The Distributional Impact of Serbia’s Taxes and Social Spending

The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent.


I. Introduction
Serbia's economic reform agenda aims to accelerate economic growth and boost convergence in living standards with the European Union.To support that process, macroeconomic sustainability and fiscal policy for inclusive growth are an important part. 2 Serbia achieved notable fiscal consolidation efforts in 2015-2019 to bring down high public debt and stabilize its macroeconomic foundations, and economic growth and labor market improvements during this period benefitted the population.Yet more structural reforms are needed to boost private sector dynamism and jobs, improve welfare of citizens, and strengthen the middle class in Serbia.In 2019, 23.2 percent of Serbia's population was still considered at risk of poverty, defined as living below 60 percent of the median per-adult-equivalent income.The COVID-19 pandemic and economic crisis brought earlier progress to a halt and underscored the need for a green, resilient, and inclusive recovery.
In this context, understanding the impact of the fiscal system on inequality and poverty is key to inform policy choices.To better understand how the government redistributes resources and the potential equity and distributional impacts of any future fiscal reform, it is useful to have a comprehensive assessment of the redistributive effect of the current tax-benefit system.The equity impacts should be considered along with other goals of fiscal policy, such as efficiency and sustainability.This paper's key research question is to analyze the incidence of fiscal policy in Serbia, that is, the impact of government spending and tax collection on households and on poverty and inequality.The results will shed light on the extent to which the government redistributes resources, which can serve as an input into the public policy debate about whether a tax or a benefit, or what combination, is desirable.To our knowledge, this is the first paper that analyzes and presents a comprehensive assessment of how the distribution of household income is affected by each type of tax or social transfer and by the whole system in Serbia.The paper advances on the previous literature as follows.Earlier work has looked at the impact of direct taxes and direct transfers using SRMOD, a version of the EUROMOD model (Randjelovic and Zarkovic, 2013) and estimated the effects of fiscal instruments on income inequality using a decomposition approach (Krstic and Zarkovic, 2017).However, these studies did not include indirect taxes, which are a major source of revenue collection, and did not look at poverty impacts.This paper incorporates these aspects and, in addition, includes inkind transfers such as education and health, which are major components of social spending, in assessing impacts on inequality.
This paper follows the Commitment to Equity Assessment (CEQ) approach 3 and, to our knowledge, is the first to apply the methodology and generate new findings for Serbia.Compared to other approaches, using the CEQ has the benefits of assessing the comprehensive fiscal system as a whole, assessing impacts for various types of households, and comparison with the many countries where this same approach has been applied.The paper uses the household survey data available at the time of analysis, namely the 2017 Survey of Income and Living Conditions (SILC), with 2016 income data, and 2016 budget and administrative data from the Statistical Office of Serbia, Ministry of Finance, Ministry of Education, Science and Technological Development, and Ministry of Health.It examines the following main questions: How much redistribution and poverty reduction are accomplished through social spending, subsidies, and taxation in Serbia?How progressive are revenue collection and government spending?Based on the data available, the analysis covers fiscal components that represent about 65 percent of direct taxes (83 percent of contributions), 50 percent of indirect taxes revenues, and 82 percent of spending.As with many other CEQ studies, certain categories such as corporate tax and spending on infrastructure are excluded due to the difficulty of assigning the benefit or burden to individual households based on the available information.
Our findings show that Serbia's fiscal system reduces inequality, though it has a povertyincreasing effect.The system is redistributive, reducing the Gini coefficient of income from 0.38 at market income to 0.35 at final income, including the benefit of public expenditures on education and health.However, poverty, measured as per-capita income below $5.5/day in 2011 PPP, is estimated to increase by about 7 percentage points, from 15.3 percent in market income (including pensions as deferred income) to 22.4 percent in consumable income after reflecting direct and indirect taxes and transfers.Serbia's social benefits are inequalityreducing.Direct taxes in Serbia are overall slightly progressive, and their impact on lowering inequality is smaller than that of direct transfers and the impact observed in other countries.The flat personal income tax gives little progressivity while regressivity of Value-Added Tax (VAT) with respect to current income increases inequality.With respect to consumption when looking at welfare over the life cycle, the incidence of VAT is much less regressive.Future analyses can assess the impacts of potential changes to a tax or a benefit, or a combination of several changes, in the system.Simulating such changes would demonstrate further use of this analysis for policy making.
The paper is organized as follows.The next section describes the system of taxes and social spending.Section III presents the data, assumptions, and methodology in estimating the impact, combining household survey, national accounts, and administrative data.Sections IV and V show the results regarding the impact of the net fiscal system and of each fiscal intervention, respectively.The following section concludes with direction for further use of the analysis.

II. The structure of taxes and social spending in Serbia Fiscal revenue
The public finance system in Serbia consists of: ) and local-level state owned enterprises (utility companies).The structure of tax revenues in Serbia is presented in Table 1, and Appendix 3 includes detailed description of the taxes and spending system.Indirect taxes and social contributions make up the bulk of fiscal revenue.In 2016, indirect taxes made up approximately 49.8 percent of total tax receipts of the general government, direct taxes (personal and corporate income tax) 15.5 percent, while social security contributions 34.7 percent.Two main sources of tax revenues are value added tax (VAT) and social security contributions (SSC), together accounting for 64 percent of total tax collections.Our analysis is focused on the major tax items, including personal income taxes, SSC, VAT, and excise duties. 4hese taxes account for 92 percent of total tax revenues in Serbia in 2016.

Direct taxes and social security contributions
Personal income tax (PIT) revenues amounted to 3.8 percent of GDP in 2016.PIT in Serbia applies to wages/salaries, income from self-employment (including entrepreneurs and farmers), and income from authorship and intellectual property rights.The Serbian tax system does not provide for the option of joint, synthetized taxation, which means that all sources of income are assessed separately.Taxation of wages/salaries is the main component of the PIT system, accounting for 73 percent of PIT revenue.
The employment PIT rate in Serbia is flat at 10 percent,5 and it applies to wage income exceeding a personal allowance of RSD 11,604 per month (approximately 19 percent of average gross wage) in 2016.Income from self-employment is also subject to 10 percent PIT.Income from capital, rental income and capital gains are taxed at 15 percent.Income from authorship and intellectual property rights are taxed at 20 percent.Old-age pensions and social benefits are exempted from taxation.Individuals whose total annual net income from employment, self-employment, authorship and intellectual property rights and other income exceeds a threshold (three times the average annual wage in Serbia) need to pay annual income tax at the rates of 10 percent (on taxable income between three and six times the average annual wage in Serbia) and 15 percent (on income above six times the average annual wage in Serbia).No Sources: Authors' calculations based on Ministry of Finance Note: "Included in analysis" refers to the items that can be identified in the household data for this analysis and the amount that can be estimated and reflected in the analysis.
Social security contributions (SSC) represent a large source of public revenues in Serbia, at 13 percent of GDP (pension contributions -9.15 percent of GDP, health contributions -3.4 percent of GDP and unemployment contributions -0.6 percent of GDP).The social security system in Serbia consists of three components: pension and disability insurance, health insurance, and unemployment insurance.This is a Bismarckian social security system financed from mandatory contributions payable on personal income.In 2016, the contribution rate for pension and disability insurance was 26 percent (14 percent employee and 12 percent employer), health insurance 10.3 percent (5.15 percent employee and 5.15 percent employer), and unemployment insurance contributions 1.5 percent (0.75 percent employee and 0.75 percent employer).All social contributions are paid on employment income and also on part of self-employment income ("entrepreneur's salary").Income from authorship and intellectual property rights is subject to pension contributions, while health insurance contributions are paid only if taxpayer is not employed.Health insurance contributions are also paid on pensions (the payment is made by the Pension Fund -see Appendix 3).

Indirect taxes
Indirect taxes are the largest source of revenue, of which VAT is the largest single component, amounting to 11.2 percent of GDP (25.6 percent of total tax revenues) in 2016.The standard VAT rate levied on most goods and services in Serbia is 20 percent.The reduced VAT rate of 10 percent is applied to basic food (bread, milk, meat, etc.), water, natural gas, drugs, books, concerts tickets, culture/art magazines, etc. (see Appendix 3). 6cise duties generated revenues of 6.5 percent of GDP in 2016.Excise duties are levied on tobacco and tobacco products, coffee, alcoholic drinks, and oil/oil derivatives.Starting from August 2015, excise duties are also charged on electricity.

Social spending
Overall public expenditures in Serbia amounted to 47.3 percent of GDP in 2016.Spending on social transfers (contributory and non-contributory benefits) was 15.2 percent of the GDP, out of which the largest share was for contributory social insurance (12.8 percent of GDP).Spending on education was 3 percent of GDP and health 5.9 percent of GDP (Table 2).Thus, total social spending (social transfers, education, and health) was 24 percent of the GDP.The analysis includes major social spending measures and covers 82 percent of social spending.
The main non-contributory social assistance programs that can be analyzed in the available data include the following.This description pertains to the year of analysis (2016), noting some of the changes introduced in 2018.The child allowance is conditional on school enrollment for those older than the age of 7. Parental allowance refers to a one-off payment to the mother for giving birth to the first child, and payment in 24 equal monthly installments for giving birth to the second, third, and fourth child.With the law changes in 2018, the government increased the allowance amounts and extended the benefit period for the third and fourth child.Now the allowance can be received until child reaches 10 years of age.Finally, changes to the wage compensation during maternity leave increased the length of the employment period prior to the leave that is taken into account for the calculation of the wage compensation.Also, the benefit was extended to the group of mothers in vulnerable employment (working in agriculture, on flexible contracts).4. War veterans benefits.There are many care benefits for veterans, disabled veterans, and their family members.The major ones are: cash compensations for disabled members of the military, personal compensation for disability acquired at war paid by post offices, personal compensation for disability acquired at war paid through the municipalities, wage subsidies for working veterans, cash payments to military veterans living abroad, annual compensations, and benefits for civilians who have acquired disability at war. 5. Local government benefits.Many local governments provide additional benefits in the form of cash assistance to low-income groups.Some of those benefits are included in the analysis in the form of one-off assistance, heating allowances, utility and education allowance (Table 2).
In Belgrade, which has the largest social assistance program, most spending goes to one-off cash payments to pensioners with lowest incomes and subsidies for utility bills.Also, there is cash assistance to war veterans and their associations.Cash supplements are provided to parents of newborn children.There are also food packages for the poor, and benefits for students (grants, subsidies for schoolbooks).Regarding in-kind transfers, social spending on health amounted to 5.9 percent of GDP in 2016.
Health care in Serbia is provided on a near universal basis through a compulsory health insurance for all employed and self-employed individuals and their families.The state also provides insurance for pensioners, unemployed and their family members, workers from lossmaking state-owned enterprises, refugees, and certain vulnerable groups (e.g., single parents).
According to data from the Health Insurance Fund, only around 3 percent of the total population is not covered by health insurance.Mandatory health insurance covers the whole basic range of health services.The central government (Ministry of Health) or local governments (for primary protection) pay for investments in facilities and equipment from general tax revenues.
The Health Insurance Fund covers current expenditures (salaries, utility expenses, medical supplies, etc.) and is mainly financed through health contributions.
Spending on education amounts to 3 percent of GDP.Preschool (for 5 to 6-year-olds) and primary school (up to age 15) are mandatory.Most students are in free public schools.The public education system has the largest enrollment with 91 percent of the children enrolled in pre-school programs, 99 percent in primary schools, 98 percent in secondary schools, and 87 percent at higher education institutions.Preschools are financed by the local self-governments, so the parent fee depends on the extent to which daycare is subsidized by the local government.
In the capital, Belgrade, 80 percent of the full price is subsidized by the city government that, since 2015, started to subsidize daycare in private pre-school institutions as well.

Data sources
This analysis combines household survey data with data from the national accounts, Ministry of Finance public finance data, and reports of the Statistical Office of the Republic of Serbia.
The household surveys used as the basis for the analysis are the Survey on Income and Living Conditions (SILC) for 2017 (which refers to 2016 income) and Household Budget Survey (HBS) for 2016.The SILC is the most comprehensive survey used in the European Union countries to collect microdata on income, poverty, social exclusion and living conditions.The SILC has been the basis for officially monitoring household welfare, poverty, and social inclusion in Serbia since 2013.The HBS contains both income and expenditure data, but the level of details on income components and benefits received is less than the SILC, and the sample size is smaller.Therefore, this analysis is primarily based on the SILC while household consumption information in the HBS is used to estimate the impacts of indirect taxes.Appendix 4 explains how value added tax and excises are estimated from the HBS and imputed for each household in the SILC data.

Approach
This paper follows the Commitment to Equity (CEQ) approach developed by Lustig ( 2018) to analyze the distributional impacts of fiscal interventions.For each household, we define the following income concepts and calculate them for each household in the survey data: 1) Market income includes gross income from labor (wages and income from selfemployment, both from informal7 and formal employment), capital income, income from private transfers, self-production consumption as well as imputed rent a. Labor income includes: income from dependent employment, different types of self-employment (entrepreneurship, agriculture, professional practices, household help, selling or reselling products, private services and other business activities) b.Capital income includes: income from rent, dividends and interest c.Income from private transfers includes: net (amount received minus amount paid) financial assistance and alimony paid by the household, remittances and private pensions d.Imputed rent includes estimated value of monthly rent that the house-owners would pay if they had to rent their living space e. Self-produced consumption is the market value of goods produced for own consumption deducted from expenses incurred in the process of production of goods for own consumption f.Other income includes reported income for which no source was reported and labor income for children under 16 years of age 2) Market income plus pensions is calculated based on the Market income (1) from which the total amount of pension contributions is deducted, while pensions (old-age, disability, survivors, agriculture and abroad) are added 3) Net market income is based on Market income plus pensions (2) from which we deduct direct taxes, health and unemployment insurance contributions 4) Disposable income is calculated as the sum of the Net market income (3) and direct transfers 5) Consumable income is calculated as the difference between Disposable income (4) and indirect taxes and indirect subsidies (heating and utility allowance) 6) Final income is calculated as the sum of Consumable income (5) and in-kind health and education benefits Comparing the income distribution across these different income concepts helps to evaluate the impacts of the fiscal system and of specific fiscal intervention on inequality and poverty.Market income in this analysis includes imputed rent value for owner-occupied housing to enable cross-country comparison, following the Commitment to Equity method in the Guidebook.The poverty and inequality estimates differ slightly from official estimates based on the SILC, which do not include imputed rent in the income aggregate.Appendix 2 includes a robustness analysis for when imputed rent is excluded, where the direction of the results remains valid even though the levels can differ.
The treatment of pensions and pension contributions is often not straightforward.There is no clear consensus: on the one hand, contributory pensions can be treated as personal savings or deferred income; on the other hand, pensions can be treated as government transfers.Pensions are contributory in Serbia, though the Pension Insurance Fund also receives some transfers from general tax revenues to make up for shortfalls in revenues.Contributions still form the main source of pension financing and usually account for approximately two thirds of the Pension Fund expenditures.Therefore, we present the main results based on treating pensions as deferred income and contributions treated as savings while also showing the effect on inequality when pensions are treated as transfers.8

Assumptions
This methodology evaluates the first-order effects, keeping everything else constant, and does not take into account behavioral responses to fiscal policies.It assumes that the incidence of direct taxes and contributions is borne by the income earner, and the burden of indirect taxes borne by consumers.In addition, this type of standard fiscal incidence analysis does not capture the long-term effects of excise duties on products with negative health consequences such as tobacco and alcohol.
PIT and SSC are simulated based on reported income data, using tax legislation parameters described in Appendix 3. The results depend on the accuracy of reported income data, in particular the extent of informality, and reflect the de jure impact of tax policy based on statutory rates.
Indirect taxes (VAT and excise duties) were estimated by applying statutory rates to the detailed consumption data in the HBS.For excise duties, we apply statutory rates to the reported consumption of alcohol, tobacco, fuel and coffee, to estimate the burden of these excises across households.Following the survey-to-survey imputation technique (see Appendix 4), we assign indirect tax burdens to households in the SILC.For both direct and indirect taxation, we employ statutory rates.The paper's main results will show the incidence of indirect taxes with respect to current income, based on household income as the main measure of welfare, as used by Serbia's official poverty and inequality indicators.Appendix 1 explores an alternative question about the incidence of indirect taxes over the life cycle with respect to household consumption as a proxy for permanent income, using adjustments by the income-consumption ratio.The estimated value of indirect taxes that can be included in the analysis reflects the extent of information available in the HBS.In most countries, total consumption of households in the household surveys (HBS) represents a fraction of household consumption reported in the National Accounts.In this application of the methodology to Serbia, the fraction of indirect taxes that can be captured in the analysis (Table 1) closely reflects the fraction of consumption that is captured in the HBS (close to 60 percent, see also Appendix 4).Due to the lack of an available input-output table in Serbia, this analysis has not been able to estimate indirect effects of indirect taxes.
On the spending side, the SILC includes detailed information on who received payments from contributory and non-contributory social protection programs.Among contributory benefits, the data includes all types of pensions (old-age, disability, survivor's, agricultural, pension from abroad, pensions from private schemes), unemployment benefit, and sickness benefit.The non-contributory benefits included in the survey are financial social assistance, child allowance, parental allowance, foster care benefits, war veteran benefits, severance payments, one-off assistance paid by local governments, heating and utility bills allowance.Wage compensation during maternity leave is not defined as such in the data, but we use the question "Did you receive any pregnancy-puerperal benefits" for this purpose.Caregiver allowance is placed in one variable, so it is not possible to distinguish what is contributory and received through the Pension Fund and what is noncontributory and received from the Ministry of Labour, Employment, Veteran and Social Affairs.We used the reported values from the SILC database for all the benefits.9 Assigning the value of public spending on education follows the government cost approach.Since the education system is mostly public, we calculate per beneficiary input costs by level of education as total government spending divided by the number of pupils in each level of education.This per pupil expenditure is then assigned as education benefit to each student.
We use the number of students from the SILC data for tertiary education since it includes post graduate students.For the other levels of education, the number of students from the SILC is almost equal to the numbers from administrative data.For health expenditures, since access to services is near universal, we follow the cost of insurance approach and assign per capita expenditures (obtained by dividing total expenditure by the total population in Serbia) for each person in the dataset.10

Impacts on inequality
Government intervention through taxes, social benefits, and in-kind transfers reduces inequality by almost 10 percent, or 0.03 Gini points.Without fiscal intervention, the Gini coefficient of market income inequality (when pensions are treated as deferred income) in Serbia stands at 0.385.After government intervention in the form of direct and indirect taxes, social contributions, social benefits and in-kind transfers (healthcare and education), inequality falls to 0.351 (Figure 2).The largest contribution to the reduction of inequality comes from in-kind transfers (in health and education), which is reflected in the fall of the Gini from 0.385 for consumable income to 0.351 for final income.The effects of direct taxes (personal income tax and social security contributions) and social transfers (contributory and non-contributory benefits) are slightly smaller, bringing the Gini down from 0.385 for market income to 0.361 for disposable income.Indirect taxes (VAT and excise duties) tend to increase inequality with respect to current income, which is reflected in the increase of the Gini from disposable to consumable income.Over the life cycle, the incidence of indirect taxes with respect to household consumption as a proxy for permanent income is much less regressive, as discussed in Appendix 1.

Figure 2. Serbia: Change in inequality (Gini coefficient) from market to final income
Source: Own estimations using HBS and SILC data Note: The poverty and inequality estimates differ slightly from official estimates based on the SILC because imputed rent is included in the income aggregate, and the direction of the results is robust to whether imputed rent is included.
The reduction of inequality in Serbia due to fiscal intervention is smaller than that in several countries in Central and Eastern Europe, some neighboring countries, and Latin American countries of similar income level where the CEQ analyses have been conducted.When pensions are treated as deferred income, the scale of inequality-reducing redistribution through the tax and transfer system in Serbia of 0.034 Gini points is lower than in Albania (0.052), Montenegro (0.072), Romania (0.071), and Croatia (0.102). 11It is also lower than in Armenia, Poland, Belarus, the Russian Federation, Ukraine, and several Latin American countries with similar income level as Serbia.Strict comparability across countries can be difficult due to the use of income versus consumption as the main welfare aggregate to represent disposable income.
As previously mentioned, the main analysis in Serbia treats pensions as deferred income since Serbia has a pay-as-you-go pension system and the entitlement to pensions is linked to payment of contributions during employment.Figure 2  Direct taxes and social benefits in Serbia are less equalizing than in a few Central and Eastern Europe (CEE) countries where similar analysis has been conducted.Based on the CEQ analysis for other countries 12 , and disaggregating effects by policy instruments, the equalizing effects of direct taxes and social benefits in Serbia are somewhat stronger than in non-European developing countries, but still considerably smaller than in those CEE economies.The marginal contribution of all direct transfers in Serbia is to reduce the Gini (when pensions are treated as deferred income) by 0.0139.In Russia, Romania, and Croatia, they reduce inequality by 0.022-0.024points.Direct transfers in Armenia and Albania also have a higher capacity to reduce inequality, with the marginal contribution of 0.019 and 0.17, respectively.At the same time, the impact of indirect taxes in Serbia on inequality is close to the average of the CEE countries and non-EU developing countries with available CEQ results.
The equalizing effects of in-kind transfers in Serbia is comparable with the effects in those CEE countries.The effects of health and education spending in Serbia are considerably stronger than in non-European developing countries which may be the result of its free and universal primary and secondary schooling and almost universal healthcare scheme.

Impacts on poverty
Government intervention through taxes and benefits is associated with an increase in poverty in Serbia, mostly due to indirect taxes.Using household income and standardized international poverty lines of $1.9, $3.2, and $5.5/day (2011 PPP), the results suggest that the poverty rate decreases after direct taxes and social benefits are taken into account but increases upon the imposition of indirect taxes.In other words, social transfers cannot fully offset the burden of taxes on poverty.Table 3 shows that the share of the population whose consumable income is below a poverty line is higher than that for market income, except for the extreme poverty line of $1.9/day.The same trend applies if poverty is anchored on the national relative poverty line at 60 percent of median equivalized income.
Similarly, poverty gap and poverty severity are reduced when direct taxes and social benefits are introduced, but the effects are reversed after considering indirect taxes.This finding is robust across different absolute poverty lines and the relative poverty line.Since households do not observe the monetary value of health and education transfers, these items are not included in the computation of the poverty headcount, following standard practice.The finding that fiscal policy is estimated to increase poverty in Serbia is similar to that in most available analyses in Europe and Central Asia.Only in some countries in the CEQ sample, such as Russia and several countries in Latin America, poverty is reduced or remains mostly unchanged after accounting for direct and indirect taxes and transfers. 13Poverty, measured as per-capita income below $5.5/day in 2011 PPP, is estimated to increase by about 7 percentage points in Serbia, from 15.3 percent in market income (including pensions as deferred income) to 22.4 percent in consumable income.In Armenia, poverty increases by 6.6 percentage points while in Romania 1.5.Other studies using slightly different international poverty lines also indicate poverty increases after fiscal interventions in Albania, Montenegro, Croatia, and Poland.

V. Progressivity and marginal effects of individual taxes and transfers
This section discusses how each component of the tax and benefit system contributes to the system's overall impacts on poverty and income inequality.The influence of individual fiscal interventions may be different from the influence of the overall fiscal system.Both the size and progressivity of a tax or expenditure item determine its marginal effects on inequality and poverty measures.For example, a highly progressive but small government transfer may have a small marginal effect on inequality and, therefore, small redistributive impacts.We first show how taxes and transfers affect household income across the distribution to illustrate their distributional impact.We then analyze the marginal effect of each individual tax and benefit component on the Gini coefficient and the poverty headcount.
Figure 4 shows different tax and transfer categories as a share of households' pre-fiscal income (market income plus pensions) across income deciles.The shares for the bottom first decile are large because their pre-fiscal incomes (denominators) in this decile are small.Direct transfers and in-kind education and health benefits are progressive: they make up a larger share of household income among the poor than among the rich.Direct transfers represent a sizable share of household income mainly for the first income decile, reflecting the good targeting, especially of the financial social assistance program.Direct taxes and contributions represent a smaller share of income in the bottom deciles.They are progressive, with their share in the income of the top 10 percent is five times that for the bottom 10 percent.With respect to current income, indirect taxes are regressive, accounting for larger shares of income among the lower income deciles.
The poorest households are net beneficiaries of the fiscal system.Households in the first decile of the income distribution receive a net benefit from the combined fiscal system on a cash basis and including in-kind transfers.For this decile, the positive effects of transfers outweigh the large negative effect of indirect taxes.The net cash position of all the other income deciles is negative.This means that the majority of individuals were net payers to the budget in 2016 since the taxes they paid exceeded the cash benefits received.4. The extent to which individual tax or expenditure items are progressive is shown by their concentration coefficients or Kakwani index.To assess the extent to which a particular fiscal intervention is equalizing, we compare inequality with and without this tax or transfer.This difference is called the marginal effect, that is, the change in market income due to adding or subtracting only the given benefit or tax from it.
• Concentration coefficient: an index that ranges from -1 to 1, derived from the concentration curve that relates the cumulative percentage of households (from poor to rich ranked by market income) on the horizontal axis and the cumulative percentage of tax (transfer) paid (received) by each centile.The coefficient of concentration is used jointly with the pre-fiscal income Gini to construct the Kakwani index.
• Kakwani index: A Kakwani index for taxes will be positive (negative) if a tax is globally progressive (regressive).A Kakwani index for transfers is positive if a transfer is progressive in relative terms.For taxes, the index is defined as the difference between the concentration coefficient of the tax and the Gini for pre-fiscal (market) income.For social transfers, it is defined as the difference between the Gini for market income and the concentration coefficient of the transfer.• Marginal contribution: The marginal contributions of each tax and transfer component to reducing the Gini and the poverty headcount, respectively, show the redistributive effect and poverty-reducing effect.It is calculated as the difference between the Gini coefficient or the poverty rate with and without particular tax or transfer.In case of a positive value, there is a redistributive effect of a tax or a transfer and a decline in the Gini coefficient.A positive marginal contribution to poverty reduction captures a poverty-reducing effect.It is useful to calculate the marginal contribution since a progressive tax is not necessarily equalizing, depending on certain characteristics of the fiscal system, and the marginal contribution can show impacts on reducing poverty.

Direct and indirect taxes
Table 4 shows that direct taxes and social contributions in Serbia have a small contribution to reducing inequality.They are slightly progressive, as shown by the positive Kakwani coefficient.They are also inequality-reducing.However, the marginal contributions of direct taxes to redistributive effect of 0.005 is smaller than that observed in other countries, for example, 0.007 in Albania (2015) and 0.03 in Croatia (2014). 15The flat personal income tax with a tax allowance gives little progressivity.Since taxes and contributions take away from market income, they are poverty increasing, as shown by the negative marginal contribution to poverty reduction (poverty is measured using the US$5.5/day in 2011 PPP poverty line).
On the other hand, with respect to current income, indirect taxes are regressive (negative Kakwani index).They contribute to an increase in both poverty and inequality.Both VAT and excise taxes are regressive.Excises have a smaller negative impact on poverty and inequality than VAT since they are smaller in size.Appendix 1 analyzes the incidence of indirect taxes with respect to consumption, i.e., looking at welfare over the life cycle.Indirect taxes taken together in Serbia are relatively unequalizing with their marginal contribution to reducing inequality of -0.027 Gini points.While the effect is smaller than that for Croatia, it is more unequalizing than many Latin American and developing countries.16

Social spending
All contributory and non-contributory benefits are progressive.They contribute to reducing both poverty and inequality, as shown in the positive marginal contributions of these items (Table 4, which shows the results for the main social benefits as captured in the data).Among social benefits, child allowance and financial social assistance have the biggest inequalityreducing and poverty-reducing effects.These programs are well targeted to the poor.Almost 93 percent of the financial social assistance and 90 percent of the child allowance is concentrated at the bottom 40 percent of the distribution of market income (including pensions as deferred income).Half of all the spending on parental allowance goes to bottom 40 percent of the population in terms of market income (Figure 5). 17However, the marginal effects of social benefits on poverty and inequality-reducing are constrained by the relatively small coverage of the most targeted transfers.The marginal contributions of all direct transfers (excluding contributory pensions) are to reduce the Gini by 0.0139 points and the poverty headcount by 0.0216.
It should be noted that the different benefit programs may have different intentions.Not all programs necessarily have a redistributive or poverty-reducing objective, for example, sick leave and wage compensation during maternity leave.Almost 40 percent of the wage compensation during maternity leave goes to top 40 percent of the market income distribution.

Figure 5. Concentration of social benefits (by market income plus pensions quintiles)
Source: Own estimations using HBS and SILC data Education and health in-kind benefits have inequality-reducing effects.As noted earlier, we assume that the benefit received by the individual is equal to the amount spent per capita.Figure 6 shows that overall education spending is progressive, with primary education being the most redistributive (Table 4).The concentration coefficient for all levels of education except tertiary is negative, suggesting that a larger share of expenditures goes to the lower deciles.However, these results do not capture heterogeneity in the quality of services.For example, the PISA 2018 results 18 suggest that in reading, socio-economically advantaged students outperformed those from more disadvantaged background by 73 score points.Though this is still smaller than the average difference between the two groups across OECD countries (89 score points), the performance gap in Serbia had increased since 2009 when it was 66 points (87 points on average across OECD countries).Health spending is equally distributed across the income quintiles, given the way health benefits are assigned, but is progressive in relative terms since it makes up a larger share of income at the bottom of the distribution. 17This result is based on quintiles of market income plus pensions per capita that excludes pregnancy and maternity allowance, which may differ from other studies looking at targeting based on quintiles of disposable income.

Figure 6. Concentration of education and health (by market income plus pensions quintiles)
Source: Own estimations using HBS and SILC data

VI. Concluding remarks
The analyses presented in this paper show that Serbia's fiscal system as a whole reduces inequality, though it has a poverty-increasing effect.The system is redistributive, reducing the Gini coefficient of income by 0.03 points, from 0.38 at market income to 0.35 at final income, when taxes, transfers, and in-kind benefits in education and health are taken into account.The inequality-reducing impacts of the fiscal system in Serbia is somewhat smaller than what is observed in other countries in Central and Eastern Europe and Latin America where similar analysis has been applied.However, social transfers cannot fully offset the burden of taxes on poverty, resulting in a net poverty-increasing effect of these measures combined.On the expenditure side, social transfers in Serbia are well-targeted to the poor and reduce inequality by 0.014 Gini points.Their poverty and inequality-reducing impacts are constrained by the relatively small coverage of the most targeted transfers, i.e., the poverty-targeted financial social assistance and child allowance.Therefore, expanding coverage of these targeted programs can increase the poverty reduction and redistributive impacts of social transfers.
On the revenue side, direct taxes in Serbia are overall slightly progressive.Their impact on lowering inequality is much smaller than that of direct transfers.The flat personal income tax gives little progressivity while the regressivity of VAT with respect to current income increases inequality.Poverty, measured as per-capita income below $5.5/day in 2011 PPP, is estimated to increase by about 7 percentage points, from 15.3 percent in market income (including pensions as deferred income) to 22.4 percent in consumable income after reflecting direct and indirect taxes and transfers.The same trend applies if poverty is anchored on the national relative poverty line at 60 percent of median equivalized income.The poorest households are net beneficiaries of the fiscal system.
This analysis of fiscal incidence in Serbia provides a basis for future analyses to assess the potential impacts of changes to a tax or a benefit in the system, or a combination of simultaneous policies.Simulating such changes would demonstrate further use of this analysis for policy making.Understanding poverty and distributional impacts helps identify potential winners and losers from fiscal reforms, inform how to protect the poor and vulnerable in case of adverse short-term impacts, and help build support and momentum for reforms.

APPENDIX 1 Incidence of indirect taxes from a life-cycle perspective: VAT corrections
The main results presented in the paper are based on household income as the main measure of welfare (as used for official welfare statistics in Serbia), thus showing the distributional impact of taxes with respect to current household income.This is considered relevant for policy making purposes in the short to medium term.An alternative question, explored further in this Appendix, is to understand the incidence of taxes over the life cycle with respect to household consumption as a proxy for permanent income.Assessing the incidence of indirect taxes from a life-cycle perspective is possible through adjustments by the income/consumption ratio, which is the VAT correction as per the Commitment to Equity guidebook.
For the permanent income concept of welfare, standard life-cycle theory argues that consumption is a good proxy.With respect to consumption, taxes at a fixed rate on consumption like VAT should in theory be a neutral tax over the life cycle.The other motivation for adjustments comes from negative consumable income observations.In cases when the reported consumption is much higher than reported income, either due to underreporting of income, dissaving, or borrowing, some of the households with consumption much higher than reported income end up with negative consumable income.In the case of SILC 2017 data for Serbia, the number of those with negative consumable income is very small, only 0.2 percent.
Adjustments take the form of scaling VAT values by the income/consumption ratio.After computing the amount of indirect taxes paid as in the main results scenario (using the imputation explained in Appendix 4), we multiply this by the income/consumption ratio to arrive at the adjusted indirect taxes paid in this alternative scenario.
In the main results, as showed in Figure 2, consumption taxes (VAT and excise duties) tend to increase inequality, which is reflected in the increase of the Gini from 0.361 for disposable income to 0.385 for consumable income.After the implementation of VAT corrections, reflecting the lifetime perspective, the Gini coefficient for consumable income is unaffected.The effect of indirect taxes on poverty is smaller with VAT corrections (Table A1.3) compared to the baseline scenario (Table 3).At the $1.90 PPP/day poverty line, the poverty rate after indirect taxes increases to 3.5 percent while in the baseline scenario to 4.9 percent.The results based on other poverty lines ($3.2 PPP/day, $5.5 PPP/day or relative poverty line) show that indirect taxes increase poverty in the model with VAT corrections but by less compared to the scenario without corrections.After the VAT adjustments, the exact net cash position for the first six income deciles looks better than in the main scenario, due to the mechanical adjustment applied.The deciles are intact as based on market income plus pensions.Poorer households in Serbia tend to be of larger size and have more children.Households in the poorest income decile are still net beneficiaries of the fiscal system, with a positive net cash position while it remains negative for the rest of the deciles.After the implementation of VAT corrections, indirect taxes account for only slightly larger shares of income among the lower deciles.They are less regressive with respect to consumption, reflecting the lifetime perspective, than with respect to current income.A1.4 shows that after the adjustments indirect taxes are still regressive, given the negative Kakwani index, but much less than in the case without adjustments: for VAT, the index is -0.01, compared to -0.119 in the main results.Excises are still more regressive than VAT, as in the case with no corrections, and the Kakwani index is -0.05, compared to -.017.Indirect taxes after VAT corrections continue to contribute to an increase in inequality and poverty but to a smaller degree than in the case with no adjustments.

APPENDIX 2 Robustness checks on imputed rent
The results in the main paper included imputed rent in the market income, according to the CEQ methodology guidelines.Imputed rent is also known as "income from owner-occupied housing." 19The SILC survey asks families who own their homes to report the amount they think they would be paying in rent for the same dwelling, or for how much they would rent it out.This Appendix analyzes the imputed rent values in the SILC data, which appear reasonable when compared to actual rental prices, and presents the results in an alternative scenario of excluding imputed rent from market income for robustness checks.
According to the SILC data, only about 4 percent of the households (and 4 percent of all individuals) pay rent for their apartments/houses.The remaining households are either homeowners (in about 80 percent of the cases) or are living in their homes free-of-charge (16 percent).Imputed rent in the SILC takes into account both the returns to home ownership, in the sense that the main residence is an asset, as well as the in-kind transfers accruing to those who live free-of-charge.
Imputed rent is registered based on the question "To rent an apartment you live in or similar to yours, what do you think you would have to pay on the monthly level?"The respondents had an option to respond in RSD, EUR, CH or USD.All the answers were recoded to RSD using the average exchange rate for 2016 and aggregated to yearly levels.Given the number of households who do not pay rent and the relatively sizeable amount of rent, imputed rent represents a significant portion of both market (about 15 percent) and disposable income (about 16 percent).
Descriptive statistics indicate that the non-zero average imputed monthly rent is 15.060 RSD (about 122 EUR), which corresponds to the average renting prices in Serbia.The range is also plausible, ranging from 20 RSD (people who do not think that it is possible to rent their home at all) to about 84000 RSD (about 680€).Detailed analysis suggests that the rent in the SILC data is higher in Belgrade than in other regions and higher in urban than in rural areas, which is also in line with expectations.
Without imputed rent, the broad results of impacts on inequality remain similar.The Gini coefficient for all income concepts is larger than in the main results scenario with imputed rent, but the direction of impact remains as shown in Figure 2. The result that the fiscal system increases poverty is robust.In the scenario without imputed rent, all poverty indicators are higher than in the main scenario with imputed rents.But the main results of impacts on poverty are similar (Table A2.3).The findings regarding the incidence of taxes and social transfers as assessed against their share in the market income (plus pensions) remain robust.The exact shares, however, differ when imputed rent is not part of the market income since market income is lower than in the main results with imputed rents (Figure A2.6).As shown in Table A2.4, the Kakwani index for transfers is higher when imputed rent is not included in the income aggregate, so transfers look more progressive compared to when imputed rent is included.This is due to, as displayed in Figure A2.6, an increase in the shares of social transfers relative to market income without imputed rent.The marginal contribution of all direct transfers (excluding contributory pensions) to the redistributive effect is somewhat increased whereas the marginal contribution to a poverty headcount is almost unchanged.
The Kakwani index for all direct taxes remains positive but is somewhat lower than in the main results, meaning that its progressive feature is less when imputed rent is not part of the market income.The share of direct taxes relative to market income increases much more for the first income decile compared to the rest of the income distribution when market income concept lacks imputed rent.
The marginal contribution of all direct and indirect taxes to poverty headcount is higher than in the main results, meaning that without imputed rent, their contribution to increase in poverty is larger.The marginal contribution of all indirect taxes to redistributive effect is also higher after the elimination of the imputed rent from market income, whereas for direct taxes, it remains almost unchanged.The concentration of social benefits is almost the same as in the case with imputed rent.The bottom 40 percent of the income distribution have the same share in public spending on child allowance and financial social assistance.Only the share of spending on unemployment benefit going to the bottom 20 percent is higher by 10 percentage points when imputed rent is not included in the market income.
4. Income from capital Taxpayer Individual earning income from interests, dividends, distribution of profit, return to investment units are obliged to pay tax on capital income.

Tax base
The tax base is gross amount of income.

Tax rate
Tax rate is flat and amounts to 15%.

Tax reliefs
Interest income based on investments in central or local government bonds and securities of the central bank, as well the interest income based on bank deposits in dinars, are exempted from taxation.

Taxpayer
Individual earning income from rent of immovable property is obliged to pay tax on rental income.

Tax base
The tax base is gross amount of income.

Tax rate
Tax rate is flat and amounts to 20% in 2018, while in 2015 it amounted to 15%.

Tax reliefs
There are no tax reliefs.

Taxpayer
Individual earning capital gains from sale of shares/stakes, other securities, immovable property or authorship and industrial property rights, are obliged to pay capital gains tax.

Tax base
The tax base is the positive difference between the selling price and acquisition price.The selling price is a price stated in the selling contract or the market price assessed by the Tax Administration, if contractual price is lower than the market price.

Tax rate
Tax rate is flat and amounts to 15%.

Tax reliefs
Capital gains generated based on transfer of property through inheritance or between the first line relatives are exempted from taxation.Also, capital gains generated by selling immovable property in which individual resides is exempted from capital gains tax, if sale proceedings are reinvested within 90 days in other immovable property in which individual will reside.

Other incomes Taxpayer
Individual earning income from free-lance contracts, rent of movable property, lotteries, life insurance, as well individuals earning income from sports, etc. are obliged to pay tax on other income.
Free-lance contract income is subject to pension and disability insurance contributions (26%).If taxpayer is not employed, she/he also needs to pay the health insurance contributions on this income (10.3%).

Tax base
The tax base is gross income.In case of lottery winnings, there is non-taxable amount on each winning up to RSD 11,165.In case of free-lance contract, the tax based is the gross income decreased by standard deductions of 20% (of gross income).

Tax rate
Tax rate is flat and amounts to 15% on income from life insurance and 20% on other incomes.

Annual income tax
In addition to this, tax residents whose total annual income exceeds stipulated threshold are obliged to pay annual income tax as well.
Following types of income are included in the annual income tax b: salary, income from selfemployment, income from royalties and authorship rights and related rights, real estate income, income from renting movable property, income from insurance of individuals, income of sportsmen and other income.
Non-taxable threshold for annual income tax purposes is three times average annual salary in Serbia in the respective year (RSD 2,201,220 in 2015).
Individual's total annual net income, exceeding this limit, may be decreased by standard deductions:

•
Personal deduction (40% of the average annual gross salary in Serbia -RSD 293,496 in 2015); and • For dependent family members (15% of average annual gross salary in Serbia -RSD 110,061 in 2015 per dependent family member).
The total amount of standard deductions cannot exceed 50% of taxable income.If two or more family members are obliged to pay annual income tax deductions for dependents may be used only by one family member.If one family member is using part of these deductions remaining (unused amount) may not be transferred to the other family member.
Individual's total net annual income, exceeding stipulated (non-taxable) threshold and decreased by the abovementioned deductions, is subject to annual income tax at following schedule: • 10% on taxable income between 3 and 6 average annual salaries in Serbia and • 15% on income above 6 average annual salaries in Serbia).
Annual income tax is to be paid by individual (i.e.individual's employer does not have any duties related employee's annual income tax).Annual income tax return is to be submitted with the Tax Administration by 15 May of the year following the year in which income is generated.

SOCIAL SECURITY CONTRIBUTIONS
Social security contributions comprise of: a) pension and disability insurance contributions, b) health insurance contributions and c) unemployment insurance contributions.
There is no compulsory private pension or health insurance in Serbia (there is only voluntary private pension and health insurance).

Pensions and disability insurance
The pension system in Serbia is based on two pillars: a) PAYG mandatory pension and disability insurance scheme, and b) supplementary, voluntary private scheme.

Payers
All employees, self-employees, earners of free-lance income or authorship, industrial and other property rights are obliged to pay pension insurance contributions, under PAYG scheme.

Base
The base for calculation is employee's gross wage (net wage + wage tax + social security contributions on behalf of employee There is also the maximum annual base for calculation of social security contributions, equal to five times average annual gross salary in Serbia, in year in which contributions are paid.In 2016, the maximum annual base amounted to RSD 3,808,4400.Therefore, if individual's annual gross wage and other types of income, which are subject to social contributions in total exceed this annual threshold, that individual may claim refund of social security contributions paid on income exceeding the annual threshold.
Self-employed individuals pay social insurance contributions by themselves, on their personal salary as defined by the Personal Income Tax Law.
The same rule on the base for calculation of social security contributions apply also to health insurance and unemployment insurance contributions.

Rates
Since August 2014, pension and disability insurance contributions are paid at the total rate at 26% (14% -on behalf of employee and 12% -on behalf of employer).
Before August 2014, the total rate of contributions for pension and disability insurance amounted to 24%, split half-half between employee and employer.

Health insurance Payers
All employees and self-employees are obliged to pay health insurance contributions on their income.Also, earners of free-lance income or authorship, industrial and other property rights income are obliged to pay health contributions on their income, if they are not employed.

Base
Health insurance contributions are payable on the same base as pension and disability insurance contributions.

Rates
Since August 2014 health insurance contributions are payable at the total rate of 10.3% (5.15% on behalf of employer and 5.15% on behalf of employee).
Before August 2014, the total rate of contributions for health insurance amounted to 12.3%, split half-half between employee and employer.

Insurance of retired and unemployed individuals
Retired people participate in health insurance scheme based on their retirement status.The Pension Fund is obliged to pay health insurance contributions at the statutory rates (for health insurance of retirees) to the Health Insurance Fund.Contributions are to be calculated on the amount of pensions.
Unemployed individuals (registered with the National Employment Service) are covered with health insurance, based on their unemployment status.The National Employment Service is obliged to pay health insurance contributions, at the statutory rates, on the amount of unemployment benefit, paid to registered unemployed individuals.

Unemployment insurance Payers
All employees are obliged to pay health insurance contributions on their income.For the purpose of calculation of quantitative amounts of consumption of cigarettes, the following weighted average prices per pack can be used: in 2015 -RSD 197.5, in 2016 -211.9,  in 2017 -RSD 224.60 and in (Jan-July) 2018 -RSD 242.30.

Excise duties on coffee
Consumption of coffee is subject to excise duties.The tax rates are defined in the specific form (RSD per kg), presented in the Table A3.4.Considering that the roasted coffee is the most commonly type of coffee consumed in Serbia, the rates of tax on roasted coffee are presented.

Excise duties on electricity
Starting from August 1 st , 2015 excise tax on electricity is introduced.The tax rate is 7.5%.The tax base is the net price of electricity.

Excise duties on oil derivatives
Excise duties are to be paid on consumption of petrol, diesel, natural gas, heating oil and other derivatives.The tax rate is stated in the specific form and indexed annually (Table A3.5).Propane-butane is exempted from excise duties.The only substantial difference in the average values is in the group 4 -Housing, Water, Gas, Electricity and Other Fuels.The difference arises due to exclusion of the imputed rent from the expenditures in this group (it cannot be a subject of taxes), and results in about 500 RSD lower calculated (weighted) average for this group than the one presented in the official SORS reports.Consequently, our estimate of the total monthly expenditure -58,500 RSDis also by about 500 RSD lower than the one from the official SORS report (59,052).
There are two VAT rates: one standard rate (20%) and one reduced rate (10%).The VAT Law stipulates the finite list of goods and services which are taxed at the reduced rate, while supply of all other goods and services is taxed at the standard rate.
To estimate VAT, we first extract the expenditure on products and services which are taxed with a reduced rate from the expenditure on other items within that COICOP group.We then apply the 10 and 20 percent VAT rates to the respective product groups to calculate total VAT.
The average consumption from the HBS represents about 58 percent of the household consumption based on the statistics from the National accounts. 22The estimated value of VAT based on the HBS data amounts to about 268 billion RSD (yearly level), or 59 percent of the VAT reported in the official statistics 23 .This suggests that the discrepancy between our estimate of VAT and the one reported in the official statistics is almost entirely due to underestimation of the household consumption by HBS.
Excises are partially charged ad valorem, and partially based on the quantities of the goods.Diary data, which contain information on COICOP groups 1 (Food and non-alcoholic beverages) and 2 (Alcohol Beverages and Tobacco), beside the expenditures, also provide the quantities of the goods purchased, which are, similarly to expenditures aggregated at the household level and multiplied by 2. For the products which are not collected in the diaries, the quantities are not directly available but are calculated as the ratio between the expenditure of these products and the average prices of these products in 2015.
Ad valorem excises were calculated by multiplying the expenditure of the goods with the respective percentage rate, while the excises based on quantity were calculated by multiplying the quantities with the respective per unit rates.
The total estimated value of excises based on the HBS data is 87.4 billion RSD and they represent about 37 percent of the excises reported in the administrative data24 .Lower value of precision of excise estimate (compared to VAT) could be due to the fact that the consumption of excised goods is not socially desirable (e.g.cigarettes and alcohol), so people tend to underestimate the consumption of these goods more, compared to other goods.

Imputation procedure consists of three steps
We first use the data from HBS to estimate the share of VAT and excises in the total household income conditional on income.We use natural logarithm and add powers of natural logarithm, until we reach the specification which does not increase explanatory power of the model  =   ⁄ =  0 +  1 ln() +  2 ln() 2 +  3 ln() 3 + ⋯ (1)  =   ⁄ =  0 +  1 ln() +  2 ln() 2 +  3 ln() 3 + ⋯ (2) We drop extreme values by first dropping VAT and excise shares higher than 2 and additionally use the procedure suggested by Cameron and Trivedi (2010)  The results suggest that level and square of the log income per capita are sufficient to predict VAT and excises and that inclusion of the other powers increases the explanatory power of the model only marginally.The values of R squares are higher than the ones estimated for Croatia using the same methodology (0.2429 and 0.0632 for VAT and excises respectively).
In the second step, we use the coefficients from the specifications (2) for both VAT and excise shares and apply these coefficients to the SILC disposable income to obtain the imputed values of VAT and excise shares, which we then multiply by household specific income to obtain the total imputed values of VAT and excises for all households in the SILC database.

Figure 1 :
Figure 1: Income concepts in the CEQ fiscal incidence analysis

Figure 4 .
Figure 4. Distributional Impact of the Tax and Benefit System in 2016

Figure A1. 6 .
Figure A1.6.Distributional Impact of the Tax and Benefit System, with VAT corrections

Table 1 : Serbia: General Government Revenue, 2015
Sources: Authors' calculations based on Ministry of Finance

Table 2 . General government spending, 2016
Sources: Authors' calculations based on Ministry of Finance data

Decline in inequality (Gini coefficient) from market to final income
also shows the impact on inequality when pensions are treated as government transfers instead of deferred income.In that scenario, market income inequality in Serbia amounts to 0.482, which is somewhat lower than in Albania (0.49), Croatia (0.51), and Montenegro (0.53).The overall reduction in inequality due to fiscal intervention in Serbia is equivalent to 0.131 Gini points from market income to final income when old-age pensions are treated as transfers but again lower than in most of the countries in Figure3(0.24 in Poland, 0.22 in Croatia, 0.17 in Romania, 0.15 in Albania).

Table A2 .4. Serbia: Marginal Contributions to Reducing Inequality and Poverty, without imputed rent
). Serbian Law on Contributions for Mandatory Social Insurance stipulates minimum and maximum base for calculations of social security contributions.Minimum base for calculation of social security contributions equals 35% of average wage in the Republic of Serbia, paid out in previous quarter (published by the Statistical Bureau of the Republic of Serbia).According to the official information, minimum base for calculation of (monthly) social security contributions in Serbia, in 2016 was ranging from RSD 21,552 to RSD 22,400.Therefore, if employee's monthly gross salary is below this threshold, social security contributions are calculated on this threshold.Maximum base for calculation of social security contributions equals five times average gross wage in Serbia, according to the latest information of the Statistical Bureau of the Republic of Serbia.Further to the official information, maximum (monthly) base for calculation of social security contributions in 2016 was ranging from RSD 278,815 to RSD 353,815.Therefore, if employee's monthly gross wage exceeds this amount, social security contributions are paid only on this amount (gross salary exceeding this limit is subject to salary tax only).

Table A4 .1: Comparison of weighted average values of consumption from the data and consumption reported by HBS.
to eliminate influential observations which fall within the first or last percentile of the distribution.In all regression estimates we use population weights.The results of different specifications of models 1 and 2 are presented in the next table.