Report No: AUS0001090 . Uganda Poverty Monitoring and Analysis Detour from the Poverty Reduction Path: Uganda Poverty Update Note – Uganda National Household Survey 2016/1 . July 2019 . POV (Poverty and Equity Global Practice) World Bank Group . . © 2019 The World Bank 1818 H Street NW, Washington DC 20433 Telephone: 202-473-1000; Internet: www.worldbank.org Some rights reserved This work is a product of the staff of The World Bank. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of the Executive Directors of The World Bank or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Rights and Permissions The material in this work is subject to copyright. Because The World Bank 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. Attribution—Please cite the work as follows: “World Bank. 2019. Detour from the Poverty Reduction Path: Uganda Poverty Update Note – Uganda National Household Survey 2016/17. © World Bank.” All queries on rights and licenses, including subsidiary rights, should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. 2 Main messages1 • Progress against poverty in Uganda remains fragile. Between 2012 and 2016 the poverty rate in Uganda increased moderately reaching 21.4 percent in 2016, a 1.7 percentage point increase that resulted in around 1.4 million Ugandans slipping into poverty. This is the result of the overall economic slowdown observed since 2012, coupled with a severe drought that affected the country for the better part of 2016 and 2017. During the period 2012 to 2016, the 40 of the distribution benefitted less from the economic progress, resulting in a more unequal distribution of consumption. The regional gap in living conditions persists, with the Northern and Eastern region lagging behind the rest of Uganda along many dimensions; despite the progress observed in the North. On a positive note, consumption increased for the ultra-poor, those in the lowest 10 percent of the distribution. • Reducing the vulnerability of rural households to adverse shocks will be fundamental to ensure sustained progress against poverty reduction going forward. The overall increase was mainly explained by the increase in rural poverty, which went up from 22.8 to 25.3 over the four-year period. These events reflect the high degree of vulnerability of rural households to weather shocks, and underscores the need to modernize agricultural production, through drought resistant seeds, investments in irrigation technologies and extension services, and engage these households in agroindustrial activities. The expansion of existing safety net programs (currently concentrated in the Northern region and exhibiting low coverage rates) and the introduction of new ones, can help reduce the negative effects of adverse shocks to vulnerable households and reduce poverty. This should be done in a way that does not compromise the macroeconomic stability of the country, taking into consideration its limited fiscal space. • Enabling households to increase their participation in non-agricultural economic activities is also fundamental to increase the living standards of the population. The vulnerability of rural households to weather shock and the stagnation of progress in urban areas (poverty has remained unchanged at around 9 percent), underlines the need to increase the demand for household labor outside the agricultural sector and in more productive sectors. This requires important investments to improve the education outcomes and skills of the population, that should be complemented with policies focused on firm and job creation in high productivity activities. • Further improvement of the non-monetary indicators of poverty and the reduction of regional disparities require ensuring access of all households to basic services of adequate quality. This involves an effective service delivery system at the local level (in particular of education -secondary level-, health and agricultural extension services), important infrastructure investments at the central level (mainly in electricity and sanitation) and a strong coordination between the two. It also entails keeping the policy focus on the less developed areas of the Uganda. 1This policy note was written by Carolina Mejía-Mantilla (Senior Economist, GPV01) and Johanna Fajardo-Gonzalez (Consultant, GPV01), with most of data work was completed by Dan Pavelesku (Consultant, GPV01). The report benefitted from guidance and comments from Pierella Paci (Practice Manager, GPV01), Nobuo Yoshida (Lead Economist, GPV01) and Allen Dennis (Program Lead, EA1DR). Special thanks to James Muwonge (UBOS) and Vincent Ssennono (UBOS) for their collaboration with data, queries, and for the various discussions. 3 Table of Contents I. Background ........................................................................................................................................... 6 II. Recent trends in poverty in Uganda ..................................................................................................... 8 a) Regional patterns of poverty ...................................................................................................... 12 b) Revisiting the poverty methodology in Uganda ......................................................................... 15 III. Incidence of progress and inequality trends................................................................................... 16 a) Evolution of inequality indicators ............................................................................................... 19 IV. What explains the changes in poverty in Uganda........................................................................... 21 a) Demographic change .................................................................................................................. 21 b) The role of growth versus changes in the distribution ............................................................... 22 c) The role of poverty reduction within geographical areas versus internal migration ................. 24 d) Structural pattern of poverty changes ........................................................................................ 24 e) Decomposing the poverty decline in the Northern region ......................................................... 25 V. Evolution of non-monetary poverty indicators .................................................................................. 27 VI. What factors affect households’ consumption in Uganda? ........................................................... 31 VII. Who are the poor in Uganda?......................................................................................................... 33 VIII. Conclusions ..................................................................................................................................... 37 Going forward ......................................................................................................................................... 38 References .................................................................................................................................................. 40 4 List of Figures Figure 1. Recent trends in Uganda ................................................................................................................ 7 Figure 2. Poverty Trends in Uganda: 2005-2016 ........................................................................................ 10 Figure 3. Additional measures of poverty in Uganda ................................................................................. 11 Figure 4. Geographical distribution of the poor ......................................................................................... 13 Figure 5. Weather shocks and increases in poverty rates .......................................................................... 14 Figure 6. Real annualized consumption growth, 2012-2016 ...................................................................... 18 Figure 7. Real annualized consumption growth by regions, 2012-2016 .................................................... 19 Figure 8. Real annualized consumption growth by regions, 2012-2016 .................................................... 20 Figure 9. Decline in fertility has contributed to poverty reduction ............................................................ 22 Figure 10. Decomposing poverty change in Uganda 2012-2016 ................................................................ 26 Figure 11. Non-monetary poverty indicators ............................................................................................. 29 Figure 12. Child mortality and nutrition indicators .................................................................................... 30 List of Tables Table 1. Growth-redistribution decomposition .......................................................................................... 23 Table 2. Decomposing poverty change in Uganda 2012-2016 ................................................................... 25 Table 3. What factors are associated with higher per capita consumption ............................................... 32 Table 4 Poor versus non-poor (and all population) socio-economic conditions ....................................... 35 Table 5 Socio-economic conditions by region ........................................................................................... 36 5 I. Background Uganda’s progress in reducing poverty during the late 1990s and early 2000s was remarkable. Between 1992 and 2005, the national poverty rate decline at an impressive pace of 1.9 percentage points per year, one of the fastest declines in Sub-Saharan Africa and the world.2 The restoration of peace and stability in the country, coupled with a series of economic liberalization reforms under the leadership of President Museveni and the response of households and firms to the new economic and political order (Collier and Reinikka, 2003; World Bank 2007), resulted in a period of high economic growth. This, in turn, improved the incomes of households across the board, reducing poverty and improving the living conditions of the entire population. While economic growth slowed down between 2005 and 2012, it particularly benefitted the poorest segments of the population. This allowed for poverty reduction to continue, albeit a slower pace than before -poverty declined at a pace of 1.6 percentage points per year during this six-year period-.3 The strong pro-poor consumption growth in rural areas, particularly between 2009 and 2012 was rooted in the expansion of agricultural income aided by peace in northern Uganda, more integrated regional economic markets (particularly with the East African Community countries), and favorable weather conditions (World Bank, 2016). Moreover, 2009 to 2012 was the only period since 2002 in which consumption growth benefited the bottom 40 percent of the distribution more than the top 60 percent, lowering inequality. Since 2012, economic growth in Uganda continued to slow down, particularly in per capita terms. The average annual growth of real Gross Domestic Product (GDP) between 2012 and 2016 was 4.5 percent, a marked deceleration compared to 6.4 between 2009 and 2012 and, particularly, compared to 8.2 between 2005 and 2009 (see Figure 1a). While the slowdown is observed in all sectors of the economy, from agriculture to services and the manufacturing activities (see Figure 1b), the stagnation of the agricultural sector (at growth rates lower than 2 percent) is particularly worrying, considering that the agricultural sector still employs around 70 percent of the labor force. Moreover, the economic slowdown is more pronounced in per capita terms, due to persistently high fertility rates. GDP per capita grew only 1.1 percent between 2012 and 2016, considerably lower than the population growth rate of around 3.3 percent for that same period, and a significant decline compared to the 4.5 per capita growth experienced in the period 2005 and 2009 (see Figure 1a). Not only has growth faltered, the responsiveness of poverty reduction to growth has also declined. Since 2009 it seems that the same measure of economic growth rate results in a lower decline of the poverty rate, compared to what was observed between 2000 and 2009; a concerning result. The poverty- 2 Poverty monitoring in Uganda is possible because the Government of Uganda (GoU) has invested in a high-quality series of household surveys to document progress in well-being since 1993. The Uganda Bureau of Statistics (UBOS) has conducted comparable household surveys every three to four years. For this note, we use the UNHS 2005/06, 2009/10, 2012/13 and 2016/17, referred as 2005, 2009, 2012 and 2016 data points in this note. 3 For an in-depth analysis of poverty and inequality trends between 2005 and 2012, see the World Bank 2016 Poverty Assessment Report. 6 growth elasticity, the elasticity of the poverty headcount rate under the international poverty line to GDP per capita growth, was -0.5 for the period 2009 to 2016, considerably lower than -1.24 for 2000 to 2016. While it is generally true that there are diminishing returns in poverty reduction as countries progress, the poverty elasticity of Uganda is lower than the average for Sub-Saharan Africa and that of Ethiopia and Kenya, and considerably lower than that of Ghana (Figure 1c). The implication is that even if economic growth recovers over the medium term, additional and complementary measures should be taken to reduce poverty at the same pace as in the two decades. Figure 1. Recent trends in Uganda a. GDP per capita shows a downward trend b. Slowdown is observed in all sectors 11.3 12 Annual % growth 7.9 7.8 8 7 9.4 6.3 6.2 5.4 5.4 5.4 4.8 4.6 4 3.1 3.4 4 5.6 5.2 2.1 5.1 4.8 3.8 3.6 3.9 2.9 3.1 2.7 2.8 1.9 2.3 0 1.6 2010 2011 2012 2013 2014 2015 2016 2017 GDP Agriculture Industry Services c. Poverty to GDP per capita elasticity has declined 0 South Ethiopia Poverty to GPD per capita elasticity Saharan Uganda Africa Ethiopia Kenya -0.5 -1 Uganda -1.5 Ghana -2 -2.5 2000-2016 2009-2016 Note: poverty rate used based on the international poverty line and the latest survey available for each of the countries. Source: Word Bank WDI indicators (or own calculations based on them) Moreover, a severe drought affected Uganda for the better part of 2016 and 2017. The country’s rainy seasons (October-December 2016 and March-May 20174) performed poorly and the monthly rainfall for most of the country -except for the Northern region as will be discussed in Section II- was considerably below historical averages. The limited water availability impacted crop and livestock production, 4 The Karamoja region in the North presents a unimodal rather a bimodal rainfall pattern. It goes from April until October. 7 negatively affecting the income of rural households (FEWSNET, 2017). Poor rural households are particularly vulnerable to weather shocks. A recent study estimated that while a 10 percent reduction of rainfall reduces crop income for all rural households by 19 percent, the decline reaches 25 percent for rural households in the bottom 40 percent (Hill and Mejia-Mantilla, 2016). Thus, it is not surprising that the drought was accompanied with an increase in poverty in rural areas, where it is estimated that around 60 to 65 percent of the rural labor force for 2016/17 is engaged in subsistence agriculture (Uganda National Household Survey - UNHS - 2016/17). This note documents the trends in poverty and inequality in Uganda with a focus on the period 2012 to 2016. It first explores how various indicators of monetary poverty have evolved over time and if these trends vary by urban/rural areas and by region. This is complemented with a description of the movements in several non-monetary poverty indicators. Secondly, the note analyzes who has benefitted from the lackluster economic growth, and how this has translated into different measures of inequality. It then aims to disentangle some of the forces behind the changes in poverty, mostly resorting to decomposition exercises. Finally, it provides an account of who are the poor households and what are their living conditions, particularly compared to non-poor households. II. Recent trends in poverty in Uganda While poverty increased between 2012 and 2016, it is still lower than it was a decade ago. Despite the recent increase, the proportion of the population living below the national poverty line5 in Uganda has shown a declining trend over the last decade (Figure 2a). Between 2005 and 2012 the poverty rate declined at about 2.7 percentage points per year, reaching 19.7 percent in 2012. However, this trend reversed by 2016 as a result of the severe draught that affected most of the country for the better part of 2016 and 2017 and of the overall slow economic growth, as documented in Section I. Between 2012 and 2016 poverty incidence increased moderately reaching 21.4 percent in 2016, a 1.7 percentage point increase that resulted in around 1.4 million Ugandans slipping into poverty. The increase is more pronounced when poverty is measured with the international poverty line. Under this benchmark, the proportion of the population living in poverty rose from 35.9 percent in 2016 up to 41.6 percent in 2016 (Figure 2a). Compared to other countries in the region, poverty in Uganda is higher than in Ghana and Ethiopia and slightly higher than in Kenya, while still lower than Rwanda and Tanzania (Figure 2b). It must be noted, however, that recent data points are not available for countries that were also affected by the 2016/2017 drought, such as Kenya and Ethiopia, which likely increased poverty incidence as in Uganda. Poverty incidence increased by 2.5 percentage points in rural areas and remained stagnant in urban centers. Poverty in rural areas went up from 22.8 percent in 2012 to 25.3 percent in 2016, while it remained essentially unchanged at around 9.3 percent in urban areas (the 0.1 increase reported is not 5 See Box 1 for details. 8 statistically significant, as seen in Figure 2c). This represents an increase in the number of rural poor of around 1.3 million since 2012. This confirms the high degree of vulnerability of Ugandan rural households to weather shocks that has been pointed repeatedly by several studies (World Bank 2018, EPRC 2013, Hill and Mejia-Mantilla 2016). As for urban areas, poverty incidence in urban Uganda has been stagnated around 9 percent since 2009, suggesting that the wellbeing of urban households has not improved in almost a decade. Looking closely into poverty incidence in urban areas, the urban poverty rate excluding Kampala fluctuated around 10 percent between 2012 and 2016, while the poverty rate in Kampala increased from 0.7 percent to 2.6 percent, signaling a deterioration of monetary living conditions in the capital. While the gap between rural and urban poverty rates has declined over time, poverty remains a rural phenomenon. Despite the progress observed since 2005, the proportion of the poor that reside in rural areas remains at around 90 percent (Figure 2e). Thus, the number of rural poor at 7.2 million in 2016, is significantly higher the number of urban poor at 0.9 million, which is not surprising given that 76 percent of the population still resides in rural areas (Figure 2f). Also, despite the fact that the urban poverty rate has remained stagnant, the number of urban poor went up by 0.2 million between 2012-2016, reflecting the increase in the urban population from 22.6 to 24.3 percent. The depth and severity of poverty in Uganda, additional measures of poverty, have not changed noticeably between 2012 and 2016. The poverty gap, a measure of the intensity of poverty defined as the average percentage shortfall of consumption with respect to the poverty line (as a proportion of the poverty line)— remained unchanged at 5 percent (Figure 3a). Similarly, the severity of poverty index, which gives more weight to those furthest below the poverty line, fell slightly from 2 in 2012 to 1.9 percent in 2016. These results represent positive news, as they indicate that the poor have not fell deeper into poverty during this period and that inequality amongst the poor has decreased to some extent, despite the overall increase in inequality that will be discussed in detail in Section III. The trends for poverty depth and severity for both urban and rural areas follow closely the national trend, as seen in Figure 3b and c. Consumption levels of the ultra-poor increased between 2012 and 2016. As shown in the density graphs of Figure 3e, in 2016 the consumption per adult equivalent of the ultra-poor (those at the lower end of the consumption distribution, below the 10 percentile) was higher than that in 2012. This helps to explain why despite an increase in poverty, the depth of poverty remained unchanged and why has inequality among the poor declined slightly since 2012. It also highlights that the increase in poverty is the result of the decline in the living conditions of the households bunched just below the poverty line (see Figure 3e). 9 Figure 2. Poverty Trends in Uganda: 2005-2016 a. Headcount poverty rate increased in 2016 b. Increase is more pronounced under the international poverty line (U$1.90 a day 2011 PPP) 60% 55.4% 100 86 50% 77 43.5% 75 67 Proportion of population 41.6% 62 64 60 40% 35.9% 56 55 Proportion of population 49 31.1% 50 44 44 42 30% 37 36 34 36 24.5% 27 25 19.7% 21.4% 20% 25 12 10% 0 2005 2009 2012 2016 2000 2005 2010 2013 2005 2015 2000 2007 2011 2004 2010 2015 2005 2012 2002 2005 2009 2012 2016 National Poverty Rate International Poverty Rate (US$1.9 PPP) Rwanda Kenya Tanzania Ethiopia Ghana Uganda c. The poverty rate increased mainly in rural areas d. Number of poor in Uganda rose in 2016 34.2% 2005 2009 2012 2016 35% Proportion of the population 27.2% Uganda 8.4 7.5 6.6 8.1 25.3% 22.8% 25% Residence Rural 7.9 7.1 5.9 7.2 13.7% 15% 9.1% 9.3% 9.4% Urban 0.6 0.4 0.7 0.9 Region 5% Central 1.3 0.9 0.4 0.9 Eastern 2.5 2.2 2.4 3.5 -5% Northern 3.5 2.8 3.1 2.6 2005 2009 2012 2016 Western 1.4 1.6 0.6 1.1 Rural Urban e. Nearly 1 out of 10 poor people live in rural areas f. Three out of four Ugandas reside in rural areas 100% Proportion of poor population 100% 6.8% 5.6% 15.4% 15.0% 22.6% 10.7% 10.7% 24.3% Proportion of population 80% 80% 60% 60% 93.2% 94.4% 89.3% 89.3% 84.6% 85.0% 40% 40% 77.4% 75.7% 20% 20% 0% 0% 2005 2009 2012 2016 2005 2009 2012 2016 Rural Urban Rural Urban Source: Own calculations based on UNHS 2016/17. 10 Figure 3. Additional measures of poverty in Uganda a. Poverty gap and severity remain unchanged b. Poverty gap urban and rural follow similar trend 10% 15% 9% 9% Poverty gap, severity 8% 7% 9.7% 7% 10% Poverty gap 5% 5% 7.6% 6% 6.0% 6.3% 5% 3.5% 4% 5% 3.5% 2.8% 2.5% 3% 2.0% 1.9% 1.8% 2.0% 2% 1% 0% 2005 2009 2012 2016 2005 2009 2012 2016 Poverty gap Severity of poverty Rural Urban c. Severity of poverty 2.000e-06 e. Kernel density of household expenditure 6% 5% 1.500e-06 3.9% 4% Poverty severity 3.1% 1.000e-06 3% 2.4% 2.3% 2% 5.000e-07 1.4% 0.9% 0.6% 0.6% 1% 0 0% 0 200000 400000 600000 800000 1000000 2005 2009 2012 2016 2009 2012 Rural Urban 2016 Note: real poverty line in red. Real consumption in 2016 prices per adult equivalent. Source: Own calculations based on UNHS 2016/17. Box 1. Measuring poverty 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 in order 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 Consumer Price Index has been used to update this poverty line. Source: Appleton et al (1999) 11 Box 2. Measures of Poverty Poverty is usually measured through consumption or income, aggregated at the household level, and a poverty line. The poverty line indicates the minimum level of welfare required for a healthy living. Healthy living is usually defined as being able to afford the food covering the minimum required caloric intake and additional non-food items. The following three poverty measures are commonly used to assess poverty: Incidence of poverty (headcount rate): The headcount index for the incidence of poverty is the proportion of individuals in the population living below the poverty line. The incidence of poverty can also be calculated for non-monetary measures of well- being by the share of population, which does not reach a defined threshold. Depth of poverty (poverty gap): The depth of poverty indicates how far off in average poor households are from the poverty line. It captures the mean consumption shortfall relative to the poverty line across the whole population. It is obtained by adding up all the shortfalls of the poor (considering the non-poor are having a shortfall of zero) and dividing the total by the population. Thus, the depth of poverty shows the total resources needed per capita to eliminate poverty assuming that all poor individuals would obtain exactly the shortfall between their consumption and the poverty line. Poverty severity (squared poverty gap): The poverty severity considers the distance separating the poor from the poverty line (the poverty gap) as well as the inequality among the poor. Conceptually, poverty severity puts a higher weight on households/individuals, who are further below the poverty line. Source: Foster, et al. (2013). a) Regional patterns of poverty Poverty increased in all regions of Uganda except for the Northern region, and 75 percent of the poor reside in the Eastern and Northern regions. The increase in poverty incidence between 2012 and 2016 was more pronounced in the Eastern region (11.2 percentage points) compared to the Central and Western regions (4.1 and 2.7 percentage points, respectively, see Figure 4a), likely linked to the more favorable weather conditions for the latter seen in Figure 1d. Currently, the Eastern region has the highest poverty incidence, at 35.7 percent, concentrating almost 44 percent of Uganda's poor 2016 (up from 37 percent in 2012, see Figure 4b). The Northern region follows with a poverty rate of 32.5 percent, well above the national poverty incidence of 21.4 percent, but experiencing a decline since 2012 (Figure 4a). In 2016, these two regions concentrated 75 percent of the poor population in Uganda, despite the fact only 28 percent of the total population resided there. The decline in poverty in the Northern region occurred both in urban and rural areas. Between 2012 and 2016, rural poverty decreased from 46.5 percent to 36 percent, while urban poverty more than halved reaching 13.6 percent (Figure 4d). Traditionally, the Northern region has had the lowest welfare levels, partly explained by the long-lasting conflict between the government and the Lord Resistance Army (LRA), which curbed capital and human accumulation in the region for over two decades. The pacification of the region, around the period 2006-2008, together with favorable food and weather conditions contributed to a considerable reduction in the poverty rate through the 2005-2012 period, reaching 43.7 percent in 2012. The trend continued, and poverty dropped to 32.5 percent by 2016. 12 Favorable weather conditions in the Northern region, compared to the rest of the country, help explain this pattern. As can be seen in Figure 5a, the rainfall patterns varied significantly across regions for the year 2016 (the same is true for 2017, not pictured here), with null / little rainfall deficits in the Northern region and more pronounces deficits in the rest of the country, particularly in the Western and Eastern regions. Given the high dependency on households on agricultural production for their livelihood, rainfall deficits seem to be highly spatially correlated (albeit not perfectly) with increases in poverty at the sub- county level, as shown in Figure 5. An additional factor that seemed to have contribute to the poverty reduction in the Northern region was the increase in transfers (which includes both remittances and social security benefits) received by households in this region and explored further in Section IV. Figure 4. Geographical distribution of the poor a. Poverty rates increased in all but the Northern region b. Distribution of poor by region 70% 100% 10.3% 13.5% 60.7% 17.0% 21.3% 60% Proportion of the population 80% Proportion of the population 50% 46.2% 31.6% 43.7% 46.6% 38.5% 60% 37.7% 40% 35.9% 35.7% 32.5% 24.3% 30% 24.5% 20.5% 40% 21.8% 43.6% 20% 16.4% 29.0% 8.7% 29.3% 36.9% 10.7% 11.4% 20% 8.8% 10% 4.7% 15.4% 11.6% 11.3% 6.2% 0% 0% Central Eastern Northern Western 2005 2009 2012 2016 Central Eastern Northern Western 2005 2009 2012 2016 c. Distribution of population by region d. Poverty in Northern region 100% 80% Proportion of the population 90% 24.0% 23.5% 64.2% Proportion of the population 25.9% 25.5% 80% 60% 70% 49.0% 46.5% 19.7% 20.0% 21.1% 20.9% 39.7% 60% 40% 36.0% 50% 29.8% 40% 25.2% 29.6% 29.7% 26.2% 19.7% 30% 13.6% 20% 20% 29.2% 26.5% 25.8% 27.5% 10% 0% 0% 2005 2009 2012 2016 2005 2009 2012 2016 Central Eastern Northern Western Rural Urban Source: Own calculations based on UNHS 2016/17. 13 Figure 5. Weather shocks and increases in poverty rates a. Rain deficit (SPEI indicator)- 2016 b. Changes in poverty rate at the sub-county level Source: Own calculations based on UNHS 2016/17. SPEI (Standardized Precipitation and Evaporation Index) – from https://spei.csic.es/. Regionally focused economic policies and the expansion of social protection programs, more specifically direct income support programs, can help reduce the regional gap. The spatial concentration of poverty in the Northern and Eastern regions underscores the importance of policies that allow all regions to reap the benefits of economic growth, and that reduce regional gaps in terms of living standards. The recent efforts of the Government of Uganda in the Northern region through social protection programs such as NUSAF, School Feeding Program and Mother and Child Health and Nutrition, particularly in the Karamoja districts, should be commended. A recently fiscal incidence study completed by the World Bank and the Commitment to Equity Institute indicates that while social protection programs in Uganda are progressive fiscal instruments that reduce inequality and poverty their effect on reducing poverty and inequality is minor because they are concentrated in just one region (Northern) and their scope and coverage is limited (World Bank 2019). The expansion of these programs within the Northern region and to other regions, particularly the Eastern region, would greatly benefit the poor and would help to reduce the overall poverty incidence of the country. 14 b) Revisiting the poverty methodology in Uganda6 The national poverty line in Uganda was established using data from 1993 and it is very likely that it no longer reflects the consumption patterns of the poor. As mentioned in Box 1, the national poverty line was set based on an in-depth analysis of the pattern of food and non-food consumption among Uganda’s poor in 1993 (Appleton et al., 1999). It is expected that consumption patterns have changed considerable since 1993. The share of consumption that the poor spend on non-food items is around 20-30 percent higher in 2016 than in 1993, when the poverty line was set. For example, in 1993 no household owned a mobile phone, yet today around 50 percent of the poor households in Uganda own mobile phones and purchase credit to make and receive calls. Moreover, relative prices of food items have changed substantially since 1993 and households likely adjusted their food consumption patterns in response. The prices of food items in the food poverty line basket seem to have risen faster than the CPI, on average. Under the current methodology, the CPI index has been used to adjust the poverty line for the changes in prices over the years. However, in the case of Uganda, the CPI index only covers urban centers, disregarding price changes in the rural parts of the country. While the current poverty methodology adjusts for differences in prices between urban and rural areas (across the four regions) during the period in which data are collected (usually 12 months), this might not be enough to capture changes in price differentials between the survey years (usually a three to four-year period). Analysis of household data up to 2012 shows that the price of food items in the food poverty line basket may have risen faster than the CPI on average, which suggests that cost of the poverty line basket is underestimated under the current methodology (World Bank 2016). Finally, the national poverty line much lower than international poverty line. Uganda has different poverty lines for different regions to account for the fact that the cost of living varies across different parts of the country (see Box 1 for more details on how poverty is measured in Uganda). When these poverty lines are converted into 2011 PPP they vary from US$1.36 to US$1.55, in other words 72 to 82 percent of the international extreme poverty line of US$1.90. The international extreme poverty line is designed to capture the average national poverty line among the world’s poorest countries (most of them neighbor countries in Africa), so the fact that Uganda’s poverty lines are much lower, suggests that the poverty line in Uganda is perhaps too low and is not adequately capturing the cost of the basic needs for the poor. It is appropriate to revisit the current poverty measurement methodology in Uganda, in order for the poverty line to adequately reflect the current consumption patterns of the poor and to make sure that the price adjustments over time preserve the real value of the poverty basket. The GoU is aware of this issue and is considering a revision for the UNHS 2021/22 wave. This is an important initiative that should start right away and should take into consideration alternative methodologies to account for price differences between geographical regions and over time, including the price changes in rural areas. 6 This section relies is based on World Bank (2016). 15 III. Incidence of progress and inequality trends Consumption growth was lowest (negative) for households between the 10th and 40th percentiles of the distribution, which explains the increase in the poverty rate. To better understand what parts of the distribution have benefited most from the economic growth that Uganda experienced between 2012 and 2016, growth incidence curves (GIC) that display annualized real consumption growth over the entire distribution are presented (Figure 6). Consumption growth7 for the poorest 10 percent of households was positive and higher than the average growth for the entire distribution. This fact helps explain why the severity of poverty (which gives more weight to poor households furthest from the poverty line) fell slightly at the same time that poverty was increasing. This also highlights the fact that what explains the increase in poverty is the negative consumption growth for households between the 10th and 40th percentiles of the distribution (Figure 6a). Excluding the ultra-poor (those below the 10th percentile), consumption was higher for households in the upper end of the distribution. Consumption growth rates become positive for households above the 45th percentile, with rates above the average growth rate for households between the 55th and 70th percentiles, and for those households at the very top. This resulted in higher inequality, as discussed below. A plausible explanation for this pattern of consumption growth is that better-off households can cope better with negative shocks, such as droughts, and are engaged at higher proportions in the sectors that drove overall economic growth, mainly financial and communication services. Overall, the incidence of progress benefitted the upper end of the distribution in rural areas while it benefitted the lower end in urban areas. The growth incidence curve for rural areas mimics that of the national level. It shows negative consumption growth for households between the 7th and 55th percentile and positive consumption growth for households at the top of the distribution (Figure 6b), which explains the poverty increase in rural areas. As for urban households, growth rates were more homogenous across the distribution, and the average annualized rate was 1 percent. While the pattern was slightly pro-poor, lowering the severity of poverty in rural areas, this was not enough to bring down the poverty headcount rate. Incidence of economic progress between 2012 and 2016 was less favorable for the bottom 40 of the distribution than in previous periods. The annualized growth rate of per adult equivalent consumption for the bottom 40 percent of the population was close to zero (0.05 percent), much lower than the 3.3 and 3 percent growth observed for the 2005-2009 and 2009-2012 periods respectively (Figure 6d). Moreover, progress disproportionally benefitted better-off households. The premium indicator (calculated as the difference between the growth rate of the bottom 40 percent and the average growth rate for the whole distribution) was negative at -0.5 percentage points. This represents a major shift from 7 Consumption per adult equivalent expressed in real terms (2016 prices). 16 the broad-based pattern of economic progress observed during 2009-12, which as explained before, allowed Uganda to make progress against poverty despite lower overall economic growth. Uganda has been less successful than other countries in the region in boosting shared prosperity.8 The annualized growth rate of per capita consumption in 2011 PPP dollars (to allow for international comparisons) for the bottom 40 was -0.3 percent between 2012 and 2016. This compares unfavorably to other countries in the region like Mozambique, Ethiopia or Tanzania which experienced rates of 1.5, 1.7 and 5.4 percent respectively, in recent periods (Figure 6e). However, it must be noted once more that the latest available surveys for these countries do not capture the effects of the 2016/17 drought, and thus, the actual shared prosperity estimate likely lay closer to Uganda’s number. Consistent with the regional poverty trends of poverty, there are marked differences in the incidence patterns of consumption growth across the different regions. Overall, while average consumption growth between 2012-2106 was negative for the Central and Eastern regions (at about -1 percent), it was positive for the Northern and Western regions. Moreover, the growth incidence curves show that the decline was relatively evenly distributed in the case of the Central region and was less pronounced for richer households in the Western region (Figure 7a and Figure 7b). In marked contrast, the Northern region shows positive annualized growth rates for all percentiles of the distribution, and a pro-poor pattern, which explains why it was the only region for which poverty declined during this period (Figure 7c). Finally, while the Western region also displays a positive annualized growth rate of 1 percent, it disproportionally benefitted the upper end of the distribution (Figure 7d). As a matter of fact, consumption fell for the bottom households, consistent with the poverty increase. 8The shared prosperity indicator is defined as the growth of the (unadjusted) per capita consumption of the bottom 40 percent of the distribution. In the case of Uganda, this consumption aggregate is different from the one used to derive the poverty estimates, as the latter is in per adult equivalent term as opposed to per capita terms and is adjusted differently for inflation. 17 Figure 6. Real annualized consumption growth, 2012-2016 a. Growth incidence curves – National b. Growth incidence curves – Rural 6.0 4.0 4.0 2.0 Annual growth rate Annual growth rate 2.0 (%) (%) 0.0 0.0 -2.0 -2.0 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Per capita expenditure percentiles Per capita expenditure percentiles c. Growth incidence curves – Urban d. Shared prosperity over time 10.0 6% 2.5 Premium (percentage points) 2.0 2.0 4.5% 5.0 Annualized growth rate ( %) 1.5 4% 3.3% 3.0% 1.0 Annual growth rate 0.5 (%) 0.0 -0.5 0.0 2% 1.0% -0.5 0.5% -1.0 -5.0 -1.1 0.0% 0% -1.5 2005/09 2009/12 2012/16 -10.0 0 10 20 30 40 50 60 70 80 90 100 Bottom 40 Total pop. Premium (right axis) Per capita expenditure percentiles e. International comparison shared prosperity 6% 5.4% 4.8% 5% Annualized per capita consumption growth of B40 of the population, % 4% 3% 2% 1.5% 1.7% 1% 0% -1% -0.3% Uganda Mozambique Ethiopia Rwanda Tanzania (2012-16) (2008-14) (2010-15) (2010-13) (2007-11) Source: a to d Own calculations based on UNHS 2016/17; e. Word Bank WDI indicators (or own calculations based on them). Note: the dotted horizontal line depicts the zero line, while the solid horizontal line depicts the average for the entire distribution. 18 Figure 7. Real annualized consumption growth by regions, 2012-2016 a. Growth incidence curves – Central Region b. Growth incidence curves – Eastern Region 10.0 5.0 5.0 0.0 Annual growth rate Annual growth rate (%) (%) 0.0 -5.0 -5.0 -10.0 -10.0 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Per capita expenditure percentiles Per capita expenditure percentiles c. Growth incidence curves – Northern Region d. Growth incidence curves – Western Region 15.0 15.0 10.0 10.0 Annual growth rate Annual growth rate (%) (%) 5.0 5.0 0.0 0.0 -5.0 -5.0 0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 Per capita expenditure percentiles Per capita expenditure percentiles Source: Own calculations based on UNHS 2016/17 Note: the dotted horizontal line depicts the zero line, while the solid horizontal line depicts the average for the entire distribution. a) Evolution of inequality indicators Inevitably, the pattern of pro-rich consumption growth brought along an increase in inequality. Inequality, as measured by the Gini index, increased from 0.40 in 2012 to 0.42 in 2016 (Figure 8a), returning to the levels observed in 2009. This places Uganda in a moderate level of inequality compared to other countries in the region (Figure 7b), similar to that of Tanzania, Kenya and Ghana, and much lower than Mozambique’s index of 0.54.The increase in inequality in confirmed by other measures of inequality such as the Theil index (with a parameter α=−1, which emphasizes inequality for lower incomes) and the 90/10 and 75/25 ratios (Figure 8c). Under the Theil index, inequality increased from 0.29 in 2012 to 0.34 in 2016 while the evolution of the 90/10 and 75/25 ratios show that inequality rose more at the extremes rather than in the middle of the consumption distribution. Inequality is primarily explained by differences within urban and rural areas and within regions, rather than by differences between these groups. A decomposition analysis of the Theil index allows for a better understanding of the nature of inequality and how it has changed over time. In 2016, about 87 percent of 19 the inequality can be attributed to differences within rural and urban households (Figure 8d), rather than to differences between rural and urban households.9 This trait of inequality has not changed significantly since 2005. Figure 8. Real annualized consumption growth by regions, 2012-2016 a. Inequality increased between 2012 and 2016 b. Uganda places in a moderate level of inequality in Sub-Saharan Africa 0.5 0.43 0.42 60 54.0 0.41 46.3 0.40 50 45.1 0.4 40.8 42.8 0.36 37.8 38.6 39.1 40 0.34 0.4 0.32 30 0.29 20 0.3 10 0.3 0 0.2 2005 2009 2012 2016 Gini Theil c. Increase in inequality is also confirmed by the 90/10 d. Inequality is more prevalent within urban and rural and 75/25 ratios areas (decomposition Theil index) 7 0.35 5.63 5.72 5.86 0.30 0.29 6 5.44 0.30 0.27 0.24 5 0.25 4 0.20 2.42 2.36 2.34 2.50 3 0.15 2 0.10 0.05 0.06 0.05 0.04 1 0.05 0 - 2005 2009 2012 2016 Within-group inequality Between-group inequality p90p10 p75p25 2005 2009 2012 2016 Source: Own calculations based on UNHS 2016, World Bank, World Development Indicators Box 3. 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 the entire 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. 9 A similar analysis that disaggregates by region shows that inequality is mainly rooted in the differences within regions rather than across regions. 20 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 consumption / 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 consumption in the 90th and 10th percentile. Given the nonresponse issues in Nairobi, we opt for the p75/p25 ratio, which is the average consumption ratio in the 75th and 25th percentiles. Finally, the Atkinson index introduces value judgements about the degree of inequality aversion prevalent in the society, which is expressed by the choice of an inequality aversion parameter. The higher this parameter, the more emphasis is placed on the lower tail of the distributions and the changes experienced there. Source: World Bank’s Poverty Handbook IV. What explains the changes in poverty in Uganda It is possible to shed some light on the drivers behind the observed changes in the poverty incidence of Uganda between 2012-2016 by using decomposition exercises. This section examines the role of demographic progress, growth versus redistribution, the progress in urban / rural areas and the different regions as well as the population shift amongst them. It also explores the relative importance of economic sectors, and thus it implicitly shades light on the effect of weather shocks in the poverty levels. a) Demographic change Despite recent progress, the total fertility rate (TFR) in Uganda remains one of the highest in Africa. In the last decade, the country has made progress in reducing fertility, bringing the TFR down from 6.2 in 2009 to 5.6 in 2009. However, this figure is well above the Sub-Saharan African average of 4.8 (Figure 9a).10 As discussed above, the issue with high fertility rates is that the level of economic growth required to underpin to poverty reduction (taking into account the limited scope of redistributive policies in the country) is high, and thus, moderate economic growth, while contributing to poverty reduction, may not be enough going forward. Moreover, even if the poverty headcount rate goes down, the actual number of the poor may still increase. Having said that, the progress observed between 2012 and 2016 prevented a larger increase in the poverty rate, as total household consumption declined by around 2.7 percent. The reduction in fertility has contributed to poverty reduction since 2005. Between 2005-2009 it was coupled by spectacular growth in total household consumption, and poverty declined at an annualized rate of 5.8 percent. Between 2009-2012 it was coupled with very pro-poor growth, resulting in a 6.9 percent annual decline. 10 This exclude upper-middle income and upper-income countries. 21 During the period of focus of this note, 2012-2016, the decline in the size of Ugandan households helped to offset the large drop in real total household consumption observed in Figure 9b. The result was a slight increase in the per adult equivalent consumption of around 0.5 percent per year (see Figure 9c). Thus, had it not been for the improvements in fertility, the reduction in the wellbeing of the population would have been much worse. This underscores the importance of strengthening recent efforts by the GoU and the international community to continue addressing this challenge by empowering women. Figure 9. Decline in fertility has contributed to poverty reduction a. Fertility rate in Uganda remains high b. Per-adult-equivalent consumption vs. household consumption 1,500 4,300 7 Fertility rate (births per women) 6.2 4,200 6.0 6 5.5 1,000 4,100 5.6 5.3 Thousands Thousands 5.3 5.1 5.1 4,000 5.0 5 4.6 4.7 4.8 4.3 4.3 500 3,900 4.2 4.2 3.9 4.0 3,800 4 0 3,700 3 2005 2009 2012 2016 2009 2012 2016 Per AE annual expenditure, constant prices Sub-Saharan Africa Uganda Rwanda Tanzania Ethiopia Ghana Household annual expenditure, constant prices, right axis c. Per-adult-equivalent expenditure vs. adult- equivalence and household size 1,600 1,388 1,418 6.0 5.19 1,349 1,400 1,133 4.93 5.0 4.80 Number of members 1,200 4.43 4.0 1,000 Thousands 800 3.93 3.72 3.0 3.62 3.38 600 2.0 400 1.0 200 0 0.0 2005 2009 2012 2016 Per AE annual expenditure, constant prices Sum total of AE scales, right axis Household size, right axis Source: Own calculations based on UNHS 2016/17 b) The role of growth versus changes in the distribution Rising inequality has curbed poverty reduction both at the national level and in rural areas . Changes in poverty in Uganda can be decomposed into “pure growth” and “distribution” components to determine how each factor contributed to the change in the poverty rate observed. Between 2012 and 2016, the increase in the mean average household consumption11 (growth effect), a poverty reducing force, was not enough to offset the increase in inequality (distribution effect), traditionally a poverty increasing force. 11 Mean average households per-adult equivalent consumption. 22 The distribution effect alone would have increased poverty by 3.84 percentage points between 2012 and 2016 had the growth effect not reduced poverty by 2.16 percentage points (Table 1). The predominance of the distribution effect is also visible when analyzing the poverty increase in rural areas, where the rise in inequality would have resulted in a 5.08 percentage point rise if it had not been for the growth factor. Table 1. Growth-redistribution decomposition National Rural Urban Central Eastern Northern Western Contribution Growth -2.16 -2.58 -0.24 0.64 7.05 -13.18 -2.76 Distribution 3.84 5.08 0.32 3.42 4.08 2.02 5.43 Total change in p.p. 1.68 2.50 0.08 4.06 11.13 -11.16 2.67 Percentage Growth -128% -103% -303% 16% 63% 118% -104% Distribution 228% 203% 403% 84% 37% -18% 204% Source: Own calculations based on UNHS 2016 Box 4. What lies in the decomposition of changes in poverty? In this chapter the results of two decomposition methods are presented. The first method is the Datt-Ravallion approach, which isolates the growth and redistribution effects associated with the decline in poverty over the period of analysis. Conceptually, this decomposition is based on the idea that that a measure of monetary poverty can be expressed as the product of mean consumption and a parameterized Lorenz curve. Keeping the Lorenz curve constant gives the distribution neutral growth that would drive the average increase in consumption across the population, for instance, raising the levels of consumption of all households by the same rate. The other part is derived from holding the mean consumption constant (a mean-preserving redistribution) to capture the change in the shape of the consumption distribution driven by, for instance, a faster growth in the consumption of the poorest relative to the consumption growth of the richest (Datt and Ravallion, 1992). The second is the Ravallion and Huppi (1991) decomposition method, that quantifies how much poverty reduction among mutually exclusive groups, or movement between these groups, accounts for national poverty reduction . More specifically, the analysis decomposes changes in poverty over time into “intra-group effects” (poverty changes within sectors, within provinces, or within urban and rural areas, while assuming no changes in the distribution of the population across groups), “inter -group effects” (allowing for changes in the distribution of the population between groups keeping poverty rates constant) and an “interaction” term that can be interpreted as a measure of the correlation between the population shifts and the intra-group changes in poverty. Under both methods, a counterfactual scenario is used, and estimates are made as to 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, be it a distribution-neutral consumption growth, the amount of poverty reduction that took place within a sector (as if the distribution across sectors had not changed), or the amount of poverty reduction that took place as a result of people moving between groups. Source: World Bank’s Poverty Handbook Only the Northern region experienced a substantial positive effect from the increase in mean household consumption. The Central and Eastern regions experienced both a negative effect distribution effect and a negative growth effect. The Western region, which also increased its poverty rate, was mostly affected by an increase in inequality, offset in about half by higher consumption. For the Northern region, the growth effect alone would have reduced poverty by 13.18 percentage points between 2012 and 2016, 23 bringing the poverty headcount down to 30.5 percent rather than the actual 32.5 percent. Inequality, however, prevented this result from happening. c) The role of poverty reduction within geographical areas versus internal migration As expected, the increase in poverty among rural households accounts for almost all the poverty increase observed at the national level. A decomposition exercise (between urban and rural households) confirms that the rise in poverty at the national level has been mainly driven by the increase in poverty amongst rural households. It accounts for around 115 percent of the overall increase in national poverty during 2012 and 2016 and was slightly counterbalanced by the effect of rural to urban migration in reducing poverty. In other words, if the rural population share had not declined between 2012 and 2016, poverty would have increased by slightly more (by around 0.23 percentage points more, see Table 2a). In terms of the regional contribution to the increase in poverty, the decline of the poverty headcount in the North and inter-regional migration helped to off-set the overall poverty increase. Every region contributed differently to the changes observed in the poverty headcount in Uganda, and such regional contributions are related to both changes in the regional economic performance and to movements across regions. The Eastern province, for which poverty increased considerably from 24.5 percent in 2012 to 35.7 percent in 2016, account for the largest share of the increase12, followed by the Central and Western regions, which also experienced poverty increases (Table 2b). Had it not been for the decline in poverty observed for the Northern region and for the intra-regional migration, the increase in poverty derived from these three regions would have amounted to almost 5 percentage points as opposed to the actual 1.7 percentage points increase. d) Structural pattern of poverty changes: role of weather shocks Households engaged in the agricultural sector account for the majority of the poverty increase between 2012 and 2016, which suggests that the drought is the main factor behind the poverty increase. When looking at the contribution of different sectors to the change in poverty, the focus is placed to the sector of employment of the household head. Households engaged in the agricultural sector explain the large majority of the overall increase in the national poverty rate between 2012 and 2016 (Table 2c). This is not a surprising result taking into account the drought conditions that affected this sector in 2016 and 2017, as discussed in Section I.13 While households engaged in the industrial sector and “other sectors” (different from agricultural, industrial and services) also help explain the overall increase in poverty, the magnitude of their contribution is considerably smaller, close to 0.3 percentage points. 12 As a matter of fact, all else equal, the increase in poverty observed in the Eastern region would have increased the national poverty rate by 3.3 percentage points. 13 All else equal, the increase attributed to households whose households head was employed in agricultural adds to a 3.7 percentage point increase in the national poverty rate. 24 On the contrary, the services sectors and the population shifts amongst sectors were factors that contributed to reduce poverty during this period. As in many other African countries, such as Kenya and Mozambique, the services sector has increasingly allowed Ugandan households to diversify their economic activities outside agriculture. In addition, this sector has grown faster than the economy as a while since 2015. Thus, this sector was a poverty reducing factor during the period 2012-2016 albeit the magnitude was small: households working in this sector contributed with a 0.2 percentage-point reduction. In addition, intra-sector shifts, likely from agriculture to the services sector, was also a source of poverty reduction during this period, contributing (all else equal) with a decline of around 1.1 percentage points (Table 2c). Table 2. Decomposing poverty change in Uganda 2012-2016 a. Urban/rural decomposition of poverty change b. Regional decomposition of poverty change Population Absolute Population Absolute Percentage Percentage Component share in change Component share in change change change 2012 (ppts) 2012 (ppts) Total Intra-sectoral Total Intra-sectoral effect 1.96 116.13 effect 2.63 156.12 Central region 25.79 1.05 62.25 Rural 77.39 1.94 115.08 Eastern region 29.69 3.30 196.30 Urban 22.61 0.02 1.05 Northern Region 21.05 -2.35 -139.58 Population-shift Western region 23.45 0.63 37.14 effect -0.23 -13.67 Population-shift effect -0.70 -41.53 Interaction effect -0.04 -2.46 Interaction effect -0.25 -14.59 Total change 1.68 100 Total change 1.68 100 c. Sectoral decomposition of poverty change Population Absolute Percentage Component share in change change 2012 (ppts) Total Intra-sectoral effect Agriculture 66.54 3.13 172.36 Industry 8.92 0.08 4.63 Services 17.00 -0.21 -11.47 Other 7.53 0.26 14.47 Population-shift effect -1.10 -60.69 Interaction effect -0.35 -19.30 Total change 1.68 100 Source: Own calculations based on UNHS 2016 e) Decomposing the poverty decline in the Northern region Contrary to what occurred in the rest of Uganda, poverty in the North declined from 43.7 percent of the population in 2012 to 32.5 percent in 2016. For this region, the Datt-Ravallion decomposition between the “growth” and the “distribution” components shows that the growth component (or the increase in the mean average household consumption) was large enough to off-set the adverse effect of the distribution component (as shown in Table 1). Moreover, as was the case at the national level, the 25 decomposition exercises for the Northern region suggest that changes are driven mainly by households residing in rural areas, which in 2016 accounted for about 80 percent of the change (Figure 10a). Also, all economic sectors seemed to have contributed to poverty reduction, including the agricultural sector. As a matter of fact, the contribution of this sector was a little over 50 percent (see Figure 10b) related to the favorable weather conditions experienced by this region, as shown in Figure 5a. Additional transfers in 2016, which includes remittances and social protection transfers (the data does not allow to separate these), may have also contributed to the poverty reduction. Another one of the factors that seemed to contribute to the poverty reduction in the Northern region was the increase in transfers received by households. Between 2012 and 2016, the proportion of households receiving transfers (which includes remittances, social security and assistance, and family allowances) increased from 34 to 55 percent (Figure 10c).14 The Eastern region saw a less noticeable rise of 7 percentage points, while the proportion declined for the other two regions. More importantly, northern households reported a twofold increase in the value of transfers, as reported by heads of household (Figure 10d). This is a remarkable increase compared to the other regions, for the Eastern region, the increase was only 8 percent, while households in the Central and Western regions experienced a decline in the income from transfers by 37 and 43 percent, respectively (Figure 10d). As a result, the proportion of income from transfers in the North rose from only 5.7 percent of the total income in 2012 to 12.7 percent in 2016. Figure 10. Decomposing poverty change in Uganda 2012-2016 a. Northern region – urban/rural decomposition of poverty b. Northern region – Sectoral decomposition of poverty change change Population- Interaction Population-shift shift effect, Interaction effect, -0.6 effect, 3.4 effect, -1.9 7.2 Other, 3.8 Urban, Services, 24.3 17.7 Agriculture , 55.8 Rural, 78.2 Industry, 12.1 14 For the bottom 40 percent of households in the Northern region, the increase was of 20 percentage points. 26 c. Proportion of households receiving transfers increased in d. The value of transfers also increased for the North the North $1,200,000 80% Proportion of households 65% $1,000,000 USh 2016 prices 60% 55% 49% 50% $800,000 43% 42% 45% 46% 41% 36% 36% $600,000 40% 34% $400,000 20% $200,000 $- 0% Central Eastern Northern Western Central Eastern Northern Western 2009 2012 2016 2009 2012 2016 Source: Own calculations based on UNHS 2016 V. Evolution of non-monetary poverty indicators Contrary to what was observed for monetary poverty, Uganda experienced progress in most non- monetary wellbeing indicators during the period 2012 to 2016. Net enrollment in secondary level increased around 8 percentage points and almost reached 30 percent in 2016. The prevalence of malnutrition among children under 5 years of age declined, maternal and child mortality rates continued to fall, and life expectancy continued to improve. The notable exception is the fall in the net enrollment rates in primary education. While non-monetary wellbeing indicators are negatively correlated with poverty (less-poor countries tend to have better non-monetary indicators and, over time, for a given country non-monetary indicators tend to improve as poverty declines), it is possible for them to move in different directions. Non-monetary indicators (such as life expectancy and mortality rates) tend to be more stable than monetary poverty. This does not imply that shocks, such as droughts, have no impact on these indicators15 but it may take longer for the effects to be noticeable, and if the shock is temporary, these effects may go unnoticed. Moreover, in the case of Uganda, the data collection for the nutrition and health indicators took place at least half a year before that for poverty and thus, may not fully reflect the extent of the shock. Finally, as discussed previously, poverty seems to have increased in 2016 because households close to the poverty line slipped into poverty (per capita consumption increased for the ultra- poor), and these households are likely more capable than those at the very end of the distribution to cope with it. Enrollment in primary school decreased moderately between 2012 and 2016 but it remains close to the sub-Saharan African average. Net primary school enrollment rates16 decreased slightly from 82.3 in 2012 15 In fact, several studies in Africa (and in Uganda) have shown that droughts have adverse consequences for education and health outcomes. For example, Hyland and Russ (2019) for 19 sub-Saharan African countries, Baez et al. (2018) for Mozambique, Hill and Mejia-Mantilla (2017) for Uganda, and Randell and Gray (2016) for Ethiopia. 16 This is the ratio of children of official school age who are enrolled in primary school to the population of the corresponding official school age for primary school, which is 6-12 years in Uganda. 27 to 79.5 in 2016, continuing a declining trend observed since 2005 (Figure 11a). Despite the drop, the rate in Uganda lies slightly above the average for Sub-Saharan Africa (at 78.8 percent). The fall occurred for both boys and girls, and the rate for females remains slightly higher than that for boys. Interestingly, female net enrollment is higher in Uganda compared to the region (80.3 versus 76.4 percent), while male enrollment is slightly lower (78.7 versus 81 percent). The trends in net primary enrollment are probably linked to the recent decline in primary education expenditure per student (from 7.1 percent of GDP per capita in 2011 to around 5.5 in 2016). Net secondary school enrollment rates17 improved between 2012 and 2016, although they remain lower than in the rest of region. The net secondary school enrollment rates increased from 21.7 to 27.8 percent between 2012 and 2016 (Figure 11b). The progress is observed for both young men and women, albeit enrollment is still lower for women by around 4 percentage points. Despite the progress, these rates remain relatively low when compared to the rest of Africa. The average net secondary enrollment rate for sub-Saharan African countries is 34.5 percent, around 7 percentage points above than in Uganda. This is a concern as low secondary enrollment results in low human capital, and for women it is highly correlated with child marriage and early pregnancy. Maternal mortality rates declined, and life expectancy increased, by almost two years, during the period of interest. Maternal mortality in Uganda fell from 395 deaths per 100,000 live births in 2012 to 343 in 2016 (Figure 11c). This rate is well below the average for sub-Saharan African countries, which stands at 547 maternal deaths in 2016. However, the rate is not lower than what is expected for the level of poverty of Uganda (Figure 11d)18. There was also a progress in terms of life expectancy, increasing by 1.6 years since 2012 to reach 60 years in 2016 (Figure 11e). This increase is observed for both men and women, even though women tend to live 5 years longer than men. These figures are consistent with what is observed in other countries in sub-Saharan Africa, where overall life expectancy is 60.4 years, but slightly below what is expected for the level of poverty of Uganda (Figure 11f). Uganda continued to make progress in reducing childhood mortality, including neonatal, infant and under-five mortality. Neonatal, infant and under-five mortality rates have decreased substantially between 2005 and 2012 (Figure 12a). This trend continued throughout 2012 to 2016: neonatal mortality reached 20.9 (per 1,000 live births) in 2016, infant mortality dropped to 37 (per 1,000 live births), and under-five mortality rate stood at 51.6 (per 1,000 live births). In this regard, Uganda compares positively to the rest sub-Saharan Africa, whose average rates were 27.8, 53,1 and 78.3 respectively for 2016. This is consistent with the fact that the under-five mortality rate in Uganda is below what is expected for its level of poverty (at the international poverty line) as seen in Figure 12b. 17 This is the ratio of children of official school age who are enrolled in secondary school to the population of the corresponding official school age for secondary school, which is 12-18 years in Uganda. 18 The graph shows a scatter plot of the maternity rates versus international poverty rates for all countries for which there is information in the last 5 years. The green line corresponds to the linear fitted trend. 28 Figure 11. Non-monetary poverty indicators a. Net Primary school enrollment has recently declined b. Net secondary school enrollment improved for women and men 100 40 36.4 34.5 34.3 32.3 32.6 83.2 83.2 82.4 29.0 30.3 82.3 83.5 81.0 27.8 80.3 81.1 78.8 80.5 30 26.7 79.5 78.7 78.1 76.4 78.2 75.5 75.7 21.7 23.0 72.8 20.6 75 20 Rate, % Rate, % 10 50 0 All Female Male All Female Male All Female Male All Female Male Uganda SSA Uganda SSA 2009 2012 2016 2012 2016 c. Maternal mortality rates are below the average for d. Maternal mortality is what is expected for the poverty sub-Saharan Africa level of Uganda Deaths per 1,000 live births 800 717 642 587 600 547 504 432 395 400 343 200 0 Maternal mortality rate Maternal mortality rate 2005 2009 2012 2016 e. Life expectancy is higher for both women and men f. Life expectancy is slightly low for the level of poverty of Uganda 64 62.1 62.1 62 59.9 60.3 60.4 60 59.9 58.3 58.3 58.7 57.7 5… 57.6 58 56.4 56.7 Years 56.3 56.2 56 54.7 54.8 54 52 50 All Male Female All Male Female Uganda SSA 2009 2012 2016 Source: a and b, Uganda Bureau of Statistics; c to f, Word Bank WDI indicators 29 Figure 12. Child mortality and nutrition indicators a. Childhood mortality rates are below the average for b. Under-five mortality is low for the poverty level of sub-Saharan Africa Uganda 140 126.1 120 108.8 105.4 Deaths per 1,000 live births 100 92.0 83.4 78.3 78.3 80 68.0 67.3 67.5 60.6 53.9 51.6 36.2 53.1 60 26.3 45.1 25.1 32.9 37.0 30.6 40 27.8 23.7 20.9 20 0 Neonatal Infant Under five Neonatal Infant Under five Uganda SSA 2005 2009 2012 2016 c. Low prevalence of wasting (low weight for height) d. Wasting is low for what it is expected for the level of and high prevalence of stunting (low height for age) poverty of Uganda 45 38.7 40 34.2 35 % of children under 5 28.9 30 25 20 15 10 6.3 4.3 3.6 5 0 2006 2012 2016 Wasting Stunting e. Stunting is also low for the level of poverty of Uganda Source: Author’s calculations based on Word Bank WDI indicators. 30 Incidence of stunting and wasting dropped between 2012 and 2016, but the former remains a challenge. The incidence of wasting, which measures low weight for height, stands at 3.6 percent in 2016, decreasing from 4.3 percent in 2012 (Figure 12c). In addition, the rate of wasting is also low given the country’s international poverty rate (Figure 12d). This is encouraging, as wasting is a strong predictor of mortality among children under five. However, stunting, defined as low height for age and an indicator of chronic malnutrition for children under five, remains prevalent. In 2016, close to 28.9 percent of children presented this condition (Figure 12c), albeit this level is lower than what the poverty rate would have predicted (Figure 12e). Given the negative long-term consequences of malnutrition on education and labor market outcomes, more attention should be placed on the underlying causes. VI. What factors affect households’ consumption in Uganda? This section reports the results from multi-variate regression analyses to determine what characteristics (demographic, socio-economic, geographical) are associated with higher levels of consumption per capita The advantage of this method is that it allows to analyze the association between consumption and the different household characteristics, assuming the rest of the characteristics as fixed.19 We explore if these characteristics (demographic, socio-economic -related to the household head, geographical) are positively or negatively correlated with household consumption and thus, with poverty (see Table 3). More numerous households and those with a larger share of children and seniors tend to have a lower per capita consumption. The empirical analysis suggests that in Uganda, an additional household member is associated with a decrease in per capita consumption of around 7 percent (see Table 3).20 Similarly, the proportion of dependents (children aged 0-14 and adults 65+) is also negatively associated with consumption, given that these individuals are less likely to engage in economic activities that increase household income. On the contrary, having more women in the household seems to increase consumption, although the magnitude is rather small. Households headed by older people or women tend to consume more, while those headed by divorced or widowed consume less. On average, per capita consumption is higher for households with older and female household heads: a 10-year increase in age is associated with a 6.2 percent increase in consumption in the first case and a 10 percent increase in the second (see Table 3). On the contrary, households with divorced (or separated) or widowed heads have lower consumption levels and thus, are more vulnerable of falling into poverty. 19 We use OLS regressions where the dependent variable is the log of annual total per adult equivalent household consumption (including food and non-food consumption). Fixed effects for region, sub-region and strata are added in each separate regression. Similar results were obtained using a linear probability model where the dependent variable is the household poverty status. 20 In the regression that controls for differences across regions, the effect is calculated as: [exp (-0.068)-1] * 100 = -6.6 (see regression coefficient associated to household size in column 1 in Table 3). 31 Table 3. What factors are associated with higher per capita consumption (1) (2) (3) OLS Dependent variable: Log of real per adult equivalent consumption Demographic Household size -0.068*** -0.071*** -0.073*** (0.005) (0.005) (0.005) Share of women in the household 0.001** 0.001* 0.001* (0.000) (0.000) (0.000) Share of household members 0 to 14 years old -0.004*** -0.003*** -0.003*** (0.000) (0.000) (0.000) Share of household members 65+ years old -0.002*** -0.002** -0.001** (0.001) (0.001) (0.001) Household head characteristics Gender: Female (Ref. Male) 0.006*** 0.006*** 0.006*** (0.001) (0.001) (0.001) Age in years 0.096*** 0.099*** 0.098*** (0.027) (0.027) (0.026) Marital status (Ref. Married): Never married 0.004 -0.004 -0.012 (0.037) (0.036) (0.037) Divorced/Separated -0.068** -0.075** -0.081*** (0.032) (0.032) (0.031) Widowed -0.079** -0.095** -0.106*** (0.037) (0.037) (0.035) Industry classification in primary job (Ref. Agriculture): Industry 0.111*** 0.124*** 0.112*** (0.033) (0.033) (0.033) Services 0.321*** 0.318*** 0.306*** (0.024) (0.024) (0.024) Other 0.191*** 0.189*** 0.182*** (0.037) (0.036) (0.035) Highest education level attained (Ref. None): Primary incomplete 0.194*** 0.171*** 0.171*** (0.029) (0.029) (0.028) Primary complete 0.388*** 0.345*** 0.344*** (0.033) (0.033) (0.033) Secondary incomplete 0.433*** 0.402*** 0.396*** (0.034) (0.033) (0.033) Secondary complete 0.556*** 0.523*** 0.519*** (0.063) (0.062) (0.060) Post-secondary but not university 0.744*** 0.698*** 0.674*** (0.056) (0.056) (0.054) University 1.118*** 1.093*** 1.051*** (0.073) (0.072) (0.073) Assets/Geographic Area of residence (Ref. Rural): Urban 0.186*** 0.149*** 0.146*** (0.022) (0.022) (0.022) Ownership of land 0.065*** 0.058*** 0.065*** (0.018) (0.018) (0.019) Constant 13.579*** 13.709*** 13.830*** (0.053) (0.067) (0.229) Region Fixed Effects Yes Strata Fixed Effects Yes Sub-region Fixed Effects Yes Observations 6,746 6,746 6,746 R-squared 0.492 0.514 0.542 Robust standard errors in parentheses Coefficient significant at *** 1%, ** 5%, * 10% confidence level. Source: Own calculations based on UNHS 2016/17 32 As expected, households whose head is engaged in sectors other than agriculture or has a higher educational level, have higher consumption. The main sector of occupation provides an indication of the main source of income of the household. Consumption tends to be higher for households engaged in the industry/manufacturing, services or other sectors compared to those engaged in agriculture. The association is larger for the services sector, with an estimated 36-38 percent increase in real consumption relative to the agriculture sector. Similarly, the education gradient is positive and large. Relative to a household whose head has no education, a household whose head has incomplete primary consumes close to 20 percent more. The education premium increased with the level of education and completed secondary education is associated with a 68-74 percent increase in consumption. Finally, living in urban areas and owning land increases consumption. Land ownership is associated with 6.7 percent increase in consumption, relative to non-owners, while urban households consume on average, 16 to 20 percent more than rural households, consistent with the higher poverty headcount rates among the latter. VII. Who are the poor in Uganda? Raising the living standards and enhancing the economic opportunities of the poor requires identifying what factors are constraining their ability to engage productively in the economy. More specifically, it is useful to compare the poor and the non-poor along different dimensions, such as demographic characteristics, human capital accumulation, access to basic services and other socio-economic traits, and identify which particular areas requires attention. Similarly, comparing the different regions along these dimensions can help explain, and help tackle, the spatial gap observed in the poverty. Households living in poverty are larger and have higher dependency ratios. Consistent with the regression estimates presented in Table 3, poor households have around 1.3 more members than non- poor households and a larger dependency ratio (see Table 4).21 Interestingly, the latter is associated with a higher share children (ages 0-14) rather than a higher share of elderly (ages 65 and up). As for the spatial differences, the Eastern region -which has the highest poverty rate - has both the largest household size, at 6.5 members, and the largest dependency ratio, at 55 percent (see Table 5). Poor households have lower levels of human capital, measured by adult literacy rate and educational attainment of the household head. The adult literacy rate, which measures the percentage of household members ages 15 and above who can both read and write, is larger among the non-poor (see Table 4). The literacy rate among the poor was 57.3 percent in 2016, about 24 percentage points lower than for the non-poor, limiting the income generating potential of the poor. At the regional level, the Northern region lags considerably behind the rest of the country. While the national literacy rate is 73.4 percent, the figure for the Northern region was 64.7 percent (Table 5). Similarly, the heads of poor households have very low levels of education—only 24.6 percent reported a level of completed primary and higher 21This is the ratio of dependents--people younger than 15 or older than 64--to the working-age population--those ages 15-64 within the household. 33 (compared to almost 52 percent for the non-poor). When looking at difference by regions, a marked contrast between the Central region and the rest is observed: about 62 percent of the heads in this region have primary education or more, compared to the 40-45 percent observed for the rest (see Table 5). Living conditions and ownership of basic assets are limited for the poor. By 2016, almost 30.9 of poor households live in dwellings with good walls22 and good floors23, while that was the case for 57.5 percent and 45 percent of non-poor households respectively (Table 4). Asset ownership is also low among poor households, except for house and land ownership: almost 92 percent of poor households own a house, and 48 percent of poor households own land. This is likely linked to the urban/rural poverty divide observed in the country. Interestingly, cell ownership is significantly low for poor households (52.6 percent versus 81.8 percent for non-poor households) and Northern households, which may curb the use of mobile technologies, such as mobile money. Given that this has been one of the crucial factors for poor households in Kenya to diversify their economic activities outside the agricultural sector, contributing to reduce poverty, this represents an important area for action (World Bank 2018). Access to improved sanitation and electricity remains low among the poor. Overall access to improved sanitation24 is low for Uganda, at 40.6 percent of households. Moreover, access is considerably lower for poor households (18.1 percent) compared to non-poor households (46.8 percent, as shown in Table 4). Electricity coverage is even more limited and only 1 in 4 Ugandan households has access.25 It also varies substantially by poverty status: 12.9 percent of poor households report having access to electricity, in contrast with 45.8 percent of the non-poor. On the contrary, both poor and non-poor households enjoy moderately high access to improved water 26, an area where Uganda has made important progress since 2000. About 77 percent of households report having access to improved water sources and there are no major differences between the poor and the non-poor. Also, contrary to the other services, access to improved water sources is higher in the Eastern and Northern regions compared to the other two regions (see Table 5). Given that the majority of poor reside in rural areas, the majority of poor household heads are engaged in the agricultural sector. As mentioned in previous sections, nearly 9 out of 10 of poor households live in rural areas compared to just seven in ten non-poor households (Table 4). Thus, the majority of household heads from poor households are employed in the agricultural sector. More specifically, about 81 percent of heads of poor households report agriculture as the sector of their primary occupation, while that is only the case for 54 percent of non-poor heads. These results suggest that accessing economic 22 These are walls made with bricks, wood panels, or concrete/cement/stone. 23 These are floors made with bricks, wood planks, polished wood/tiles or cement. 24 The World Health Organization (WHO/UNICEF Joint Monitoring Programme) de fines ‘improved’ sanitation as using facilities such as sewer connections, septic system connections, pour-flush latrines, ventilated improved pit latrines and pit latrines with a slab or covered pit. 25 Electricity sources include national grid, solar, personal generator or community/thermal plant. 26 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). 34 activities outside of agriculture seems to enhance household income. This is also exemplified by the fact that in the Central region (region with the lowest poverty rate at 8.8 percent), only 38 percent of household heads work in agriculture. In the other three regions, between 63 and 71 percent of household heads work in agriculture. Table 4 Poor versus non-poor (and all population) socio-economic conditions Difference Socio-economic condition Average for all Average for Poor Average for Non-Poor (Poor - Non-Poor) Demographic Dependency ratio 51.9% 58.7% 50.1% 8.7 *** Children share 48.7% 55.9% 46.7% 9.2 *** Elderly share 3.2% 2.9% 3.3% -0.5 ** Household size 5.9 6.9 5.6 1.3 *** Literacy rate 73.4% 57.3% 77.8% -20.5 *** Dwelling Good roofs 0.5% 0.1% 0.5% -0.4 *** Good walls 51.8% 30.9% 57.5% -26.6 *** Good floors 37.8% 11.3% 45.0% -33.7 *** Assets and services Owns house 80.8% 91.7% 77.8% 13.9 *** Owns lands 42.2% 47.7% 40.4% 7.2 *** Owns radio 46.9% 27.3% 52.2% -24.9 *** Owns TV 16.1% 0.4% 20.4% -20.0 *** Owns cell 75.6% 52.6% 81.8% -29.2 *** Owns fridge 4.9% 0.0% 6.2% -6.2 *** Owns computer 3.0% 0.0% 3.8% -3.8 *** Owns stove 1.2% 0.0% 1.6% -1.6 *** Owns motorcycle 8.6% 1.8% 10.5% -8.7 *** Access to electricity 38.8% 12.9% 45.8% -0.3 *** Access to improved sanitation 40.6% 18.1% 46.8% -28.6 *** Access to improved water sources 77.1% 78.1% 76.8% 1.3 Household head characteristics Age of head of household (years) 43.41 44.73 43.05 1.7 *** Household head is male 73.6% 73.4% 73.7% -0.3 Marital status of head of household: Married 78.6% 80.2% 78.2% 2.01 ** Never Married 3.5% 1.6% 4.0% -2.40 *** Divorced/Separated 7.5% 6.8% 7.7% -0.91 * Widowed 10.5% 11.5% 10.2% 1.30 ** Education level of head of household: No education 12.8% 21.6% 10.3% 11.30 *** Primary Incomplete 40.7% 53.7% 37.1% 16.7 *** Primary Complete 18.5% 12.4% 20.2% -7.7 *** Secondary Incomplete 18.7% 11.1% 20.8% -9.7 *** Secondary Complete 2.2% 0.4% 2.7% -2.4 *** Post-secondary, no university 4.4% 0.7% 5.5% -4.7 *** University 2.7% 0.0% 3.4% -3.4 *** Industry of head of household: Agriculture 59.3% 80.9% 53.6% 27.4 *** Industry 9.4% 6.8% 10.1% -3.3 *** Services 22.3% 8.4% 26.0% -17.6 *** Other 9.0% 3.9% 10.3% -6.4 *** Geographic Residence in rural areas 75.7% 89.3% 72.0% 17.4 *** Source: Own calculations based on UNHS 2016 Note: Reported are the coefficients from a linear probability model where the dependent variable is binary (poor=1, otherwise=0). The stars denote the significance values from standard unconditional t-tests of differences between means. Significant coefficients at *** 99%, ** 95%, *90%. 35 Table 5 Socio-economic conditions by region Socio-economic condition Central region Eastern region Northern region Western region Demographic Dependency ratio 47.8% 54.9% 54.3% 51.2% Children share 45.5% 51.1% 51.2% 47.6% Elderly share 2.3% 3.9% 3.1% 3.6% Household size 5.5 6.5 5.9 5.8 Literacy rate 86.8% 70.7% 64.7% 72.5% Dwelling Good roofs 1.2% 0.1% 0.2% 0.4% Good walls 86.0% 50.4% 25.3% 37.1% Good floors 72.4% 26.7% 15.9% 27.9% Assets and services Owns house 63.4% 88.6% 87.7% 85.8% Owns lands 27.5% 40.3% 68.3% 37.7% Owns radio 53.3% 39.0% 38.0% 57.4% Owns TV 38.7% 6.1% 4.2% 11.3% Owns cell 91.2% 71.4% 56.8% 81.4% Owns fridge 13.9% 1.2% 1.1% 1.7% Owns computer 6.2% 0.9% 2.6% 2.0% Owns stove 3.3% 0.2% 0.2% 0.7% Owns motorcycle 11.5% 4.3% 7.2% 11.1% Access to electricity 62.3% 25.5% 19.0% 42.6% Access to improved sanitation 76.4% 23.8% 26.0% 29.5% Access to improved water sources 75.0% 88.8% 81.9% 61.3% Household head characteristics Age of head of household (years) 41.16 43.29 41.53 42.82 Household head is male 73.3% 78.1% 71.2% 76.2% Marital status of head of household: Married 75.8% 83.8% 81.1% 80.0% Never Married 5.0% 1.8% 2.1% 3.7% Divorced/Separated 11.2% 6.8% 5.5% 6.3% Widowed 8.0% 7.6% 11.3% 10.0% Education level of head of household: No education 7.9% 7.9% 15.9% 14.6% Primary Incomplete 29.7% 50.1% 43.7% 41.3% Primary Complete 18.2% 17.1% 21.1% 19.9% Secondary Incomplete 26.9% 19.0% 14.0% 15.4% Secondary Complete 5.0% 1.5% 0.9% 1.6% Post-secondary, no university 6.3% 3.2% 3.2% 5.1% University 6.1% 1.1% 1.2% 2.1% Industry of head of household: Agriculture 38.2% 69.1% 71.0% 63.7% Industry 13.1% 6.5% 8.1% 9.4% Services 35.9% 16.7% 13.8% 19.9% Other 12.8% 7.7% 7.1% 7.1% Geographic Residence in rural areas 55.0% 86.8% 85.3% 81.8% Source: Own calculations based on UNHS 2016 Note: Reported are the coefficients from a linear probability model where the dependent variable is binary (poor=1, otherwise=0). The stars denote the significance values from standard unconditional t-tests of differences between means. Significant coefficients at *** 99%, ** 95%, *90%. 36 VIII. Conclusions Between 2012 and 2016 the poverty rate in Uganda increased moderately, reaching 21.4 percent in 2016, a 1.7 percentage point increase that resulted in around 1.4 million Ugandans slipping into poverty. This is the result of the overall economic slowdown observed since 2012, coupled with a severe drought that affected most of the country for the better part of 2016 and 2017. The increase was mainly explained by the increase in rural poverty, which went up from 22.8 to 25.3 over the four-year period. Even though the poverty rate remained stagnant around 9 percent in urban Uganda, the number of urban poor increased explained by the rise in the urban population, now at 24.3 percent of the population. These trends reflect the high degree of vulnerability of households in the face of adverse shocks, such as weather shocks. Poverty increased in all regions of Uganda, except for the Northern region, but the poor remain concentrated in the Eastern and Northern regions. The increase in poverty incidence between 2012 and 2016 was more pronounced in the Eastern region, which currently has the highest poverty headcount rate at 35.7 percent. While poverty declined in the Northern region, traditionally the poorest of the four regions, at 32.5 percent in 2016 is still significantly higher compared to the national level. More importantly, 3 out of 4 of those who live in the poverty reside in the Northern and Eastern regions, illustrating that poverty remains geographically concentrated. Thus, public policies with a regional focus to increase the livings standards and enhance the resilience of households, like those recently introduced by the Government of Uganda (GoU) in the Northern region, should be supported and expanded. Consumption levels of the ultra-poor (those below the 10th percentile) increased between 2012 and 2016. Consumption growth for the poorest 10 percent of households was positive and higher than the average growth for the entire distribution. This fact helps explain why the depth of poverty remained largely unchanged and why the severity of poverty (which gives more weight to poor households furthest from the poverty line) fell slightly at the same time that poverty was increasing. It also highlights the fact that the increase in poverty was the result of the decline in the living conditions of the households bunched just below the poverty line and will likely not be permanent. The bottom 40 of the distribution benefitted less from the economic progress, resulting in a rise in inequality. The annualized growth rate of consumption for the bottom 40 percent of the population was close to zero (0.05 percent), a reversal of what Uganda experienced between 2009 and 2012. Thus, inequality increased, as confirmed by several indicators that measure how evenly is consumption distributed along the distribution. The Gini coefficient went up from 0.40 to 0.42 between 2012 and 2016, while the Theil index increased from 0.29 to 0.34. Moreover, the 90/10 and 75/25 ratios show that inequality rose more at the extremes rather than in the middle of the consumption distribution. Households in rural areas and those engaged in agriculture account for the bulk of the poverty increase. Between 2012 and 2016, close to 80 percent of the total increase is accounted for by households residing in rural areas, consistent with the poverty trends. Similarly, households engaged in the agricultural sector 37 account for almost 65 percent of the rise. While in some instances, weather shocks benefit net food sellers (as opposed to net food buyers), this is not the case in Uganda given that the majority of rural households consume most of what they produce. In the case of the poverty decline of the Northern region, two factors seem to have contributed to the decline: favorable weather and an increase in transfers received by households. Contrary to what was observed for monetary poverty, Uganda experienced progress in most non- monetary wellbeing indicators during the period 2012 to 2016. Net enrollment in secondary level increased from 21.7 to 27.8 percent between 2012 and 2016 and is slightly higher for girls compared to boys. The incidence of wasting and stunting (for children under 5 years old) also declined during this period, reaching 4.3 percent and 28.9 percent in 2016. Similarly, maternal and child mortality rates continued to decline, and life expectancy continued to improve. Despite the progress, may challenges remain. Secondary enrollment remains lower than in the rest of Sub-Saharan Africa and primary enrollment rates has been declining since 2009. Also, the fact that nearly one in three children under five is chronically malnourished is a major concern, given the negative consequences of malnutrition of human capital accumulation and adult productivity. Several factors continue to limit the income generating capacity of those living in poverty, including restricted access to basic services. Poor households are larger and have higher dependency ratios, mainly linked to a larger share of children rather than by a larger share of elders. Moreover, they present lower education levels, worse living conditions, and own less productive assets. Importantly, only 1 in 5 poor have access to improved sanitation and only 1 in 10 have access to electricity. All the above restricts the ability of poor households to participate in the economy and enhance their income. Going forward Considering the acceleration of the economy and the good performance of the agricultural sector in 2018, it is likely that Uganda’s poverty levels are currently back to those observed in 2012 or even slightly lower. Nonetheless, the events of the period 2012 to 2016 are an important reminder of several key lessons to ensure that moving forward Uganda remains in a poverty reducing path. • Reducing the vulnerability of rural households to adverse shocks will be fundamental to ensure sustained progress against poverty reduction. This requires investments in modernizing agricultural production and decreasing its reliance on rainfall, through drought resistant seeds and investments in irrigation technologies among other measures, as well as increasing their participation in higher value-added agricultural activities. It also requires a strong system of agricultural extension services to support small scale farmers across all different regions of Uganda. • In addition, more should be done to enable households to increase their participation in non- agricultural economic activities. Increasing the demand for household labor outside this sector and in more productive sectors is a key step forward. This requires important investments to 38 improve the education outcomes and skills of the population, that should be complemented with policies focused on firm and job creation in high productivity activities. • The expansion of existing safety net programs (currently concentrated in the Northern region and characterized by a low coverage) or the introduction of new ones can help reduce the negative effects of adverse shocks to vulnerable households. This should be done in a way that does not compromise the macroeconomic stability of the country and that takes into consideration its limited fiscal space. • Ensuring the access of all households to basic services, such as secondary education, sanitation and electricity, is an important component of improving the wellbeing of the Ugandan population. This involves an effective service delivery system at the local level (in education, health and extension services for example), important infrastructure investments at the central level (in electricity and sanitation for example) and a strong coordination between them. Without this, it is hard to ensure sustained progress in poverty reduction and address regional disparities. 39 References Appleton et al., 1999. “Changes in Poverty in Uganda, 1992-1997”. Center for the Study of African Economies (CSAE), University of Oxford, WPS/ 99.22. Baez Ramirez, J.; Caruso, G., & Niu, C. 2018. “Extreme Weather and Poverty Risk: Evidence from Multiple Shocks in Mozambique.” Policy Research Working Paper, No. WPS 8667. Washington, D.C.: World Bank Group. 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. FEWSNET 2017. “Drought Conditions Exacerbated by Above-Average Surface Temperatures Over Eastern Horn”. East Africa Seasonal Monitor. Foster, J., Suman, S., Lokshin, M., and Sajala, Z. 2013. “A Unified Approach to Measuring Poverty and Inequality: Theory and Practice. Washington, DC: The World Bank. Hill, R., & Mejia-Mantilla, C. (2016). “With a Little Help: Shocks, Agricultural Income, and Welfare in Uganda.” World Bank Policy Research Papers, No. 7935. Washington, DC: The World Bank. Hyland, M., & Russ, J. (2019). “Water as Destiny–The Long-Term Impacts of Drought in Sub-Saharan Africa.” World Development, 115, 30-45. Randell, H., & Gray, C. (2016). “Climate Variability and Educational Attainment: Evidence from Rural Ethiopia.” Global Environmental Change, 41, 111-123. World Bank, 2016. “Farms, Cities and Good Fortune: Assessing Poverty Reduction in Uganda from 2006 To 2013.” Poverty and Equity practice, Africa region. Washington DC, World Bank Group. World Bank, 2018. “Reflecting on a Decade of Progress and the Road Ahead: Kenya Poverty and Gender Assessment 2015/16.” Poverty and Equity Practice, Africa region. Washington DC, World Bank Group. World Bank, 2019 - forthcoming. “Impact of Fiscal Policy on Poverty and Inequality in Uganda: Fiscal Incidence Analysis Using the UNHS 2016/17” Poverty and Equity Practice in collaboration with the Commitment to Equity Institute. Washington DC, World Bank Group. 40