Tax-benefit systems, income distribution and work incentives in the European Union

Our analysis makes use of EUROMOD, the tax-benefit microsimulation model for the European Union, which provides cross country comparable measures of direct taxes and social insurance contributions liabilities, as well as cash benefit entitlements for the household population of all 27 EU Member States. EUROMOD is a static microsimulation model, meaning that behavioural responses to policy reforms are not taken into account and socio-demographic changes are not considered. Instruments which are not simulated in EUROMOD are taken directly from the data, when available. See Sutherland and Figari (2013) for further information.

The underlying micro-data used for the simulations in EUROMOD reported here come from the 2008 European Union Statistics on Income and Living Condition (EU-SILC).³ Exceptions to the use of EU-SILC 2008 concern France, where EU-SILC 2007 is used; Malta, where EU-SILC 2009 is used; and the UK where the 2008/09 Family Resource Survey (FRS) is used.

Currently, policy years 2005 to 2010 are covered in EUROMOD, for the purpose of tax-benefit simulations.⁴ In this study, the tax and benefit rules used are those in place on the 30th of June 2007, when we refer to 2007 systems, and 30th of June 2010, when we refer to 2010 systems. For most countries, the 2007 policy system matches the income reference period of the underlying micro-data, since EU-SILC 2008 reports income from the previous year.

When simulated policy systems refer to a year that is different from the underlying micro-data income reference period (e.g. 2010 policy systems for all countries), non-simulated income variable are updated using country and income specific updating factors.⁵ An example can be useful to clarify this point: between one year and another wages changes, and not taking this into account might cause serious problems in the simulation of various instruments (e.g. income tax and social insurance contributions); EUROMOD captures the changes of the non-simulated monetary variables updating their values according to suitable indicators (e.g. wage growth rate for wages).

In most countries, simulation of taxes and benefits is based under the assumption of full compliance with policy rules and 100% benefit take-up. Under such assumptions, simulations tend to overestimate the amounts of taxes and benefits and hence underestimate inequality of disposable income. The extent to which these problems occur varies across countries. For this reason, adjustments have been made in some cases. Non take-up of certain benefits has been modelled for Belgium, Estonia, Greece, France, Ireland, Poland, Romania and the UK, while tax evasion has been accounted for in Bulgaria, Greece and Italy.⁶ While accounting for benefit non take-up and tax evasion is reasonable for a cross-country comparison of income inequality, the assumption of full compliance is more adequate for the study of the “intended effect” of the tax and benefit system on labour market incentives. For this reason, adjustments for tax evasion and benefit non take-up are not taken into account for the calculation of METRs.

In our analysis, equivalised⁷ household disposable income is used to evaluate household resources. Disposable income is the result of adding original household income to any social transfers and deducting direct taxes and social insurance contributions. Original household income includes earnings, income generated by assets, private pensions, investment and capital income, and transfers from other households. Taxes include income tax and other direct taxes. Social insurance contributions refer to employee and self-employed insurance contributions and social transfers include means-tested and non means-tested benefits, as well as public pensions.

In order to analyse the redistributive effect of tax-benefit systems, we use the standard Gini coefficient to measure income inequality. Similar to previous studies using EUROMOD (Immervoll et al. 2006; Paulus et al. 2010; Avram et al. 2012), we measure the redistributive effect of different tax-benefit instruments, by comparing Gini coefficients before and after taxes and/or benefits.

It is important to stress that interactions between instruments are not accounted for in the analysis, when different tax-benefit components are excluded from disposable income. The exercise consists therefore of a static decomposition. For instance, we ignore that excluding benefits might cause a decrease in taxes, as some benefits are taxable; and that the absence of non means-tested benefits in the systems might result in an increase of recipients of means-tested benefits. This is equivalent to the usual methodology used for analysis directly from micro-data (Mahler & Jesuit, 2006; Fuest et al. 2010).

Marginal Effective Tax Rates (METRs) are a measure of the incidence of the tax and benefit system on a marginal increase of earnings. They are, in particular, an indicator of the proportion of a marginal increase in earnings that is taxed away due to social insurance contributions, taxes and loss of benefit entitlement. This feature makes METRs a popular indicator of the incentives faced by a worker on the intensive margin of labour supply.⁸

Because of the complexity of tax and benefit systems, it is possible that a variation in earnings not only affects the disposable income of the individual whose earnings change, but also benefit entitlements or tax burden of other members of the household. Therefore, METRs for each individual are evaluated taking into consideration taxes paid by, and benefits paid to, all members of a household and affecting household current cash disposable income.⁹

Individual level METRs are calculated according to the following formula:

where the numerator measures the change in household disposable income before $\left(Y_{HH}^0\right)$ and after $\left(Y_{HH}^1\right)$ the increase in individual earnings (E_i), the denominator is equal to the increase in earnings itself.

The calculations are performed by EUROMOD through the following steps: first, household disposable income is calculated; then individual earnings are increased in turn for each earner in the household and the corresponding household disposable incomes are computed. METRs are hence calculated and assigned to the individual whose earnings have been increased. Consider the following example: assume two earners, i and j, are part of a household. Earner specific marginal effective tax rates are calculated as follows. First, household disposable income $\left(Y_{HH}^0\right)$ is calculated. Earnings of the first earner (i) are then increased by the margin, holding constant the earnings of the other earner (j) and any other household characteristics. The new household disposable income $\left(Y_{HH}^1\right)$ is hence computed simulating new tax liabilities and benefit entitlements for all household members and METRs for the first earner (i) are calculated. The same procedure is then applied to the second earner (j).

Moreover, disposable income can be decomposed as the arithmetical sum of original incomes (O) (Incomes before any tax and transfer), benefits and pensions (B) minus taxes (T) and social insurance contributions (S). Equation (1) can hence be rewritten as:

Since the change in original incomes is equal to the change in earnings, the expression can be further rewritten as:

where the first component reports the reduction in benefit and pensions at the household level as a percentage of the earnings increase; the last two components report the increase in taxes and social insurance contributions as a percentage of the earnings increase.

In the context of this analysis, the marginal increase applied to earnings is set equal to 3% of total earnings. The choice of this margin is due to the fact that extra 3% earnings roughly corresponds to an extra working hour per week for a worker working 40 hours per week.

In theory marginal tax rates are usually expected to take values between 0% and 100%. Many aspects of tax and benefit systems make METRs outside this interval a possibility. It is possible, for example, to have METRs above 100 if the marginal increase in earnings causes the loss of some benefit entitlement. In some cases the reduction in benefit overcomes the increase in earnings, causing household disposable income after the increase in earnings to be lower than the household disposable income before it. Negative values of METRs might arise from tax allowances or benefit entitlements that are only paid to people with income above a given threshold; as reported in Immervoll (2004), those with income just below the threshold will see their household disposable income increase more than earnings do, resulting in a negative marginal effective tax rates.

Being an indicator of the incentive faced at the intensive margin of labour supply, Marginal Effective Tax Rates are only calculated for those reporting positive earnings in the underlying survey data. Following Immervoll (2004), we choose to focus our calculation on a sub-sample of earners: those aged 18-64. Excluding people below and above the age thresholds allows us to focus on incentives faced by the working age population. We further restrict this sub-sample to those with more than 1 unit of national currency of monthly earnings.¹⁰ Finally, in order to reduce the risk of our calculations being biased by “outliers”, we exclude from our calculations the top percentile of the METR distribution if the METR is above 150% and the lowest percentile if the METR is negative. Although METR values of observations considered “outliers” are plausible, we exclude them from our analysis as they are likely to bias our calculation of average METRs, especially when deciles specific measurements are performed.

Before analysing the effect of tax-benefit instruments on income distribution and work incentives, it is worth considering the extent to which different tax-benefit components vary in size on average across countries in Europe. Figure 1 shows the relative size of five tax-benefit components for EU-27, where the average size of each income component is measured as a percentage of average household disposable income.¹¹ Taxes and social insurance contributions are shown as negative values as they represent deductions from disposable income.

Figure 1

Download asset Open asset

Note that the size of the redistributive system in each country could be roughly measured by the total length of the bars in Figure 1, which represents the sum of all tax and benefit components. Using such indicator, we observe a substantial variation of the size of tax-benefit systems across EU countries. Cyprus ranks particularly low in terms of the size of the redistributive system, while Nordic countries (Denmark, Sweden and Finland) and continental countries such as the Netherlands, Belgium, Germany and Austria, as well as Hungary, are the highest ranked. These results are consistent with previous studies using EU-SILC and EUROMOD (Immervoll et al. 2006; Paulus at al. 2009; Fuest et al. 2010; and Avram et al. 2012).

In terms of deductions, income taxes are in general more important than social insurance contributions, with some exceptions such as Czech Republic, Greece, the Netherlands, Slovakia and Slovenia. Nordic countries (Denmark, Sweden and Finland) and Belgium present particularly high shares of income taxes, while Czech Republic, Greece, Malta, Cyprus and Slovakia rank among the ones with lowest shares. Taking into account both income taxes and social insurance contributions together, taxation ranges between 12% of disposable income in Cyprus and 52% in Denmark.

From the benefits side, public pensions represent by far the largest component, except in Anglo-Saxon countries (Ireland and the UK). The relative size of pensions is particularly important in Austria, Poland, Italy and Hungary, where they account for over 30% of household disposable income. Besides Ireland and the UK, the Netherlands also has a small share of public pensions, which is related to the prominent role of private and occupational pension schemes in these countries. In general, non means-tested benefits represent only a small part of disposable income, except in Denmark and Ireland where they amount to more than 10%. Means-tested benefits represent, on average, an even smaller share of disposable income (less than 5%), except in the UK and Ireland. Among other countries, where means-tested benefits represent an important component of social benefits, we observe the Netherlands, France and Finland.

The relative size of different tax-benefit components varies for individuals in different parts of the income distribution. In general, taxes tend to be progressive and some types of benefits are targeted to groups with low incomes. In order to consider such variation, we replicate the previous analysis for the bottom and top deciles of the income distribution in the EU-27 countries.

Figure 2 presents the relative size of tax-benefit components for households in the bottom decile of the distribution. As expected, the relative size of benefits is more important among this group of the population. In 20 out of the 27 countries analysed, benefits together with public pensions represent more than 60% of household disposable income. Even in Greece, where the share of social benefits and pensions is the lowest, it amounts to a considerable 44% of disposable income. Besides pensions, means-tested benefits represent now an important share of disposable income, ranging between 3.5% in Cyprus and 51.4% in Romania. Non means-tested benefits are in general lower, except in Sweden, Belgium, Hungary and Denmark.

Figure 2

Download asset Open asset

Taxes and social insurance contributions represent a much lower share of disposable income in the bottom decile of the distribution. Some exceptions are Denmark and Sweden where many benefits are taxed; and the UK, where most of the tax consists of Council Tax (which is offset by a benefit for those on low incomes). Few countries are also characterised by high shares of social insurance contributions. This is the case of the Netherlands, where contributions are payable on benefits; Hungary, where this is related to high contributions paid by the self-employed who are under a simplified taxation regime; Poland and Romania.

Turning to the top of the income distribution, Figure 3 shows that the relative size of taxes and benefits is reversed, with respect to the picture of the bottom decile. As expected, now taxes and social insurance contributions represent the main income components. In all countries, the share of taxes and social insurance contributions is greater than the share of benefits received. The only exception is Cyprus, where public pensions¹² are an important component of household income at the top decile. Total deductions (direct taxes plus social insurance contributions) are particularly important in Belgium and the Netherlands, where they represent over 60% of disposable income, while in Cyprus they account for less than 20%. The share of social benefits consists mainly of public pensions. The lowest share of social benefits is found in the Netherland and the UK, while Hungary, Italy and France have social benefits of over 30% of disposable incomes, which mostly corresponds to public pensions.

Figure 3

Download asset Open asset

One of the roles of tax-benefit systems is to reduce income inequality (Boadway & Keen, 2000). In this section, we analyse the extent to which the redistributive effect of taxes and benefits, varies across European countries. Figure 4 presents Gini coefficients for household original income (before benefits receipt and tax-contribution deductions), original income with public pensions and disposable income for EU-27 countries. Each income measure is equivalised using the modified OECD scale. Countries are ranked by Gini coefficients of disposable income. Confidence intervals for Gini coefficients, calculated using bootstrapping techniques (with 1000 replications), are also included in the graph.

Figure 4

Download asset Open asset

Inequality in terms of household disposable income is low (Gini between 0.21 and 0.25) in Nordic countries (Denmark, Sweden and Finland), some continental countries such as Belgium, France, Luxembourg and Austria, as well as in Slovakia, Czech Republic, Slovenia and Hungary. Among countries with high inequality in disposable income (Gini between 0.31 and 0.34) we observe Southern countries, such as Italy, Greece and Portugal; Eastern European countries such as Bulgaria, Latvia, Lithuania, Poland and Romania, and the UK.

The difference between Gini coefficients for original income and disposable income represents the total redistributive effect of tax-benefit systems. Tax-benefit systems have a significant effect on reducing inequality, although to a different degree across countries. In most countries, public pensions have a stronger effect in reducing income inequality than the remaining tax-benefit instruments combined. The only exceptions are the Netherlands, Ireland, the UK; and to a lesser extent Denmark. Comparing Gini coefficients of original income plus pensions and disposable income, we observe that tax-benefit systems play an important role in reducing inequality in Belgium, Germany and Anglo-Saxon countries (UK and Ireland). The smallest effect of tax-benefit systems (excluding pensions) is found in Cyprus, Estonia, Latvia and Lithuania.

In order to have a better idea of the contribution of different tax-benefit components to reducing inequality, we calculate Gini coefficients excluding each instrument individually from disposable income. Figure 5 shows the absolute change in Gini coefficient of disposable income when each instrument of the tax and benefit system is excluded. As previously said, our results do not take into account interactions between instruments within the tax-benefit system, when different components are excluded in order to recalculate Gini coefficients.

Figure 5

Download asset Open asset

In most countries, direct taxes tend to be the most important instrument in reducing inequality ranging from an absolute change of about 0.02 in Slovakia to 0.07 in Belgium. In Ireland and the UK, means-tested benefits play a major role on income redistribution, accounting for an absolute change in Gini of 0.075 points. Other countries, where means-tested benefits dominate the effect of direct taxes are France, Malta and Romania. Non means-tested benefits, on the other hand, are particularly important in Denmark, Sweden, Belgium and Hungary. Finally, in most countries social insurance contributions reduce inequality but in a rather small degree, while in very few cases (Netherlands, Spain, Malta, Cyprus and Romania) inequality increases, which could be related to contributions being payable on benefits or to the presence of low caps on contributions (Avram et al. 2012).

3.2.1 Changes in income inequality across 2007 and 2010 policy systems

In this subsection, we briefly describe the effect of policy changes between 2007 and 2010 on income inequality. For this purpose, 2010 policies are simulated with the same input data used for the analysis of income inequality in 2007. Labour market characteristics are, hence, kept fixed with the aim of isolating the effect of policy changes. As described in Section 2.1, an updating process of the monetary variables not simulated in EUROMOD is performed. As such, the analysis presented in this section represents a static microsimulation exercise, where behavioural responses to policy changes are not accounted for.

Figure 6 presents Gini coefficients in policy years 2007 and 2010. In most countries income inequality falls or remains the same. Latvia, Ireland and Estonia are the countries where inequality decreases the most, while Hungary and Poland present the highest increase (around one percentage point). Note that, by abstracting from changes in population characteristics and behavioural responses, the changes in Gini coefficients estimated by EUROMOD represent the effect of policy changes on income inequality. However, some of the changes across countries and over time are also affected by the way incomes are uprated (Jara, Sutherland et al., 2013). In this sense, our estimates are measuring the effects of policy changes relative to income growth.

Figure 6

Download asset Open asset

The analysis presented above described the extent to which different tax-benefit systems in the EU redistribute income and affect work incentives. The aim of this section is to analyse whether a trade-off between redistribution and work incentives is observed across countries in the EU. Are countries with low levels of income inequality characterised by weak work incentives? We use the results from previous sections in order to discuss the relationship between income inequality and work incentives across EU countries.

While the distributional analysis has been conducted for the population as whole, METRs calculations are based on earners in working age only. The choice of focusing on different populations is justified by the fact that tax-benefit systems affect income distribution across the whole population, but they can create incentives and/or disincentives for specific groups (earners in this case). In making the link between redistribution and work incentives we want to keep this specificity and be able to analyse whether countries with low (high) income inequality for the population as a whole are characterised by weaker (stronger) work incentives for those currently in work.

In general, reducing income inequality and providing incentives to work can be considered as two objectives governments need to balance, when designing tax-benefit policies (Adam & Browne, 2010). The relationship between income redistribution and work incentives has been previously analysed, at a national level. For instance, Adam et al. (2006) found a significant negative relation between METRs and Gini coefficients over time in the UK. More generally, the trade-off between equity and efficiency has been widely analysed in the framework of optimal income taxation (see, among others, Mirrlees, 1971; Saez, 2001; Saez, 2002).

Figure 14 plots Gini coefficients and mean Marginal Effective Tax Rates for EU-27 countries in 2007. The figure shows that countries with high income inequality are usually characterised by strong work incentives. Income inequality is negatively correlated with mean METR, with a correlation of −0.41.

Figure 14

Download asset Open asset

In order to further investigate the relationship across EU countries, we compare changes in income inequality and work incentives across policy years 2007-2010. By abstracting from population changes and behavioural responses we attempt to capture the effect of policy changes on income inequality and work incentives. In particular, our aim is to study whether countries in which policy reforms decrease (increase) income inequality, present higher (lower) mean METR. Such pattern is depicted in Figure 15, where a negative correlation between differences in Gini coefficients and differences in mean METR is observed between 2007 and 2010. The equity-efficiency trade-off seems to describe the situation of most countries in the EU. For instance, in countries such as Hungary, the 2010 tax-benefit system increases income inequality while decreasing mean METR. The opposite is observed for countries such as Ireland, where inequality decreases while work incentives are weakened. It is worth noticing, however, that many countries do not follow this pattern. A decrease in both, inequality and mean METR is observed in Bulgaria and Denmark. Other exceptions are Germany, where work incentives strengthen with no substantial change in inequality; and Estonia and the UK where inequality decreases with no particular changes in work incentives.

Figure 15

Download asset Open asset

Despite being a mere association, Figures 14 and 15 highlight the existence of a negative relationship between METRs and Gini coefficient, suggesting the presence of a trade-off between work incentives and income inequality.

See also Bargain et al. (2012) for comparative evidence on labour supply elasticities in 17 European countries and the US. The study makes use of EUROMOD to enhance cross-country comparability between European countries

See also OECD, 2007.

In most countries, the UDB version of EU-SILC is used. However, for Greece, Spain, Italy, Austria and Slovakia the national version of the EU-SILC is used because it includes more variables at the necessary level of detail. For some other countries, additional national variables are added to UDB SILC.

The policy systems of the 27 EU member states have been modelled in EUROMOD through a tree stage process; as a consequence, 2005 systems are modelled for nine countries; 2006 systems for eighteen; and 2007-2010 system for all the 27 EU countries.

See country specific EUROMOD Country Reports for information about the updating process.

See country specific EUROMOD Country Reports for more information.

Disposable income is equivalised using the “modified OECD” equivalence scale, assigning a weight of 1 to the first adult, 0.5 to each subsequent adult and 0.3 to each child aged below 14.

Together with incentive in increasing the number of hours worked, METRs capture also the incentive in increasing effort (or wage) for the same labour supply.

It should be noted that EUROMOD simulations are usually based on yearly incomes averaged over the 12 months. This could introduce some degree of imprecision in the simulation of some policies affecting METRs if they are related to shorter period income. Social assistance benefits are, for example, typically affected by this limitation.

The choice of 1 unit of national currency of monthly earnings reduces the risk of having “too small” denominators and the risk of our results being driven by rounding factors. Since part of our analysis focuses on a comparison between 2007 and 2010 values of METRs, we imposed that the condition of having average monthly earnings above 1 unit of national currency must be met both with 2007 and 2010 earning values.

Note that EUROMOD offers the advantage of decomposing household disposable income into even more detailed components (see Avram et al. 2012).

Note that the definition of public pensions used here corresponds to the EU-SILC definition of old age benefits, including lump sum payments at retirement age, which could affect the income distribution.

The absence of the benefit withdrawal component in the average METRs decomposition for Greece is due to the lack of a broad social assistance benefit and to the fact that the remaining means tested benefits are implemented using thresholds or brackets. In these cases the extra earnings produce a variation in the benefit received only if those income limits are crossed, generating a sizeable reduction in the benefit amount, or even its loss. The very large values of the METRs following this process, as well as the very small number of cases for whom this happens, make very likely for them to be classified as “outliers” and not included in the analysis.

The process of extending and updating EUROMOD is financially supported by the Directorate General for Employment, Social Affairs and Equal Opportunities of the European Commission [Progress grant no. VS/2011/0445]. We acknowledge the contribution of all past and current members of the EUROMOD consortium and in particular our colleagues at ISER who worked on the development of the version of EUROMOD used in this paper: Silvia Avram, Paola De Agostini, Francesco Figari, Horacio Levy, Jekaterina Navicke, Alari Paulus, Olga Rastrigina and Iva Tasseva. We would also like to thank Holly Sutherland and an anonymous referee for valuable comments on previous versions of the paper. The usual disclaimer applies. We are grateful for access to microdata from the EU Statistics on Incomes and Living Conditions (EU-SILC) made available by Eurostat under contract EU-SILC/2011/55, the Italian version of the EU-SILC (IT-SILC) made available by ISTAT, the Austrian version of the EU-SILC made available by Statistics Austria, the Lithuanian version of the EU-SILC (PGS) made available by the Lithuanian Department of Statistics, the Slovak version of the EU-SILC made available by the Statistical Office of the Slovak Republic, the Spanish version of the EU-SILC made available by the Spanish Statistical Institute, variables from Bulgarian version of the SILC made available by the National Statistical Institute, variables from the Greek SILC Production Database (PDB) made available by the Greek Statistical Office and the Family Resources Survey (FRS), made available by the UK Department of Work and Pensions (DWP) through the UK Data Archive. Material from the FRS is Crown Copyright and is used with permission. Neither the DWP nor the Data Archive bears any responsibility for the analysis or interpretation of the data reported here. An equivalent disclaimer applies to all other data sources and their respective providers cited in this acknowledgement.

1. Version of Record published: April 30, 2013 (version 1)

© 2013, Jara and Tumino

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.