the country. counterparts. Thus, teaching is not the only reason Schools in poor areas are less likely to have a PTA. why male teachers show up in school. It would be Indeed, while 46.6 percent of schools in the poorest interesting to understand what they could be doing quintile have a functioning PTA, the corresponding in school when they are not in classrooms. figure for the richest quintile is 55.2 percent (Annex 2, Table A2.1). By contrast, seven out of ten schools 251. The learning environment in classrooms is have a functioning School Management Committee again much better in richer areas. Schools (SMC), and there is no apparent relationship in richer areas are more likely to have a library, between welfare levels and the availability of electricity, and work displayed on the walls, among an SMC in a school, probably because SMCs are others (Figure 7.1). At the national level, serious mandatory. challenges remain when it comes to classroom environment, especially in line with the country’s 253. Inspectors tend to often visit schools that are ambition to become a middle-income country in located in better-off areas. The likelihood of a the near future. Indeed, the availability of a library, school receiving the visit of an inspector during electricity, or displayed material is still very low. For the school year is close to one for most schools. instance, only 8.8 percent of schools have a library. This is true across regions, regardless of welfare The corresponding figure for electricity is only levels. The only exception is the Western region 10.8 percent. It is worth noting that connectivity to where up to 11 percent of schools did not receive electricity, while perhaps not the most essential the visit of inspectors. The number of inspections element for student learning, is important to carried out at schools is, however, correlated with operationalize the ‘skilling Uganda’ agenda toward welfare. Inspectors tend to often visit schools that the use of ICT and appropriate vocational training. are located in better-off areas more. Here again, issues related to the fact that poor areas are more 252. Institutional aspects of the management of the likely to be hard to reach/hard to stay areas may be schools show a mixed message across welfare at play in that visiting these areas is more costly for distribution. At the national level, three out of five inspectors (Office of the Prime Minister 2012). Classroom Blocks - Aduku ss 112 Figure 7.1: Inputs for primary schools by welfare and subregion Pupil-classroom and Pupil-teacher Ratio Teachers Absenteeism Classroom environment54 Source: Staff calculations using the 2013 SDI and the UNHS 2012/13 surveys. 7.2 Quality of inputs at the health center level 254. Sick people in poor areas are more likely to with five health providers consulting six outpatients face overcrowding and long queues while each, on a daily basis. visiting their health centers. The poorer the 255. Unlike the education sector, there is no area, the higher the patient caseload55 (Figure 7.2 apparent correlation between health workers’ and Annex 2, Table A2.2). Looking at the median, a absenteeism and welfare.56 At the national level, health provider in the poorest quintile consulted six outpatients per day, against only three outpatients it is estimated that excluding off duty, absenteeism for the richest quintile. Health workers in the rate is high at 42 percent. The incidence of health Northern region were the busiest and received workers’ absenteeism is quite similar across six outpatients on a daily basis. The Eastern and welfare quintiles. Results from a t-test show no Western regions also had high patient caseloads statistically significant difference by quintile. There are important disparities across regions. Health 54. An index representing the quality of the classroom environment is estimated using factorial analysis. This index represents a weighted average of the various classroom characteristics (see Annex 2, Table A2.1 for the detailed list) with the weights for each variable directly derived from the data to maximize the explanatory power of the index.55. See the previous footnote for education on how absenteeism measures are estimated. 55. Patient caseload is defined as the average number of outpatient visits a health worker attends to per working day. 56. See the previous footnote for education on how absenteeism measures are estimated. 113 workers are more likely to be absent in the Central facilities in the poorest quintile have a disciplinary region: half of the health workers were absent committee, compared to 17 percent for the top when excluding off duty. If the Central region is quintile. excluded from the analysis, then absenteeism of 258. In Uganda, most of the public facilities are push health workers negatively correlates with welfare facilities. Most public facilities (90 percent) are and differences between the poor and the rich are push facilities, which means that they receive drugs statistically significant. Thus, remoteness (hard to centrally. By contrast, most private facilities are pull reach/hard to stay) is a driver of health workers’ facilities, which means that they order their drugs. absenteeism. On the other hand, the Central region For public facilities, drugs are centrally managed being the one with the higher rate of absenteeism by the NMS. The NMS purchases drugs in bulk and is something to explore further. Probably, available handles the logistics of distribution across the and appealing opportunities to diversify and country. It also retrieves expired drugs for proper increase income sources are playing a role here. disposal. This dichotomy between public and 256. Contrary to the education sector, absenteeism private providers is driving the story behind drugs. of workers in the health sector is gender 259. The push system used by public facilities seems neutral. Female health workers have the same to be effective in that availability of essential probability to be absent as their male counterparts. drugs is higher in public facilities. The six tracer This finding holds, even if one excludes the Central drugs were indeed available in 46 percent of public region. More analysis is needed to understand facilities, but only in 15 percent of private facilities. the underlying factors of absenteeism, but at The issue of lack of availability in private facilities least finding different patterns in the health and may be related in part to ‘for profit’ behavior, in education sectors shows that particular actions that little gain is to be obtained from these basic might be needed for specific sectors to curb medicines. The presence of private pharmacies in absenteeism in the country. areas where private health facilities operate and 257. Disciplinary and quality assurance committees the comparative advantage of pharmacies in the are more likely to be present in poor areas. drug business may be another factor explaining the Institutional aspects of the management of health low availability of these drugs in private facilities. facilities show a mixed message across welfare Perhaps surprisingly, the Northern region is the distribution. On average, about half of the health region with the highest availability of tracer drugs. facilities reported the presence of functioning Efficiency of the NMS coupled with interventions of Health Facility Management Committees. Very few nongovernmental organizations may be a reason health facilities have a procurement committee for this. Among the six tracer drugs, the measles or an audit committee (only 5.9 and 6.2 percent, vaccine had the highest stock-out rate. respectively). As a consequence, issues related to 260. The poorest localities are also those with very proper financial and resources management can be limited availability of basic infrastructure problematic. Disciplinary Committees are available and equipment in health facilities.57 The only in one out of five health facilities. The share of health facilities with a quality assurance committee availability of basic infrastructure and equipment is also low (12.6 percent). Disciplinary and quality is positively correlated with community welfare. assurance committees are more likely to be present For example, health facilities in richer areas are in poor areas. For instance, 37 percent of health more likely to have electricity and piped water, 57. The SDI survey collected information on the availability of electricity, piped water, toilets, ambulance, microscope, weighing scale, blood pressure machine, thermometer, malaria test kit, HIV test kit, etc. (see Annex A7.3 for a detailed list). 114 as expected. Only one in ten health facilities has centers. This probably explains the fact that a low a functioning ambulance, again mostly in richer proportion of women delivered in formal health areas. Surprisingly, the availability of telephone facilities under the attendance of specialized (landline and mobile phone) remains low in health workers, despite high rate of attendance most facilities. All health facilities in the richest for antenatal care and the mama kit program. The quintile have an adult weighing scale, while the Northern and the Eastern regions, which happened corresponding figure for the poorest quintile is to be the poorest, tend to have very limited 58 percent. Maternity waiting centers (antenatal availability of infrastructure and equipment in their rooms) are available in only 23.9 percent of health health facilities. Figure 7.2: Inputs for health facilities by welfare and subregion Health workers absenteeism Push or a pull facility and drugs availability Caseload (median) Infrastructure availability index58 Source: Staff calculations using the 2013 SDI and the UNHS 2012/13 surveys. 58. The share of teachers with minimum content knowledge was observed based on results of a customized teacher test administered to Primary 4 mathematics/numeracy and English teachers. The English test results were for teachers teaching English, and the mathematics test results were for teachers teaching mathematics. The tests were based on items from the curricula being taught in Uganda (World Bank 2013). 115 7.3 Knowledge and behavior of teachers 261. There is a clear, positive relationship between 262. Female teachers perform better in English, teacher knowledge and community welfare. while male teachers perform better in Teachers’ knowledge of the subjects they teach mathematics. Female teachers scored 56 percent is low, as are pedagogical skills to transform in English while their male counterparts score 53 their knowledge into quality teaching.59 On percent on average. Although this difference seems average, teachers scored 59 percent and 64 percent small, it is statistically significant. With regard to in the English and numeracy/mathematics tests, mathematics, male teachers performed better respectively (Figure 7.3). Teacher knowledge than females, scoring 60 percent compared to 53 increases with community welfare. For instance, percent for female teachers. This difference is also teachers in the poorest quintile of communities statistically significant. scored 56 percent and 59 percent in the English and 263. There are no significant differences in teachers’ numeracy/mathematics tests. The corresponding pedagogical knowledge across community figures for the richer quintile of communities are 62 welfare quintiles. Results from a Wald test suggest and 68 percent. The difference between the poorest similar pedagogical knowledge across the board. and the richest quintile is statistically significant for Estimation results suggest that overall, pedagogy English and mathematics. In line with the positive skills are disappointingly low, as reflected in the correlation between teacher knowledge and average score of 25 percent on the pedagogy test community welfare, the Northern region is also the and only 7 percent of teachers scored above 50 region where teacher scores are the lowest for both percent. the English and numeracy/mathematics tests. Figure 7.3: Primary school teachers’ assessment by welfare quintiles Source: Staff calculations using the 2013 SDI and the UNHS 2012/13 surveys. 59. The share of teachers with minimum content knowledge was observed based on results of a customized teacher test administered to Primary 4 mathematics/numeracy and English teachers. The English test results were for teachers teaching English, and the mathematics test results were for teachers teaching mathematics. The tests were based on items from the curricula being taught in Uganda (World Bank 2013). 116 7.4 Knowledge and behavior of health workers 264. The accuracy of diagnostics is lower in is positively correlated to community welfare. For poor areas, especially for acute diarrhea, example, for the richest quintile, 84.6 percent of pneumonia, diabetes mellitus, and pulmonary providers were able to properly manage post- tuberculosis (PTB).60 Only one in four health partum hemorrhage (Figure 7.5). The corresponding workers was able to diagnose all five tracer figure for the poorest quintile is 67.6 percent. conditions. The diagnostic assessment shows Regionally, the worst performance is registered that health workers perform very poorly on acute in the Eastern and Western regions where only diarrhea. Less than half (47 percent) were able to 48.6 and 52 percent, respectively, of providers are properly diagnose acute diarrhea. Performance on able to properly manage neonatal asphyxia. The pneumonia and diabetes mellitus is also very low, knowledge gap between these two regions and with only 60 percent able to accurately diagnose other regions regarding neonatal asphyxia is very each of these diseases. For all the diseases, health big. In other regions, at least 74 percent of providers workers’ knowledge increases with welfare (Figure were able to properly deal with neonatal asphyxia. 7.4). For those in the poorest quintile, only 16 266. Male health workers exhibit better knowledge percent were able to accurately diagnose the five of the common diseases as well as better tracer conditions. The corresponding figure for management of neonatal asphyxia and post- the richest quintile is 39.6 percent. The biggest partum hemorrhage. About 35 percent of male knowledge gap across welfare quintiles is revealed health workers were able to diagnose all the five through diagnosis of pneumonia. Among health cases. Meanwhile, only 13 percent of female health workers in the richest quintile, 85.3 percent were workers were able to do so. One out of four female able to properly diagnose pneumonia, against workers (24 percent) was not able to properly only 44.5 of those in the poorest quintile. The manage any of the neonatal asphyxia and post- knowledge gap across quintiles is also big (double partum hemorrhage conditions. Meanwhile the digit) for acute diarrhea, diabetes mellitus, and corresponding figure for males is only 8 percent. PTB. Diagnostic accuracy was significantly higher in Kampala and lower in the Northern region. For example, in Kampala, 41 percent of the providers were able to accurately diagnose all the five tracer conditions. In the Northern region, only 11 percent of the providers were able to do so. 265. There is no clear correlation between community welfare and management of neonatal asphyxia, but proper management of post-partum hemorrhage increases with community welfare. Only half (54.3 percent) of the providers were able to properly manage maternal and newborn complications (post- partum hemorrhage and neonatal asphyxia). Proper management of post-partum hemorrhage 60. Health worker knowledge and quality of care were assessed using two indicators of process (adherence to clinical guidelines in five tracer conditions and management of maternal and newborn complications—as measured in the vignette interviews) and one indicator of outcomes (diagnostic accuracy in the five tracer conditions at the end of the vignette interviews). Three of the tracer conditions were childhood conditions (malaria with anemia, acute diarrhea with severe dehydration, and pneumonia), and two were adult conditions (PTB and diabetes mellitus). Two other conditions were included: post-partum hemorrhage, the most common cause of maternal death during birth; and neonatal asphyxia, the most common cause of neonatal death during birth (World Bank 2013). 117 Figure 7.4: Share of health workers giving the correct diagnostic (5 tracer conditions) Source: Staff calculations using the 2013 SDI and the UNHS 2012/13 surveys. Figure 7.5: Share of health workers giving the correct diagnostic for post-partum hemorrhage and neonatal asphyxia Source: Staff calculations using the 2013 SDI and the UNHS 2012/13 surveys. 7.5 Outcomes at the school level 267. Learning outcomes are strongly and positively average score on the non-verbal reasoning part correlated with community welfare. The pupil of the test was 57 percent. There is substantial assessment consisted of three parts: English, variation in learning outcomes across community numeracy, and non-verbal reasoning.61 Overall, welfare (Figure 7.6). For example, pupils in the pupils answered 47 percent of questions on the richest quintile scored 66 percent overall while test correctly. The average score for English was those in the poorest quintile scored only 34 percent. 46 percent and for numeracy was 43 percent. The The largest gaps are observed for English, where 61. Learning outcomes were measured for grade 4 pupils. Outcome for health facilities are more complex to measure, hence the SDI survey did not attempt to collect such information. This section therefore focuses on student outcomes only. The objective of the pupil assessment was to assess basic reading, writing, and arithmetic skills. The test was designed by experts in international pedagogy and based on a review of primary curriculum materials from thirteen African countries, including Uganda (see Johnson, Cunningham, and Dowling 2012). The pupil assessment also measured non- verbal reasoning skills on the basis of Raven’s matrices, a standard IQ measure that is designed to be valid across different cultures. 118 pupils in the richest quintile answered 69 percent extent, management. Teacher absenteeism reduces of questions correctly versus only 31 percent of student performance. Better teacher behavior pupils in the poorest quintile. The knowledge gap leads to better student performance, as does a across the welfare distribution is also important better score of the teacher in English and numeracy for numeracy. Students in wealthier communities tests. The econometric results also suggest that performed better, which could be related to the boys perform better than girls do, particularly in fact that as discussed earlier, schools in wealthier mathematics, and non-verbal reasoning. communities had better inputs related to the 269. These results are consistent with expectations classroom environment, teacher absenteeism, and and have the following implication: improving pupil-teacher and pupil-classroom ratios, among the quality of inputs could bring substantial others. gain in learning outcomes. The results suggest 268. The low quality of inputs is affecting the that improvements in the quality of teaching performance in poor communities. The and the knowledge base of teachers could bring determinants of pupils’ performance is assessed substantial gains in student performance, especially using econometric modeling (Annex 2, Table in poor areas. A reduction in pupil-teacher ratio A2.7). A wide range of factors can affect the ability as well as better school infrastructure would also of children to learn in school. Previous work for bring gains, although these are likely to be smaller, Uganda suggests that children from disadvantaged and may be more costly to achieve in terms of backgrounds are less likely to fare well. However, budgetary resources. Although one should be school-level factors also play a role (Mulindwa and careful not to infer causality, it could also be that Marshall 2013). Using the SDI and UNHS surveys it strengthening the inspection regime would bring appears that performance is affected by a variety of gains as well, while by contrast PTAs and SMCs factors, including pupil-teacher ratio, inspections, seem to have less of a beneficial impact, possibly school/classroom environment, and, to some because how well they function matters more. Figure 7.6: Pupil assessment (score) Source: Staff calculations using the 2013 SDI and the UNHS 2012/13 surveys. 119 7.6 User satisfaction with facilities 270. Poorer communities are more likely to be the opportunity costs of staying organized for a satisfied with the services that they are sustained period could be really high for the poor. receiving, even though it is clear from the 273. The contrast between the objective measures analysis based on the SDI survey that the of inputs from the SDI survey and the measures level of inputs and their quality is higher in of satisfaction from the UNHS raises questions better-off communities. The perceived quality for the effectiveness of community-based of service is negatively correlated with community monitoring and the demand for accountability. welfare (Figure 7.7). The likely explanation is If the population in poor communities has low that poor communities are so deprived that expectations or is not exposed enough to good their expectations are low. This leads them to be quality services to be able to assess quality, it is more easily satisfied with the services they get. not clear that it can effectively lobby for quality By contrast, better-off communities have higher services. For social accountability mechanisms to expectations and therefore are more demanding be effective, additional measures may be needed about quality and less satisfied, even if objectively to enable disadvantaged communities to properly they are getting comparatively better services. monitor the services they receive. The issue is not 271. Low expectations in poor communities can specific to Uganda, and there are examples of social be a problem for social accountability. Social accountability initiatives with mixed results (Fox accountability is an approach toward building 2015). Issues of political economy may also have to accountability that relies on civic engagement, in be considered for social accountability measures which citizens participate directly or indirectly in to work (Joshia and Houtzagerb 2012). Overall, demanding accountability from service providers in a context where poverty and expectations are and public officials. Social accountability generally a problem, more needs to be done for social combines information on rights and service delivery accountability to be effective. These findings with collective action for change. In Uganda, social are in line with existing literature. For example, accountability has emerged as an important tool in Svensson et al. (2015) conducted an experiment the fight for better governance and service delivery. on community-based monitoring of absenteeism Examples include U-report, Barazas, and the versus head teachers monitoring. They found that Uganda Participatory Poverty Assessment Process. local monitoring improves teacher attendance but only when the head teacher is responsible for 272. Besides low expectations, there are several monitoring and there are financial incentives for other hypotheses for this observation. First, teachers at stake. Moreover, they also found that it could just be lack of information to the poor of parents generate significantly less reliable reports what their options or choices are. For example, the than head teachers do. The results in this chapter supply of private facilities may not be available for further echo the importance of information as the poor. Second, poor people just cannot hold highlighted by (Reinikka and Svensson 2005). providers accountable because either they cannot They conducted an experiment that shows that observe provider quality or they do not have making information on budget allocation available the power. Third, there also exists the possibility to the beneficiaries, reduces corruption and elite that the poor could be threatened if they engage capture, and ends up having a positive impact on in organizing themselves. Fourth and finally, enrollment and educational outcomes. 120 Figure 7.7: Inputs and user satisfaction by welfare quintiles in education sector Absenteeism, pupil per classroom per teacher Teacher and pupil knowledge Satisfaction and health workers’ absenteeism Satisfaction, drugs, and knowledge Satisfaction and child mortality Satisfaction and maternal health Source: Staff calculations using the 2013 SDI survey, UNHS 2012/13, and UDHS 2011. 121 7.7 Conclusion 274. Poorer communities tend to have services of 275. The contrast between the objective measure lower quality, but are more satisfied with the of quality and the perceived quality raise services that they are receiving. Low quality has implications for social accountability of inputs in poor communities negatively affects mechanisms. If populations in poor areas have outcomes such as student learning. The poor are low expectations, their ability to monitor quality more likely to be satisfied with the service that they is weakened. Apart from the demonstration of are getting, although objective measures from the the need to improve inputs for education and SDI survey suggest that it should be the opposite. health facilities in Uganda, one of the implications This implies that the poor are so deprived that their of the analysis is that for social accountability expectations are low, and they tend to be happy mechanisms to be effective, additional measures with the little service that they can get. Conversely, must probably be taken to enable populations in the non-poor tend to have higher expectations and poor communities to ask for and obtain higher therefore will be more demanding about quality quality services. and will be less satisfied, even if objectively they are getting the best service in the country. 122 REFERENCES Ali, Daniel Ayalew, Frederick H. Bowen, Klaus W. Deininger, and Marguerite Felicienne Duponchel. 2015. “Investigating the Gender Gap in Agricultural Productivity: Evidence from Uganda.” Policy Research Working Paper, No. WPS 7262, Washington, DC. Appleton Simon. 1996. “The gender wage gap in three African countries,” Economics Series Working Papers WPS/1996- 07, University of Oxford, Department of Economics. Appleton Simon. 2003. “Regional or National Poverty Lines? The Case of Uganda in the 1990s,” Journal of African Economies, Centre for the Study of African Economies (CSAE), vol. 12(4), pages 598-624. Appleton Simon, Emwanu Tom, Kagugube Johnson and Muwonge James. 1999, “Changes in Poverty in Uganda, 1992- 97”, Centre for the Study of African Economies, University of Oxford, WPS/99.22. Ardington, Cally, Anne Case, and Victoria Hosegood. 2009. “Labor Supply Responses to Large Social Transfers: Longitudinal Evidence from South Africa.” American Economic Journal: Applied Economics 1 (1): 22–48. Azam, J.-P., and F. Gubert. 2006. “Migrants’ Remittances and the Household in Africa: A Review of Evidence.” Journal of African Economies 15 (Supplement 2): 426–62. Bandiera, Oriana and Goldstein, Markus and Rasul, Imran and Burgess, Robin and Gulesci, Selim and Sulaiman, Munshi. 2010. Intentions to participate in adolescent training programs: evidence from Uganda Journal of the European Economic Association, 8 (2-3). 548-560. ISSN 1542-4774. Bandiera, Oriana, Robin Burgess, Markus Goldstein, Niklas Buehren, Selim Gulesci, Imran Rasul, and Munshi Sulaiman. 2015. “Women’s Empowerment in Action: Evidence from a Randomized Control Trial in Africa.” Economic Organisation and Public Policy Discussion Papers, EOPP 50, The London School of Economics and Political Science. Beegle, Kathleen, Joachim De Weerdt, and Stefan Dercon. 2011. “Migration and Economic Mobility in Tanzania: Evidence from a Tracking Survey.” Review of Economics and Statistics 93 (3): 1010–33. Berge L., Oppedal I., Bjorvatn K., Tungodden B. 2014. “Human and Financial Capital for Microenterprise Development: Evidence from a Field and Lab Experiment”, Management science, Vol. 61, issue 4. Binswanger-Mkhize, H., and S. Savastano. 2014. “Agricultural Intensification: The Status in Six African Countries.” World Bank Policy Research Working Paper 7116. Björkman-Nyqvist Martina, Svensson Jakob, Yanagizawa-Drott David. 2012. “Can Good Products Drive Out Bad? Evidence from Local Markets for (Fake?) Antimalarial Medicine in Uganda”, Center for International Development at Harvard University, Working Paper No. 242. Blattman Christopher, Green Eric P., Jamison Julian, Lehmann M. Christian, and Annan Jeannie. 2016. “The Returns to Microenterprise Support among the Ultrapoor: A Field Experiment in Postwar Uganda”, American Economic Journal: Applied Economics 2016, 8(2): 35–64. Blattman, C., N. Fiala, and S. Martinez. 2014. Generating skilled employment in developing countries: Experimental evidence from Uganda. Quarterly Journal of Economics. Bold Tessa, Kaizzi Kayuki C., Svensson Jakob, Yanagizawa-Drott David. 2015. “Low Quality, Low Returns, Low Adoption: Evidence from the Market for Fertilizer and Hybrid Seed in Uganda”, Harvard Kennedy School, Faculty Research 123 Working Paper Series. RWP15-033. Bold, T, K Kaizzi, J Svensson and D Yanagizawa-Drott. 2015. “Low Quality, Low Returns, Low Adoption: Evidence from the Market for Fertilizer and Hybrid Seed in Uganda”, Harvard Kennedy School WP. Brockerhoff, Martin, and Hongsook Eu. 1993. “Demographic and Socioeconomic Determinants of Female Rural to Urban Migration in Sub-Saharan Africa.” International Migration Review 27 (3): 557. Bryan, Gharad, Shyamal Chowdhury, and Ahmed Mushfiq Mobarak. 2014. “Under-Investment in a Profitable Technology: The Case of Seasonal Migration in Bangladesh.” NBER Working Paper No. 20172, National Bureau of Economic Research. Campos, Francisco; Goldstein, Markus; McGorman, Laura; Munoz Boudet, Ana Maria; Pimhidzai, Obert. 2015. Breaking the Metal Ceiling: Female Entrepreneurs Who Succeed in Male-Dominated Sectors. Policy Research Working Paper No. 7503. World Bank, Washington, DC. © World Bank. Canning, David, Sangeeta Raja, and Abdo S. Yazbeck, Eds. 2015. Africa’s Demographic Transition: Dividend or Disaster?, Africa Development Forum Series, a Co-Publication of the Agence Française de Développement and the World Bank, Washington, D.C. Carletto, G., S. Gourlay, and A. Winters. 2015. “From Guesstimates to GPStimates: Land Area Measurement and Implications for Agricultural Analysis.” Journal of African Economies 24 (5): 593–628. Chant, Sylvia H. 1992. Gender and Migration in Developing Countries. Belhaven Press. Christiaensen, Luc, Punam Chuhan-Pole, and Aly Sanoh. 2013. “Africa’s Growth, Poverty and Inequality Nexus - Fostering Shared Prosperity.” Forthcoming. Collier, Paul and Reinikka, Ritva. 2003. “Uganda’s Recovery: The Role of Farms, Firms, and Government”. World Bank Publications, The World Bank, number 13850, February. de Brauw, Alan, Valerie Mueller, and Tassew Woldehanna. 2013. “Does Internal Migration Improve Overall Well-Being in Ethiopia?” Ethiopia Strategy Support Program II No. 55, International Food Policy Research Institute. De Mel, S., McKenzie, D., Woodruff, C. 2008. “Returns to Capital: Results from a Randomized Experiment.” Quarterly Journal of Economics 123(4): 1329–72. Deininger, K. and J. Okidi. 2003. “Growth and Poverty Reduction in Uganda: 1999-2000: Panel Data Evidence”, Development Policy Review 21(4): 481-509. Dorosh, Paul, and James Thurlow. 2012. “Can Cities or Towns Drive African Development? Economy-Wide Analysis for Ethiopia and Uganda.” Working Paper Series UNU-WIDER Research Paper, World Institute for Development Economic Research (UNU-WIDER). Etang and Tsimpo. 2016. “Beyond Income Poverty: Non-Monetary Dimensions of Poverty in Uganda”, The World Bank, Background Paper for the Uganda Poverty Assessment. Fafchamps, M., and R. V. Hill. 2008. “Price Transmission and Trader Entry in Domestic Commodity Markets.” Economic Development and Cultural Change 56: 729–66. ———. 2015. “Redistribution and Group Participation: Comparative Experimental Evidence from Africa and the UK.” National Bureau of Economic Research Working Papers 21127. Ferrone, Lucia, and Gianna Claudia Giannelli. 2015. “Household Migration and Child Educational Attainment: The Case of Uganda.” IZA Discussion Paper Series No. 8927, IZA. Fiala Nathan. 2015. “Rethinking the microfinance model: Returns to subsidized microcredit for male and female 124 entrepreneurs in Uganda”. Working paper, poverty-action.org. Finlayson, Mary. 2009. “The Impact of Teacher Absenteeism on Student Performance: The Case of the Cobb County School District.” Dissertations, Theses and Capstone Projects, Paper 4, Kennesaw State University. Firpo, S., Fortin, N. and Lemieux, T. (2009). “Unconditional Quantile Regressions.” Econometrica 77(3): 953-973. Fox, Jonathan A. 2015. “Social Accountability: What Does the Evidence Really Say?” World Development 72: 346–61. Fox, Louise, and Obert Pimhidzai. 2011. “Is Informality Welfare-Enhancing Structural Transformation?” Policy Research Working Paper 5866. Gable, Susanna, Hans Lofgren, and Israel Osorio-Rodarte. 2014. Country Development Diagnostics Post-2015: Uganda. Washington, DC: World Bank. ———. 2015. Trajectories for Sustainable Development Goals: Framework and Country Applications. Washington, DC: World Bank. Golumbeanu, Raluca; Barnes, Douglas. 2013. Connection charges and electricity access in Sub-Saharan Africa. Policy Research Working Paper; no. WPS 6511. Washington, DC: World Bank. Gross, Elena, and Rebecca Ntim. 2014. “Evaluating the Impact of Internal Remittances on Financial Inclusion: Evidence from Ugandan Household Data.” Mimeo. University of Bayreuth. http://www.ael.ethz.ch/downloads/2014/ Papers/Gross_and_Ntim.pdf. Gutierrez, Eva; Choi, Tony. 2014. “Mobile Money Services Development: The Cases of the Republic of Korea and Uganda”. Policy Research Working Paper; No. 6786. World Bank, Washington, DC. © World Bank. Hanushek, Eric, and Ludger Woessman. 2012. “Do Better Schools Lead to More Growth? Cognitive Skills, Economic Outcomes and Causation.” Journal of Economic Growth. Harris, John R., and Michael P. Todaro. 1970. “Migration, Unemployment and Development: A Two-Sector Analysis.” The American Economic Review 60 (1): 126–42. Hill Ruth and Mejia-Mantilla Carolina. 2016. “Welfare, Income Growth and Shocks in Uganda”, The World Bank, Background Paper for the Uganda Poverty Assessment. Hill Ruth Vargas. 2010. “Liberalisation and Producer Price Risk: Examining Subjective Expectations in the Ugandan Coffee Market,” Journal of African Economies, Centre for the Study of African Economies (CSAE), vol. 19(4), pages 433-458, August. Ilahi, Nadeem, and Saqib Jafarey. 1999. “Guest-Worker Migration, Remittances and the Extended Family: Evidence from Pakistan.” Journal of Development Economics 58 (2): 485–512. Jakiela, P., and O. Ozier. 2015. “Does Africa Need a Rotten Kin Theorem? Experimental Evidence from Village Economies.” The Review of Economic Studies 1–38. Jedwab Remi, Christiaensen Luc, and Gindelsky Marina. 2015. “Demography, Urbanization and Development: Rural Push, Urban Pull and... Urban Push?”, World Bank Policy Research Working Paper No. 7333. Johnson, D., A. J. C. Cunningham, R. Dowling. 2012. Professional Teaching Competencies across 17 African Countries – A Meta Analysis of Published Governmental Expectations. New York: World Bank. Joshia, A, and P. P. Houtzagerb. 2012. “Widgets or Watchdogs?” Public Management Review 14 (2). Kaminski, Jonathan, and Luc Christiaensen. 2014. “Household Welfare in Uganda 2005–2009.” Mimeo, World Bank. Kassie, Menale, Bekele Shiferaw, and Geoffrey Muricho. 2011. “Agricultural Technology, Crop Income, and Poverty Alleviation in Uganda.” World Development 39 (10): 1784–95. 125 Kilic T., A. Zezza, C. Carletto, and S. Savastano. 2014. “Missing(ness) in Action: Selectivity Bias in GPS-Based Land Area Measurements.” Policy Research Working Paper 6490, World Bank, Washington, DC. Kleemans, Marieke. 2015. Migration Choice under Risk and Liquidity Constraints. Washington, DC: World Bank. Lekfuangfu Nuarpear Warn, Machin Stephen, and Rasul Imran. 2012. “IGC Rapid Response Project:Report on Uganda’s Returns to Schooling”, International Growth Centre, Working paper No F-43007-UGA-1. Lewis, W. Arthur. 1954. “Economic Development with Unlimited Supplies of Labour.” The Manchester School 22 (2): 139–91. Loayza, Norman and Claudio Raddatz. 2010. “The Composition of Growth Matters for Poverty Alleviation.” Journal of Development Economics 93 (1): 137–151. Massey, Douglas S., Joaquin Arango, Graeme Hugo, Ali Kouaouci, Adela Pellegrino, and J. Edward Taylor. 1993. “Theories of International Migration: A Review and Appraisal.” Population and Development Review 19 (3): 431–66. McGillivray, M., and M. Clarke, eds. 2006. Understanding Human Well-Being. New York: United Nations University Press. McKenzie, David, and Hillel Rapoport. 2007. “Network Effects and the Dynamics of Migration and Inequality: Theory and Evidence from Mexico.” Journal of Development Economics 84 (1): 1–24. McKenzie, David J., and Marcin Jan Sasin. 2007. “Migration, Remittances, Poverty, and Human Capital: Conceptual and Empirical Challenges.” World Bank Policy Research Working Paper No. 4272. Mensah Edouard and O’Sullivan Michael. 2016. “Moving Out and Up: Panel Data Evidence on Migration and Poverty in Uganda”, The World Bank, Background Paper for the Uganda Poverty Assessment. MoFPED. 2014. “Poverty Status Report: Structural change and poverty reduction in Uganda”, Economic Development Policy and Research Department. Mukwaya, Paul, Yazidhi Bamutaze, Samuel Mugarura, and Todd Benson. 2012. “Rural-Urban Transformation in Uganda.” IFPRI-University of Ghana Understanding Economic Transformation in Sub-Saharan Africa Conference, Accra, Ghana. International Food Policy Research Institute. Mulindwa, I. N., and Jeffery H. Marshall. 2013. Improving Learning in Uganda, Volume 2: Problematic Curriculum Areas and Teacher Effectiveness: Insights from National Assessments. World Bank Studies. Washington, DC: World Bank. Mulumba, Deborah, and Wendo Mlahagwa Olema. 2009. Policy Analysis Report: Mapping Migration in Uganda. Project Report for IMMIS – African Migration and Gender in Global Context: Implementing Migration Studies. Kampala: Makerere University. Muto, M. 2012. “The Impacts of Mobile Phones and Personal Networks on Rural-to-Urban Migration: Evidence from Uganda.” Journal of African Economies 21 (5): 787–807. Mwesigye, Francis, and Tomoya Matsumoto. 2013. “Rural-Rural Migration and Land Conflicts: Implications on Agricultural Productivity in Uganda.” GRIPS Discussion Paper No. 13-17, National Graduate Institute for Policy Studies, Tokyo. Nagler Paula and Naude Wim, 2014, “Non-Farm Entrepreneurship in Rural Africa: Patterns and Determinants”, IZA DP No. 8008. Nikoloski, Zlatko, Luc Christiaensen, and Ruth Hill. 2015. “Household Shocks and Coping Mechanism: Evidence from Sub-Saharan Africa.” Mimeo, World Bank. 126 Nyqvist, Martina Björkman. 2012. “Income Shocks and Gender Gaps in Education: Evidence from Uganda.” Journal of Development Economics 105 (2013): 237–253. Office of the Prime Minister. 2012. “Evaluation of the Government of Uganda’s Response to Absenteeism in the Public Service.” Mimeo, Kampala: Government of Uganda. Ravallion M. and Huppi M. 1991. “Measuring changes in Poverty: a methodological case study of Indonesia during an adjustment period”. World Bank Economic Review Vol 5, pp57-82. Ravallion, Martin and Benu Bidani. 1994. “How robust is a poverty line?” World Bank Economic Review 8(1): 75-102. Reinikka, R., and J. Svensson. 2005. “Fighting Corruption to Improve Schooling: Evidence from a Newspaper Campaign in Uganda.” Journal of the European Economic Association 3 (2/3): 259–67. Rosenzweig, Mark R., and Oded Stark. 1989. “Consumption Smoothing, Migration, and Marriage: Evidence from Rural India.” The Journal of Political Economy 905–26. Saunders, P. 2005. The Poverty Wars: Reconnecting Research with Reality. Sydney: UNSW Press. Seff, I., D. Jolliffe, K. Kafle, and J. Kastelic. 2014. “Dynamics of Wellbeing in Tanzania 2008–2013: Analysis Based on Three Waves of the National Panel Survey.” Background paper prepared for Tanzania Poverty Assessment Report (May 19). Sheehan, M., and C. Barrett. 2014. “Understanding the Agricultural Input Landscape in Sub-Saharan Africa: Recent Plot, Household, and Community-Level Evidence.” World Bank Policy Research Working Paper 7014. Ssewanyana, Sarah, and Kasirye, Ibrahim. 2012. “Poverty and Inequality Dynamics in Uganda: Insights from the Uganda National Panel Surveys. 2005/6 and 2009/10.” EPRC Research Series 94. Ssewanyana, Sarah, and Kasirye, Ibrahim. 2013. “The Dynamics of Income Poverty in Uganda: Insights from the Uganda National Panel Surveys of 2009/10 and 2010/11”, EPRC, Occasional Paper No. 35. Stark, Oded, and David E. Bloom. 1985. “The New Economics of Labor Migration.” The American Economic Review 173–78. Strobl, E., and M.-A. Valfort. 2015. “The Effect of Weather-Induced Internal Migration on Local Labor Markets. Evidence from Uganda.” The World Bank Economic Review 29 (2): 385–412. Svensson J, and Yanagizawa D. 2009. “Getting Prices Right: The Impact of the Market Information Service in Uganda”, Journal of the European Economic Association 7 (2‐3), 435-445. Svensson, J., A. Zeitlin, T. Bold, and K. Bjorvatn. 2015. Improving Services Delivery: An Assessment of Absenteeism Data. Ideas for Growth, International Growth Center. Tacoli, Cecilia. 1998. “Rural-Urban Interactions: A Guide to the Literature.” Environment and Urbanization 10 (1): 147–66. Tsimpo, C., and Q. Wodon. 2014a. “Workforce Development and Returns to Education in Uganda: Key Findings.” Mimeo, Washington, DC: World Bank. ———, eds. 2014b. Residential Electricity in Uganda. Washington, DC: World Bank. UBOS (Uganda Bureau of Statistics). 2006. Uganda National Household Survey 2005/2006 Report. Kampala: Uganda Bureau of Statistics (UBOS). ———. 2010. Uganda National Household Survey 2009/2010 Report. Kampala: Uganda Bureau of Statistics (UBOS). ———. 2011. Uganda National Panel Survey 2009/10: Wave I Report on the Key Findings. Kampala: Uganda Bureau of Statistics (UBOS). 127 ———. 2013a. Uganda National Panel Survey 2010/11: Wave II Report. Kampala: Uganda Bureau of Statistics (UBOS). ———. 2013b. Uganda National Panel Survey 2011/12: Wave III Report. Kampala: Uganda Bureau of Statistics (UBOS). ———. 2014a. Uganda National Household Survey 2012/2013 Report. Kampala: Uganda Bureau of Statistics (UBOS). ———. 2014b. National Population and Housing Census 2014: Provisional Results. Kampala: Uganda Bureau of Statistics (UBOS). Uganda Bureau of Statistics (UBOS) and ICF International Inc. 2001. Uganda Demographic and Health Survey 2000/2001. Kampala, Uganda: UBOS and Calverton; Maryland: ICF International Inc. ———. 2007. Uganda Demographic and Health Survey 2006. Kampala, Uganda: UBOS and Calverton; Maryland: ICF International Inc. ———. 2012. Uganda Demographic and Health Survey 2011. Kampala, Uganda: UBOS and Calverton; Maryland: ICF International Inc. UNDP (United Nations Development Programme). 2010. Human Development Report 2010. United Nations Development Programme. New York, NY: Palgrave Macmillan. UNICEF (United Nations International Children’s Emergency Fund). 2007. Progress for children: A world fit for children statistical review, UNICEF, New York. Van Rooyen, C., Stewart, R. & De Wet, T. 2012. “The Impact of Microfinance in Sub-Saharan Africa: A Systematic Review of the Evidence”. World Development, 40, 2249-2262. Wodon, Q., M. C. Nguyen, and C. Tsimpo. 2016. “Child Marriage, Early Childbirth and Education.” Uganda Education Technical Assistance Series, Policy Note No. 7. WHO and UNICEF. 2006. Core Questions on Drinking-Water and Sanitation for Household Surveys. Geneva: WHO. World Bank. 2007. Country Economic Memorandum for Uganda. Washington, DC: World Bank. ———. 2008. The Welfare Impact of Rural Electrification: A Reassessment of the Costs and Benefits, Independent Evaluation Group. Washington, DC: World Bank. ———. 2011. “Weather Index Insurance for Agriculture: Guidance for Development Practitioners.” Agriculture and Rural Development Discussion Paper No. 50, World Bank, Washington, DC. ———. 2012. “Uganda: Promoting Inclusive Growth: Transforming Farms, Human Capital, and Economic Geography.” Report No. 67377–UG, World Bank, Kampala. ———. 2013. Education and Health Services in Uganda: Data for Results and Accountability. Survey Delivery Indicators (SDI) report. ———. 2015. Uganda Systematic Country Diagnostic Report: Boosting Inclusive Growth and Accelerating Poverty Reduction. Washington, DC: World Bank. ———. 2016a. Poverty in a Rising Africa. Washington, DC: World Bank. ———. 2016b. Country Economic Memorandum for Uganda. Washington, DC: World Bank. Y. Cho, M. Honorati. 2013. “Entrepreneurship programs in developing countries: A meta-regression analysis”, Mimeo Social Protection and Labor, The World Bank, Washington (2013). 128 ANNEXES 129 ANNEX 1: EXPLORING PATTERNS OF FOOD AND NON-FOOD CONSUMPTION OVER TIME, METHODOLOGY 1. Appleton et al. (1999) examined spending on basic non-food items to estimate the share of food in the national poverty line following an approach adopted from Ravallion and Bidani (1994). Appleton et al. investigated the expenditure patterns of people who are at the food poverty line by regressing the food share of household i (si) on region/urban-rural dummy, demographic characteristics, and the ratio of adult equivalent consumption expenditure (Yi) to food poverty line zf (and its square): si = b0 + b1 ln(Yi / z f ) + b2 ln(Yi / z f ) 2 + ∑ j = 2 f j D j + b3wi + ei 8 (1) ...where ei is the error term, Dj is dummy for the four regions urban/rural (central urban is excluded and serves as a reference group), wi is the demographic characteristics of household i including household size, head’s gender, zf and proportion of boys/girls of different age groups in the household, and is the food poverty line, which is U Sh 21,258. 2. The estimation result for equation 1 is presented in Table A1.1. Column 2 of Table A1.1 shows the estimated coefficient by Appleton et al. (1999), using 1997/98 data. Column 3 of this table presents estimates of equation 1 using UNHS 2012/13 data. 3. Table 1.4 presents the share of spending on basic non-food items in total consumption expenditure using the national average demographic characteristics of these households ( wm ). For households residing in region j, the predicted non-food share is given by 1 − (b + f + b w ) .62 0 j 3 m 62. In equation 1, Central Urban is omitted. The interaction between the demographic characteristics of the ‘food poor’ households and the 9 = 1 − ( b0 + b3wm ) corresponding coefficients in Table A1.1, that is, b3wm , is 0.071. Therefore, non-food share for Central Urban is 0.5 . In other regions, the share of non-food expenditure is estimated by 1 − ( b0 + f j + b3wm ) . 130 Table A1.1: Regression of food share   1997/98 2012 Coef. t-stat Coef. t-stat Log consumption per capita divided by food poverty line 0.060 (11.9) 0.01 (1.64) Square of log consumption per capita divided by food poverty line −0.053 (−19.84) −0.04*** (−14.97) Central rural −0.119 (−15.26) 0.09*** (11.67) East rural −0.052 (−6.46) 0.18*** (20.88) East urban 0.044 (5.480) 0.09*** (7.94) North rural −0.031 (−3.65) 0.19*** (22.49) North urban 0.029 (3.52) 0.10*** (9.68) West rural −0.020 (−2.50) 0.21*** (27.88) West urban 0.066 (8.44) 0.12*** (9.71) Household size 0.008 (1.54) 0.00 (0.00) Male-headed household −0.006 (−1.05) −0.01* (−1.94) The following variables are as proportion of household size: Boys aged <6 years 0.071 (3.99) 0.12*** (6.54) Boys aged 6–12 years 0.052 (2.62) 0.11*** (5.47) Boys aged 13–17 years 0.041 (1.92) 0.06*** (2.83) Men aged 60+ 0.082 (5.33) 0.14*** (6.83) Girls aged <6 years 0.089 (4.81) 0.11*** (6.18) Girls aged 6–12 years 0.047 (2.34) 0.11*** (5.59) Girls aged 13–17 years 0.022 (1.0) 0.02 (0.73) Girls aged 18–59 years 0.056 (4.41) 0.08*** (5.22) Women aged 60+ 0.075 (4.32) 0.17*** (8.90) Constant 0.55 (60.55) 0.34*** (22.68) Observations 4,962 6,888 R-squared 0.255 0.43 Note: t-statistics in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1 131 132 ANNEX 2: ADDITIONAL TABLES ON SERVICE DELIVERY AND WELFARE Table A2.1: Primary schools—resources and management   Provider Location Region Welfare National Public Priv. Urb. Rur. Central Eastern Kampala Northern Western Q1 Q2 Q3 Q4 Q5 Basic resources Pupils per classroom 86.4 35.0 72.4 71.4 47.0 90.5 44.2 136.0 53.4 116.5 88.3 51.0 53.1 57.9 71.7 Pupil to teacher ratio 47.6 22.2 35.5 42.4 28.8 49.6 21.8 61.0 34.4 58.0 51.5 33.4 32.4 31.0 41.0 Pupil to textbook ratio 24.2 92.4 36.7 26.4 14.9 34.2 43.2 91.3 22.7 48.6 63.6 54.0 14.5 13.4 27.8 Teacher absenteeism (%) School absence rate 27.4 13.2 14.2 26.3 21.8 26.0 10.2 34.7 18.5 36.9 25.1 17.2 22.6 18.0 23.8 Classroom absence rate 57.0 39.8 46.7 54.1 46.7 59.7 42.9 69.5 42.1 71.1 52.8 42.5 47.5 49.0 52.6 Reasons for absence Not his/her shift 1.8 5.7 0.0 2.7 0.9 0.0 0.0 4.8 4.8 3.8 2.3 4.0 0.0 1.2 2.4 Sick 17.0 21.0 21.7 16.9 18.6 18.3 15.2 14.9 18.3 11.8 17.2 28.7 14.6 22.8 17.5 Maternity leave 3.9 5.6 4.1 4.1 8.0 3.0 2.1 3.0 2.5 3.9 2.0 3.6 4.1 7.3 4.1 In training 2.3 1.5 0.3 2.4 0.8 2.8 2.9 1.9 3.4 0.7 5.3 5.0 1.1 0.5 2.2 Field trip 29.7 30.3 45.0 27.7 31.3 21.4 34.3 30.2 39.1 27.6 27.8 37.4 25.4 35.0 29.8 Funeral 2.9 0.0 0.0 2.8 4.8 0.6 0.0 3.6 1.2 2.2 1.8 1.8 6.4 0.0 2.5 Other approved ab- 1.1 0.0 0.0 1.1 3.6 0.0 0.0 0.0 0.2 0.0 0.0 0.3 0.8 4.1 0.9 sence Gone to retrieve salary 24.6 25.0 21.2 25.1 23.2 46.3 43.4 11.5 9.6 26.8 32.1 4.6 29.9 21.2 24.7 On strike 10.1 7.6 2.4 10.8 4.7 7.1 2.1 21.6 6.4 16.4 10.9 2.3 7.2 5.2 9.8 Other 6.7 3.2 5.2 6.3 4.2 0.4 0.0 8.5 14.5 6.8 0.5 12.3 10.5 2.6 6.2 Total 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 Classroom environment Library corner/books for 9.4 6.9 31.2 5.4 11.1 11.2 36.6 0.0 7.4 1.1 1.0 11.6 4.2 28.2 8.8 pupils Blackboard/whiteboard 99.7 100 100 99.7 100 100 100 100 99.3 100 100 98.9 100 100 99.8 in class Chalk/marker to write 97.8 99.0 98.8 98.0 100 99.7 100 98.8 94.6 99.7 97.2 94.7 99.2 100 98.1 Working electricity 6.0 24.8 32.3 7.6 17.7 11.9 55.6 3.1 4.6 5.4 2.2 6.0 15.7 26.9 10.8 connection Children’s work on walls 8.5 19.6 15.3 10.7 24.3 3.4 33.5 0.0 10.6 2.6 1.4 17.1 13.6 24.0 11.3 Other materials on walls 38.5 44.2 86.8 33.2 56.5 34.4 93.5 33.1 29.3 30.1 16.6 33.5 53.6 70.3 39.9 Hygiene in the class- 77.9 70.8 91.1 73.9 70.2 61.8 100 85.9 86.3 74.9 69.4 84.6 71.2 80.8 76.1 room Blackboard good for 86.3 90.4 89.8 87.0 87.6 72.3 100 97.3 93.4 87.5 84.1 92.2 85.1 87.9 87.4 reading Enough light for reading 95.8 95.6 98.0 95.4 92.7 93.1 90.3 98.1 99.8 94.7 95.7 95.1 95.1 98.3 95.7 Enough light for reading 90.7 90.2 88.1 90.9 85.5 82.0 79.0 97.3 99.3 90.5 91.3 94.4 90.1 86.1 90.6 at back Classroom environment index Mean 0.336 0.388 0.526 0.324 0.405 0.311 0.625 0.300 0.337 0.289 0.259 0.357 0.367 0.493 0.349 Median 0.268 0.268 0.409 0.268 0.337 0.268 0.599 0.268 0.268 0.268 0.268 0.268 0.332 0.441 0.268 Inspection, SMC, PTA Share with functioning 69.1 38.9 69.0 60.2 58.2 60.4 46.2 46.3 73.8 46.6 54.6 72.6 78.4 55.2 61.3 PTA Share with functioning 76.2 50.8 78.8 68.3 54.8 66.3 66.4 77.9 81.5 69.9 69.9 76.3 65.4 65.9 69.6 SMC At least one inspection 98.1 82.5 100 93.2 92.1 99.2 100 98.5 89.0 99.1 85.4 99.3 89.5 97.0 94.1 (%) Number of inspections, 5.3 3.6 7.7 4.4 4.0 6.0 5.5 3.9 5.1 4.1 4.6 6.4 3.0 6.2 4.8 mean Number of inspections, 4 4 6 3 3 4 3 3 5 3 3 5 3 4 4 median Source: Staff calculations using the 2013 SDI and the UNHS 2013 surveys. 133 134 Table A2.2: Health facilities—caseload, workers, management, and drug availability   Provider Location Region Welfare National Public Priv. Urb. Rur. Central Eastern Kampala Northern Western Q1 Q2 Q3 Q4 Q5 Caseload Mean 10 2 2 8 7 8 1 6 10 8 5 11 3 3 6 Median 6 2 1 5 4 5 1 6 5 6 4 5 2 3 3 Absenteeism rate (%) Absent including off duty 51.2 52.2 50.7 52.4 59.3 52.2 50.4 54.2 48.0 54.8 55.1 46.7 51.2 64.2 51.6 Absent excluding off duty 44.0 38.6 36.9 45.8 51.0 43.0 36.3 48.3 42.3 49.1 50.4 37.4 37.5 48.3 42.0 Reason for being absent Sick/maternity 8.2 4.1 4.2 8.4 5.9 8.7 4.7 11.0 5.6 11.3 7.7 3.5 5.4 0.0 6.5 In training/seminar 11.7 5.7 5.4 12.3 23.6 4.4 5.0 8.5 13.0 8.4 7.6 17.1 6.7 28.9 9.3 Official mission 7.4 4.6 4.9 7.3 6.9 14.8 1.9 8.0 4.9 11.9 12.0 3.6 2.5 13.0 6.2 Approved absence 13.8 18.7 17.7 14.3 21.4 13.4 17.8 11.7 13.4 11.8 17.9 15.6 17.1 5.4 15.8 Not his/her shift 26.1 44.8 45.5 24.4 30.6 30.6 46.2 21.1 23.4 19.7 20.6 32.3 45.4 48.0 33.8 Doing fieldwork 0.3 0.3 0.0 0.5 0.0 1.1 0.0 0.0 0.4 0.8 0.2 0.4 0.0 0.0 0.3 Not approved absence 20.7 6.8 8.8 20.0 6.5 21.9 9.5 18.8 21.8 21.8 17.0 19.4 9.6 3.5 15.0 Gone to retrieve salary 0.7 0.0 0.0 0.7 0.0 0.7 0.0 0.7 1.1 0.7 0.4 1.2 0.0 0.0 0.4 Outreach 2.1 2.9 0.9 3.6 1.0 1.9 0.0 6.2 6.0 4.0 4.9 2.6 0.7 0.0 2.4 Other 8.9 12.2 12.6 8.4 4.0 2.7 14.7 14.1 10.5 9.6 11.6 4.4 12.6 1.2 10.2 Total 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 Management Health facility management committee 73.9 29.2 19.3 72.3 84.8 67.7 11.7 67.0 70.9 75.6 74.9 67.7 20.8 85.8 52.3 Finance committee 13.9 23.3 12.3 22.2 37.7 15.6 9.2 34.4 14.0 21.4 18.7 24.1 13.6 39.5 18.4 Procurement committee 2.1 9.9 4.7 6.6 20.0 3.0 5.4 6.6 0.6 6.4 4.1 9.6 4.7 0.0 5.9 Audit committee 2.6 9.9 5.4 6.6 14.1 5.5 5.5 9.0 1.5 6.7 1.0 12.4 5.0 12.6 6.2 Disciplinary Committee 22.8 25.2 13.4 30.3 27.8 31.9 11.8 52.3 16.1 37.8 25.0 26.0 14.7 17.1 24.0 Quality Assurance Committee 13.6 11.6 11.9 13.0 4.9 18.6 9.7 22.1 10.6 17.0 13.4 13.6 9.3 12.6 12.6 Push or pull facility Push 90.1 6.9 12.2 74.7 44.0 73.9 5.7 89.5 77.0 86.6 74.5 65.1 14.2 43.5 54.0 Pull 3.2 83.7 82.7 16.0 37.4 19.7 90.9 3.8 14.4 8.8 11.2 24.8 80.4 56.5 38.1 Both 6.7 9.4 5.1 9.3 18.6 6.4 3.5 6.7 8.6 4.6 14.4 10.1 5.4 0.0 7.9 Total 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 Drugs availability (6) 46 15 19 38 23 24 17 69 40 41 32 44 17 39 31 Availability by drug ACT 90.9 89.1 91.4 89.1 89.3 81.1 94.7 89.1 91.6 82.6 89.8 97.6 90.7 74.7 90.0 Cotrimoxazole 80.2 89.1 87.2 82.8 74.2 72.5 87.6 89.9 92.7 80.5 89.9 84.0 84.5 82.9 84.5 Measles vaccine 76.1 33.5 29.6 71.3 73.9 61.7 24.3 87.5 66.5 70.9 65.5 79.6 30.3 65.8 55.6 ORS sachets 78.8 81.5 81.9 79.0 74.3 67.4 81.7 90.7 85.4 77.0 82.4 77.4 82.0 87.4 80.1 Depo-Provera 98.5 66.0 81.6 83.6 61.1 85.2 82.2 94.2 87.5 90.3 78.0 85.1 79.8 73.1 82.8 Fansidar 84.4 70.8 70.5 82.3 88.7 80.4 68.2 94.4 73.9 86.8 81.0 78.0 71.2 87.4 77.9 Source: Staff calculations using the 2013 SDI and the UNHS 2013 surveys. Note: ACT = Artemisinin combination therapy. 135 136 Table A2.3: Health facilities—infrastructure and equipment   Provider Location Region Welfare National Public Priv. Urb. Rur. Central Eastern Kampala Northern Western Q1 Q2 Q3 Q4 Q5 Infrastructure/equipment Electricity 55.6 88.6 90.0 60.4 80.6 47.7 92.7 58.4 63.5 49.7 64.5 68.5 88.4 85.8 71.5 Piped water 12.9 50.3 60.2 13.3 19.8 8.3 65.1 4.1 23.0 5.6 22.1 16.6 56.3 57.4 31.0 Flush toilet for outpatients 4.1 34.5 48.9 0.5 1.1 0.0 55.8 1.3 1.4 0.0 0.0 1.6 46.4 22.4 18.7 Flush toilet for staffs 2.6 16.8 20.4 2.8 3.5 2.8 21.9 1.3 5.2 1.0 4.9 3.3 19.2 22.4 9.4 Functioning land phone 4.8 15.8 18.6 5.0 14.3 4.4 19.1 1.6 4.7 3.1 3.0 11.8 16.5 17.1 10.1 Functioning cellular 6.8 25.2 26.8 8.9 8.5 23.3 26.3 0.5 6.7 5.2 7.1 16.2 25.0 25.2 15.6 Functioning shortwave 2.0 0.9 0.0 2.4 0.0 3.1 0.0 4.0 1.6 3.4 1.9 1.9 0.0 0.0 1.5 radio Functioning computer 5.5 22.9 26.5 6.3 6.9 4.0 28.2 9.3 8.4 5.2 10.4 5.7 24.2 39.5 13.9 Access to email or internet 2.6 16.7 20.7 2.6 6.0 3.2 21.8 0.7 3.7 1.1 5.0 3.5 18.7 35.0 9.4 Functional ambulance 2.8 17.3 18.7 4.4 9.9 2.4 18.9 0.0 8.7 0.6 5.7 8.2 17.5 26.8 9.8 Maternity antenatal room 24.5 23.3 14.0 29.9 32.0 38.8 11.5 30.9 20.6 24.1 28.3 40.2 13.6 29.7 23.9 Adult weighing scale 70.4 80.6 78.5 73.4 86.4 64.5 77.0 68.2 79.8 58.3 79.1 88.5 76.3 100 75.3 Thermometer 74.0 92.5 94.6 75.8 94.2 60.0 96.0 79.4 78.8 67.2 84.4 77.3 93.6 100 82.9 Child weighing scale 78.1 48.3 36.4 80.3 73.4 85.1 28.8 76.1 83.4 77.2 90.5 84.0 33.4 95.5 63.8 Stethoscope 79.9 93.2 93.1 82.2 82.5 73.2 95.0 84.1 88.3 76.6 93.1 80.8 91.3 100 86.3 Infant weighing scale 48.6 33.5 28.3 49.3 79.1 48.2 22.2 32.5 46.9 38.1 51.3 54.6 31.8 59.4 41.3 Microscope 32.2 63.9 64.6 37.1 53.8 32.9 65.4 29.1 41.0 24.5 45.2 44.0 63.2 56.1 47.5 Glucometer 18.3 50.5 56.9 19.9 39.8 11.6 57.0 27.0 19.9 15.6 22.6 25.7 53.0 60.6 33.8 Malaria Test kit 80.8 62.0 54.6 82.1 76.1 66.5 53.6 87.4 90.5 75.9 84.7 83.0 58.1 57.7 71.7 Urine Dip kit 28.7 47.3 50.5 29.9 46.4 24.4 50.3 34.1 27.8 24.9 40.9 28.4 47.9 48.0 37.6 HIV Test kit 50.4 67.3 73.4 49.6 71.2 44.6 74.8 47.3 46.2 40.7 57.7 49.1 73.5 82.9 58.6 Tuberculosis test kit 20.7 11.3 17.2 15.6 15.2 22.8 14.1 20.2 11.9 16.5 16.4 17.4 15.2 30.8 16.2 Autoclave 15.7 22.9 30.7 12.3 11.4 11.1 30.0 20.0 14.5 14.3 20.1 6.6 27.8 25.3 19.2 Electric boiler/steamer 1.5 10.0 12.7 1.3 2.8 4.1 13.3 0.0 0.5 1.7 0.5 1.9 11.9 0.0 5.6 Electric dry heat sterilizer 1.6 12.6 16.7 1.0 4.2 0.0 17.8 0.7 2.3 0.4 0.5 2.2 15.8 14.2 6.9 Non-electric pot 51.6 59.1 57.1 54.1 58.8 52.2 57.0 44.1 59.6 43.0 64.0 60.6 55.4 50.8 55.2 Incinerator 9.9 12.6 7.9 13.2 14.0 2.4 6.7 26.8 14.4 14.8 12.4 11.1 8.6 9.7 11.2 Facilities index Mean 0.187 0.347 0.381 0.193 0.262 0.171 0.389 0.187 0.207 0.156 0.228 0.217 0.363 0.417 0.264 Median 0.133 0.275 0.287 0.159 0.213 0.123 0.284 0.157 0.163 0.118 0.215 0.170 0.281 0.329 0.208 Source: Staff calculations using the 2013 SDI and the UNHS 2013 surveys. 137 138 Table A2.4: Assessment of teacher knowledge and teaching quality   Provider Location Region Welfare National Public Priv. Urb. Rur. Central Eastern Kampala Northern Western Q1 Q2 Q3 Q4 Q5 Teacher knowledge English 59 58 63 58 62 60 64 56 56 56 57 59 60 62 59 Numeracy 64 66 70 63 70 66 69 58 61 59 62 64 67 68 64 Pedagogy 26 24 28 25 28 24 29 26 23 26 23 23 26 28 25 Overall 45 45 48 45 49 45 49 43 43 44 43 45 47 48 45 Teaching quality Mean 0.671 0.665 0.746 0.658 0.642 0.634 0.754 0.705 0.699 0.646 0.690 0.746 0.597 0.667 0.670 Median 0.704 0.734 0.757 0.700 0.686 0.687 0.785 0.702 0.784 0.667 0.746 0.785 0.592 0.722 0.707 Source: Staff calculations using the 2013 SDI and the UNHS 2013 surveys. Table A2.5: Assessment of health worker knowledge   Provider Location Region Welfare National Public Priv. Urb. Rur. Central Eastern Kampala Northern Western Q1 Q2 Q3 Q4 Q5 Diagnostics Acute diarrhea 45.3 50.6 58.2 39.2 45.3 41.0 61.0 28.8 41.5 36.1 40.7 38.3 59.2 50.6 47.5 Pneumonia 56.3 65.9 75.4 48.4 54.3 58.3 74.8 42.5 50.1 44.5 57.2 47.7 73.7 85.3 60.2 Diabetes Mellitus 57.8 66.6 76.6 49.7 51.2 51.9 77.2 57.7 50.4 49.1 55.2 50.0 74.5 62.3 61.4 PTB 86.9 89.5 91.9 84.9 91.2 91.7 90.7 77.4 84.0 81.6 87.5 87.4 91.0 94.8 87.9 Malaria/Anemia 95.3 97.9 96.6 96.1 99.9 98.7 96.0 95.6 93.2 96.4 94.5 97.1 96.6 97.8 96.3 Share correc7 All 5 Cases 25.9 28.7 39.4 17.5 17.1 21.7 41.3 11.3 21.8 16.4 14.4 20.0 39.7 39.6 27.1 Exactly 4 Cases 27.2 34.9 38.1 24.3 26.8 35.5 36.4 27.6 18.8 24.2 37.6 20.7 34.9 26.0 30.3 Exactly 3 Cases 21.4 21.5 12.0 28.8 42.2 14.4 11.6 26.6 29.6 23.7 26.6 30.1 14.3 24.3 21.5 Exactly 2 Cases 16.1 9.0 5.8 18.9 8.9 20.3 5.4 21.2 18.4 23.3 11.9 20.1 6.1 7.8 13.2 Only 1 Case 6.8 4.7 1.7 9.3 5.0 7.2 1.7 12.5 9.2 11.4 8.8 7.2 2.1 0.0 6.0 No Case 2.6 1.1 3.0 1.2 0.1 0.9 3.5 0.6 2.1 1.0 0.6 1.9 3.0 2.2 2.0 Total 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 Clinical knowledge PPH 68.6 76.0 71.7 71.5 69.7 74.7 71.2 74.2 69.3 67.6 72.3 71.5 72.9 84.6 71.6 Respiratory 62.8 73.6 75.5 60.7 74.0 48.6 80.5 74.5 52.3 65.3 50.6 57.0 79.4 58.3 67.2 Share correct All 2 Conditions 48.1 63.3 58.0 51.4 58.2 40.0 61.6 64.0 46.3 53.6 41.9 47.8 62.4 55.3 54.3 Only 1 Condition 35.3 23.0 31.2 29.5 27.4 43.4 28.4 20.8 29.0 25.8 39.0 32.9 27.5 32.3 30.3 None of the condi- 16.7 13.7 10.8 19.1 14.5 16.7 9.9 15.3 24.7 20.6 19.1 19.3 10.1 12.4 15.5 tions Total 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 Source: Staff calculations using the 2013 SDI and the UNHS 2013 surveys. Note: PTB = Pulmonary Tuberculosis/Pneumonia/Chronic Bronchitis. PPH = Post-partum Hemorrhage. Respiratory = Respiratory distress syndrome/Birth asphyxia. 139 140 Table A2.6: Assessment of student performance   Provider Location Region Welfare National Public Priv. Urb. Rur. Central Eastern Kampala Northern Western Q1 Q2 Q3 Q4 Q5 English 43 67 63 42 68 35 85 35 54 31 29 55 62 69 46 Numeracy 42 53 48 42 52 40 56 37 47 38 37 47 50 50 43 Non-verbal reasoning 56 62 60 56 62 56 67 51 59 52 56 60 59 60 57 Overall score 43 65 61 43 66 37 80 36 54 34 31 55 61 66 47 Source: Staff calculations using the 2013 SDI and the UNHS 2013 surveys. Table A2.7: Correlates of pupil achievement (Probit model)   Model 1: Model 2: nu- Model 3: Non-verbal Model 4: overall Model 5: Model 6: nu- Model 7: Non-ver- Model 8: overall English meracy reasoning score English meracy bal reasoning score Coef t Coef t Coef t Coef t Coef t Coef t Coef t Coef t School absence rate −0.264* 0.154 −0.219*** 0.071 −0.000 0.099 −0.231* 0.125 −0.317** 0.155 −0.265*** 0.079 −0.030 0.123 −0.392** 0.173 Classroom absence rate −0.303** 0.123 −0.063 0.056 −0.065 0.076 −0.245** 0.098 −0.263** 0.123 −0.056 0.062 −0.062 0.095 −0.286** 0.134 There is a PTA −0.009 0.063 0.018 0.028 0.120*** 0.040 −0.004 0.050 0.001 0.063 0.019 0.028 0.120*** 0.040 0.000 0.051 There is a SMC 0.070 0.064 0.010 0.029 0.044 0.042 0.060 0.051 0.074 0.064 0.013 0.029 0.047 0.041 0.067 0.050 Number of inspections 0.009* 0.005 0.010*** 0.002 0.007** 0.003 0.009** 0.004 0.009* 0.005 0.010*** 0.002 0.007** 0.003 0.009** 0.004 Index classroom environ- 1.072*** 0.148 0.278*** 0.061 0.081 0.088 0.828*** 0.116 1.062*** 0.148 0.277*** 0.061 0.084 0.088 0.826*** 0.116 ment Index teacher behavior 0.371*** 0.137 −0.031 0.065 0.044 0.089 0.292*** 0.111 0.366*** 0.137 −0.029 0.065 0.054 0.089 0.293*** 0.111 Teacher score English 0.992*** 0.249 0.156 0.161 0.804*** 0.207 0.917*** 0.249 0.119 0.161 0.743*** 0.208 Teacher score numeracy 0.426*** 0.079 0.306*** 0.108 0.061 0.138 0.408*** 0.080 0.273** 0.108 0.018 0.139 Age −0.190*** 0.021 −0.014 0.009 −0.027** 0.012 −0.146*** 0.017 −0.189*** 0.021 −0.013 0.009 −0.025** 0.012 −0.145*** 0.017 Age squared 0.002*** 0.000 0.000 0.000 0.000* 0.000 0.001*** 0.000 0.002*** 0.000 0.000 0.000 0.000 0.000 0.001*** 0.000 Girl −0.054 0.047 −0.123*** 0.021 −0.066** 0.029 −0.067* 0.037 −0.054 0.047 −0.121*** 0.021 −0.066** 0.029 −0.068* 0.037 Has breakfast 0.157*** 0.050 −0.013 0.023 0.001 0.031 0.111*** 0.041 0.147*** 0.050 −0.016 0.023 −0.004 0.031 0.102** 0.041 Public school −0.626*** 0.072 −0.328*** 0.032 −0.171*** 0.044 −0.545*** 0.057 −0.531*** 0.081 −0.293*** 0.035 −0.107** 0.047 −0.458*** 0.064 Urban area 0.146 0.095 0.063 0.041 0.131** 0.058 0.133* 0.077 0.137 0.095 0.063 0.041 0.132** 0.058 0.131* 0.076 Region Central ref. ref. ref. ref. ref. ref. ref. ref. ref. ref. ref. ref. ref. ref. ref. ref. Eastern −0.784*** 0.132 −0.319*** 0.061 −0.370*** 0.086 −0.699*** 0.111 −0.705*** 0.133 −0.291*** 0.061 −0.322*** 0.087 −0.635*** 0.111 Kampala 0.267* 0.137 −0.037 0.050 0.061 0.078 0.144 0.099 0.291** 0.138 −0.035 0.050 0.063 0.078 0.155 0.100 Northern −0.333** 0.150 −0.259*** 0.068 −0.452*** 0.095 −0.354*** 0.125 −0.249 0.152 −0.231*** 0.069 −0.399*** 0.096 −0.288** 0.126 Western −0.145** 0.068 0.001 0.032 −0.159*** 0.043 −0.119** 0.056 −0.122* 0.069 0.009 0.032 −0.144*** 0.043 −0.100* 0.057 Welfare quintile Q1 ref. ref. ref. ref. ref. ref. ref. ref. ref. ref. ref. ref. ref. ref. ref. ref. Q2 −0.157** 0.079 −0.156*** 0.036 0.008 0.048 −0.145** 0.063 −0.141* 0.079 −0.150*** 0.036 0.013 0.049 −0.129** 0.063 Q3 0.233* 0.142 −0.036 0.066 −0.112 0.090 0.159 0.120 0.208 0.142 −0.054 0.066 −0.145 0.090 0.124 0.120 Q4 0.631*** 0.143 0.049 0.065 −0.288*** 0.093 0.488*** 0.120 0.612*** 0.143 0.033 0.065 −0.315*** 0.093 0.460*** 0.120 Q5 0.590*** 0.144 0.033 0.066 −0.206** 0.091 0.447*** 0.121 0.555*** 0.145 0.012 0.066 −0.244*** 0.091 0.405*** 0.121 Pupil to teacher ratio −0.005*** 0.002 −0.002*** 0.001 −0.004*** 0.001 −0.004*** 0.001 (PTR) Constant 1.310*** 0.308 0.062 0.135 0.586*** 0.197 1.053*** 0.261 1.445*** 0.308 0.120 0.136 0.709*** 0.199 1.215*** 0.262 Number of observations 3,565 3,576 3,546 3,546 3,555 3,566 3,536 3,536 141 Source: Staff calculations using the 2013 SDI and the UNHS 2013 surveys. Note: 0.01 - ***; 0.05 - **; 1 - * 142 For more information, please visit: www.worldbank.org/uganda Join the discussion on: http://www.facebook.com/worldbankafrica http://www.twitter.com/worldbankafrica http://www.youtube.com/worldbank 144