El presente Reglamento será obligatorio en todos sus elementos y di rectamente aplicable en cada Estado miembro.
CATEGORÍA 1 — MATERIALES ESPECIALES Y EQUIPOS CONEXOS 1A Sistemas, equipos y componentes
N. B.: VÉASE TAMBIÉN EL ARTÍCULO 1B102.
Reducing unemployment is a common public policy goal which becomes especially important during a period of economic crisis. For this reason, unemployment insurance (UI) policies aim to provide a social safety net while limiting moral hazard to promote re- employment and reduce time in unemployment. In this context, most studies analyze how the generosity ofUIsystems in terms of potential benefit duration (PBD) and replacement rates (i.e. benefit levels) affects re-employment outcomes (Solon 1985,Katz & Meyer 1990b,Card & Levine 2000,Kolsrud et al. 2018, etc.). However, this focus neglects another typical channel of leaving unemployment: the transitions into self-employment (SE). This post-unemployment outcome which accounts for 10-15% of the labor force in the member countries of theOrganization of Economic Co-operation and Development (OECD)1 is economically relevant since one quarter of all new firms is started out of unemployment each year.2 Therefore, given the potential of successful startups to sus- tainably create additional employment or boost innovation, and because self-employment is a typical trajectory for individuals to exit unemployment, it is important to understand the role ofUIbenefits on the transition from unemployment to self-employment. This is necessary to complete the analysis of how the design of unemployment benefits affects all relevant post-unemployment outcomes (and not only dependent employment), and thus may lead to more efficient unemployment policies.
The second chapter of my doctoral thesis aims to shed light on this issue by analyzing howUIbenefit levels affect the probability of unemployed individuals to become self- employed, as well as their actual unemployment duration before transitioning to self- employment in comparison to the case of re-employment.3 Exploiting reform-driven exogenous variation inUIbenefit levels, we are among the first to estimate the causal effect ofUIbenefits (holdingPBDfixed) on total employment, and decompose the overall effect into the causal effects on transitions from unemployment to self-employment and to re-employment. Since most other studies investigate increases inUIgenerosity, our focus on analyzing a reduction inUIbenefit levels is also novel within this field of research.
To answer the research questions mentioned above, we focus on the Spanish UI system and use comprehensive social insurance data linked with income tax data from theContinuous Working Life Sample (Muestra Continua de Vidas Laborales) (MCVL). In particular, we prepare the administrative data to include so far inaccessible information on self-employment because this is necessary to analyze our variables of interest.
1Spain is particularly interesting because its self-employment rate is among the highest in theEuropean Union (EU). Spain’s self-employment rate has been between 16.4% and 17.9% within the last decade (OECD 2018).
2In Spain, around 30-50% of founders between 2005 to 2017 have been unemployed before starting their firms. 3Concerning the notation in this chapter: with self-employment, we refer to the labor market status to distinguish unemployment, employment and self-employment. Within the labor market status of self-employment, the term founder refers to the person starting a firm which covers both firms with and without employees. The term entrepreneur is used to focus on a founder who continues to run a firm after having started it. The term startup refers to the act of starting a firm and is used as a synonym for new firm.
First, we document the evolution of all relevant labor market status flows over the busi- ness cycle (2005-2017) in Spain, and analyze the relevance of flows from unemployment into self-employment. Second, our causal analysis focuses on the direct link between the reduction ofUIbenefits and the probability to become self-employed (compared to becoming re-employed) in a setting where thePBDschedule of theUIsystem remained fixed. In 2012, the Spanish government implemented a labor market reform which led to a sharp change inUIbenefits: it decreasedUIbenefit levels by 10 percentage points (from a replacement rate of 60% to one of 50%) for all eligible individuals with aPBD surpassing six months. This quasi-experimental set-up allows us to exploit exogenous variation in our explanatory variable of interest, theUIbenefit level, in order to estimate the causal effect of a reduction ofUIbenefits on the probability to become either self- employed or re-employed as dependent employee. In this context, we decompose the total reform effect on the average actual unemployment duration into the effect on individuals who become either self-employed or re-employed, and we calculate distinctUIbenefit level duration elasticities. We apply both a Difference-in-Differences (DiD) approach and a Regression Discontinuity Design (RDD) to estimate our causal effects. TheDiD estimation allows us to study not only the average treatment effect, but also the dynamic reform effects on treated relative to untreated individuals over the unemployment spell. This enables us to analyze the behavioral responses toUIbenefits with respect to both job search and startup efforts. TheRDDapproach, which relies on the time interval between theUIentry date and the sharp reform cutoff date, confirms the internal validity of our identification strategy: it shows that manipulation around the cutoff date is not an issue because the reform could not be anticipated due to its unexpected implementation.
Regarding the causal effects of the reform, we find that in response to the cut inUI benefit levels the self-employment probability is estimated to be rather unaffected in the short run. In the medium and long run, this effect tends to become negative. On the contrary, the probability of finding a job is rather positively affected in the short run while flattening out in the medium and long run. The total employment effect is thus rather slightly positive in the short run, but attenuates towards zero after two years. These results clearly show a behavioral response of the affected individuals. In response to the reform, treated unemployed individuals increase their search intensity to find employment beforeUIbenefits drop after six months. This explains the increase in the short-run employment probability and its decline after the first six months. Instead, when we consider self-employment as an additional exit channel out of unemployment, this response seems to be much smaller. In relative terms, the self-employment probability declines compared to the job-finding probability.
We estimate that theUIbenefit duration elasticity is around 0.4 (in theDiDsetting) and 0.5-0.66 (in theRDDsetting) for those becoming re-employed, which is slightly higher compared to findings in other studies of the literature. Interestingly, we find theUI benefit duration elasticity to be smaller (around 0.26-0.33 in theDiDsetting and 0.11-0.38 in theRDDsetting) for those transitioning from unemployment to self-employment.
Thus, UI benefit levels affect the actual unemployment duration of unemployed individuals no matter whether they become re-employed or self-employed. However, the effect is stronger for re-employment than self-employment, i.e. reducingUIbenefit levels reduces actual UI duration more for those transitioning into employment than those becoming self-employed. On the macro level, our results suggest that the cut in UIbenefit levels shifts the transition from unemployment towards employment rather than self-employment. Finally, our descriptive analysis illustrates that in response to the reform new firms are predominantly created in the service sector, whereas the share of startups in the industry and construction sector declines. This indicates that in addition to the effects on the extensive margin, the quality of self-employment may be affected. Less generous UI benefits may not only decrease transitions from unemployment to self-employment, but also increase the share of necessity-driven entrepreneurship among previously unemployed founders. Therefore, we try to disentangle the causal reform effect on different measures of self-employment quality to obtain evidence for the potential welfare effect. However, as both of our quasi-experimental models produce only insignificant results, we cannot confirm the causal nature of the welfare effects. Thus, more research is needed to assess the potential welfare implications ofUIbenefits on self-employment.
This chapter of my dissertation relates to three strands of the literature, and makes three contributions. First, we contribute to the entrepreneurship literature by provid- ing evidence on the role ofUIbenefits for entrepreneurship (Evans & Leighton 1989a,
Levine & Rubinstein 2017, etc.). Using administrative data from social insurance and tax authorities in Spain, we enable analyzing the unemployment exit channel into en- trepreneurship over time (and business cycle), and show that inflows into self-employment from unemployment can account for up to 50% of all new businesses (in times of crisis).
Second, this project adds to the literature in public economics on the optimal design of unemployment insurance; especially, by providing evidence for the effect ofUIbenefits on the transitions into re-employment and self-employment. The public economics literature has discussed several aspects concerning the optimal design of unemployment insurance policies, i.e. the level of benefits and their eligible duration (Schmieder et al. 2012,2016, Kolsrud et al. 2018, etc.). Its focus has been on investigating effects on subsequent employment outcomes, predominantly re-employment wages (Schmieder et al. 2016,Nekoei & Weber 2017). Results suggest that increases inUIbenefit levels lead to increases in actual unemployment duration. However, the effects of longer actual unemployment on re-employment wages are disputed. For instance,Nekoei & Weber
(2017) argue that longer PBD can either induce delay in job acceptance and simply subsidize leisure, or improve job opportunities through promoting a longer search that results in job matches of higher quality. WhileNekoei & Weber(2017) find that the latter positive effect dominates in Austria,Schmieder et al.(2016) report negative effects of unemployment durations on re-employment wages in Germany. We contribute to the debate by providing first evidence on the causal effect of cutting unemployment benefit
levels on self-employment. Thus, this chapter complements the analysis ofUIbenefits regarding post-unemployment outcomes (Jäger et al. 2019). We show thatUIbenefits generate a fiscal externality (Lawson 2017) through the transition from unemployment to self-employment. This should be taken into account for the optimal design ofUIsystems.
Third, we contribute to the literature on (un)intended consequences of economic crisis politics. In fact, the labor market reform that we analyze was one of the policies to deal with the Spanish crisis and was supposed to reduce unemployment under the pressure of fiscal consolidation. Thus, we contribute to the literature on extendingUI generosity during times of crisis which has mostly focused on the US (e.g.Farber et al. 2015,Card et al. 2015) because we provide evidence on how the non-standard response of cuttingUIbenefits in a crisis period affects both re-employment and self-employment. Therefore, we also contribute to the limited literature on reducingUIgenerosity instead of increasing it. For instance,Rebollo-Sanz & Rodríguez-Planas(2020) orDoris et al.
(2018) find that cuts inUIbenefits can increase the job-finding rate. Instead, we are the first to investigate the effect of a cut inUI benefits on the self-employment (start-up) rate. Moreover, this chapter relates to the work ofHombert et al.(2020) who exploit a French reform in 2002 which lowered the downside risk to start a business. They find that more self-employment is created when more social security is provided. In contrast, our contribution is to analyze the causal effect of providing less security (lessUIbenefits) on self-employment (instead of employment), and on theUIbenefit duration elasticity.
The case of Spain is especially interesting as its external validity is higher compared with inference from other European countries with good data access on self-employed individuals (e.g. Scandinavian countries whose labor markets are smaller). Our research questions appear to be of high relevance in times of high unemployment rates (as in Spain and other European countries during the 2010s).4 Moreover, we can learn about the general bias created in studies which focus only on employment and give insights on the full picture of inflows into self-employment.
The chapter proceeds as follows. Section 2.2provides the theoretical background in relation to the literature and discusses potential determinants of the self-employment probability. Section 2.3illustrates the data used and presents a descriptive analysis of the Spanish labor market and all labor market flows over time (2005-2017). Section 2.4 describes the institutional setting of the unemployment benefit system in Spain, as well as the investigated labor market reform on which our identification strategy relies. In Section 2.5, we explain our estimation methodology and its underlying assumptions. Section 2.6presents results, whileSection 2.7discusses their interpretation (with respect to the theory) and potential policy implications (welfare). Section 2.8concludes.
4Current active labor market policies in Europe increasingly subsidize unemployed individuals to start their own businesses. Especially in Spain, such policies have been used to address the high (youth) unemployment rates after the economic crisis of 2007/2008. For instance, in 2013 the Spanish government launched the Strategy of Entrepreneurship and Youth Employment 2013-2016. This program aimed at promoting self-employment among the unemployed youth through reductions in social security contributions (González Menéndez & Cueto 2015).