Labor demand response and labor supply incentives: Evidence from firm outcomes in the context of the German Mini-Job Reform

Transcription

1 Labor demand response and labor supply incentives: Evidence from firm outcomes in the context of the German Mini-Job Reform Gabriela Galassi European University Institute October 9, 2016 PRELIMINARY AND INCOMPLETE Abstract Tax benefits for low-earnings workers are part of many welfare-to-work policies. The theory of tax incidence predicts that employers can appropriate at least part of tax benefits targeted to workers, and hence might react strongly. Whereas existing studies have focused mainly on labor supply responses to welfare-to-work policies, I analyze the labor demand response by exploring plant-level outcomes around the Mini-Job Reform in Germany in 2003, which entailed a significant expansion of tax benefits for marginal employment. Descriptive evidence suggests that firms did not cut gross wages directly to reap the benefits of the lower tax wedge, but created new vacancies targeting workers in the low tax regime. Using a regression based approach I document that plants with a high intensity of low skilled workers prior to the reform expanded relative to low intensity firms. Interestingly, this expansion was not biased towards the type of workers targeted by the tax benefits (low skilled and part-time workers). On the other hand, plants that were initially less exposed to the reform because they had a lower intensity of low-paid workers substituted their employment towards more marginal employment without expanding at the same pace. Both phenomena can be explained by firms trying to capture savings in labor costs after the reform. JEL Classification: H20, H24, H32, E24, E64, I38, J23, J38 Keywords: tax benefit, low-earnings workers, firm outcomes

2 1 Introduction Tax policies aimed at redistributing income towards low-earnings workers while stimulating labor force participation have become very popular in the last decades. The most studied examples are the Earned Income Tax Credit (EITC) in the US, introduced in 1975, and the Working Families Tax Credit (WFTC) in UK, implemented in the late 1990s. The usual design involves tax credits to workers up to certain family earnings limit, which vary according to household composition. Whereas these policies have proven effective in general for stimulating labor supply of low-earnings workers in line with one of their main goals (see for example, Eissa and Liebman 1996 for the US, Blundell 2006 for UK), whether firms react has been overlooked by the literature. 1 In particular, firms might find beneficiaries of the tax policy more attractive than non-beneficiary workers given the lower tax burden. Tax wedge reduction is shared between employer and employee depending on elasticities of demand and supply according to the theory of tax incidence. Studies on the effect of the expansion of the EITC and the WFTC on wages have shown downwards pressure on wages of benefitted workers consistent with this theoretical prediction (Leigh 2010, Rothstein 2008, Azmat 2008). Hence, if firms can save costs on beneficiaries, these workers might be more attractive for them. In this paper I analyze the response of firms to a tax policy reducing the burden on lowincome workers. Recent studies analyzing earnings distribution around kinks and notches introduced by tax schemes 2 have documented that the amount of bunching is higher than if the only relevant response comes from labor supply (Tazhitdinova 2016, Gudgeon and Trenkle 2016 and Haywood and Neumann 2015, Chetty, Friedman, Olsen, and Pistaferri 2011). Different explanations have been proposed, such as adjustment costs and technological constraints (Chetty, Friedman, Olsen, and Pistaferri 2011, Haywood and Neumann 2015), differences in benefits outside the wage package at each side of kinks or nothces (Tazhitdinova 2016) and intra-firm adjustment (Gudgeon and Trenkle 2016). However, direct evidence on firms outcomes and behavior is still missing. This paper focuses on the Mini-Job Reform, which took place in Germany in April 2003 as part of a comprehensive labor market reform (Hartz), and which represented a significant expansion of tax benefits for low-income earners. The so-called mini and midi jobs in Germany involve exemptions and subsidies to Social Security Contributions (SSC) for workers earning up to a certain amount per month: e400 in the mini-jobs and e800 in the midi-jobs. The literature refers to these jobs as marginal, as opposed to regular employment (defined by higher earnings and full SSC). Workers holding a marginal job represent approximately 25% of total employment in the private sector, as shown in Figure (1). This magnitude is relevant, as compared to other forms of atypical employment such as temporary jobs, which amount to 10%. Both the media and the public discourse have warned about the potential risk that this type of policies which make low-paid jobs more attractive increase precariousness in the labor market without net job creation. Insights on 1 The basic assumption in the labor supply analysis, prevalent in the literature about tax benefits for lowincome earners, is that labor demand is perfectly elastic, disregarding general equilibrium effects. 2 In the taxation jargon, kinks refer to changes in the marginal tax rate whereas notches emerge when the average tax rate changes. They are both non-linearities in the budget sets introduced by the tax design. 1

3 Figure 1: Share of workers who hold a mini or midi-jobs in total employment (17-65 years old).3 Share of employed workers (%) Mini jobs Midi jobs Note: SIAB data, using definitions before and after the reform (main and secondary jobs). these claims require understanding firms behavior. The empirical analysis is divided in two parts. In the first part, I explore the hypothesis of intra-firm adjustment, i.e. a worker already employed in the neighborhood of the threshold for defining mini or midi jobs faces a reduction in the gross wage in order to be eligible for the lower tax rate. Descriptive evidence from administrative data on a 2% representative sample of the individuals in social security records (SIAB) 3 suggests that intra-firm adjustments of the type described are not the prevalent way of adjustment. Among the workers with monthly earnings close to the e400 threshold, the proportion pulled down from higher gross earnings levels is relatively low (15%), and it is higher for those who changed establishment (36%). This observation does not lend support to a direct adjustment to the new threshold by cutting the wage of workers within firms. In the second part of the analysis, I use linked employer-employee data for Germany (LIAB) and I estimate the effect of the expansion of tax benefits by the Mini-Job Reform on workplace-level outcomes. The empirical strategy consists of comparing outcomes of establishments highly exposed and non-exposed to workers earning below the threshold im- 3 These data include employment episodes as well as spells of benefit reception of unemployment insurance and assistance and of registration as job seekers or participant of active labor market policies, all of which require notification to the social security system. 2

4 posed by the reform. This methodology is inspired by the minimum wage literature which analyzes the effect on firm behavior (Harasztosi and Lindner 2016 and Machin, Manning, and Rahman 2003). Preliminary findings suggest that highly exposed firms do not increase their intensity in either marginal or part-time employment compared to non-exposed firms. However, the aggregate share of this type of employment increases after the reform, suggesting that nonexposed establishments are becoming relatively more intense in marginal and part-time employment. On the other hand, total employment, which moves similarly for highly and non-exposed workplaces before the reform, picks up for the former in the years following the reform, accompanied with higher turnover (both inflows and outflows). Together with the previous result, this implies that workplaces with higher exposure expand in terms of workforce after the reform, not particularly relying in jobs with tax benefit. The other side of the coin is that non-exposed workplaces are not growing in terms of employment and becoming more parttime and marginal employment intensive, providing an indication of substitution of regular employment in this kind of establishments. Interestingly, establishments do not seem to differ in the behavior in terms of the intensity in unqualified and female workers, groups typically affected by tax benefits. The number of vacancies with respect to the workforce increases more in highly exposed workplaces after the reform, and labor costs decrease simultaneously with a raise in total amount of investment in physical capital. The overall evidence on firms outcomes is suggestive of an expansion of establishments more intensive in low-wage employment, possible due to labor costs savings coming from a lower tax wedge. It is worth noting that the expansion in terms of labor does not seem to be through subsidized work at the expense of regular employment, being no signs of substitution in these workplaces. On the other hand, establishments non-intensive in lowpaid employment appear to substitute regular employment after the expansion of benefits for low-paid jobs. The rest of the paper is organized as follows. In the next section I explain the contribution to the related literature. Next, I detail the institutional context of the Mini-Job Reform. I then explain the empirical strategy used in the paper, and the data. I present the results and finally I conclude and set future research paths. 2 Literature review The existing literature about tax benefits for low-earnings workers in general and the Mini- Job Reform in particular has focused on the effect on labor supply and employment, as it is one of the main objectives to affect these variables. However, the evidence regarding demand side response is still scarce, and this paper aims at contributing to it. A related strand of literature provides evidence that firms share at least partially the wedge reduction implied by tax credits. Azmat (2014) provides evidence that gross wages of eligible workers is reduced compared to similar non-eligible workers when the benefit increases and becomes salient in 1999 for the Working Families Tax Credit (WFTC) in the UK. Employers appropriate 38% of the tax credit by reducing wages. Leigh (2010) exploits the 3

5 state level variation in EITC amount in the US in , and estimates that when the generosity of the tax credit increases in 10%, hourly wages fall by 5% for high school dropouts and 2% for high school graduates. Similarly, Rothstein (2010) simulates the effects induced by the mid-1990s expansions of the EITC, and assuming labor demand is not perfectly elastic, shows that 55% of the benefits given to low-skilled women. All these findings are consistent with the theory of tax incidence, by which employer and employee share the economic incidence of a tax reduction. Whereas the results in Leigh (2010) and Rothstein (2010) hint in the direction of higher labor supply induced by the tax credits, which depresses equilibrium wage, Azmat (2014) claims the main effect comes from the salience of the benefit, and direct wage cuts done by firms. Independent of the mechanism, evidence that employers appropriate part of workers benefits entailed by tax credits hints in the direction of firms facing incentives to respond to this type of policies. In the Mini-Job case, a number of studies suggest the effect on labor supply are lower than for the EITC, mainly focused on the composition of workers (Akyol, Neugart, and Pichler 2013, Jacobi and Kluve 2007, and Eichhorst and Zimmermann 2007 for surveys, Fertig and Kluve 2006, Caliendo and Wrohlich 2010, Bargain, Caliendo, Haan, and Orsini 2010 and Steiner and Wrohlich 2005). However, by design employers know who are the beneficiaries (workers earning below certain gross amount a month) and hence there is salience, suggesting that firms can react directly as shown in Azmat (2014). Another piece of evidence in favor of firms responding to tax benefits comes from the literature using earnings distributions. In particular, for the Mini-Job case, Tazhitdinova (2016) extends the traditional framework for estimating labor supply elasticity using bunching at tax notches and kinks to the case of large changes in marginal and average tax rates. Computing the elasticities using administrative data on social security records for Germany between 1999 and 2010, she documents larger estimates than findings for US and other countries using the bunching methods. Acknowledging that singularities of labor supply incentives in the German setting and potential optimization frictions are not enough to explain the difference entirely, the author provides evidence of firm bunching : individuals with lower incentives to bunch at kink or notches locate at these points because firms offer this type of jobs. To rationalize these findings, she adapts the framework of partial equilibrium tax incidence with job search costs and endogenous hours constraints in Chetty, Friedman, Olsen, and Pistaferri (2011) for the Danish income tax case. The model predicts that labor supply responses are bigger if firms have the statutory incidence of the taxes instead of workers. The author provides evidence that even if the statutory incidence of the low tax wedge implied by the mini-jobs is on the worker s side, firms face lower fringe benefits and higher flexibility at extensive and intensive margin for mini-jobs, being these rationales for a stronger response. Similar approach and findings are in Gudgeon and Trenkle (2016) and Haywood and Neumann (2015), also hinting on the direction of firm responding to tax benefits allocated to workers. These studies provide indirect evidence of incentives for firms to respond when workers have tax benefits. This paper instead analyzes directly firms outcomes for the Mini-Job case in Germany, exploiting the expansion of tax benefits introduced by the 2003 reform. Another related strand of literature analyzes displacement effects of active labor market policies, in particular job seeker assistance (Blundell, Meghir, Costa Dias, and Van Reenen 4

6 2004, Ferracci, Jolivet, and van den Berg 2010, Pallais 2014, Gautier, Muller, van der Klaauw, Rosholm, and Svarer 2015). The evidence is mixed when comparing eligible and ineligible workers in areas affected and not affected by the program. Crepon, Duflo, Gurgand, Rathelot, and Zamora (2013) alert about the possibility of biased estimates when comparing local labor markets targeted by a policy. Instead, they use a two-step randomized design to study displacement effects of job seeker assistance program in France where participants are young and educated. They exploit that both eligible job seekers and treated employment areas are randomly selected, and show that the positive impacts on the probability of obtaining a long-term of stable job of eligible participants compared to non-eligible ones in treatment areas are outweighed by a negative impact on non-eligible workers in treated areas compared to control areas. This paper considers the possibility of substitution effects, though I do not count with a geographic randomization, since the Mini-Job design is homogeneous nationwide. A somewhat related literature focused in the case of the Mini-job Reform in Germany some studies has estimated the parameters of the labor demand for heterogeneous labor using a flexible cost function framework (Freier and Steiner 2007, Addison, Bellmann, Schank, and Teixeira 2008, Jacobi and Schaffner 2008). Particularly, Jacobi and Schaffner 2008 studies the evolution of the estimated elasticity of substitution between unskilled and skilled labor using the industry level data from the LIAB, finding no changes after the reform. This paper contributes to this literature as well. Differently to this structural approach, I do not rely on functional forms assumptions for the production function and explore instead direct firm dynamics outcomes and changes in labor composition within firms, exploiting the variability in the intensity on low-paid workers. 3 Institutional Context The Mini-Job Reform was part of a wider set of policies, the so-called Hartz reform, which were gradually implemented between 2003 and The main goal was to increase labor market participation, by providing incentives to both the labor supply and demand. The explicit objectives in the legislation are to (1) reduce unemployment, (2) increase competitivity, and (3) stimulate both labor supply and demand. In this paper I focus on the second set of policies (Hartz II or Mini-Job Reform), one of the most controversial reforms was Hartz II, which consisted in a set of exemptions and subsidies in social security contributions (SSC) paid by workers with low-earnings ( marginal workers as the literature has referred to), introduced in April Before the reform, mini-jobs existed already in Germany. However, they had been restricted to employment with a maximum of 15 hours a week and a gross monthly wage of e325 or less, provided it was the only source of income for the worker. Mini-jobbers were exempted from SSC and income tax, while the employer paid 22% tax on gross wages. Earnings from several mini-jobs were added up and the resulting amount was subject to SSC. Above the threshold, regular SSC (21% for both workers and employers) plus income tax kicked in. The the reform consisted in the extension of the wage limit to e400 and the elimination of the hour limit. Employers SSC were raised to 25%. In addition a phase out category 5

7 (midi-job) was introduced for monthly gross wage between e400 and e800, where the SSC increase linearly for the worker while it is the normal rate for the employer, and income tax applies normally. Secondary jobs were allowed to qualify as mini or midi-jobs if complying with the earnings limits, benefiting from a lower tax burden. The change in the definition of mini-jobs led to an expansion from approximately 4 million in 2003 to more than 7 million in 2007 according to official statistics. Adding up midi-jobs, they are 8.5 million or 25% of total private employment. While workers in midi-jobs are entitled to the same benefits as the rest of employees in Germany, mini-jobbers have health insurance, paid holidays and other benefits such as maternity leave, but they do not have pension entitlements. Regarding income taxation, the amount coming from mini-jobs is exempted on an individual basis. 4 Finally, mini-jobbers earning up to e165 are also entitled to unemployment insurance and social assistance. As figure (2) shows, the institutional design is reflected in the shape of the earnings distribution. There is a considerable amount of bunching around the upper limit of monthly gross earnings defining mini-job. Moreover, the location of the bunching shifts from e325 before 2003 to e400 after. There is not a comparable amount of bunching for the e800 defining midi-jobs. Figure (3) shows there is not a similar change in the components of earnings, i.e. hours and hourly wage, given the institutional design does not impose limits with this respect, only in terms of monthly earnings. Overall, this suggests that the presence tax exemptions in mini-jobs represents a strong incentive for both workers and firms to bunch at the e400 a month threshold. As a consequence of the institutional design, the German labor market comprises what has been referred in the literature as regular and atypical employment (see for eg. Eichhorst and Tobsch 2013, Keller and Seifert 2012). Regular employment is defined as a combination of the following characteristics: full-time, high payment, permanent employment contract, and complete integration into social security system. On the other hand, atypical employment is characterized by any of the following features: part-time, mini-jobs and midi-jobs, temporary and agency employment. In this paper, I will consider a particular type of atypical employment, the mini and midi-jobs, referred as marginal employment, one of the main components of atypical forms of employment (25% within the private sector, compared to 10% temporary workers). The rest of the employment will be referred as regular employment in the rest of the paper, even though some forms of atypical employment are also included in it (temporary work with earnings above e800 a month for example). The table (1) shows the composition of the working age population in Germany in 2005, after the reform, as well as average hours and earnings of each employment type. Minijobs are comparable to regular part-time employment in average hours (14 and 13 a week respectively), though the dispersion is much higher (the standard deviation is 12 compared to 5.7 for regular part-time jobs). However, average hourly earnings of regular part-timers 4 The individual base of the tax benefit entailed by the mini-jobs is different from the tax credits in the US and UK, in which the benefit is computed on the family income declaration. The joint declaration for couples in the Germany income tax is important though because it generates strong disincentive for secondary earners to work above the e400 monthly level, above which the higher family tax rate applies. 6

8 Figure 2: Monthly gross earnings in main job (25-55 years old) Frequency Euros/month Note: SIAB data, daily wage converted to monthly. 7

9 Figure 3: Weekly working hours and hourly net earnings in all occupations Density Density Hours/week Hours/week Density Density Hourly net wage Hourly net wage Note: G-SOEP data. 8

10 double mini-jobbers (19 and 10 respectively). For midi-jobs, average hours worked, 26, are below the full-time workers, 41, while double those of part-timers. The dispersion is higher as well (13.9, 9.5 and 6.7 of standard deviation respectively). The hourly net earnings are comparable to full-time workers in terms of mean (8 and 9) while they are more disperse (the standard deviation is 16.4 and 4.2 respectively). Finally, 2% of the working age population holds a mini or midi-job as a secondary employment. This group resembles full-time workers, except for a higher hourly earnings for those whose second job is a midi-job. Table 1: Composition of working age population (17-65) in Germany, Prop. (%) Hours/ week Hr. wage Month. gross earn. Inactive 15.1 Studying 3.4 Unemployed 5.0 Irregularly employed 2.4 Part-time ,640 (5.68) (21.20) ( ) Full-time ,634 (9.50) (4.16) ( ) Mini-job (main) (12.00) (25.59) (161.51) Mini-job (secondary) ,456 (13.97) (6.59) ( ) Midi-job (main) (13.86) (16.41) (166.63) Midi-job (secondary) ,123 (16.96) (12.38) ( ) Total ,021 (15.21) (11.94) ( ) Note: G-SOEP data. Hours worked, hourly net earnings and monthly gross earnings are reported only for those who have a positive declaration (excluding zeros and missing). Considering the transitions between 2002 (before the reform) and 2005, shown in table (2), mini-jobs come mainly from regular full-time (19%) and part-time (17%) employment, and a non-negligible fraction (13%) comes from unemployment, as it is shown in the Column totals. Inflows from regular employment to midi-jobs is stronger. Considering the destination of workers according to their origin (Row totals), mini and midi-jobs absorb 10% and 2% respectively of unemployed, though the big majority of jobless workers either remain in their original state (60%) or transit directly to regular employment (27%). Similar observation emerges for regular workers: 16% of part-timers become mini or midi-jobs, compared to 77% who end in regular employment; the numbers for full-timers are 3% and 90% respectively. Hence, even if mini and midi-jobs represent an important part of the labor market, they do not seem to replace regular part-time work. Finally, table (3) shows that marginal workers have particular characteristics. They are mainly women, low educated workers, married, secondary earners, in poor households. Both pre and post-reform labor outcomes of marginal workers suggest they are less experienced than regular workers, with shorter tenure, in part-time occupations and with lower hourly earnings. After the reform, the gap in terms of part-time incidence is widened be- 9

11 Table 2: Transitions 2002/2005, years old Row totals Unemployed Mini-job Midi-job Regular PT Regular FT Regular other Total Unemployed 60% 10% 2% 5% 22% 0% 100% Mini-job 13% 60% 5% 9% 13% 1% 100% Regular PT 7% 9% 7% 66% 11% 0% 100% Regular FT 7% 2% 1% 3% 87% 0% 100% Regular other 12% 5% 2% 11% 64% 6% 100% Total 12% 8% 2% 13% 64% 0% 100% Column totals Unemployed Mini-job Midi-job Regular PT Regular FT Regular other Total Unemployed 48% 13% 10% 4% 3% 18% 10% Mini-job 7% 51% 14% 5% 1% 17% 7% Regular PT 8% 17% 40% 74% 3% 19% 14% Regular FT 36% 19% 35% 17% 92% 33% 68% Regular other 0% 0% 0% 0% 0% 14% 0% Total 100% 100% 100% 100% 100% 100% 100% Note: SIAB data. tween marginal and regular workers. Table 3: Characteristics of unemployed, mini and midi-jobbers, and regular workers Unemployed Mini-job Midi-job Regular employment Demographics and socio-economics Women Age Secondary or less Apprenticeship Higher education Married Single parent Head of HH* Per capita annual HH income* 9,764 9,911 10,665 14,247 In 2002 Employment experience Tenure in job Part-time Full-time Weekly working hours* In 2005 Part-time Full-time Hourly net wage* Weekly working hours* Note: SIAB data, except for marked with *, for which G-SOEP data is used. 4 Empirical Strategy In this section, I describe the data used. I also provide some descriptives and explain the main empirical strategy in the analysis. 4.1 Data I use the social security records provided by the Institute for Employment Research of the German Federal Employment Agency (IAB). The period under analysis is

12 Information on mini-jobs is available only since 1999, where there was a modification to the mini-job legislation, and the end of the window aims at avoiding introducing the effects of the major economic crisis in I use the Sample of Integrated Labor Market Biographies (SIAB) and the Linked Employer-Employee (LIAB). Access to the data is provided on-site at the Research Data Centre (FDZ) of the IAB, with subsequent remote access. The weakly anonymous version of the SIAB the data provides information on employment, job search and unemployment benefit reception for a 2% sample of wage-earners (excluding civil servants, self-employed and family workers) in Germany, comprised in the Integrated Employment Biographies (IEB). The data include official information on employment subject to social security (since 1975), marginal employment (since 1999), unemployment (since 1975), social benefits (since 1975), registered jobseekers (since 2000), and participants in employment or training programs (since 2000). It provides labor market biographies on more than 1.5 individuals from 1975 to 2010, representing more than 80% of all employed persons in Germany. Employment histories consist of notifications to the social security system done by employers in the event of hirings, terminations, switches in the contribution group or health insurance company of the employee, or changes in the payroll system of the employer. Worker information includes age, gender, education (only for employment spells), marital status (only for spells of benefit reception of job search), daily wages and benefits and intensity of employment (part-time/full-time). Workplace information is registered in the Establishment History Panel, which aggregates information from individual social security records by establishment ID resulting in more than 300,000 establishments. It is a cross-sectional dataset, and covers West Germany since 1975 and East Germany since The BHP contains information about the branch of industry and the location of an establishment, the employment structure given by the number of employees liable to social security and marginal part-time employees, disaggregated by gender, age, occupational status, qualification and nationality, median wages for full-time employees, gross worker flows and establishments creation and closure. The LIAB combines official social security data on individuals from the IEB with establishment data from the IAB Establishment Panel/Betriebspanel, merged through an establishment identifier available in both datasets. The IAB Establishment panel begins in 1993 in West Germany and is extended to the East Germany in The sample comprises approximately 11,000 establishments and is periodically refreshed to account for establishment deaths and births. It is stratified by establishment size, industry and region, it oversamples large establishments and excludes unipersonal and informal firms. The (weighted) sample is representative of the population of establishments with employees liable for social security contributions (and marginal part-time workers since 1999). The participating establishments are surveyed on a large number of employment policy-related subjects, including standard topics covered annually and wave specific topics. Information includes employment development, business policy and performance, investment, collective bargaining, personnel structure and recruitment, remuneration, and working time. Data preparation includes some adjustments such as imputation of education levels (which are only available for employment records), elimination of parallel spells to consider only the main job with the highest payment, and transformation of spell data in the SIAB into regular frequency data, either annual or quarterly. 11

13 Figure 4: Monthly gross earnings in secondary job (25-55 years old) Frequency Euros/month Note: SIAB data, daily wage converted to monthly, employment spells parallel to main job. 4.2 Descriptive Analysis The first observation in the data, as shown in figure (2), is that there is a strong incentive for either workers or firms to bunch at the mini-job threshold of e400. Given there is not a comparable bunching at the midi-job e800 threshold, the empirical analysis is based only on mini-jobs for now. Bunching analysis suggest that workers respond to kinks and notches in the tax design by bunching at these points. However, there is indirect evidence that firms also bunch at these points. 5 One observation consistent with this hypothesis is that even if before the Mini-Job Reform mini-jobs held as secondary employment were not entitled to the same tax benefits, there is a considerable amount of bunching at the prevalent e325 threshold for secondary occupations in 2002, as figure (4) shows. Bunching at the threshold becomes more pronounced in 2005, after the reform given secondary jobs are entitled to the same benefits as main occupations. However, this is an indication that even if workers do not have an incentive to bunch at e325 in 2002, firms might find it profitable to create employment close to the threshold. A natural question is whether firms bunch a the threshold by adjusting wages of workers already in their payroll, or by creating new vacancies to benefit from the tax design and potentially dismissing workers who are not entitled for the benefits. Descriptive evidence in 5 This point is already made in Tazhitdinova (2016), Gudgeon and Trenkle (2016) and Haywood and Neumann (2015). 12

14 table (4) favors the second hypothesis. Looking at workers close to the mini-job threshold before the reform in 2002, two thirds remain below the new threshold in 2004 after the reform. 29% profit from a gross wage increase but up to the new e400 threshold, whereas less than 15% surpasses this threshold (very few, 4%, to the midi-job range, between e400 and e800). Upward mobility in terms of wage is stronger for leavers (those who change establishment, 29% of the total) than stayers (who do not change establishment, 71%). Considering workers who end up close to the mini-job threshold after the reform in 2004, more than one third were already in mini-jobs under the pre-reform definition. Only 15% are pulled down from above the e400 threshold. This proportion more than doubles for workers who change establishment. Even though this evidence is merely suggestive, it is consistent with firms not being able to adjust wages of workers already in their payroll, but creating new vacancies close to the threshold for benefiting from the reform. Table 4: Workers close to the mini-job monthly gross earnings threshold Earnings bracket Mini-jobs in 2002, by earnings in 2004 Earnings in 2002, of mini-jobs in 2004 Total 19,253 18,667 Inactive Unemployed (0, 325] (325, 400] (400, 800] more than Not changed establishment out of total of these (0, 325] (325, 400] (400, 800] more than Changed establishment out of total of these (0, 325] (325, 400] (400, 800] more than Note: SIAB data. Close to the mini-job threshold x means in (x 75,x] for x {325,400} for 2002 and 2004 respectively. Inactive are considered those individuals who are in the data in 2002 and then are not in 2004 (first column), or viceversa (second column). To shed some light on which are the mechanisms behind the bunching, table (5) shows that 12% of workers close to the e400 threshold in 2004 reduce the hours worked by changing from full-time in 2002 to part-time in 2004, and 24% change the occupation. This transitions are stronger among workers who change establishment. Of workers who remain in the same establishment, the adjustment downwards in terms of gross wage to bunch at the e400 threshold is stronger for small (21%) and large (23%) workplaces than in medium (14%), as it is shown in table (6). It is reasonable in the sense mini-jobbers close to the threshold are concentrated in the latter (83%) compared to small 13

15 Table 5: Workers close to the mini-job monthly gross earnings threshold in 2004 (e400) Changes with respect to 2002 Full-time to Part-time Change occupation Total Of those who do not change establishment Of those who change establishment Note: SIAB data. Close to the mini-job threshold e400 means in (325,400] for and big establishments (7% and 10% respectively), where it might be difficult to adjust the wages selectively, and adjusting them for all might be too costly. Table 6: Workers close to the mini-job monthly gross earnings threshold in 2004 (e400) Of those who do not change establishment Small Medium Large Out of total (0, 325] (325, 400] (400, 800] more than Note: SIAB data. Close to the mini-job threshold e400 means in (325,400] for For workers who change workplace and end up close to the mini-job threshold, table (7) shows that a considerable proportion change industry (71.5%) or occupation (80%). These workers potentially loose in terms of human capital, and hence it is not surprising that this transitions are stronger among the group in which there is sorting down in terms of wage. Table 7: Workers close to the mini-job monthly gross earnings threshold in 2004 (e400) Of those who change establishment Change industry Change ind. or occup. Out of total of those who increased earnings of those who did not increase earnings Note: SIAB data. Close to the mini-job threshold e400 means in (325,400] for To sum up, the evidence is suggestive of firms not favoring direct gross wage cuts as a way of taking advantage of the lower labor costs entailed by the mini-jobs, particularly in those in which jobs close to the threshold are predominant. Firms instead seem to be opening new vacancies close to the threshold. Workers accept jobs in this segment either reducing hours or changing completely the type of job for which they have experience, suffering hence a loss in terms of gross wage. Table (8) presents establishment level information by marginal employment intensity (quintiles of share of mini-jobs) after the reform (2005). Workplaces which are more intensive in the use of low-paid jobs are slightly younger and smaller, they have a higher proportion of women, lower fraction of full-time employees and pay the lowest wages. All these features are to be expected given the characteristics of mini-jobbers. There are a few observations that go against expectations. First, marginal employment intensive workplaces are not more intensive in unqualified workers (with low level of education) before the reform, and the share of unqualified workers declines after the reform. The same happens when 14

16 considering unskilled labor (according to the task workers perform). Turnover of this type of labor (inflows and outflows) is not relatively higher in these workplaces, and it decreases by This is unexpected given that mini-jobs are associated with low skilled workers. Secondly, establishments with the higher share of marginal employment reduce the gap between wages in the 75 percentile and 25 percentile of full-time workers between 2002 and 2005, i.e. they become more equal. If there was selection of mini-jobs into bad establishments, which pay the lowest wages and are more unequal, I would expect to see the opposite. Table 8: Establishment characteristics by quintiles of share of marginal employment in Establishment age N. employees Share female employees Share unqualified employees Share unqualified employees Share unskilled employees Share unskilled employees Share unskilled inflows Share unskilled inflows Share unskilled outflows Share unskilled outflows Median wage FT employees P75/P25 wage FT employees P75/P25 wage FT employees Note: SIAB data. Marginal employment corresponds to mini-jobs according to the Establishment History Panel. Evolution in time of the establishments in the IAB Establishment Panel 6 show that the total number of workers decreases, more pronounced after the 2003 reform, whereas marginal employment shows the opposite behavior increasing particularly after the reform (figure (5)). Establishment-level share of part-time employment also displays an increasing trend accelerated by the reform, whereas the proportion of low-qualification workers has a sustained decline in time (figure (6)). In the next subsection, I describe the regression analysis performed in the paper, aimed at providing further insights on the evolution of firm outcomes associated to the Mini-Job Reform. In most of the figures presented before, it is apparent that the Mini-Job Reform (red vertical line) is an important factor, but not the only one, affecting employment trends in the period. It is worth noting that 2004 and 2005 also seem to have an effect on total employment and its components, probably due to the Hartz IV reform. It took place in 2005, but was anticipated since 2004, and consisted in a reduction of the income support to workers once they exhaust the unemployment benefit. This policy change interacted 6 Figures for the IAB Establishment Panel, Longitudinal Model. 15

17 Figure 5: Total number of workers and of marginal employees in LIAB Total n. of workers 8 N. marg. part time workers N. workers N. workers Note: LIAB Longitudinal Model (unweighted). 16

18 Figure 6: Share of part-time and low qualified employment out of total employment in establishments of LIAB Part time Low qualification Share Share Note: LIAB Longitudinal Model (unweighted). 17

19 probably with the Mini-Job Reform, given low-paid workers, who have a higher probability of unemployment, saw their benefits and hence reservation wage reduced. I will comment on the results acknowledging this interaction for 2004 on. In the figures before the 2008 crisis also seems to have affected, but this is excluded from the following analysis given the observation window stops in Regression Analysis The regression framework is inspired in the minimum wage literature which analyzes the effect on employment (Harasztosi and Lindner 2016, Machin, Manning, and Rahman 2003). The idea is to exploit the pre-reform variation in the intensity of low-paid employment to compare highly exposed to non-exposed establishments. Highly exposed firms are mechanically more affected by the reform, in the sense that the tax burden on the total payroll is lower and they can profit from cost reductions. The equation estimated is: y jt = α j + λ t + β t FA 03 j + γx jt + ε jt (1) where y jt stands for establishment j outcome in period t, α j is establishment fixed effect, λ t is year fixed effect, FA 03 j is the fraction of workers in establishment j in year 2002 (before the reform) paid below the 2003 mini-job threshold of e400, and X jt is a set of time-varying controls at the establishment level. Establishment outcomes include the share of marginal and part-time employment, total employment, fraction of unqualified and female workers, vacancies, total labor costs, inflows and outflows of workers, sales and investment. The coefficient of interest is β t, which is computed for each year (the baseline is 2003, the year of the reform for which the value is 0). X jt includes district and industry (3 digit code). The assumption is that after controlling for heterogeneity at the establishment level, both time-invariant (establishment fixed effects) and observable time-varying (controls) and common macroeconomic shocks (year fixed effects), β t captures the effect of the Mini-Job Reform through lower labor costs implied by lower tax burden on employees that highly exposed firms enjoy compared to non-exposed firms. In the following section, estimates of β t are presented for different establishment outcomes. 5 Results Figure (7) shows the estimates for β t with respect to atypical employment. Considering that both marginal and part-time employment increase after the reform, the results here suggest that non-exposed establishments are driving the aggregate trend, particularly in the case of part-time for which the coefficients are statistically different from zero since And this is the case even before the reform. The results point in the direction of a convergence in marginal and part-time employment intensity, with non-exposed workplaces becoming more intensive in these forms of atypical employment compared to highly exposed establishments. Highly exposed firms, for which the growth rate of total employment was the same as in non-exposed firms before 2003, expand relatively since 2003 as shown in figure (8). The coefficient is statistically significant for total employment in 2004 and The relative expansion does not seem relying on the type of workers more associated to mini-jobs, such 18

20 Figure 7: Intensity in marginal employment (left) and part-time employment (right) year year as low qualified and women (figure (9)). Note: LIAB Cross-Sectional Model (weighted). A higher turnover (inflows and outflows of workers) accompanies the relative employment growth in highly exposed establishments compared to non-exposed (figure (10)), which supports the previous observation that firm and worker bunching does not occur through intra-firm adjustment, but through workers changing workplace. Consistently, figure (11) shows a higher number of vacancies to total employment since 2003 for highly exposed establishments. It is interesting that these units experience a reversal in the relative trend of their labor costs, which were growing faster than in non-exposed workplaces before 2003 and start growing slower (though the coefficients are not statistically significant) after. This result is consistent with the expectation that units with a higher proportion of workers below the mini-job threshold get mechanically a reduction in labor costs due to the lower tax burden on employees. However, it is surprising because this firms are increasing employment relatively. Costs savings through economic incidence of lower tax burden seems to dominate higher wage bill from new employment. Finally, figure (12) shows the evolution of sales and investment. Even if sales do not seem to increase in highly exposed establishments compared to non-exposed establishments, which is counterintuitive given the relative increase in employment, investment bill goes up in relative terms. This is suggestive of expansion of expansion not only in employment but also in physical capital. 19

21 Figure 8: Total employment: liable to social security (left) and including non-liable to social security (right) year year Note: LIAB Cross-Sectional Model (weighted). Overall, evidence from the establishment panel points in the direction of a relative scale increase in workplaces with a high pre-reform intensity in low-paid employment, made possible by a big costs savings potentially coming from the lower tax wedge on the workforce. Particularly interesting is the fact that this expansion is not disproportionately done through marginal, low skilled or part-time employment, typically associated to mini-jobs. On the other hand, non-exposed firms become more intensive in these forms of employment, potentially displacing regular employees given their employment is decreasing as compared to highly-exposed workplaces. 6 Conclusions This paper provides an empirical exploration using administrative social security records on firm response to the expansion of an in-work benefit materialized by tax reductions to lowpaid workers. Focusing on the case of the Mini-Job Reform in Germany in 2003, I begin by discussing the importance of considering how firms react to a policy aimed at improving workers earnings in the lower tail of the distribution. Different from the literature so far, which has focused on labor supply responses to this type of interventions or has provided indirect evidence of firm reaction by acknowledging supply responses are not enough to explain the total effect, this paper provides a direct analysis of establishment level outcomes. 20

22 Figure 9: Intensity in unqualified (left) and women (right) year year Note: LIAB Cross-Sectional Model (weighted). The Mini-Job Reform is particularly interesting because even if the goals were similar to well-known policies such as the Earned Income Tax Credit in the US or the Working Familise Tax Credit in the UK, there are a number of singularities accruing to it. First, the benefit is visible to employers, allowing them to respond directly. Second, official numbers reflect a big impact in the labor market, being 1 out 4 workers in marginal employment after the reform, as compared to 1 out of 10 in temporary contracts. Descriptive analysis zooming in on the workers close to the threshold relevant for categorizing as mini-jobbers suggests that the dominant strategy for firms is not to directly cut gross wages of the existing work force, but to create new employment (and potentially dismiss part of the workers) in order to take advantage of the tax benefit. Workers suffer (gross) wage losses either by reducing the number of hours worked or by losing specific human capital by changing the tasks in which they are specialized. Using the approach in the minimum wage literature to differentiate highly-exposed and nonexposed establishments with respect to the policy, I perform a regression analysis on several workplace-level outcomes. The findings point in the direction of a relative expansion in terms of labor and capital in highly-exposed production units, potentially possible thanks to the labor costs savings derived from the reform. Interestingly, the expansion does not seem biased towards the type of workers (low skilled, part-time) targeted by the tax benefits. The other side of the coin is an intensification in marginal employment in non-exposed firms, 21

23 Figure 10: Worker inflows (left) and outflows (right) year year Note: LIAB Cross-Sectional Model (weighted). possibly to take profit of the expansion in tax benefits. Given this firms are contracting relatively, this result hints on the direction of a substitution of regular employment in these units. These results are interesting on their own because they suggest that incentives targeted to workers are not innocuous to the demand side in the labor market. They also shed some light with respect to the public discourse regarding the pervasive effect of the Mini-Job Reform, deemed as the cause of a spread in precarious employment in Germany. The findings here suggest that tax benefits are not negative per se, do not lead automatically to a substitution of regular employment by marginal employees. Optimizing firms use costs savings to create new employment, both in the marginal and the regular segments. However, a further understanding of the implications for welfare of firm reactions to these type of benefits calls for the development of a theoretical framework, work in progress at the present moment. 22