Labour force participation elasticities and the move away from a flat tax: The case of Slovakia

Senaj, Matus; Siebertova, Zuzana; Svarda, Norbert; Valachyova, Jana Article Labour force participation elasticities and the move away from a flat tax: The case of Slovakia IZA Journal of European Labor Studies Provided in Cooperation with: IZA – Institute of Labor Economics Suggested Citation: Senaj, Matus; Siebertova, Zuzana; Svarda, Norbert; Valachyova, Jana (2016) : Labour force participation elasticities and the move away from a flat tax: The case of Slovakia, IZA Journal of European Labor Studies, ISSN 2193-9012, Springer, Heidelberg, Vol. 5, Iss. 19, pp. 1-26, https://doi.org/10.1186/s40174-016-0069-y This Version is available at: https://hdl.handle.net/10419/195014 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ ORIGINAL ARTICLE Open Access Labour force participation elasticities and the move away from a flat tax: the case of Slovakia Matus Senaj 1 , Zuzana Siebertova 1,3* , Norbert Svarda 1,2 and Jana Valachyova 1 * Correspondence: [email protected] 1 Council for Budget Responsibility, Bratislava, Slovakia 3 Present Address: I. Karvaša 1, 813 25 Bratislava, Slovakia Full list of author information is available at the end of the article Abstract This paper provides a microeconometric analysis of labour force participation elasticities in Slovakia where we study the elasticity with respect to a unique tax reform whereby the flat tax was backtracked and replaced by a progressive tax. By estimating a probability model for labour force participation, we show that the low-skilled and females are groups that are particularly responsive to changes in income taxes and transfers. We perform a microsimulation analysis of two scenarios of flat-tax regime abolishment. We find that the recent departure from the flat-tax system in Slovakia in 2013, which introduced two tax brackets in personal income taxation, only negligibly reduced the average probability of being economically active at the extensive margin. A more significant average effect has been found in a hypothetical scenario with a similar fiscal revenue impact, simulating a departure from the flat-tax system by reintroducing five tax brackets. We show the different impacts of the two distinct scenarios of abolishing the flat tax on selected subgroups of the population. JEL Classification: H31, H53, I38, J21 Keywords: Labour force participation elasticity, Extensive margin, Microsimulation, Flat tax 1 Introduction This paper examines the link between labour force participation and changes in the tax system. As argued by Meghir and Phillips (2010), the impact of taxation on work incentives is one of the principal sources of inefficiency that may arise in a tax system. The fundamental issue is to assess how sensitive individuals’work incentives are to changes in taxes and benefits. An analysis of labour supply behaviour is therefore a key element when evaluating the reforms of tax and transfer systems and the impact of different policies on changes in tax revenues, employment, and wealth redistribution. The way how labour force participation responds to the work incentive/disincentive effects of taxation and welfare programmes has attracted a lot of interest in both labour and public economics, and extensive research has resulted in numerous empirical results. For an overview of the literature that connects labour supply to income taxes and social benefits, see, among others, surveys by Meghir and Phillips (2010), Moffitt (2002), and Blundell and MaCurdy (1999). IZA Journal of European Labor Studie s © The Author(s). 2016 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Senaj et al. IZA Journal of European Labor Studies (2016) 5:19 DOI 10.1186/s40174-016-0069-y In this paper, a case study of the recent moving away from the flat-tax system in Slovakia is performed. The idea of introducing a flat-tax regime was widespread among Central and Eastern European (CEE) countries, including Slovakia, at the beginning of the 2000s. Effective from 2004, the system of graduated personal income tax rates in Slovakia was simplified to a flat-tax rate of 19%. Only a few EU countries abandoned the flat tax and returned to the more progressive system by reintroducing tax brackets. These countries currently include Slovakia and the Czech Republic. Using the detailed microsimulation model and estimates of participation decisions (labour supply elasticities at the extensive margin), we quantify the effects of the tax system reform valid from 2013 that resulted in a marginal move away from the flat tax and to a positive first-round fiscal effect leading to an increase in revenues by 0.4% of GDP. This reform has been characterized by the re-introduction of two tax brackets and the unification (and increase) of the assessment basis for different social and health insurance contributions. In addition, we perform a simulation of the counterfactual scenario abolishing the flat-tax regime by introducing five tax brackets. In broad terms, this scenario simulates the personal income tax system valid in Slovakia before the flat-tax reform in 2004. By simulating these two scenarios with the same simulated first-round revenue effect, we show that the departure from a flat tax can have a different impact on selected population subgroups. The literature on the microeconometric estimations of labour supply elasticities is vast. A comprehensive overview in relation to the progress in the field of microsimulation models focused on labour supply, and methodological approaches can be found in Aaberge and Colombino (2014), who identify three main methodologies that have been adopted for modelling labour supply. A so-called reduced form approach embodies the hypothesis that labour supply (namely the observed hours of work) is a function of an exogenous net wage and net income. In general, it is not a precise representation of dependence, mainly due to the non-linearities of budget constraint. Moreover, corner solutions are usually ignored. The structural “marginalist”approach works with the conditions for a constrained maximum of the utility function; see, among others, a presentation of the methodology by Hausman (1981). The solution is obvious when convex budget sets and a consumption-leisure setup in the utility function are assumed. However, the method tends to be cumbersome if more complicated non-convex budget constraints are formulated. As a response, a discrete choice framework based on the concept of a “random utility maximization”presents an often-used alternative. This approach, introduced originally by van Soest (1995), has become rather standard in recent years. The utility maximization problem of individuals is reduced to a choice among a discrete set of options (yielding different utilities) such as working full time, working part time, or not working at all. Being inactive thus presents one of the alternatives, and the extensive and intensive margins could be directly estimated so that labour supply decisions could be evaluated even in the presence of non-convexities in budget constraints. Currently, a number of empirical studies conclude that an extensive margin is much more important than an intensive one. Existing studies usually evaluate labour supply elasticities of some special demographic subgroups (e.g. single individuals, married women, and couples). They usually find that wage elasticities are larger for women than for men. Looking at the magnitude of the estimated elasticities, the variation of the Senaj et al. IZA Journal of European Labor Studies (2016) 5:19 Page 2 of 26 results found in the literature is sizeable. As noted by Bargain et al. (2014), differences across studies arise due to the distinct methodologies applied, including the underlying datasets used (administrative versus survey data) as well as the periods of study. An overview of recent estimates of labour supply elasticities in the US economy can be found in Chetty et al. (2013) and in McClelland and Mok (2012). For an overview of recent empirical evidence on labour supply elasticities in Europe and the USA, see Bargain et al. (2014). A brief survey and critique of different methods of the estimation of labour force participation elasticities can also be found in Heim (2008). However, despite the multitude of methodologies and information covered by existing studies, analyses focusing on CEE countries are rather scarce, and the case of Slovakia has been covered only in one paper so far. Chase (1995) compared labour force participation and wage elasticities between the communist and post-communist regimes in Slovakia and the Czech Republic. He showed that women’s participation in the labour market was higher under communism and concluded that the effects of changes in earnings are smaller in Slovakia compared to the Czech Republic. This is probably a result of the slower transformation of the Slovak economy. Looking at countries bordering Slovakia, Benczur et al. (2014) studied the labour supply at the extensive margin in Hungary. They modified an existing structural approach originally proposed by Hausman (1981) by taking the effects of the tax and benefit system directly into account. As regards the participation decision, they showed that wages, taxes, and transfers have a stronger influence on the participation decision of individuals that are older, low-skilled, married women, or women of child-bearing age. Galuscak and Katay (2014) followed the same methodology and provided empirical estimates for the Czech Republic, which are close to those reported for Hungary. Another analysis focused on the Czech Republic was performed by Bicakova et al. (2011), who provided estimates of participation probabilities separately for males and females by using a probit model. Compared to the study by Galuscak and Katay (2014), the estimated wage semi-elasticities of labour supply are substantially smaller, even though they are larger for women compared to men. Our estimates of participation elasticities are based on a model of labour supply where both taxes and social transfers are simultaneously taken into account. We estimate a labour supply model following the methodological approach introduced by Benczur et al. (2014). The behavioural response is based on the rationale of utility maximization, and using the classification provided by Aaberge and Colombino (2014), it can be seen as a “marginalist”approach. The model covers in minute detail the joint effects of tax and benefit systems on individuals’net income. Using this modelling strategy, individual participation probabilities are determined by comparing incomes in two states: being in the labour force and being out of it. A key component of this approach is to precisely evaluate the disposable income of an individual, including nonlabour income and social transfers received by a household in both states. In order to do so, the concept of the gains to work of an individual is introduced and defined as the difference between the net wage and the amount of welfare benefits lost due to taking up a full-time job. Employing this microeconometric method allows us to evaluate how the Slovak taxbenefit system can affect work incentives at the extensive margin. We document that participation probabilities are generally dependent on the level of net income and Senaj et al. IZA Journal of European Labor Studies (2016) 5:19 Page 3 of 26 non-labour income, including social transfers. We find that a 1% increase in gains to work increases the probability of economic activity by 0.08 percentage points for males and 0.12 percentage points for females. Our findings are broadly in line with the results usually reported in the literature that frequently demonstrate that elasticities are large for women and very small for men. Taking into account tax and transfer system details valid from 2010 to 2012, a 1% increase in non-labour income decreases the probability of labour force participation by 0.04 percentage points for both genders. Policy initiatives likely to increase financial incentives to work should result in higher participation rates. Our results also show that, in line with findings for other countries, the lowskilled and females are the groups that are particularly responsive to changes in taxes and transfers. A major advantage of this method is that it allows an ex ante assessment of the counterfactual tax and transfer system reforms and permits an evaluation of specific government interventions and policies. The essential part of this modelling approach is the SIMTASK module, a microsimulation model of the Slovak tax and transfer system described in detail in Siebertova et al. (2016). This tool enables us to simulate individual tax liabilities and benefit entitlements in detail according to valid legislation or hypothetical reform. Labour supply models are extensively used in the literature to assess the effects of proposed tax system reforms. In such studies, the (hypothetical) introduction of a flat income tax and its impact on the supply of labour is frequently analysed, for example, Decoster et al. (2010) studied the introduction of a flat tax in Belgium and found that a flat-tax system could potentially increase labour supply. The introduction of linear taxation in Germany was examined in Beninger et al. (2006) and Fuest et al. (2008). Beninger et al. compared the effects in a computed manner by using a unitary and collective labour supply model. Fuest et al. used a behavioural microsimulation model and concluded that flat-tax reform could potentially increase employment although the magnitude of the increase was very small. Duncan and Sabirianova Peter (2010) analysed the Russian flat-tax reform of 2001 by using the difference-in-difference regression approach. As a reaction to tax changes, they identified an increase in the distribution of hours worked and that the reform increased the probability of finding a job. Compared to the studies mentioned above, we perform a kind of “reverse”analysis where we study the effects of departure from the flat-tax system. In our setup, the baseline is the flat-tax system valid in 2012 in Slovakia and we study the effects of reintroducing the tax brackets. By performing a microsimulation analysis of two scenarios, we show that a different way of moving away from the flat-tax system may have a different impact on labour supply decisions. We find that the recent departure from the flat-tax system in Slovakia effective from 2013 slightly reduced the average probability of being economically active. Although the hypothetical scenario of the abolition of the flat-tax system would have a higher average impact on the probability of being economically active, we show that the impact on participation probabilities in the two scenarios differs for selected population subgroups. In our analysis, we investigate the immediate or “day-after”effects of two reforms. A long-run general equilibrium analysis will be performed in a separate paper, since the discussion and execution of these issues is beyond the scope of the present study. The rest of the paper is organized as follows: the next section provides an overview of the developments in the Slovak labour market and briefly describes the reforms to Senaj et al. IZA Journal of European Labor Studies (2016) 5:19 Page 4 of 26 the tax-benefit system; Section 3 presents the modelling approach that was employed in this article; Section 4 follows with a data description and definition of variables used in the model; Section 5 depicts a short introduction of the Slovak tax and benefit system; Section 6 discusses the main results on estimated labour supply elasticities and provides tax reform simulations; and Section 7 offers a conclusion. In the Appendix, we list the definitions of the main variables and present the detailed results of our estimations. 2 Labour market outcomes and policies in Slovakia The empirical evidence on labour supply behaviour in transition and post-transition countries is limited. From the historical point of view, labour force participation was obligatory in the CEE countries that experienced communist regimes. In general, after the change of regimes at the beginning of the 1990s and during the transition period, when national economies changed from planned to market-based ones, a continual withdrawal from the labour force was detected in labour markets in all CEE countries. Participation and employment rates in Slovakia reached their lowest in the early 2000s. Later, in the period of economic growth, an increase in both rates was observed; they started to decline again in 2009 as a consequence of the global economic crisis. The situation in post-transition Slovakia to 2012 can be characterized by participation rates (see Fig. 1) being permanently below the EU-27 average but still somewhat high compared to Hungary and Poland. Low activity rates in Slovakia persist, especially for labour market entrants and for individuals with low qualifications. Participation rates of the youth and low-skilled (low-educated) workers are excessively low, even compared to neighbouring countries (see Fig. 2). The Slovak tax-benefit system experienced major changes over the last decade. Both tax and social transfer systems were considerably modified 1 in 2004, when Slovakia became the first among Central European countries to implement a flat income tax scheme. It followed the cases of the Baltic States, where the flat tax was introduced in the mid-1990s and Russia in 2001. Afterwards, other CEE countries also followed this flat-tax track, among them Ukraine (in 2004); Georgia and Romania (in 2005); Albania, Bulgaria, and the Czech Republic (in 2008); Bosnia and Herzegovina (in 2009); and Hungary (in 2011). As in most of the mentioned countries, the implementation of the 19% flat income tax (both personal and corporate) in Slovakia was supplemented by additional reform changes, including the modified definition of the tax base, social and Fig. 1 Participation rates in Slovakia and selected countries, 1998–2012 Senaj et al. IZA Journal of European Labor Studies (2016) 5:19 Page 5 of 26 health insurance contributions, indirect taxation (the VAT rate was unified at 19%), and tax administration. In countries with established market economies (like those in Western Europe), there was not such a strong requirement for overall reforms as there was in transition countries, and as a consequence, the demand for the introduction of the flat tax has not been so appealing. As pointed out by Fuest et al. (2008), flat-tax reform is unlikely to take place in Germany due to its questionable distributional impact and limited efficiency effects. After the economic crisis, most countries needed to increase their revenues due to growing deficits, and increasing taxes, whether income or indirect, looked like an appealing tool. Currently, most of the CEE countries that introduced flat-tax regimes still use them. However, Ukraine abandoned the flat tax in 2011. Slovakia, the Czech Republic, and Montenegro returned back to the progressive system in 2013 by introducing a second tax bracket in the personal income tax scheme. It is noteworthy that the threshold for the higher tax rate in both Slovakia and the Czech Republic is sufficiently high and only applies to a small fraction of taxpayers in both countries. In Slovakia, the levels of social transfers were also effectively significantly cut from 2004 in order to increase work incentives. A report conducted by the World Bank (2012) has shown that these reforms have considerably improved work incentives for low-income workers. However, this improvement has mainly been achieved due to a reduction in transfer levels. The tax-benefit system currently valid in Slovakia seems to encourage work more than the system valid before reforms in 2004. On the other hand, low-wage part-time work is still not sufficiently attractive for those who are eligible to receive the material needs benefit. The Slovak transfer system is restrictive especially for labour market entrants and low-skilled workers employed in low-paid jobs. Structural changes both in the tax and transfer systems that followed from 2005 to 2012 were minor and are well documented in Porubsky et al. (2013). 3 Methodology In this section, we set up the microeconometric model of labour supply behaviour. We present an approach where taxes and transfers are explicitly taken into account. This extension of the standard labour supply model leads to the specification of a probit model that relates labour participation probabilities to the gains to work from working full time, non-labour income, and other individual characteristics. Finally, we show that participation elasticities can be derived analytically when using this methodology. Fig. 2 Participation rates by education and age in 2012 Senaj et al. IZA Journal of European Labor Studies (2016) 5:19 Page 6 of 26 3.1 A specification of the model of participation decision The labour supply decision of individuals is usually modelled as a utility maximization problem formulated as a consumption-leisure trade-off 2 : max c;l uc;1−lðÞ ð1Þ subject to the budget constraint cþw1−lðÞ¼wþNY;ð2Þ where cstands for consumption, wis wage, lis labour, and NY is other non-labour income, including the income of other household members and government transfers. Note that the budget constraint includes the disposable income of the whole household; thus, the income of other members also affects the labour supply decision of an individual. The total time endowment between work and leisure is normalized to 1, so (1 −l) denotes leisure. Using this modelling framework, taxes regulate the decision to supply labour through their impact on net market wages and non-labour net income. When employing the standard utility function 3 characterized by strictly positive marginal utilities, the optimality condition is determined by first-order conditions:wu0 c c;1−lðÞ¼u0 1−lc;1−lðÞ. An individual will participate if the utility from working will exceed the utility from not working. In this theoretical framework, non-participation in work results from the corner solution of the model (Hausman, 1981). Note that if an individual does not work, the optimal consumption equals c= NY. The reservation wage is the lowest wage rate at which the worker will be willing to accept a particular job, i.e. working non-zero hours, and in this setup, it can be expressed as wres ¼u1−l 0NY;1ðÞ u0 cNY;1ðÞ ¼NYψχ: ð3Þ An individual takes up a job if the offered wage exceeds his reservation wage w≥w res , or put differently, logw≥ψlogNY + logχ. Assume that individuals differ in their preferences so that relation logχi¼Z0 iαþεiholds. Z i is a vector of observable preferences that affect an individual decision to work, and εieN0;σ2 ε  is the error term independently and normally distributed among individuals. Given the assumption of the normality of the error term, the probability that an individual supplies labour can be estimated using the standard probit specification Pr activityi¼1  ¼ΦγlogwiþZ0 iα−ψlogNYi  ;ð4Þ where Φ(∙) stands for the standard normal cumulative distribution function. The early generation of static models of labour supply, represented essentially by the approach of Hausman (1981), were capable of only partially representing the effects of tax and transfer policies on household budget sets. Relying on tangency conditions, the Hausman model is restricted to the case of (piecewise) linear and convex budget sets. As argued by Benczur et al. (2014), this assumption is particularly restrictive if certain benefits expire immediately after taking up a job and the wage earned for the first few hours does not reward this discrete downward jump in transfers. Senaj et al. IZA Journal of European Labor Studies (2016) 5:19 Page 7 of 26 In the next step, we methodologically follow the approach presented in Benczur et al. (2014). Adding taxes and social transfers to the model leads to a redefinition of the reservation wage at the cost of the participation decision of an individual needing to be constrained to a full-time job. The participation decision is defined by comparing the utility derived from working full time and the utility from being inactive and receiving full social transfers. Taking into account the corresponding budget constraints, estimating the probability of being economically active yields a probit equation. Considering the binomial probit can be supported by the fact that in Slovakia the most typical form of employment is full-time employment. As is shown by statistics from Eurostat, Slovakia is a country with one of the lowest shares of workers in Europe who are employed part time. In 2012, the share of part-time workers was only 4% as opposed to 20% in the EU-27 on average. 4 A similar situation has been documented in Hungary and the Czech Republic. To derive formal expressions, we first introduce the concept of the gains to work (or effective net wage) variable GTW i of the individual i, defined as the annual net wage w i minus the difference between social benefits if not working and social benefits if working: GTWi¼b wi−SBNW−SBW  ¼b wi−ΔSB;ð5Þ where the term in parentheses expresses the amount of social benefits lost when working and the net wage b wiis computed from the predicted gross wage. Since income from employment is naturally unobservable for those who are unemployed or inactive, we use Heckman’s sample selection methodology (Heckman, 1979) to predict gross wages. 5 In order to obtain a consistent vector of gains to work GTW and reduce the division bias, we use the predicted values of gross wages for every individual in our sample (also for the employed), as this is common in the labour supply literature: see, for example, Bargain et al. (2014) and Breunig and Mercante (2010). To construct the vector GTW, a microsimulation tool is needed. The SIMTASK tax and benefit calculator is used to compute net wages from gross wages and simulate the amount of social benefits an individual is entitled to when working (SB W ) and when not working (SB NW ), taking into account the individual’s characteristics as well as the characteristics of the corresponding household. In our implementation, considering the details of the tax and transfer system, social benefits that enter the variable GTW include the meanstested material needs benefit and its supplements allocated at the household level. 6 The second variable of principal interest to us is the non-labour income NY i of the individual i, which is defined as a sum of three components, namely the social benefits that an individual is entitled to when not working, the non-labour income of all household members (including individual i), and the net labour income of other members of the household. Non-labour income covers pensions, income from property, dividend payments, and family-related benefits (eligibility does not depend on whether a parent works or not) and the unemployment benefit (we assume that this transfer does not affect the decision to work—it is a contributory benefit and expires after 6 months). Note that the construction of the variable NY i also needs a microsimulation tool. Using the notation of the standard labour supply model presented above, the budget constraint of an individual that does not work can be written as follows: c=NY,1−l=1, and the utility is given as u(NY, 1). When working full time (l * ), the budget Senaj et al. IZA Journal of European Labor Studies (2016) 5:19 Page 8 of 26 Prime-age males with small children under 3 years of age are identified as the subgroup with the smallest semi-elasticity. On the contrary, females with small children are the group with the highest responsiveness, being ten times higher than that of males in the same category. Overall, the presented results suggest that policies that make work pay would lead to an increase in participation. The low-skilled and females are the groups that are more responsive to changes in taxes and transfers. This implies that labour market policies, namely tax and transfer system reforms that are aimed at boosting economic activity, should be primarily targeted at low-educated individuals and women. A comparison of estimates obtained for Slovakia to those published for neighbouring countries can be found in Table 4. It appears that the magnitudes of Slovak estimates are lower compared to those for Hungary or the Czech Republic (Benczur et al. 2014, Galuscak and Katay, 2014), suggesting a lower sensitivity of individuals to changes in labour and non-labour incomes. On the other hand, Bicakova et al. (2011) provided Table 3 Marginal effects by selected subgroups dy/dx Std err dy/dx Std err Females, age 25–50 Males, age 25–50 Gains to work (logGTW) 0.109 0.010 Gains to work (logGTW) 0.056 0.006 Non-labour income (logNY) −0.036 0.003 Non-labour income (logNY) −0.025 0.002 Single females, age 25–50 Single males, age 25–50 Gains to work (logGTW) 0.123 0.012 Gains to work (logGTW) 0.085 0.010 Non-labour income (logNY) −0.041 0.004 Non-labour income (logNY) −0.038 0.004 Females w. child <3 years, age 25–50 Males w. child <3 years, age 25–50 Gains to work (logGTW) 0.236 0.023 Gains to work (logGTW) 0.021 0.005 Non-labour income (logNY) −0.078 0.007 Non-labour income (logNY) −0.010 0.002 Females, age 50+ Males, age 50+ Gains to work (logGTW) 0.108 0.010 Gains to work (logGTW) 0.080 0.009 Non-labour income (logNY) −0.036 0.003 Non-labour income (logNY) −0.035 0.003 Note: Probit estimates are computed using full sample, and average marginal effects are evaluated at subgroups. Bootstrapped standard errors, 5000 replications Table 4 Marginal effects: a comparison with neighbouring countries Females Males All individuals Slovakia (authors’calculations) Gains to work (logGTW) 0.12 0.08 0.10 Non-labour income (logNY) −0.04 −0.04 −0.04 Czech Republic (Bicakova et al. 2011) Effective net wage 0.06 0.01 Other income −0.04 −0.01 Czech Republic (Galuscak and Katay 2014) Gains to work 0.27 Non-labour income −0.10 Hungary (Benczur et al. 2014) Gains to work 0.29 Non-labour income −0.30 Note: Average marginal effects (Slovakia); marginal effects at sample means (Czech Republic, Hungary) Senaj et al. IZA Journal of European Labor Studies (2016) 5:19 Page 15 of 26 estimates for the Czech Republic that are substantially smaller in magnitude for males as well as for females. Estimates lie in the (rather wide) range of already published labour supply elasticities. However, this comparison should only be taken as indicative due to the differences in the methodologies used. 6.2 Tax reform scenario simulation Using the SIMTASK microsimulation model and the model of labour force participation decision estimated above, we conducted a policy analysis of static and behavioural effects of two tax system reforms. As a baseline, the tax and transfer system valid in 2012 in Slovakia was taken. We performed a microsimulation of two scenarios. Firstly, we simulated the effects of adopting the legislation valid since January 2013, which includes a marginal departure from the flat tax and results in higher revenues. Secondly, we estimate the impact of a hypothetical abolition of the flat-tax regime with the same simulated fiscal impact as in the first scenario. Although revenue-neutral scenarios are usually analysed in the academic literature, in our setup, we preferred to simulate the reform with the same first-round fiscal effect that is directly comparable with the first scenario. The “2013 scenario”directly assessed the effect of recent changes in Slovak legislation, including the marginal deviation from a flat tax and a significant increase in social security contributions. The two tax brackets of PIT were introduced so that incomes were taxed at the 19% tax rate as before, and an additional 25% rate was applied to those earnings exceeding a threshold value. The higher rate applies approximately to the top 2% of earners. Moreover, this scenario includes a significant increase in the maximum assessment base for social security and health care contributions as well as an increase in the burden for income from agreement contracts. To solely assess the effects of changes in PIT legislation, government transfers and other system parameters that enter the computations in SIMTASK (for example, the minimum subsistence level and the minimum wage) were fixed to the level valid in 2012. The “hypothetical scenario”simulated the effect of reintroducing the tax brackets that were valid before the flat-tax reform in 2004. Five tax brackets with rates of 10, 20, 28, 35, and 38% were defined as in 2003; their thresholds were updated according to the growth of the average nominal wage between 2003 and 2012. As this elementary setup of tax rate regime would result in a decline in tax revenues, a further hypothetical measure needed to be applied to make the fiscal effect of the reform comparable to the 2013 scenario. Specifically, the basic tax allowance is reduced by two thirds. Firstly, we looked at the static or “day-after”effect of the two scenarios. In particular, the change in individual tax burden and in households’disposable income was assessed under the assumption that people do not change their behaviour. The behavioural aspect was analysed afterwards using the estimates of the probit model of the labour force participation decision. Figures 3 and 4 depict the first-round effects of the analysed tax reforms in terms of changes in individual marginal and average effective tax rates. It can be clearly seen that simulated changes affect individuals across the whole income distribution in positive as well as negative ways. The variability arises mainly from various combinations of incomes (labour and non-labour) and the fact that the individual income components Senaj et al. IZA Journal of European Labor Studies (2016) 5:19 Page 16 of 26 might be considered differently in tax liability computations (in particular, the entitlement to apply different tax allowances). In the 2013 scenario, the individuals in the upper tail of the distribution face a positive change in their marginal as well as their average effective tax rates. This is mainly the result of an increase in the maximum assessment base for social security and health care contributions. Individuals with income exceeding the pre-reform values of the maximum assessment base pay higher contributions, which at the same time decreases their tax liability. After the threshold of the new maximum assessment base is reached, both effective tax rates are solely influenced by the newly introduced second tax rate. For most of the earners in the lower part of income distribution, the marginal and average effective tax rates stay unaffected in the 2013 scenario. Effective tax rates increased for those with income from agreement contracts, whose burden was affected by legislation. While before the reform the incomes from agreement contracts 10 were only subject to a 1.05% rate to be paid for social insurance contributions and were taxed at a rate of 19%, since 2013, the regular income from agreement contracts have been burdened at the same rate as employment income. Additionally, due to tightened Fig. 3 Simulated change in individual effective tax rates in scenario 2013 Fig. 4 Simulated change in individual effective tax rates in the hypothetical scenario Senaj et al. IZA Journal of European Labor Studies (2016) 5:19 Page 17 of 26 eligibility conditions for spousal tax allowance valid from 2013, an increase in the tax burden could be observed for affected individuals. On the other hand, a decrease in the tax burden was identified for three small groups affected by the specific change in legislation. In the hypothetical scenario, the individuals in the upper tail of the distribution face a positive change in their marginal and average effective tax rates, which is expected as the tax rates for incomes in the second to fifth tax brackets increased while the decline in the tax allowance additionally led to an increase in their burden. A decline in effective tax rates was expected for low-income earners, being hypothetically taxed at the 10% instead of the 19% rate. However, those with earnings below the new basic tax allowance would have no tax liability after the deduction of the tax allowance, similarly as in the baseline scenario. For those with earnings above the new basic tax allowance, the effect of the lowered allowance prevails over the effect of the lower tax rate, thus leading to an increase in the tax burden. The burden would decrease mainly in those cases when the income is not eligible for a tax allowance deduction, e.g. in the case of working old-age pensioners or people with a prevailing income from capital property. These results can be contrasted to the findings of Krajcir and Odor (2005), who analysed the 2004 Slovak flat-tax reform. They showed that an increase in the non-taxable allowance was an important factor that allowed the tax system to remain moderately progressive and made the reform revenue-neutral, leading to a modest net income decrease for certain groups of workers with below average earnings. Following this, the impact of legislative changes on labour supply behaviour was analysed. Using the SIMTASK microsimulation model, key income variables (gains to work and non-labour income) were computed for the tax and transfers system setup valid in the baseline and in the two scenarios. Given the semi-elasticities estimated by the probit model of participation probability, we can quantify the extent of change by comparing the probabilities of individuals’participation decisions in the baseline and the scenarios. The individual responses to analysed tax regime changes, i.e. changes in the individual participation probabilities, are presented in Fig. 5. Individuals with a higher labour income are less responsive to changes in the tax and welfare system despite the fact that they face an increase both in the METR and AETR. In line with the literature as well as our model estimates, suggesting that extensive margin decisions are taken at the lower end of income distribution, Fig. 5 shows the highest changes in probabilities for those earning less than the average wage. Our approach allows us to compare the impact on participation probabilities for arbitrarily defined population subgroups, thus allowing us to assess to what extent they are affected by the reform. The response in average participation probabilities of specific subgroups in the two scenarios is reported in Table 5. It turns out that the average probability of participation decreases only negligibly (by 0.05 percentage points) in the 2013 scenario. The detailed results suggest that the probability of participation would decrease for almost all of the selected subgroups. Low earners (the first income quintile) and individuals with a lower education responded with the highest magnitudes. This can be explained by pointing out that these individuals often have income from agreement contracts (which after the reform were Senaj et al. IZA Journal of European Labor Studies (2016) 5:19 Page 18 of 26 burdened more and in the same way as income from employment) or no labour income at all. Individuals with either a low or no labour income but with a partner that faced a drop in disposable income could as a family become recipients of the material needs benefit or could have their material needs benefit increased. As a result, this translates into a decrease in participation probability in the behavioural model. A positive reaction to legislation changes was identified among young people (15–24 years old) and females with children under 3 years of age. Both of these groups are specific; individuals frequently do not have a labour income (being students and mothers on parental leave) and their own disposable income did not substantially change due to Fig. 5 Simulated response in participation probabilities Table 5 Simulated response in the average probabilities of participation Baseline participation probability in percent Change of baseline in p.p. Scenario 2013 Hypothetical scenario Population 15–64 68.28 −0.05 −0.11 Age 15–24 39.63 0.03 −0.27 Age 25–50, female 76.26 −0.07 −0.18 Age 25–50, male 89.57 −0.06 −0.09 Age 50+ 60.97 −0.08 0.03 Female with child under 3 years, age 25–50 25.39 0.02 −0.18 Male with child under 3 years, age 25–50 96.40 −0.06 −0.01 Elementary education, age 25–50 66.60 −0.33 −0.27 Secondary education, age 25–50 84.51 −0.03 −0.14 Tertiary education, age 25–50 82.28 −0.10 −0.11 Gross wage quintile—Q1 (below 315 euro), age 25–50 62.31 −0.45 −0.31 Gross wage quintile—Q2 (below 538 euro), age 25–50 85.43 0.13 −0.46 Gross wage quintile—Q3 (below 731 euro), age 25–50 92.33 −0.14 0.00 Gross wage quintile—Q4 (below 1005 euro), age 25–50 94.58 −0.11 0.00 Gross wage quintile—Q5 (above 1005 euro), age 25–50 97.94 −0.03 −0.04 Source: authors’calculations Senaj et al. IZA Journal of European Labor Studies (2016) 5:19 Page 19 of 26 the simulated reform. In the behavioural model, the average participation probability in these groups increased mainly due to the lower disposable income of other working family members. In the hypothetical scenario, the average probability of labour market participation would decrease by 0.11 percentage points. Similarly to the previous scenario, the detailed results show no special pattern among particular subgroups. A positive response was observed for persons over 50 years of age. This is influenced by the fact that oldage pensioners would not be worse off by a lowered tax allowance but would benefit from a significant decrease in the tax rate. On the other hand, young individuals would decrease their labour participation by 0.27 percentage points. The same applies for the level of income earned. Low earners with high values of estimated participation elasticities are more responsive to changes to the tax system than those people with higher earnings. The individuals belonging to the third and fourth quintile would not change their participation decision under the simulated tax system. For policymakers, it is of crucial importance to assess the distributional impacts of every proposed tax change. The fundamental question that may arise here is which subgroups of the population would benefit and which would lose after the reform. Table 6 presents the income inequality measures for the analysed scenarios. For both scenarios, the changes in the disposable income of households are simulated, and as a result, the income inequality measures can be compared in two states (before and after the reform). The impacts of the simulated scenarios differ in terms of income inequality. The Gini coefficient, which is the most commonly used measure of inequality, suggests a slight increase in inequality in the 2013 scenario. The reason is that this tax change would increase the tax wedge for agreement contracts, which are located in the lower deciles of the income distribution and among couples with a lower income. On the other hand, the implementation of the hypothetical scenario would decrease income inequality. This would result from a combination of new tax rates and a substantial decrease in tax allowances. A more detailed insight into inequality impacts is offered by income share ratios (S90/S10, S80/S20, and S60/S40). We documented that the income share of the top 10% of the population compared to the income share of the bottom 10% would increase in both scenarios. The S90/S10 ratio would increase from 4.85 to 5.34 in the 2013 scenario and to 5.20 in the hypothetical scenario. Like the change in the Gini coefficient, the S80/S20 ratio would indicate a decrease in inequality for the hypothetical scenario. The simulated decrease in the S80/S20 ratio for the 2013 scenario is counterintuitive. Having in mind a slight increase in the Gini coefficient, we would Table 6 Impact on income inequality Baseline index Scenario indices Scenario 2013 Hypothetical scenario Gini index 24.52 24.59 24.19 S90/S10 ratio a 4.85 5.34 5.20 S80/S20 ratio 3.60 3.51 3.54 S60/S40 ratio 2.34 2.35 2.32 Source: authors’calculations a Inter-decile income share ratios are the ratios of total income received by the top 10, 20, or 40% of the population to that received by the bottom 10, 20 or 40%. Measures are based on disposable household income and equalized by the modified OECD equivalence scale Senaj et al. IZA Journal of European Labor Studies (2016) 5:19 Page 20 of 26 have expected the opposite effect. Finally, if we compare the top 40% and the bottom 40% of the population, the income shares would remain stable after the implementation of either scenario. 7 Conclusions In this paper, we studied the elasticity with respect to a unique tax reform in Slovakia whereby the flat tax was backtracked and replaced by a progressive tax. We used a labour supply model that takes into account both taxes and transfers to estimate the semi-elasticities of labour force participation decisions. The advantage of this model is in its ability to conduct an ex ante analysis of changes in the tax and welfare system. As an interesting case study, a move away from the flat-tax system which had been valid in Slovakia until 2012 was analysed. In particular, a probit model for labour force participation decisions was estimated, and the results were extensively discussed. This analysis shows several clear results. We identify a significant individual responsiveness to changes in labour and non-labour income. It turns out that the results are qualitatively comparable to those reported for mature market economies as well as for neighbouring countries in the region (the Czech Republic and Hungary): the highly responsive groups of the population are the low-skilled and females. Therefore, labour market policies aimed at boosting economic activity should concentrate on increasing marginal gains to work, especially for loweducated individuals and women. We performed a policy analysis of the first-round and behavioural effects of two scenarios. Both of them simulate the departure from the flat-tax system valid in 2012. However, the setup and the details of the two scenarios differed. The simulations of both scenarios confirmed that the responsiveness of labour supply to legislative changes was marginal for people with high earnings. On the contrary, the highest changes in participation probabilities were faced by individuals with below average earnings. By simulating the two different tax reform scenarios, we demonstrated that the departures from the flat-tax regime with the same fiscal revenue effect have a comparable impact on the average probability of participation. However, the impacts on selected subgroups are different. In the case of real reform in 2013, an increase in tax revenues was accompanied by a slight decrease in the average probability of being economically active by 0.05 percentage points. Individuals with agreement contracts, for whom the tax burden increased significantly, were prominent among the discouraged. In the second scenario, which simulated a hypothetical departure from the flat-tax system by reintroducing five tax brackets together with a significant reduction in the basic tax allowance, labour participation probability was shown to decrease by 0.11 percentage points. The most discouraged groups here would be low earners, individuals with a lower level of education, and women. Finally, we showed that the two simulated scenarios also differed in terms of their consequences for income inequality. Endnotes 1 Reform has been set up as revenue-neutral; see Brook and Leibfritz (2005). Due to data limitations, no quantitative evaluation of this reform has been reported. Senaj et al. IZA Journal of European Labor Studies (2016) 5:19 Page 21 of 26 2 The notation is based on the model presented by Benczur et al. (2014). 3 Let us assume that the utility function is an additively separable CES function considered in the form c1−ψ−1 1−ψþχ1−lðÞ 1−Φ−1 1−Φ. 4 This is also justified in the underlying SK-SILC survey. Less than 2% of respondents in 2012 defined their economic status as “working part time”. 5 In Heckman’s framework, the model consists of two equations: selection and regression. In our implementation, the wage equation contains the degree of urbanization of a region where a person resides (dummy) and regional dummy variables (eight regions). These two variables are intended to capture differences in the regional economic environment and thus present a control for the activity indirectly. In addition, we include human capital characteristics, such as a quadratic form of years of experience and three educational groups. The group of exclusion restrictions consists of characteristics that affect the probability of being employed with the assumption that they have no direct effect on gross wages. These include other forms of income available in the household, the quadratic form of age, and unfavourable health conditions. Controls for family status include dummies like being a parent of a child (younger/older than 3 years of age). We also set a control for having a working partner and being single, married, or divorced. Finally, dummies as a control for working students and pensioners were also included. The estimated coefficients were mostly in line with findings that can be found in the literature; we found a statistically significant effect of selection. The detailed estimation results are available upon request. 6 Our approach to the construction of the GTW variable differs from the setup used by Benczur et al. (2014) and by Galuscak and Katay (2014). They constructed GTW (using the microsimulation tool) for workers and estimate GTW using the Heckman selection model for non-workers. 7 Income semi-elasticity (η) of labour force participation is defined as η¼∂Pr activityðÞ¼1 ∂GTW GTW , implying that the marginal effect of wages on the probability of economic activity can be expressed as MFX ¼∂Pr activity ¼1ðÞ ∂logGTW ¼γ φγlogGTW −ψlogNY þΖ0αðÞ,whereφ(∙) denotes the standard normal density function. The estimated effect should be interpreted so that a 1% rise in gains to work leads to the increase of the probability of supplying labour by 0.01 × MFX. 8 The income elasticity (ε) of labour force participation is defined as ε¼∂Pr activity ¼1ðÞ ∂GTW GTW Pr activity ¼1ðÞ and can be calculated as ε¼η Pr activity ¼1ðÞ , knowing the values of semi-elasticity ηand predicted probability of activity Pr(activity = 1). 9 The income elasticity (ε) of labour force participation is defined as ε¼∂Pr activity¼1ðÞ ∂WW Pr activity ¼1ðÞ and can be calculated as η Pr activity ¼1ðÞ , knowing the values of semi-elasticity ηand predicted probability of activity Pr(activity = 1). 10 Agreement contracts were a popular form of temporary employment contract before the reform in 2013. In our estimation sample (related to the pre-reform period from 2010 to 2012), a higher share of individuals with this type of contract were observed in the subsample of students, among females, and among younger age cohorts. Detailed descriptive statistics is available upon request. Senaj et al. IZA Journal of European Labor Studies (2016) 5:19 Page 22 of 26 Appendix Table 7 List of variables Active Binary indicator that equals 1 if the person is economically active in the income reference period. Employed Binary indicator that equals 1 if the person is employed in the income reference period. Gains to work (logGTW) Variable defined as annual net wage minus the difference between social benefits if not working and social benefits if working. Non-labour income (logNY) Variable defined as a sum of two components, namely non-labour income of all household members (for example, pensions, income from property, parental allowance, unemployment benefit, dividend payments) and labour income of other members of the household. Female Binary variable that equals 1 if the person is woman, 0 if man. Age Variable indicating the person’s age. Years of work experience Variable representing the person’s work experience in years. Education group dummies 3 binary variables are created based on ISCED classification (EDU: primary [reference cat.], EDU: secondary, EDU: tertiary). If the person belongs to a group according to his highest degree awarded, the corresponding binary variable equals 1, otherwise 0. Chronic disease Binary indicator that equals to 1 if the person reports a chronic/long-standing disease. Parent with child under 3 years Binary indicator that equals to 1 if the person is a parent of a child that is younger than 3 years. Parent with child over 3 years Binary indicator that equals to 1 if the person is a parent of a child that is over 3 years old. Student Binary indicator that equals to 1 if the person is a student, 0 otherwise. Pensioner Binary indicator that equals to 1 if the person is a pensioner, 0 otherwise. Working partner Person has a working partner. Married Binary indicator that equals to 1 if the person is married, 0 otherwise. Separated, divorced, or widowed Binary indicator that equals to 1 if the person is separated, divorced, or widowed, 0 otherwise. Degree of urbanization 3 binary variables are created based on the number of inhabitants of the area where the person resides (dense [reference category], average, sparse). If the person belongs to a group according to the degree of urbanization of his residence, the corresponding dummy variable equals 1, otherwise 0. Regional dummies 8 binary variables are created based on NUTS3 classification (REG: Bratislava [reference cat.], REG: Trnava, REG: Trencin, REG: Nitra, REG: Zilina, REG: Banska Bystrica, REG: Presov, REG: Kosice). If the person belongs to a group, the corresponding binary variable equals 1, otherwise 0. Senaj et al. IZA Journal of European Labor Studies (2016) 5:19 Page 23 of 26 Table 8 Descriptive statistics of the estimation subsample and original SK-SILC 2010–2012 Dataset Subsample for estimation SK-SILC 2010–2012 Variable Mean Std. dev. Mean Std. dev. Active 0.6 0.5 0.5 0.5 Employed 0.6 0.5 0.5 0.5 Gains to work (in euros, monthly) 519.2 168.6 437.6 237.4 Log of gains to work 6.2 0.3 6.1 0.5 Non-labour income (in euros, monthly) 1110.3 632.1 1133.4 633.9 Log of non-labour income 6.8 0.6 6.9 0.6 Male 0.5 0.5 0.5 0.5 Female 0.5 0.5 0.5 0.5 Education: primary 0.1 0.4 0.3 0.4 Education: secondary 0.7 0.5 0.6 0.5 Education: tertiary 0.2 0.4 0.2 0.4 Age 42.4 16.5 39.9 21.0 Years of experience 19.0 14.8 17.4 15.5 Chronic disease 0.3 0.4 0.3 0.4 Parent with child under 3 years 0.1 0.2 0.0 0.2 Parent with child over 3 years 0.3 0.5 0.2 0.4 Pensioner 0.2 0.4 0.2 0.4 Student 0.1 0.4 0.1 0.3 Working partner 0.4 0.5 0.3 0.5 Family: married 0.5 0.5 0.4 0.5 Family: separated, divorced, or widowed 0.1 0.3 0.1 0.3 Density: dense 0.2 0.4 0.2 0.4 Density: average 0.3 0.5 0.3 0.5 Density: sparse 0.4 0.5 0.5 0.5 Region: Bratislava 0.1 0.3 0.1 0.3 Region: Trnava 0.1 0.3 0.1 0.3 Region: Trencin 0.1 0.3 0.1 0.3 Region: Nitra 0.1 0.3 0.1 0.3 Region: Zilina 0.1 0.3 0.1 0.3 Region: Banska Bystrica 0.1 0.3 0.1 0.3 Region: Presov 0.2 0.4 0.2 0.4 Region: Kosice 0.1 0.4 0.1 0.4 Sample size 37,960 46,191 Senaj et al. IZA Journal of European Labor Studies (2016) 5:19 Page 24 of 26