Eur J Law Econ https://doi.org/10.1007/s10657-018-9583-x

The effect of occupational licensing deregulation on migrants in the German skilled crafts sector Petrik Runst1

 Springer Science+Business Media, LLC, part of Springer Nature 2018

Abstract Occupational licensing on the national level reduces labor market prospects of individuals with a low likelihood of fulfilling the licensing requirements. Such regulation has the potential to adversely affect the labor market integration of foreign-born citizens and can be an obstacle to the free movement of labor toward its most productive uses. Before the backdrop of increased levels of migration into Germany, and the discussion about harmonizing labor standards in Europe, this paper empirically examines the effects of the deregulation of occupational licensing in the German crafts sector on the proportion of migrants working in this sector. The results suggest that the deregulation has increased the proportion of migrants among self-employed as well as employed craftsmen in the fully deregulated trades. Keywords Occupational licensing  Migrants  Germany  Common market  Deregulation JEL Classification D45  K20  L51

Electronic supplementary material The online version of this article (https://doi.org/10.1007/s10657018-9583-x) contains supplementary material, which is available to authorized users. & Petrik Runst [email protected] 1

Economics Department, Institute for Small Business Research, University of Go¨ttingen, Heinrich-Du¨ker-Weg 6, 37073 Go¨ttingen, Germany

123

Eur J Law Econ

1 Introduction Occupational licensing represents a legal market entry barrier where a government mandates certain conditions under which professionals may enter. Individuals are typically required to undergo a specific training, pass a government test, or obtain a university degree. Occupational licensing has been justified by the intention to ensure certain quality standards. Certification, on the other hand, is an alternative means of achieving the same end. Under voluntary certification, professionals may choose to fulfill the necessary steps in order to get certified, and thereby obtain human capital and provide higher quality to customers, or they may choose not to pursue this route. Certificates lower the search costs for customers who try to differentiate between high or low quality suppliers (Fredriksen et al. forthcoming). While ‘‘economists have long favored certification over licensure’’ (Svorny 2000, 306; also see Domin and Marciano 2013) there are some theoretical reasons that speak in favor of licensure over certification in certain situations (Shapiro 1986; Svorny 2000; Law and Marks 2009; Philipsen 2009). First, negatives externalities may arise in connection to lower quality services. For example, improper installation of heating equipment may cause fires that could potentially affect neighboring buildings. Second, if asymmetric information problems are severe buyers may receive lower quality than expected (Law and Marks 2009). Finally, even without information asymmetries, nor negative externalities, certain costly forms of licensing may still come about as a result of rent seeking by public or private officials (see Ogus and Zhang 2006; Branstetter et al. 2014). Williams (1982) stated that licensing costs are disproportionately borne by minority groups. If minorities have a harder time to acquire the skills and educational credentials prescribed by licensing requirements, they will be underrepresented in licensed markets. Some recent empirical contributions confirm that labor market regulations in general (Feldman 2009; Feldmann 2003) and licensing regulations in particular (Dorsey 1983; Federman et al. 2006; Gomez et al. 2015; McDonald et al. 2015) affect minorities disproportionately by increasing the costs of entering licensed professions. One recent study however (Law and Marks 2009), finds that licensing has not made it harder for minorities who are covered by licensing laws in the United States to enter a number of professions during the late nineteenth and mid-twentieth century. The authors argue that licensing can even help minorities by overcoming asymmetric information problems. They state that ‘‘In all markets, segregated or otherwise, for which worker quality was difficult to ascertain, licensing may have provided information about quality that reduced the extent of statistical discrimination and increased employment opportunities for minority workers’’. It should however be noted that this information may just as well be provided by a credible certification scheme, and does therefore not strictly require a licensing scheme. Pashigian (1979), Kleiner et al. (1982) and Kleiner (2015) show that US interstate migration is reduced by licensing requirements; restricting the free flow of labor towards its most productive uses. While the latter set of papers examines within-country mobility, a similar process may unfold on the international level as

123

Eur J Law Econ

strict licensing rules could deter foreigners from entering the country. For example, while citizens of countries within the European Union are generally permitted to work in any of the member countries occupational licensing may still constitute a barrier to labor mobility. Hawthorne (2015) shows that a streamlined system of immigrant qualification recognition can significantly increase employment rates within the first years after arrival. Occupational licensing has been steadily increasing in Europe and North America after WWII. Particularly in the US, this expansionary tendency has not ceased after 1990 (Kleiner 2006; Kleiner and Krueger 2013). On the other hand, the common market area of the European Union has led to an attempt to harmonize and reduce national licensing policies in order to enable the free movement of labor between countries (European Commission 2013). In addition, there have been various national attempts to liberalize labor markets. In Germany, the so-called ‘‘Hartz Reforms’’ of 2004 selectively abolished licensing requirements for one group of occupations (B1-trades) in the German crafts sector, whereas keeping another group of trades in that sector under full licensure (AC-trades). A final group was partially deregulated (A-trades). This quasi experiment allows me to study the effects of deregulation on labor market participation of migrants. Of course, group assignment of trades is not random and there are multiple potential threats to causal inference that will be addressed below. In order to accurately sort crafts occupations into one of the three groups (fully deregulated/B1, partially deregulated/A, or regulated/AC), a selection algorithm has been developed that improves upon previously existing classification schemes (Rostam-Afschar 2014). A detailed description of the selection algorithm as well as a list of crafts occupations and their assigned group can be found in ‘‘Appendix’’. I use repeated cross-sections (2000–2010) of German microcensus data, which enables me to identify the groups of craftsmen and craftswomen covered by occupational licensing laws that have been differentially affected by the reform. I employ difference-in-differences regressions in order to estimate the effect of the reform on the likelihood of a migrant working in one of the three distinct occupational groups (AC, A, and B). The results suggest that licensure deregulation increases the share of migrants among both self-employed individuals and employees. The debate about licensing as a barrier to migration between countries has become more prominent in Europe in recent years because labor mobility is an essential component of the European Single Market (European Commission 2013). In addition, most developed countries experienced increased levels of migration due to a surge in armed conflicts (OECD 2017). Since self-employment constitutes an important component of migrant labor market participation (Baycan-Levent and Nijkamp 2009) entry barriers generated by licensure can have negative effects on migrant integration. The results presented in this paper suggest that deregulating labor markets contributes to increased levels of labor market participation of immigrants. In contrast, one recent paper by Law and Marks (2009) provides empirical evidence for positive effects of occupational licensing on the labor market integration of migrants under conditions of severe asymmetric information problems (about the quality of workers).

123

Eur J Law Econ

Finally, this paper does not investigate the potential negative effects of deregulation on native and migrant income (see Gomez et al. 2015) which would have to be weighed against the positive effects of increased labor market participation. Gomez et al. (2015) use Canadian panel data (SLID) in order to study the effects of licensing on migrant labor market outcomes. They find that immigrants are significantly less likely to be in a regulated profession than observationally equivalent non-immigrants. They also show, however, that earnings premia from being in a regulated profession are higher for migrants than native born Canadians. This paper contributes to the literature in the following manner. First, the effects of licensing deregulation on the integration of immigrant workers have hardly been studied. While there are a handful of studies on the effects of regulation, the reverse process has not merited sufficient attention. As we have seen above, the general trend towards more regulation does not generate many instances of deregulation reforms that can be examined. Second, the recent paper by Law and Marks (2009) presents historical evidence for a novel theoretical conjecture, i.e. that licensing can actually increase immigrant integration into the labor market. As this is opposed to the older view presented by Williams (1982), the topic ought to receive increased attention by empirical researchers. Third, the paper presents a novel classification scheme which allows researches to identify a section of the German economy (skilled crafts), thereby enabling future studies on the topic.

2 Ethnic diversity and the deregulation of the german trade and crafts code 2.1 The reform of the trade and crafts code In Germany, 93 trades belong to what is legally defined as the crafts sector, which comprises about five million professionals (Federal Statistical Office 2016). These trades are governed by the so-called Trade and Crafts Code (TCC, Handwerksordnung). Between 1953 and 2004, the legislation required the head of a crafts company to hold a Meister-degree. The Meister is the highest degree of vocational training. In order to acquire it a professional must first undergo basic training (typically 3 years) and become a Geselle. This first stage of training is comprised of practical learning in a private company as well as taking classes at vocational colleges. After having become a Geselle the crafts professional may take additional training and pass associated exams in order to become a Meister. This second stage of training involves occupation specific knowledge as well as knowledge about business management and pedagogy (because Meisters are permitted to train craftsmen) (Mueller 2014). The regulation has changed after 2004 as part of the so called ‘‘Hartz-Reforms’’ intended to decrease unemployment, whereby 53 so-called B1-trades such as brewers, interior decorators and musical instrument makers are now fully deregulated and no longer subject to any educational requirements (HwO §7.1). Instead, individuals can obtain the Meister-certification voluntarily.

123

Eur J Law Econ

Some trades such as bakers, butchers and car mechanics have been partially deregulated (A-trades hereafter), meaning that experienced employees without a Meister-degree are permitted to start a business (HwO §7b, Altgesellenregel). An experienced employee is defined as a person who has worked in the field for 6 years or more. In addition, an experienced individual must have worked in a managerial position for 4 years or more. A managerial position is defined as a position with executive decision-making power. A potential business owner without a sufficient degree may also hire a company manager who possesses a Meister degree in order to start the business (Betriebsleiterregelung). The decision whether or not an A-trade business can be opened by an experienced craftsman without a Meisterdegree lies with the crafts chambers, i.e. the representatives of the incumbent firms. Overall, the extent of the deregulation in the A-trades must be regarded as much more modest that in the fully deregulated B1-trades. The political process through which certain crafts trades have been sorted into the fully or partially liberalized groups (B1 and A) can be reconstructed to some degree. The governing coalition proposed the bill about the reform of the trade and crafts code in 2003 (Bundestag 2003a) after which it was criticized and rejected by the Federal Assembly, the upper house of the German parliament (Bundestag 2003b). The original bill exclusively relied on the criterion of hazardousness. If a trade was deemed to be potentially hazardous to third parties it would remain fully regulated or would only be partially deregulated. The critical statement by the Federal Assembly however, introduced the additional criterion of vocational training, according to which a trade would not be fully deregulated if it made significant contributions to vocational training in Germany. Because of this gridlock between parliament and assembly, the bill was finally discussed in the mediation committee on December 10, 2003 after which many of the trades intended for deregulation would remain regulated or would be only partially deregulated. The Federal Administrative Court has upheld the criteria of hazardousness and vocational training as a sorting mechanism in 2011 although the minutes of the negotiations provide evidence for interest group lobbying1 (see Bundestag 2011; Bulla 2012). Since the reform only affected certain occupations, the effects of removing occupational licensing (and replacing it with a certification scheme) on migrants can be examined. The fully deregulated trades (B1) and the partially deregulated trades (A) represent the treatment groups. The still regulated trades (AC) and all non-crafts occupations represent the control groups. See Table 1 for a summary of the treatment and control groups. We expect the reform impact to be more pronounced for treatment group B1 than A because of the differential degrees of deregulation.

1

One member of parliament made the following statement concerning a trade which was intended to be deregulated: ‘‘Surgical device mechanics [Chirurgiemechaniker] play an important role in my local election district. Many of them vote for SPD [Social Democratic Party]’’ (Bulla 2012).

123

Eur J Law Econ Table 1 The natural experiment Group

Requirements pre-2004

Requirements post-2004

Group

AC-trades

Meister

Meister

Control

A-trades

Meister

Meister

Treatment

Exceptions: Altgesellen, Betriebsleiter B1-trades

Meister

No requirements

B2-trades

No requirements

No requirements

Treatment –

Non-crafts

Varied

Varied

Control

This table describes the requirements for starting a business in the German crafts sector before and after 2004. A full list of trades for each category can be found at the Federal Association of Skilled Crafts (ZDH, Zentralverband des Deutschen Handwerks). B2-trades could, in addition to AC-trades, theoretically serve as a control group. However, individuals within B2-trades cannot be clearly identified in the data set

2.2 Theoretical background and hypotheses Occupational licensing could affect the share of migrants in the German crafts sector in two ways. The first (direct) mechanism pertains to self-employed individuals. The second (indirect) mechanism pertains to employees. The lower educational attainment of migrants is one of the primary obstacles for entering the labor market (Borjas 2014; Constant und Zimmermann 2006, 297). Migrants who desire to work in the crafts sector in Germany face the problem that the specific type of training (dual-training system) which leads to a crafts degree (Geselle or Meister) is not available in most countries of origin. As a result, Migrants in Germany could not immediately start a business in the regulated parts of the crafts sector before 2004. If migrants decide to undergo training after they have arrived in Germany, they may be disadvantaged by language deficits and insufficient schooling prerequisites, which may obstruct their way into the training system. One may argue that German speaking immigrants from Austria and Switzerland may find it easier to obtain the Meister degree than other immigrants. However, our sample contains only a negligible fraction of crafts professionals from these two countries (\ 1%). In addition, the annual education report provided by the Federal Statistics Office (Bildungsbericht 2016, 173–178) as well as the ministry for migration and refugees (BAMF 2008) show that migrants are more likely to study in lower tier secondary schools (‘Hauptschule’) and are less likely to study in middle or upper tier secondary schools (‘Realschule’ and ‘Gymnasium’) during the period under consideration. We thus expect migrants to have a lower probability of having a Meister degree. In fact, according to the nationally representative data set used in this analysis (described below), only about 5% of all migrants in the crafts sector hold a Meister degree, whereas about 14% of German craftsmen do. These statistics are almost identical to what can be found in other representative data sets (Mu¨ller 2015, 93).

123

Eur J Law Econ

Thus, before 2004, migrants have been disproportionally excluded from entrepreneurial activity in the crafts sector. After the removal of these entry barriers the proportion of migrants among self-employed craftsmen can therefore be expected to rise. Hypothesis 1 The deregulation of occupational licensing in the German crafts sector caused an increase in the proportion of self-employed migrants. As the deregulated crafts market exhibits lower entry barriers the number of companies will increase. This basic theoretical prediction has been supported by recent empirical work (Rostam-Afschar 2014; Runst et al. forthcoming). The increase in market entry is mostly caused by companies, which are owned by craftsmen without a Meister degree as they were the ones previously not permitted to enter (Mueller 2014, 2015). If hypothesis 1 is correct, and the proportion of craft companies owned by migrants rises, social network effects can lead to an increase in the share of migrant employees as well. This is because migrant business owners are most likely acquainted with other migrants and sociological studies suggest that social networks play an important role in job search processes (see Granovetter 1973; Bian 1997). We can expect the increase in the share of migrant employees to be more pronounced among untrained individuals than highly trained individuals (i.e. Meisters) as the latter group was not restricted to enter the market before 2004. Hypothesis 2 The deregulation of occupational licensing caused an increase in the proportion of migrants among crafts employees. The percentage of migrant worker has been used as a measure of integration. An increase in this statistic after deregulation can been interpreted as evidence for the explanation presented above, i.e. that regulation represents labor market barriers for immigrants due to insufficient training. However, an additional mechanism may be at work. The deregulation may also cause German worker to selectively leave the deregulated sector, with immigrant workers replacing them. This is likely to be caused by a fall in wages due to the removal of entry barriers. Despite methodological problems concerning the classification of crafts occupations, there is some empirical evidence for a fall in wages (Lergetporer et al. 2016; Damelang et al. 2018). Thus, an increase in the proportion of migrants in the crafts, as indicated by hypothesis 1 and 2, could be explained by both insufficient training and a selective withdrawal of German workers as a result of lower wages.

3 Data and methods 3.1 German microcensus data and sample design The empirical analysis is based on German microcensus data, provided by the Federal Statistical Office. It is an annual and representative 1% sample of all households in Germany. The questions are designed to gather demographic and labor market information about all individuals with a legal residence permit in Germany. The survey does not follow the same individuals each year, thus it is

123

Eur J Law Econ

organized as a repeated cross section. Only about 15% of working adults are craftsmen, but since the annual scientific use files contain about 490,000 individuals, the sample size is nevertheless sufficiently large. The mandatory nature of the census survey guarantees a low rate of item-non-response for most questions. I use pooled data for the years of 2000 until 2010. As this analysis pertains to labor market participation, individuals younger than 18 or older than 66 are excluded from the sample. Furthermore, all non-craftsmen are dropped (except for the robustness specifications in which non-craftsmen serve as the control group). After these preliminary steps have been taken, the sample contains about 25,000 craftsmen of working age per year. The data set contains information on the migration background of individuals, including the year of immigration into Germany, their nationality, and whether they possess a German citizenship. For the purpose of the present analysis, migrants are defined as individuals who immigrated to Germany during their life-time, regardless of whether they have acquired the German citizenship. Thus, this category includes European and non-European immigrants. Three quarters of immigrants in the sample arrived after 1980 (see Fig. 1). There are three categories that could potentially serve as a control group. First, six mostly health related crafts trades (AC hereafter) remain fully regulated. The disadvantage of this control group relates to its small size, which is why all regressions results are validated by an additional specification with an alternative control group. Second, the so called B2-trades have never been subject to licensing regulations. However, individuals within B2-trades cannot be clearly identified in the data set (see ‘‘Appendix ’’ online for details on the crafts classification), i.e. any attempt to generate a B2-group will necessarily include a large proportion of individuals not working in the crafts, generating a mixed group of crafts and noncrafts individuals. Finally, individuals working in non-crafts occupations can be used as a control group. This group is quite large and heterogeneous. Without adjusting for missing variables, there are roughly 190,000 working non-crafts individuals in the sample per year, compared to about 25,000 working crafts individuals. Because of its large size, there may be a number of confounding factors which affect the migrant share in certain subgroups of the non-crafts group. On the other hand, these factors may cancel each other or they may not be sizable enough to affect the large non-crafts group overall. 3.2 Distinguishing crafts and non-crafts occupations If one is to assess the implications of a particular policy change in the crafts sector, it is paramount that the treatment groups only comprise individuals within this sector. It must not contain individuals in the agricultural, industrial or any other sector of the economy, all of which have not been directly affected by the 2004 reform of the trades and crafts code. The data set does not contain direct information on whether a professional works in the crafts sector. However, craftsmen can be distinguished from non-craftsmen on the basis of the occupational classification code (KldB1992).

123

Eur J Law Econ

Fig. 1 Histogram. Year of Immigration. (Craftsmen between age 18 and 66). Source: German microcensus, 2000–2010

Rostam-Afschar (2014) already developed a crafts-classification procedure based on occupation codes in the microcensus (KldB1992). I analyzed this list in detail because it constitutes an important attempt to make the microcensus data utilizable for studies focusing on the German crafts sector (see Runst et al. forthcoming). According to federal statistics office data on the overall population of craftsmen in Germany (Feuerhake 2012), less than 20% of all craftsmen are situated in the B1trades. However, according to RA, the B1-trades comprise about 38% of all crafts individuals (RA 2014, Table 2, p. 1083). After thorough examination it must be concluded that the demarcation chosen by RA is too broad: while it certainly includes many of the occupations that German craftsmen would practice, it also contains a large proportion of non-crafts individuals who are unaffected by the policy reform (see ‘‘Appendix’’). This inclusion of non-craftsmen is especially pronounced in RA’s fully deregulated B1-group and explains its disproportionately large size in his sample. Therefore, I rely on an improved classification system that is still based on the occupation codes of the microcensus (KldB1992) yet also uses additional information, allowing me to exclude a number of non-craft workers by following the criterion that a crafts-occupation must comprise 60% or more craftsmen. The selection algorithm which identifies craftsmen and sorts them into one of the three categories (control: AC; treatment: A, B1) is described in detail in ‘‘Appendix’’. Descriptive statistics for key variables, such as the proportion of women or the proportion of individuals in the A or B1 sector, have been generated for the two different classification schemes and were compared with crafts statistics provided by the German Federal Statistics Agency (Statistisches Bundesamt), which is based on administrative accounts on all existing (crafts) companies. According to our

123

Eur J Law Econ Table 2 Weighted averages by treatment and control groups in pre- and post-reform samples A

B

AC

Treatment group

Treatment group

Control group

Pre

Post

Pre

Post

Pre

Post

Migrant

0.12

0.17

0.15

0.23

0.06

0.09

Self-employed

0.11

0.13

0.16

0.20

0.18

0.18

Female

0.10

0.11

0.36

0.35

0.41

0.45

Age

37.28

38.28

41.16

41.99

37.03

38.69

Realschule

0.22

0.28

0.19

0.24

0.44

0.48

Fachabitur

0.02

0.03

0.03

0.04

0.06

0.08

Abitur

0.03

0.04

0.07

0.09

0.15

0.18

Geselle

0.60

0.65

0.56

0.62

0.49

0.56

Meister

0.14

0.14

0.10

0.09

0.25

0.27

University degree

0.01

0.01

0.02

0.03

0.02

0.02

5k–20k

0.21

0.21

0.21

0.21

0.21

0.20

20k–100k

0.23

0.25

0.25

0.26

0.26

0.27

100k–500k

0.12

0.12

0.14

0.14

0.14

0.14

[ 500k

0.08

0.09

0.10

0.11

0.10

0.12

92,005

125,396

16,756

24,129

4243

5869

Secondary schooling

Professional qualification

City size

Observations

All number are weighted by survey weights provided in the microcensus data set. Cleaners are omitted from the sample of B-trades. Realschule enables individuals to enter vocational training, whereas Fachabitur and Abitur permit students to obtain university education. A Geselle-degree is acquired after about 3 years of vocational training. The Master-degree represents the highest form of vocational training in the crafts sector

novel classification, the share of individuals in B1-trades is 16%, thus being much closer to the share of individuals in B1-trades in the overall population of craftsmen according to official statistics (Feuerhake 2012). 3.3 The development of migrant shares over time and potential confounding policies Figures 2, 3 and 4 plot the share of migrants for the fully deregulated trades (B1), the partially-deregulated trades (A), the still regulated trades (AC) and all non-crafts occupations. Figure 2 displays the proportion of migrants among self-employed craftsmen for the four groups. First, let us turn toward the fully deregulated B-trades. The share of migrants between 2000 and 2004 is below 10%. It moves up to more than 20% in the year 2010. A similar development can be observed in the partially deregulated A-trades, where the share of migrants increases from roughly 5–12%. The still

123

Eur J Law Econ

Fig. 2 The share of self-employed migrants over time. Source German microcensus data, 2000–2010

Fig. 3 The share of migrant employees over time (untrained employees). Source: German microcensus data, 2000–2010

regulated AC-trades and the non-crafts occupations do also display an increase in the share of migrants. This change may be the result of the expansion of the European Union, which permits eastern European individuals to migrate to

123

Eur J Law Econ

Fig. 4 The share of migrant employees over time (trained employees only). Source: German microcensus data, 2000–2010

Germany if they start a business,2 which is why some of the regression specifications below drop observations if the individual was born in Eastern Europe and migrated to Germany in the year 2004 or after. However, even after removing this group, the shift toward higher shares of migrants after 2004 is still visible across all four groups (AC, A, B1, and non-crafts). Other policy instruments were introduced as a part of the ‘‘Hartz Reforms’’ and came into effect in the year 2003 (Existenzgru¨ndungszuschuss) and 2006 (Gru¨ndungszuschuss). These policies granted subsidies to entrepreneurs, and were targeted at formerly unemployed individuals (see Caliendo and Ku¨nn 2011, pp. 313–314; but also Rostam-Afschar 2014, p. 1076). As these subsidies were found to have significant effects on entrepreneurship (Caliendo and Ku¨nn 2011) they can potentially explain the general increase in migrants among the selfemployed across all four occupation groups in Fig. 2. As the unemployment rate among migrants in Germany is higher than among native-born individuals (BA 2016) migrants benefit more strongly from a targeted entrepreneurship subsidy for the unemployed. The subsidies may also confound the analysis of causal effects of the crafts deregulation if the subsidies do not uniformly affect all four groups (AC, A, B1, and non-crafts) but systematically affect treatment groups relatively more strongly (A, B1) than control groups (AC, non-crafts). However, according to Mu¨ller (2006, p. 73) data collected by the German crafts chambers show that the proportion of 2

After the year 2011, migration was fully liberalized. Individuals who are born in any member state of the EU may freely relocate to any other member state and can be legally employed. After 2004, a person was permitted to engage in entrepreneurship in Germany if he or she is a citizen of a country within the European Union.

123

Eur J Law Econ

subsidized entrepreneurs is the same across different crafts groups. May-Strobl (2005) estimates that the proportion of subsidized entrepreneurs in the fully deregulated B1-trades was actually smaller than the one in AC and A together, even though B1-trades display the most pronounced increase in the share of migrants. Thus, the subsidies do not increase self-employment relatively more in treatment groups and do therefore not confound the analysis. In contrast, as subsidies seem to have been less effective in treatment groups one could argue that the DiD estimates are a lower bound estimate of the true reform effect. Figure 3 displays the share of migrants among untrained employees. These professionals do neither possess a Geselle-degree, which is typically obtained after a 3-year apprenticeship, nor do they have an advanced Meister-degree. In fully deregulated B1-trades, the share of migrants increases by almost 20 percentage points between 2004 and 2010. The partially deregulated A-trades display an increase by 10 percentage points during the same time period. Finally, the still regulated AC trades and non-crafts occupations display a small increase in the share of migrants as well. The share of migrant employees amongst trained employees is plotted in Fig. 4. The proportional increase in the share of migrants applies to all groups. The impact of the crafts deregulation on the migrant share of trained employees is therefore doubtful. As will be seen below, the regression analysis confirms the intuition that the reform increases the migrant share among untrained employees, whereas the evidence for an increasing migrant share among trained employees is much weaker. Legislators selected occupations into group AC, A, and B1 according to two criteria. According to the parliamentary protocols, occupations deemed hazardous or accident-prone and occupations with a large number of trainees have not been fully deregulated. They are categorized as either AC or A (Federal Assembly 2003; Mueller 2006). Thus, treatment is not random and a simple cross section estimation procedure cannot be applied. Instead, I run difference-in-differences specifications. This will be appropriate in the case of non-random treatment as long as the difference in migrant shares between groups remains constant over time in the absence of the treatment effect. 3.4 Estimation procedure The empirical strategy exploits the deregulation of occupational licensing in 2004 as a quasi-experiment, keeping in mind that the lack of true randomization poses several threats to causal inference that need to be dealt with. I use data from 2000 to 2010 for the two occupational treatment groups (A and B1). AC-trades have not been deregulated and serve as the control group. In other specifications, the noncrafts sector serves as the control group. A difference-in-differences (DID) approach is employed. DID regressions contrast the difference in Yi across groups, before and after the policy change. This approach does not require panel data. Repeated cross section data works well if the group composition remains identical across time periods (Blundell and Costa 2009). The present analysis follows Rostam-Afschar (2014)

123

Eur J Law Econ

and Runst et al. (forthcoming), who have also employed DID methods in order to estimate the effect of crafts deregulation on new business formation. The validity of the common trends assumption determines whether we can interpret the results as causal or merely suggestive. This question is discussed in detail in Sect. 4.2. The dependent variable of interest (Yi) is the likelihood of being a migrant. It is equal to one if an individual migrated to Germany at some point in their life and zero otherwise. While it may appear peculiar to estimate the determinants of being a migrant as a person cannot become a migrant if born in Germany, the method is chosen to identify the factors that positively (or negatively) affect the likelihood that a randomly selected individual is an immigrant. Thus, a positive and significant coefficient of the B1 and Post-2004 interaction term signifies that (the difference in) the probability that a randomly selected individual in the B1 trades to be a migrant has increased after 2004 (in relation to the control group). In an alternative specification, the individual level data is collapsed by occupation in order to obtain a more familiar specification in which each row represents one occupation and year. The individual’s characteristics, as for example the set of dummy variables which captures the highest obtained education and professional qualification, are thereby regenerated as group means (i.e. the share of individuals with a Meister degree by occupation). The sample size is reduced to 60 crafts observations (occupations) per year. The results of the occupation level specifications are very similar to the results of the individual level specifications. A linear probability model is used. The dummy variables Ai ; Bi denote an individuals’ affiliation to one of the treatment groups, whereas the dummy Post2004i indicates an individual in microcensus wave 2005 up to 2010. The average treatment effect on the treated group is given by the coefficient of the interaction between the treatment and post-policy dummy ðb5 ; b6 Þ. This interaction represents the comparison of the difference across groups before and after the policy change. Yi ¼ b1 þ b2 ðBi Þ þ b3 ðAi Þ þ b4 ðPost2004i Þ þ b5 ðBi  Post2004i Þ þ b6 ðAi  Post2004i Þ þ b7 Xi þ ei ;

ð1Þ

where ‘‘i’’ represents individuals, Xi represents a number of control variables. It includes age (also age2, in order to allow for a non-linear relationship), a number of dummy variables for each type of secondary schooling degree (such as Realschule, Fachabitur, and Abitur), a dummy denoting the completion of basic (Geselle) and advanced (Meister) vocational training, and a dummy for having obtained a university degree. Furthermore there are dummy controls for all occupations, state of residency, city size, crafts branch and years. The errors are clustered by occupation as suggested by Bertrand, Duflo and Mullainathan (2004). Age can be expected to exert a negative influence on the proportion of migrants as the overall proportion of migrants has increased over time. Thus a randomly selected individual who is older is likely to be a migrant. As was discussed above, since migrants’ educational credentials are somewhat lower than that of individuals born in Germany, the expected association between education and the likelihood

123

Eur J Law Econ

that a randomly selected individual is a migrant is also negative. I also expect that the more recent year dummies are positively related to the dependent variables as the number of migrants has increased over time. The impact of the state dummy variables reflects the geographic dispersion of migrants and we can expect a higher proportion of migrants in states that are relatively more affluent. Similarly, migrants are less likely to settle down in rural areas as opposed to larger cities and this relationship should be captured by the city size dummy variables. There is no reason to expect a systematic relationship between the female dummy variable and the independent variable. 3.5 Descriptive statistics Table 2 displays the weighted averages for all variables by treatment group and preand post-reform period. The post-reform share of migrants is higher than the prereform share in treatment group A (by 5%) and B (by 8%). In contrast, the share of migrants in control group AC rises only by 3 percentage points. The share of women is lowest in A-trades, which includes the construction sector, and highest in ACtrades, which are mostly health-related (opticians, orthopedic shoe makers, dental technicians, etc.). Secondary schooling and Meister-training is generally higher among AC-trades. Rates of self-employment increase in both treatment groups in the post-policy period as a direct result of lower entry restrictions (see RostamAfschar 2014; Runst et al. forthcoming). The proportion of professionals with basic vocational training (Geselle) appears to increase uniformly across all three categories in the post-policy period.

4 Results and robustness checks 4.1 Regression results Table 3 displays the regression results for self-employed individuals. Overall, the interaction term between treatment group and the post-2004 dummy is statistically significant, suggesting that the deregulation of occupational licensing has led to an increase in the proportion of self-employed migrants in the crafts sector. Specification (1) shows an increase of 8 percentage points in fully deregulated B-trades and increase of 4 percentage points in partially deregulated A-trades. However, the interaction term ‘A 9 Post’ does only display significant coefficients if control group AC is used. Thus the results are not robust and it must be concluded that the partial deregulation did not affect the migrant share among the selfemployed in A-trades. The coefficient for the interaction term (B 9 Post) diminish by one percentage point if all control variables are included in specification (2). It is important to display results without cleaners because it is doubtful whether this occupational category belongs to the crafts sector (see ‘‘Appendix’’ for more detail). As cleaners are omitted from the sample in specification (3), the treatment effect for B-trades and A-trades become lower (0.05 and 0.03 respectively) but remain significant.

123

Eur J Law Econ Table 3 Regression results: self-employed craftsmen No controls

All controls

No cleaners

No cleaners and no post-2004 migrants

Coeff. (1)

SE

Coeff. (2)

SE

Coeff. (3)

SE

Coeff. (4)

SE

B1 9 Post

0.08**

0.03

0.07***

0.02

0.05**

0.02

0.05***

0.02

A 9 Post

0.04**

0.02

0.03*

0.02

0.03*

0.02

0.03*

0.01

Post

0.02*

0.01

0.07***

0.02

0.07***

0.02

0.06***

0.01

B

0.08**

0.03

0.15***

0.03

- 0.05**

0.02

- 0.05**

0.02

A

0.03*

0.01

0.00

0.01

0.01

0.01

0.01

0.01

Female

- 0.03

0.02

- 0.05**

0.02

- 0.04**

0.02

Age

- 0.02*

0.01

- 0.01

0.01

- 0.01

0.01

Age squared

0.00

0.00

0.00

0.00

0.00

0.00

Realschule

- 0.03***

0.01

- 0.03***

0.01

- 0.03***

0.01

Fachabitur

- 0.03**

0.01

- 0.04***

0.01

- 0.04***

0.01

Abitur

- 0.01

0.02

- 0.02*

0.01

- 0.02**

0.01

Geselle

- 0.11***

0.02

- 0.10***

0.03

- 0.10***

0.03

Meister

- 0.15***

0.03

- 0.14***

0.03

- 0.14***

0.03

University

- 0.03

0.03

- 0.03

0.03

- 0.03

0.03

Controls Occupation

No

Yes

Yes

Yes

State

No

Yes

Yes

Yes

City size

No

Yes

Yes

Yes

Branch

No

Yes

Yes

Yes

Cleaners

Yes

Yes

No

No

Post-2004 migrants

Yes

Yes

Yes

No

2

R (%)

2.80

16.58

15.12

14.65

N

35,679

35,679

34,621

34,560

Sample contains

*, **, ***Denote a 10, 5, and 1% significance level. German Microcensus data for the years 2000–2010 has been used The dummy ‘B’ is equal to one if the individual’s occupation belongs to the fully deregulated treatment group, ‘A’ stands for the partially deregulated treatment group The control group is ‘AC’, i.e. all occupations that have not been deregulated at all (the control group). Robust standard errors, clustered by occupation, are reported in the table

It may be objected that the inclusion of Eastern European countries in the EU after 2004 has caused the increase in the proportion of migrants in the crafts. This would only be problematic if it affected treatment and control groups differently. While the free movement of employees did not initially apply to eastern European countries, eastern Europeans did already then have the right to work in Germany if they started a business. Therefore, specification (4) omits all East European migrants who entered Germany after 2004, and all cleaners from the sample. The

123

Eur J Law Econ

results remain identical to the ones in specification (3). It can be concluded that, in accordance with hypothesis 1, the deregulation of occupational licensing increased the proportion of migrants among self-employed craftsmen by three percentage points in the case of partial deregulation and five percentage points in the case of full regulation. Table 4 displays regression results for self-employed craftsmen by gender and hours worked. The upper part of the table shows results if control group AC is used, whereas the lower part uses a non-crafts control group. There are robust results in the case of part-time women and full-time men in B1-occupations. Relying on the lower estimates in the lower half of the table, the migrant share increases by 6 and 4.5%, respectively. As other studies find that female entrepreneurship is often motivated by a desire to balance commercial and family work (Minniti and Wim 2010) this may explain why migrant women enter part-time self-employment rather than full-time self-employment. Table 5 displays the results for self-employed individuals in the collapsed sample. The migrant share in treatment group B1 increases by 6.3%. In specification Table 4 Regression results (self-employed craftsmen) by gender and hours worked Part-time (less than 25 h/week)

Full-time (more than 30 h/week)

Male

Male

Coeff. (1)

Female SE

Female

Coeff. (2)

SE

Coeff. (3)

SE

Coeff. (4)

SE

0.053

0.062**

0.028

0.131**

0.062

Control group: AC B1 9 Post

–

0.184***

A 9 Post

–

0.113***

0.042

0.020

0.021

0.081

0.060

Post

–

- 0.031

0.044

0.087***

0.023

0.001

0.063

B

–

- 0.044**

0.021

- 0.023

0.020

0.086***

0.020

A

–

- 0.109*

0.056

0.002

0.016

0.102***

0.029

R2 (%)

–

10.9

16.6

21.1

N

701

1249

19,440

3599

Control group: non-crafts individuals B1 9 Post

- 0.071

0.044

0.059**

0.029

0.045*

0.023

0.021

0.019

A 9 Post

0.010

0.029

- 0.005

0.012

0.000

0.012

- 0.026***

0.004

Post

0.111***

0.017

0.115***

0.009

0.092***

0.008

0.088***

0.007

B

0.019

0.030

- 0.004

0.021

0.036**

0.015

0.091***

0.013

A

0.072***

0.023

0.045***

0.010

0.097***

0.008

0.069***

0.005

R2 (%)

14.03

11.32

13.2

10.3

N

7555

16,421

100,629

38,581

*, **, ***Denote a 10, 5, and 1% significance level. German microcensus data for the years 2000–2010 has been used. All regressions exclude migrants from eastern European countries who arrived in Germany in the year 2004 and after. Specification (1) for self-employed males is not reported due to low sample size, particularly for the control group. The interaction terms are not significant. Robust standard errors, clustered by occupation, are reported in the table

123

Eur J Law Econ Table 5 Regression results from the collapsed data set Self-employed

Employees

Coeff. (1)

Coeff. (2)

Coeff. (3)

Coeff. (4)

0.063***

0.054***

0.030***

0.049***

(0.014)

(0.020)

(0.0092)

(0.011)

0.037**

- 0.017**

- 0.0013

- 0.0019

(0.015)

(0.0079)

(0.0078)

(0.0041)

B

- 0.11

0.18***

0.35***

0.27***

(0.092)

(0.019)

(0.057)

(0.021)

A

- 0.049**

0.10***

0.19***

0.083***

dB1 9 dPost dA 9 dPost

(0.022)

(0.023)

(0.026)

(0.019)

Female

0.45**

- 0.00022

- 0.14**

- 0.017

(0.18)

(0.030)

(0.057)

(0.046)

Age

0.013

- 0.025

- 0.073***

- 0.0078

(0.015)

(0.022)

(0.019)

(0.010)

- 0.00021

0.00022

0.00095***

0.000082

(0.00017)

(0.00023)

(0.00025)

(0.00013)

- 0.048

- 0.024

0.0065

0.064*

Age squared Realschule

(0.050)

(0.029)

(0.043)

(0.034)

Fachabitur

0.034

- 0.048

- 0.16

- 0.15***

(0.098)

(0.045)

(0.13)

(0.055)

Abitur

0.025

- 0.035

0.10

- 0.071*

(0.070)

(0.037)

(0.083)

(0.042)

Lehre

- 0.19**

- 0.092***

- 0.083*

- 0.035

(0.078)

(0.027)

(0.043)

(0.034)

Meister

- 0.26***

- 0.15***

- 0.032

0.040

(0.070)

(0.029)

(0.074)

(0.034)

University

0.021

- 0.048

0.013

- 0.012

(0.18)

(0.034)

(0.26)

(0.0090)

Control group

AC

Non-crafts

AC

Non-crafts

2

R (%)

32.0

37.0

53.0

57.0

N

600

3208

624

3573

*, **, ***Denote a 10, 5, and 1% significance level. German microcensus data for the years 2000–2010 has been used. The data has been collapsed, i.e. each row no longer represents an individual. Instead one row represents an occupation (i) for one particular year (t). The dependent variable measures the share of migrants working in occupation (i) and year (t). Occupations with fewer than 30 individuals have been dropped. Frequency weights (i.e. the number of individuals in each occupation and year) have been employed in the regression. Robust standard errors, clustered by occupation, are reported in the table

(2), in which non-crafts individuals serve as the control group, the migrant share rises by 5.4%. The results for treatment group A are not robust. The migrant share increases by 3.7% in specification (1) but actually falls by 1.7% in specification (2).

123

Eur J Law Econ

Overall, there is evidence in favor of hypothesis 1. The full deregulation of the B1 crafts trades increased the share of migrants among the self-employed. The effect is visible mostly for part-time female and full-time male self-employed individuals. There is no robust effect in the partially deregulated A-trades. Let us turn to hypothesis 2. The results for the employee sample can be found in Tables 5, 6, 7 and 8. Table 6 contains the basic specifications using control group AC. Before omitting cleaners and without the full set of controls, the coefficients of the interaction term are equal to 0.02 for A-trades and 0.09 for B-trades [specification (1)]. After omitting cleaners and adding controls [specification (3)], the B-trade coefficient is lower (0.04) but still statistically significant and the

Table 6 Regression results: crafts employees No controls

All controls

No cleaners

No cleaners and post2004 migrants

Coeff. (1)

SE

Coeff. (2)

SE

Coeff. (3)

SE

Coeff. (4)

SE

B1 9 Post

0.09***

0.01

0.08***

0.02

0.04**

0.02

0.04**

0.02

A 9 Post

0.02**

0.01

0.01

0.01

0.01

0.01

0.01

0.01

Post

0.03***

0.01

0.06***

0.01

0.06***

0.01

0.05***

0.01

B

0.18***

0.03

0.22***

0.01

0.15***

0.02

0.15***

0.02

A

0.05***

0.01

0.11***

0.01

0.10***

0.01

0.10***

0.01

Female

–

- 0.05***

0.01

- 0.04***

0.01

- 0.04***

0.01

Age

–

0.06***

0.01

0.06***

0.01

0.06***

0.01

Age squared

–

0.00***

0.00

0.00***

0.00

0.00***

0.00

Realschule

–

0.01

0.02

0.00

0.01

0.00

0.01

Fachabitur

–

0.03

0.02

0.01

0.01

0.01

0.01

Abitur

–

0.09**

0.04

0.05***

0.02

0.05***

0.02

Geselle

–

- 0.14***

0.01

- 0.14***

0.01

- 0.14***

0.01

Meister

–

- 0.18***

0.02

- 0.18***

0.02

- 0.18***

0.02

University

–

0.18***

0.02

0.14***

0.02

0.14***

0.02

Occupation

No

Yes

Yes

Yes

State

No

Yes

Yes

Yes

City size

No

Yes

Yes

Yes

Branch

No

Yes

Yes

Yes

Cleaners

Yes

Yes

No

No

Post-2004 migrants

Yes

Yes

Yes

No

R2 (%)

5.24

17.20

11.67

11.62

N

252,859

252,859

209,366

209,269

Controls

Sample contains

*, **, ***Denote a 10, 5, and 1% significance level. German microcensus data for the years 2000–2010 has been used. Robust standard errors, clustered by occupation, are reported in the table

123

123

0.01

0.01

0.05***

0.10***

FachAbitur

0.00

Yes

Yes

State

City size

Branch

No

Yes

Cleaners

Post-2004 migrants

Sample contains

Yes

Yes

Occupation

Controls

Yes

No

Yes

Yes

Yes

Yes

Yes

No

Yes

Yes

Yes

Yes

0.06***

0.01

0.00

0.00***

0.06***

- 0.04***

0.10***

0.15***

0.05***

Yes

No

Yes

Yes

Yes

Yes

0.14***

0.02

0.01

0.01

0.00

0.01

0.02

0.01

0.02

0.02

0.03* 0.01

University

- 0.02***

- 0.07***

- 0.04

0.00***

0.08***

- 0.07***

0.16***

- 0.07***

0.07***

0.03 0.02

- 0.18***

0.02

0.01

0.01

0.00

0.00

0.01

0.01

0.01

0.01

0.06** 0.05***

Coeff. (4)

Meister

0.08***

0.02

0.00

0.00***

0.02***

- 0.04***

0.10***

0.02

0.05***

0.02 0.01

SE

[ 30 h/week

- 0.14***

0.01

0.02 0.01

Coeff. (3)

No training

Geselle

Abitur

0.01

0.00**

Realschule

0.01

0.01

0.01

Age squared

A

0.02

0.01

- 0.01

- 0.01**

B

- 0.03**

- 0.08**

Post

0.03

0.01

Female

0.02*

A 9 Post

Age

0.05*

0.02*

B1 9 Post

SE

Coeff. (2)

Coeff. (1)

SE

Intermediate training

Advanced training

Table 7 Regression results: crafts employees by training level and hours worked

0.03

0.02

0.01

0.02

0.01

0.01

0.00

0.01

0.01

0.01

0.01

0.01

0.01

0.02

SE

Yes

No

Yes

Yes

Yes

Yes

0.17***

- 0.18***

- 0.15***

- 0.05***

- 0.02

0.00

0.00***

0.08***

- 0.11***

0.02

- 0.14***

0.11***

0.02

0.07**

Coeff. (5)

\ 25 h/week

0.07

0.03

0.03

0.02

0.02

0.01

0.00

0.02

0.04

0.03

0.03

0.02

0.02

0.03

SE

Eur J Law Econ

17,233

- 0.01

dA 9 dPost

0.01

0.03 0.00

0.01

137,925

7.10

0.01

0.01 0.02**

0.03*

43,855

16.57

Coeff. (3)

No training

0.01

0.02

SE

0.00

0.01

196,639

11.54

Coeff. (4)

[ 30 h/week

0.01

0.01

SE

0.01

0.07**

10,143

16.68

Coeff. (5)

\ 25 h/week

0.01

0.03

SE

*, **, ***Denote a 10, 5, and 1% significance level. German microcensus data for the years 2000–2010 has been used. Advanced training corresponds to a so called Meister degree, which represents the highest crafts degree obtainable. It permits the holder to start a business in a regulated trade. Intermediate training corresponds to a Geselle degree. Robust standard errors, clustered by occupation, are reported in the table

0.02

dB1 9 dPost

Robustness check: non-crafts control group

6.38

N

SE

Coeff. (2)

Coeff. (1)

SE

Intermediate training

Advanced training

R2 (%)

Table 7 continued

Eur J Law Econ

123

Eur J Law Econ Table 8 Regression results (crafts employees) by gender and hours worked Part-time (less than 25 h/week)

Full-time (more than 30 h/week)

Male

Male

Coeff. (1)

Female SE

Coeff. (2)

SE

Coeff. (3)

Female SE

Coeff. (4)

SE

Control group: AC B1 9 Post

0.071

0.062

0.060*

0.036

0.028

0.023

0.060***

0.022

A 9 Post

- 0.000

0.037

0.025

0.019

0.026

0.022

0.012

0.016

Post

0.113**

0.044

0.115***

0.022

0.056**

0.022

0.067**

0.016

B

- 0.351***

0.069

- 0.180***

0.034

- 0.058***

0.014

0.188***

0.017

A

- 0.075**

0.033

0.115***

0.014

0.084***

0.012

0.098***

0.011

R2 (%)

24.2

13.0

11.2

11.3

N

3055

6298

167,275

22,7792

Control group: non-crafts B1 9 Post

0.061*

0.036

0.069***

0.018

0.005

0.009

0.051***

0.015 0.008

A 9 Post

0.010

0.014

0.012

0.008

0.003

0.007

0.002

Post

0.099***

0.008

0.100***

0.007

0.078***

0.005

0.070***

0.004

B

0.138***

0.026

0.151***

0.010

0.110***

0.007

0.185***

0.008

A

0.196***

0.008

0.070***

0.004

0.111***

0.004

0.104***

0.007

R2 (%)

15.4

10.7

12.5

9.83

N

48,884

269,275

982,871

636,282

*, **, ***Denote a 10, 5, and 1% significance level. German microcensus data for the years 2000–2010 has been used. All regressions exclude migrants from eastern European countries who arrived in Germany in the year 2004 and after. Robust standard errors, clustered by occupation, are reported in the table

A-trade-treatment effect is no longer significant. Finally, post-2004 eastern European migrants are again dropped from the sample in specification (4). Although most eastern European citizens were not permitted to work in Germany before 2011, some individuals entered the German labor market if they fulfilled certain conditions. For example, family members of self-employed foreigners, individuals who lived in Germany for more than 3 years and some individuals willing to undergo vocational training were already permitted to enter. The coefficients of the interaction terms are equal to 0.04 for B-trades and not significantly different from zero for A-trades. Table 7 displays results for an advanced training sample (Meister), mid-level vocational training (Geselle), and an untrained sample (specifications 1, 2, 3 respectively). It also shows results for a part-time sample (specification 4) and a full-time sample (specification 5). The results suggest that the deregulation of occupational licensing increased the proportion of migrants among untrained employees in both A-trades and B-trades (see specification 3). If we use the alternative control group of non-crafts individuals, these results remain robust (see bottom of the table).

123

Eur J Law Econ

Specification (1) seems to suggest that the reform also affected the proportion of migrants among highly trained employees, although the results are only significant at the 10% level. However, once we use the alternative control group of non-crafts individuals the coefficient is no longer significant. Specifications (4) and (5) show that the increased migrant share is visible only in the case of part-time employees in B1-trades. Table 8 provides further details. The upper part shows results for control group AC, the lower part for control group non-crafts. The two robust results which emerge from the table pertain to part-time and full-time female employees. Using the coefficients from the bottom part of the table the share of migrants increased by 7% for part-time and 5% amongst full-time female employees. There is no evidence for an increase in the migrant share among employees in A-trades. Let us return to Table 5, in which results are presented which are based on the collapsed sample. Specification (3) suggests that the migrant share among employees in B-trades increased by 5.6% while there is no increase for employees in A-trades. Specification (4) suggests that that the migrant share among employed craftsmen in B-trades increased by 5%. Again, there is no increase in the migrant share among A-trades. Overall, there is evidence in favor of hypothesis 2. The reform increased the likelihood of migrants to work as untrained employees in the crafts but not among highly trained employees (Meister) or mid-range training levels (Geselle). The results seem to be strongest in size and most robust in the case of female employees. Finally, Tables 9 and 10 explore further dimensions of the data set. The results in Table 9 are split between a sample of migrants who arrived not more than 5 years ago and a sample of less recent arrivals. The deregulation effect is only visible in the latter sample. This result dovetails with other findings about the slow pace of labor market integration of migrants in Germany indicated by long periods of initial unemployment (see Kogan 2010). Thus, despite the deregulation of occupational licensing, there seem to be other factors that hinder the integration of recent migrants into the labor market. Table 10 shows regression results by age at the time of immigration. The reform of occupational licensing has hardly had any influence on migrants who immigrated as children. Instead it increased the share of migrants who immigrated as adults. The results suggest the regulatory framework did not restrict the occupational choices of migrants who grew up in Germany as they, most likely, have had the chance to acquire the educational credentials that are necessary for entering the regulated part of the crafts sector. On the other hand, regulations did restrict occupational choices of adult migrants until the year 2004. 4.2 Difference-in-differences-assumptions and robustness checks In order to further assess whether the results of the previous section can be seen as causal or merely suggestive, the models are re-estimated, varying the specification and the definition of variables. While the reformed trade and crafts code came into effect in 2004, microcensus questionnaires were filled out in April already. Thus it is plausible to expect changes

123

Eur J Law Econ Table 9 Regression results by year of arrival Recent arrivals (5 years or less)

Older arrivals (more than 5 years since immigration)

Coeff. (1)

Coeff. (2)

SE

SE

Self-employed B1 9 Post

0.007

0.014

0.050

0.022**

A 9 Post

0.002

0.006

0.025

0.014*

Post

- 0.008

0.006

0.065

0.014***

B

- 0.024

0.014*

- 0.031

0.020

A

- 0.006

0.004

0.010

0.011

Female

- 0.012

0.004***

- 0.039

0.017**

Age

- 0.005

0.004

0.013

0.008

Age squared

0.000

0.000

0.000

0.000*

Realschule

- 0.007

0.002***

- 0.020

0.005***

Fachabitur

- 0.008

0.005***

- 0.030

0.011***

Abitur

- 0.001

0.004

- 0.024

0.009***

Geselle

- 0.032

0.009***

- 0.085

0.024***

Meister

- 0.038

0.009***

- 0.119

0.030***

University

- 0.017

0.007**

- 0.021

0.030

R2 (%)

6.4

12.1

N

32,134

34,100

Employees B1 9 Post

- 0.010

0.003***

0.043

0.019**

A 9 Post

- 0.008

0.003***

0.024

0.012*

Post

- 0.004

0.002*

0.066

0.012***

B

0.010

0.003***

- 0.066

0.014***

A

0.018

0.001***

0.075

0.008***

Female

- 0.013

0.003***

- 0.030

0.007***

Age

0.021

0.004***

0.055

0.006***

Age squared

- 0.001

0.000***

- 0.001

0.000***

Realschule

- 0.003

0.001*

0.005

0.007

Fachabitur

0.008

0.003***

0.007

0.008

Abitur

0.013

0.004***

0.051

0.014***

Geselle

- 0.037

0.006***

- 0.111

0.013***

Meister

- 0.043

0.007***

- 0.151

0.017***

University

0.065

0.022***

0.121

0.024***

R2 (%)

3.4

9.9

N

185,756

202,626

*, **, *** Denote a 10, 5, and 1% significance level. German Microcensus data for the years 2000–2010 has been used. Results are robust to marginally changing the cut-off value between recent and older arrivals. Robust standard errors, clustered by occupation, are reported in the table.

123

Eur J Law Econ Table 10 Regression results by age at immigration Immigrated as a child (age 17 or less) Coeff. (1)

Immigrated as adult (age 18 or more) Coeff. (2)

B1 9 Post

0.01*

0.05***

A 9 Post

0.01

0.02**

Post

0.03***

0.04***

Self-employed

B

- 0.01**

0.01

A

- 0.01

0.01

R2 (%)

5.5

12.6

N

35,456

36,437 0.05***

Employees B1 9 Post

0.01

A 9 Post

0.01

0.02**

Post

0.04***

0.02*

B

0.05***

0.08***

A

0.03***

0.07***

R2 (%)

6.1

12.1

N

232,619

240,017

*, **, ***Denote a 10, 5, and 1% significance level. German Microcensus data for the years 2000–2010 has been used. Results are robust to marginally changing the cut-off value between recent and older arrivals. Robust standard errors, clustered by occupation, are reported in the table

to be fully visible in the 2005 survey. In order to assess alternative cut-off points, and following Rostam-Afschar (2014), the year 2004 is dropped, or, alternatively, coded as belonging to the post policy period. Dropping the 2004 observations does not affect the regression results for the self-employed sample nor the employee sample, whereas re-classifying the year 2004 as belonging to the post-policy period lowers all coefficients by about one percentage point. These changes may be seen as evidence for a small adjustment period before changes in the proportion of migrant craftsmen become visible. I also test whether other confounding factors that might have affected the treatment, but not the comparison group, exist. Placebo tests are run by restricting the sample period to 2000–2004 and by pretending the policy intervention took place in 2003 (or 2004). Thus, I generate new interaction terms between the treatment groups and a post-2003 (or post-2004) dummy. If any of the interaction terms are significantly different from zero, we must conclude that additional factors might be at work which have selectively affected one of the three groups prior to the reform in 2004, thereby rendering the parallel trends assumption invalid. After rerunning the final specifications in Tables 3 and 6 it can be concluded that the risk of confounding factors appears to be small (see Table S5 in online supplementary material). None of the interaction terms is significantly different from zero.

123

Eur J Law Econ

The validity of a DID hinges on the common trends assumption. Visual inspection of Fig. 2 suggests that prior to the reform, the proportion of migrants moves roughly parallel. However, there is one exception. The mean share of migrants among the self-employed in AC-trades seems to roughly double between the year 2000 and 2001, after which it returns to a value closer to its original level of the year 2000. The development of the migrant share in the second control group (non-crafts) does not exhibit such an erratic behavior, which suggests that regression results can only be interpreted as causal effects in the specifications which use a non-crafts-control-group. The common trends assumption is also investigated by replacing the single interaction term (treatment group 9 the post policy dummy) with several interaction terms for each year after 2000 (Bx2001; Bx2002; etc.), also including year fixed effects and all control variables. If any of the interaction terms before 2004 are significant, the common trend assumption must be rejected. In addition, if the annual interaction terms after 2004 differ from each other, this would indicate changing treatment effectiveness over time. Such a situation could for example come about if individuals need time to adjust to the change in labor market regulation. The results of these regressions can be seen in Table 11.

Table 11 Regression results: common trends assumption Control group: AC

Control group: non-crafts

Self-employed sample Male and full time

Employee sample Untrained individuals

Self-employed sample Male and full time

Employee sample Untrained individuals

A

A

B

A

A

B

B

B

Interaction year 2000

–

–

–

–

–

–

–

–

2001

0.01

0.01

0.00

- 0.00

0.01

0.01

0.01

- 0.01

2002

- 0.02*

- 0.02

- 0.01

0.00

- 0.01

0.00

0.00

- 0.01

2003

0.01

- 0.01

- 0.02*

- 0.02

0.01

0.00

0.01

- 0.01

2004

0.02**

0.01

- 0.02

0.01

0.01

0.01

- 0.01*

- 0.01

2005

0.00

0.04

0.02

0.07**

0.00

0.04

0.00

0.03*

2006

0.03

0.03

0.02

0.06**

0.00

0.01

0.01

0.05**

2007

0.02

0.07*

0.01

0.04

0.01

0.06*

0.01

0.01

2008

0.02

0.05*

0.01

0.06**

0.01

0.04

0.00

0.03*

2009

0.05*

0.10**

0.04**

0.05**

0.02

0.09*

0.02*

0.01

2010

0.05**

0.09***

0.06**

0.09***

0.01

0.05*

0.00

0.04*

*, **, and ***Denote a 10, 5 and 1% significance level respectively. The self-employed sample is restricted to full-time and male (see Table 4, specification 3). The employed sample is restricted to untrained individuals (see Table 7, specification 3). Each line represents the interaction term for the treatment dummy (A or B) and one particular year (2001–2010). All control variables and year fixed effects have been included in the regression model. If the common trends assumption is valid, the pre2004 interaction terms must not be different from zero

123

Eur J Law Econ

The common trends assumption seems to be fully satisfied when we use the noncrafts-control group but the situation is not entirely clear for control group AC. There appears to be an increase in the proportion of migrants in A-trades in the year 2004. This is not problematic as the year 2004 must be seen as a transition year in which the deregulated rules are already in place. There are two incidents however, in which the pre-2004 interaction term coefficient is significant. Thus, the common trends assumption is not fully satisfied when we use control group AC, and the results cannot be interpreted as causal but merely suggestive. In contrast, when we use the non-crafts-control-group, the results can be interpreted as causal because none of the pre-reform interaction terms are significant. Therefore, I suggest interpreting all results carefully. Throughout the paper, I only present results as being robust, if they are significant in both the AC-control and non-crafts-control specifications. Table 11 also shows that the impact of the reform generally gets stronger as time progresses. In particular, there seems to be a 3-year lag for the self-employed group after which the deregulation effect becomes visible. There is no lag for the group of employees.

5 Conclusion and discussion The reform of the German trades and crafts code in 2004 removed occupational licensing requirements necessary to enter certain segments of the crafts sector. This paper empirically quantified the effects of this reform on migrants, which generally face higher hurdles to enter the labor market. Individuals that are less likely to obtain vocational training or university education are disproportionately adversely affected by licensing requirements. It is found that the reform of occupational licensing in 2004 led to an increase in the proportion of migrants among self-employed in the fully deregulated B-trades. The effect is visible mostly for part-time female and full-time male self-employed individuals. There is no robust effect in the partially deregulated A-trades. The reform also had an impact on the migrant share among crafts employees. The reform increased the likelihood of migrants to work as untrained employees in the crafts but not among highly trained employees (Meister) or mid-range training levels (Geselle). The results seem to be strongest in size and most robust in the case of female employees. Again, there is no robust effect in the partially deregulated A-trades but only in the fully deregulated B1-trades. The debate about occupational licensing has yet again become more prominent in recent years as the European Commission seeks to encourage labor mobility and economic growth by reducing national occupation regulations (European Commission 2013). In addition, Germany and other European countries have experienced increased levels of immigration during the last three decades. The successful

123

Eur J Law Econ

integration of foreign-born citizens becomes an increasingly important topic for political debate. My results suggest that occupational licensing has kept migrants from entering into crafts professions before the deregulation of 2004. The removal of occupational licensing has increased the integration-capacity of foreign born workers into the German labor market, whereby income inequalities between ethnic groups have been most likely affected as well. The analysis presented here did not address the question of whether the increase in the proportion of foreign-born citizen crowds out employment of German born citizens. As Koch and Nielen (2016) have found no positive employment effects of the 2004 reform, it can be stated that immigrants have replaced non-immigrants in the crafts sector to some degree. As the results have shown, this is particularly the case among untrained craftsmen. The analysis has also not addressed the question of whether migrants in the crafts have been affected by a decrease in income as a result of the deregulation. Indeed, there is some first empirical evidence for a fall in wages (Lergetporer et al. 2016; Damelang et al. forthcoming) due to the reform.

Appendix: Classification of crafts trades The following procedure was used to identify individuals who are working in the crafts sector by using the microcensus occupation codes (KldB1992). In a first step, information was gathered on all training occupations and their classification codes (KldB 1992) Training occupations are different from occupations but are nevertheless associated with a particular crafts trade. This was achieved by consulting the official classifications of the ZDH and the Federal Institute for Vocational Education and Training and included present as well as predecessor occupations (Bundesinstitut fu¨r Berufsbildung, BiBB 2012). In a second step, I used data provided online by BiBB concerning the information about how many apprentices within one occupational field are trained either within crafts companies or non-crafts (mainly industrial) companies. Subsequently, I computed a proportion of crafts apprentices within each occupational code. To exclude occupation codes with a high proportion of non-crafts workers, I used the information on the proportion on crafts trainees and dropped codes if this proportion was less than 60%. Lowering or increasing this cut-off point by up to 20% hardly affects the classification as most occupations contain either a very low or a high proportion of craftsmen. Observations were also removed if occupations could not be clearly marked as either an A or B occupation. This method is not error-proof as it assumes that the proportion of crafts trainees strongly correlates with the proportion of crafts employees. However, this method removes occupation codes from the analysis that most probably contain very low proportions of crafts workers. For example, while the KldB code 141

123

Eur J Law Econ

(‘‘Chemiebetriebswerker’’, chemical plant employee) may seem a good proxy for the B-trade of ‘‘Wachszieher’’ (candle maker), according to my results less than 1% of individuals in the occupation of chemical plant employee are actually craftsmen. The classification scheme implies that most of the individuals in that occupation are industrial workers such as chemical production specialists, chemical technicians or pharmaceutical technicians. In a last step, I scrutinized the occupation of building cleaners (KldB code: 934). The occupation comprises about 45% of all individuals in the deregulated B-trades in the microcensus dataset. Owing to its large size, it potentially biases any general conclusions about B-trades. After a thorough inspection, it is doubtful if the occupational group of cleaners in the microcensus data perfectly matches the TCC trade of cleaners. For example, while official company registration data by the Federal Statistics office of Germany points to a sharp increase in market entry in that trade after 2004 (Mueller 2006), no such trend can be established in the microcensus data. The proportion of selfemployed cleaners in the microcensus only increases from 1.6 (2004) to 2.3% (2011). Upon request, employees of the Research Data Centers of the German States confirmed our suspicion and suggested several other classification codes under which cleaners might be found, none of which can be identified as crafts trades based upon our classification scheme. According to the documentation for an older occupation classification system (KldB1975), there are about seven activity profiles coded as 933 or 934 (cleaners). The classification scheme in the microcensus (KldB1992) merges these codes into one code (934). According to the crafts classification scheme recently developed by the Federal Employment Agency (BAA 2014), only three of these seven occupations belong to the crafts sector. I therefore do not include cleaners in the analysis. Doing so, however, increases the effects size as well as the statistical significance of our results, suggesting that the share of migrants rose more strongly amongst cleaners (see Table 12).

123

123 303 804 307 372

315

Dental technician

Chimney sweep

Orthopaedic technician

Orthopaedic Bootmaker (also contains very few individuals from B occupation Bootmaker)

Hearing aid acoustician (also contains very few individuals from occupation Radio- and television technician) 100.00

100.00

100.00

100.00

100.00

261 392 510/511 484/441 332 501 145 481 264/267/268 391 101 287 482

Plumber Confectionist Painter and laquerer Oven and air heating manufacturer Ropemaker Joiner Mechanic for tyres and vulcanization Plasterer Installer and heating manufacturer Baker Stonemasons Coachbuilder Thermal and acoustic insulation fitter

485

Glazier

266

901

Hairdresser Refrigeration mechanic

312

310

Electrical technician Electrical technician

488

300

221

KldB code

Roof tiler

Gunsmith (also contains precision mechanic)

Precision mechanic

304

Dispensing optician

100.00

TCC trade title

KldB code

TCC trade title

Fraction of crafts trainees (%)

A

AC

Table 12 Classification of crafts occupations by Runst et al. (KldB1992 titles)

98.45

98.51

99.09

99.33

99.36

99.53

100.00

100.00

100.00

100.00

100.00

100.00

100.00

100.00

100.00

100.00

100.00

100.00

100.00

100.00

10.00

Fraction of crafts trainees (%)

Eur J Law Econ

487

Carpenter

442

Bricklayer and concretor

316

317

Communication technician

Electrical technician

461

Roadbuilder

131

466

Wellbuilder

Glass blower and glass apparatus builder

512 295

Surgical instrument mechanic (dropped, cannot be separated from cutter mechanic, 295)

Electrical engine manufacturer Painter and Laquerer

318 313

Automotive technician

441

401

Butcher

311

443

Scaffolder

Electrical technician

254

Metal worker

Bricklayer and concretor

281 506

Boat builder

282

Mechanic for agricultural and construction machinery Automotive technician

KldB code

TCC trade title

Fraction of crafts trainees (%)

TCC trade title

KldB code

A

AC

Table 12 continued

8.79

22.07

38.82

47.20

54.06

67.36

70.50

70.85

74.93

77.90

80.36

84.79

88.83

91.81

92.24

92.92

94.71

95.01

95.34

Fraction of crafts trainees (%)

Eur J Law Econ

123

123 359 483 378 302 305 354 374 185

Embroiderer

Tile, slab and mosaic layer

Furrier

Gold- and silversmith

Organ and harmonium builder

Milliner

Saddler

Basket maker, wood turner, wood carver, wooden toy maker

71.61

82.47

87.75

90.71

90.47

93.27

93.42

96.55 Printer

Metal and bell founder

Miller

Cooper

Glass finisher, precision optician

98.75

486

99.65

491

Flexographer

Screed layer

0.3

Embroiderer, weaver Instrument maker

Parquet layer, interior decorator

Wax Chantler

100.00 100.00

Cast stone and terrazzo maker

Vessel and equipment constructor

Brewer and maltster

Screen printer

Printer

141

Signpost builder

100.00

100.00

100

100.00

(This is a small crafts trade. Has been deleted bc of overlap with the industrial occupation of chemical production specialist)

259 839

Shutter builder/advertisment maker

305 294

Engraver

934

Cleaner (see ‘‘Appendix’’ for more details)

Luthier

837

Photographer

Galvaniser

305

Bowyer

100.00

TCC trade title

KldB code

TCC trade title

Fraction of crafts trainees (%)

B continued

B

Table 12 continued

171

201

435

423

135

174

173

305

341

112

252

421

175

234

KldB code

1.18

4.46

5.08

6.10

7.36

8.51

12.98

74.47

9.43

16.43

17.29

20.19

26.73

34.93

Fraction of crafts trainees (%)

Eur J Law Econ

351 308 295 121 931 502 514 305 358 178

Costume tailor

Watchmaker

Cutter mechanic (dropped, cannot be separated from surgical instrument mechanic, 295)

Ceramist

Textile cleaner

Model builder

Glass and China painter

Piano and harpsichord builder

Sailmaker

Bookbinder

35.38

41.47

50.00

54.38

55.54

58.10

69.61

70.50

80.08

80.56

TCC trade title

KldB code

TCC trade title

Fraction of crafts trainees (%)

B continued

B

Table 12 continued

KldB code

Fraction of crafts trainees (%)

Eur J Law Econ

123

Eur J Law Econ

References BAMF. (2008). Schulische Bildung von Migranten in Deutschland. Integrationsreport, Teil 1. Bundesministerium fu¨r Migration und Flu¨chtlinge, working paper 13. Baycan-Levent, T., & Nijkamp, P. (2009). Characteristics of migrant entrepreneurship in Europe. Entrepreneurship and Regional Development, 21(4), 375–397. Bertrand, M., Duflo, E., & Mullainathan, S. (2004). How much should we trust difference-in-difference estimates? Quarterly Journal of Economics, 119(1), 249–275. Bian, Y. (1997). Bringing strong ties back in: Indirect ties, network bridges, and job searches in China. American Sociological Review, 62(3), 366–385. Bildungsbericht. (2016). H-Bildung und migration. In Bildung in Deutschland. Ein indikatorengestu¨tzter Bericht mit einer Analyse zu Bildung und Migration, Bertelsmann, Bielefeld. ISBN 978-3-76395742-2. Blundell, R., & Costa, D. M. (2009). Alternative approaches to evaluation in empirical microeconomics. Journal of Human Resources, 44(3), 565–640. Borjas, G. J. (2014). Immigration economics. Cambridge: Harvard University Press. Branstetter, L., Lima, F., Taylor, L. J., & Venancio, A. (2014). Do entry regulations deter entrepreneurship and job creation? Evidence from recent reforms in Portugal. The Economic Journal, 124(577), 805–832. Bulla, S. (2012). Ist das Berufszulassungsregime der Handwerksordnung noch verfassungsgema¨ß? Gewerbearchiv, 12, 470–476. Bundesagentur fu¨r Arbeit [BA]. (2016). Hintergrundinformation. Auswirkungen der Migration auf den deutschen Arbeitsmarkt. Nu¨rnberg. ¨ nderung der Handwerksordnung und anderer Bundestag [German Parliament]. (2003a). Gesetz zur A handwerksrechtlicher Vorschriften. Bundestagsdrucksache, 15(1206), 4–46. ¨ nderung der Bundestag [German Parliament]. (2003b). Stellungnahme des Bundesrates zum Gesetz zur A Handwerksordnung und anderer handwerksrechtlicher Vorschriften. Bundestagsdrucksache 1. Bundestag [German Parliament]. (2011). Protokolle des Vermittlungsausschusses des Deutschen Bundestages und des Bundesrates fu¨r die 13. bis 15. Wahlperiode (1994 bis 2005): DVD-Edition der… mit Materialien zur Erschließung, C. H. Beck. ISBN 978-3406617997. Caliendo, M., & Ku¨nn, S. (2011). Start-up subsidies for the unemployed: Long-term evidence and effect heterogeneity. Journal of Public Economics, 95, 311–331. Constant, A., & Zimmermann, K. (2006). The making of entrepreneurs in Germany: Are native men and immigrants alike? Small Business Economics, 26, 279–300. Damelang, A., Haupt, A., & Abrahan, M. (2018). Economic consequences of occupational deregulation. Natural experiment in the German crafts. Acta Sociologica, 61(1), 34–49. Domin, J.-P., & Marciano, A. (2013). How much should professional markets be regulated? An introduction. The European Journal of Comparative Economics, 10(2), 153–157. Dorsey, S. (1983). Occupational licensing and minorities. Law and Human Behavior, 7(2), 171–181. European Commission. (2013). Communication from the Commission to the European Parliament, the Council and the European Economic and Social Committee on Evaluating National Regulations on Access to Professions, Brussels 2013, COM 2013/676 final 02 October 2013. Federal Assembly [Bundesrat]. (2003). Stenografischer Bericht, 795. Sitzung, Berlin, Friday Dec 19, 2003. Upper House of the German parliament. Federal Statistical Office. (2016). Produzierendes Gewerbe. Unternehmen, ta¨tige Personen und Umsatz im Handwerk—Jahresergebnisse 2013. Fachserie 4, Reihe 7.2. Wiesbaden. Federman, M. N., Harrington, D. E., & Krynski, K. J. (2006). The impact of state licensing regulations on low-skilled immigrants: The case of Vietnamese Manicurists. American Economic Review, 96(2), 237–241. Feldman, H. (2009). The effects of hiring and firing regulation on unemployment and employment: Evidence based in survey data. Applied Economics, 41(19), 2389–2401. Feldmann, H. (2003). Labor market regulation and labor market performance: Evidence based on survey among senor business executives. Kyklos, 56(4), 509–540. Feuerhake, J. (2012). Handwerksza¨hlung 2008. Wirtschaft und Statistik, 1, 51–62. Fredriksen, K., Runst, P., & Bizer, K. (forthcoming). Masterful meisters? Voluntary certification and quality in the German crafts sector. German Economic Review.

123

Eur J Law Econ Gomez, R., Gunderson, M., Huang, X., & Zhang, T. (2015). Do immigrants gain or lose by occupational licensing? Canadian Public Policy, 41(supplement 1), 80–97. Granovetter, M. (1973). The strength of weak ties. American Journal of Sociology, 78(6), 1360–1380. Hawthorne, L. (2015). The impact of skilled migration on Foreign Qualification Recognition reform in Australia. Canadian Public Policy, 41(1), 173–185. Kleiner, M. M. (2006). Licensing occupations. Ensuring quality or restricting competition?. Kalamazoo: W.E. Upjohn Institute for Employment Research. Kleiner, M. M. (2015). Border battles: The influence of occupational licensing on interstate migration. Employment Research Newsletter, 22(4), 4–6. Kleiner, M. M., Gay, R. S., & Greene, K. (1982). Barriers to labor migration: The case of occupational licensing. Industrial Relations, 21(3), 383–391. Kleiner, M. M., & Krueger, A. (2013). Analyzing the extent and influence of occupational licensing on the labor market. Journal of Labor Economics, 31(2/2), 173–202. ¨ konomische Effekte der Liberalisierung der Handwerksordnung von Koch, A., & Nielen, S. (2016). O 2004, WISO Diskurs, 5/2016. Kogan, I. (2010). Anew immigrants—Old disadvantage patterns? Labour market integration of recent immigrants into Germany. International Migration, 49(1), 91–117. Law, M. T., & Marks, M. S. (2009). Effects of occupational licensing laws on minorities: Evidence from the progressive era. Journal of Law and Economic, 52, 351–366. Lergetporer, P., Ruhose, J., & Simon, L. (2016). Labor market effects of entry barriers to selfemployment: Evidence from deregulating the German crafts sector. Ifo working paper—Ifo Center for the Economics of Education, Ifo Institute. May-Strobl, E. (2005). Die Ich-AG als neue Form der Existenzgru¨ndung aus der Arbeitslosigkeit. IfM Bonn. McDonald, J. T., Warman, C., & Worswick, C. (2015). Immigrant selection systems and occupational outcomes of international medical graduates in Canada and the United States. Canadian Public Policy, 41(Suppl 1), 116–137. Minniti, M., & Wim, N. (2010). What do we know about the patterns and determinants of female entrepreneurship cross countries? The European Journal of Development Research, 22(3), 277–293. Mueller, K. (2006). Erste Auswirkungen der Novelierung der Handwerksordnung von 2004, Go¨ttinger Handwerkswirtschaftliche Studien, 74. Duderstadt: Mecke-Druck. Mueller, K. (2014). Stabilita¨t und Ausbildungsbereitschaft von Existenzgru¨ndern im Handwerk, Go¨ttinger Handwerkswirtschaftliche Studien, 94. Duderstadt: Mecke-Druck. Mueller, K. (2015). Strukturentwicklung im Handwerk, Go¨ttinger Handwerkswirtschaftliche Studien, 98. Duderstadt: Mecke-Druck. OECD. (2017). International migration outlook 2017. Paris: OECD. Ogus, A., & Zhang, Q. (2006). Licensing procedures in developing countries: Should they be part of the set-up process? International Journal of Public Administration, 29(12), 1091–1108. Pashigian, P. B. (1979). Occupational licensing and the interstate mobility of professionals. The Journal of Law and Economics, 22(1), 1–25. Philipsen, N. (2009). Regulation of liberal professions and competition policy: Developments in the EU and China. Journal of Competition Law & Economics, 6(2), 203–231. Rostam-Afschar, D. (2014). Entry regulation and entrepreneurship: A natural experiment in German craftsmanship. Empirical Economics, 47, 1067–1101. Runst, P., Thoma¨, J., Haverkamp, K., & Mu¨ller, K. (forthcoming). A replication of ‘Entry regulation and entrepreneurship: A natural experiment in German craftsmanship’. Empirical Economics. Shapiro, C. (1986). Investment, moral hazard, and occupational licensing. The Review of Economic Studies, 53(5), 843–862. Svorny, S. (2000). Licensing, market entry regulation. Encyclopedia of law and economics. Williams, W. E. (1982). The state against Blacks. New York, NY: New Press.

123