Report No: AUS0001847 . Ethiopia Ethiopia programmatic knowlegde service Fiscal Incidence Analysis for Ethiopia . July 2020 . POV . . © 2017 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: “Mesfin, W. and Gao, J. 2020. Fiscal Incidence Analysis for Ethiopia. © 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. Fiscal Incidence Analysis for Ethiopia * Wondimagegn Mesfin Jia Gao June 2020 Abstract This study analyzes the distributional impacts of government taxes and public expenditure in Ethiopia using detailed and nationally representative survey data combined with administrative data. Using the Commitment to Equity (CEQ) methodology, the study finds that Ethiopia’s fiscal policy is (1) equalizing in reducing market income inequality by 2.1 percentage points and (2) poverty-reducing, reducing the national poverty headcount rate by 5.6 percentage points. On the revenue side, personal income taxes and indirect taxes are progressive and equalizing. Direct transfers through the Productive Safety Net Program (PSNP) and Humanitarian Food Aid (HFA) are progressive, pro-poor, equalizing, and poverty- reducing. Of the subsidies we consider, wheat and kerosene subsidies are progressive, but the electricity subsidy is unambiguously regressive, and no subsidies are pro-poor. Public spending on education and health is progressive and poverty-reducing but achieved little redistributive effects. While spending on primary education is pro-poor, spending on tertiary education is regressive. Going forward, broadening the income tax base would help to ensure pro-poor tax revenue generation and expand public spending, which would ultimately maximize the welfare impacts of the fiscal policy. Keywords: taxes and public expenditure, fiscal incidence, poverty, inequality, redistribution, CEQ, Ethiopia JEL Codes: H22, D31, I38 Acknowledgements: This report is produced under the close supervision and guidance of Tom Bundervoet (Senior Economist, EAEPV) and Christina Wieser (Senior Economist, EAEPV). The authors would also like to thank Gabriela Inchauste (Lead Economist, ELCPV) and Pierella Paci (Practice Manager, EAEPV) for their guidance. The report has benefited a lot from comments of three peer reviewers: Miguel Sanchez (Senior Economist, EAEM2), Emily Weedon (Senior Social Protection Economist, HAES1) and Rafika Chaouali (Lead Financial Management Specialist, EAEG1). The team has also received additional comments from the education and health team. We are specifically grateful to Fitsum Zewdu (ET Consultant, HAEE1), Kebede Feda (Senior Economist, HAEE1) and Paul Jacob Robyn (Senior Health Specialist, HAEH1) for their comments. The authors are grateful to Berhe Mekonnen (Economist, EAEPV), Zerihun Getachew (Research Analyst, EAEM2) and Obert Pimhidzai (Senior Economist, EAEPV) for their helpful comments and support. Any remaining errors are our own. Executive Summary Ethiopia has made remarkable progress in reducing poverty in recent years as a result of its robust economic growth. While the proportion of the population living below the national poverty line fell from 38.7 percent in 2004/05 to 23.5 percent in 2015/16, the proportion of the population living in poverty remains comparatively high. Ethiopia’s poverty rate is still high for a lower-middle income country, a group that Ethiopia aspires to join in the near future. The growth observed over the past decade has largely been due to large public sector infrastructure investments and the poverty elasticity of growth between 1997 and 2016 was low at -0.33. Although Ethiopia has the lowest level of income-inequality in Africa, income inequality has increased over time. Challenges also remain including the increasing rural-urban gap and lack of growth at the bottom income distribution. This begs the obvious question: what can be done to ensure economic growth is pro-poor in Ethiopia. Fiscal policy is shown to be key to reduce inequalities and fund public services which provide opportunities for all. This study assesses the distributional consequences of Ethiopia’s fiscal system—system of taxes and transfers—using data from the latest household consumption expenditure and welfare monitoring surveys combined with administrative data and the Commitment to Equity (CEQ) methodology. The study decomposes the contributions of individual tax and spending measures, and in doing so, the report provides a unified framework for measuring transfer and tax progressivity and incidence as well as their impacts on poverty and inequality. Overall, Ethiopia’s tax and transfer system reduces inequality with each intervention. The Gini index, a measure of inequality, of the market income—the income before the government has any influence on the income distribution through its tax and spending policies—is 0.348. It falls to 0.328 after direct taxes and transfers are paid but remains at 0.328 after indirect taxes and indirect subsidies. The Gini index further falls to 0.327 after public education and health spending is considered. The results of the Fiscal Incidence Analysis show that direct taxes (and transfers) played a substantial role in reducing income inequality, while indirect taxes and indirect subsidies do not. However, transfers on in-kind education and health achieved little redistributive effects. Overall, Ethiopia achieves little poverty reduction through direct taxes and direct transfers while indirect taxes and indirect subsidies increase poverty. This study finds that the combined system of taxes and transfers (but excluding in-kind education and health transfers) decreases headcount poverty rate by only 0.15 percentage points. Both direct and indirect taxes increased poverty, but by less than 1 percentage point. Direct transfers, indirect subsidies and in-kind transfers contribute to poverty reduction but with different marginal contributions. Direct taxes on personal incomes are progressive. More than seventy percent of the direct (personal income) tax incidence falls upon the richest 10 percent of the population. This could be because the poor are less likely to hold formal sector jobs in Ethiopia and earnings generated in the formal sector are most likely to be taxed. In addition, personal income tax in Ethiopia is subject to progressive rates, starting from a 10 percent marginal tax rate to 35 percent for higher income groups. At the same time, direct taxes have very limited effects on poverty as the tax burden is concentrated in the upper quintiles and few of these households are pushed below the poverty line. However, direct taxes reduce the Gini index by about 1.39 i percentage points and the greater share comes from personal income taxes. Agricultural income taxes and land use fees are regressive and their effects on poverty and inequality appear to be small. In part, this reflected the tendency for agricultural households to be poorer than non-agricultural households. It is also a result of the fact that the tax is largely levied according to landholding size, which might not be an accurate proxy for income. The direct transfer programs considered in this study—the Productive Safety Net Program (PSNP) and Humanitarian Food Aid (HFA)—are progressive and pro-poor. They also tend to contribute to reducing inequality and poverty reduction. Direct transfers reduce the headcount poverty rate by 1.1 percentage points and income inequality by 0.46 percentage points. The effective targeting performance of the two programs could have contributed to their progressivity and pro-poorness. Overall, close to 58 percent of the benefits are captured by the poorest 40 percent of the population. While their targeting performance contributes to positive welfare impacts on the poor, they reach only a small fraction of the population, limiting their effectiveness in reducing poverty and inequality. Indirect taxes—Value Added Tax (VAT) and excise taxes—are progressive but their impact on inequality is marginal. However, they tend to increase poverty by 0.6 percentage point. Compared to VAT, excise taxes appear to be more equalizing and poverty increasing. Excise taxes on beer and fuel are mildly progressive, but excise taxes on tobacco products and alcohol are ambiguous (neither progressive nor regressive). Wheat and kerosene subsidies are progressive but not pro-poor. Although positive, their effect on inequality and poverty are small. The electricity subsidy is regressive and unequalizing but contributes to poverty reduction. While the regressivity of electricity subsidy suggests that it benefits the rich more than the poor, its positive effect on poverty indicates that it helps to reduce urban poverty since electricity access and use is more common among urban households. Public education spending is progressive in relative terms but progressivity declines with increasing levels of education. Most importantly, public spending on primary education is progressive and pro-poor. This implies that a disproportionately larger share of children from poor households benefit from public primary education compared to children of higher income households where the uptake of private primary education is higher. Secondary education (including TVET) are progressive only in relative terms and public university (tertiary) education is regressive due to low levels of enrollment among the poor. The later result is also an indication that the in-kind benefits of spending at higher levels of the education system increasingly benefit the better-off. Public health spending on health care is progressive but not pro-poor. This could be related to Ethiopia’s poor being less likely to seek outpatient care but more likely to get services from public providers, particularly lower-level facilities such as health centers and health posts. These lower-level facilities have lower unit costs compared to government hospitals where the uptake of outpatient care is lower. Ethiopia’s flagship Health Extension Program (HEP) has also contributed to improved access to universal primary health services in the country at free cost. ii Four potential policy areas are recommended for improving the fiscal space and making it more pro-poor. (1) Ensuring pro-poor tax revenue generation by reducing the burden of direct taxes on the poor and by broadening the tax base, both for employment and agricultural income taxes. This could be accompanied by simplifying the income tax system and minimizing the adverse effects of possible “bracket-creep”—inflation would cause income/earnings to rise and enter into higher tax brackets. As a result, financing public spending through taxation would be possible without worsening poverty. (2) Expanding the geographic coverage of the Productive Safety Net Program (PSNP) to poor geographic areas currently excluded and increasing its size would help to improve its poverty and inequality reducing effect. This demands additional financing from the government. Given reduction in donor support, the government would need to finance this from its domestic resources. Taxation remains the main tool for financing social protection. (3) Redirecting (electricity) subsidy spending to direct transfers that could more efficiently and effectively benefit the poorest; however, this shift would need to ensure that the (urban) poor with access to electricity receive compensation. This is important since the GoE is currently introducing a new tariff and ceasing the electricity subsidy. (4) Though regressive, tertiary education will have benefits on long term economic growth through fostering technological convergence (absorption and innovation). An attempt should be made to increase the number of students from lower deciles to complete primary and secondary education to have a chance at entering tertiary education. This could also be complemented with redirecting spending to lower schooling in the near term to benefit larger portion of the population. Low performing woredas and regions could be targeted to effectively promote equity. Furthermore, partially subsidizing university education through a stipend system to poorer students could also be considered. However, this could be done in such a way that it does not create burden on the government. iii 1. Introduction Ethiopia has registered rapid, sustained and robust economic growth since the early 2000s. Ethiopia’s macroeconomic trajectory since 2000 is characterized by sustained and robust economic growth as well as a significant change to the structure of the economy (National Planning Commission Ethiopia, 2017). GDP grew by an average of 9.1 percent per year in real terms between 2000/01 and 2016/17. Examining this performance across five-year intervals shows that average annual GDP growth accelerated from 6.6 percent between 2000/01 and 2004/05 to 10.3 percent between 2010/11 and 2014/15. Ethiopia continued to post strong economic growth between 2010/11 and 2015/16; the economy grew at a rate of over 9 percent per year, that results in a 38 percent increase in per capita GDP levels during this period (World Bank, 2020). Before 2016, GDP grew at an annual rate between 8 percent and 11 percent for more than a decade – making Ethiopia one of the fastest growing countries in the world. Large public investment in infrastructure and sustained progress in the agriculture and services sector were the main drivers of this growth (Endale, Pick, & Woldehanna, 2019). In 2015/16, Ethiopia’s economy grew by 8.0 percent, its worst performance since 2002/03. This was attributed to the impact of the El Niño drought on the agricultural sector. For about five years before 2016, the agriculture sector had been registering an average annual growth of 6.6 percent, before it achieved only a 2.3 percent growth in that year. However, this marked slowdown in agricultural output did not slow GDP growth by as much as it would have done in the past. This is due to the growth of the industrial and service sectors since 2000. Nevertheless, this does not alter the fact that the 2015 El Niño drought put the health and livelihoods of millions of rural households at severe risk and required a massive increase in humanitarian relief to avert negative impacts on households (Endale et al., 2019). The strong and sustained economic performance has been accompanied by positive trends in poverty reduction. Poverty declined from 45.5 percent in 1995/96 to 29.6 percent in 2010/11 and reduced further to 23.5 percent in 2015/16 (Figure 1.1). Although Ethiopia has reasonably low poverty rates compared to its structural peers, the extent to which growth (per capita GDP) translates into poverty reduction as indicated by the “poverty-elasticity of growth” has also been low. The poverty-elasticity of growth was - 0.33 between 1997 and 2016, implying that a one percent increase in per capita GDP was associated with a 0.33 percent decrease in poverty rates. The semi-elasticity for the same period was -0.19, indicating that a one percent increase in per capita GDP is accompanied by 0.19 percentage points reduction in poverty (World Bank, 2020). Moreover, the absolute number of people living in poverty has not declined at the same rate as the poverty headcount, mainly due to strong population growth over this period (OECD, 2017). The number of people living in poverty increased from 25.6 million in 1996 to 27.5 million in 2005 then declined to 21.0 million in 2016 (Endale et al., 2019). While the strong GDP growth led to stronger poverty reduction in urban areas , the reduction in poverty (particularly between 2010/11 and 2015/16) was substantially slower in rural areas, parts of which were affected by the El Niño drought (National Planning Commission Ethiopia, 2017). Although poverty is higher in rural areas than in urban areas (25.6 percent versus 14.8 percent in 2015/16), the rapid urbanization in 1 the country and rural-urban migration are more likely to cause an increase in the number of poor individuals in urban areas (OECD, 2017). Figure 1.1. Trends of poverty incidence (national Figure 1.2. Poverty severity trends by location, headcount rate), 1995/96-2015/16 1995/96-2015/16 45.5% 47.5% 44.2% 45.4% 6.0% 5.3% 39.3% 38.7% 36.9% 35.1% 4.6% 33.2% 5.0% 30.4% 29.6% 5.1% 25.7% 25.6% 4.0% 4.5% 23.5% 3.2% 3.1% 4.1% 3.9% 2.7% 3.0% 3.1% 14.8% 2.0% 2.7% 2.7% 2.8% 2.6% 1.0% 1.4% 0.0% 1995/96 1999/00 2004/05 2010/11 2015/16 1995/96 1999/00 2004/05 2010/11 2015/16 National Rural Urban National Rural Urban Source: HCES 1995/96, 1999/00, 2004/05, 2010/11, Source: HCES 1995/96, 1999/00, 2004/05, 2010/11, 2015/16 2015/16 The severity of poverty increased between 2004/05 and 2010/11 in both rural and urban areas, suggesting that the ultra-poor did not benefit from the growth of the economy over this period (Figure 1.2). Between 2010/11 and 2015/16, however, the poverty gap and severity declined in rural areas but not by as much as in urban areas. This suggests the need for a stronger focus on poverty reduction programs in rural areas to ensure rural populations do not fall further behind. The national income Gini coefficient remained around 30 percent for nearly a decade, indicating that the rapid economic growth in the country has been fairly inclusive. With a Gini coefficient of 0.33 in 2016, Ethiopia remains to be one of the most egalitarian countries in the world (UNDESA, 2019). Inequality in Ethiopia, as indicated by the Gini index, has increased from 0.29 in 1995/96 to 0.30 in 2010/11 and further to 0.33 in 2015/16 (Figure 1.3). Inequality has modestly increased between 2010/11 and 2015/16 due to the increasing gap between urban and rural areas and higher consumption growth in the upper parts of the welfare distribution (World Bank, 2020). The data also show that income inequality is higher in urban Ethiopia than in rural areas, and this could be due to higher income growth in the upper parts of the distribution (World Bank, 2020). 2 Figure 1.3. Trends of income inequality (Gini index) by Although the fast and sustainable economic location, 1995/96-2015/16 growth has led to substantial poverty reduction, 0.5 a quarter of the population still live below the 0.44 0.38 0.38 national poverty line in 2015/16 and inequality 0.4 0.37 0.34 0.33 increased in recent years as a result of the urban- 0.29 0.30 0.30 0.27 0.28 0.27 0.28 rural gap in consumption growth. Moreover, the 0.3 0.26 0.26 degree to which growth translates to poverty 0.2 reduction—as measured by the poverty elasticity of growth—was low in Ethiopia. 0.1 Between 1997 and 2016, a one percent increase 0 in per capita growth was accompanied by a 0.33 1995/96 1999/00 2004/05 2010/11 2015/16 percent decrease in poverty, on average. The low National Rural Urban poverty elasticity of growth could be explained by the large public sector infrastructure Source: HCES 1995/96, 1999/00, 2004/05, 2010/11, investments that were behind the strong and 2015/16 sustained economic growth over the past decade, that may not always have direct pro-poor growth impacts. Even as the government’s commitment to poverty reduction remains strong, considerable fiscal deficit and growing debt burden could undermine the fiscal space to expand public spending to the poor. Fiscal policy is shown to be key to reduce inequalities and fund public services that can help provide opportunities for all. In this environment, the question on adequate public spending and pro-poor fiscal policy making is paramount to achieve Ethiopia’s goal of reducing poverty and inequality, both in the present and in the long term. This study analyzes the distributional impacts of taxation and policies on public expenditure in Ethiopia by addressing the following key questions: (1) How does fiscal policy redistribute income and reduce poverty and what is the combined impact of government taxes and expenditures on inequality and poverty? (2) How equalizing, progressive and pro-poor are the different fiscal interventions implemented by the government (direct and indirect taxes, direct transfers, subsidies and in-kind education and health transfers)? (3) What is the contribution of specific taxes and transfers to the overall reduction in poverty and inequality? (4) Is there room for an increased role for the fiscal policy in improving the well-being of the poorest? We combine household survey data from the 2015/16 Household Consumption Expenditure Survey (HCES) and Welfare Monitoring Survey (WMS) collected by the Central Statistical Agency (CSA) of Ethiopia with administrative data related to national income and public finance accounts from the Ministry of Finance and Economic Cooperation (MoFEC) for the same period and other secondary data sources, applying the Commitment to Equity (CEQ) methodology (Higgins & Lustig, 2016; Lustig, 2018). CEQ 3 facilitates a comprehensive assessment of how taxation and public spending affect the welfare of different households, individuals, and socioeconomic groups; an important consideration as those who bear the burden of taxes (economic incidence) are often not those who pay taxes (statutory incidence). Using the CEQ methodology, poverty and inequality indicators are traced across five income concepts: market income, net market income, disposable income, consumable income, and final income. The findings of the study aim to inform policymakers and development practitioners on the redistributive impacts of fiscal policy or fiscal reforms. From a policy perspective, the findings will help to identify areas that could improve the efficiency and pro-poor nature of public spending and tax administration. The results will provide an evidence base for and bring an equity lens to the decision-making process surrounding tax and spending policy reforms. This is particularly important for Ethiopia as the government has recently undertaken several tax reforms, such as a revision of the personal income tax rule in 2016, the Excise Tax Proclamation in 2020, and the implementation of an electricity tariff reform in 2019. The remainder of the paper is structured as follows. Section 2 provides an overview of the fiscal structure of Ethiopia with a focus on taxes and (pro-poor) expenditure. Section 3 discusses the methodology applied in this study, with a focus on the CEQ framework, the poverty and inequality metrics, measures for assessing the distributional impacts of fiscal policy. It also discusses the data used in the analysis with a brief discussion of assumptions and caveats of the study. Section 4 presents the findings of the study, starting from findings of a broader scope and moving towards more specific findings of certain sectors. Section 5 compares the results of this fiscal incidence analysis with the results of a previous analysis of fiscal incidence in Ethiopia undertaken in 2010/11. The last section concludes and provides recommendations for the fiscal policy dialogue. 2. Fiscal Structure 2.1. Taxes Over the previous decades, the Government of Ethiopia (GoE) has implemented various tax policy and administrative reforms to increase its domestic revenue collection. Despite these reforms, revenue collection in general and tax revenue collection in particular remain relatively low (Waiswa, Fekade, & Lake, 2019). Total tax revenue increased at an average annual rate of 11.5 percent (in 2010/11 prices) between 2010/11 and 2015/16. During the same period, annual tax revenue averaged 78.2 percent and 86.1 percent of total government revenue and total domestic revenue, respectively (Endale et al., 2019). This points to the importance of taxes relative to other sources of revenue. Although total tax revenue has increased in absolute terms, it has declined slightly as a percentage of GDP, from 12.7 percent in 2010/11 to 12.5 percent in 2015/16. Table 2.1 presents the tax revenue profile of Ethiopia for 2015/16. 4 Table 2.1. Tax revenue by source in Ethiopia (2015/16) ETB Share in total tax Share in Tax Revenue category (millions) revenue (%) GDP (%) Total tax revenue 190,519.7 100 12.5 Direct taxes 71,843.9 37.7 4.7 Personal income tax 25,744.7 13.5 1.7 Corporate income tax 36,536.4 19.2 2.4 Agricultural income and rural land use fee 706.0 0.4 0.0 Rental income 1,368.8 0.7 0.1 Other direct taxes a 7,488.0 3.9 0.5 Indirect taxes 118,675.8 62.3 7.8 Domestic indirect taxesb 55,952.8 29.4 3.7 Import duties, surcharges, and import taxes 62,722.9 32.9 4.1 Source: MoFEC (2017) a Other direct taxes include withholding income tax on imports, interest income tax, capital gain tax, urban land lease fee, and other income. b Domestic indirect taxes include local value added, excise, and other sales taxes on domestic goods and services. Ethiopia’s tax-to-GDP ratio was 12.5 percent in 2016, much lower than the average of 19.1 percent among African countries and over 25 percent in developed economies (IMF, 2013; Moore, 2013). It is also lower than the tipping point of 15 percent, which is acknowledged to be associated with a significant acceleration in economic growth and development, and below which economies tend to struggle to function and to provide basic services (Waiswa et al., 2019). The average ratio of tax revenue to GDP was only 12.2 percent between 2010/11 and 2015/16, which was much lower than Ethiopia’s Growth and Transformation Plan target of 15 percent of GDP by 2014/15. If existing trends continue and no significant new measures are adopted, the government’s aspiration to raise the tax-to-GDP ratio to 17 percent by the end of 2020 (GTP II target) is unlikely to be realized (UNDP, 2016). The low level of tax revenues significantly reduces the scope for fiscal redistribution (Hirvonen, Mascagni, & Roelen, 2018). Low tax collection performance in Ethiopia could be due to the country’s large informal sector (IMF, 2013), poor tax-paying culture, and shortcomings in tax administration (Waiswa et al., 2019). 2.1.1. Direct taxes Direct taxes represent 38 percent of the total tax take and 4.7 percent of GDP in 2015/16 (Table 2.1). Although the collection of direct taxes is typically low for lower-income countries (Besley & Persson, 2009), direct tax collection in Ethiopia is remarkably high relative to its level of GNI per capita (Hill, Inchauste, Lustig, Tsehaye, & Woldehanna, 2016). Direct taxes in Ethiopia are levied on personal (employment) income, business profits, dividends, interest on deposits and royalties, and agricultural income and land uses (see Appendix A.1 for details on specific tax rates). Personal income taxes (PIT), which are imposed on formal salary income, pensions, interest and dividends, and income from self-employment, are the major direct taxes that contributed 13.5 percent of total revenue in 2015/16 (Table 2.1). This suggests that revenues from PIT are an important source of income for the GoE. About half of the direct tax is collected from business profits. According to the old income tax 5 schedule (the 2002 proclamation which is used in this analysis), the payroll tax rate is progressive with exemption for income less than Ethiopian Birr (ETB) 150 per month, starting at 10 percent and increasing along different income brackets to a maximum of 35 percent which is levied on income of more than ETB 5,000 per month (Table A.2 in Appendix A). As of July 2016, the government adopted a new income tax law (Income tax (amendment) proclamation No. 608/2016) that has dramatically increased all income tax thresholds. The reform raised the exemption threshold for PIT from ETB 150 per month to ETB 600 per month. Changes were also made to the upper thresholds; e.g., the threshold of the monthly employment income to which the 35 percent tax rate is applied was raised from ETB 5,000 to ETB 10,000. Similarly, the portion of annual rental income on which the 35 percent tax is levied was increased from ETB 60,000 to ETB 138,000 (Hirvonen et al., 2018). By doing this, the reform aimed to relieve the poorest from paying income tax and generally to decrease the tax burden for all, especially for individuals with lower incomes. Moreover, it may also improve the perceived equity of the system, thus potentially encouraging compliance among employees and businesses (Endale et al., 2019). Agricultural income taxes and land use fees is levied for the use right of land in both urban and rural areas depending on the size of landholdings (Table A.4 in Appendix A).Despite being a predominantly agricultural society, agricultural income taxes and land use fees account for less than 1 percent of the total tax revenue in Ethiopia. 2.1.2. Indirect taxes Indirect taxes contributed about 62 percent of the total tax revenue and 7.8 percent of the GDP in 2015/16 (Table 2.1). Indirect taxes from import duties and taxes make up a bulk of the indirect taxes (53 percent) and the total tax revenue (33 percent). This points to the country’s reliance on indirect taxes and international trade, a typical feature in other low- and middle-income economies (Besley & Persson, 2009). Import duty rates vary depending on the type of commodity. There are six bands on import duty with a maximum rate of 35 percent. Exemptions from import duties or other taxes levied on imports are granted for raw materials that are necessary to produce export goods and selected investment items. In addition to import duty, a 10 percent surcharge on imported consumer goods was introduced in 2007 and has been implemented to date. Domestic indirect taxes that include Value Added Tax (VAT) and excise taxes comprise about 47 percent of the indirect taxes and 29 percent of the total tax revenue. Goods and services in Ethiopia’s VAT regime are either standard rated at 15 percent, zero-rated, or exempt. Goods and services charged with zero VAT rate include export of goods or services (to the extent provided in the regulations), the rendering of transportation or other services directly connected with international transport of goods or passengers as well as the supply of lubricants and other consumable technical supplies taken on board for consumption during international flights. Goods and services exempted from VAT include unprocessed food items, education, medicine, water, kerosene, electricity, financial services, and transport. Excise taxes are levied on a broad group of commodities that are deemed to be either luxuries or harmful to health, such as alcoholic beverages, tobacco, electronics, textiles, garments, motor vehicles, and other 6 goods and services whether imported or produced locally. Various levels of excise tax are levied depending on the type of commodities, ranging from 10 percent (for textile products) to 100 percent (for perfumes, alcohol, tobacco, and high-power personal vehicles). As of February 2020, a new excise tax proclamation was adopted, replacing a more than two-decades old regime. The new excise tax proclamation (Proclamation 1186/2020) changes the tax base from the cost of production to the ex- factory or manufacturers’ selling price. It also broadens coverage to new items and removal of excise tax from some items. The purpose of the reform is to broaden the excise tax base and raise more revenue, enhance the progressivity of the tax system, and correct for negative externalities. 2.2. Government Spending Public expenditure in Ethiopia comprises spending by federal, regional, and other lower-level administrative units (Endale et al., 2019). The sources of funding for public expenditure are tax revenue, non-tax revenue, domestic credit, external assistance and external loans (World Bank, 2016). Total public expenditure (federal and regional governments combined) was about 281 billion ETB in 2015/16 (Table 2.2). It increased at an average annual rate of 9.8 percent between 2010/11 and 2015/16 (Endale et al., 2019). While public expenditure has increased in absolute terms, it has fluctuated as a percentage of GDP, falling from 18.6 percent of GDP in 2010/11 to 16.6 percent of GDP in 2012/13 before recovering to 18.4 percent of GDP in 2015/16 (Endale et al., 2019). The share of capital spending in total spending was slightly larger (51.3 percent) than recurrent spending (48.7 percent) in 2015/16. Table 2.2. Ethiopia: General Government Expenditure, 2015/16 Share in government Share in GDP ETB (Million) expenditure (%) (%) Total General Government Expenditure 280,893 100.0 18.4 General Services 55,359 19.7 3.6 Economic Development 117,204 41.7 7.7 Agriculture 31,411 11.2 2.1 PSNP 6,036 2.1 0.4 Food security 4,206 1.5 0.3 Urban Development and Housing 10,800 3.8 0.7 Road 41,993 14.9 2.7 Other a 12,462 4.4 0.8 Social Development 98,094 34.9 6.4 Education 67,859 24.2 4.4 Health 21,760 7.7 1.4 Labor and Social Welfare 1196 0.4 0.1 Otherb 7,278 2.6 0.5 Otherc 10,236 3.6 0.7 Indirect subsidies (off-budget) 5,598 2.0 0.4 Source: Ministry of Finance and Economic Cooperation (MOFEC, 2017) and HCES/WMS (2015/16) 7 a other includes trade, tourism, industry, mines and energy, transport and communication, etc. b other includes culture & sports and rehabilitation c other includes interest & charges, external assistance, and miscellaneous expenses. Public spending is guided by Ethiopia’s GTP and is particularly targeted to the pro-poor sectors identified in the GTP (Endale et al., 2019; Hill et al., 2016). The pro-poor sectors cover both the social sector and economic sectors namely education, health, agriculture (and food security), natural resources, and roads. In 2015/16, the share of total poverty-targeted expenditure was about 66 percent of the total government spending and 12 percent of GDP (National Planning Commission, 2017; Waiswa et al., 2019). On the spending side, education accounted for the largest share (24 percent), followed by roads (15 percent), and agriculture (11 percent). Health spending accounts for about 8 percent of the general government budget. The public spending structure shows that the GoE promotes an expenditure policy that aims at expanding education and health service that could play a pivotal role in promoting inclusive growth and poverty reduction. The relatively higher government expenditure-to-GDP ratio (18.4 percent) compared to the tax-to-GDP ratio (12.5 percent) shows that tax revenue is unable to fully finance public spending and the government continues to rely on external assistance to cover budget deficits. 2.2.1. Direct transfers Ethiopia is globally recognized as a leading example to put in place one of the largest social protection programs, replacing humanitarian emergency response. In 2005, the GoE launched one of the largest social protection systems in Africa – its flagship Productive Safety Net Program (PSNP)—which provides predictable safety net support to chronically food insecure people in chronically food insecure rural areas (Endale et al., 2019).1 The PSNP has two components: direct support (unconditional) and public works (conditional on providing labor).2 The number of public works (PW) beneficiaries reached rose from about 6 million in 2005/06 to close to 7 million in 2015/16 (National Planning Commission Ethiopia, 2017). The combined coverage of direct support and public works beneficiaries of 8 million individuals in 2016 represents less than half of the country’s poor population (Endale et al., 2019). In 2015/16, the PSNP budget was 13.8 billion ETB and the total expenditure was 7.6 billion ETB, of which 5.5 billion ETB were transfers to households. Ethiopia has been affected by repeated drought shocks in recent times. In response, the GoE has supported affected households and individuals with emergency relief aid. The average number of beneficiaries per month rose to 7.4 million in 2015/16. During some months of 2015/16, the number of individuals receiving humanitarian relief exceeded 10.2 million, or some 10 percent of the population. Likewise, total expenditure on emergency assistance increased from 9.3 billion ETB in 2009/10 to 13.1 1 See Endale, K. Pick, A. Woldehanna (2019); Hirvonen, Mascagni, & Roelen (2018); World Bank (2020); Coll-Black et al. (2013); and Gilligan et al. (2010) for detailed discussion of PSNP and Food aid, their coverage and targeting and poverty and food security impacts. 2 Direct support (DS) beneficiaries are households enrolled in the PSNP that have no labour capacity, such as children, the elderly and people with disabilities, as well as those who cannot participate in public works programs without jeopardizing their ability to care for children. While the Ministry of Agriculture and Natural Resources (MoANR) oversees the implementation of public works (PW) programs undertaken under the PSNP, the Ethiopian Ministry of Labour and Social Affairs (MoLSA) is responsible for designing and managing transfers to DS (Endale et al., 2019). 8 billion ETB in 2015/16. Humanitarian Food Aid (HFA) or assistance accounted for the majority of the spending, with an average annual share of 78 percent, followed by health and nutrition support and supplementary feeding. Expenditure on HFA in 2015/16 (in 2010/11 prices) is reported to be 10.2 billion ETB (National Planning Commission Ethiopia, 2017). Targeted social programs are believed to be much more effective in terms of the impact and the cost to address the needs of the most vulnerable (World Bank, 2016). Both PSNP and HFA are well-targeted to the poor: PSNP and HFA transfers are generally received by households that are poor, report food shortages, have lower assets (including livestock), have poor access to infrastructure, and live in dryer places with less vegetation and less suitable for rainfed agriculture (World Bank, 2020). Despite an overall successful targeting of PSNP and HFA towards the poor, there are inclusion errors in both programs (mainly for HFA) and substantial disparities between needs and allocations. This suggests that there is a scope to improve the geographical targeting of the PSNP, by targeting households in non-PSNP districts that host a larger share of the poor. Appendix A.3 provides detail discussion of the two programs, including coverage and targeting. 2.2.2. Public spending on education Ethiopia’s education system comprises three years of pre-primary education, eight years of primary (primary grades 1-6, middle school/junior secondary – grades 7-8), four years of secondary (first cycle grades 9-10, and second cycle grades 11-12), and three to five years of tertiary education.3 Over the past decade, Ethiopia has made remarkable progress in extending education to its population, expanding from 10 million students in 2005/06 to about 25 million in 2015 (MoE, 2017; UNICEF, 2017b). Of this, significant improvement is noted for pre-primary and primary education cycle one (grades 1-4) but not much for cycle two (grades 5-8). Enrollment has also increased substantially in urban secondary schools and in higher educational institutions but declined in Technical and Vocational Education and Training (TVET).4 According to the Ministry of Education Statistical Abstract (MoE, 2017), about 96 percent of the enrollment in primary education and 95 percent of the enrollment in secondary education is in public schools. The private sector accounts for only 4 percent of the primary school enrollment, 5 percent of secondary school enrollment, and 15 percent of university enrollment. Private education is mostly concentrated in big urban cities such as Addis Ababa. Public education spending has increased from 21.6 billion ETB in 2009/10 to 67.9 billion ETB in 2015/16. In terms of types of expenditure, spending in the education sector is predominantly recurrent in nature (60.7 percent). In 2015/16, the education sector captures the largest share of the total public spending at 24.2 percent (Table 2.2), reflecting the strong commitment of the GoE to educational development 3Ministry of Education (MoE), Education Statistics Annual Abstract (2016/17). 4The past two decades (1998/99 – 2015/16) have witnessed a major expansion of the public schooling system in Ethiopia, with primary enrollment (Grades 1-8) rising from 5.7 to 20 million, and secondary enrollment (Grades 9-12) increasing from 0.5 to 2.4 million. Up to the late 2000s, the bulk of the spending was directed at financing general education, with an imbalance for TVET and higher education (UNICEF, 2017b). 9 (UNICEF, 2017b).5 This indicates that Ethiopia met the global benchmark of 20 percent of the national budget spent on education as put forth by “the Education for All” initiative (UNICEF, 2017b). Ethiopia’s public education system absorbs a comparatively large share of total GDP, around 4.4 percent in 2015/16, yet, Ethiopia falls short of meeting the global target of 6 percent of GDP allocated to education. As of 2015, the share of government expenditure on primary education, secondary education, and tertiary education were 27.5 percent, 18 percent, and 48 percent, respectively. Even though Ethiopia spends significant public resources on all major levels of the education system, public expenditure is skewed in favor of tertiary education. The per student recurrent spending in higher education is significantly higher compared to that of primary education (26 times) and that of secondary education (10 times) (World Bank, 2016). However, because close to half of the education spending goes to the higher education subsector, the per capita spending at the primary education level (in constant purchasing power parity) is US$118, significantly lower compared to the SSA average of US$586.6 Enrollment rates decrease drastically across the levels of education, partly as a result of the structure of the education system, resulting in escalating levels of per student spending. 2.2.3. Public spending on health Ethiopia’s steady growth trajectory in the last decade alongside the GoE’s commitment towards implementing pro-poor health policies and strategies, resulted in improved health outcomes (UNICEF, 2017c). The country has made good progress to achieve Universal Health Coverage (UHC) by expanding access to a range of health services. The Health Extension Program (HEP)–Ethiopia’s flagship community health program–is among the innovative strategies introduced to improve basic health services in the country. The HEP was launched in 2003 with a view to achieving universal coverage of primary health care to the entire population by delivering a package of basic and essential promotive, preventive, and curative health services (UNICEF, 2017c).7 The public health system in the country focuses on expanding primary health care services and strengthening the implementation of the HEP to deal with the burden of communicable diseases, pre- natal and maternal conditions, and malnutrition. Among the services provided free of charge through the government health system are maternal health care, immunization, and emergency child health services. The national directives also outline the abolishment of user fees for health posts and for some services at health centers (UNICEF, 2017a). Although substantial improvements have been achieved, Ethiopia’s has yet a long way to go to achieve universal health coverage. According to the 2015/16 WMS (CSA, 2016), 5 It should be noted that due to linkages across sectors, spending in sectors such as health and road construction (especially rural roads) has spillover effects, positively affecting education outcomes. For instance, healthier children are more likely to have higher school attendance rates, while rural roads facilitate access to education services (UNICEF, 2017b). Such sectoral interlinkages of welfare effects were outside the scope of this analysis. 6 UNESCO Institute of Statistics (UIS) (2015). 7 Through the HEP, there are service packages that are provided free of charge. The services include childhood vaccinations, family planning, prevention and treatment of malaria, treatment of diarrhea and pneumonia in under five children. Other public health services not covered in HEP are provided through marginal user fees. To protect the poor against the financial burden of user fees, there are fee waiver and exemption systems for services at the public health centers and hospitals. 10 71.4 percent of households are within less than five kilometers from the nearest health post; 47.3 percent within five kilometers from a health center; and 14.5 percent within five kilometers of a hospital. National on-budget health expenditure executed by the Federal Ministry of Health (FMoH) and its affiliated sub-national level Bureaus of Health (BoHs) has increased from 1.3 billion ETB in 2005/06 to 21.8 billion ETB in 2015/16. Per capita health spending amounted to only ETB 268 (US$12) in 2015/16, which is significantly below the US$86 per capita that the World Health Organization (WHO) estimates as the minimum amount needed to provide essential health services in Sub-Saharan Africa (SSA) (UNICEF, 2017c). Data from MoFEC show that public health expenditure accounts for 7.7 percent of the total government spending or 1.4 percent of GDP in 2015/16 (Table 2.2).8 This indicates that Ethiopia has not met the Abuja Declaration target of 15 percent of total government expenditure spent on health. The recent COVID-19 impacts could also constrain the capacity of the GoE to meet its target in the near future. The low spending on health in Ethiopia also implies a considerable reliance on external assistance and personal out-of-pocket (OOP) health care fees that could undermine access for low-income households. The share of household out-of-pocket (OOP) payments of the total health expenditure over the last two decades remained high and stands at about 31 percent in 2016/17 (UNICEF, 2017c; World Bank, 2019). It also underlines the need to increase the share of the budget allocated to the health sector to ensure free provision of health care for the most vulnerable groups. Promotion of health insurance schemes, including the provision of formal health care insurance, which could be viable option to reduce high levels of personal OOP health care charges, remains in its very early stages in Ethiopia. 2.2.4. Indirect subsidies Price subsidies on social services, other than education and health, are limited in Ethiopia. Ethiopia removed the fuel subsidy in 2008 to stabilize escalating prices but government subsidies on electricity, kerosene, and wheat remain intact. These subsides are considered to be expenditures for off-budget operations and not included in general government finance. Electricity subsidies, the primary indirect subsidy, are administered through the operations of the Ethiopian Electric Power Corporation (EEPCo). They totaled an estimated 4.3 billion ETB (equivalent to 0.28 percent of GDP). Kerosene is subsidized through the Oil Stabilization Fund the Oil Stabilization Fund and amounted to 1.2 billion ETB (0.08 percent of GDP). Wheat subsidies are provided to urban consumers (Addis Ababa) primarily through the Ethiopian Grain Trade Enterprise (EGTE). Survey estimates show that they amount to about 21 million ETB (0.001 percent of GDP). Overall, the financing envelop for these subsidies is about 0.4 percent of GDP. 3. Methods and Data 3.1. The CEQ Framework This Fiscal Incidence Analysis (FIA) seeks to assess the biggest winners and losers from the different fiscal instruments applied by the government upon which domestic resource mobilization and government spending depend (Lustig, 2018; OECD, 2017). FIA facilitates our understanding of the overall impact of 8The UNICEF (2015b) budget brief reports a public health spending of 24.5 billion ETB, while MOFEC (2017) reports 21.8 billion ETB. 11 Ethiopia’s fiscal policy on poverty and inequality by demonstrating the extent to which individuals along the income distribution are either net payers or net beneficiaries from the fiscal system—the system of taxes and transfers. To this end, the study applies the Commitment to Equity (CEQ) methodology (Lustig, 2018; Lustig & Higgins, 2012), which stipulates various methods of assigning burdens and benefits to households to analyze the impacts of taxes and transfers on poverty and inequality. It is based on the premise that taxation and expenditure should always be looked at in unison. At the core of the CEQ method is the application of different income concepts. Theoretically, starting from market or pre-fiscal income, the burden and benefits of distinct components of the tax and transfer system should be added consecutively to obtain disposable income, consumable income, and final income. However, since our main data source does not include market income data, our construction of the CEQ income concepts starts with disposable income and works backward towards constructing market incomes and forward to constructing final income. The welfare measure used based on our main data source (the Household Consumption Expenditure Survey), aggregate household expenditure, is conceptually most closely related to disposable income. Figure 3.1 depicts the relationship between the five income concepts constructed for this analysis and helps to illustrate how they are used to analyze the distributional effects of fiscal policy. Market income Market income refers to income before the government has any influence on the income distribution through its tax and spending policies. Thus, “market income” is viewed as the “pre-fiscal” income (income before any transfers or taxes of any kind have been added). It comprises pretax wages, salaries, income earned from capital assets (rent, interest, or dividends), and private transfers. Put in other words, market income includes all earned and unearned income except government transfers and contributory pension receipts. There is some debate on how to treat social security contributions (pensions); the CEQ methodology provides two options: “Public Contributory Pensions as Deferred Income” (PDI) or “Public Contributory Pensions as Government Transfer” (PGT). In the former, pensions are considered as deferred compensation for previous employment (earned income) and are not part of fiscal policy. In the latter, public security contributions are treated as direct tax on incomes and the resulting income from the system are treated as a transfer. In the Ethiopian context, the social security system provides income security in old age, disability, or death only to public servants. As such, social security contributions are considered as savings for formal sector workers (deferred income), and thus part of market income (Hill et al., 2016). 12 Figure 3.1. CEQ Income Concepts Net Market Income As stated earlier, market income is calculated as disposable income plus all direct taxes and less all direct transfers. Given that direct taxes and direct transfers often have very different distributional impacts, it can be helpful to consider their influence separately. Thus, net market income is computed as an intermediate income concept between market and disposable income, as the difference between Source: Lustig (2018) market income and direct taxes. Since our main data sources do not include market income, we calculate direct taxes using information from disposable income, direct transfers, and direct tax rates (personal income tax and employee social security contribution). Section 3.5 and Appendix A.1 provide details about direct tax (income tax) calculations. Disposable Income With market income data available, disposable income can be computed as market income less personal income taxes and employee contributions to social security plus direct cash and near-cash benefits (e.g., direct cash transfers). Disposable income is a cash income available after government has taken away direct taxes (such as personal income tax) and has distributed direct transfers (such as cash transfers). As discussed above, our calculation of market income and other income concepts rely on disposable income which we assume to be equal to aggregate consumption expenditure. Consumption expenditure is computed using detailed consumption data included in the HCES. The survey includes consumption of food and nonfood items, regardless of its source (purchase on the market, own production, or gifts). The consumption data are collected for different households in different periods, spanning the past months (July 8, 2015 to July 7, 2016), by randomly allocating sampled households to different months (World Bank, 2020). The nominal consumption expenditure is adjusted for price variations across space (reporting levels) using spatial price deflators. Consumable (post-fiscal) income Beyond influencing nominal cash income through direct taxes and direct cash transfers, many government policies affect households’ real income indirectly by altering the prices that they pay. Consumable income is a post-fiscal income measure of how much individuals actually consume—that is, the net cash position of households after the intervention of taxes and cash transfers. It is calculated by subtracting the value of direct and indirect taxes paid from the sum of market income and the value of subsidies and direct transfers received. Alternatively, it can be calculated as disposable income less indirect taxes (such as VAT, 13 import duties, and excise taxes) plus indirect subsidies (such as electricity subsidy governments provide to electricity generators and distributors). Final income Governments also effect the income distribution of households or individuals through the provision of free or subsidized services such as health and education. Final income is calculated as consumable income plus the value of these in-kind benefits, less any co-payments, user fees and participation costs for those services. Moving from consumable to final income highlights the effect of public spending on health and education on poverty and inequality. All the income concepts are expressed in ETB per adult equivalent scales per year. 3.2. Inequality and Poverty Measures After computing the CEQ income concepts, inequality indices are calculated at each income concept to assess the redistributive effect of taxes and transfers. Inequality is measured using the Gini index, a widely measure of income inequality. The index ranges from 0 (perfect equality where all individuals earn the same) to 1 (maximal inequality where 1 person earns all the income in the country).9 It computes the difference between all available income pairs in the data and calculates the total of all absolute differences. This total is then normalized by dividing it by population squared times mean income (Hirvonen et al., 2018). We trace how inequality involves as different transfers and taxes are added to and subtracted from income by comparing inequality at different income concepts. For instance, comparing market and disposable income inequality shows how much redistribution is achieved by direct transfers and taxes. Comparing disposable and consumable income inequality shows how much redistribution is achieved by indirect subsidies and taxes (Lustig, 2018). We assess the impact of the fiscal system on poverty by tracing the change in poverty headcount across the different income concepts. Our calculation of poverty is based on the popular Foster– Greer– Thorbecke (FGT) class of poverty measures (Foster, Greer, & Thorbecke, 1984) and includes the headcount index (a measure of the proportion of the population that is poor), the poverty gap ratio (a measure of the depth of poverty – the aggregate poverty deficit of the poor relative to the poverty line), and poverty severity (the a measure of the (squared) proportional shortfall from the poverty line). The poverty rates presented in this study are based on the national poverty line, defined based on a food basket that is required to achieve the minimum daily caloric requirement of 2,200 kilocalories per adult in Ethiopia and adding a nonfood consumption component. The food basket was determined in 1996 and has not been updated since. National poverty lines after 1996 are obtained by updating the 1996 poverty line, which was 1,075 ETB per adult equivalent per year in 1996, using the national Consumer Price Index (CPI). The poverty line for 2011 was set at 3,781 ETB per adult equivalent per year in December 2010 prices and was determined by costing the items in the original food basked at prevailing prices and doing a similar adjustment for nonfood consumption (World Bank, 2020). For 2015/16, the 2011 poverty line was inflated using the GDP deflator, resulting in a poverty line of 7,184 ETB and a food (extreme) poverty line of 3,781 9 Descriptive statistics are presented for Theil Index and r90/10 measures of inequality. 14 ETB per adult equivalent per year in December 2015 prices (Planning and Development Commission, 2018; World Bank, 2020). 3.3. Incidence, Progressivity, and Pro-poorness The incidence of taxes and transfers is determined by inspecting the distribution of the share of taxes paid, or transfers received as a proportion of income. The progressivity of taxes (transfers) is assessed using the tax (transfer) redistribution approach that compares the cumulative distribution (cumulative concentration shares) of their burden (benefit) with the cumulative distribution of market income and the cumulative share of total population ranked by income (Duclos & Araar, 2006). We illustrate this in figure 3.2 using market income Lorenz curve and concentration curves for taxes and transfers. Lorenz curves are used to make unambiguous comparisons about whether inequality falls as a result of a fiscal system. It maps the cumulative share of market income (or any income of interest) on the vertical axis against the cumulative share of the population, ordered by market (sometimes called ‘benchmark or ‘reference’) income on the horizontal axis. Because the horizontal axis is reranked with each income concept, the Lorenz curve is an anonymous measure by definition.10 A 45-degree line represents perfect equality (Gini coefficient equals 0) and a Lorenz curve that lies flat over the horizontal axis until the last person and vertical for the last person in the population is the line of perfect inequality (Gini coefficient equals 1). Typically, the estimated market income Lorenz curve lies between these two extremes (Lustig, 2018). Concentration curves (sometimes called “quasi– Lorenz curves”) are important descriptive and normative tools for capturing the impact of tax and transfer policies. They map the cumulative share of taxes paid or benefits received from a particular category of taxes or transfers on the vertical axis against the cumulative share of the population, ordered by pre-fiscal income, on the horizontal axis (Duclos & Araar, 2006). The concentration curves graph basically includes the 45-degree line, the pre-fiscal income Lorenz curve, and concentration curves for different categories of transfers and taxes. The progressivity of a tax or transfer can be determined by comparing its concentration curve to the market income Lorenz curve. Whether a progressive transfer is progressive in absolute terms or in relative terms, is, in turn, determined by comparing the concentration curve to the 45-degree line. Concentration indices are also computed from concentration curves, and used to (i) compute aggregate indices of progressivity and vertical equity and (ii) decompose the inequality at a particular income concept into a sum of the concentration of the components of that income (Duclos & Araar, 2006). Another standard measure of progressivity of a particular fiscal intervention is the Kakwani coefficient calculated by subtracting the intervention’s concentration coefficient from the market-income Gini. Progressive interventions have positive Kakwani coefficients, and regressive ones have negatives coefficients (Higgins & Lustig, 2016). 10Its non-anonymous analog would be the concentration curve of each income definition with respect to the market income rankings. 15 Figure 3.2. Progressivity of taxes and transfers (diagrammatic A tax is globally progressive (regressive) if representation) the proportion paid–in relation to market income–increases (decreases) as income rises, or when its concentration curve lies everywhere below (above) the market income Lorenz curve (Higgins & Lustig, 2016; Lustig, 2018). A necessary but not sufficient condition for this is that the concentration coefficient is positive and larger (smaller) than the market income Gini. This is equivalent to saying that the Kakwani index (defined for taxes as the tax concentration coefficient minus the market income Gini) will be positive (negative) if a tax is everywhere Source: Higgins & Lustig (2016); Lustig (2018) progressive (regressive). If the concentration curve of the tax crosses the market income Lorenz curve (from above or from below and for any number of times), the tax will be ambiguous (i.e., neither progressive nor regressive). Therefore, the progressivity of a tax will be determined using concentration curves, not concentration coefficients or Kakwani indices alone. However, taxes are not everywhere progressive in practice since some households would manage to evade the tax while another household with slightly lower income and another with slightly higher income do not. A tax will be neutral (in relative terms) if its distribution (its concentration curve) coincides with the distribution of the market income Lorenz curve. A necessary but not sufficient condition for this is that the concentration coefficient is equal to the market income Gini, or equivalently, the Kakwani index is equal zero (Lustig, 2018). A transfer is globally (everywhere) progressive if the proportion received—in relation to market income— decreases as income rises, or when its concentration curve lies everywhere above the market income Lorenz curve (Higgins & Lustig, 2016; Lustig, 2018). There are two types of progressive transfers: absolute and relative. A transfer will be progressive in absolute terms (“pro-poor”) if the amount received increases as income rises. A transfer whose concentration curve lies everywhere above the 45-degree line is globally progressive in absolute terms or pro-poor. A necessary but not sufficient condition for this is that the concentration coefficient of the transfer is negative, or equivalently that the Kakwani index (defined for transfers as the market income Gini minus the transfer’s concentration coefficient) is positive and higher than the market income Gini. Again, transfers in practice are usually not everywhere progressive because someone might not receive the transfer while a slightly poorer and a slightly richer person both do. A transfer will be progressive in relative terms if the proportion received in relation to market income decreases as income rises but not so the amount of the transfer, or equivalently, when its concentration curve lies everywhere between the market income Lorenz curve and the 45-degree line. A necessary but not sufficient condition is that the concentration coefficient is positive and lower than the market income Gini, or equivalently that the Kakwani index is positive if a transfer is progressive in relative terms. If the concentration curve of a transfer crosses the 45-degree line (from above or below and any number of times) but still lies everywhere above the market income Lorenz curve, it is unambiguously progressive, 16 but we cannot say whether it is unambiguously progressive in absolute terms, even if its concentration coefficient is negative (Lustig, 2018). A transfer is globally regressive if the proportion received–in relation to market income–increases as income rises, i.e., when its concentration curve lies everywhere below the Lorenz curve. However, transfers will not be everywhere regressive in practice (Lustig, 2018). A transfer will be unambiguously regressive if its concentration curve lies everywhere below the market income Lorenz curve. A necessary but not sufficient condition for this is that the concentration coefficient of the transfer is positive and greater than the market income Gini, or equivalently, that the Kakwani index is negative. If the concentration curve of the transfer crosses the market income Lorenz curve, we cannot unambiguously say that the transfer is progressive or regressive. We determine the progressivity of transfers using concentration curves—and not concentration coefficients or Kakwani indices alone. A transfer will be neutral (in relative terms) if its distribution (its concentration curve) coincides with the market income distribution (market income Lorenz). A necessary but not sufficient condition for this is that the concentration coefficient is equal to the market income Gini, or the Kakwani index equals zero. To gauge whether a particular tax or transfer is equalizing or unequalizing, we use the “marginal contribution” of that tax or transfer to inequality–its “redistributive effect”. The “redistributive effect” or the marginal contribution equals the difference between market-income Gini coefficient (inequality without the tax or transfer of interest but including all the other components of the fiscal policy in place) and the relevant ending income concept Gini (the inequality with all the components including the one whose effect we are considering). The tax or transfer is equalizing (unequalizing) if the difference is positive (negative), i.e., inequality is higher (lower) without the tax or transfer of interest than with it (Lustig, 2018). If the difference equals zero, the tax or transfer is “neutral”, suggesting no effect on inequality or poverty. A tax or transfer is poverty reducing (increasing) if its marginal contribution to poverty reduction– “poverty reducing effect” calculated as the difference in the poverty headcount for the income concept with and without the intervention–is positive(negative). Even a fiscal system that reduces poverty and inequality and is progressive can make a substantial portion of the poor worse off or better off (Higgins & Lustig, 2016). This startling result occurs because poverty indicators are anonymous in the sense that we do not know whether a particular individual with a set post-fiscal income had a lower or higher pre-fiscal income. We assess the extent to which the prevailing combination of taxes and transfers system makes some of the poor poorer and some of the non-poor poor using a measure of “fiscal impoverishment” (Higgins & Lustig, 2016). A measure of the extent of fiscal impoverishment helps to assess anti-poverty policies in tandem with the taxes used to finance them. 3.4. Data This assessment is based on comprehensive household surveys as well as administrative data on taxes and transfers. The surveys ensure representative sample of individuals in the country, which helps to better understand the distributional consequences of taxes and public expenditures. This study utilizes data from the 2015/16 HCES and WMS collected by the CSA of Ethiopia, the main data sources used by the GoE to monitor its poverty reduction strategies. The HCES has detailed information on household 17 consumption of food and nonfood items, whereas the WMS has detailed information on socio-economic outcomes of individuals and access to various services, including education and health. Data from the HCES and WMS are combined with data from the national income and public finance accounts obtained from MoFEC. These accounts provide the public revenue and expenditures corresponding to the 2015/16 Ethiopian fiscal year. Data are also gathered from various other sources. We use the most recent (2015/16) Social Accounting Matrix (SAM) for Ethiopia, from which we have constructed an Input-Output (I-O) matrix that is used to estimate the indirect effects of indirect taxes (Mengistu et al., 2019). We combine consumption item data from HCES and tax schedules from the Ethiopian Revenue and Customs Authority (ERCA) with sectors in the I-O matrix. 3.5. Assumptions Several assumptions are made regarding the CEQ income concepts calculations and the different methodological approaches employed in this analysis. The main assumption is related to the fact that HCES, our main data source, does not report income data. Therefore, aggregate consumption expenditure is assumed to be equal to “disposable income”, and taxes and transfers are subtracted or added to obtain the other CEQ income concepts described earlier. Taxes Regarding the economic incidence of taxes, we assume that direct taxes are borne entirely by the income earner and indirect taxes entirely by the consumer (Hill et al., 2016). This assumption is important since most markets in Ethiopia would not be competitive, a quintessential condition in developing countries. There is also no clear information regarding the degree to which monopolies or oligopolies shift indirect taxes to consumers; it could be either greater or less than 100 percent depending on the functional form of the demand function (Fullerton & Metcalf, 2002). Direct Taxes Since the HCES does not provide information on taxes, allocating taxes across households is challenging. Using consumption data to compute and allocate taxes across households is equally challenging since taxes are calculated separately for different types of incomes, such as wage and self-employment, rather than being applied to the aggregate income amount (Hirvonen, Mascagni, & Roelen, 2016). Personal income taxes are imputed using information from HCES about sector of employment of individuals, household disposable income and direct transfers, and income tax rates. We first estimate the share of wage and self-employment income earners based on information regarding sector of employment given in HCES (Hill et al., 2016; Hirvonen et al., 2018). After a series of intermediate calculations, we compute wage and self-employment taxes at the household level.11 Consistent with other conventional tax 11We deduct direct transfers from disposable income instead of applying taxes (and employee social security contributions) on disposable income because: Net income tax = Market income – Direct taxes = Disposable income - Direct transfers. Then, we multiply the resulting value by the share of wage earners to arrive at labor income. Using the “dirtax” Stata routine, we com pute the wage tax. Similar procedure is followed for calculating self-employment tax. Our method is not without a limitation. Note that, by doing this, we might overestimate the total tax payments for those households that have more than one wage earner 18 incidence analyses, we assume that the economic burden of PIT is borne by the income earner. It is assumed that self-employed and employees of the informal sector do not pay taxes. Agricultural income taxes and rural land use fees are calculated on the basis of landholding size reported in the WMS data. The tax schedule for agricultural income taxes and land use fee is set by regional and local governments. Although the rate varies across locations, very similar per hectare tax rates are used regardless of land size. We use the tax rates along with the size of the land holdings to estimate the amount of agricultural tax and land use fee paid by each household. Although the WMS provides land holdings in standardized units, the data are noisy. We winsorize at the top 1 percent to deal with outliers. Appendix A.1 provides further information about the calculation of direct taxes for this analysis. Indirect Taxes Indirect taxes (excise and VAT) are simulated using data on households’ reported consumption of the corresponding items. Households do not pay indirect taxes explicitly, but they are reflected in the prices they pay for taxed goods and services. The total exercise captures both the direct effects (i.e. VAT or excises paid directly by households when purchasing manufactured and specific imported goods subject to this tax) and the indirect effects of taxes. We assume that agricultural products, own production for consumption, free collection of firewood, water, are non-excisable and VAT is not levied on them. Estimating the indirect taxes paid by a household when purchasing a particular product is complicated by variable tax rates, significant tax evasion, and the fact that some of these taxes fall on intermediate products that will induce effects on price of other products. This latter problem is especially important for fuel excises because fuel is consumed as intermediate goods and fuel excises will have indirect effects on product prices. To deal with this in our analysis, we use the 2015/16 social accounting matrix (SAM) for Ethiopia (Mengistu et al., 2019) and applied the method described in Chapter 7 of Lustig (2018) to calculate the indirect taxes. We calculate both the direct and indirect effects of fuel excises on the final prices of all goods and services by tracing their impact through the I-O table. We then map the sectors in the SAM to each item in the HCES expenditure module, applying the effective (direct + indirect) tax rates from the SAM to the corresponding expenditure items. In terms of tax evasion, the compliance gap for excises is estimated to be large, particularly for road fuels and cigarettes. There usually is, however, also underreporting of this type of consumption items in the household survey (alcohol, cigarettes, and others). As such, the exercise does not alter the tax rate to adjust for evasion.12 Transfers Estimation of the benefits of direct transfers (PSNP and HFA) is applied based on the methodology used in the recent poverty assessment for Ethiopia (World Bank, 2020). PSNP and HFA beneficiaries are identified using detailed information in the 2015/16 HCES. The beneficiary status from HCES and and/or one self-employed. By taxing those incomes at the household level as a sum, rather than separately, we apply a higher marginal rate and only one exempt threshold. However, the majority of households in our data have only one wage earner and/or one self-employed (Hirvonen et al., 2018). 12 The “costpush” Stata routine is used to estimate the indirect effects of VAT and excise taxes. 19 household size information are combined with prevailing wage rates, support type and eligibility to calculate PSNP transfers. The report also calculates HFA benefits assuming they are on average 15 percent higher than PSNP. The targeting performance of both programs is also assessed using a range of monetary and non-monetary welfare indicators. Appendix A.3 gives detailed information about the two direct transfer programs including beneficiaries, expenditure, coverage, and targeting mechanisms and performance. Allocation of public education and health spending to households is based on a “cost-of-production” or “government cost” approach.13 Public education spending (obtained from administrative data) is allocated to households in which members are enrolled in public schools. The WMS provides information on educational enrollment by level (primary, secondary including TVET, and tertiary) and type (public vs. private institutions). The allocation is made on a per-student basis and by education level, using the number of students estimated from the survey. For health spending, in-kind benefits are estimated by equally distributing the public expenditure to all households that use public health institutions. Appendix A.4 provides details regarding the allocation of in-kind benefits from public education and health spending to households in the survey. Indirect subsidies are estimated using item-level HCES data, which provide households’ consumption of wheat, kerosene, and electricity. We apply the subsidy per kilogram, liter, and kilowatt-hour for each good, respectively to estimate the total value of the subsidy received by the household. Appendix A.5 gives details about subsidy rates and the calculation of subsidies for this analysis. 3.6. General Caveats Along with the above assumptions, the caveats of this analysis are important. Some of the limitations are inherent to the CEQ methodology and others are related to data availability. There are important limitations of the analysis that are common in this type of analysis. First, the analysis does not consider behavioral, lifecycle, or general equilibrium effects. Public expenditure on health and education may have an inherent investment element, the inclusion of which would likely imply different conclusions about long-run inequality and poverty dynamics. However, the CEQ methodology does not attempt to estimate the long-run benefit or investment value created by public expenditures on health and education services. Rather, it provides a point-in-time economic analysis of the distribution of public expenditures and the burdens created by government taxation (Lustig, 2018). While this is a caveat, the CEQ framework is not well placed to address these trade-offs between consumption now and consumption later, but it only helps to assess the current impacts of public expenditures on poverty and inequality. In connection to this, policymakers will have to decide between public spending that immediately benefits the population in the form of higher levels of economic welfare (e.g., direct cash transfers) and public investment that will increase economic welfare only over time (e.g., 13 This approach is also known as the “classic” or “nonbehavioral approach,” and it amounts to asking the following question: how much would the income of a household have to be increased if it had to pay for the free or subsidized public service at the full cost to the government? 20 public education). Furthermore, the analysis provides information about the average incidence, not the incidence at the margin. Second, as in much of the literature on poverty and inequality analysis, the study ignores the intra- household distribution of consumption. Third, the analysis does not consider differences in the quality of education or health care services delivered by the government across income groups. Fourth, important taxes (e.g., taxation of corporate income and international trade, property taxes) and spending (e.g., infrastructure) that are part of the government budget are not considered in this analysis. Due to data limitations and inconsistencies between survey and administrative data, this analysis attempts to cover all the taxes and transfers that can be plausibly allocated directly to households. It covers personal income taxes, agricultural income taxes, and domestic indirect taxes (VAT and excise taxes) that account for about 43 percent of the total government revenue. It also covers direct cash transfers, subsidies, and public spending on education and health that account for about 36 percent of the total government spending in 2015/16. Nevertheless, the imbalance that results from covering different shares of the tax and spending side is not unusual for CEQ based assessments. While this might result in a lop-sided view of the overall effect on poverty or inequality, the comprehensiveness of the assessment means that it will still serve as a baseline from which more narrow questions about the tax and transfer system can be answered. 4. Distributional Impact of Taxes and Spending 4.1. Overall Impact on Inequality The impact of the fiscal system on inequality is analyzed using the difference in the Gini coefficient across the different income concepts (Table 4.1). The market income Gini coefficient for Ethiopia is 0.348, lower than the simple worldwide unweighted average of 0.38 and the average for SSA of 0.44 during the same period (UNDESA, 2019). The market income inequality in Ethiopia is low relative to other countries including Tanzania (0.379 in 2011/12), Uganda (0.43 in 2015/16), and Kenya (0.36 in 2015/16). As we move from market income to final income, the Gini coefficient drops from 0.348 to 0.327—a decline of 2.1 percentage points—indicating the inequality reducing effect of the overall fiscal system. The results of this study demonstrate that fiscal policy in Ethiopia is an important tool to affect income distribution and reduce income inequality in the country. Table 4.1. Poverty and Inequality Indicators in Ethiopia by CEQ Income Concept (2015/16) Market Net market Disposable Consumable Final income income income income income Poverty: Absolute poverty Headcount 0.237 0.242 0.232 0.236 0.182 Gap 0.069 0.071 0.066 0.067 0.048 Severity 0.029 0.030 0.027 0.028 0.019 Extreme poverty Headcount 0.049 0.051 0.043 0.045 0.028 21 Gap 0.010 0.010 0.008 0.009 0.005 Severity 0.003 0.004 0.002 0.003 0.001 Inequality: Gini 0.348 0.332 0.328 0.328 0.327 Theil 0.233 0.204 0.199 0.199 0.202 r 90/10 4.592 4.475 4.326 4.334 4.097 Source: Own calculations based on HCES and WMS (2015/16) and Administrative Data. Moving across the table from market income to disposable income, there is evidence that direct transfers and direct taxes jointly reduce market income inequality by about 2 percentage points. However, there is no significant change in the Gini coefficient as we move from disposable income to consumable income. The transition from consumable income to final income shows a small reduction in the Gini coefficient of only 0.1 percentage points, indicating that, although in-kind benefits (such as spending on education and health care) are expected to further reduce income inequality through their effect on future earnings, health condition, ability to work, and living standards, their aggregate effect on inequality remains low. Figure 4.1. Gini coefficient by CEQ income concepts and country 0.450 0.430 0.420 0.420 0.410 0.400 0.379 Ethiopia 2015/2016 0.358 0.357 0.362 0.345 Kenya 2015/2016 0.350 0.331 0.336 Tanzania 2011/12 0.348 0.328 0.332 0.328 0.328 0.327 Uganda 2015/16 0.297 0.250 Market income Net market Disposable Consumable Final income income income income Source: Own calculations based on HCES and WMS (2015/16), Administrative Data, and Data from the CEQ institute. Note: Gini coefficient is not calculated at net market income for Kenya. The inequality-reducing effect of taxes and transfers between market income and final income in Ethiopia is comparable with other countries in the region. Much of the reduction in inequality is achieved by direct taxes, while direct transfers do not achieve much in reducing inequality. The contribution of direct taxes in reducing inequality (as we move from market income to net market income) in Ethiopia (1.6 percentage points) seems to be in the middle range between that of Tanzania (2.1 percentage points) and Uganda (1 percentage points). Inequality in Kenya, Uganda, and Tanzania is reduced through direct taxation and transfers, ranging from a decline in the Gini by about 1 percentage point in Uganda to 2.2 percentage points in Kenya and 2.6 percentage points in Tanzania (Figure 4.1). Nevertheless, inequality barely 22 increases in Ethiopia and decreases in Kenya between disposable income and consumable income. Only Tanzania and Uganda achieve a reduction by 1.2 and 1 percentage points, respectively, moving from disposable to consumable income. The incidence analysis of the comprehensive fiscal policy (all taxes and transfers including in-kind transfers) shows that the net impact of the fiscal policy is progressive, with all but the top 10 percent (the richest) receiving more benefits relative to their market incomes than the taxes they pay (Figure 4.2).14 The poorest received about 26 percent of the benefits (all transfers minus all taxes) relative to their market income compared to the taxes they pay. The bottom 20 percent receive 44 percent of the benefits relative to their market incomes. The results indicate that fiscal policy reduced inequality in Ethiopia. Figure 4.2. Incidence of Taxes and Transfers and Net Fiscal Benefit by Market Income Decile Share of market income (%) 30% 20% 10% 0% -10% Poorest 3 8 Richest 2 4 5 6 7 9 Income deciles Direct Taxes All Direct Transfers Indirect Subsidies Indirect Taxes In-kind Education In-kind Health Dir. transf. net of dir. tax. Total net benefit Source: Own calculations based on HCES and WMS (2015/16) and Administrative Data Note: Net benefit is calculated as the difference between all transfers (direct subsidies and in-kind transfers) and all taxes (direct + indirect). 4.2. Overall Impact on Poverty Beyond the redistributive effects of a fiscal system on equality, it is also equally important to better understand its impact on poverty, particularly as results may not go in the same direction. Put in other words, an inequality-reducing fiscal system could still increase poverty. In this study, we measure the effect of the fiscal policy on poverty using the typical indicators such as the headcount ratio at different income concepts (Table 4.1). This study finds that the combined system of taxes and transfers (but excluding in-kind education and health transfers) decreases headcount poverty rate by only 0.15 percentage points. Overall, Ethiopia achieves little poverty reduction through direct taxes and direct transfers while indirect taxes and indirect subsidies increase poverty. The poverty headcount rate drops from 23.8 percent at market income to 23.2 percent at disposable income (Table 4.1). This suggests that direct transfers offset the poverty increasing effects of direct taxes. 14The shares of the taxes, transfers and net benefit are calculated by dividing the values by the market income at each decile. Therefore, the shares are in relative to the market income. 23 In moving from market income to consumable income, the results show that taxes, cash transfers, and subsidies jointly slightly increase the poverty headcount from 23.8 percent (at market income) to 23.9 percent (at consumable income). This indicates, on the one hand that government transfers and subsidies do not make up for the impact of direct and indirect taxes around the poverty line, on the other hand, we see that government subsidies do not offset the poverty increasing effects of consumption taxes.15 Figure 4.3. Poverty headcount (national) ratios by income concepts across countries 0.400 0.348 0.294 0.285 0.282 0.300 0.242 0.25 0.237 0.232 0.236 Ethiopia 2015/2016 0.237 0.182 0.200 Tanzania 2011/12 0.2139 0.2144 0.2142 Uganda 2015/16 0.100 Disposable (pre-fiscal) Consumabl Net market income e income income Market income Final income Source: Own calculations based on HCES and WMS (2015/16), Administrative Data and Data from the CEQ Compared to other SSA countries, the poverty reducing effects of direct taxes and direct transfers are moderate in Ethiopia (0.6 percentage points). Cross-country comparisons suggest that the change in the poverty headcount ratio in going from market income to disposable income is often limited. In Tanzania for example, a reduction of 1.2 percentage points is achieved when moving from market income to disposable income, while Uganda manages to reduce poverty only by a tenth of a percentage point (Figure 4.3). The limited poverty reducing effect of direct taxes and direct transfers could be due to low cutoff income for personal income tax, high levels of informality, and relatively low coverage of the direct transfers programs.16 In going from disposable income to consumable income, poverty headcount increased by 0.4 percentage points, suggesting the poverty-increasing effects of consumption taxes are not offset by subsidies. This finding is not uncommon among countries in SSA. However, compared to other SSA countries (e.g., Tanzania and Uganda), the poverty increasing effects of indirect taxes and subsidies in Ethiopia is modest (Figure 4.3). Our results also show that the poverty gap and poverty severity are lower for disposable income than market income. But they are also lower for consumable income than for market income. Although this indicates that the fiscal policy reduces the incidence, depth, and severity of poverty, it can be misleading 15 Standard CEQ convention suggests that poverty rates are not calculated after in-kind health and education transfers because households may not be aware of the actual amount spent on their behalf and may not value this spending as much as they would a direct cash transfer (Hill et al., 2016). 16 In Ethiopia, the largest social protection program, the PSNP, covered 13 percent of the country’s poor in 2016. 24 as standard poverty measures could fail to capture the extent to which the tax and benefit systems fiscally impoverish the poor (Higgins & Lustig, 2016; Lustig, 2018). To assess the overall effect of the fiscal policy on the poor, we use a “fiscal impoverishment headcount index” that measures the percentage of the population impoverished by the fiscal system as a proportion of the post-fisc-poor (Higgins & Lustig, 2016). Table 4.2 summarizes the impoverishment indices at the national poverty line, using three income- concept comparisons: from market to disposable income, from market to consumable income, and from market to final income. Table 4.2. Fiscal impoverishment by fiscal policy in Ethiopia, 2015/16 Fiscally Fiscally impoverished as impoverished as % % of populationa of post-fisc poorb Market income to disposable income 14.1 60.9 Market income to consumable income 18.2 77.2 Market income to final income 2.96 16.3 Source: Own calculations using HCES and WMS (2015/16) and Administrative Data. Note: aFiscally impoverished as share of population represents the proportion of the non-poor population (at market income) that became poor at the ending income concept; bfiscally impoverished as share of post-fisc poor is the percentage of the population that were pre-fisc poor (at market income) and became poorer at the ending income concept. Moving from market income to disposable income, we observe that direct taxes impoverished 14 percent of the non-poor and made 61 percent of the poor even poorer, even when taking direct transfers into account. Moving from market income to consumable income—when all taxes and transfers excluding in- kind education and health benefits are considered—the fiscal policy still further impoverishes 77 percent of the poor and makes 18 percent of the non-poor poor. This finding, although startling, is not uncommon for low-income countries that have experienced fiscal impoverishment despite the reduction in poverty and inequality due to the tax and transfer system, including Ethiopia’s own past. Results from the fiscal incidence analysis in 2010/11 shows that 28.5 percent of the population and over 80 percent of the post- fisc poor experience fiscal impoverishment (Higgins & Lustig, 2016). Fiscal impoverishment has also been significant in other SSA countries (Mejia-Mantilla, Fajardo-Gonzalez, Goldman, Jellema, & Renda, 2019).17 The proportion of the population fiscally impoverished in Uganda (2016) was 22.8 percent and 50.9 percent in Tanzania (2011). These countries also experienced higher rates of fiscal impoverishment among the post-fisc poor: 92.5 percent in Uganda (2016) and 98.6 in Tanzania (2011). 4.3. Progressivity, Marginal Contributions, and Pro-Poorness of Taxes and Transfers As discussed in section 2, the progressivity of taxes and transfers is measured using the Kakwani coefficient and their “marginal contribution” to inequality and poverty reduction. Tables 4.3 presents the marginal contributions of taxes and transfers to inequality and poverty reduction. 17 These values are based on the $1.25 PPP [2005] poverty line, and with respect to market income plus pensions. 25 Table 4.3. Marginal Contribution of Taxes and Transfers to Inequality and Poverty Reduction Marginal contribution Kakwani Fiscal intervention Redistributive effect Poverty reduction coefficienta a (change, pp) effect (change, pp)a Consumable income 2.006 0.147 Direct taxes 0.42 1.392 -0.484 Personal income tax 0.43 1.409 -0.429 Agricultural income & land use fee -0.28 -0.017 -0.020 Direct transfers 0.57 0.459 1.100 PSNP 0.58 0.270 0.552 HFA 0.55 0.175 0.467 Indirect subsidies -0.09 -0.053 0.186 Electricity -0.17 -0.071 0.155 Kerosene 0.19 0.018 0.024 Wheat 0.10 0.000 0.000 Indirect taxes 0.01 0.030 -0.602 VAT 0.03 0.003 -0.019 Excise 0.01 0.027 -0.565 Final income 2.115 5.599 Direct taxes 0.42 0.013 -0.347 Direct transfers 0.57 0.004 1.042 Indirect subsidies -0.09 -0.001 0.118 Indirect taxes 0.01 0.000 -0.541 In-kind transfers 0.17 0.001 5.452 Education 0.13 -0.004 3.402 Primary 0.47 0.009 2.125 Secondary 0.22 0.002 1.139 Tertiary -0.15 -0.014 0.176 Health 0.28 0.005 1.897 Source: Own calculations based on HCES and WMS (2015/16) and Administrative Data a Note: pp = percentage point; positive (negative) Kakwani coefficient implies that the tax/transfer is progressive (regressive); positive (negative) marginal contribution means inequality/poverty reducing (increasing). Moving from market income to consumable income, overall, the results show that direct taxes and indirect taxes are progressive. Among the direct taxes, personal income tax is redistributive but poverty increasing. However, the agricultural income tax and land use fee is regressive and both, inequality- and poverty-increasing. Direct cash transfers are progressive in absolute terms. Based on their marginal contribution, they are also redistributive (reducing inequality by 0.46 percentage points) and poverty reducing (reducing poverty headcount by 1.1 percentage points). Among the direct transfers, PSNP appears to be slightly more welfare-enhancing than HFA. The results also show that indirect taxes and subsidies are regressive. However, their redistribute effect is considerably small. Indirect taxes are poverty increasing. In-kind health and primary education are equalizing and poverty-reducing. However, there is 26 heterogeneity across levels of education, with primary education being strongly progressive and redistributive. In contrast, tertiary education is regressive and unequalizing. 4.3.1. Taxes Direct Taxes Direct taxes, particularly personal income tax, are progressive. Personal income taxes are allocated to individuals in the survey based on the assumptions about formal-sector employment. Their progressivity is assessed based on comparisons between Lorenz and concentration curves. The poorest 40 percent of the population (in terms of per adult equivalent market income) account for about 20 percent of the market income but (red line in Figure 4.4) and about 5 percent of the direct taxes (blue line in Figure 4.4). In contrast, about 71 percent of the incidence of direct taxes is borne by the richest 10 percent of the population (Figure 4.4). Figure 4.4. Incidence and Progressivity of Taxes by Market Income Decile in Ethiopia 100% CUMMULATIVE PROPORTION OF TAX 90% 80% 70% 60% 50% (%) 40% 30% 20% 10% 0% 0 1 2 3 4 5 6 7 8 9 10 MARKET INCOME DECILE Market Income Lorenz PIT Agri Tax Direct taxes Population shares Source: Own calculations based on HCES and WMS (2015/16) and Administrative Data PIT account for about 0.5 percent of the total market income among the poorest decile and their share increases to 8 percent in the top decile (orange line in Figure 4.4). This could be the result of both limited access to formal-sector jobs among the poor and the progressivity of the tax system. The analysis of the incidence of personal income taxes shows that PIT is progressive, as the main burden of these taxes is carried by the rich (more than 70 percent is paid by the top 10 percent). However, lack of revisions to the tax brackets for more than a decade, and the relatively low threshold of the first tax bracket, suggest that Ethiopia levies more taxes on the lowest income households compared to other countries.18 Moreover, the prevalent larger underemployment and limited formal employment in Ethiopia limits the progressivity of direct taxes and tax revenue collection. Analysis of the redistributive effect of direct taxes shows that they are the main instrument of redistributive policy affecting income inequality. Direct taxes are more equalizing than direct transfers as 18The recent tax reform of July 2016 could not be considered in this analysis as the HCES/WMS data were collected before the tax rule was revised. 27 they lower the Gini index by 1.4 percentage points compared with 0.5 percentage points as a result of direct transfers. However, the poverty headcount increases with direct taxes by around 0.5 percentage points. PIT is progressive and equalizing, but it is also poverty-increasing (Table 4.3) because any monthly personal income above ETB 150 (or ETB 1,800 per year) is taxed.19 This threshold is much lower than the poverty line of ETB 7,184 per adult equivalent, implying that the poor are effectively paying income taxes. Increasing this minimum cutoff would reduce the direct tax burden on the bottom deciles and the subsequent loss in tax revenue could easily be offset by slightly higher PIT rates for individuals in higher deciles (Hill et al., 2016; Hirvonen et al., 2018). Agricultural income tax and land use fee is regressive, as well as both inequality and poverty increasing (Table 4.3). This can partly be explained by the fact that agricultural households are likely to be poorer than nonagricultural households (Hill et al., 2016). In addition, agricultural income tax rules and land use fees are decided by regional (and local) governments and are levied according to landholding size. As such, they do not capture income from agricultural production or income from other productive assets for assessing the tax base. The contribution of agricultural taxes and land fees to the overall tax collection could be scaled up by introducing simple ways for estimating and collecting taxes. The agricultural income tax base can also be broadened by defining and accurately capturing commercial agriculture. Indirect Taxes Our results show that indirect taxes are mildly progressive. VAT is regressive and excise taxes are ambiguously progressive. This result is not surprising since indirect taxes are often only mildly progressive or even regressive in low income countries (Bastagli, 2015). This could result from higher tax rates being applied to those goods that are consumed more by poor households. As a result, indirect taxes played little redistributive role, despite their higher contribution to tax revenue (Table 2.1) and the tendency to finance a large part of social spending (Hill et al., 2016). Indirect taxes, however, are also poverty increasing, with the national poverty headcount rate rising by 0.60 percentage points as a result of indirect taxes (Table 4.3). The burden of indirect taxes seems to be distributed almost proportionally to market income (Figure 4.5). For instance, the bottom 40 percent account for 18 percent of the VAT burden and capture about 20 percent of the market income. Regarding excise taxes, the bottom 40 percent, which account for 20 percent of market income also pay about 21 percent of all excise taxes. The excise taxes are calculated using the old schedule.20 19 This is according to the tax schedule of 2002. The new tax rule (July 2016) increases the lower threshold to 600 ETB per month or 7200 ETB per year, which is slightly higher than the national poverty line of 7184 ETB per adult equivalent per year. 20 As stated earlier, the new excise tax policy (adopted as of February 2020) changes the excise tax calculation base from cost- based assessment to manufacturer’s selling price. Excise tax rates are also changed for most excisable items. Moreover, it changes the coverage of items and removes excise taxes for some items. 28 Figure 4.5. Incidence and Progressivity of Indirect Taxes by Market Income Decile, in Ethiopia CUMMULATIVE SHARE OF TAX (%) 100% 80% 60% 40% 20% 0% 0 1 2 3 4 5 6 7 8 9 10 MARKET INCOME DECILE Market Income Lorenz VAT Excise Indirect Taxes Population shares Source: Own calculations based on HCES and WMS (2015/16) and Administrative Data Among the excise taxes included in this analysis, excise tax on beer and fuel is progressive (yellow and blue lines in Figure 4.6). However, excise on alcohol (orange line in Figure 4.6) is ambiguous (neither progressive nor regressive) because its concentration curve crosses the market income Lorenz. Excise on alcohol is initially regressive and appears to be progressive after the fifth decile. Likewise, excise tax on tobacco is initially progressive before it turns regressive around the fourth decile. The poorest 10 percent account for only 2.8 percent of the market income, yet, pay 1.2 percent of the tobacco excise tax. The concentration curve for tobacco excise tax eventually crosses the Lorenz curve so that the poorest 40 percent account for 21 percent of the tobacco excise tax, a larger share than their 20 percent in market income. This suggests lower relative spending among the poor and higher relative spending among the middle quintiles. Our study also shows that excise taxes on tobacco products are ambiguous. Tobacco taxes are often assessed as regressive as low-income households tend to allocate a larger share of their budgets to the purchase of tobacco products. Yet, we also have to consider the negative health effects of consuming tobacco products. Given shortened life expectancy, higher medical expenses and other negative externalities associated with tobacco consumption, higher tobacco excise tax rates could serve as effective policy tools to reduce the adverse economic effects, particularly among the low-income households. Figure 4.6. Incidence and Progressivity of Excise Taxes by Market Income Decile in Ethiopia 100% CUMMULATIVE PROPORTION OF 80% 60% TAX (%) 40% 20% 0% 0 1 2 3 4 5 6 7 8 9 10 MARKET INCOME DECILE Market Income Lorenz Alcohol excise Tobacco excise Beer excise Fuel excise Excise Population shares Source: Own calculations based on HCES and MS (2015/16) and Administrative Data 29 4.3.2. Social Spending As noted earlier, poverty tends to be higher when moving from disposable to consumable income, indicating poverty increasing effects of indirect or consumption taxes. The results also show that taxes contribute to fiscal impoverishment. Although taxes are poverty increasing and hurt the poor in cash terms, they are also the main source of funding for cash transfers and access to education and health for the poor (Gebru et al., 2018; Hill et al., 2016). This section discusses the incidence and distributional effects of spending on direct transfers, indirect subsidies, and in-kind transfers for education and health services. Results from the assessment of the progressivity and pro-poorness of public spending show that direct transfers (PSNP and HFA) and public spending on primary education are progressive and pro-poor (Figure 4.7). This suggests that the poorest households benefit the most from these programs (transfers) and the share of these benefits are high for the poorest decile and tend to fall at higher market income deciles. Figure 4.7. Progressivity and Pro-Poorness of Public Spending in Ethiopia Wheat and kerosene subsidies, public spending on education (including secondary education) and health are also progressive as the share of these benefits tend to fall with a rise in income. They are, however, not pro-poor as they do not benefit the households at the lowest tail of the market income distribution more in absolute terms compared to those in the top decile. In contrast, electricity Source: Own calculations based on HCES and WMS (2015/16) and subsidy and public spending on Administrative Data tertiary education are unambiguously regressive, an indication that they disproportionately benefit the better-off compared to the poor. Direct Transfers Direct transfers through the PSNP and HFA programs are progressive, equalizing, and pro-poor. Our data show that, overall, more than 58 percent of the benefits go to the bottom 30 percent. In terms of generosity, however, direct transfers (from the PSNP and HFA) make up about 19 percent of the market income of the poorest deciles. They reduce poverty rates by 1.1 percentage points and market income inequality by 0.46 percentage points. Among the cash transfer programs, PSNP appears to be more progressive and pro-poor than HFA (Table 4.3 and Figure 4.7). While both programs direct similar share of the benefits to the poorest decile (19 percent for PSNP and 18 percent for HFA), the share that goes to the richest decile is higher for HFA (7 percent) than for PSNP (4 percent). This suggests that PSNP is better targeted to the poor than HFA. The 30 recent poverty assessment report for Ethiopia also documents that both cash transfers are well-targeted to the poor, with the targeting performance of PSNP being slightly better (World Bank, 2020). PSNP is also shown to have clear targeting rules and identification of beneficiaries that results in lower targeting errors than HFA (Coll-Black et al., 2011, 2013; Gilligan, Hoddinott, Kumar, & Taffesse, 2010). The targeting performance of Ethiopia’s PSNP is comparable or slightly better than the targeting performance of similar programs elsewhere. PSNP also appears to be more effective in reducing poverty and inequality than HFA. This difference, although not significant, could be due to the higher size of PSNP and its better targeting performance (Hirvonen et al., 2018; World Bank, 2020). This finding is in line with the results of PSNP external evaluations and the findings of the literature on HFA targeting in Ethiopia (Hill et al., 2016; Hirvonen et al., 2018; World Bank, 2020).21 Despite the good performance of PSNP in terms of targeting, its limited geographical coverage means that many poor households are not covered by the program and its redistributive effect are still limited. A recent study shows that re-channeling PSNP funds to the poor, regardless of their location (that includes woredas/districts where PSNP is not currently operational), would help to further reduce poverty, although it will not significantly affect inequality (Hirvonen et al., 2018). Indirect Subsidies Although subsidies are expected to promote equity, we find that some indirect subsidies in Ethiopia benefit the richer Ethiopians (Figure 4.7 and 4.8). Our results show that even though none of the subsidies in Ethiopia are pro-poor, wheat and kerosene subsidies are progressive. This is primarily because poorer households consume less electricity, kerosene, and wheat than richer households (World Bank, 2016). Wheat and kerosene make up a larger share of income and spending among poorer households than among richer households, and consequently, these two subsidies are progressive in relative terms. Importantly, they are also equalizing and poverty-reducing with the impact of kerosene subsidy being higher than that of wheat subsidy. The electricity subsidy is unambiguously regressive and unequalizing. In Ethiopia, poor households are unlikely to have access to electricity and electricity consumption rises substantially with increasing income, thus largely benefitting non-poor households. The richest 30 percent of the population received 70 percent of the electricity subsidy, while the poorest 30 percent received merely 9 percent. Although the electricity subsidy is regressive and unequalizing, it is also poverty reducing. The results regarding electricity subsidy present tradeoffs from a policy perspective. Our results show that electricity subsidies are captured by higher-income households, reinforcing inequality. Thus, such subsidies might distort resource allocation by encouraging excessive consumption at least in the short run. Relatedly, such kind of subsidies might also aggravate fiscal imbalance and crowding-out priority public spending instead of protecting consumers. Nonetheless, the electricity subsidies also have a 21 There could also be other mechanism through which PSNP contributes to welfare such as improving nutrition, the acquisition and protection of productive assets, increasing agricultural productivity, and spillover effects on health and education (Endale et al., 2019). 31 positive effect on poverty reduction—concentrated in urban areas—which still makes electricity reforms challenging as it might hurt the urban poor. The current energy tariff reform, however, maintains the lifeline tariff, alleviating the impacts of the reform on the urban poor. Figure 4.8. Concentration Curves for Indirect Subsidies in Ethiopia 100% Cummulative proportion of subsidy 80% 60% 40% 20% 0% 0 1 2 3 4 5 6 7 8 9 10 Market income decile Market Income Lorenz electricity kerosene wheat Indirect Subsidies Population shares Source: Own calculations based on HCES and WMS (2015/16) and Administrative Data In-Kind Transfers This study assesses the benefits of in-kind transfers in the form of public spending on education and health using government expenditure data on the various forms of education and health services. As stated earlier, in-kind transfers in education and health have substantial poverty reducing benefits although their inequality reducing effect is low. However, the use of these services is not universal and many of the poor are still excluded. Education Public education spending is progressive in relative terms (brown line in Figure 4.9). The bottom 40 percent of the population capture about 20 percent of the market income and about 30 percent of the benefits of spending on public education. This result is driven by spending on primary education, with the poorest 40 percent capturing 48 percent of the public spending on primary education. Spending on primary education as a proportion of market income is higher for poorer households: the poorest decile received 12 percent of the value of public spending on primary education compared with a 5.3 percent share that goes to the richest decile. As a result, primary education spending is progressive in absolute terms or pro-poor. 32 Figure 4.9. Concentration Curves for Education and Health Spending 100% Cummulative proportion of spending 80% 60% 40% 20% 0% 0 1 2 3 4 5 6 7 8 9 10 Market income decile Market Income Lorenz Primary Secondary Tertiary Education Health Source: Own calculations based on HCES and WMS (2015/16) and Administrative Data Public education spending in primary education is pro-poor in Ethiopia for two reasons. First, the share of school-age children (between the ages of 6 and 17) is higher among the poor as more than 45 percent of the children in primary school are from the bottom 40 percent (Figure 4.10). This suggests that the poor benefit disproportionately from public spending on primary education (conditional on children being in school). Second, the poor are more likely to be enrolled in public schools than their wealthier counterparts, particularly at the primary level (Figure 4.10). Secondary education spending is progressive in relative terms and equalizing as it makes up a larger share of market income for poor households (3.7 percent) than for rich households (0.8 percent). Though progressive, it is not pro-poor; richer households receive a larger share of the spending on secondary education (13.8 percent) than the poorest households (6.1 percent), because children from poor households are less likely to progress to secondary school. In contrast, spending on tertiary education is regressive and unequalizing. About 44 percent of the benefits of public spending on tertiary education is received by students in the richest 10 percent of the population, while only 2.5 percent of the spending goes to the poorest decile. The regressive nature of spending on tertiary education is due to the low levels of primary and secondary education completion rates among the poor that result in low tertiary education enrollment (MoE, 2017). Primary and secondary education completion, and thus tertiary education enrollment is low among children from poorer backgrounds who make slower progress through school and are more likely to drop out without completing even primary school. As long as primary and secondary completion rates for the poor do not improve, public spending on tertiary education will remain regressive as a high share of the education budget (about 46 percent) is allocated to tertiary education. 33 Figure 4.10. Education enrollment by levels and income deciles 50.0% 45.0% 40.0% 35.0% 30.0% SHARE OF ENROLLMENT 25.0% 20.0% 15.0% 10.0% 5.0% 0.0% 1 2 3 4 5 6 7 8 9 10 MARKET INCOME DECILE Primary Secondary Tertiary Public school Source: Own calculations based on HCES and WMS (2015/16) and Administrative Data In-kind transfers through public spending on education have a significant effect on poverty reduction and a small effect on reducing inequality. While the substantial poverty reducing effects of in-kind education benefits come from public spending on primary education, spending on tertiary education also contributes to the negative inequality effect of education spending. Our results show that the contribution of primary and secondary education spending to poverty reduction is significant, but their contribution to inequality reduction is considerably low. Spending on tertiary education reinforces income inequality and its marginal contribution to poverty reduction is less than that of primary and secondary education spending. Health Public health spending is progressive but not pro-poor. The in-kind health benefits received by the poorest households as a share of their market income are relatively high (6.5 percent) compared to the share captured by the richest households (1.2 percent). However, these expenditures are not pro-poor as only 7 percent of the health spending is concentrated in the poorest decile, while 13.5 percent is captured by the richest decile (Figure 4.9). The bottom 40 percent account for 38 percent of the benefits. Overall, the incidence of public health spending is progressive only in relative terms. This could be because the poorest are less likely to consult health providers and seek care due to cost or financial reasons than richest households. According to recent studies, among the lowest income quintile group, 55 percent did not utilize outpatient services, and 45 percent did not utilize inpatient services for financial reasons (World Bank, 2016, 2019). But conditional on the uptake, they are more likely to consult public facilities, particularly lower-level facilities such as health posts and health centers. In the case of Ethiopia, (rural) households receive a range of health services through the health extension agents that are present in almost all villages (Hill et al., 2016). Furthermore, our study shows that public health spending is equalizing (reduce income inequality by 0.47 percentage points) and poverty reducing (reduce poverty headcount by about 1.9 percentage points). The progressivity and redistributive effects of public health spending, 34 however, would be limited due to low government spending in the sector and limited access to high- quality health services by the poor. 5. Comparison to the 2010/11 Incidence Analysis We compare the results of the FIA of 2015/16 with FIA of 2010/11 undertaken by Hill et al. (2015 and 2016). For this purpose, we divide the comparison into three parts: (i) the incidence of poverty and inequality by CEQ income concepts in 2010/11 and 2015/16; (ii) the redistributive and poverty reducing effects of fiscal instruments (taxes and transfers) in 2010/11 and 2015/16; and (iii) the progressivity and pro-poorness of government spending in 2010/11 and 2015/16. Both fiscal incidence analyses utilized similar metrics for inequality (Gini coefficient). However, the 2011 study used the international poverty line of 1.25 USD per adult equivalent (in 2005 PPP terms) while this study uses the national poverty line (7,184 ETB per adult equivalent per year) for assessing distributional impacts. 5.1. Poverty and inequality by CEQ income concepts We find a similar pattern in poverty incidence (national headcount is used for comparison in this section) by income concepts in 2010/11 and 2015/16 (Figure 5.1). In both periods, poverty headcount at disposable income is lower than poverty headcount at market income, and the prevalence of poverty is in line with the official poverty rates in Ethiopia (29.6 percent in 2010/11 and 23.5 percent in 2015/16). In both periods, poverty headcount at disposable income is lower than poverty headcount rate at the market income, indicating the benefit of direct transfers in offsetting potential poverty increasing effects of direct taxes. However, poverty headcount at consumable income is higher than the one at market income in 2010/11; however, the headcount poverty rate at consumable income is lower than the poverty headcount at market income in 2015/16. Poverty headcount is not calculated at the final income in 2010/11. Figure 5.1. Poverty headcount (national) rates by CEQ Figure 5.2 Inequality by CEQ income concepts in 2011 income concepts in 2011 and 2016 and 2016, Ethiopia 31.2% 32.2% 30.2% 32.4% 0.360 35.0% 0.348 30.0% 25.0% 0.340 0.332 20.0% 24.2% 0.328 0.328 0.327 23.7% 23.2% 23.6% 15.0% 0.322 18.2% 0.315 10.0% 0.320 5.0% 0.305 0.302 0.302 0.0% 0.300 CONSUMABLE NET MARKET DISPOSABLE MARKET INCOME FINAL INCOME INCOME INCOME CONSUMABLE NET MARKET DISPOSABLE MARKET INCOME INCOME FINAL INCOME INCOME INCOME INCOME 2011 2016 2011 2016 Source: HCES/WMS (2015/16) and Administrative Data Source: HCES/WMS (2015/16), Administrative Data, for 2015/16 and Hill et al. (2016) for 2010/11 and Hill et al. (2016) 35 Comparing income inequality in 2010/11 and 2015/16, we find similar patterns across the different income concepts (Figure 5.2). In both periods, market-income Gini are higher than the Gini-coefficients at the other income concepts. Moreover, Gini coefficients are lowest at the final income in both periods. The results suggest that fiscal policy in Ethiopia has played an important income redistribution and inequality reducing role in 2011 and 2016. Although taxes and transfer exert different redistributive effects in the two periods, direct taxes exerted greater redistributive effects in both periods. 5.2. Progressivity and pro-poorness of transfers In figure 5.3, we compare the progressivity and pro-poorness of government transfers in 2011 and 2016. Overall, the patterns are very similar in the two periods. Public spending on primary education, PSNP transfers, and HFA appear to be progressive and pro-poor. Wheat and kerosene subsidy, public spending on education, and secondary education spending are progressive but not pro-poor. Electricity subsidy and public spending on tertiary education are regressive in both periods. Despite these similarities, direct transfers (PSNP and HFA) were more progressive in 2011 than in 2015. Nonetheless, electricity subsidy and public spending on tertiary education are more regressive in 2016 than in 2011. The former result suggests that electricity subsidies continue to benefit the rich to a larger extent than the poor, although there is no change in the tariff structure. The regressivity of public spending on tertiary education could be due to the increase in enrollment in tertiary (university) education (MoE, 2017) and the share of education expenditure on tertiary education from 2011 to 2015. Figure 5.3. Progressivity and pro-poorness of transfers, 2011 vs Comparing the progressivity of taxes 2016 (direct and indirect) in the two periods, we do not find significant differences in Wheat subsidy the patterns. Direct taxes, particularly Kerosine subsidy PIT, are progressive in 2010/11 and Electricity subsidy Health 2015/16. However, agricultural income Tertiary education tax and rural land use fee are Secondary education regressive in both periods. Indirect Primary education taxes are progressive in 2010/11 and Education 2015/16. Food Aid 5.3. Marginal contributions to PSNP inequality and poverty reduction -0.6 -0.4 -0.2 0 0.2 0.4 0.6 The results of the comparison of the 2011 marginal contributions of taxes and 2016 transfers (evaluated at consumable Linear (Market Income Gini 2016) income with respect to market income) Linear (Market Income Gini 2011) are presented in Table 5.1. The combined effects of taxes and transfers Source: Own computation based on HCES and WMS (2015/16) and Administrative Data for 2015/16 and Hill et al. (2016) for 2010/11. on reducing inequality is very similar in 2011 (1.9 percentage points) and 2016 36 (2.0 percentage points). Moving from market income to consumable income, the results show that fiscal policy (direct and indirect taxes, direct transfers, and indirect subsidies) significantly reduced market income inequality by about 2 percentage points. However, it increases poverty by 2 percentage points in 2011, but it reduces the headcount poverty rate by 0.15 percentage points in 2016. Table 5.1. Marginal Contribution of Taxes and Transfers to Inequality and Poverty Reduction Marginal contribution Fiscal intervention Gini coefficient (pp) Poverty headcount (pp) 2010/11 2015/16 2010/11 2015/16 Consumable income 1.9284 2.0060 -1.9550 0.1473 Direct taxes 0.7162 1.3923 -1.1723 -0.4843 Personal income tax 0.7216 1.4095 -1.0127 -0.4293 Agricultural income & land use fee -0.0132 -0.0167 -0.0938 -0.0203 Direct transfers 1.1812 0.4591 2.0676 1.0998 PSNP 0.9925 0.2702 1.6274 0.5521 HFA 0.1616 0.1750 0.5634 0.4671 Indirect subsidies -0.0330 -0.0528 0.3564 0.1859 Electricity -0.0455 -0.0712 0.2257 0.1552 Kerosene 0.0098 0.0184 0.1196 0.0244 Wheat 0.0022 0.0002 0.0097 0.0000 Indirect taxes 0.3391 0.0301 -3.6542 -0.6019 Final income 2.2072 2.1151 2.3172 5.5990 Direct taxes 0.7157 1.2673 -0.9589 -0.3474 Direct transfers 1.1032 0.4125 2.2000 1.0418 Indirect subsidies -0.0437 -0.0508 0.3910 0.1178 Indirect taxes 0.3364 0.0164 −3.6349 -0.5410 In-kind transfers — 0.1091 n.a. 5.4517 Education -0.0392 -0.3679 n.a. 3.4021 Primary 0.5242 0.8631 n.a. 2.1255 Secondary -0.0353 0.1670 n.a. 1.1386 Tertiary -0.5321 -1.4000 n.a. 0.1758 Health 0.3063 0.4662 n.a. 1.8967 Source: Hill et al. (2016) for 2010/11, and HCES and WMS (2015/16 ) for 2015/16 Note: n.a. = not included in analysis; — = not calculated. pp = percentage points; positive(negative) pp implies inequality/poverty reducing(increasing). When the monetized value of in-kind education and health transfers are considered (moving from market income to final income), the inequality reducing effects of the fiscal policy is of similar magnitude in both periods (about 2 percentage points). However, the fiscal policy contributes to poverty headcount reduction by about 2.3 percentage points in 2010/11 and by 5.6 percentage points in 2015/16. 37 When disaggregating the results by tax and transfer types, overall, inequality reducing effects of direct taxes (especially PIT) is higher in 2016 than in 2011; direct taxes become more redistributive and less poverty increasing in 2015/16 than in 2010/11. The milder negative effect of the PIT and indirect taxes on poverty in 2015/16 compared to 2010/11 could relate to improvements in households’ purchasing power and/or how far they are from the poverty line. Direct transfers (both PSNP and HFA) appear to be less redistributive and less poverty reducing in 2015/16 than in 2010/11. The lower distributional power and poverty reduction benefits of PSNP could be related to the decline in PSNP spending (in real terms) which did not follow similar trajectory with the number of beneficiaries (Endale et al., 2019). This reflects the erosion in safety net benefits due to inflation that results in reduction in the effectiveness of the program in reducing poverty. Indirect taxes and indirect subsidies also appear to be less equalizing and less poverty reducing in 2015/16 than in 2010/11. Indirect taxes appear to be less equalizing and less poverty increasing in 2015/16 than in 2010/11. Indirect subsidies became more equalizing and less poverty reducing in 2015/16 than in 2010/11. 6. Conclusion and policy implications This study investigates the distributional impacts of taxes and government spending in Ethiopia using survey data combined with administrative data. The study utilizes the Commitment to Equity (CEQ) methodology. Overall, the results show that fiscal policy that entails taxation and transfers (except in-kind education and health transfers) reduces inequality by about 2 percentage points and national poverty headcount rate by 0.15 percentage points. The results also highlight the importance of in-kind education and health benefits in poverty reduction by offsetting the poverty increasing effects of direct and consumption taxes. The overall incidence analysis shows that poor households are net beneficiaries of the fiscal policy that considers in-kind education and health transfers. However, the fiscal system made 12 percent of the poor poorer and pushed about 3 percent of the ex-ante non-poor population to poverty after all taxes and benefits are considered. This reflects that the transfers and benefits directed to those poor households do not compensate for the direct and indirect taxes they have paid. Beyond the aggregate impacts of the fiscal system, our results also highlight heterogenous effects of individual taxes and transfers on inequality and poverty. Taxes are progressive and equalizing, but their progressivity could be further enhanced. Among the direct taxes, personal income taxes are found to be progressive and equalizing, though it also exacerbates poverty. This is mainly due to a low cutoff income for personal income tax, prior to the 2016 reform. Agricultural income taxes and land use fees levied mostly on rural households are regressive and exacerbate poverty. In part, this reflects the fact that agricultural households tend to be poorer than non-agricultural households. It is also the case because a larger share of these taxes come from rural households who are also poor compared to urban households. Consumption taxes (VAT and excise taxes) are progressive and equalizing, but poverty increasing. This could be the case because indirect taxes are generally levied equally regardless of income level, rendering them a burden on the poor. On the spending side, direct transfers are progressive, pro-poor, and have been effective in reducing poverty and inequality. Ethiopia’s flagship Productive Safety Net Program (PSNP) and HFA contribute to 38 poverty reduction. The PSNP is more effective (slightly) than HFA. The effectiveness of direct transfers is associated with solid targeting performance (especially for PSNP) so that a large fraction of the benefits is captured by the poor. Despite this, the relatively small geographic coverage of cash transfer schemes in Ethiopia seem to limit their contribution in poverty and inequality reduction. This implies that expanding the geographic coverage of direct transfers, specifically PSNP, to other districts that host a large share of the poor could further improve their welfare enhancing benefit. Although indirect subsidies are meant to benefit the poor, our study demonstrates that they end up benefiting households in the top deciles of the market income distribution. This is particularly true for the electricity subsidy that appears to be unambiguously regressive. None of the subsidies considered in our study are pro-poor. Wheat and kerosene subsidies are progressive in relative terms with small effect on inequality and poverty reduction. Although highly regressive, electricity subsidy also has poverty reducing benefits. Since access to electricity is limited in rural areas, the poverty reducing effects of electricity subsidy is more pronounced in urban Ethiopia. These findings, to some extent, are a signal that policy action to remove electricity subsidy might hurt the urban poor, though safeguarding of the lifeline tariff partially alleviates this concern. Public spending on education and health is overall progressive and equalizing, but not pro-poor. Public spending on primary education is the most progressive, equalizing, and poverty reducing followed by spending on secondary education. Nonetheless, public spending on tertiary education is regressive although it contributes to poverty reduction. The regressivity of tertiary education spending is mainly due to low tertiary education enrollment among the poor that result from low primary and secondary education completion rates. Public spending on health is also progressive but not pro-poor. This could be due to limited use of curative health services by the poor. Redirecting spending from higher-level public health facilities to primary care facilities might improve progressivity and benefit the poor. In light of the findings of the study, we highlight the following policy options that would help to improve the contribution of the fiscal policy to poverty and inequality reduction by mitigating the negative impacts of taxes and transfers. These policy actions could help create a larger fiscal space for pro-poor spending by strengthening tax collection and improving efficiency of public spending. (a) Ensuring pro-poor tax revenue generation by reducing the burden of direct taxes on the poor and by broadening the tax base, both for employment and agricultural income taxes. This could be accompanied by simplifying the income tax system and minimizing the adverse effects of possible “bracket-creep”—inflation would cause income/earnings to rise and enter into higher tax brackets. As a result, financing public spending through taxation would be possible without worsening poverty. (b) Expanding the geographic coverage of the Productive Safety Net Program (PSNP) to poor geographic areas currently excluded and increasing its size would help to improve its poverty and inequality reducing effect. This demands additional financing from the government. Given 39 reduction in donor support, the government would need to finance this from its domestic resources. Taxation remains the main tool for financing social protection. (c) Redirecting (electricity) subsidy spending to direct transfers that could more efficiently and effectively benefit the poorest; however, this shift would need to ensure that the (urban) poor with access to electricity receive compensation. This is important since the GoE is currently introducing a new tariff and ceasing the electricity subsidy. (d) Though regressive, tertiary education will have benefits on long term economic growth through fostering technological convergence (absorption and innovation). An attempt should be made to increase the number of students from lower deciles to complete primary and secondary education to have a chance at entering tertiary education. This could also be complemented with redirecting spending to lower schooling in the near term to benefit larger portion of the population. Low performing woredas and regions could be targeted to effectively promote equity. Furthermore, partially subsidizing university education through a stipend system to poorer students could also be considered. However, this could be done in such a way that it does not create burden on the government. 40 References Bastagli, F. (2015). Bringing taxation into social protection analysis and planning. ODI Working Paper, 24. Besley, T., & Persson, T. (2009). The origins of state capacity: Property rights, taxation, and politics. American Economic Review, 99(4), 1218–1244. https://doi.org/10.1257/aer.99.4.1218 Bird, R. M., & Miller, B. D. (1989). The Incidence of Indirect Taxes on Low-Income Households in Jamaica. Economic Development and Cultural Change, 37(2), 393–409. Cardenas, H., & Whittington, D. (2019). The consequences of increasing block tariffs on the distribution of residential electricity subsidies in Addis Ababa, Ethiopia. Energy Policy, 128(February), 783–795. https://doi.org/10.1016/j.enpol.2019.01.033 Coll-Black, S., Gilligan, D. O., Hoddinott, J., Kumar, N., Seyoum Taffesse, A., & Wiseman, W. (2011). Targeting food security interventions in Ethiopia: The productive safety net programme. In ESSP II Working Paper 24. Coll-Black, S., Gilligan, D. O., Hoddinott, J., Kumar, N., Seyoum Taffesse, A., & Wiseman, W. (2013). Targeting food security interventions in Ethiopia: The productive safety net programme. ESSP Research Note 26, (June). CSA. (2016). Welfare Monitoring Survey (2016) Statistical Report: Basic Population Characteristics, Education, Health, Child Care And Breast Feeding. Duclos, J.-Y., & Araar, A. (2006). Measuring Progressivity and Vertical Equity: Measurement, Policy and Estimation with DAD. Poverty and Equity: Measurement, Policy and Estimation with DAD. Endale, K., Pick, A., & Woldehanna, T. (2019). Financing Social Protection in Ethiopia: A Long-term Perspective. OECD Development Policy Papers, 15. Foster, J., Greer, J., & Thorbecke, E. (1984). A Class of Decomposable Poverty Measures. Econometrica, 52(3), 761–766. Fullerton, D., & Metcalf, G. E. (2002). Tax Incidence. In NBER WORKING PAPER SERIES (No. Working Paper 8829). https://doi.org/10.1201/noe1420052756.ch385 Gebru, M., Remans, R., Brouwer, I., Baye, K., Melesse, M. B., Covic, N., … Vandenberg, M. (2018). Food Systems for Healthier Diets in Ethiopia Toward a Research Agenda. Gilligan, D. O., Hoddinott, J., Kumar, N., & Taffesse, A. S. (2010). Targeting food security interventions when “everyone is poor”: The case of Ethiopia’s Productive Safety Net Programme (Vol. 53). https://doi.org/10.1017/CBO9781107415324.004 Higgins, S., & Lustig, N. (2016). Can a poverty-reducing and progressive tax and transfer system hurt the poor? Journal of Development Economics, 122, 63–75. https://doi.org/10.1016/j.jdeveco.2016.04.001 Hill, R., Inchauste, G., Lustig, N., Tsehaye, E., & Woldehanna, T. (2016). Fiscal Incidence Analysis for Ethiopia. In The Distributional Impact of Taxes and Transfers: Evidence from Eight Low- and Middle- Income Countries (pp. 79–112). 41 Hirvonen, K., Mascagni, G., & Roelen, K. (2016). Linking taxation and social protection: Evidence on redistribution and poverty reduction in Ethiopia (No. Working Paper 61). Hirvonen, K., Mascagni, G., & Roelen, K. (2018). Linking taxation and social protection: Evidence on redistribution and poverty reduction in Ethiopia. International Social Security Review, 71(1), 3–24. https://doi.org/10.1111/issr.12159 IMF. (2013). Translating economic growth into higher living standards: Inclusive growth in Ethiopia. IMF Country Reports, 13/309, (October). Lustig, N. (2018). Commitment to Equity Handbook - Estimating the Impact of Fiscal Policy on In equality and Poverty (N. Lustig, Ed.). CEQ INSTITUTE AT TULANE UNIVERSITY (New Orleans) and BROOKINGS INSTITUTION PRESS (Washington, D.C.). Lustig, N., & Higgins, S. (2012). Fiscal Incidence, Fiscal Mobility and the Poor: A New Approach. Martinez-Vazquez, J. (2001). The Impact of Budgets on the Poor: Tax and Benefit Incidence. (August). Mejia-Mantilla, C., Fajardo-Gonzalez, J., Goldman, M., Jellema, J., & Renda, H. (2019). Impact of Fiscal Policy on Poverty and Inequality in Uganda: Fiscal Incidence Analysis Using the UNHS 2016/17. (October). Mengistu, A. T., Woldeyes, F., Dessie, E., Ayalew, Z., Yeshineh, A., Mainar-Causapé, A. J., & Ferrari, E. (2019). Ethiopia Social Accounting Matrix 2015/16 (Vol. 16). https://doi.org/10.2760/974668 MoE. (2017). Federal Democratic Republic of Ethiopia - Education Statistics Annual Abstract 2008 E.C. (2015/16). Moore, M. (2013). Obstacles to Increasing Tax Revenues in Low Income Countries (No. 15). National Planning Commission. (2017). Ethiopia’s Progress Towards Eradicating Poverty: An Interim Report on 2015/16 Poverty Analysis Study. (September). National Planning Commission Ethiopia. (2017). Ethiopia’s Progress Towards Eradicating Poverty: An interim report on 2015/16 Poverty Analysis Study. Addis Ababa, Ethiopia. OECD. (2017). Social Protection in East Africa: Harnessing the future. In Social Protection in East Africa. https://doi.org/10.1787/9789264274228-en Planning and Development Commission. (2018). Poverty and Economic Growth in Ethiopia (1995/96- 2015/16). UNDESA. (2019). Income inequality trends: the choice of indicators matters. United Nations Department of Economic and Social Affairs, Social Development Brief, 8(December). UNDP. (2016). Performance and Prospects of Tax Collection in Ethiopia. UNICEF. (2017a). Ethiopia - Health Sector Development Program. Retrieved from http://www- wds.worldbank.org/external/default/WDSContentServer/WDSP/IB/2007/08/20/000020439_20070 820105110/Rendered/PDF/ICR0000383.pdf UNICEF. (2017b). National Education Sector Budget Brief: 2006-2016. 2006–2016. 42 UNICEF. (2017c). National Health and Nutrition Sector Budget Brief: 2006-2016. Waiswa, R., Fekade, S., & Lake, A. (2019). Practices, Challenges and Prospects for Public Sector Taxation in Ethiopia. African Tax Administration Paper 10, (June). World Bank. (2016). Ethiopia public expenditure review 2015. World Bank. (2019). Ethiopia Public Expenditure Review: Health. World Bank. (2020). Ethiopia Poverty Assessment: Harnessing Continued Growth for Accelerated Poverty Reduction. Washington DC. 43 Appendix A. Methodological Assumptions The principal taxes currently in place in Ethiopia are personal income tax (PIT), corporate income tax (CIT), turnover tax, value-added tax (VAT), excise tax, stamp duty, withholding tax, royalty tax and dividend tax. Table A.1 summarizes the types of taxes and their rates. Table A.1. Principal taxes and tax rates in Ethiopia Principal Taxes Tax rate Corporate income tax 30% Turnover Tax (TOT) 2% and 10% Excise tax 10% - 100% Customs duties 0% - 35% Income tax from employment 0% - 35% Export tax — Withholding taxes 2% Value added tax (VAT) 15% A.1. Direct Taxes This analysis focuses on personal income tax (wage and self-employment income taxes) and agricultural income tax and rural land use fees – among direct taxes applicable in Ethiopia. Personal Income Tax (PIT) Personal income tax is payable as per Proclamation No.286/2002, and it is taxed based on a progressive rate structure with seven tax brackets. The first ETB 150 of monthly personal income is exempted from payment of income tax. Monthly income of ETB 151 and above is subject to marginal tax rates that increase progressively from 10 percent to 35 percent (Table A.2). Table A.2. Personal Income Tax Rates/Month Tax rate and Employment Income per month Marginal Tax Tax bracket Rate (%) From ETB to ETB 1 0 150 exempt 2 151 650 10 3 651 1400 5 4 1401 2350 20 5 2351 3550 25 6 3551 5000 30 7 Over 5000 — 35 Source: Proclamation 286/2002 The recent income tax reform (July 2016) raised the exemption threshold for personal income tax (PIT) from ETB 150 per month to ETB 600 per month to ease the burden of taxation on low-income households (Hirvonen et al., 2016). Changes were also made to the upper thresholds: the threshold of the monthly employment income to which the 35 percent tax rate is applied is raised from ETB 5,000 to ETB 10,000 (Income tax (amendment) proclamation No. 608/2016). Similarly, the portion of annual rental income on which the 35 percent tax is levied was increased from ETB 60,000 to ETB 138,000. This analysis uses the 44 tax brackets from 2002. For other businesses (including self-employment), the tax rates are summarized in Table A.3. Table A.3. Taxable Business Income Tax Rates Employment Income per month Tax bracket Tax rate From ETB To ETB 1 0 1800 exempt 2 1801 7800 10 3 7801 16800 5 4 16801 28200 20 5 28201 42600 25 6 42601 60000 30 7 Over 60000 - 35 A standard assumption in the CEQ framework is that taxes paid by the employer are shifted to workers in the form of lower wages. Hence, disposable income–in this case, the value of consumption–was taken as a point of departure and the tax code was then applied to arrive at market income (Lustig, 2018). We assume that disposable income reported in the HCES data is always income net of taxes and employee social security contributions, but it includes direct transfers. The employee social security contribution for formal sector employees is a flat rate of 7 percent. Therefore, we estimate the household-level PIT by applying the income tax schedule on the disposable income of individuals who were employed by formal private or public organizations. Rural individuals were assumed not to be formally employed (Hill, Inchauste, Lustig, & Tsehaye, 2015). To distribute the total PIT to wage and self-employment taxes, we calculate the share of wage earners and self-employed in the households using HCES data. For self- employed individuals, we applied the business tax schedule to determine PIT. Agricultural Income Tax and Rural Land Use Fee Income from agricultural activities are taxes based on proclamations issued by regional states, on the basis of landholding size. The tax schedule for this tax and fee is based on the “Rural Land Use Payment and Agricultural Income (Tax Proclamation No. 131/2007)”. Although tax rules vary from region to region because they are set by regional and local governments, many of the main tax schedules levy similar constant per hectare tax rates and fees regardless of land size (Hill et al., 2015). Table A.4. Land Use Fee and Agricultural Income Tax Schedule in Ethiopia Land size (hectare) Rural land use fee (Br) Income tax (Br) Total (Br) < 0.5 15 Exempted 15 0.5–1 20 20 40 1–2 30 35 65 2–3 45 55 100 3–4 65 70 135 4–5 90 100 190 >5 120 140 260 Source: Rural Land Use Payment and Agricultural Income Tax Proclamation No. 131/2007 Note: Tax rates (for the income) are different for farmers that depend on rain and for those who produce using rain and irrigation. Since irrigation coverage is too small (<5%), all taxes are calculated using these rates. 45 A.2. Indirect Taxes Indirect taxes considered in this analysis include VAT and excise taxes. In the analysis, it was assumed that households report the value of purchases inclusive of taxes. Further, it was assumed that the burden of VAT is shifted entirely to consumers, so that consumers bear the tax burden in proportion to their purchases of taxable items (Bird & Miller, 1989; Martinez-Vazquez, 2001). Value Added Tax (VAT): VAT replaces the old business tax system of commodity and service taxes including the sales tax and the withholding tax. VAT is levied at a rate of 15 percent of the value of every taxable transaction by a registered person and all imports of goods and services other than those VAT exempted items in Ethiopia. Some goods and services are charged with zero VAT rate. These include export of goods or services to the extent provided in the regulations, the rendering of transportation or other services directly connected with international transport of goods or passengers as well as the supply of lubricants and other consumable technical supplies taken on board for consumption during international flights. A wide range of products and services are exempt from VAT: 1. Financial service, 2. Imported raw materials used for production of exportable produce 3. Raw materials and packaging materials purchased domestically and used for exportable produce 4. Local or Foreign currencies and warranty distribution or importation except for cents and medals research services, 5. The import of Gold for the presentation to the National bank of Ethiopia, 6. Religious or spiritual related services given by religious institutions, 7. Educational services given by educational institutions and childcare given by kindergartens, 8. Electricity, kerosene and water supplies (does not include water processed by Factories), 9. Except for different services or commission fees, goods or services presented by postal service institutions as per the authority given by its establishment proclamation, 10. Transportation Services, 11. Education, Books, 12. Permit, license and certification payments, 13. If 60 percent of the employees are disabled the goods and services supplied by the institution employing these disabled individuals, 14. Food items, 15. Goods like sealing plastic bags, sewing materials and fertilizers for making Insecticide-treated bed nets for the prevention of malaria, 16. Transactions of pickles, wet blues and crust made by leather processing factories, 17. The import of chemically processed clothes used for the sewing of Insecticide-treated bed nets for the prevention of Malaria, 18. Government imported wheat, 19. Palm oils used for food, 20. Sale of Milk and bread, 21. Drugs, medical supplies and equipments, 22. Agricultural fertilizers, pesticide chemicals, selected seed (improved seeds and seedlings), 23. Pension fee Services 24. The sale of Airplane tickets by travel agencies. 25. “Injera” 26. Publication and printing of books. 27. Sale of Processed leather to Shoe factories by leather processing factories. 28. Manufacturing of Stoves Source: Ethiopian Revenues and Customs Authority (ERCA) 46 Excise Tax: Excise tax in Ethiopia is imposed on a range of selected consumer goods, whether produced locally or imported, such as, luxury goods and goods hazardous to health. The excise tax rate ranges from 10 percent for textiles and most other goods, to as high as 100 percent for alcoholic beverages (Table A.6). It is payable in addition to VAT, on goods mentioned under the schedule of the proclamation when imported or when produced locally at the rate prescribed in the schedule. The base of computation of excise tax is the cost of production for locally produced goods and the cost, insurance and freight (CIF value) for imported goods. For locally produced goods, excise tax will be paid by the producer and by the importer for imported goods. Table A.6. Excise Tax Rate for Different Product Types S.No. Type of Product Tax Rate (%) 1 Any type of sugar/In solid form excluding Molasses 33 2 Drinks All types of soft drinks/except Fruit/ Juices 40 Powder soft drinks 40 Water bottled or canned in a factor 30 Alcoholic drinks All types of beer & stout 50 All types of wine 50 Whisky 50 Others alcoholic drinks 100 3 All types of pure Alcohol 75 4 Tobacco & Tobacco products Tobacco leaf 20 Cigarettes, Cigar, Cigarillos, pipe Tobacco snuffs and other tobacco products 75 5 Salt 30 6 Fuel-Super Benzene, Regular Benzene, Petrol, Gas-online and other motor spirits 30 7 Perfumes and toilet waters 100 8 Textile and Textile products Textile fabrics, knitted or woven of natural silk, Rayon, nylon wool or other similar 10 material Textile of any type partly or wholly made from cotton which is gray, white, dyed or 10 printed, in pieces of any length or width /except mosquito net and "Abudgedi"/ and including blankets, bed sheets, counterpanes, towels, table clothes and similar articles Garments 10 9 Disk washing machines of a kind for domestic use 80 10 Washing machines of a kind for domestic purpose 30 11 Video decks 40 12 Television and video cameras 40 13 Television broadcast receivers whether or not combined with gramophone, radio, or 10 sound receivers and reproducers 14 Motor passenger cars, station wagons, utility cars, and land rovers, tips pickups, similar vehicles/including motorized caravans/ whether assembled, tighter watt gaur appropriate initial equipment. Up to 1,300 C.C 30 From 1,301 C.C up to 1800 C.C 60 Above 1,800 C.C 100 15 Carpets 30 16 Asbestos and Asbestos products 20 47 17 Clocks and watches 20 18 Dolls and toys 20 Source: Excise Tax Proclamation No. 307/2002, Ethiopian Revenues and Customs Authority (ERCA): www.erca.gov.et A.3. Direct Transfers This analysis focuses on two cash transfer programs: productive safety net program (PSNP) and HFA. While PSNP aims to cover the chronically poor and food-insecure, HFA aims to support populations with acute food needs, mainly due to shocks. HFA entails in-kind and cash transfers to individuals or households for the purpose of increasing the quantity and/or quality of food consumption in anticipation of, during, and in the aftermath of a humanitarian crisis. According to HCES (2015/16), the total number of PSNP beneficiaries is 7,388,895 which is slightly lower than the government figures of 7,985,832.22 The absolute coverage of PSNP was 8.7 percent in 2016. It covers 13 percent of the poorest quintile and 4 percent of the higher-income groups of the Ethiopian population in 2016. According to government budget estimates, the total budget for PSNP was 13.8 billion ETB. Total expenditure on PSNP in 2015/16 was 7.6 billion, of which about 5.5 billion was transferred to households. The number of HFA beneficiaries vary wildly from year to year. The number of beneficiaries increased from about 2.8 million in 2010/11 to about 10 million in 2015/16, mainly due to the 2015 El Nino drought. Though accurate data on spending on emergency relief is difficult to come by given its largely off-budget nature, it is estimated that over US$1 billion was spent on HFA during the 2015/16 drought year (World Bank, 2020). The incidence analysis for direct transfers (PSNP and HFA) presented in this study is based on the 2016 HCES. Using the survey information, we define PSNP/HFA beneficiary those that were beneficiary during the survey year (2015/16). Although the survey identifies households that benefited from the PSNP and households that receive HFA, it does not provide information on the actual amounts received by the households. This analysis used the PSNP and HFA transfers calculated and used in the Poverty Assessment for Ethiopia (World Bank, 2020). For PSNP, annual value of transfers is calculated by aggregating the benefits from the two PSNP modalities/beneficiaries: Direct Support (DS) or Public Works (PW). Since HCES does not distinguish between DS and PW beneficiaries, the PSNP benefits are calculated by assigning PSNP households to DS or PW modality by constructing a statistical model based on the 2016 Ethiopian Socio-Economic Survey (ESS) that contains information on whether a PSNP household is in DS or PW (World Bank, 2020). The annual value of benefits received by a PW households is computed by multiplying the daily wage rate by the number of days worked per month, and this is multiplied by 6 since PW beneficiaries work only for 6 months per year. Likewise, the annual value of transfers to DS is calculated by multiplying the daily wage rate by the number of eligible days per month and by 12. Since there is no comprehensive data regarding HFA transfers, the HFA benefit is calculated with an assumption that they are on average 15 percent higher than PSNP benefits (ibid). The targeting performance of PSNP and HFA is assessed based on a variety of indicators of well-being. The PSNP beneficiary targeting for PSNP was clear. Targeting happens in three stages. First, the Federal Ministry of Agriculture (MoA), following consultations with the regions, defines the caseload per woreda 22 PSNP IV (Revised) Annual Work Plan and Budget for EFY 2008 (2015/16) p. 31, p. 36 48 based on historical receipt of HFA in the woreda. Second, woredas determine kebele selection and caseloads. Finally, households are selected at kebele level though community-based targeting. Overall, PSNP is well targeted towards the poor. About one-third of the PSNP beneficiaries were in the poorest quintile in terms of consumption in 2016, and 60 percent of the beneficiaries were in the bottom 40 percent. In terms of binary poverty status, 39 percent of the PSNP beneficiaries were below the national poverty line in 2016. PSNP is also well targeted on other non-monetary indicators of living standards. PSNP beneficiaries are more likely to report food shortages, have lower assets (including livestock), have limited access to infrastructure, and live in dryer places with less vegetation and less suitable for rainfed agriculture. Despite this solid targeting performance, the following issues remain: First, there is substantial under-coverage, with PSNP covering only 13 percent of the poor in 2016. Second, there was considerable inclusion of the upper quintile: close to 10 percent of beneficiaries were in the top consumption quintile in 2016, and 21 percent were in the upper 40 percent of the consumption distribution. Third, regional distribution of PSNP beneficiaries does not seem to align to any indicator of monetary welfare or food security. HFA targeting was also progressive in 2016 with 31 percent of the HFA beneficiaries coming from the bottom consumption quintiles and another 28 percent from the second quintile. Almost 30 percent of the HFA beneficiaries in 2016 were in the top two quintiles in terms of pre-food-aid consumption, indicating a fair degree of leakage (fit was 21 percent for PSNP). In terms of binary poverty status, 37 percent of the HFA beneficiaries were below the national poverty line in 2016. Compared to non-beneficiaries, HFA beneficiaries live in more remote places, are more likely to face food shortages, have low durable assets, and live in less green places. Despite an overall good targeting of HFA towards the poor, there were substantial inclusion errors, with a considerable share of better-off households benefitting from HFA. This is mainly due to targeting of HFA in woredas where PSNP is not active. There was also large unmet demand due to substantial disparities between needs and allocations. A.4. Public Education and Health Expenditure The benefits of public education expenditure were allocated to households using a cost-of-production approach - by dividing total education spending at a given level of the education system by the number of students enrolled in the corresponding. The unit costs of primary, secondary, and tertiary education were obtained by dividing the total public spending by the number of students enrolled in each level. The Welfare Monitoring Survey (WMS) data are used to elicit enrollment in education by level (primary education, secondary education and TVET, and tertiary education) and type of provider (public or private). Public education spending for 2015/16 is obtained from MoFEC. Education expenditure at the three levels are computed using data from the administrative data on education expenditure and the share of expenditure at each level (figure A.1). The monetized value of the in-kind education transfer at the household level is determined by multiplying the number of children enrolled in primary, secondary, and tertiary education in 2015/16 by the unit costs, considering only education provided by the government. 49 Figure A.1. Distribution of total public education However, there could be several caveats spending by level regarding the allocation of per-student 2% education expenditure to individual households using the cost-of-production approach. The cost of production method implicitly assumes that 27% 46% the cost of production equals the value of consumption, an assumption that will almost 18% always result in an overestimate of the effect of public education expenditures on economic welfare. Moreover, it is assumed that the value 3% 4% Pre-primary Primary of services is constant across users. This Secondary Post-secondary non-tertiary assumption is violated if, for instance, in the case TVET Tertiary of education, students from poor families attend public schools that have fewer resources. Dividing total public education spending by level to the number of students estimated from the survey will likely result in an overestimation of the redistributive effect. This is because they are obtained from the government budget or cost of providing these services reported in the national accounts. For allocating in-kind health benefits, total public health spending from MoFEC is distributed to all individuals who received public health services as recorded in the WMS. This study does not distinguish between curative and preventive health service spending. Thus, the total health spending is distributed to all households equally. A.5. Indirect Subsidies The price subsidies included in the study are in the form of cross subsidies on three products: electricity, kerosene and wheat. Water, housing and agricultural subsidies are not included due to data limitations. Electricity: Electricity production and distribution is entirely undertaken by the Ethiopian Electric Power Corporation (EEPCO), an enterprise owned by the federal government. The government has a policy of universal subsidy to the electricity sector, and the tariff rates have been very low compared to other countries and in relation to the cost of production. The electricity tariff and subsidy rates depend on the amount of electricity consumed. Tariff rates are progressive with the increase in kwh used, aimed at easing the burden on poor households that are usually believed to use low amount of electricity per household (Table A.7). Electricity service coverage increased from 41 percent in 2009/10 to 60 percent in 2014/15. Table A.7. Tariff and Subsidy for Household Electricity Consumption in Ethiopia Tariff w/o Monthly Tariff Subsidy Tariff (Block) subsidy USD consumption (Kwh) (Br/kWh/mo.) (Br/kWh/mo.) (Br/kWh/mo.) 1 0-50 0.273 0.967 0.694 0.0132 2 51-100 0.356 0.967 0.611 0.0172 3 101-200 0.499 0.967 0.468 0.0241 4 201-300 0.55 0.967 0.417 0.0266 50 5 301-400 0.567 0.967 0.400 0.0274 6 401-500 0.588 0.967 0.379 0.0284 7 > 501 0.694 0.967 0.273 0.0336 Source: Cardenas & Whittington (2019) Kerosene: Prices of different petroleum products are determined and regularly revised by government mainly based on import cost, taxes and contribution to the ‘petroleum price stabilization fund’. The petroleum stabilization fund is an account built by charging petroleum products during normal periods and the fund is used to smoothen domestic prices during international oil prices shocks. Currently, petroleum in general is not subsidized in Ethiopia. Within petroleum products, however, there is a cross- subsidy, kerosene being subsidized by other petroleum products mainly benzene. Kerosene is subsidized in the form of exemption from VAT and direct funding from the petroleum stabilization fund. In 2015/16 (first quarter), kerosene was subsidized from the oil stabilization fund by ETB 3.72 per liter which is equivalent to 26 percent of the Addis Ababa retail price (14.13 ETB/liter). Wheat: The GoE distributes imported grain in Addis Ababa at a subsidized price to reduce the effect of food inflation on the urban poor. This is primarily done through the Ethiopian Grain Trade Enterprise (EGTE). However, the transfer is not targeted, and sales are rationed to all households of the city through local administrative units (kebeles). Wheat was subsidized by government to reduce the effect of food inflation on the urban poor through a program of import and distribution of wheat in Addis Ababa at a subsidized price. The estimated subsidy in 2015/16 was 262.21 ETB per quintal, which is about 32 percent of the cost. This transfer will be imputed by distributing the total transfer/subsidy to all households of Addis Ababa since there was no targeting in the program. 51