Do Social Insurance Taxes Hinder Entrepreneurial Activity? *

Transcription

1 Do Social Insurance Taxes Hinder Entrepreneurial Activity? * Youssef Benzarti, Jarkko Harju and Tuomas Matikka February 14, 2018 Abstract This paper estimates the eect of social insurance taxes on entrepreneurial activity. We use a unique institutional setting in Finland that allows us to leverage exogenous variation in the compulsory social security taxes of entrepreneurs, along with rich rm and individual level administrative data for all entrepreneurs. We nd that entrepreneurs substantially reduce their contributions to the social insurance scheme when they can more freely decide the level of contributions. We observe no eects on the rm-level business activity measures of the aected entrepreneurs but they seem to receive lower total income from the rm after the reform. Keywords: Social security, entrepreneurs, private savings, economic activity JEL classication codes: H25, H32, H55 * First draft. All errors are our own. Benzarti: UCLA and NBER, benzarti@econ.ucla.edu; Harju: VATT Institute for Economic Research (Helsinki, Finland) and CESifo, jarkko.harju@vatt.; Matikka: VATT and CESifo, tuomas.matikka@vatt..

2 1 Introduction Sometimes dubbed the engine of growth, entrepreneurship plays a central role in modern economies. In the US, for example, new businesses account for 20 percent of total gross job creation (Decker et al. 2014) While entrepreneurs can be very successful and accumulate massive amounts of wealth, entrepreneurship remains one of the most risky lines of activity and can result in large wealth losses (see, for example, Hall and Woodward [2010]). For this reason, mandating retirement savings for this population can prove out to be a rst order welfare improvement: without a retirement safety net, entrepreneurs face substantial old age risk. However, the marginal value of resources for entrepreneurs can be substantial given how cash-constrained they often are. Mandating retirement savings, while reducing old age risk, could signicantly aect entrepreneurial activity. More broadly, the eect of taxes on entrepreneurial activity is a central question as it matters for both tax policy and economic growth. 1 Yet, there is limited evidence on how taxes aect entrepreneurial activity. This is mainly due to the fact that tax variation that aects entrepreneurs in a an exogenous way is scarce. Instead research has focused on cross-country comparisons of the level of taxes and entrepreneurial activity and found negative correlations between these two variables. 2 While informative at the macro level, the question of whether taxes reduce entrepreneurial activity remains open. In this paper, we exploit a unique setting in Finland which allows us to use exogenous variation in the amount of social security taxes entrepreneurs have to pay, along with rich rm and individual level administrative data on the full population of Finnish entrepreneurs, to causally estimate the eect of social security taxes on entrepreneurial activity. While the social insurance tax rates are the same across all rm owners, the base over which the tax rate applies is not. Contributing into social insurance is optional above a certain threshold of ownership share: if an entrepreneur owns more than 50% (30% in later years) of the rm she is operating, she has full discretion over how much social security tax to pay. Below the ownership share threshold, the entrepreneur cannot choose her contribution rate. We use a standard dierences-in-dierences strategy and exploit the change in the ownership share to assess how removing the social security contribution mandate aects entrepreneurial activity by comparing entrepreneurs with 30-50% pre-reform ownership share (treatment) to entrepreneurs with 50-70% pre-reform ownership share (control). We have three main ndings. First, as entrepreneurs cross the ownership share threshold, they substantially reduce their contribution to the social insurance scheme, implying that the social insurance mandate is binding for entrepreneurs below the threshold. Second, we can track what entrepreneurs do with the money saved. We nd some evidence that treated owners use the money saved in lower social security contributions 1 See for example Schumpeter [2013] and more recently Baumol et al. [2007]on the importance of entrepreneurship for growth. 2 See for example Djankov et al. [2010]who nd strong negative relationships between the level of corporate income taxes and entrepreneurship using data from 85 countries. 1

3 to increase long-term investments in their rm. Third, we do not detect any changes in business activity measured by total output of rms or net prots. However, we nd a rather puzzling result that the total income from their rm decrease among treated entrepreneurs. This change is mostly explained by a decrease in wages rather than dividends. This paper is the rst to show that social contributions and, more generally, taxes aect entrepreneurial activity, using a quasi experimental design and micro level data. It contributes to a public nance literature that analyzes the behavioral responses of entrepreneurs to tax incentives. Cullen and Gordon [2007] for example, use time series variation in tax rates to estimate the eect taxes have on risk taking by entrepreneurs. Gentry and Hubbard [2000] use a discrete choice model to estimate the eect of tax progressivity on entrepreneurial entry and nd negative eects. This paper is also related to a corporate nance literature that estimates the effect of regulation and government intervention on entrepreneurial activity. Hombert et al. [2014] for example estimate the eect of unemployment insurance provision on entrepreneurial entry. Other papers use cross country comparisons to assess the eect of regulations on entry, such as Desai et al. [2005] or Klapper et al. [2006]. Our ndings are also related to a literature that estimates the eect of taxes on business activity. While that literature has shown compelling evidence of these eects, there are no papers that specically focus on entrepreneurs. Yagan [2015]for example, nds no eect of the 2003 US dividends tax cut on investment, while we nd that investment increases as social insurance contribution decrease. In addition, our results are related to the elasticity of taxable income (ETI) literature where the largest elasticities are commonly found for entrepreneurs and self employed [Saez et al. 2012]. We contribute to this literature by showing that real entrepreneurial activity is unlikely to be distorted by taxes, suggesting that the main margin of responses for entrepreneurs in the ETI literature is evasion and/or avoidance. Finally, this paper is related to a literature that estimates the crowding out of private savings by government mandated and subsidized saving plans. Bernheim [2002] summarizes older research on the topic and concludes that it is inconclusive. More recently, Chetty et al. [2014] makes use of rich data and a compelling design and nds passive investors tend to invest more when nudged into it, while investments by active individuals tend to be crowded out by subsidies. In this paper, we consider a population that is unlikely to be passive and who has access to a wide range of saving opportunities and yet, we still nd that crowding is small and mostly reduces long term investment. This suggests that even if passive savers are made active, subsidized or mandated savings are unlikely to lead to large crowding out eects. This paper is structured as follows. Section 2 describes the institutions, and section 3 explains the methods and data. The results are presented in section 4, and section 5 concludes the paper. 2

4 2 Institutions The Self-employed Persons' Pension Act (YEL 3 ) sets the basis for the pension and social security of entrepreneurs in Finland. The YEL insurance applies to all selfemployed workers whose work does not take place within the scope of an employment or service contract. On top of this requirement, the self-employed individual must meet the following conditions: he/she has to be years old, business operations must have been ongoing for at least four months, and the estimated earnings from work amount to the rm have to be at least 7, per year (in 2016). 4 The YEL insurance also applies to all partners of limited partnerships, and to the owners of limited companies who own, alone or together with family members, 50% of the company. In addition, the owners who hold a leading position in limited companies and own over 30% of the company's shares are considered to be self-employed. These above conditions are mandatory and entrepreneurs cannot opt out from the YEL insurance system. If the above conditions are not met, the entrepreneurs are automatically subject to the TyEL insurance system 5. The TyEL insurance contribution amount is calculated on the basis of the employee's earned income, and it is nanced with the employer's and the employee's TyEL contributions. The employer has to to pay the full TyEL contribution amount, and the employee's contribution is withheld directly from his/her earnings. In this paper, we focus on the behavior of the owners of limited companies. We restrict our analysis to these owners because they fact exogenous variation in the amount of contributions which allow us to generate our main estimates. Hereafter, we refer to all entrepreneurs to which YEL rules are applied as Y owners and all other entrepreneurs that do not fulll these requirements as T owners. The contribution rates of the Y and T owners are very similar, e.g. 23.6% and 23.7% in 2016, respectively. However, Y owners can choose the base of the contribution between a minimum and a maximum amounts ( 7, ,625 in 2016), therefore eectively choosing the amount of social insurance they want to pay. 6 Therefore, the Y owners can use much more discretion in deciding the level of social security compared to the T owners. 7 For T owners instead, the earned income level and the contribution 3 Yrittäjän eläkelaki in Finnish. 4 The Self-employed Persons' Pension Act (HE 1272/2006), is available online here (in Finnish): More information in English can be found here: 5 This statutory earnings-related pension scheme is set by the Employees Pensions Act that is Työntekijän eläkelaki in Finnish. 6 Both the YEL and TyEL contribution rates, and the minimum and maximum YEL income levels vary over time and are usually determined annually. The TyEL contribution rates are shared by the employer and employee such that the employer pays larger share, approximately 18% in 2016, and the rest is paid by employees, 5.7% if the employee is younger than 53 years old and 7.2% for those 53 years old and older. In addition, the YEL contribution rate depends on the age of an entrepreneur, at the age of 5362 the contribution rate is higher, 25.1% (in 2016), compared to the normal rate of 23.6% for other Y owners. Also, the contribution rates are lower for new Y owners and the rates vary by the age, being 19.6% for 5362 years old and 18.4% for others. 7 Although, according to the law, the YEL income should reect a wage that would be paid if the work had been carried out by another equally competent person. But this provision is not binding as 3

5 rate denes their annual contributions. The estimated YEL income also directly aects other social security benets provided by the Social Insurance Institution of Finland (Kela). These are, for example, sickness allowance, maternity, paternity and parental allowances. A further requirement for receiving an unemployment allowance (from an unemployment fund or from the Kela) is that the Y owner has estimated a minimum of 12,420 YEL income per year (in 2016). This creates an additional threshold to the YEL income schedule. Two important details are important to point out. First, the YEL insurance accrues pensions with the same annual rate, provides disability and unemployment insurances, and aects benets, similarly as the TyEL insurance. Therefore, the insurance schemes are very similar to each other but Y owners have more discretion to decide the level of social security compared to T owners. Second, the insurance status does not aect the taxation of income from the rm, i.e. dividends and wages, and thus, there is no dierence in tax rates between Y and T owners with the same level of total income (wages+dividends). Ownership share threshold and the reform of Before 2011, the owners of limited companies holding leading positions and owning over 50% of the company shares, were considered to be self-employed. In 2011 this threshold was decreased to the current level of 30%. 8 However, the change in legislation included a transition period of 3 years, meaning that those entrepreneurs who own a limited company with 30-50% ownership share could retain T insurance status until the end of 2013, if wanted. Otherwise, the institutional setting remained similar over time, except for some small adjustments in contribution rates and YEL income thresholds that are typical annual changes for both insurance types during the time period we are interested in. 3 Methods and Data 3.1 Methods The institutional features of our setting oer a unique opportunity to study how the threshold aects the amount of social security contributions of entrepreneurs. More interestingly, we can use these details as an instrument for other outcomes, and thus estimate the eects of the social security on the economic activity of rms. The ownership threshold itself provides a possibility to use a regression discontinuity design (RDD) where the running variable is the share of how much the owner of a limited company owns of the rm, i.e. ownership share. The comparison of entrepreneurs would be by the ownership share threshold, 50% pre-2011 and 30% post-2011, where the identifying assumption is that entrepreneurs close to the threshold are otherwise similar but the only dierence is the insurance status. If credibly argued, this strategy it is hard to implement in practice. 8 More information about the reform (HE 135/2010) can be found here (only in Finnish): 4

6 could oer causal eects of the threshold. However, a challenge in this strategy is that entrepreneurs can aect the running variable, ownership share. This is especially possible locally very close to the threshold, and thus we do not focus on using this method. Our preferred method is the dierence-in-dierences strategy. The estimated equation is: Y i.t = α 0 + α 1 T reat i + α 2 P ost t + α 3 (T reat i P ost t ) + ε i,t (1) where Y is the outcome variable, i refers to entrepreneur/owner of a limited company, t is time, T reat is a dummy variable being 1 for the treated entrepreneurs and 0 otherwise, P ost refers to the period after the reform (2011->) and ε is the error term. In the following analysis we use several outcome variables. The rst outcome is the level of mandatory social security contributions. We also estimate the eects on several individual- and rm-level outcome variables. In particular, we investigate how the owner-level social security contributions aect the economic activity of rms using the overall annual output of rms as a measure for economic activity. Other owner-level outcome variables are total income, wages and dividend from the rm to the owner, and other rm-level outcomes are total assets, liabilities, equity and long-term investments in the rm. We dene all owners of limited companies that had ownership shares between 3150% before the reform ( ) to the treatment group. These owners had to change their insurance status from being T owners to Y owners. This was true in 2014 onward due to the three year transition period mentioned earlier. To oer causal evidence, we select a comparison group of owners that would constitute a credible counterfactual for the treatment group, essentially representing the behavior of the treatment group absent any reform. We select owners of limited companies with pre-reform ownership shares between 51-70% to represent the counterfactual as their insurance status was unaected by the reform. We provide graphical evidence that this is an appropriate control group by showing that trends are parallel before the reform. 3.2 Data We use data from two dierent sources. First, we have information about the social security contributions for both Y and T owners. For T owners all we need is the level of annual earnings and the age of an entrepreneur, both directly observable from tax data, to be able to back out the level of contributions. However, for Y owners this is not possible as their contributions depend on their self-reported YEL income that do not necessarily correlate with their true annual earnings. A further complication is that there is no individual-level register available in Finland that collects all information about the estimated YEL income levels by owners. Therefore, we have secured access to individual-level YEL income from two major Finnish employment pension companies. These data have never been used in prior research. Our coverage from these companies 5

7 exceeds 70% of all YEL insured entrepreneurs. The second data source is the owner- and rm-level tax register data from the Finnish Tax Administration. These data include information on the nancial statements and tax records of all Finnish businesses and owners. The data include tax record information on both the rm and its main owner. One unique feature of the data is that we can link the owner-level and the rm-level data together. As mentioned already earlier, we focus only on the Y and T owners of limited companies in this paper. Thus, we exclude all other entrepreneurs from the sample. We use the data for the years The owner-level data set contains important tax information for our analysis, for example, wages and dividends paid to the owner by the rm, and taxable income earned from other sources, detailed information about capital income from dierent sources (e.g. dividends from listed rms), investment fund shares (stocks+prots separately) and voluntary pension savings. In addition to personal income and savings, the owners can use rm accounts to accumulate pension savings or to invest in their rm as a response to the changes in their social security contributions. Our rm-level data contain information on dierent assets categories such as long-term investments, shares and dividend income and real-estate investments. The ownership share is not readily available in the data set, and thus to be able to dene the treatment and control groups we have to calculate it. Our owner-level data include information on the number of shares of the company the entrepreneur owns and the rm-level data contain the number of total shares in the company. Using these information we can calculate the ownership share for each owner and use it to divide owners into treatment and control groups. One potential issue in measuring the ownership share, essential for our analysis, is that rms might have shares of dierent value. However, this is rather unlikely for relatively small rms that are in the focus of this paper. In addition, we use an alternative measure for ownership share to check the robustness of our results by using the sum of total dividends received from the rm and the total payout of dividends by the rm to calculate ownership share. Table 1 shows the pre-reform descriptive statistics for our sample by treatment status. The treated owners contribute more to the pension fund before the reform. The average annual dierence in pension contributions is almost 2,000 euros, that is over one-third larger payments among treated owners compared to control group. However, there is no clear dierence in the average total income from the rm to the owner, approximately 55,000 euros per year for both owner types. Also, the rm-level characteristics are similar across treatment status. These statistics give us a rst piece of support that the treatement and control groups are comparable. 6

8 Owner-level variables Pension contrib. Total income Wages Dividends % of males Age Treatment: T owners (3150% ownership share pre-2011) Mean Median N Control: Y owners (5170% ownership share pre-2011) Mean Median N Firm-level variables Turnover Total assets Total liabilities Equity No. of empl. Treatment: T owners (3150% ownership share pre-2011) Mean Median N Control: Y owners (5170% ownership share pre-2011) Mean Median N Notes: Table presents the descriptive statistics. The sample includes only rms belonging to the treatment or control group fro time period Table 1: Descriptive statistics: by treatment status 4 Results 4.1 Graphical evidence Social security contributions. We rst focus on showing changes in (annual) social security contributions graphically. Figure 1 shows the average changes in the log public insurance payments relative to year 2010, one year before the reform. The Figure illustrates that the pre-reform trends are very similar across groups and just after the reform, in 2011, the level of social security contributions decreases rapidly among the treated entrepreneurs. Therefore, it seems that our comparison between groups is appropriate and the discretion to more freely decide the level of social security decreases the contributions when the insurance status changes from T to Y owners. Also, the eect is occurs right after the reform and the transition period between does not aect the level of contributions. Business activity. We measure business activity by examining the total output of rms, namely turnover. We believe that annual turnover is the best measure for business activity in our empirical setting as it summarizes the size of the total activities of rms. Thus, if there were changes in business activity, this measure should show an increase after the reform among treated rms. However, we also use other measures to approximate the business activity of these entrepreneurs. The upper-left panel of Figure 2 shows the changes in log rm-level turnover for treatment and control groups over time relative to year There seems to be no 7

9 Figure 1: Social security contributions over time: Treatment and control groups dierence in pre-reform trends although the eect of the global downturn seem to have aected both groups in 2009 and There is also no dierence in turnover after the reform. Therefore, our rst graphical examination suggest that the social contributions do not aect the overall business activity. We have data on other rm-level variables that enables us to study other interesting behavioral margins. The upper-right panel of Figure 2 depicts the changes in log annual prots relative to year The annual net prots include all prots that a rm can distribute to its owners. This can be seen as an alternative measure for business activity and similarly as for turnover we should observe an increase among treated rms after the reform if there are any changes in business activity. Again the Figure shows a neat pre-reform trends and a dip in prots during the nancial crisis (2009/2010) but no change in trends between groups after the reform. Another alternative measure for how well the rm is doing, is the distributed gross total income (including both wages and dividends) to its main owner. We expect that the reform would increase the aggregate income from the rm, if the pension contributions would matter for these owners. However, surprisingly, the lower-left panel of Figure 2 shows a decrease for treated owners that is against our hypothesis and the traditional view of how these types of taxes aect the behavior of entrepreneurs. Finally, the lowerright panel of Figure 2 depicts that the decrease in total income is explained by a decrease in wages, implying that smaller compulsory social security contributions decrease the wage income from the rm. This is a surprising result as for the treated owners the level 8

10 Figure 2: Business activity of rms that are owned by Y and T owners of annual wages does not aect the amount of social security contributions any more after the reform, and thus wages should have rather increased after the reform. 4.2 Regression estimates To quantify the eects of the reform, we estimate equation (1) on the log of the outcome variables. Table 3 shows the results. Column (1) indicates a clear decrease, 17 percent, in public insurance payments among the treatment group compared to the control group after the reform. Column (2) shows a signicant, 7 percent, decrease in total income from the rm to the owners that seems to be mainly driven by a decrease in wages (column (4)). Columns (5) and (6) in Table 3 show that there is no economically nor statistically signicant eects on turnover or net prots. Overall, the results suggest no eect of the decrease in social insurance contributions on economic activity. In addition, we do not nd any eects on. However, we nd some weak evidence that long-term investments increase among treated rms after the reform (column (10)) which, at least partially, explains where the money saved in lower social security contributions are invested. 9

11 Owner-level variables Firm-level variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) VARIABLES Pension contrib. Total income Dividends Wages Turnover Net prots Total assets Liabilities Equity Long term inv. DD estimate *** *** *** * (0.015) (0.012) (0.020) (0.016) (0.023) (0.024) (0.019) (0.026) (0.020) (0.102) Post *** 0.109*** *** 0.223*** 0.076*** *** 0.039** *** 0.085*** 0.404*** (0.011) (0.009) (0.014) (0.012) (0.016) (0.020) (0.016) (0.021) (0.016) (0.081) Treatment 0.077*** 0.068*** 0.044*** 0.052*** *** 0.057*** 0.083*** 0.047*** (0.013) (0.010) (0.016) (0.014) (0.019) (0.017) (0.014) (0.019) (0.014) (0.076) Constant 8.414*** *** 8.954*** *** *** *** *** *** *** *** (0.009) (0.007) (0.012) (0.010) (0.013) (0.014) (0.011) (0.016) (0.012) (0.061) N 113, , , , ,733 93, , , ,286 10,066 R Notes: The time period in these specication is , except in Column (1) where we have data only until Heteroscedasticity consistent standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1. Figure 3: DD results 10

12 5 Conclusions In this paper, we provide evidence of how public social insurance taxes aect entrepreneurial activity. We nd that entrepreneurs substantially reduce their contributions to the social insurance scheme when they can more freely decide their level of contributions, implying that the social insurance mandate is binding for entrepreneurs. We nd suggestive evidence that treated owners use the money saved in lower social security contributions to increase long-term investments in their rm. We observe no eects on the the rm-level business activity measures of the aected entrepreneurs. We also show that the aected entrepreneurs receive lower total income from their rm after the reform, due to a decrease in wages rather than dividends. References William J. Baumol, Robert E. Litan, and Carl J. Schramm. Capitalism, and the Economics of Growth and Prosperity Good Capitalism, Bad Yale University Press, Douglas B. Bernheim. Taxation and saving. Handbook of public economics, 3: , Raj Chetty, John N. Friedman, Søren Leth-Petersen, Torben Heien Nielsen, and Tore Olsen. Active vs. passive decisions and crowd-out in retirement savings accounts: Evidence from denmark. The Quarterly Journal of Economics, 129(3): , Julie Berry Cullen and Roger H. Gordon. Taxes and entrepreneurial risk-taking: Theory and evidence for the us. Journal of Public Economics, 91(7): , Ryan Decker, John Haltiwanger, Ron Jarmin, and Javier Miranda. The role of entrepreneurship in us job creation and economic dynamism. The Journal of Economic Perspectives, 28(3):324, Mihir Desai, Paul Gompers, and Josh Lerner. Institutions, capital constraints and entrepreneurial rm dynamics: Evidence from europe. NBER Working Paper Series, June Simeon Djankov, Tim Ganser, Caralee McLiesh, Rita Ramalho, and Andrei Shleifer. The eect of corporate taxes on investment and entrepreneurship. American Economic Journal: Macroeconomics, 2(3):3164, Joel Slemrod Emmanuel Saez and Seth H. Giertz. The elasticity of taxable income with respect to marginal tax rates: A critical review. Journal of Economic Literature, 50 (1):350, doi: 11

13 William M. Gentry and R. Glenn Hubbard. Tax policy and entry into entrepreneurship. American Economic Review, 90(2):283287, Robert E. Hall and Susan E. Woodward. The burden of the nondiversiable risk of entrepreneurship. American Economic Review, 100((3)):116394, doi: / aer Johan Hombert, Antoinette Schoar, David Sraer, and David Thesmar. Can unemployment insurance spur entrepreneurial activity? NBER Working Paper No , November Leora Klapper, Luc Laeven, and Raghuram Rajan. Entry regulation as a barrier to entrepreneurship. Journal of nancial economics, 82(3):591629, Joseph A. Schumpeter. Capitalism, socialism and democracy. Routledge, Danny Yagan. Capital tax reform and the real economy: The eects of the 2003 dividend tax cut. The American Economic Review, 105(12): ,