Income redistribution through taxes and social beneﬁts: the case of Slovenia and Croatia

The article analyses the redistributive effect attained by personal income tax, social security contributions and social benefits in Slovenia and Croatia. The redistributive effect is decomposed first to reveal progressivity and horizontal inequity effects, and further to show contributions of different tax and benefit instruments. Even though both countries started from the same socioeconomic background two decades ago, the current results reveal divergence that is a consequence of diverse development during this period. The results indicate that Croatia experienced significantly higher pre-fiscal income inequality and lower redistributive effect than Slovenia. Horizontal inequity effects, though, were higher in Slovenia than in Croatia. In both countries, the means-tested social benefits exerted an over-proportionate influence on vertical effects, suggesting a strong impact of the welfare state on income position of their residents, but also induced a large amount of horizontal inequity. In Slovenia, the non-means-tested benefits slightly increased income inequality.


Introduction
Social security programs provide income assistance in the form of social benefits to individuals and their families in the case of unemployment, work injury, maternity, sickness, old age, or permanent earning inability. They are financed by social security contributions, as well as by other taxes. In this article, we focus our research on tax and benefit systems consisting of social benefits, social security contributions and personal income tax (PIT).
Given their size in modern states, tax and benefit systems have a significant influence on income distribution. Generally, tax and benefit systems reduce income differences between high-income and low-income households. Income inequality reduction caused by tax and benefit systems is called the redistributive effect, and it is equal to the difference between pre-tax-and-benefit (or pre-fiscal) income inequality and post-tax-and-benefit (or post-fiscal) income inequality.
A more detailed analysis of the redistributive process reveals its complexity. Usually, one distinguishes between vertical equity and horizontal inequity, perceiving them as two opposite "forces". Vertical equity is achieved through inequality reduction between richer and poorer units (households, individuals). Horizontal inequity, on the other hand, emerges due to inequality augmentation between equally well-off units. These notions are built into the framework of decomposing the redistributive effect into two main parts: the vertical effect, representing fulfilment of the vertical equity principle, and the horizontal effect, measuring the horizontal inequity.
Due to different designs, various tax and benefit instruments have different consequences for vertical equity and horizontal inequity of the overall tax and benefit systems. The means-tested social benefits are particularly designed to help the poorest individuals or households, while the non-means-tested social benefits are dispensed irrespectively of the recipient's personal or household income. One could expect that the contribution of means-tested social benefits to vertical equity is larger than the contribution of non-means-tested social benefits, while the latter will contribute relatively more to horizontal inequity.
While social security contributions are typically proportional to their tax base, PIT systems are usually designed to create larger relative burden on higher income earners. However, social security contributions may also achieve inequality reduction due to their usual payers -employed people -receiving on average higher incomes than nonactive people that do not pay social security contributions. In the present article, we analyse redistributive effects of tax and benefit systems in Slovenia and Croatia, two neighbouring countries that share a similar socioeconomic background. Both countries achieved independency in 1991 after the breakdown of Yugoslavia, and started the transition towards market economy from the same tax and benefit system of excommon state.
In Croatia, the overall income inequality increased between 1998 and 2002 due to increased inequality of wages, in spite of better-targeted social transfers that suppressed the overall inequality increase (cf. Nestić, 2005). This was confirmed by Bićanić et al. (2010), who examined wage inequality and wage differentials for the period from 1970 to 2008. They concluded that average wage differentials by education and vocational training increased through the mid-1990s and then stabilized. The stabilized income inequality in the period 2002-2004 is also evident from the World Bank reports (cf. World Bank, 2007). On the other hand, the overall inequality of pre-tax income in Slovenia increased as well during the last two decades (Stanovnik and Verbič, 2005;Stanovnik and Čok, 2009), while the inequality of after-tax income remained fairly stable during 1991-2009 due to changes in the PIT system (Stanovnik and Verbič, 2012).
Developments in both countries during the last two decades were reflected in the level of pre-and post-fiscal income inequality. As the results of this article reveal, the inequality of pre-fiscal and post-fiscal income was higher in Croatia than in Slovenia. However, the aim of this article is neither to investigate the political and economic mechanisms that have caused an increase of inequality of pre-fiscal income during the last two decades nor to focus on the mechanisms that have created the current tax and benefit systems in both countries, but to apply recently developed methodology for the measurement of redistributive, vertical and horizontal effects, and to demonstrate their usefulness in cross-country comparisons of the role of fiscal systems in reducing income inequality.
To compare the performance of tax and benefit systems in Slovenia and Croatia, we employed the following decompositions of the redistributive effect. First, the Duclos et al. (2003) model decomposes the redistributive effect into vertical, classical horizontal inequity, and reranking effects. For certain combinations of parameters, the Duclos et al. (2003) model then becomes equivalent to the well-known Kakwani's (1984) decomposition of the redistributive effect, which contains only the vertical and reranking effects. Finally, all the effects obtained from the Kakwani's (1984) decomposition were further decomposed, following Urban (2012), to reveal contributions of individual tax and benefit instruments to the vertical, reranking and redistributive effects. To the best of our knowledge, this is the first application of the Duclos et al. (2003) and Urban's (2012) methodology in comparative analyses of tax and benefit systems. The results show significant differences between the two countries in almost every respect. Croatia had a higher level of post-fiscal income inequality than Slovenia, which was a consequence of higher inequality of pre-fiscal income, as well as of a less redistributive tax and benefit system.
The article proceeds as follows. In Section 2, we briefly provide the measurement models decomposing the redistributive effect. Section 3 delivers an overview of tax and benefit systems in Slovenia and Croatia, followed by a description of income definitions and databases used in the empirical analysis. In Section 4, we present and discuss our results based on the methodology and data sources presented in the latter two sections. The final section concludes with the main findings.

Decompositions of the Redistributive Effect
The starting point of our analysis is the redistributive effect, i.e. the change of income inequality induced by the fiscal system, where post-fiscal incomes, , are equal to prefiscal incomes, N X , minus taxes, T , plus benefits: (1) In measurement terms, we set that () () I XI N Δ =−, where represents the redistributive effect, while ε ≠ . Finally, the Atkinson-Gini inequality index is calculated as follows: where X μ is the mean pre-fiscal income. Post-fiscal income inequality, () I N , is obtained analogously, using the quantiles of post-fiscal income distribution.
The Duclos et al. (2003) model decomposes the redistributive effect as follows: .
The vertical effect, , represents the potential redistributive effect or the reduction of inequality that would be achieved by the counterfactual, horizontally equitable system. The discrepancy between potential and actual redistributive effect is divided into the classical horizontal inequity effect, , which measure two different m inequity ect (C) measures horizontal inequity emerging from violation of the "classical horizontal equity principle", which says that equals should be treated equally. The latter effect (R) evaluates horizontal inequity arising from the infringement of the "no-reranking principle", requiring that fiscal process does not change ranks of income units in transition from pre-to post-fiscal income. On the other hand, When ε = , utilities are identical to incomes: . Therefore, ((|) , 0 ) ( UNq p Nq = across all p and (| ) Nq p, and it follows that The consequence for the Duclos et al. (2003) model is that the classical horizontal inequity effect collapses to zero, and the decomposition (4) can be rewritten as: As me e racting taxes from pre-fiscal incom TB = −+. Urban (2012) 2 Duclos and Araar (2006) showed that the S-Gini index of inequality is equal to , i.e. to deviations of incomes from the mean, weighted by the ν are called "single parameter Gini" indices or (more conveniently) the "S-Gini" indices. 3 Originally, all these indices were defined for 2 ν = .
decomposes marginal changes of K V ν , APK R ν and ν Δ , to show how each tax and benefit instrument contributes to these effects. This model the Lerman and Yitzhaki's (1985) decomposition overall income inequality into contributions of income sources, and is applied to the terms from the Kakwani's (1984) decomposition of the redistributive effect. It raises the following question: if actual values of each tax and benefit were changed independently from each other by some small factor e , what would be the shares of this tax or benefit in total changes of vertical, reranking and redistributive effects? These shares are given on the right hand sides of the following three equations, for changes of is rooted in of K V ν , APK R ν and ν Δ , respectively: , g than e of redistributive effect will be g the share of reranking e butiv effect obtained for the "overall ind or the purpose of the present article, pre-fiscal income was defined as the sum of ubject to PIT) and ensions. Income from own-use production and transfers of money and goods from tri ces". F market incomes from different sources (both subject and non-s p other persons were not included. Taxes were assembled into three groups: employers' social security contributions, employees' social security contributions and the PIT. Benefits were divided into means-tested and non-means-tested social benefits. Public pensions were part of pre-fiscal income and were not included into benefits. Post-fiscal income was obtained according to expression (1). Figure 1 presents pre-fiscal and postfiscal sample incomes in Slovenia (in terms of mean pre-fiscal income). 4 Table 1 further provides a very brief overview of the tax and benefit system in Slovenia and Croatia. According to the PIT system in Croatia, capital incomes (capital gains, dividends, subject to taxation. Income from employment, domestic pensions nd and interest) were not a income from self-employment were subject to progressive tax schedule with four tax brackets, with marginal tax rates of 15%, 25%, 35% and 45% in 2007. 5 Tax schedule was then applied to the tax base, obtained as the difference between the overall income and the amount of personal allowance and other tax allowances. For some income sources (e.g. income from contractual work and rental income), different flat rates were applied, ranging from 15% to 45%. The surtax is obtained as a percentage of PIT, with the rates ranging from 0% to 18%, set by the local municipalities. Unit of taxation was an individual taxpayer.
Employers' and employees' social security contributions were equal to 17.2% and 4 Graphic representation of sample incomes in Croatia is omitted due to similarity to the Slovenian sample; the distribution of sample incomes in Croatia is somewhat less dispersed, though. 5 Submission of a tax file can be obligatory (in certain pre-defined conditions and for certain groups of taxpayers) or voluntary (when a taxpayer wants to use some allowances and deductions). Since 2010, there are three tax brackets with marginal tax rates of 12%, 25% and 40%.
20% of gross wage, respectively. Combined social security contribution rate for the self-employed was equal to 35%. 6 Pension insurance system contained two mandatory ns), business income, income from Slovenia Croatia pillars: "solidarity" pension insurance and "capitalized accounts". Majority of the insured were participating in both pillars, paying 15% of the gross wage to the former and 5% to the latter. In our analysis below, we treated only the social security contributions to the "solidarity" pillar as taxes, while the outlays to "capitalized accounts" were equivalent to personal savings. Slovenia had a similar PIT system, though with several differences. PIT in Slovenia was set at the individual level (as in Croatia), and levied on six categories of income: income from employment (including pensio agriculture and forestry, income from rents and royalties, income from capital, and other income accruing to persons liable to tax. Current personal income tax code was adopted in 2007, when a 20% flat tax rate for income from capital income was introduced. Other, non-capital income was subject to progressive tax schedule with three tax brackets and marginal tax rates of 16%, 27% and 41%. Rate of social security contributions (without any ceiling) was set at 22.1% for employees and 16.1% for employers. The self-employed paid an overall rate of 38.2% by themselves. We chose year 2007, as it is the most recent year for which comparable data for both countries were available. Since then, tax and benefit system in neither country has changed substantially. Unit of analysis was a household. Incomes were deflated by an equivalence factor, obtained by using the "modified OECD scale",   Duclos et al. (2003) and Kakwani (1984) models.   Duclos and Lambert's (2000) model. Thus, the first and the last scenarios are the "extreme" ones -the former lacking the classical h ontal i y term, and in the latter the reranking term being absent. On the other hand, scenarios 2 and 3 contain both the reranking and classical horizontal inequity effects.
Comparing the results from Table 2 for Slovenia and Croatia across the different scenarios, several important conclusions can be drawn. First, Croatia ha oriz nequit d higher prefiscal income inequality than Slovenia, as reflected in ˆ( ) expressed as percentage of ˆ( ) x i I X , suggests that the rela istributive effect countries was of similar magnitude, but slightly higher in Slovenia. Consequently, postfiscal income inequality wa higher in Croatia than in Slovenia by about 10%, as indicated by ˆ( ) n i tive red in two s also I N . Second, horizontal inequity was significantly larger in Slovenia, as can be seen by inspecting the combined value of ˆ( , ) C ε ν and ˆ( , ) R ε ν , which ranges from 15.5% to 19.3% of the redistributive effect in Slovenia, and from 9.8% to 13.1% of the redistributive effect in Croatia. Figure 2 illustrates the same results for a larger range of values ε. The columns of the histograms are divided into two parts: the bottom part represents the share of the redistributive effect in pre-fiscal income inequality, i.e. the ratio of ˆ( , ) ε ν Δ to ˆ( )  Slovenia achieved a larger relative redistributive effect than Croatia, but the difference decreased with ε, so that for the share of ˆ( , ) was higher for Croatia. The differences between the two countries were much more noticeable in the case of horizontal inequity; it can be easily observed that Slovenia had significantly larger horizontal inequity than Croatia. Taking the redistributive effect and horizontal inequity together, we also see that the vertical effect was larger in Slovenia.
Next, Figure 3 looks more closely into the relationships between different effects, obtained by the Duclos et al. (2003) decomposition. Again, we use the property that the vertical effect is a sum of the redistributive effect, classical horizontal inequity and reranking effects, trying to reveal the share of each in the vertical effect.
The share of overall horizontal inequity in the vertical effect, measured as ˆˆ  Thus, one of the main conclusions from the above comparison of two countries is that the Slovenian tax and benefit system created much larger horizontal inequity than the Croatian tax and benefit system, while the redistributive effect was only slightly higher in Slovenia. How could this be explained, i.e. what tax and benefit instruments were responsible for these results? The answers are provided by applying the Urban's (2012) decomposition of vertical, reranking and redistributive effects, with the results presented in Table 3 and Figure 4. Notes: sscer -employers' social security contributions, sscee -employee's social security contributions, pit -personal income tax, nmt -non-means-tested benefits, and mt -means-tested benefits; mcVK, mcRAPK, and mcRE -decompositions of marginal changes of the Kakwani vertical, Atkinson-Plotnick-Kakwani and redistributive effects, according to expressions (7), (8) and (9), respectively; % pfinc = percentage of pre-fiscal income.
Source: Own calculations.
irst observe the relative sh shown in the bottom Let us f ares of taxes and benefits in pre-fiscal income, row of Table 3. In both countries, the overall size of taxes was much higher than the size of benefits. Nam ly, in Croatia (Slovenia)  Social security contributions had the strongest role in creation of reranking. Taken together, they decreased the potential redistributive effect by 29.9 (16.1) percentage oints in Slovenia (Croatia p these taxes were paid in cr ted much more reranking (8.8% of mcRE) than Croatian PIT (2.2% of mcRE).
On the other hand, Croatian benefits induced a lot more reranking (5.8% of mcRE, combined) than benefits in Slovenia (2.3% of mcRE). Another large difference between two countries was the role of non-means-tested benefits. In Croatia they made a relatively large contribution to the redistributive effect (5.8% of mcRE), while in Slovenia they increased inequality (-2.4% of mcRE).  Notes: sscee -employee's social security contributions, sscer -employers' social security contributions, pit -personal income tax, nmt -non-means-tested benefits, and mt -means-tested benefits.
Source: Own calculations.

Concluding Remarks
Even though Croatia and Slovenia share a similar background, they have developed in different ways during the last two decades. A major consequence of this divergent development is higher pre-fiscal income inequality in Croatia. Both countries also differ in the characteristics of their tax and benefit systems. In this article, we focused on the analysis of the redistributive effects of tax and benefit systems consisting of employers' and employee's social security contributions, personal income tax, and a wide range of means-tested and non-means-tested social benefits.
ition efit system created a much larger ertical effect than the Croatian tax and benefit system. However, the former system al effect, wi e redistributive effect was only slightly higher in lovenia. In both countries, the ov rall size of taxes was much higher than the size of benefits. Namely, in Croatia (Slovenia) taxes made 26% (33.3%) of pre-fiscal income, against the share of benefits equal to 2.5% (2.7%) of pre-fiscal income. The reason of this disproportion is clear: social security contributions were used to finance the outlays of the pension and health systems, but the respective benefits obtained by households from these systems were not covered by the analysis.
Moreover, the contributions of taxes and benefits were evaluated on the post-fiscal income margin, answering to the question: if actual values of each tax and benefit are changed independently from each other by some small factor , what are the shares of this tax or benefit in total marginal changes of vertical, reranking and redistributive effects? Despite their small share in comparison to taxes, the benefits created about one fifth of the overall redistributive effect in Slovenia and Croatia. However, there was a large difference between the impact produced by means-tested and non-means-tested ocial benefits in the two countries. In Slovenia, the vertical effect of non-means-tested sted social ertical effect, albeit relatively smaller than the vertical effect of efits.
IT was the largest contributor to the vertical effect in both of pre-fiscal income in Slovenia, it accounted for 55.6% of For the purpose of our analysis, we employed a series of decompositions of the redistributive effect that, to the best of our knowledge, has not been done before in this manner. First, we decomposed the redistributive effect into vertical, classical horizontal inequity, and reranking effects. Additionally, we performed the well-known Kakwani's decomposition of the redistributive effect, which contains only the vertical and reranking effects. Finally, all the effects obtained from the Kakwani's decompos were uniquely decomposed to reveal contributions of individual tax and benefit instruments to the vertical, reranking and redistributive effects. In particular, this is the first application of the Duclos et al. (2003) and Urban's (2012) methodology in comparative analyses of tax and benefit systems.
The results reveal that the Slovenian tax and ben v also induced much more horizontal inequity, which cancelled the advantage in vertic th the final result that th e S e s social benefits was close to zero, indicating that these benefits were on average almost proportional to pre-fiscal income. On the other hand, in Croatia non-means-te benefits did induce a v eans-tested social ben m On the revenue side, P ountries. Absorbing 9.1% c the vertical effect; in Croatia the PIT took up 5.9% of pre-fiscal income and was responsible for 38.5% of the vertical effect. Social security contributions contributed significantly to the vertical effect; in Slovenia (Croatia) they accounted for 63.9% (61.6%) of the vertical effect. Social security contributions also introduced quite a lot of reranking in both countries. This result was not unexpected, because we know that these taxes were paid prevalently by the employees and not by pensioners and other inactive groups that made a large part of population in these countries. The contribution of PIT e Slovenian tax and benefit system. ića (2006) Poverty and Equity: Measurement, Policy and E to reranking in Croatia was surprisingly low, amounting to only 2.2% of the redistributive effect, compared to 8.8% of the redistributive effect in Slovenia. Reranking caused by Slovenian non-means-tested social benefits further aggravated their negative contribution to redistributive effect. However, the contribution of benefits in Croatia to reranking was six times higher than in Slovenia, evidencing another divergence between the two systems.
Taking into account these results, we can conclude that Slovenia revealed lower pre-fiscal income and post-fiscal equality than Croatia. Even though Croatian tax and benefit system contributed substantially to the equalisation of post-fiscal income, it was still slightly less redistributive compared to th Croatian PIT induced much less horizontal inequity than the Slovenian PIT, while the opposite was true for benefits.