Work Incentives in Europe

Transcription

1 Work Incentives in Europe Work in progress Do not quote nor circulate! Charlotte Bartels Cortnie Shupe February 15, 2016 Abstract Many advanced economies implemented labor market reforms over the past decades in order to encourage higher labor force participation rates. In this paper, we measure work incentives inherent in tax-benefit systems by computing Participation Tax Rates (PTRs) across the EU and subsequently identify the impact of these disperse PTRs on the probability of job take-up. We find that the PTR-level is negatively associated with the probability of being employed, i.e., higher work incentives increase the participation probability. The effect is larger for women than men. However, it remains to be estimated (as soon as we have PTRs for several years and for more countries), how a change of the PTR affects the change in labor market status. Following this last step, we will estimate participation elasticities by gender. JEL Classification: H24, H31, J22, J65 Keywords: tax-benefit-systems, work incentives, Europe Charlotte Bartels (cbartels@diw.de) and Cortnie Shupe (cshupe@diw.de) are affiliated with DIW. This paper uses the EU-SILC database which is made available by EUROSTAT.

2 1 Introduction Within the European Union (EU), most countries share the commonality of large gaps between labor market participation and unemployment rates of the high and low-skilled. Creating work incentives for the latter proves increasingly important, as the percentage of non-working individuals with low educational attainment continues to rise and the most recent financial crisis exacerbated this trend. 1 A wide range of empirical studies showing that low-income and low-skilled individuals are particularly responsive in their decision whether or not to take up work further stress the importance of work incentives inherent in the tax and transfer system. 2 As a consequence, tax-benefit system miscalculations at the extensive margin create particularly high efficiency costs. Additionally, the costs of low-skilled inactivity include expenses on out-of-work benefits, foregone taxes and social security contributions, poorer health, well-being losses and failures in social inclusion. Macroeconomic studies using aggregate data have claimed that cross-country differences in income tax rate changes can largely explain the vast differences in labor supply across the OECD (see for example Ohanian et al. 2008; Prescott 2004). However, macro approaches using aggregate data have not been able to disentangle the impact of differences in policies from institutional and cultural differences across countries, rendering the identification of labor supply determinants that are necessary for policy recommendations difficult. In contrast, microeconomic studies emphasize the importance of heterogeneity of elasticities across different groups, taking into consideration skill levels, gender, marital status and the presence of small children. This paper takes a microeconometric approach to identifying the causal impact of tax-benefit system s inherent work incentives for job take up across the EU, controlling for country fixed effects such as cultural norms or tastes for work and leisure. We measure work incentives at the extensive margin by computing Partic- 1 The percentage of non-working, low-skilled individuals increased from 39.5 percent in 1995 to over 50 percent in 2015 (OECD 2005, 2015; ECB 2013) 2 Research on labor supply elasticities finds that behavioral responses are higher on the extensive margin than on the intensive margin, particularly for low-income and low-skilled individuals. For overviews, see Meghir and Phillips (2010) or Chetty et al. (2013). 1

3 ipation Tax Rates (PTRs) using the microsimulation model EUROMOD thereby taking into account the institutional variation across EU countries. We place a specific focus on the low-skill sectors and provide a decomposition of which aspects of the PTR are driving labor supply disincentives. Both the growing theoretical literature and the empirical results on the size of the response at the extensive margin triggered a number of studies estimating PTRs. EU cross-country studies based on the tax-benefit simulation model EUROMOD for the year 1998 are Immervoll et al. (2007), Immervoll et al. (2009) and O Donoghue (2011). Kalísková (2015) uses EUROMOD of the years to estimate PTRs for women. Country studies on PTRs are, e.g., Dockery et al. (2011) for Australia, Adam et al. (2006) and Brewer et al. (2008) for UK, Pirttillä and Selin (2011) for Sweden as well as Bartels and Pestel (2015) for Germany. PTR estimates can be used to obtain an empirical measure of an extensive elasticity. To the knowledge of the authors, only two recent studies estimate the effect of the PTR as a work incentive measure on the participation probability. They find that a 10 %-point decrease of the PTR raises the participation probability by 2 %-points (Kalísková; 2015) or %-points (Bartels and Pestel; 2015), respectively. According to Kalísková (2015) her estimate implies an extensive elasticity for women of which is rather low compared to other empirical results. Our contribution is threefold. Firstly, we offer empirical evidence for the relevance and exogeneity of using lags of the PTR to instrument for tax-benefit incentives and their effect on the labor supply of the long-term unemployed. Secondly, we identify the driving components of these incentives, distentangling the effects of taxes, benefits and social security contributions. Thirdly, we provide cross-country estimates for extensive labor supply elasticities. 2

4 2 Methodology 2.1 Measuring Participation Tax Rates The PTR represents a measure of work disincentives from optimal tax theory literature and is based on the assumption that an individual i faces a binary choice between the two labor market states E employed or U unemployed. The PTR measures the change in household net taxes from labor market state E to U as a proportion of individual earnings in labor market state E. 3 Net taxes T paid by the household h are income taxes t h including social security contributions reduced by benefits b h. An annual PTR is denoted as where y E h is gross household income, T (ye h P T R ih = T (ye h ) T (yu h ), (1) y E,w i ) is household net taxes and ye,w i is individual labor earnings if the individual is in labor market state E. Gross household income can be calculated as the sum of labor earnings, asset income, private transfers, private pensions and social security pensions of all household members. y U h gross household income and T (yh U ) is household net taxes if the individual is in labor market state U when individual labor earnings are zero. If household net taxes are equal for both labor market states, then the PTR is zero and incentives to take up work are not distorted. is In reality, however, a welfare state providing income support in state U usually leads to t U h < bu h resulting in T (yh U ) < 0 as unemployment benefits will surpass taxes paid for the declined household income yh U. In sum, the change in net taxes will be positive in presence of a welfare state and the PTR will be higher than zero for most individuals. The higher the PTR, the more generous income support programs reduce the financial gain from working. The PTR will equal one if the change in net taxes T (y E h ) T (yu h ) (numerator) is equal to individual earnings y E,w i (denominator). In this case, no financial gain arises from working. Finally, if out-of-work income support exceeds earnings, then the PTR can be even greater than one. 3 Specifically, state E in this study will refer to working and U indicates not working, which includes inactivity. 3

5 In order to obtain a PTR for all individuals in the labor force independent of their observed labor market status E or U, we simulate the non-observed state. For this simulation, we make the assumption that a change in one partner s labor supply behavior, i.e., giving up or taking up a job, neither triggers a compensating labor supply reaction by other household members nor changes in household income from other non-labor sources. This assumption reflects standard procedure in the PTR literature (see, e.g., Immervoll et al.; 2007). For those observed in U, we estimate individual earnings ŷ E,w i compute y E h = yu h + ŷe,w i using a standard Heckman regression (Heckman; 1979) and accordingly. Following the calculation of household gross income described above, we then apply the tax-benefit rules of the respective year to obtain household taxes t h and public transfers b h for both states E and U in such a way that ensures consistent assumptions regarding deductions and other special tax and transfer rules. example, household taxes paid in state U are the sum of income tax t U,inc h on the basis of y U h and social security contributions su j on spouse s earnings y E,w j For assessed if the spouse j is working in E. Household public transfers are the sum of unemployment benefits, unemployment assistance, maternity benefits, social assistance, housing allowances and child benefits. A potential increase in benefits when changing from E to U will mostly occur for unemployment benefits, unemployment assistance, social assistance and housing allowances. All simulations are based on EUROMOD version G2.0+, which is a microsimulation model developed and maintained by country-specific institutions in 27 European countries. See Sutherland and Figari 2013 for details 2.2 Estimation Strategy In our regression analysis, we test the extent to which lower PTRs are associated with an increased likelihood of working or taking up work, respectively. First, we test if high work incentives, i.e. low PTRs, contribute to raise probability of labor market participation (E). The binary outcome variable is one if individual i is 4

6 employed in period t (E it ). We estimate the following regression model: P (E it ) = γp T R it + X itβ + µ t + ɛ it (2) The coefficient γ captures the effect of a PTR on the likelihood of labor market participation. Controls are captured by X it and include age, education, experience, health, industry and occupation, household type, number of children as well as the presence of a child under 6 years. Year fixed effects capture business cycle fluctuations affecting labor demand and are denoted by µ t. The error term is denoted by ɛ it. Second, we test if increased work incentives, i.e. reduced PTRs, contribute to raise the likelihood to take up work (U E). The binary outcome variable is one if individual i switches from long-term non-participation in period t 2 (U it 2 ) to participation in period t (E it ). The main explanatory variable of interest is the PTR-change between period t 1 and t, i.e., P T R it = P T R it P T R it 2. We estimate the following regression model: P (U it 2 E it ) = γ P T R it + X itβ + α i + µ t + ɛ it (3) The coefficient γ captures the effect of a PTR-change on the likelihood of taking up work and is expected to be negative, i.e., a decrease (increase) is associated with a higher (lower) likelihood of labor market participation. We estimate both equations with ordinary least squares (OLS) and logit. The second equation is estimated in an individual fixed effects (FE) framework exploiting individual variation around an individual time-invariant fixed effect denoted by α i, which captures unobserved heterogeneity, such as preferences for leisure or innate ability affecting the employment status. 5

7 3 Data For our regression analysis, we draw on EU-SILC longitudinal data from , which provide ex-post harmonized and comparative household-level statistics on labor and income variables across the European Union. The annual survey covers a representative sample of private households, retaining each household for a maximum of 4 years on a rotational basis. The panel component of EU-SILC permits the observation of transitions into and out of the labor market and many other individual- and household-level changes over time for 23 EU countries. 4 Currently, the EUROMOD microsimulator runs exclusively with the input of EU-SILC crosssectional data. For most member states, the cross-sectional and panel data files share an identical household base. Nevertheless, due to data protection concerns in some member states, it is not possible to directly link and identify the households in the cross-section with those in the panel dataset. Therefore, we calculate PTRs using the cross-sectional data and apply propensity score matchig to link these observations to their nearest neighbor in the panel dataset for each country and each year. Our sample includes individuals in their prime working age, between 25 and 54 years old. We restrict the sample to these ages because individuals younger than 25 likely face a decision between education and work rather than between employment and inactivity, which presents the focus of this paper. Likewise, beginning at age 55, individuals in many countries may choose between (early) retirement and employment rather than employment and inactivity. Furthermore, we exclude students, pensioners, permanently disabled persons and those in compulsory military service. Our final sample consists of approximately 424,000 individuals in 260,000 households over four years, whereby the number of observations varies greatly between years as a consequence of the rotational design of the EU-SILC longitudinal 4 Countries include: Austria (AT), Belgium (BE), Bulgaria (BG), Cyprus (CY), Czech Republic (CZ), Denmark (DK), Estonia (EE), Spain (ES), Finland (FI), France (FR), Hungary (HU), Ireland (IE), Italy (IT), Lithuania (LT), Luxembourg (LU), Latvia (LV), Malta (MT), Netherlands (NL), Poland (PL), Portugal (PT), Slovenia (SI), Slovakia (SK) and United Kingdom (UK) and in the following will be referred to by the aforementioned abbreviations. Excluded countries did not submit panel data to the EU-SILC project. 6

8 data Descriptive Statistics Table 1 depicts labor market participation rates across the 23 EU member states for which EU-SILC longitudinal data exist. 6 Participation rates vary substantially between countries from roughly 65 percent of the working-age population to over 95 percent for both men and women. This study will investigate the role of the divergent tax- and transfer systems in explaining these differences. Figure 1 demonstrates that transition rates for the long-term unemployed and inactive individuals are particularly low in all countries examined. 7 However, some variation does exist even in this relatively immobile section of the population, both within and across countries, and we exploit this variation in our analysis. 5 Each year 25 percent of the sample rotates out of the dataset and is replaced by new observations, each of which is then observed for 4 years, absent attrition. Additionally, the number of participating countries and their sample size has increased over the years. 6 The table does not distinguish between full- and part-time work, but rather considers whether the individual s main current activity is gainful employment regardless of the number of hours worked. 7 Ireland in 2011 and 2012 is an outlier in this respect, with a transition rate above 12 percent, and was excluded in the depiction. Its rate drops to 3 percent again in The financial crisis that hit this country particularly hard in the years immediately prior to this period likely explains this fluctuation, as unemployment numbers and durations temporarily increased. 7

9 Figure 1: Transition Rates from Unemployment to Employment Transition Rate (UE) AT BE BG CY CZ DK EE ES FI FR HU IS IT LT LU LV MT NL PL PT SE SI SK UK Graphs by country Year Note: Transition is defined as a change in individual employment status from long-term inactivity (at least one year) in t 2to predominantly employed (at least 6 months) in the reference year t. The sample includes individuals aged 25-54, excluding students, pensioners, the permanently disabled and those in compulsory military service. Rates describe weighted means per country. Source: EU SILC longitudinal data (2011,2013), own calculations. 8

10 Table 1: Labor market participation by country Male Female Male Female Male Female Male Female Male Female Male Female AT BE BG CY CZ DK EE ES FI FR HU IE IS IT LT LU LV MT NL PL PT SI SK UK Note: The sample includes individuals aged 25-54, excluding students, pensioners, the permanently disabled and those in compulsory military service. Rates describe weighted means per country. Source: EU SILC longitudinal data (2011,2013), own calculations. 4 Outlook to results Table 2 shows average PTRs by gender and country that we have estimated so far. In most countries, the average PTR is between 30% and 40%. Women have substantially or slightly lower PTRs than men. This can be explained by two factors: First, women tend to earn less than men such that the income tax wedge between employment and unemployment is lower than for men rendering a lower PTR. Second, women more frequently interrupt their careers for parental leave and often work in the low income sector where they gain less potential claims on out-of-work benefits than men. Regression results for Equation 2.2 are presented in Table 3. We find that the 9

11 Table 2: Participation tax rate by country Male Female Male Female Mean SD Mean SD Mean SD Mean SD AT BE BG CZ Note: The sample includes prime working-aged individuals aged 25-54, excluding students, pensioners, the permanently disabled and those in compulsory military service. Later versions of this paper will include years Source: EU SILC data ( ) and EUROMOD, own calculations. PTR has a negative and statistically significant effect on the probability of being employed. The size of the effect is larger for women than for men. Overall, belonging to an older age group, being low-skilled, and suffering from bad health all decreases the probability to be employed, whereas experience increases the probability. Having children and being married has a positive effect for men, but a negative effect for women. However, individual fixed effects such as taste for work is not controlled for in this framework. However, so far the results include only data for 2010 in five countries (AT, BE, BG, CZ, EE) for which we have calculated PTRs so far. Results for Equation 2.2 will follow as soon as PTRs are calculated for several years which will enable us to identify how changes in the PTRs are associated with changes in the labor market status. 10

12 Table 3: Regression results Males Females OLS LOGIT OLS LOGIT (1) (2) (3) (4) PTR Age Age Low-skilled Experience Health Married Child in hh cons r2 a r2 w r2 p N t statistics in parentheses p < 0.05, p < 0.01, p < Note: The sample includes prime working-aged individuals aged 25-54, excluding students, pensioners, the permanently disabled and those in compulsory military service. Source: EU SILC data 2008 and 2010 and EUROMOD, own calculations. 5 Conclusion In this paper, we computed Participation Tax Rates (PTRs) across the EU as a measure for work incentives inherent in tax-benefit systems using EU-SILC data and the microsimulation model EUROMOD. In most European countries, the PTR is around 30% (to be continued). We then identified the impact of these disperse PTRs on labor supply and estimated extensive elasticities. We find that the level of the PTR is negatively associated with the probability of being employed and the effect is larger for women. However, it remains to be estimated (as soon as we have PTRs for several years and for more countries), how a change of the PTR affects the change in labor market status accounting also for individual fixed effects. For this framework, (Bartels and Pestel; 2015) find for Germany that a 10 %-point decrease of the PTR raises the participation probability by %-points. 11

13 References Adam, S., Brewer, M. and Shephard, A. (2006). The poverty trade-off. Work incentives and income redistribution in Britain, Policy Press. Bartels, C. and Pestel, N. (2015). The impact of short- and long-term participation tax rates on labor supply, IZA DP No Brewer, M., Saez, E. and Shephard, A. (2008). Means-testing and tax rates on earnings, Prepared for the Report of a Commission on Reforming the Tax System for the 21st Century, Chared by Sir James Mirrlees, Institute for Fiscal Studies. Chetty, R., Guren, A., Manoli, D. and Weber, A. (2013). Does Indivisible Labor Explain the Difference between Micro and Macro Elasticities? A Meta-Analysis of Extensive Margin Elasticities, NBER Macroeconomics Annual : Dockery, A., Ong, R. and Wood, G. (2011). Welfare Traps in Australia: Do they bite?, CLMR Discussion Papier Series No. 08/02. Heckman, J. (1979). Sample Selection Bias as a Specification Error, Econometrica 47: Immervoll, H., Kleven, H., Kreiner, C. and Saez, E. (2007). Welfare reform in European countries: a microsimulation analysis, The Economic Journal 117: Immervoll, H., Kleven, H., Kreiner, C. and Verdelin, N. (2009). An Evaluation of the Tax-Transfer Treatment of Married Couples in European Countries, IZA- Discussion Paper No Kalísková, K. (2015). Tax and transfer policies and the female labor supply in the eu, IZA DP No Meghir, C. and Phillips, D. (2010). Labor Supply and Taxes, Chapter 3 for Mirrlees Review (2009), in J. Mirrlees, S. Adam, T. Besley, R. Blundell, S. Bond, R. Chote, M. Gammie, P. Johnson, G. Myles and J. Poterba (eds), Dimensions of Tax Design: the Mirrlees Review, Oxford University Press. 12

14 O Donoghue, C. (2011). Do tax-benefit systems cause high replacement rates? A decompositional analysis using EUROMOD, LABOUR 25: Pirttillä, J. and Selin, H. (2011). Tax Policy and Employment: How does the Swedish system fare, CESifo Working Paper Series No