1.4 Summary and discussion
1.4.3 External validity
My samples consist either of establishments in the European Union, of small es- tablishments in Germany or of workers in small German establishments. Hence, it is relevant whether results are externally valid beyond these cases (Cook, Camp- bell and Shadish, 2002). For generalization, it is important to account for relevant interactions of country-level as well as establishment- and worker-level variables with EPLP. There are at least four of those interactions which are important.
First, for the effect of EPLP on the behavior of firms, differential enforcement due to governance differences is relevant (e.g. Micco and Pages, 2007; Venn, 2009; Haltiwanger, Scarpetta and Schweiger, 2014). For instance, if corruption is high, legislation on employment protection might not be followed by the employers. Sim- ilar to aforementioned studies, I also consider governance indicators (government effectiveness, rule of law, control of corruption) in Chapter 3, too. I, however, do not find that countries with a low governance indicator have a differen condition- ing effect of EPLP compared to countries with a high indicator. This is explained by the fact that differential enforcement is specifically relevant in countries with very low enforcement, which is the case in e.g. developing countries (Venn, 2009). Employing a sample of developing and non-developing countries, Micco and Pages
(2007) have an average of -0.18 in the rule of law indicator, while in my sample of European countries the rule of law average is 1.26 on a scale from -2.5 to 2.5. Thus, I expect my results on the abolishment of EPLP and on the conditioning effect of EPLP on temporary employment to hold in countries with similar levels in the governance indicator but not necessarily in countries with low levels in this indicator, such as in developing countries.
Second, the effect of EPLP may also differ with respect to the level of EPLP itself. With reference to the literature on varieties of capitalism (Hall and Soskice, 2001; Estevez-Abe, Iversen and Soskice, 2001), coordinated economies (CME), e.g. Germany, Netherlands, and mixed market economies (MME), e.g. Spain, Italy, are usually characterized by more rigid labor markets with high levels of employment protections compared to liberal market economies (LME), e.g. Ireland, United States. Indeed, EPL as measured by the OECD was 65% smaller in LME compared to CME in the period from 1990 to 2002 (Hall and Gingerich, 2009). MME were even more regulated (1.65 times) compared to CME. Hence, abolishing EPLP in CME or MME makes a strong difference in firing costs for firms compared to LME. Germany is generally categorized as a CME and is characterized by relatively strong protection. Hence, in LME, I would expect only small effects of abolishing EPLP on firms behavior with regard to temporary workers.
Third, and related to the second, firm-specific skills - as well as industry- specific skills - are considered to be important in CME (Hall and Soskice, 2001). This is related to the level of EPLP and the rational behind the investment in human capital by employees (e.g. Estevez-Abe et al., 2001; Wasmer, 2006a; Was- mer, 2006b). Hence, the substitution of permanent workers by temporary workers might be limited (Hijzen et al., 2013). This, however, would less likely be the case in LME, which are assumed to rely more on general skills (e.g. Hall and Soskice, 2001; Estevez-Abe et al., 2001; Wasmer, 2006a; Wasmer, 2006b). Hence, the effect of abolishing EPLP on temporary workers is expected to be stronger in LME such as in the United States compared to Germany. The conditioning effect of EPLP is also expected to be stronger in those countries - keeping other institutions constant. Furthermore, larger firms might rely more heavily on firm- specific skills (Pfeifer, Sch¨onfeld and Wenzelmann, 2011). This may result in limits concerning the substitution of permanent workers by temporary workers in larger
firms. I would expect that the effect of abolishing EPLP is smaller in firms which are relatively large in Germany. In Chapter 3, I do not find that the conditioning effect of EPLP is driven by large or small firms.
Fourth, wage rigidity might be relevant for the effect of EPLP on temporary employment at the establishment-level as well as on life satisfaction (via temporary employment). According to Lazaer (1990), if wages are flexible, firms could at least compensate for EPLP related transfers, e.g. severance payments, by lower wages. This is empirically supported by Leonardi and Pica (2013). They show that an increase in EPLP has a negative effect on wages of new workers, specifically when they are in a weak bargaining position, e.g. young blue-collar workers or workers in labor markets with low employment rates. If firms can compensate EPLP by lower entry wages, numerical reactions are expected to be less strong (Leonardi and Pica, 2013; Lazaer, 1990). Hence, the effect of abolishing EPLP on temporary employment might be stronger in German establishments which are larger than those in the treatment group. This is because larger German firms employ, on average, more white-collar workers, and according to Leonardi and Pica (2013), blue-collar workers are associated with low bargaining power. As already mention, in Chapter 3, I do not find that the conditioning effect of EPLP is driven by large or small firms. Furthermore, the conditioning effect of EPLP for the effect of shocks and the effect of EPLP might be an upper bound estimate as other countries beyond the European Union and beyond Germany might have labor market institutions which are more supportive in terms of wage flexibility, i.e. less collective wage bargaining and lower unemployment benefits. In developing countries, for instance, effects are expected to be less strong.
I show that workload fluctuations are only positively related to temporary employment at the establishment-level if EPLP is sufficiently large. Next to the discussion on generalization above, I show that these results are robust with respect to country subsamples, to different periods (2009 and 2004/2005), as well as to sector subsamples. Thereby, this provides empirical support that one specific set of national institutions, one specific sectoral production function, or one specific year drives the results.
Finally, if the negative effect of abolishing EPLP on life satisfaction of tem- porary workers is due to the comparison to former temporary worker colleagues
who transitioned into a permanent job, the behavior of firms in terms of trans- forming temporary into permanent contracts is important for the effect of EPLP on life satisfaction. Hence, the aforementioned arguments apply to the results of the effect of EPLP on life satisfaction, too. Additionally, from the perspective of workers, the perception of individuals of income mobility might also play a role in the effect of EPLP on life satisfaction. For instance, Senik (2008) shows that the effect of social comparison (here income) is adverse in ”old” European countries but that the relation is positive in the United States as well as in post-transition countries. Hence, in countries with a high perceived mobility, abolishing EPLP might not have a negative but rather, a positive effect on well-being of temporary workers.
Chapter 2
Employment protection reform
effects on temporary employment
2.1
Introduction
In the public debate, employment protection legislation for permanent workers (EPLP) is considered to be a structural source of high incidences in temporary1
employment, a high youth unemployment rate and low transition rates from tem- porary to permanent work. Thus, reform proposals for labor market institutions of- ten suggest a decrease in EPLP. For instance, in Portugal, Spain, Italy, and Greece, EPLP was liberalized in the aftermath of the 2007 financial crisis (OECD, 2013b). Therefore, it is highly relevant to understand the effects of a decrease in EPLP on economic outcomes.
EPLP is generally modeled via firing costs of workers with a permanent contract (e.g. Bentolila and Bertola, 1990; Cahuc et al., 2012). Boeri (2011) predicts that by increasing these costs, the share of temporary employment increases, while the effect on the unemployment level is ambiguous. This finding is in line with papers employing within-country time and subgroup variation. Centeno and Novo (2012) show that the increase in EPLP has a positive effect on the share of temporary workers at the firm-level. Boockmann and Hagen (2001) show that a decrease in 1Temporary is referred to as fixed-term contract workers (FTC) or and/or to temporary agency
workers. The former holds a direct fixed-term contract with a firm, while the latter holds a contract with an employment agency and works on a fixed-term basis for a user firm.
EPLP has a negative effect on the binary decision to employ at least one temporary worker. Literature using within-country time variation but no subgroup variation, however, show that the effect of EPLP on temporary employment is less clear (Nunziata and Staffolani, 2007; Kahn, 2010).
Building upon this literature, we investigate for the first time the symmetry of the effect of symmetric EPLP reforms (increase, decrease) on temporary em- ployment. None of the aforementioned papers analyzed, whether a decrease in EPLP has a similar effect on temporary employment as an increase in EPLP. If a decrease in EPLP affects temporary employment less strongly compared to an increase, this should be accounted for in theoretical models as well as by policy makers when they design reforms in EPLP.
Germany provides the unique opportunity to investigate a symmetric increase and decrease in EPLP. In 1999, German EPLP increased for new hires in firms with 6 to 10 full-time equivalent employees (FTE) but not for other firms. In 2004, EPLP decreased for new hires for the same group of firms. Both reforms were almost perfectly symmetric. These reforms provide within-country time and subgroup variation which we employ in a difference-in-difference approach (DID). This became a standard tool in the evaluation of EPLP2 and was also applied
in the case of German EPLP reforms (e.g. Boockmann and Hagen, 2001; Bauer et al., 2007; Boockmann, Gutknecht and Steffes, 2008). By focusing on firms which are close to the threshold (6-12 full-time equivalent employees), we conduct the DID in a regression discontinuity design (Leonardi and Pica, 2013).
The main result is that symmetric EPLP reforms have asymmetric effects in terms of statistical and economical significance. The increase in EPLP in 1999 raised the share of fixed-term workers in firms with 6 to 10 FTE by 1.7 percentage points. The decrease in EPLP in 2004 decreased the share of fixed-term workers. These effects, however, were asymmetric. The effect of the increase in EPLP was more economically (58% versus 12% of the mean) and statistically significant. The sign of the effects as well as the asymmetric pattern is robust to different definitions of the treatment group dummy, different controls and different samples. The change in share of fixed-term workers was driven by changes in the number of fixed-term workers, while the employment level remains almost unaffected. For 2For instance, Leonardi and Pica (2013), Scoppa (2010), Martins (2009), Kugler and Pica (2008).
our main results, we do not find pre-reform trend differences between treatment and control group, which supports the common trend assumption.
The article is organized as follows: In the next section, we explain the insti- tutional setting with regard to EPLP and the degree of symmetry of the reforms in EPLP. In Section 2.3, we describe the theoretical and empirical background. A section on the empirical strategy follows. It presents the identification strategy, data sources, sample selection and the definition of the treatment group dummy as well as of the dependent variables. In Section 2.5, we present some descript- ive statistics and the difference-in-difference results, and a conclusion is drawn in Section 2.6.