Fiscal Incidence in Kyrgyzstan: A Commitment to Equity Analysis 2 ACKNOWLEDGMENT This report is part of the Kyrgyz Republic Poverty and Equity Program led by Saida Ismailakhunova (Sr. Economist, Task Team Leader). The report is prepared by Saida Ismailakhunova (EECPV, Sr.Economist), Gleb Shymanovich (Consultant), and Kateryna Bornukova (Consultant). The team benefited from advice and contributions of Sarosh Sattar (EA2PV, Sr.Economist), Alan Fuchs (EECPV, Sr.Economist), (Peer Reviewers). The fiscal incidence analysis has been conducted in coordination with government counterparts, development partners with earlier PowerPoint presentations and workshops used to support a dialogue on social and tax policy reforms that are on the agenda of the Kyrgyz Republic. The report was prepared under the guidance of Bolormaa Amgaabazar (Country Manager of the Kyrgyz Republic, ECCKG), Salman Zaidi (ECA Practice Manager, Poverty and Equity GP) The financial and technical support by DFID is gratefully acknowledged. The team is grateful to the National Statistics Committee for the access to the Kyrgyz Integrated Household Survey that underlay Commitment to Equity Analysis. The team is also thankful for consultations with government officials of the Kyrgyz Republic, primarily with the representatives of the Ministry of the Economy, Ministry of Finance, Ministry of Labor and Social Development, Social Fund, the National Statistics Committee, State Tax Service. The team also benefited from presentations of the preliminary results at the University of Central Asia and comments received from its participants and Roman Mogilevsky in particular. Any errors and omissions remain the responsibility of the authors. 3 Contents Executive summary ........................................................................................................................ 7 1. Introduction ............................................................................................................................... 9 2. Fiscal interventions and risk of poverty .......................................................................................... 9 3. Methodology of the analysis ....................................................................................................... 13 3.1. Data available and CEQ approach to income concepts construction ............................................ 13 3.2. Direct taxes........................................................................................................................ 15 Taxation of employees ........................................................................................................... 15 Taxation of entrepreneurs ....................................................................................................... 17 Taxation of farmers................................................................................................................ 18 3.3. Indirect taxes...................................................................................................................... 18 VAT .................................................................................................................................... 18 Sales tax............................................................................................................................... 19 Import duties ........................................................................................................................ 20 Excises ................................................................................................................................ 20 3.4. Contributory pensions.......................................................................................................... 21 3.5. Direct transfers ................................................................................................................... 22 3.6. Indirect subsidies ................................................................................................................ 23 3.7. In-kind transfers: Healthcare................................................................................................. 24 3.8. In-kind transfers: Education.................................................................................................. 25 4. Results and discussion ............................................................................................................... 26 4.1. Main results ....................................................................................................................... 26 4.2. Distributional impact and marginal contributions of fiscal interventions ...................................... 30 4.3. Efficiency of interventions.................................................................................................... 32 4.4. Regional differences............................................................................................................ 34 4.5. Targeting the vulnerable groups ............................................................................................ 37 4.6. Cross-country comparisons................................................................................................... 39 5. Conclusions ............................................................................................................................. 41 References................................................................................................................................... 42 Annex. Incidence of pensions and benefits by deciles based on market income (pre-fiscal) concept........... 44 4 List of figures Figure 1. General government expenditure by function of government in 2016, % of GDP ...................... 11 Figure 2. General government expense in 2016, % of GDP ................................................................. 12 Figure 3. General government revenue in 2016, % of GDP ................................................................. 12 Figure 4. Construction of income concepts ....................................................................................... 15 Figure 5. Incidence of direct taxes on formally employed by deciles of population distinguished based on market and disposable income, % of income and USD PPP per day...................................................... 16 Figure 6. Incidence of direct taxes on entrepreneurs by deciles of population distinguished based on market and disposable income, % of income and USD PPP per day ................................................................ 17 Figure 7. Incidence of direct taxes on entrepreneurs by deciles of population distinguished based on market and disposable income, % of income and USD PPP per day ................................................................ 18 Figure 8. Incidence of VAT and sales tax by deciles of population distinguished based on disposable and consumable income, % of income and USD PPP per day .................................................................... 20 Figure 9. Incidence of import duties and excises by deciles of population distinguished based on disposable income, % of income and USD PPP per day ..................................................................................... 21 Figure 10. Incidence of contributory pensions by market income deciles of population, % of income and daily 2011 PPP USD............................................................................................................................. 22 Figure 11. Incidence of direct transfers by market income deciles of population, % of income and daily 2011 PPP USD .................................................................................................................................... 23 Figure 12. Incidence of indirect subsidies by disposable income deciles of population, % of income and daily 2011 PPP USD............................................................................................................................. 24 Figure 13. Incidence of indirect subsidies by disposable income deciles of population, % of income and daily 2011 PPP USD............................................................................................................................. 25 Figure 14. Incidence of indirect subsidies by disposable income deciles of population, % of income and daily 2011 PPP USD............................................................................................................................. 26 Figure 15. Distribution of gains and losses by income bins set in USD PPP ........................................... 29 Figure 16. Change from market (pre-fiscal) to disposable average regional income, by market income, daily 2011 PPP USD per capita .............................................................................................................. 36 Figure 17. Change from market (pre-fiscal) to disposable average regional income, by market income, daily 2011 PPP USD per capita .............................................................................................................. 39 Figure 18. Effects of redistribution through direct taxes and transfers across countries ............................ 40 Figure 19. Effects of redistribution by income concepts across countries; changes in Gini index ............... 40 List of tables Table 1. Main poverty and inequality indicators by income concepts ...................................................... 8 Table 2. Revenues and expenditures of state budget, % of GDP ........................................................... 10 Table 3. Revenue and expenditure of social fund in 2016, % of GDP.................................................... 11 Table 4. Poverty and inequality indicators ........................................................................................ 13 Table 5. Tariff rates of the modelled indirect taxes, % of purchaser prices ............................................. 19 Table 6. Main poverty and inequality indicators by income concepts .................................................... 27 Table 7. Decompositions of inequality changes into vertical and horizontal equity components ................ 27 Table 8. Incidence of net effects from fiscal interventions in relation to market income by deciles ............ 28 Table 9. Fiscal gains to the poor and fiscal impoverishment in relation to market income......................... 29 Table 10. Progressivity of taxes and transfers in relation to income concepts ......................................... 31 Table 11. Marginal contributions to inequality and poverty ................................................................. 32 Table 12. Efficiency measures for transfers, subsidies and in-kind programs.......................................... 34 5 Table 13. Marginal contributions of pensions and benefits to inequality and poverty reduction and their efficiency .................................................................................................................................... 34 Table 14. Fiscal programs by regions, relative to the country average ................................................... 35 Table 15. Pensions and benefits reported by region, USD cents PPP per day per capita ........................... 36 Table 16. Fiscal programs for the households by gender and family composition, relative to the country average, per capita ........................................................................................................................ 37 Table 17. Fiscal programs for the vulnerable groups by households with children, relative to the country average, per capita ........................................................................................................................ 37 Table 18. Fiscal programs for the households with low number of earners, relative to the country average, per capita.......................................................................................................................................... 38 Table 19. Fiscal programs for the households with different vulnerable groups, relative to the country average, per capita .................................................................................................................................... 38 Table 20. International comparisons by programs, Gini marginal effects and Kakwani indices ................. 41 6 Executive summary The scale of fiscal interventions into the economy of Kyrgyzstan is relatively high compared to the other countries of the Central Asia region, Caucasus countries or Southeastern Europe . The state supports economy at large extent and bears high social expenditures. As a tax base is rather narrow due to the large informal sector, it results in a significant budget deficit and a risk of public debt accumulation. Hence, there is a need to increase efficiency of the public finance in order to cut the deficit without hampering economic growth and leaving behind the vulnerable social groups. Significant public expenditure of Kyrgyzstan has not resulted in poverty reduction until 2016 . Fall in poverty risk in 2016 occurred largely in urban area and Bishkek city in particular. Inequality level is stable and rather low if measured based on consumption aggregate. It is due to the high density of distribution of population by consumption around the poverty line. Simultaneously, level of income inequality is substantially higher, stressing the role of remittances and informal employment in sustaining welfare of population. The role of fiscal policy in poverty and inequality reduction can be examined by means of CEQ analysis. It is designed to shed light solely on the redistributive properties of public expenditures and taxes. This is quite a specific, narrow view that does not take into account other factors. The core element of CEQ analysis is calculation of income concepts and estimation of related inequality indicators and absolute poverty headcount . There are four main income concepts considered within CEQ analysis: market income, disposable income, consumable income and final income. Disposable income corresponds to the consumption aggregate used by Kyrgyzstat for households’ welfare analysis. Market income is estimated by subtracting direct transfers and adding direct taxes to disposable income, either including pensions and related social contributions into analysis (pensions are considered as a government transfer, PGT approach) or not (pensions are considered as a deferred income, PDI approach). Consumable income is calculated as disposable income minus indirect taxes plus utility subsidies. In order to obtain final income in-kind transfers in the form of public expenditure on health and education are added to consumable income. Results of the CEQ analysis applied to the Kyrgyzstan integrated households survey of 2016 show that the fiscal system in Kyrgyzstan redistributes income in favor of the poorest deciles, while net donors are largely households from the richest two . Still, there are inevitably those who face impoverishment due to fiscal interventions. Around 40% of officially poor population suffers from income reduction due to fiscal policy. Redistribution by the components of the tax-benefit system as a share of final income. Deciles are based on market income 100 % of final income 80 60 40 20 0 -20 -40 -60 poorest 2 3 4 5 6 7 8 9 richest direct taxes social contributions direct transfers pensions indirect taxes indirect subsidies inkind transfers total 7 Source: own estimates based on CEQ methodology. Table 1. Main poverty and inequality indicators by income concepts Market Income Disposable Consumable Final PGT PDI Income Income Income Gini 0.316 0.234 0.214 0.219 0.211 90/10 4.623 2.812 2.519 2.603 2.495 Poverty headcount index by poverty line National extreme poverty line 11.2 1.8 0.8 0.7 National poverty line 37.4 26.1 25.4 27.5 Source: own estimates based on CEQ methodology. Direct fiscal interventions (direct taxes and direct transfers plus pensions) reduce inequality measured by Gini index from 0.316 to 0.234 and national poverty measure from 37.4% to 26.1% . This effect is high if compared to other countries with similar income level. Pensions have the largest effect on poverty reduction and inequality reduction. All other direct transfers with exception of stipends are progressive, i.e. favoring population from lower income bins. Direct taxes are largely equalizing, with exception of the land tax and partly PIT due to the rule of the minimum tax base. Indirect taxes and subsidies, on the other hand, increase inequality due to the regressive nature of indirect utility subsidies. In their turn, in-kind transfers contribute to further reduction of consumption inequality due to progressive incidence of public expenditures on education. Efficiency of social policy spending if measured as a ratio of poverty and/or inequality reduction to the volume of related expenditure is high for pensions and targeted social assistance programs. Categorical benefits provided on monthly basis appear to be efficient largely in terms of extreme poverty reduction. Privileges and stipends have very low efficiency in terms of poverty reduction, and stipends even increase inequality. Indirect subsidies are highly inefficient although elimination of electricity subsidy bears risk of substantial poverty increase. Significant share of population has disposable income only slightly exceeding poverty line, thus being sensitive to any adverse shock. Public expenditure do not target vulnerable social groups well . Households with children in general have lower than average access to direct transfers and indirect subsidies. They benefit more than average only from education expenditures. Unemployed get surprisingly low amounts of direct transfers as well, and poor access to health services. Informally employed and labour migrants benefit slightly less than population on average from the direct transfers, pensions and indirect subsidies, but they do not pay direct taxes. It implies that fiscal system redistributes in favor of informal sector, creating disincentives for carrying formal activities. Fiscal interventions reduce disparities across regions. Within the modelled flows of public revenue and expenditure, Bishkek is the net donor among the regions; the rest of the regions are beneficiaries with the benefit size dependent negatively on the pre-fiscal income. This effect is generated by direct transfers while indirect subsidies are provided largely to Bishkek. Still there are disparities across regions in coverage of direct transfers as south parts of the country and rural areas in particular benefit substantially more from them than other regions. Hence, significant public expenditure on social policy contribute a lot to inequality reduction, but efficiency of some programs is questionable either due to the size of the benefits or their regressivity. Furthermore, there are vulnerable social groups that are left behind. It stresses need to improve the system of social assistance in terms of better coverage of the households with children, especially in urban areas, and provision of support to the working-age population facing difficulties with integration into the labour market. 8 1. Introduction 1. The purpose of this paper is to apply the Commitment to Equity (CEQ) analysis to the fiscal policy in Kyrgyzstan . The CEQ analysis focuses on the redistribution effects of the taxes and expenditures, with an aim to answer two questions: (1) how much the fiscal policy decreases inequality; and (2) does the fiscal policy increase or decrease poverty? 2. Kyrgyzstan is a lower middle-income country in Central Asia, with the population of 6.2 mln. According to national definition, 25.4 per cent of the population lived below the poverty line in 2016. The economy is quite fragile and exposed to external shocks, with 30% of GDP coming from remittances, and 10% of GDP coming from a gold mine. The latest crisis hit the Kyrgyzstan economy in 2014-2016, when the Russian economy was experiencing difficulties. In the years of the crisis the poverty rates were above 30%. 3. How redistributive are the fiscal policies is a question of particular importance to Kyrgyzstan. On the one hand, relatively high budget expenditure and low tax compliance due to the high share of shadow economy lead to fiscal deficits and reliance on external financing. In this regard, the efficiency of public finance becomes of high importance. Kyrgyzstan has taken important steps to optimize its pension system (by introducing different pillars), and by reforming healthcare (which currently requires copayment schemes along with the system of subsidies). However, further steps are necessary to make the budget sustainable over the business cycle. On the other hand, a large share of population is poor (or lives just above the poverty line), underlining the importance of poverty-alleviating policies. 4. The CEQ analysis sheds light on the redistributive properties of fiscal programs, on their ability to fight poverty and inequality . This is quite a specific, narrow view which does not take into account other factors. For example, it does not take into account the fact that some fiscal programs may aim to encourage or discourage certain behavior; or, like expenditure on health and education, may pursue the development of human capital without much regard to redistribution. Moreover, the analysis focuses only on policies which come in form of taxes, transfers, subsidies or in-kind transfers (such as health or education), possibly missing other important redistribution policies. Hence, the results of the CEQ analysis may not be viewed as conclusive recommendations, but rather as a source of useful and sometimes crucial information for the optimization of the fiscal policies. The CEQ analysis is particularly relevant for the fiscal policies aimed at reducing inequality and poverty: pensions, direct transfers, subsidies, “luxury” taxes. 5. Our results show that the fiscal system in Kyrgyzstan redistributes income in favor of the first two deciles, while net donors are largely households from the last two deciles. Direct fiscal interventions (direct taxes and direct transfers plus pensions) reduce Gini inequality index from 0.316 to 0.234 and national poverty measure from 37.4% to 25.4%. Pensions have the largest effect on poverty reduction and inequality reduction. Indirect taxes and subsidies, on the other hand, increase poverty and inequality due to the regressive nature of indirect utility subsidies. Pensions and social assistance programs are well-targeted and efficient. However, the fiscal system does not target well the unemployed and families with children. 6. The paper is structured as follows. Section 2 describes the current state of the fiscal system in Kyrgyzstan, and main trends in poverty and inequality. We describe the methodology of the CEQ analysis and resulting allocation of fiscal programs in Section 3. Section 4 presents and discusses the results, and Section 5 concludes. 2. Fiscal interventions and risk of poverty 7. The scale of fiscal interventions into the economy of Kyrgyzstan is relatively high if compared to the other countries of the Central Asia region, Caucasus countries or Southeastern Europe (see Figures 1–3). The state supports economy at large extent and bears high social expenditures, which results in budget deficit, taking into account narrow tax base due to the large informal sector. Level of the state budget 9 deficit has been above sustainable for the last decade with exception of 2013 –2015 (Table 1), which implies risk of public debt accumulation. Its current burden of 58.1% of GDP (in 2016) is close to average for lower middle income countries. This level of public debt is considered manageable in the long-run by the IMF, but it is “highly sensitive to shocks to real GDP growth and the exchange rate or to failure to reduce the primary deficit over the medium term” (IMF, 2017). Hence, public debt sustainability demands increase in public finance efficiency in order to cut the deficit without hampering economic growth and leaving behind social vulnerable groups. 8. The tax structure of Kyrgyzstan state budget revenue differs from average one for the region which is rooted in specifics of its economy and taxation . First, the role of income taxes in generating revenues of state budget is rather small, compared to the other countries of the region (see Figure 3), especially if tax obligations of Kumtor (the gold mining company) are not taken into account (Table 1). One of the reasons is low tax rates for corporate and personal income taxes (10% for both). The other is significant part of shadow and informal economy, which reduces tax base and complicates tax rates increase. Second, taxes on property also play marginal role in public sector revenue due to minor volume of assets. Most of revenue comes from land tax and tax on vehicles. Third, taxes on goods and services generate reasonable volume of tax revenue, but it is still lower than in most of CEE countries. It is explained by low VAT rate and significant number of exemptions from VAT for domestic goods and services. Consequently, VAT is largely collected from imports, which is supported by high ratio of imports to domestic consumption1. In addition to VAT, goods and services are also taxed with the sales tax. Despite low rates, its cascade nature ensured significant volume of tax revenue accumulated in the state budget. Long anticipated gradual abolishment of sales tax explains reduction of its contribution to the public sector revenues in 2016 –2017. Simultaneously, the role of excises has significantly increased recently in accordance with global practices and following policy of unification of excise rates with other EEU (Eurasian Economic Union) countries. Forth, taxes on foreign trade generate significant part of state budget revenue, which is typical only for oil exporting countries of the region. Since 2015, main volume of revenue from taxes on foreign trade is generated by customs duties redistributed within EEU. Fifth, fees and copayments for education and healthcare services provided to population contribute much to the revenue of state budget, which explains significant share of non-tax revenue within the state budget of Kyrgyzstan. Another factor is significant volume of transfers provided to Kyrgyzstan by international financial organizations. Table 2. Revenues and expenditures of state budget, % of GDP 2009 2010 2011 2012 2013 2014 2015 2016 2017 Total revenue 27.65 26.33 27.23 28.02 28.69 29.81 29.83 27.43 30.29 Tax revenue 17.94 17.86 18.54 20.59 20.50 20.62 19.66 19.69 20.95 Personal income tax (on кesidents) 1.72 1.72 1.70 1.92 1.78 1.78 1.81 1.80 1.86 Corporate income tax 0.86 0.88 0.99 1.32 1.20 1.07 0.95 0.77 0.87 Revenues from patenting 0.84 0.68 0.53 0.44 0.48 0.50 0.49 0.46 0.47 Tax on income of Kumtor 0.97 1.98 2.12 1.46 1.00 1.15 1.36 1.17 1.31 Land tax 0.46 0.36 0.27 0.24 0.25 0.23 0.22 0.21 0.21 Tax on property 0.23 0.39 0.32 0.31 0.33 0.32 0.31 0.32 0.32 VAT on local products 2.28 1.92 1.87 2.30 2.26 2.19 2.13 2.51 2.55 VAT on imported products 4.41 4.71 5.25 6.00 6.21 5.96 5.59 5.74 6.60 Sales tax 1.95 1.68 1.44 1.61 1.83 1.81 1.74 1.26 0.82 Excises 0.83 0.77 0.76 0.91 1.15 1.58 1.80 1.90 1.93 Taxes on foreign trade 2.06 1.97 2.50 3.04 3.35 3.44 2.25 2.90 3.34 Other taxes 1.33 0.81 0.78 1.03 0.66 0.58 1.00 0.66 0.67 Social contributions 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.01 0.49 Transfers 5.05 3.19 3.06 1.81 2.59 2.49 2.16 2.07 2.71 Non-tax revenue 4.50 4.99 5.47 5.56 5.56 6.68 7.97 5.63 6.13 incl. income from paid services 2.22 2.10 1.92 1.96 2.23 2.36 2.49 2.29 1.96 Income from sale of non-financial assets 0.16 0.29 0.16 0.08 0.04 0.01 0.03 0.02 0.01 Total expenditures 29.10 31.21 32.01 34.54 29.35 30.27 31.26 31.81 33.63 General public services 4.14 3.72 3.54 3.53 3.03 3.11 3.27 3.27 3.46 1 In 2016 ratio of imports to domestic demand was 0.52, see http://stat.kg/ru/statistics/download/dynamic/336/. 10 Safety and defense 3.18 3.85 3.40 3.21 3.20 3.38 3.45 3.34 3.34 Economic affairs 3.98 4.90 5.44 7.15 1.08 1.10 1.23 1.29 1.30 Environmental protection 0.29 0.26 0.20 0.16 0.14 0.15 0.16 0.16 0.14 Utilities 1.21 1.14 1.01 1.11 1.00 0.99 0.96 1.06 1.09 Health 2.89 2.91 3.18 3.66 3.33 3.00 2.96 2.80 3.27 Cultrure and recreation 0.69 0.71 0.78 0.78 0.76 0.71 0.74 0.79 0.73 Education 5.71 5.44 6.38 6.98 6.11 5.60 5.91 6.37 6.54 Social policy 2.78 5.03 4.96 5.63 5.73 5.81 5.76 5.24 5.44 Purchase of non-financial assets 4.24 3.27 3.13 2.32 4.97 6.42 6.81 7.47 8.31 Proficit/deficit -1.45 -4.89 -4.78 -6.52 -0.66 -0.47 -1.43 -4.38 -3.34 Source: National Statistical Committee of the Kyrgyz Republic. 9. Consequently, structure of public sector expenditures of Kyrgyzstan also differs from average one for the CEE region (see Figures 1–2). Fiscal interventions have an extra shift on support to the economy partly on a back of the policy aimed at recovery of the southern regions suffered from the unrest of 2010 and availability of donor’s financing for investments in infrastructure . Until 2013, public support to the economy took form of direct state budget expenditures on economic affairs, while later support has been carried out through instruments accounted as a purchase of non-financial assets. Expenditures on education are also higher than average. However, their efficiency is often questioned on a ground of low functional literacy rates (see World Bank, 2014d). On contrary, expenditures on healthcare are in line with average for the region, having increased recently. Table 3. Revenue and expenditure of social fund in 2016, % of GDP Revenues Expenditures Contributions Transfers and other Total, incl. 9.24 5.80 3.56 Pension fund (PAYG), incl. 8.50 5.04 3.51 pensions, incl. compensations for electricity fees 7.72 n/a 3.10 military pensions 0.44 n/a 0.41 State funded accounts 0.28 0.28 n/a Medical insurance fund 0.41 0.43 n/a Rehabilitation of employed fund 0.05 0.05 n/a Source: Social Fund of the Kyrgyz Republic. Figure 1. General government expenditure by function of government in 2016, % of GDP Uzbekistan social protection Ukraine education Russian Federation health Moldova recreation, culture, & religion Kyrgyz Republic Kazakhstan general public services Georgia defense Belarus public order & safety Turkey economic affairs Bulgaria environment protection Albania housing & community amenities 0 10 20 30 40 50 Source: IMF, Government Finance Statistics. 11 Figure 2. General government expense in 2016, % of GDP Republic of Uzbekistan Ukraine Russian Federation Republic of Moldova Compensation of employees Kyrgyz Republic Republic of Kazakhstan Use of goods and services Georgia Republic of Belarus Consumption of fixed capital Republic of Armenia Turkey Interest expense Romania Republic of Poland Subsidies expense Macedonia, former… Kosovo Grants expense Hungary Bulgaria Social benefits expense Bosnia and Herzegovina Albania Other expense Republic of Lithuania Republic of Latvia 0 10 20 30 40 50 60 Source: IMF, Government Finance Statistics. Figure 3. General government revenue in 2016, % of GDP Uzbekistan Ukraine Taxes on income, profits, and capital Russian Federation gains Moldova Taxes on payroll and workforce Kyrgyz Republic Kazakhstan Taxes on property Georgia Belarus Taxes on goods and services Armenia Turkey Taxes on international trade and Romania transactions Poland Other taxes Macedonia, F.Y.R. of Hungary Social contributions Bulgaria Bosnia and… Albania Grants Lithuania Other revenue 0 10 20 30 40 50 Source: IMF, Government Finance Statistics. 10. Meanwhile public expenditures on social policy (including social transfers) in Kyrgyzstan are higher than in other Central Asia and Caucasus countries, but lower than in majority of Central and Eastern Europe countries. They are financed both from sources collected by Social fund and state budget – transfers from state budget constitute significant share of social fund revenues (Table 2). Contributions to the Social fund do not cover public expenditures on pensions and benefits due to prevalence of informal activity, while tax rates are rather moderate (27.25% of wage fund). Transfers from central budgets are directed to financing military pension, compensations for electricity fees as well as state basic pension obligations. 11. Significant public expenditure of Kyrgyzstan has not resulted in poverty reduction until 2016. Fall in poverty risk in 2016 occurred largely in urban area and Bishkek city in particular (from 23.5 to 9.8%, see Kyrgyzstat, 2017). As a result, the disparity in poverty risk across country increased – the highest poverty rates of 37.8 and 37% are observed in Naryan and Batken oblasts. Still this trend did not affect level of inequality. It remained rather low if measured based on consumption aggregate (see Table 3). It is explained by high density of distribution of population by consumption around the poverty line. Consequently, poverty gap is low in Kyrgyzstan and even moderate welfare change can significantly influence poverty rates. 12. At large extent low level of inequality is supported by remittances and income from informal activities. Contrary to consumption aggregate, income distribution according to KIHS indicates high 12 inequality among households. Moreover, it often had adverse dynamics to inequality based on consumption. It stresses problem of informal activity in Kyrgyzstan and high role of remittances in determining welfare of households. Table 4. Poverty and inequality indicators 2009 2010 2011 2012 2013 2014 2015 2016 Poverty rate 31.7 33.7 36.8 38.0 37.0 30.6 32.1 25.4 Urban area 21.9 23.6 30.7 35.4 28.5 26.9 29.3 18.6 Rural area 37.1 39.5 40.4 39.6 41.4 32.6 33.6 29.0 Poverty gap 6.1 7.5 7.5 7.7 7.0 5.4 5.9 4.1 Extreme poverty rate 3.1 5.3 4.5 4.4 2.8 1.2 1.2 0.8 Gini by consumption 0.245 0.251 0.227 0.216 0.219 0.209 0.208 0.203 80/20 ratio 5.0 5.1 4.2 4.2 4.1 4.0 4.0 3.9 Gini by income 0.371 0.371 0.382 0.42 0.456 0.429 0.408 0.406 Source: National Statistical Committee of the Kyrgyz Republic. 13. High risk of poverty in the country and low inequality imply that poverty rates do not differ much across different social groups. By age, share of poor population among those in working age and above working age is very close (21.8 and 19.9% respectively, see Kyrgyzstat, 2017). In previous years, working- age population was more exposed to the risk of poverty (27.9 and 21.8% in 2015), stressing weak employment opportunities in the country and effectiveness of pension system in constraining poverty risk of elderly. Especially high poverty risk is observed among population with low level of education. Graduating from secondary school guarantees poverty risk reduction from 28% (observed among those with basic secondary education) to 21.1% (Kyrgyzstat, 2017). 14. One of the most vulnerable social group to the risk of poverty is children. Risk of poverty among them is above average (31.5%), and it drastically increases for large families. Poverty rates in households with 3 children is 35.6% (54.6% for households with 4 children and 65.2% for 5 children and more; Kyrgyzstat, 2017), stressing gap in effectiveness of social policy in mitigating child poverty. 3. Methodology of the analysis 3.1. Data available and CEQ approach to income concepts construction 15. CEQ analysis is based on micro-level data on households income and expenditure. In Kyrgyzstan related data is available within the Kyrgyz Integrated Household Survey (KIHS). It is conducted by the National Statistical Committee (NSC) on quarterly basis since 2003, and covers 5,000 households. It provides information on purchase and consumption of food products, purchase of non-food products and expenditures on services. Besides, it contains information about the household composition, level of education and stance of health of its members, and living conditions. This data is also integrated with labour force survey, providing detailed information on socio-economic status of individuals. A key objective of the KIHS is to analyze living standards of population of Kyrgyzstan and provide information about effectiveness of social policy aimed at support of social vulnerable groups. 16. Official approach to analysis of households’ welfare in Kyrgyzstan is based on estimating consumption aggregate (Deaton and Zaidi, 1999). The idea is to measure all households consumption expenditures as a proxy of their living standard. Related expenditures include: • consumption of food products, purchased, received as a gift or remuneration, and produced within the household measured in market prices; • expenditures on purchase of non-food products for personal use; • expenditures on services; • notional earnings from durable goods possessed by household, accounting for their depreciation. 13 This approach of measuring welfare is rather sophisticated and demands assumptions about prices of consumed food products as well as value of durable goods. However, it is more reliable than measuring welfare based on income, as it is often underreported especially taking into account significant share of shadow economy and important role of income from labour migrants in Kyrgyzstan. Besides, consumption based approach smooths welfare fluctuations, as it spreads expenditures on durable goods over time. 17. KIHS is designed to provide representative data at the national, rural/urban, and oblast levels, including seven regions (Batken, Jalal-Abad, Issyk-Kul, Naryn, Osh, Talas and Chui oblasts) and two cities (Bishkek, Osh) . However, consumption estimates obtained from KIHS do not fully correspond to macroeconomic data of the national accounts. In 2016, the gap between micro and macro level data on household consumption was 32.6%. Survey data tends to be lower due to underreporting of some expenditures (especially alcohol and tobacco) compared to retail statistics on the one hand, and up-scaling assumptions made while estimating national accounts. Still, underreporting of consumption does not have distributional impact. 18. The core element of CEQ analysis is calculation of income concepts and estimation of absolute poverty headcount . In order to fit official poverty rates we grounded our analysis both on official approach to estimation of welfare and official absolute poverty lines, applied by the Kyrgyzstat and the World Bank (World Bank, …) . Analysis of extreme poverty in Kyrgyzstan is based on the food poverty line, estimated based on existing food basket and energy consumption of 2,100 kcal per capita per day. General poverty line is obtained adjusting food poverty line to the share of non-food products and services in basic consumption of households. In order to account for price discrepancy across the country either poverty line or households income are adjusted to average price level. In official publications adjustment is done via income concept. Application of this approach within CEQ analysis implies that all analyzed fiscal interventions should also be adjusted to the difference in price level across the regions of Kyrgyzstan. First, it will complicate allocation of fiscal interventions by households, and second, it will lead to inconsistence of public finance data and CEQ results. Hence, we applied price adjustment to the poverty line instead, which resulted in minimal discrepancy between official poverty headcount and estimated poverty level (in second digit after decimal). It also lead to the discrepancies in estimated poverty measures by international lines and those published in PovcalNet, as we did not adjust them to the spatial price differences. 19. There are four main income concepts considered within CEQ analysis: market income, disposable income, consumable income and final income. Their construction begins from choice of starting point based on data availability. Official approach to welfare analysis in Kyrgyzstan is to estimate disposable resources of households based on their consumption. Hence, official consumption aggregate (equal to disposable income) was set as a basis from which other income concepts were obtained. 20. Market income was estimated by subtracting from disposable income (consumption aggregate) direct transfers and adding direct taxes. It should be noted that there are two approaches of assigning direct taxes and transfers based on pension system of a country. Pensions can be viewed either as a government transfer (PGT) or a deferred income (PDI). In first case, it implies that social security contributions are accounted as direct taxes while pensions are added to direct transfers. In the second case, pensions and related contributions are not taken into account while estimating market income – pensions are considered as a part of both market and disposable income concepts. Taking into account that pension system in Kyrgyzstan is largely pay-as-you-go, we viewed PGT approach as a basic one. Moreover, it corresponds with approaches applied to CEQ analysis of other CIS countries which simplifies cross-country comparison (see Bornukova, Shymanovich, Chubrik 2017). 21. Estimation of other income concepts was more straightforward (see Figure 4) . Consumable income was calculated as disposable income minus indirect taxes (VAT, sales tax, excises) plus utility subsidies. In order to obtain final income in-kind transfers in the form of public expenditure on health and 14 education were added to consumable income. As a result, construction of the income concepts demanded allocation of the one third of the general government revenue and almost half of general government expenditure. Figure 4. Construction of income concepts Final income in kind transfers – direct transfers (cash and (education and healthcare + indirect subsidies in kind benefits, public expenditures) (utility subsidies) pensions*) Disposable income Consumable income Market income (total consumption) – indirect taxes + direct taxes (PIT and (VAT, excises, and sales tax) social security contributions*, land tax) Note. * Pensions and social contributions on pensions financing are included into direct transfers and taxes respectively only wi thin PGT approach. Source: own elaboration. 3.2. Direct taxes Taxation of employees 22. Personal income tax (PIT) is a key direct tax in Kyrgyzstan . A tax base for PIT paid by employed individuals is wages. In most cases the tax is paid by a tax agent, i.e. by employer. The tax rate is flat and is set at the level of 10%. However, there is a lower bound set for the tax base in the form of a minimum wage. In case employee receives formal remuneration below the minimum wage, the tax is levied on the minimum wage instead of the formal remuneration. However, there are some exclusions from this rule. In particular, minimum wage limit is not applied to people with disabilities. There are also standard tax deductions of 650 KGS available for everyone, and deduction of 100 KGS per dependent person, a member of the employee’s family. Besides there is a social deduction provided for those households who have expenditure on paying education fees. Amount of this deduction is equal to related costs but can not exceed 10% of the tax base, and it is provided only to one member of the family. Furthermore, it is provided in the form of a tax refund after provision of annual declaration and documents supporting the right to apply for this deduction. 2 23. Kyrgyz Integrated Household Survey (KIHS) provides information on personal income tax paid by a respondent, but this data is not reliable. It contains missing observations, and most of the data is reported automatically as 10% of the wage. Therefore, we imputed allocation of PIT based on information about wages reported by individuals, their socio-economic status and characteristics of households. We assumed that surveyed individuals reported net wages. It implied that tax base was calculated taking into account standard deductions and possible application of minimum wage rule. We applied deduction on dependent individuals, determining their number as difference between household size, and number of employed, farmers, entrepreneurs, and pensioners, including those receiving disability pensions. As social deduction on compensating education expenditure is provided ex post, it does not have effect on the tax base calculation. Eligibility of this deduction was estimated based on information about expenditure of households on tuition fees, while exact beneficiary of this deduction within the household was determined based on the volume of PIT burden: person with higher modelled tax payments was assumed to apply for social deduction. 2 Another deduction is a property one provided for compensating expenditures on interest payments on mortgage loans. Related information is not available in KIHS, so this deduction is not covered by analysis. 15 24. The adequacy of modelled PIT allocation depends heavily on correct determination of formally employed individuals paying taxes. Survey includes question on socio-economic status, which allows to distinguish between formally employed, and farmers, and entrepreneurs, representing informal sector; as well as questions specifying type of employment, that describe informal activity and labour migration. Application of modelled PIT obligations and deductions to labour income of individuals formally employed in Kyrgyzstan resulted in total tax obligations of KGS 10.5 bn. It is almost quarter higher than actual public revenue from PIT. Assuming correct determination of formally employed individuals, it implies significant tax evasion and wide spread practice of “envelope” wages. Therefore, we adjusted tax rate to the level of effective rate, obtained by comparison of wage fund from national accounts and PIT public revenue. Effective rate, calculated based on data of 2015, equaled to 7.88%. Consequently, modelled PIT obligations reduced to KGS 8.4 bn that fits actual PIT public revenue (KGS 8.6 bn). 25. Social security contributions collected by Social Fund are also considered as a direct tax. A tax base for social security contributions is largely the same as for PIT. These contributions are partly paid by employer at the rate of 17.25% of the tax base and partly by employee at the rate of 10%. The latter rate is reduced to 8% for working pensioners and 2% for working people with disabilities. Applying these rates to previously modelled tax base for PIT, we once again received total contributions significantly exceeding actual revenue of social security fund. Thus, we adjusted formal tax rates to effective one, which was 21.1%, according to data of 2015. As a result, the volume of allocated social contributions equaled to KGS 28.4 bn, which is close to actual public revenue of KGS 26.6 bn. 26. Incidence of PIT and social contributions by deciles reveals that tax obligations are distributed progressively, as upper deciles pay more taxes both in absolute and relative terms than lower deciles (see Figure 5) . Still, difference in tax payments (even in absolute terms) within lower deciles is rather limited, which stems from the tax legislation. Furthermore, the first decile by pre-fiscal market income faces higher tax burden than other bottom 5 deciles, as application of minimal tax base rule increases effective tax rate for households with low market income. According to our estimations, around 1.5% of households are affected by this legal provision. Especially high tax burden is observed within the 10 th decile, distinguished based on market (pre-fiscal) income. It stresses key role of formal employment as a precondition that may assure relatively high welfare of a household. Figure 5. Incidence of direct taxes on formally employed by deciles of population distinguished based on market and disposable income, % of income and USD PPP per day 6 0.8 20 2.5 5 2 0.6 15 4 1.5 3 0.4 10 1 2 0.2 5 1 0.5 0 0 0 0 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 USD PPP, disposable income (right axis) USD PPP, disposable income (right axis) USD PPP, market income (right axis) USD PPP, market income (right axis) %, disposable income %, disposable income %, market income %, market income (a) Personal income tax (b) Social security contributions Note. Market income approach provides allocation of taxes by cohorts ranked by income available for households prior to fiscal interventions in the form of direct taxes and transfers. Disposable income approach reveals allocation of taxes by population ranked by income available after fiscal interventions. 16 Source: own estimates. Taxation of entrepreneurs 27. There are different regimes of taxation of entrepreneurs in Kyrgyzstan. However, data on budget revenue suggests that most of them tend to operate based on voluntary patent. The size of patent payment depends on region of the country and type of activity. Hence, thorough modelling of these payments is complicated. Taking into account rather limited role of these taxes in the economy, we imputed related tax obligations of households based on assumed average size of the patent payment. The volume of this payment was estimated as public revenue from patents (KGS 2.2 bn) divided by total number of entrepreneurs (379.2 thsd). The volume of allocated patent revenues amounted to KGS 1.3 bn. It is lower than actual public revenue from this tax, as modelled number of tax payers, formally operating in the country, according to our estimates based on KIHS data is 216.6 thsd, which is lower than official number of entrepreneurs. 28. There are different schemes of contributing to social security fund by entrepreneurs as well. Those operating based on patent are paying social contributions linked to the average wage in the region. This rate differs from 3 to 10% depending on type of economic activity. As majority of the entrepreneurs are busy in retail or agriculture sector we applied the rate of 3%, set for entrepreneurs selling products in market places from temporary points of sale. The volume of allocated according to these scheme contributions equaled to KGS 0.9 bn, which is above data on actual collection (KGS 0.7 bn). It is related to the widespread practice of tax evasion and possibility to pay contributions as a lump sum payment having reached retirement age. Still, we do not adjust modelled payments to fit macroeconomic data, as they do not have significant impact either on welfare of population or public revenue. 29. Incidence of taxes on entrepreneurs reveals that their burden as a share of either pre-fiscal or post-fiscal income is rather equal for all deciles (Figure 6), implying that its influence on welfare is neutral in relative terms. Still, in absolute terms it is progressive, as tax contributions of upper deciles measured in USD PPP are much higher than contributions of lower deciles. Figure 6. Incidence of direct taxes on entrepreneurs by deciles of population distinguished based on market and disposable income, % of income and USD PPP per day 0.8 0.05 0.6 0.03 0.04 0.5 0.025 0.6 0.4 0.02 0.03 0.4 0.3 0.015 0.02 0.2 0.01 0.2 0.01 0.1 0.005 0 0 0 0 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 USD PPP, disposable income (right axis) USD PPP, disposable income (right axis) USD PPP, market income (right axis) USD PPP, market income (right axis) %, disposable income %, disposable income %, market income %, market income (a) Patent (b) Social security contributions Note. Market income approach provides allocation of taxes by cohorts ranked by income available for households prior to fiscal interventions in the form of direct taxes and transfers. Disposable income approach reveals allocation of taxes by population ranked by income available after fiscal interventions. Source: own estimates. 17 Taxation of farmers 30. Farmers pay social contributions from their land based on the rates that differ according to land type and region . We applied these rates to the households, where members reported to work on their own farm, accounting for the type of a land plot they have in possession. Modelled payments happened to be underestimated, but total volume of public revenue from this tax (0.3 bn) is rather negligible, so it can not cause any distortions in a tax burden allocation. Besides, there is also problem of tax evasion and practice of lump-sum payments by rural population reaching retirement age, which complicates modelling actual social contributions of the farmers. 31. Another tax paid by the farmers is a land tax. It is also levied on households, having in possession land plots, used for subsistence farming and living . However, latter are exempted from paying land tax if a plot is owned by a pensioner. The sum of tax obligations depends on type of land, its area and region. Related information is provided in KIHS, which allowed to model land tax obligations rather accurately. Total volume of estimated land tax equaled to KGS 0.44 bn, which matches public revenue from the tax land on agricultural land and land used for subsistence farming and living (KGS 0.46 bn). The volume of generated tax revenue is rather small as the tax rates, set in absolute terms, have not been revised for a decade. 32. Modelled incidence of land tax and social security contributions suggests that these taxes are regressive, as lower deciles spend bigger share of their income on paying them than upper deciles (Figure 7). In absolute terms, the tax burden is largely flat. It is determined by low income of rural population, who bears burden of paying these taxes. Figure 7. Incidence of direct taxes on entrepreneurs by deciles of population distinguished based on market and disposable income, % of income and USD PPP per day 0.4 0.014 0.8 0.005 0.012 0.004 0.3 0.01 0.6 0.008 0.003 0.2 0.4 0.006 0.002 0.1 0.004 0.2 0.001 0.002 0 0 0 0 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 USD PPP, disposable income (right axis) USD PPP, disposable income (right axis) USD PPP, market income (right axis) USD PPP, market income (right axis) %, disposable income %, disposable income %, market income %, market income (a) Land tax (b) Social security contributions Note. The land tax burden for the first decile by market income is 3.8%. Market income approach provides allocation of taxes by cohorts ranked by income available for households prior to fiscal interventions in the form of direct taxes and transfers. Disposable income approach reveals allocation of taxes by population ranked by income available after fiscal interventions. Source: own estimates. 3.3. Indirect taxes VAT 33. Key indirect tax in Kyrgyzstan is VAT. However, its rate is relatively low (12%) and there are a lot of exemptions for local producers. In particular, agriculture and food production is largely exempted from VAT. It implies that incoming VAT is compensated for this sector as well. Most of services provided by public sector are also taxed at the rate of 0. As a result, main part of VAT revenue comes from import taxation. As we can not distinguish between consumption of domestic and import products, as well as purchase of goods 18 within informal sector we applied effective VAT rates for separate goods and services. Effective rates were estimated based on input-output tables, in particular share of VAT payments in total supply of the sector in purchaser prices (see Table 4). We modelled expenditures of households, related to VAT payments assuming that this tax is levied at effective rate on all purchased food and non-food products, as well as expenditure on education, healthcare, utilities, mobile and other services. The volume of modelled VAT burden for households equalled to KGS 10.7 bn, which constitutes 27.2% of total VAT revenue. It is a reasonable volume, as (i) KIHS data covers only 67.4% of total household consumption, (ii) according to input-output tables only 48% of indirect taxes is related to domestic household consumption. 34. As food products and basic services, which constitute significant share of consumption of households from lower deciles, are largely exempted from VAT, modelled tax burden is progressive both in absolute and relative terms (see Figure 8a). The first three deciles pay much less of their disposable (pre- fiscal) income on VAT than population in general. The share of income spent by other deciles on VAT is largely the same, stressing similar structure of their consumption. Exclusion are the 9th and 10th deciles, where consumption structure changes towards less expenditure on food products. Table 5. Tariff rates of the modelled indirect taxes, % of purchaser prices Retail sales tax Expenditure of Sector form the input-output table VAT rate and tax for Import duties the household services provision Food products Food production 2.71 0.15 0.78 purchase Textile; oil, petroleum products and chemical industry; Non-food production of electronic equipment; production of non - 4.82 0.05 3.33 products metallic products Eating out Hotels and restaurants 7.03 0.86 Healthcare Healthcare 0.13 0.06 Education Education 0.81 0.10 Electricity production and distribution; gas distribution; Utilities 3.83 0.49 steam production; water distribution and sewage Mobile servicies Information and communication 6.24 11.79 Other servicies Hotels and restaurants 7.03 0.86 Note. Aggregation of the rates for non-food products and utilities was based on average consumption structure of the household of related goods and services. Source: own elaboration. Sales tax 35. In addition to VAT, Kyrgyzstan tax system also includes sales tax . It is levied on retail sales and its rate depends on the sector and payment type. Besides, some sectors, like agriculture and services provided by public sector, are exempted from sales tax in accordance with legislation. In addition to sales taxes, provision of some services is levied with special taxes. Information of related tax payments done by sectors of economy is provided in input-output tables as well. Hence, we also applied effective rates, estimated as a share of other taxes on products in total supply of the sector in purchaser prices (Table 4), to calculate sales tax burden on households. Total amount of allocated sales tax equalled to KGS 1.6 bn or 26.6% of total public sector revenue from the tax. As in the case of VAT, the underestimation is related to the volume of household consumption captured by KIHS and modelling approach, implying distribution of the tax burden across the whole economy. 36. Sales tax is less progressive than VAT, as it contains fewer exemptions. Nevertheless, if population is ranked by pre-fiscal disposable income, tax burden of sales tax increases from lower deciles to upper deciles both in absolute and relative terms (see Figure 8b). The most significant increase occurs between the 9th and 10th deciles, following changes in profile of consumption expenditure. 19 Figure 8. Incidence of VAT and sales tax by deciles of population distinguished based on disposable and consumable income, % of income and USD PPP per day 5.0 0.6 0.68 0.08 0.66 0.07 4.5 0.5 0.64 0.06 0.62 4.0 0.4 0.60 0.05 3.5 0.3 0.58 0.04 0.56 0.03 3.0 0.2 0.54 0.02 0.52 2.5 0.1 0.50 0.01 2.0 0 0.48 0 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 USD PPP (right axis) % of disposable income USD PPP (right axis) % of disposable income (a) VAT (b) Sales tax Note. Disposable income approach provides allocation of taxes by cohorts ranked by income available for households prior to f iscal interventions in the form of indirect taxes and transfers. Source: own estimates. Import duties 37. Alongside with VAT and sales tax we also modelled burden of import duties faced by households. Related information on effective tax rates for food and non-food products was obtained from the input-output tables (share of import duties in total supply of corresponding sectors in purchaser prices, see Table 4). Total volume of modelled public revenue is KGS 2.41 bn, which is 18.7% of actual import duties collection. 38. The burden of the duties is allocated rather evenly across the households if measured as a share of disposable income (see Figure 9a). The highest burden is faced by the 10 th decile, which consume more non-food products than other deciles of population. Excises 39. Excises are levied on alcohol, tobacco and fuel . Related rates are set in absolute terms per amount/volume of sold product. Data on purchase of alcohol, tobacco and fuel is provided in KIHS. However, estimation of excises based on data reported by households resulted in their volume that was much lower than actual tax collection (KGS 0.7 bn compared to 9.1 bn). On the one hand, it is related to the fact that legal entities also pay excises, especially if fuel excises are concerned. On the other hand, households tend to underreport consumption of alcohol and tobacco. According to KIHS data total expenditure of households on purchase of alcohol and tobacco products was KGS 6.6 bn. Meanwhile according to retail statistics, total volume of sales of these goods was KGS 22.7 bn. Upscaling of reported expenditure may be misleading, as it is not clear, which households reported only part of their expenditure on alcohol and tobacco, and which did not report these expenditure at all. Only one third of households reported purchases of tobacco and/or alcohol. Partly, it may be attributed to the religion based refusal of alcohol consumption, and partly to the fact that people do not want to confess in consuming unacceptable products. Still, we believe that available data provides enough information to analyze distribution of welfare effects generated by excises keeping in mind underestimation of its scale. 20 Figure 9. Incidence of import duties and excises by deciles of population distinguished based on disposable income, % of income and USD PPP per day 1.00 0.12 0.07 0.006 0.95 0.06 0.005 0.1 0.90 0.05 0.004 0.08 0.85 0.04 0.003 0.80 0.06 0.03 0.75 0.002 0.04 0.02 0.70 0.001 0.01 0.02 0.65 0 0 0.60 0 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 USD PPP, disposable income (right axis) USD PPP (right axis) % of disposable income %, disposable income (a) Import duties (b) Tobacco excises 0.14 0.012 0.25 0.016 0.12 0.014 0.01 0.2 0.1 0.012 0.008 0.15 0.01 0.08 0.006 0.008 0.06 0.1 0.006 0.004 0.04 0.004 0.02 0.002 0.05 0.002 0 0 0 0 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 USD PPP, disposable income (right axis) USD PPP, disposable income (right axis) %, disposable income %, disposable income (c) Fuel excises (d) Alcohol excises Note. Disposable income approach provides allocation of taxes by cohorts ranked by income available for households prior to f iscal interventions in the form of indirect taxes and transfers. Source: own estimates. 40. According to incidence of excises by deciles (Figure 9), fuel excises are progressive, as upper bins of population are more likely to possess personal vehicles and, consequently, spend more on fuel. Alcohol excises may be also considered as progressive, absorbing higher than average share in income of the upper 3 deciles. Excises on tobacco products appear to be regressive in relative terms. 3.4. Contributory pensions 41. Certain types of pensions in CEQ analysis could be treated in two ways: the PDI scenario treats them as contributory, and hence as realization of the income deterred in the past; the PGT scenario, on the other hand, treats pensions as transfers from the government . We will use PGT as a base scenario, and PDI as an alternative. The pensions in Kyrgyzstan have both the transfer component – the basic part and the insurance part of the pension, and the contributory component – the extra part which depends on the earnings and contributions of the retiree in the past (World Bank 2014a). The basic part of the pension is covered by the government budget. The insurance part is covered by the contributions of the current workers to the Social Fund of Kyrgyzstan. The contributory part is financed by the worker’s own past contributions and is accumulated and paid through the state or private pension funds. However, the contributory component 21 is relatively small. According to the Social Fund of Kyrgyzstan, the contributory part constituted only 3.2% of the total state pension expenditure3. Given that, we consider pensions as a government transfer rather than a deferred income. 42. The information on pensions is present in KIHS, so we did not have to rely on simulations and used the direct information in the microdata . KIHS provides information on all the types of pensions present in Kyrgyzstan: retirement age pensions, special pensions for difficult labor conditions, disability pensions, and pensions on the loss of the breadwinner; however, the full disaggregation of pensions is available only in the labor force survey portion of the data. Since the labor force survey provides the incomplete data coverage, we have to use the aggregated pensions data (with limited disaggregation by kind) from the main part of the survey. We classify the retirement age pensions and special pensions for difficult labor conditions as contributory pensions, since they depend on the worker contributions. Disability pensions and pensions on the loss of the breadwinner are classified as types of transfers. 43. The contributory pensions are highly progressive, both in absolute and in relative terms. Figure 10 below shows the incidence of contributory pensions by deciles of market income (pre-fiscal income concept in this case). As we can see, lower deciles receive more through the contributory pensions in terms of absolute (daily USD PPP) and relative values (as percentage of pre-fiscal income). Figure 10. Incidence of contributory pensions by market income deciles of population, % of income and daily 2011 PPP USD 60 3.5 50 3 2.5 40 2 30 1.5 20 1 10 0.5 0 0 1 2 3 4 5 6 7 8 9 10 USD PPP, daily (right axis) %, market income Note. The average % of pensions to market income of the first decile is very high (almost 3500%) due to high incidence of zero ma rket incomes in that decile and relatively high concentration of retirees in this decile. For clarity of presentation it is not shown in the figure. Source: own estimates. 3.5. Direct transfers 44. Kyrgyzstan offers direct social assistance in different forms. These could be either direct benefits (for example, disability pension), privileges (goods and services provided free of charge) or direct utility subsidies (a discount in the utility bill atop of the standard tariffs). For detailed discussion of social assistance programs see World Bank (2014b). The information on all three types of direct assistance is reported in KIHS. Benefits include categorical assistance, including stipends, monthly benefits to disabiled, orphaned children, senior citizens who are not eligible for a pension, mothers of seven or more children, unemployed and targeted social assistance for poor households with children. Privileges, provided to rather narrow categories of population, include free-of-charge or inkind transfers of food; transportation services; pharmaceuticals and other free services; they also include the monetary payments made in exchange for non-obtained privileges. 3The report on the execution of the Social Fund of Kyrgyzstan budget for 2016, online, http://socfond.kg/ru/about_fund/report2016/ accessed on May 10 th 2018. 22 According to KIHS, direct utility subsidies are mostly offered on natural gas and hotwater services provision, but may include discounts on other types of utilities as well. In particular, there are direct subsidies on electricity tariffs which are provided for retired as an extra payment to the pensions. Hence, this subsidy is rarely reported by the households, being perceived as a part of the pension. Figure 11. Incidence of direct transfers by market income deciles of population, % of income and daily 2011 PPP USD 12 0.4 2 0.018 0.35 1.8 0.016 10 1.6 0.3 0.014 8 1.4 0.012 0.25 1.2 0.01 6 0.2 1 0.008 0.15 0.8 4 0.006 0.6 0.1 0.004 2 0.4 0.05 0.2 0.002 0 0 0 0 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 USD PPP, market income (right axis) USD PPP, market income (right axis) %, market income %, market income (a) Benefits (b) Privileges 10 0.003 9 8 0.0025 7 0.002 6 5 0.0015 4 3 0.001 2 0.0005 1 0 0 1 2 3 4 5 6 7 8 9 10 USD PPP, market income (right axis) %, market income (c) Direct utility subsidies Note. The average % of benefits to market income of the first decile is very high (422%) due to high incidence of zero market incomes in that decile and relatively high concentration of those in need in this decile. For clarity of presentation it is not sh own in the figure. Source: own estimates. 45. As figures 11 (a)–(c) above show, all direct transfers are progressive, with higher absolute and relative contributions going to the lower deciles. The benefits are the largest intervention in size. While benefits are higher for the lower decile groups, upper deciles still get substantial amounts in benefits, suggesting a room for improvement in targeting. Privileges and direct utility subsidies are substantially smaller. A spike in privileges in the sixth decile is interesting, probably explained by a concentration of a certain target group for privileges in this decile. Direct utility subsidies seem to be very well targeted for the first market income decile, but their scale is negligible. 3.6. Indirect subsidies 46. Our analysis of indirect subsidies focuses on the indirect utility subsidies . To the best of our knowledge, this is the only significant group of tariffs to be subsidized. To calculate the obtained indirect 23 subsidies, we used the estimates of the World Bank for the cost coverage of the tariffs and the KIHS data on the utility bills. Figure 12. Incidence of indirect subsidies by disposable income deciles of population, % of income and daily 2011 PPP USD 4 0.3 3.5 0.3 3.5 3 0.25 0.25 3 2.5 0.2 0.2 2.5 2 2 0.15 0.15 1.5 1.5 0.1 0.1 1 1 0.05 0.5 0.05 0.5 0 0 0 0 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 USD PPP, daily (right axis) USD PPP, daily (right axis) %, disposable income %, disposable income (a) Heating subsidy (b) Electricity subsidy 0.8 0.06 0.7 0.05 0.6 0.04 0.5 0.4 0.03 0.3 0.02 0.2 0.01 0.1 0 0 1 2 3 4 5 6 7 8 9 10 USD PPP, daily (right axis) %, disposable income (c) Hot water subsidy Note. In this case the pre-intervention income concept is the disposable income. Source: own estimates. 47. Figures 12 (a)–(c) above show that indirect utility subsidies are regressive in absolute terms, and heating and hot water subsidies are regressive even in relative terms, benefitting more to those with high income. This is quite a common result for the indirect subsidies. People with higher incomes have lower incentives to minimize the consumption of utilities, and usually have bigger homes. This results in higher subsidies for high income deciles. Both the heating and the electricity subsidy are quite large in size. As a result, indirect subsidies is a bigger intervention than all direct transfers. 3.7. In-kind transfers: Healthcare 48. Healthcare is co-funded in Kyrgyzstan – the patients should pay co-payment fees for healthcare services. However, many citizens obtain payment exemptions (World Bank 2014c). We used the standard actual consumption approach (Bastagli 2015) to allocate health expenditure proportionately to doctor visits and hospitalizations. Hence, the healthcare expenditure is allocated only tothose households who have actually used the healthcare services in the reporting year. KIHS provides the data on the number of doctor visits and number of hospitalizations. We also deduct the co-payments and other fees using the data available in KIHS. 24 The budget data by type of use (primary+secondary and tertiary healthcare expenditure) is from GFS. As suggested by CEQ handbook (Lustig 2017), expenditures are scaled down so that the ratio of that category of health benefits in the survey to disposable income in the survey equals the corresponding ratio from national accounts (the resulting scaling factor is 0.67). 49. In relative terms health expenditure is redistributionally neutral . In absolute terms higher income deciles have higher expenditure allocated to them, and this is more pronounced for the primary and secondary healthcare. The reason behind it is probably better access to healthcare of the urban population, and also a relatively higher toll of the co-payments for the population in the lower deciles, which makes them avoid the official healthcare whenever possible. As for the tertiary healthcare, we do not see the regressivity as much, in part because the upper deciles pay more in co-payment fees (deducted from allocated expenditure). Figure 13. Incidence of indirect subsidies by disposable income deciles of population, % of income and daily 2011 PPP USD 1.4 0.16 3 0.25 1.2 0.14 2.5 0.2 0.12 1 2 0.1 0.15 0.8 0.08 1.5 0.6 0.1 0.06 1 0.4 0.04 0.5 0.05 0.2 0.02 0 0 0 0 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 USD PPP, daily (right axis) USD PPP, daily (right axis) %, consumable income %, consumable income (a) Primary and secondary healthcare (b) Tertiary healthcare Note. In this case the pre-intervention income concept is the consumable income. Source: own estimates. 3.8. In-kind transfers: Education 50. School education in Kyrgyzstan is predominantly free . Kindergartens might be private and charge for its services. College education may require co-funding from its students. We have used student/pupil status reported in the KIHS data to allocate public education expenditure. Primary and pre-primary education expenditure was proportionally allocated to children attending state kindergartens and primary schools; secondary education expenditure – two secondary school pupils; post-secondary expenditure– to post-school non-college students; and tertiary education expenditure – to college students. As suggested by handbook, expenditures are scaled down so that the ratio of that category of health benefits in the survey to disposable income in the survey equals the corresponding ratio from national accounts (the resulting scaling factor is 0.67). 51. As seen on Figure 14 below, primary and secondary education is strongly progressive, redistributing expenditure towards poorer deciles, where most of the families with children are concentrated. Post-secondary and tertiary education expenditure are regressive in absolute terms. This is quite a common finding, in line with notion that the youth from relatively reach families continues studies after school. However, the post-secondary and tertiary education expenditure are smaller than the primary and secondary, hence the educational expenditure in total could be viewed as progressive. 25 Figure 14. Incidence of indirect subsidies by disposable income deciles of population, % of income and daily 2011 PPP USD 2.5 0.06 9 0.25 8 0.05 0.2 2 7 0.04 6 1.5 0.15 5 0.03 4 1 0.1 0.02 3 0.5 2 0.05 0.01 1 0 0 0 0 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 USD PPP, daily (right axis) USD PPP, daily (right axis) %, consumable income %, consumable income (a) Primary education (b) Secondary education 0.25 0.014 1.4 0.09 0.012 1.2 0.08 0.2 0.07 0.01 1 0.06 0.15 0.008 0.8 0.05 0.006 0.6 0.04 0.1 0.03 0.004 0.4 0.02 0.05 0.002 0.2 0.01 0 0 0 0 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 USD PPP, daily (right axis) USD PPP, daily (right axis) %, consumable income %, consumable income (c) Post-secondary education (d) Tertiary education Note. In this case the pre-intervention income concept is the consumable income. Source: own estimates. 4. Results and discussion 4.1. Main results 52. Effect from fiscal interventions on inequality and poverty in Kyrgyzstan is substantial, if pension system is considered as a part of the government social safety net. Difference between poverty rates according to the market income (PGT approach) and disposable concepts is 12 p.p. Most of the poverty reduction takes place within extreme poverty (10.4 p.p.), as a part of population relies on pensions as a sole source of income. Consequently, if fiscal interventions are limited to direct taxes and transfers, they generate effect of the poverty and extreme poverty reduction by 0.7 and 1 p.p. respectively (see Table 5). Higher effect on extreme poverty stresses that system of direct taxes and transfers have an impoverishment effect over certain groups of population. Moreover, impoverishment effect of indirect taxes and transfers is even more apparent, as they lead to poverty rate increase of 2.1 p.p. 53. High density of income distribution around poverty line is evident through low rates of p overty gap. According to the national concept of poverty measurement, the gap is relatively high only at the level of market income by PGT approach. Consequently, consumption inequality indicators are also very low in 26 Kyrgyzstan for all income concepts. Gini coefficient for market income that is cleared from pensions is only 0.316. Fiscal interventions related to direct transfers, taxes and pensions reduce Gini coefficient to 0.214. Indirect taxes and subsidies results in slight inequality growth, while in-kind transfers contribute to further reduction of consumption inequality (Gini coefficients are 0.219 and 0.211 respectively). Hence, overall relatively low income level of population in Kyrgyzstan is accompanied by low consumption inequality, which only at some extent can be attributed to the influence of fiscal interventions, namely pensions. Table 6. Main poverty and inequality indicators by income concepts Market Income Disposable Consumable Final PGT PDI Income Income Income Gini 0.316 0.234 0.214 0.219 0.211 Theil Index 0.162 0.092 0.078 0.081 0.075 90/10 4.623 2.812 2.519 2.603 2.495 National extreme poverty Headcount Index 11.2 1.8 0.8 0.7 line Poverty Gap 4.9 0.3 0.1 0.1 Headcount Index 37.4 26.1 25.4 27.5 National poverty line Poverty Gap 13.1 5.1 4.1 4.6 Source: own estimates based on CEQ methodology. 54. However, fiscal policy causes reshaping of income allocation with the scale much higher than change in Gini coefficients may suggest . Fiscal interventions are associated with high reranking effect. For instance, direct transfers and taxes, including pension system, cause reduction of Gini index by 0.102 points (see Table 7). Meanwhile gross vertical effect of these measures (related to reduction of consumption within upper income bins and increase of consumption within lower income bins) is equal to Gini change of 0.159. Hence, one third of generated income redistribution affects horizontal equity – it does not reduce gap in welfare between rich and poor, but makes them change places. This effect is especially high when influence of indirect taxes and subsidies is taken into account. Vertical effect of indirect taxes and transfers is almost absent. Reynolds-Smolensky index for change between consumable and disposable income is -0.001. Consequently, Gini increase of 0.005 between consumable and disposable income happens almost fully due to reranking effect, caused by indirect taxes and transfers. The same scale of reranking effect is typical for inkind transfers, but it also generates feasible vertical effect, equaled to 0.013 Gini change between distributions of population by final and consumable income. Table 7. Decompositions of inequality changes into vertical and horizontal equity components Change to market income (PGT) Change to market income (PDI) Disposable Consumable Final Disposable Consumable Final income income income income income income Gini change with respect to market income 0.102 0.097 0.106 0.020 0.015 0.023 Vertical equity (Reynolds-Smolensky Index) 0.159 0.159 0.168 0.029 0.026 0.038 Reranking (Atkinson-Plotnick Index of horizontal equity) 0.057 0.062 0.062 0.009 0.010 0.015 Note. Vertical equity implies reduction of gap in welfare between rich and poor due to fiscal intervention. Horizontal equity implies that fiscal intervention does not influence ranking position of an individual, see Kakwani (1984). Source: own estimates based on CEQ methodology. 55. Nevertheless, distribution of welfare effects is rather progressive . Net effect of fiscal interventions steadily diminishes from low-income deciles to high-income deciles, becoming negative for the highest two quintiles (see Table 8). However, scale of this effect is substantial only if pension system is treated as fiscal intervention (PGT approach). If pension system is excluded from analysis (PDI approach), significant welfare effects are observed for the 1st and 10th deciles mainly. Hence, most of income redistribution is observed within shift from market income by PGT approach to disposable income. Changes in results related to the shifts from disposable to consumable and market income are rather minor. First, inclusion into analysis of indirect taxes and transfers reduces positive effect for low-income deciles at larger extent than negative effect for high- income deciles. Hence, these fiscal interventions affect welfare of lower deciles resulting in Gini increase. 27 Second, prevalence of positive effect for final income concept in relation to market one is explained by the difference between total size of modelled market and final income. Inclusion of public expenditure on healthcare and education into analysis (that are relatively high in Kyrgyzstan) leads to the fact that total size of final income significantly exceeds the size of other income concepts. Table 8. Incidence of net effects from fiscal interventions in relation to market income by deciles Market income by PGT approach Market income by PDI approach Disposable Consumable Final Disposable Consumable Final Income Income Income Income Income Income 1 153.2 147.2 170.8 8.6 6.1 17.9 2 40.8 38.4 52.0 3.3 1.6 11.1 3 23.6 20.0 32.8 1.8 -1.2 8.6 4 13.9 12.0 23.0 -0.5 -2.2 7.9 5 8.2 6.6 16.4 -1.1 -2.6 7.2 6 4.1 2.8 11.7 -1.7 -2.9 4.8 7 -0.3 -1.4 6.6 -2.9 -4.0 4.1 8 -4.0 -4.4 2.6 -3.7 -4.1 3.0 9 -8.3 -8.5 -2.3 -4.8 -5.0 1.4 10 -13.9 -13.9 -8.5 -6.1 -6.1 -0.3 Total 3.5 2.5 11.2 -2.2 -3.2 4.8 Source: own estimates based on CEQ methodology. 56. Still, progressive distribution of net effects of fiscal interventions does not eliminate the fact that there are those, who lose from fiscal interventions even being initially in lower deciles . There are net donors of fiscal interventions even among those with market income below the poverty line (around 4 USD PPP). Nevertheless, majority of net recipients from direct transfers and taxes are among poor population (by market income, see Figure 15a). Moreover, the size of net receipt generated by direct transfers (including pension system) is much higher than donation for population with low income bins. For extremely poor population net receipt exceeds 100% of their market income. It reduces to 20% of market income for population around national poverty line. For comparison, the burden of direct taxes riches 20% level only for population with income around USD 10 PPP. Difference in the size of net donations and contributions creates ground for reranking of population. As a result, distribution of net donors and recipients of direct transfers and taxes, if ranked by post-fiscal disposable income, is almost similar, see Figure 15b. 57. System of indirect taxes and transfers is less preferential for low-income population. Most of net donors are concentrated around pre-fiscal income bins (disposable income) close to the national poverty line (see Figure 15c). Moreover, size of net indirect benefit is positively correlated with disposable income. Inclusion of expenditures on education and healthcare into analysis, changes incidence of donors and recipients towards more progressive, see Figure 15d. 58. Hence, fiscal interventions may both reduce poverty and create it . In fact, around 40% of poor population faces welfare reduction due to income redistribution caused by fiscal policy and pension system (see Table 9). If pension system is excluded from analysis, share of poor population suffering reduction of income due to fiscal interventions increases up to 65%. Still, number of beneficiaries among initially poor population is three times higher and the size of gains surpasses the size of losses, which results in poverty gap reduction. 28 Figure 15. Distribution of gains and losses by income bins set in USD PPP Gains and losses at the level of disposable income compared to the level of market income (PGT approach) distributed by bins, a) ranked by market income b) ranked by disposable income 250 100 250 100 200 80 200 80 150 60 150 60 100 40 100 40 50 20 50 20 0 0 0 0 0.25 1.50 2.75 4.00 5.25 6.50 7.75 9.00 0.25 1.50 2.75 4.00 5.25 6.50 7.75 9.00 10.25 11.50 12.75 14.00 15.25 16.50 17.75 19.00 10.25 11.50 12.75 14.00 15.25 16.50 17.75 19.00 recipients, thsd recipients, thsd donors, thsd donors, thsd average donation, % of disposable income(right axis) average reciept, % of disposable income(right axis) average reciept, % of disposable income(right axis) average donation, % of disposable income(right axis) (c) Gains and losses at the level of consumable income (d) Gains and losses at the level of final income compared compared to the level of disposable income distributed by to the level of disposable income distributed by bins, ranked bins, ranked by disposable income by disposable income 400 8 350 16 350 7 300 14 300 6 250 12 250 5 10 200 200 4 8 150 150 3 6 100 2 100 4 50 1 50 2 0 0 0 0 0.50 1.75 3.00 4.25 5.50 6.75 8.00 9.25 0.25 1.50 2.75 4.00 5.25 6.50 7.75 9.00 10.50 11.75 13.00 14.25 15.50 16.75 18.00 19.25 10.25 11.50 12.75 14.00 15.25 16.50 17.75 recipients, thsd 19.00 recipients, thsd donors, thsd donors, thsd average reciept, % of disposable income(right axis) average reciept, % of disposable income(right axis) average donation, % of disposable income(right axis) average donation, % of disposable income(right axis) Note. Average receipt and donation are moving average for 5 income bins. Source: own estimates based on CEQ methodology. Table 9. Fiscal gains to the poor and fiscal impoverishment in relation to market income Pensions as government transfer Pensions as deferred income National extreme National moderate National extreme National moderate poverty line poverty line poverty line poverty line Disposable Consumabl Disposable Consumabl Disposable Consumabl Disposable Consumabl Income e Income Income e Income Income e Income Income e Income Fiscal impoverishment headcount, % of population 0.45 0.27 10.32 11.74 0.74 0.54 16.53 19.15 headcount, % of post-fiscal poor 54.73 37.74 40.61 42.76 89.47 74.04 65.08 69.71 fiscal impoverishment per capita, % of market income 0.02 0.03 0.69 0.88 0.02 0.05 0.55 3.59 Fiscal gains to the poor headcount, % of population 10.83 10.99 30.81 30.98 1.15 1.31 11.32 11.83 headcount, % of pre-fiscal poor 97.12 98.53 82.33 82.78 65.52 74.53 43.36 45.31 fiscal gains to the poor per capita, % of market income 4.81 4.81 9.69 9.41 0.26 0.26 1.46 1.37 29 Poverty gap reduction, pp (fiscal gains minus fiscal impoverishment per capita) 4.78 4.78 8.99 8.52 0.24 0.21 0.91 -2.22 Note. Fiscal impoverishment is a considered a situation when i) somebody who is non -poor according to the pre-fiscal income (market income) appears to be poor according to the post-fiscal income (disposable or consumable), ii) somebody poo r according to the pre- fiscal income suffers further income reduction due to fiscal interventions. Fiscal gains to the poor take place when somebody poor according to the pre-fiscal income enjoys income increase due to fiscal interventions. Source: own estimates based on CEQ methodology and Higgins, Lustig (2016). 4.2. Distributional impact and marginal contributions of fiscal interventions 59. Analysis of influence of separate fiscal interventions on inequality is based on their concentration curves and corresponding concentration coefficients. If the curve of the intervention is above the 45-degree line (i.e. concentration coefficient is negative), it is progressive and equalizing. An intervention can be equalizing even if it is regressive, i.e. being below the 45-degree line. A transfer is equalizing if its concertation curve is above Lorenz line of corresponding pre-fiscal income, implying that concentration coefficient of a transfer exceeds Gini coefficient for the income. Taxes are equalizing in case their concentration curves are below Lorenz curve of related income concept (concentration coefficient of a tax is lower than Gini coefficient). Hence, difference between Gini coefficient of income and concentration coefficient of an intervention, known as Kakwani index, is the optimal way to illustrate whether intervention is equalizing. 60. Results, presented in Table 10, suggest the direct transfers are largely progressive and, therefore, equalizing. In particular, pensions, direct utility subsidies and social assistance are most progressive, as their concentration coefficients are significantly below zero (implying that their concertation curves are far above 45-degree line). High equalizing effect and progressivity of direct utility subsidies and social assistance remains even if population is ranked by the post-fiscal disposable income, as they are means tested. Pensions on contrary lose much of their equalizing effect if population is ranked by the post-fiscal income due to their dominant role in the welfare of many households even if compared to income from employment. Other direct benefits, with exclusion of stipends, are also progressive and equalizing. Disequalizing effect of stipends is explained by better availability of higher education for population from high-income bins than from low- income bins. Privileges are also equalizing if it is compared to the distribution of the pre-fiscal (market) income. However, ranking by the post-fiscal (disposable) income indicates disequalizing effect of privileges. Despite some reforms in this area, aimed at reduction of public expenditure on privileges, they still create overlapping effect with direct benefits4. 61. Direct taxes are largely equalizing if measured both in relation to the market and disposable income. The incidence of PIT and social contributions are rather neutral even compared to the incidence of disposable income due to flat tax rates and deduction availability. Contrary to PIT and social contributions, land tax is disequalizing even despite some exemptions and diversified rates. It is rooted in relatively low income of population in rural area (employed in agriculture). 62. Indirect taxes can be considered either neutral or equalizing compared to the distribution of population by disposable income. VAT and sales tax have some equalizing effect as significant part of food products and basic services, that play higher role in purchases of low-income population, are exempted from taxation. Excises on alcohol and fuel tend to be even more equalizing, as they are consumed by upper deciles (according to KIHS data). 63. Contrary to direct transfers, indirect subsidies are regressive and disequalizing . Hot water and heating subsidy are available to urban population that is initially better off. Moreover, households from the upper deciles benefit most from them as they have better access to the utilities. The only equalizing indirect 438% of those enjoying privileges are also provided with benefits (excluding pensions). The ratio increases to 71.4% if pensions are considered as benefits. 30 subsidy is electricity subsidy, as it has special focus on rural population and population in high-altitude area that have higher risk of poverty than population in general. 64. In-kind transfers, in their turn, are largely progressive . Moreover, expenditures on primary and secondary education are not only equalizing, but also progressive. Households with children are vulnerable group in Kyrgyzstan with high risk of poverty. Hence, they benefit greatly from availability of education services. However, education system seems not to be able to eliminate poverty trap, as children from poor households do not tend to enter colleges or universities. As a result, public expenditures on post-secondary and tertiary education are disequalizing. 65. Expenditures on healthcare on average are rather neutral. However, provision of tertiary healthcare services may be viewed as one that have some equalizing effect on welfare of population, while provision of primary and secondary – disequalizing effect. It may be attributed to more equal availability of tertiary services compared to primary and secondary ones, as well as influence of copayments. Table 10. Progressivity of taxes and transfers in relation to income concepts Market income (PGT) Market income (PDI) Disposable income concentration Kakwani concentration Kakwani concentration Kakwani coefficient index coefficient index coefficient index All direct transfers incl contributory pensions -0.382 0.699 -0.319 0.553 0.167 0.048 Contributory Pensions -0.446 0.762 -- -- 0.193 0.022 Direct utility subsidies -0.341 0.657 -0.301 0.535 -0.334 0.549 Privileges -0.104 0.420 0.029 0.205 0.307 -0.092 Benefits -0.284 0.600 -0.332 0.566 0.034 0.180 Stipends 0.597 -0.281 0.560 -0.326 0.651 -0.437 Disability pensions -0.272 0.588 -0.295 0.529 0.146 0.068 Pensions on loss of breadwinner -0.198 0.515 -0.237 0.471 0.132 0.083 Social assistance -0.358 0.674 -0.464 0.698 -0.262 0.476 All direct taxes and contributions 0.504 0.188 0.380 0.146 0.309 0.095 PIT 0.489 0.173 0.387 0.152 0.307 0.093 Land tax 0.126 -0.190 0.100 -0.135 0.106 -0.108 Social contributions* 0.515 0.199 0.400 0.166 0.313 0.099 All indirect subsidies 0.222 0.094 0.340 -0.106 0.310 -0.096 Electricity subsidy 0.142 0.174 0.191 0.044 0.194 0.020 Hot water subsidy 0.353 -0.036 0.484 -0.249 0.411 -0.196 Heating subsidy 0.296 0.020 0.501 -0.266 0.436 -0.221 All indirect taxes 0.218 -0.098 0.269 0.034 0.275 0.061 VAT 0.227 -0.090 0.276 0.042 0.282 0.067 Sales tax 0.187 -0.129 0.237 0.003 0.241 0.027 Excise on alcohol 0.219 -0.097 0.307 0.072 0.339 0.125 Excise on tobacco -0.019 -0.335 0.157 -0.077 0.204 -0.011 Excise on fuel 0.274 -0.042 0.316 0.082 0.323 0.109 All net in-kind transfers 0.048 0.268 0.040 0.195 0.065 0.149 Net health transfers 0.035 0.282 0.142 0.092 0.197 0.017 Primary and secondary health expenditure 0.140 0.177 0.237 -0.003 0.271 -0.056 Tertiary (hospital) health expenditure -0.024 0.340 0.089 0.145 0.156 0.058 Net education transfers 0.057 0.260 -0.025 0.259 -0.018 0.232 Primary and pre-primary education -0.067 0.384 -0.149 0.383 -0.150 0.364 Secondary education 0.005 0.311 -0.077 0.311 -0.062 0.276 Post-secondary nontertiary education 0.308 0.008 0.239 -0.005 0.229 -0.015 Tertiary education 0.368 -0.051 0.290 -0.056 0.273 -0.059 All taxes 0.335 0.019 0.318 0.084 0.299 0.085 All net transfers and subsidies excl contributory pensions 0.036 0.280 0.060 0.174 0.136 0.079 Gini 0.316 0.234 0.214 Note. Concentration coefficient of a tax exceeding Gini coefficient of related income implies positive Kakwani index and progressivity of tax. Concentration coefficient of a transfer exceeding Gini coefficient of related income implies negative Kakwani index a nd regressivity of a transfer. Bold characters reflect income concept influenced by the intervention. * Social contributions including pension contributions for market income according to PGT approach and excluding them for PDI approach. Source: own estimates based on CEQ methodology. 31 66. The Kakwani indices reveal direction of redistribution impact of the fiscal interventions, but they do not provide information on the size of the impact interventions have on inequality or poverty. This aspect is captured by marginal contributions that show the change to inequality and local poverty measures after the application of intervention. In general, magnitude of equalizing / disequalizing effects depends on the size of interventions often more than on its progressivity / regressivity. 67. Significant expenditure on pensions make their marginal effect on poverty and inequality measures the most feasible (Table 11). Consequently, social contributions also influence poverty rates and Gini coefficient, but at lesser extent. First of all, it is explained by the size of intervention. Contributions are smaller than pensions, as they are partly financed through the budget. Second, the burden of social contributions is distributed among population more evenly than pensions. 68. Expenditure on secondary education generates effects on poverty and inequality of significant scale as well due to its size and progressivity . Benefits also contribute much to poverty and inequality reduction, which is more related to their progressivity rather than size. 69. Other interventions do not have significant influence over inequality measures . Nevertheless, VAT and PIT create some pressure on poverty risk, while electricity subsidy and tertiary healthcare expenditure contribute to its substantial reduction. Table 11. Marginal contributions to inequality and poverty Gini Moderate Poverty Extreme Poverty Privileges 0.000 0.0 0.0 Pensions 0.066 15.9 10.7 Benefits 0.012 4.0 1.2 Utility subsidy 0.000 0.0 0.0 Personal income tax 0.004 2.6 0.1 Social Contributions 0.020 5.7 0.2 Land tax 0.000 0.1 0.0 VAT 0.003 3.9 0.2 Sales tax 0.000 1.3 0.1 Excises on tobacco 0.000 0.1 0.0 Excises on alcohol 0.000 0.0 0.0 Excises on fuel 0.000 0.0 0.0 Electricity subsidy 0.001 2.6 0.1 Hot water subsidy -0.001 0.2 0.0 Heating subsidy -0.001 1.7 0.1 Primary and pre-primary education 0.003 1.2 0.1 Secondary education 0.011 4.7 0.5 Post-secondary nontertiary education 0.000 0.1 0.0 Tertiary education 0.000 0.9 0.0 Primary and secondary health expenditure 0.000 1.3 0.0 Tertiary (hospital) health expenditure 0.004 2.7 0.3 Note. Changes to inequality and poverty for the direct benefits and taxes are measured in comparison to the disposable income prior to the related intervention (i.e. market income (PGT) plus direct taxes and transfers net of analysed intervention). For indirec t subsidies and in-kind transfers changes to inequality and poverty are measured in comparison to the disposable income. Negative values for Gini mean increase of inequality due to fiscal intervention. Positive values for poverty mean increase of poverty headcount in c ase of abolishment of direct transfers, need to cover in full costs of utilities and public transportation, as well as pay for educa tion and healthcare services, and increase of poverty headcount caused by presence of taxes. Source: own estimates. 4.3. Efficiency of interventions 70. Neither Kakwani nor marginal contributions capture efficiency of the interventions . Marginal contributions are well-suited to evaluate the outcomes of social policies aimed at reducing poverty and inequality, like benefits, privileges, pensions and subsidies. But the marginal contributions miss another dimension of the fiscal interventions – their cost. While pensions have the biggest impact for poverty and inequality reduction in Kyrgyzstan, they are also the largest program in terms of cost. And we cannot compare it directly to a much smaller programs like privileges, for example. 32 71. Table 12 below lists several efficiency measures which weight the impact on poverty/inequality by the cost of the intervention. Efficiency measures are derived as the ratio of marginal contributions to the relative size of intervention, where the size is measured relative to GDP. Hence, efficiency measures represent the effect (on reduction of Gini or poverty) of 1% GDP spent on the particular intervention. The combined measure reflects the effect of 1% of GDP spent on the composite impact measure, consisting of poverty and inequality effects both weighted by 0.5. 72. According to these measures, contributory pensions still look very well, featuring one of the highest efficiency coefficients (Table 12). It is largely the same for standard old-age pensions and privileged pensions (Table 13). They have equal distribution by income bins, benefiting largely those from the first decile by market income. Other pensions provided within social insurance system have similar high efficiency of moderate poverty reduction, but they are less efficient in terms of extreme poverty and inequality reduction. It is related to differences in their incidence. Disability pensions benefit the first decile by market income at most, and upper deciles at least. However the difference in the volume of public spending on disability pensions allocated to the second and eighth deciles is rather small (see Annex), stressing that there is room for their progressivity improvement. Survival pensions are even less progressive, as upper 5 deciles by market income receive 38.7% of all survival pensions. 73. Benefits in general fare even better in terms of the “bang for the buck” compared to pensio ns. However, it is attributed largely to the monthly benefit for poor families with children that is a key instrument within social assistance system (Table 13). Its efficiency is explained by its progressivity. Errors of inclusion appear to be rather moderate, as there are practically no beneficiaries of MBPF among households from upper two deciles (see Annex). Still, there are signs of the presence of the errors of exclusion. Households from the poorest first decile receive MBPF less frequently than households from the next three deciles. It maybe rooted in the relative high costs of application for the benefit for the extremely poor households, as well as design of the benefit, missing vulnerable groups of urban population. 74. On contrary to MBPF, categorical benefits provided on monthly basis appear to be efficient only in terms of extreme poverty reduction . First, it implies that if provided to extremely poor households the size of benefit is not enough to eliminate moderate poverty (marginal effect on moderate poverty is lower than on extreme poverty). Second, in absolute terms public expenditures on monthly benefits are distributed rather equally with only moderate shift towards the first decile. Other benefits, reported by households within the KIHS, are rather efficient, benefiting almost solely first four deciles by market income (see Annex). Stipends are the only exemption, they increase inequality and have very low efficiency in terms of poverty reduction (Table 13). 75. Indirect subsidies are also inefficient. Aside from the fact that some of them increase inequality, even their poverty reduction effect is very expensive. The exception here is the electricity subsidy, which has substantial poverty-reducing effect. Despite being regressive it affects the risk of poverty as significant part of population has income slightly above the poverty line and any moderate shock may lead to their impoverishment. 5 76. Another measure of efficiency more related to how well targeted is the intervention, is the proportion of expenditure spent on bottom and on top decile . If the major purpose of the program is to fight poverty and inequality, top decile should receive disproportionately less, while bottom decile – disproportionately more. Any percentage higher than 10 for the top decile should raise red flags – it would mean the top decile gets disproportionately more than other deciles. Efficient programs, like contributory pensions or social assistance, spend below 5% of its costs on the top decile. Contributory pensions and direct 5 A shock of 2.6% disposable income reduction in the country (which corresponds to the average share of electricity subsidies in disposable income of population) leads to poverty increase of 2.3 p.p. 33 utility subsidies also predominantly go to the bottom decile, spending over 30% there. The core reason of inefficiency of stipends and indirect subsidies is their very poor targeting – they spend over 30% of their budget on the top decile, while spending relatively little or nothing on the bottom decile. Table 12. Efficiency measures for transfers, subsidies and in-kind programs Efficiency in Percentage spent Percentage Poverty Inequality Сombined on bottom decile spent on top reduction reduction decile Contributory Pensions 1.87 0.89 1.39 36.7% 2.9% All direct transfers: 2.33 0.69 1.50 20.2% 3.9% Direct utility subsidies 0.00 1.04 0.52 32.8% 4.9% Privileges 0.33 0.12 0.23 15.8% 9.9% Benefits, incl: 2.35 0.71 1.54 20.3% 3.7% Stipends 0.00 -0.81 -0.41 0.0% 33.4% All indirect subsidies 1.45 -0.29 0.58 11.3% 24.2% Electricity subsidy 2.22 0.06 1.14 8.5% 16.6% Hot water subsidy 0.29 -0.68 -0.19 12.0% 31.6% Heating subsidy 0.48 -0.68 -0.10 14.6% 32.4% Net health transfers 1.35 -0.10 0.62 8.7% 13.3% Primary and secondary health expenditure 0.77 -0.14 0.33 8.6% 18.8% Tertiary (hospital) health expenditure 1.14 -0.08 0.52 8.8% 10.2% Net education transfers 2.28 0.37 1.33 6.4% 8.8% Primary and pre-primary education 3.20 0.64 1.91 8.9% 6.0% Secondary education 2.39 0.44 1.43 6.7% 6.3% Post-secondary nontertiary education 2.24 -0.08 1.08 3.1% 16.6% Tertiary education 1.00 -0.19 0.41 3.4% 21.8% All net in-kind transfers 2.01 0.19 1.10 7.3% 10.5% All net transfers and subsidies excl contributory pensions 1.76 0.62 1.18 11.1% 13.4% All net transfers and subsidies incl contributory pensions 1.74 0.87 1.31 23.7% 8.3% Note. Efficiency is measured as a ratio of marginal effect to the size of intervention, measured in relation to GDP. Deciles are based on market income. Source: own estimates. Table 13. Marginal contributions of pensions and benefits to inequality and poverty reduction and their efficiency Marginal effect on (p.p.) Efficiency in reducation of Percentage spent Size, % of moderate extreme moderate extreme on bottom on top GINI GDP GINI poverty poverty poverty poverty decile decile old-age pensions 14.95 9.93 7.15 7.98 1.87 1.24 0.90 36.8 2.8 privileged pensions 0.98 0.63 0.44 0.51 1.91 1.23 0.86 34.8 4.5 disability pensions 1.82 0.61 0.59 1.01 1.81 0.61 0.59 23.6 4.7 pensions on loss of 0.63 0.10 0.13 0.24 2.64 0.42 0.55 19.9 3.9 breadwinner social monthly benefit (SMB) 0.06 0.10 0.04 0.06 0.96 1.61 0.64 23.1 5.1 monthly benefit for poor 0.91 0.19 0.39 0.33 2.79 0.58 1.19 9.4 0.0 families with children (MBPF) other social assistence 0.40 0.04 0.09 0.08 5.14 0.51 1.16 23.7 3.6 Source: own estimates. 4.4. Regional differences 77. The regional and rural/urban differences in fiscal programs (both transfers and taxes) were analyzed. Kyrgyzstan has 7 oblasts (Batken, Jalal-Abad, Issyk-Kul, Naryn, Osh, Talas and Chui oblasts) and 2 cities which are separate administrative entities. For each oblast we also look at the difference between urban and rural areas. Table 14 below documents average per capita expenditure on fiscal programs and per capita tax burden relative to the country average. 78. Per capita pensions are substantially higher than average in Naryn oblast, urban areas of Chui oblast – these facts are explained by the high concentration of people of retirement age in these regions. In particular, high expenditure on pensions in Naryn oblast are related to the fact that they are largely provided on privileged terms (earlier retirement, see Table 15) for high-altitude areas. Population of this region also receives substantially more survival pensions than the rest of the country. Disability pensions are more evenly 34 distributed by regions, but there is an evident shift in favour of urban population across the country. It may be explained by high application costs for disability pensions for rural population and better access to the healthcare system by urban population. 79. Per capita direct transfers are highest in Naryn oblast as well – over two times higher than country average (Table 14). It might be related to the fact that Naryn oblast has the lowest incomes in the country, and the highest poverty rate. Furthermore, it benefits from social assistance provided to the households in high- altitude regions. As a result, the Batken oblast, which also has a high poverty rate (37% versus 37.8% in Naryn oblast), does not receive direct transfers as high. The difference is driven largely by MBPF, which is received at substantial volume by rural and urban households of Naryn oblast (Table 15). In Batken oblast as well as other south parts of Kyrgyzstan (Osh oblast, Jalal-Abad oblast) the MBPF is received predominantly by rural population only. Other parts of the country benefit from MBPF at much lesser extent, although some of them (rural areas of Chui and Issykul oblasts) face relatively high poverty risks. 80. Monthly social benefits as well as other social assistance are distributed more evenly across the country still benefiting population of Naryn oblast at bigger extend than other regions (Table 15). Besides Chui urban households and Jalal-Abad rural households report to be provided with other social assistance (in the form of support from local authorities and employees) more often than population of Kyrgyzstan on average. 81. Indirect subsidies in Bishkek are 2.72 higher than the country average, in Naryn urban areas – 2.2 times higher. Rural areas across the country are getting substantially less indirect utility subsidies due to lower infrastructure access, underlining the inefficiency of this form of social support. Health transfers are substantially lower in Bishkek than in the rest of the country, reflecting higher incidence of copayments and the fact that people with higher incomes might be opting out of the state healthcare. However, low health transfers in rural areas of Issykkul and Talas, and urban areas of Jalal-Abad may have to do with low access. The size of education transfers are primarily demography-driven, with the children-abundant regions receiving more transfers. But in general the regional distribution of educational transfers is even. 82. The tax burden is quite unevenly distributed . Bishkek pays over 2 times more in direct taxes and pension contributions. The reasons are higher market incomes and lower level of informal economy. Urban areas of Issykkul, Naryn, and Chui oblast also have higher direct tax and pension contributions burden. The burden of indirect taxes is more evenly distributed, with Bishkek, urban areas of Issykkul and other relatively well-off regions paying more. Table 14. Fiscal programs by regions, relative to the country average Issykkul- Issykul - Jalal-Abad - Jalal-Abad - Naryn - Naryn - Bishkek city Osh city Urban Rural Urban Rural Urban Rural Poverty rate* 9.8 24.6 24.7 32.2 37.8 Pensions 1.17 0.89 1.18 1.12 0.72 0.79 1.69 1.46 Direct transfers 0.55 0.91 1.32 1.08 0.79 0.87 2.28 2.20 Indirect subsidies 2.72 1.12 1.65 0.42 0.73 0.35 2.20 0.38 Health 0.39 1.87 1.07 0.43 0.40 0.82 2.53 1.64 Education 0.99 0.81 1.06 1.05 1.10 1.16 1.25 1.15 Pension 2.40 0.90 1.30 0.65 1.05 0.57 1.59 0.53 contributions Direct taxes 2.24 0.95 1.29 0.66 1.08 0.62 1.54 0.53 Indirect taxes 1.24 1.11 1.30 0.91 0.85 0.82 1.34 0.82 Batken – Batken - Osh - Osh - Talas - Talas - Chui - Chui - Urban Rural Urban Rural Urban Rural Urban Rural Poverty rate* 37.0 22.0 18.1 30.3 Pensions 0.95 0.82 0.78 0.98 0.68 0.79 1.55 1.03 Direct transfers 1.18 1.22 1.21 1.15 0.69 0.98 0.97 0.91 Indirect subsidies 0.60 0.29 0.45 0.42 1.71 0.68 3.11 0.86 Health 1.38 1.14 1.20 1.23 0.30 0.23 1.90 1.40 Education 0.94 0.96 0.76 0.92 1.29 1.06 0.95 0.93 35 Pension 0.84 0.35 0.37 0.54 0.90 0.45 1.57 1.12 contributions Direct taxes 0.84 0.36 0.59 0.62 1.09 0.49 1.50 1.07 Indirect taxes 1.05 0.81 0.98 1.05 0.57 0.62 1.18 1.02 * Poverty rate is in per cent, according to national definition . Note: Pensions stand for contributory pensions. Pension contributions are defined as a share of Social Fund contributions that go es to finance contributory pensions. The rest of the Social Fund contributions is included in the direct taxes. Aside from that, di rect taxes also include PIT, and taxes on entrepreneurship and patents. Source: Kyrgyzstat (2017) for poverty rate; own estimates. Table 15. Pensions and benefits reported by region, USD cents PPP per day per capita poverty old-age privileged disability survival other social rate pensions pensions pensions pensions assistence SMB MBPF Bishkek - Urban 9.8 99.3 0.3 7.5 1.8 0.4 0.1 0.1 Issykul - Urban 15.1 100.2 0.7 14.4 3.4 0.9 2.0 0.4 Issykul - Rural 30.7 95.7 0.0 9.8 5.4 0.1 1.2 2.1 Jalal-Abad - Urban 35.2 58.9 2.5 10.2 1.1 0.4 0.4 2.1 Jalal-Abad - Rural 29.4 66.1 1.2 6.7 1.0 1.7 0.6 5.5 Naryn - Urban 27.7 13.8 130.2 14.4 9.0 3.0 3.8 7.0 Naryn - Rural 39.8 46.0 78.3 13.2 10.1 1.4 2.3 11.9 Batken - Urban 25.7 76.0 4.8 13.3 3.2 1.2 1.2 1.2 Batken - Rural 42.2 69.3 0.8 10.7 1.3 0.4 1.0 7.4 Osh - Urban 35.8 66.1 0.0 18.2 0.5 0.4 0.7 1.8 Osh - Rural 21.5 82.8 0.4 9.7 2.6 0.2 0.2 6.1 Talas - Urban 12.9 55.2 2.5 8.1 3.2 0.0 1.1 0.0 Talas - Rural 19.6 63.4 3.6 10.9 2.8 0.0 1.5 2.4 Chui - Urban 16.0 128.8 3.2 12.2 2.2 2.6 0.3 0.0 Chui - Rural 33.8 84.1 3.4 12.9 1.3 1.4 0.1 0.0 Osh city 24.1 75.2 0.2 12.4 2.2 0.3 0.6 0.0 Note. Poverty rate is in per cent, according to own estimates. Source: Kyrgyzstat (2017) for poverty rate; own estimates. 83. Hence, fiscal interventions reduce disparities across regions. Figure 16 shows how the change in average per capita income from direct fiscal interventions (a change from market to disposable income) of the region depends on its initial (pre-fiscal) average per capita market income. There is a clear negative relationship between the initial value of income and redistribution, suggesting that redistribution is equalizing across regions. The only region which becomes worse off after the direct interventions is Bishkek, which has highest market income. The region with the lowest pre-fiscal income, Naryn, also has the biggest positive change as a result of direct fiscal interventions. The rest of the regions also get the fiscal transfer proportionate to their initial income, and no outliers are present. Figure 16. Change from market (pre-fiscal) to disposable average regional income, by market income, daily 2011 PPP USD per capita 1.5 Change from market income to disposable Naryn 1 0.5 income 0 3 4 5 6 7 8 -0.5 Bishkek -1 -1.5 Market income (daily PPP USD per capita) 36 Source: own estimates. 4.5. Targeting the vulnerable groups 84. Since in the CEQ analysis the fiscal interventions are attributed to the whole household, our analysis of vulnerable groups is also household-based. In the analysis we focus on gender composition of the households; on households with children; on single-parent and single-earner households; and on households with individuals belonging to vulnerable groups. The table below (16) shows the average value of transfers and taxes per capita relative to the average value of the corresponding tax/transfer across the whole population. Here and thereafter we adopted the definition of adults and children which is standard for the World Bank poverty analysis in Kyrgyzstan: adults are everyone aged 15 or older; while children are defined as those younger than 15. Table 16. Fiscal programs for the households by gender and family composition, relative to the country average, per capita HH by adults composition Single adult households with children Majority of Equal no. of men and With three or more women Majority of men women All children Pensions 1.22 0.75 0.95 0.44 0.16 Direct transfers 1.05 1.00 0.96 1.56 1.80 Indirect subsidies 1.12 0.86 0.97 1.89 1.05 Health 1.05 0.98 0.97 1.13 0.88 Education 1.06 1.27 0.84 1.24 1.37 Pension contributions 1.01 1.08 0.96 1.04 0.50 Direct taxes 1.00 1.07 0.97 1.01 0.49 Indirect taxes 1.00 1.02 0.99 1.22 0.89 Source: own estimates. 85. The major differences between predominantly female and predominantly male households stem from demographic reasons: households with a lot of elderly tend to be female-dominated due to higher life expectancy of women. It shows up in the significant uptick of pensions for the households with female majority among adults. These households also have better access to the indirect subsidies. Households where the majority of adults are men enjoy higher education expenditure, probably due to lower average household age. Single parent households (these households predominantly have a female head – in 95% of the cases) obtain higher amounts of direct transfers as they often have low incomes and qualify for targeted social assistance. They also have higher education expenditures assigned to them due to higher share of children in the household. Single parent households with multiple children face lower per capita direct taxes since they have a lower proportion of earners in the household. Table 17. Fiscal programs for the vulnerable groups by households with children, relative to the country average, per capita HH by no. of children Three or more No children One child Two children children Pensions 2.19 0.81 0.67 0.58 Direct transfers 1.44 0.93 0.75 0.95 Indirect subsidies 1.94 1.00 0.77 0.54 Health 1.34 0.94 0.95 0.86 Education 0.74 1.14 0.96 1.11 Pension contributions 1.72 1.13 0.80 0.58 Direct taxes 1.68 1.13 0.81 0.59 Indirect taxes 1.48 1.10 0.88 0.70 Source: own estimates. 86. As we can see in Table 17, households with children in general have lower than average access to direct transfers and indirect subsidies. Usually these are relatively young families, residing in smaller homes and not qualifying for social aid. These families, however, pay lower direct and indirect taxes per capita. 37 This is especially true for families with 3 or more children. This result is driven by the fact that the proportion of earners (taxpayers) in these households is lower. 87. Households with a low number of earners is a separate group, which normally pays low taxes and recieves a lot of transfers. The households with no earners are predominantly retiree households, and they recieve relatively high transfers, including pensions (see Table 17). They also (almost) do not pay direct taxes, although indirectly they are taxed as average. Surprisingly, they also recieve high indirect subsidies. High level of health expenditure is attrbuted to those households since they have higher demand for healthcare, and also often are exempt from copayment. Single-earner households in general are not significantly different from average. No earner households with children enjoy high transfers and low taxes. This is different for single-earner households with children: the amount of transfers they recieve is close to average, while relatively low indirect tax amounts suggest lower consumption. Table 18. Fiscal programs for the households with low number of earners, relative to the country average, per capita HH by no. of earners Single earner, with No earners Single earner No earners, with children children Pensions 3.36 1.04 1.21 0.73 Direct transfers 2.19 1.12 2.01 1.00 Indirect subsidies 1.62 1.20 0.73 0.92 Health 1.63 1.15 1.18 1.06 Education 0.86 0.96 1.13 1.02 Pension contributions 0.09 0.92 0.09 0.70 Direct taxes 0.12 0.92 0.11 0.71 Indirect taxes 0.99 1.02 0.73 0.89 Source: own estimates. 88. In Table 19 we focused on the households with the following vulnerable groups: persons with disabilities, informally employed, labor migrants and unemployed . The choice to view the informally employed as a vulnerable group could be questionable, given that they forgo mostly the future transfers (pensions) by avoiding the today’s taxes. Furthermore, they may opt to contribute to the pension system via a lump-sum payment reaching retirement age. However, it is interesting to see if they enjoy a similar access to other transfers. Table 19. Fiscal programs for the households with different vulnerable groups, relative to the country average, per capita HH with different vulnerable individuals Person(s) with Informally Farmers and disabilities Labor migrants Unemployed employed entrepreneurs Pensions 1.19 0.53 0.75 0.68 0.70 Direct transfers 8.46 0.88 0.64 0.83 0.94 Indirect subsidies 1.11 0.58 1.61 0.77 0.72 Health 2.23 1.22 0.51 1.12 0.88 Education 0.69 1.14 0.93 0.91 0.94 Pension contributions 0.65 0.32 1.06 0.50 0.42 Direct taxes 0.70 0.35 1.04 0.53 0.61 Indirect taxes 1.14 1.25 1.01 0.99 0.99 Source: own estimates. 89. Persons with disabilities receive direct transfers 10 times higher than the rest of the population (this includes disability pensions, privileges and social assistance). They also receive higher health transfers due to higher demand for health services. Due to lower or absent market income and lower rates of PIT and social contributions, households with persons with disabilities pay lower direct taxes, and same indirect taxes as the rest of the population. 90. Informally employed do have lower access to the direct transfers, pensions and indirect subsidies. They also have average access to health and education. They do pay almost two times less direct 38 taxes, but their indirect tax burden is only insignificantly lower than average. In general, they benefit from public transfers and contribute to the tax system at the same scale as farmers and entrepreneurs, which stresses rather beneficiary status of the latter. Labour migrants pay even less taxes, but at the same time they benefit from direct transfers, pensions and indirect subsidies at much lower extent than population on average. Unemployed get surprisingly low amounts of direct transfers, and poor access to health services. Nevertheless, unemployed pay only slightly lower direct and indirect taxes. Hence, they appear to be net losers from the system of fiscal interventions. 91. Distortions in incidence of fiscal interventions related to the social-economic status of a person are rooted in the prevalence of informal sector in the economy . Simplified tax regime for farmers and entrepreneurs and wide-spread practice of “envelope” wages result in low tax burden for informally employed. Lower tax burden is observed only for households with labour migrants. At the same time, direct transfers play more important role in the income of households with informally employed members than in the households with formally employed or unemployed ones. Moreover, formally employed carry most of the burden related to personal income taxation and social security contributions. Households with unemployed, being largely integrated into the formal sector, also carry significant burden of direct taxation and social contributions, although social assistance do not explicitly target them. Hence, there is redistribution effect from the formal sector of economy to the informal sector, which is rooted in the pension system provisions and limited development of targeted instruments of social assistance. Figure 17. Change from market (pre-fiscal) to disposable average regional income, by market income, daily 2011 PPP USD per capita 10 25 9 8 20 7 6 15 5 4 10 3 2 5 1 0 0 formally infromally labour unemployed formally infromally labour unemployed employed employed migrants employed employed migrants (incl. (incl. entrepreneurs entrepreneurs and farmers) and farmers) direct taxes, % of market (pre-fiscal) income pension contributions, % of market income direct transfres, % of disposable (post-fiscal) income received contributory pensions, % of disposable income Source: own estimates. 4.6. Cross-country comparisons 92. Figure 18 below shows how the direct taxes and transfers redistribute income in different countries. In particular, the figure shows decline in Gini when we go from market to disposable income concepts (both PDI and PGT scenarios), capturing the equalizing effects of direct transfers and direct taxes. The sample of countries is limited to those where the CEQ analysis was performed. 39 Figure 18. Effects of redistribution through direct taxes and transfers across countries 0.25 0.2 Gini reduction 0.15 0.1 0.05 0 Dominican Republic… Tanzania (2011) Indonesia (2012) EU-28 (2011) Costa Rica (2010) Honduras (2011) Armenia (2011) Colombia (2010) Russia (2010) Ecuador (2011) Ethopia (2011) Tunisia (2010) Argentina (2012) Uruguay (2009) Kyrgyzstan (2016) United States (2011) Georgia (2013) Sri Lanka (2010) Guatemala (2011) Belarus (2015) Jordan (2010) Brazil (2009) Bolivia (2009) Peru (2009) Mexico (2010) Chilie (2013) Ghana (2013) El Salvador (2011) South Africa (2010) Gini PGT Gini PDI Source: Lustig (2016), Lopez-Calva et al. (2017), Bornukova et al. (2017), and own estimates. 93. The effects of fiscal redistribution are quite sizable in Kyrgyzstan compared to other countries. The countries, which are ahead in the equalizing effects of fiscal policy, also have substantially higher GDP. In other words, for its level of GDP Kyrgyzstan is quite successful in equalization efforts compared to other countries. 94. As in other post-Soviet countries, in Kyrgyzstan pensions dominate the effect , as can be seen in the difference between PGT (with pensions) and PDI (without pensions) effects on Gini. The non-pension equalization effect (Gini PDI) in Kyrgyzstan is lower than in Belarus, Russia or Armenia, mostly due to the absence of the unconditional childcare subsidies and lower percentage of GDP spent on social assistance altogether. Figure 19. Effects of redistribution by income concepts across countries; changes in Gini index 0.040 0.020 0.000 -0.020 -0.040 -0.060 -0.080 -0.100 -0.120 -0.140 -0.160 from market to disposable from disposable to consumable from consumable to final Georgia, 2013 Armenia, 2011 Russia, 2010 Belarus, 2015 Kyrgyzstan, 2016 Source: Cancho and Bondarenko (2017), Younger and Khachatryan (2017), Lopez-Calva et al. (2017), Bornukova et al. (2017), and own estimates. 95. In all post-Soviet countries covered by CEQ analysis the redistributional effects are concentrated in the direct transfers (including pensions) and direct taxes . The changes from disposable to consumable income (due to indirect taxes and subsidies) are relatively small, but increase inequality (see Figure 19). The effects of in-kind transfers (health and education) are moderate, but equalizing. Kyrgyzstan, however, stands out due to the relatively small effects of in-kind transfers. 40 Table 20. International comparisons by programs, Gini marginal effects and Kakwani indices Marginal contributions to Gini reduction from Kakwani index to market income market to final Kyrgyz- Kyrgyz- Georgia Armenia Russia Belarus Georgia Armenia Russia Belarus stan stan Redistributive effect (total) 12.4 11.4 18.5 18.0 10.6 Direct taxes 2.2 2.1 2.8 2.3 2.6 0.182 0.097 0.035 0.018 0.188 Direct transfers 10.0 9.2 14.5 11.9 7.2 0.706 0.660 0.775 0.691 0.699 old-age pensions 5.7 6.7 11.8 10.0 6.0 0.658 0.641 0.791 0.757 0.762 Indirect taxes -1.4 0.1 0.0 -0.3 0.3 -0.230 -0.129 -0.197 -0.221 -0.098 Indirect subsidies 0.0 0.0 -- 0.2 -0.7 0.372 0.381 -- 0.280 0.094 In-kind education 2.0 1.4 1.0 0.542 0.519 0.260 3.4 4.3 0.429 0.401 In-kind health -0.8 0.4 -0.2 0.636 0.499 0.282 Source: Cancho and Bondarenko (2017), Younger and Khachatryan (2017), Lopez-Calva et al. (2017), Bornukova et al. (2017), and own estimates. 96. Kyrgyzstan has more progressive direct taxes than other countries in the region . This results in higher redistributive effects of taxes. The marginal contribution of direct transfers to inequality reduction is the lowest in Kyrgyzstan, partly due to the relatively low size of non-pension benefits. While pensions are the main driver of inequality reduction in Kyrgyzstan, relative to Belarus and Russia pensions appear to have comparatively low impact on Gini. However, it might be attributed to the initially low level of inequality according to the pre-fiscal market income concept, compared to other countries, which limits possible scale of redistribution effects of direct transfers. 97. As Table 20 shows, the low effect of the in-kind transfers on Gini reduction is due not to the size of the in-kind spending, but to the fact that it is less equalizing , as reflected by Kakwani index. This is true both for health and education. 98. In contrast to other countries of the region, indirect subsidies in Kyrgyzstan have evident regressive nature and generate inequality growth . Meanwhile, indirect taxes are distributed more neutrally than in other countries. 5. Conclusions 99. Measures of fiscal policy in Kyrgyzstan create significant redistribution effects given their scale and level of economic development of the country . Most of the effect comes from direct fiscal interventions, as they reduce Gini inequality index from 0.316 to 0.234 and national poverty measure from 37.4% to 25.4% (comparing disposable income to market income). Pensions have the largest effect on poverty reduction (by 15.9%), and inequality reduction (by 0.066 of Gini). Indirect subsidies and taxes on contrary slightly increase poverty and inequality due to regressive nature of utilities subsidies, while in-kind transfers contribute to moderate inequality reduction. 100. The net recipients of fiscal redistribution are the first two deciles by market income, while the higher deciles are net donors. Nevertheless, around 40% of post-fisc poor population faces welfare reduction due to income redistribution caused by fiscal policy and pension system. This result suggests there is a room to improve accessibility of social transfers and tax exemptions. In particular existence of minimum tax base for PIT and social contributions implies that effective tax rates for low income households might be higher than nominal. 101. Despite its large size, pensions are one of the most efficient programs, providing high marginal effects on poverty and inequality per each som spent . The social role of pensions system is very high, but its sustainability is questionable. Hence, any future pension reform should be carried out only after careful consideration of its consequences. 102. Social assistance programs are very efficient, especially targeted ones. The expansion and better accessibility of these programs could be considered. As of 2016 significant part of vulnerable households were 41 not covered by the system of monthly benefits for poor families with children due to design of eligibility criteria. At the same time, errors of inclusion appeared to be modest as households from top deciles were not provided with targeted assistance. Categorical benefits and pensions provided based on disability or loss of the breadwinner are also quite efficient. Still, there are signs of their low accessibility for rural population in comparison to urban areas. 103. On contrary, indirect utility subsidies are expensive and very inefficient due to poor targeting. Over 30% of the heating and hot water subsidies goes to the top income decile. Electricity subsidy is also regressive, but at lesser extent. At the same time, indirect utility subsidies are quite large in size – volume of public expenditure on utility tariffs subsidization is similar to the expenditure on all direct transfers (excluding pensions, see World Bank (2019)). As there is a significant share of population with income close to the poverty line, it implies feasible poverty reduction effect of the utility subsidies. Analysis of distributional impact of energy tariff increase in World Bank (2019) reveled that households with three and more children, households in high-altitude areas and energy-poor households in Bishkek might be the most vulnerable to the tariff increase. Hence, results of the CEQ analysis supports reduction of indirect utility subsidies, but it should be accompanied by introduction of targeted compensatory measures for vulnerable groups. One of the solutions is extension of the Bishkek’s housing and utility program targeted at subsidizing utiliti es for poor households.6 104. The fiscal system, as of 2016, did not target well unemployed and households with children . Poverty risk for households with children is substantially higher than average in Kyrgyzstan, and provision of monthly benefits for poor households with children was not enough to mitigate this risk. It was designed to alleviate extreme poverty and had focus on households in high altitude areas, thus covering only a fraction of poor households with children. Unemployed were not provided with meaningful measures of social assistance, although their households have been often contributing to the budget via direct taxes. For comparison, those informally employed benefit from the system of social support at similar extent as population on average, paying very low direct taxes. It implies that fiscal system generates redistribution from formal sector to informal. It stresses need for revision of principles of taxation of informal sector and support provision for unemployed. 105. Fiscal interventions appear to reduce regional disparities. Bishkek is the net donor among the regions; while the rest of the regions are beneficiaries with the benefit size dependent negatively on the pre- fiscal income. The most benefiting region is Naryn, which is rooted in the privileges and special scale-up coefficients applied for the programs provided to population in high altitude areas. References Bastagli, F. (2015) “Bringing Taxation into Social Protection Analysis and Planning.” Working Paper 421 (London: Overseas Development Institute). Bornukova, K., Shymanovich, G. and A. Chubrik (2017). “Fiscal Incidence in Belarus : A Commitment to Equity Analysis”. Policy Research Working Paper; No. 8216. World Bank, Washington, DC. Cancho, C., & Bondarenko, E. (2017). “The Distributional Impact of Fiscal Policy in Georgia ”. The Distributional Impact of Taxes and Transfers: Evidence from Eight Low-and Middle-Income Countries, 113-48. Deaton, A., Zaidi, S. (1999) Guidelines for Constructing Consumption Aggregates for Welfare Analysis, Working Papers from Princeton, Woodrow Wilson School - Development Studies. 6 For details see World Bank (2019). Please note, the KIHS does not allow a separate analysis of the Bishkek program given the limited number of observations. 42 Kyrgyzstat (2017), “The standard of living of the population of the Kyrgyz Republic” , 2012-2016. National Statistical Commitee of the Kyrgyz Republic, Bishkek, 2017. Lustig, N. (2016) “Fiscal Redistribution in Low and Middle Income Countries.” Chapter 8 in N. Lustig (editor) Commitment to Equity Handbook. A Guide to Estimating the Impact of Fiscal Policy on Inequality and Poverty, Tulane University and the World Bank, June 2017. Lopez-Calva, L. F., N. Lustig, M. Matytsin and D. Popova (2017) “Who Benefits from Fiscal Redistribution in Russia?”, in The Distributional Impact of Fiscal Policy: Experience from Developing Countries, edited by Gabriela Inchauste and Nora Lustig. (Washington: World Bank). Higgins, S. and N. Lustig (2017). “Allocating Taxes and Transfers and Constructing Income Concepts: Completing Sections A, B, and C of the CEQ Master W orkbook”. Chapter 7 in Commitment to Equity Handbook. A Guide to Estimating the Impact of Fiscal Policy on Inequality and Poverty , edited by Nora Lustig, Tulane University and the World Bank, June 2017. World Bank (2014a) Kyrgyz Republic Public Expenditure Review Policy Notes: Pensions. (Washington: World Bank). World Bank (2014b) Kyrgyz Republic Public Expenditure Review Policy Notes: Social Assistance. (Washington: World Bank). World Bank (2014c) Kyrgyz Republic Public Expenditure Review Policy Notes: Health. (Washington: World Bank). World Bank (2014d) Kyrgyz Republic Public Expenditure Review Policy Notes: Health. (Washington: World Bank). World Bank (2014e) Kyrgyz Republic Public Expenditure Review Policy Notes: Strategic Seting. (Washington: World Bank). World Bank (…) A Methodological Manual: Poverty Measurement in the Kyrgyz Republic, World Bank Group and National Statistical Committee of the Kyrgyz Republic. World Bank (2019). Kyrgyz Republic: Distributional Impact of Energy Tariff Reform, Poverty and Equity Global Practice, Europe and Central Asia Region, Report. Younger, S., & Khachatryan, A. (2017). “Fiscal incidence in Armenia”. The Distributional Impact of Taxes and Transfers: Evidence from Eight Low-and Middle-Income Countries, 43-78. 43 Annex. Incidence of pensions and benefits by deciles based on market income (pre-fiscal) concept 70 3.5 4.5 0.20 4.0 0.18 60 3.0 3.5 0.16 50 2.5 0.14 3.0 40 2.0 0.12 2.5 0.10 30 1.5 2.0 0.08 1.5 0.06 20 1.0 1.0 0.04 10 0.5 0.5 0.02 0 0.0 0.0 0.00 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 USD PPP per day (right axis) USD PPP per day (right axis) % of disposable income % of disposable income (a) old-age pensions (b) privileged pensions 6.0 0.30 1.4 0.06 5.0 0.25 1.2 0.05 1.0 4.0 0.20 0.04 0.8 3.0 0.15 0.03 0.6 2.0 0.10 0.02 0.4 1.0 0.05 0.2 0.01 0.0 0.00 0.0 0.00 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 USD PPP per day (right axis) USD PPP per day (right axis) % of disposable income % of disposable income (c) disability pensions (d) survival pensions (pensions on loss of breadwinner) 0.50 0.016 2.5 0.08 0.45 0.014 0.07 0.40 2.0 0.012 0.06 0.35 0.30 0.010 1.5 0.05 0.25 0.008 0.04 0.20 0.006 1.0 0.03 0.15 0.004 0.02 0.10 0.5 0.05 0.002 0.01 0.00 0.000 0.0 0.00 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 USD PPP per day (right axis) USD PPP per day (right axis) % of disposable income % of disposable income (e) social monthly benefit (f) monthly benefit for poor families with children 44 0.50 0.020 0.45 0.018 0.40 0.016 0.35 0.014 0.30 0.012 0.25 0.010 0.20 0.008 0.15 0.006 0.10 0.004 0.05 0.002 0.00 0.000 1 2 3 4 5 6 7 8 9 10 USD PPP per day (right axis) % of disposable income (g) other social assistence Note. Deciles are determined based on the pre-fiscal market income, while effect is measured compared to the post-fiscal income. Source: own estimates. 45