Labour market deregulation and apprenticeship training: A comparison of German and Swiss employers

Abstract

Many extensions of classical human capital theory regard labour market rigidities as a prerequisite for firms to invest in general training. From this perspective, the German labour market reforms since 2003 should have reduced their willingness to support the apprenticeship training system. This article demonstrates that, on the contrary, German firms did not abandon the training system but instead changed their training strategies after the implementation of the labour market reforms. We analyse the new training strategies that German firms deployed to cope with the increased labour market flexibility engendered by the labour market reforms. Switzerland, where no such reforms occurred, serves as the counterfactual. The results demonstrate that German firms successfully reduced the net costs of training by involving apprentices in more work and reducing non-productive tasks.

Keywords

Apprenticeship training cost–benefit difference-in-differences matching estimator labour market reforms Germany Switzerland

Introduction

In the contemporary literature on the economics of training, the existence of extensive training systems – such as the apprenticeship system in Germany, where firms are willing to invest considerable sums in the general skills of their employees – has been used as an illustration of the importance of labour market frictions. Conversely, the absence of strict labour market regulations has been used to explain the absence of such firm behaviour in Anglo-Saxon labour markets. If apprenticeship training systems and tight labour market regulation always come in tandem, and are therefore mutually dependent, then deregulating the labour market would clearly affect firms’ willingness to invest in training. Acemoglu and Pischke (1999b), pioneers in extending Becker’s theory of human capital, feared exactly this when writing shortly before the adoption of the labour market reform agenda by the Schröder government in Germany:

naturally, in practice, increased frictions will have a number of allocative costs, such as lower employment … [but] in any case, the implications of labour market frictions on training are worth bearing in mind when suggesting labour market reforms. For example, proposals for reducing union power and removing regulations in the German labour market, which are on the current agenda, could have unforeseen consequences regarding the German apprenticeship system, where employers pay for the general training of their workers. (pp. 548–549)

In this article, we analyse the impact of the German labour market reforms on the behaviour and strategies of German firms with respect to apprenticeship training. Identifying a strictly causal effect of the labour market reforms is not within our scope. However, we provide and test different explanations for the observed developments in the German training system. In particular, we undertake a comparison with Switzerland, a country with an almost identical training system that did not change its labour market regulations during the observation period. For the first time, we make use of four German and Swiss datasets at different points in time. In this way, our study extends the existing literature by restricting the reasons for changes in training behaviour in Germany to nationally specific factors. In contrast to the predictions of Acemoglu and Pischke, the probability of training in Germany did not decrease after the government relaxed labour market rules: the share of German firms with at least one employee providing training remained at approximately 24 percent between 2000 and 2007.¹ This observation is only compatible with the theoretical predictions if firms were able to change their training strategies and reduce their up-front investments in training. This makes Switzerland a most attractive comparator: first, because the Swiss labour market has traditionally been deregulated, forcing Swiss firms to train apprentices differently from their German counterparts (Mühlemann et al., 2010). Furthermore, the difference-in-differences approach allows us to rule out global changes influencing the apprenticeship training systems in both countries.

Our findings show that German firms substantially reduced their up-front net investments in apprentices at establishment level to limit the potential loss should skilled workers leave the firm which trained them. As the binding training regulations did not allow establishments to reduce their gross investments in training, they were forced to increase the productive contributions of trainees to achieve a reduction in the net costs of training. The productive contributions of apprentices per trainee and year increased in Germany (in contrast to Switzerland) on average by €2000 from 2000 to 2007. The establishments were able to increase substantially the productive contributions of trainees by involving them in more skilled work, while reducing the number of days that trainees had previously used for non-productive tasks. We also consider other explanations for our results, such as the decline of union power in Germany. Although we cannot provide clear causal evidence, our results suggest that the labour market reforms rather than other developments influenced the training strategy of German establishments.

The article is structured as follows: we next offer a brief summary of the theoretical literature on the relationship between labour market frictions and establishment investments in general skills. We then describe the apprenticeship training systems in Germany and Switzerland and their labour market regulations. Subsequently, we provide information on the changes in labour market regulation that took place in Germany from 2000, then describe the data and present our empirical strategy. After presenting our empirical results and several robustness checks as well as tests of alternative explanations, we offer a conclusion.

Labour market regulation and training behaviour

In classical human capital theory (Becker, 1962), firms have to pay part of the cost of firm-specific skill development and employees have to pay the total cost of general skill development.² These theoretical predictions are based on the assumption of competitive labour markets in which workers cannot be paid below their marginal productivity. Therefore, an employer could never recoup investments in skills that are productive across a number of employers (general skills) by paying the trained worker below his/her marginal productivity. However, it has been shown that many German employers do make considerable net investments in general skills (Von Bardeleben et al., 1995). Acemoglu and Pischke (1998, 1999a, 1999b) therefore extended classical human capital theory in an attempt to reconcile theory with empirical observations. They dropped the assumption that labour markets are perfectly competitive; instead, and more realistically, they assumed that labour markets are characterized by varying degrees of friction. In their model, these frictions lead to a compressed wage structure, in which the gap between productivity and wage levels increases with training. Therefore, skilled workers can be paid below the marginal product of their labour and rents for firms increase through training. As a result of this compressed wage structure, firms are able to recoup investments in training.

Labour market frictions may have several causes, and labour market regulations, such as employment protection legislation (increasing firing costs), may be just one but an important source of such frictions (Acemoglu and Pischke, 1999a: 135–136). Regulations such as specific rules concerning the dismissal of workers or limitations on temporary work (which allow employers to screen potential employees) increase the costs of bad employer–employee matches. To avoid mismatches, firms have to invest more in the search of new recruits, and hence bear higher hiring costs. High hiring costs make a hire-and-fire policy costly and therefore reduce the number of job vacancies, which also means that the probability that a competitor will poach a firm’s trainees is reduced. This in turn leads the training firm, at the margin, to pay wages below marginal productivity and thereby recoup prior net investments in training. Therefore, higher transaction costs induced by employment protection are one of the sources for a compressed wage structure (Acemoglu and Pischke, 1999a: 121), which is in turn an incentive for firms to train and accept net costs in doing so.

Another source of a compressed wage structure is information asymmetries between firms about the ability of the worker (Acemoglu and Pischke, 1998, 1999a, 1999b). Firms which undertake training know about the true ability of their former apprentices and can therefore retain the best trainees. As a consequence, the average ability of former trainees in the external labour market is lower than the average ability of those who stay. As workers cannot signal to other employers their true ability, the incumbent employer can exert a monopsony power and set wages for former apprentices below their marginal product, leading to a compressed wage structure. Therefore, in a labour market with high firing costs, apprenticeship training serves also as a screening device.

Instead of a costly apprenticeship training programme, probationary periods or temporary work contracts could be used to screen future workers more efficiently. Although the screening of future workers by probationary periods is not perfect – Ichino and Riphahn (2005) and Riphahn and Thalmaier (2001) show that there is a tendency for workers to have higher absenteeism once they have secured a permanent position after the probation has ended – the possibility to test candidates during probation or with a temporary work contract should help firms to detect at least the non-productive worker before having to offer them a permanent contract. Productive workers might pretend to be motivated, but it would be hard for non-productive workers to mimic good behaviour and high productivity over a long period (temporary work contracts, in contrast to probation, can last from several months up to several years). To the extent that firms can use temporary work contracts to screen future workers, the screening benefit of an apprenticeship training programme loses its relevance and therefore firms would be less willing to accept net costs of such a training programme for the sole purpose of screening.

In summary, while labour market regulations allow firms to earn a rent on their training investments after the training period has ended in competitive labour markets training only takes place if firms can pay trainees below their productivity during the training period. As a consequence, the greater the frictions in a labour market, the higher the share of firms that are willing to accept the net costs of training at the end of the training period (and vice versa).

This allows us to formulate the hypothesis that in two countries with considerably different degrees of strictness in their labour market regulations, (ceteris paribus) firms will exhibit differences in the observed investments in training at the end of the training period. Furthermore, if labour market regulations are relaxed, then net investments in training (per trainee) are either reduced, or if this is not possible (e.g. because of externally imposed training regulations), the training intensity in the economy would decrease.

One training system, different labour market regulations, different training strategies

In the following section, we present the German and Swiss apprenticeship systems and the labour market regulations in both countries.

Apprenticeship training

Apprenticeship training has a long tradition in both Germany and Switzerland. The two systems are highly comparable in terms of training regulations, requirements for entering an apprenticeship and the amount and type of skills acquired. Apprenticeship programmes in both countries are labelled ‘dual’ because learning takes place at two different learning sites: the workplace and the vocational school. Firms and apprentices sign a binding training contract for a fixed duration and at a predetermined wage, which is a fraction of the wage of a skilled worker. With more than half of a cohort of young adults entering the dual apprenticeship system after compulsory schooling in Germany and two-thirds in Switzerland, it is also the single most important educational pathway at the upper secondary level.

By signing the contract, firms commit to an occupation-specific curriculum to provide apprentices with general and occupation-specific skills in the workplace. Conveying formal education is costly to the firm, as it has to pay wages for apprentices and for training personnel, as well as material and machinery used in the training process (gross costs of training). However, apprentices also perform tasks that would otherwise have to be performed by skilled or unskilled workers and therefore generate benefits for the training firm (productive contribution). In addition to productive tasks, firms can also allocate non-productive tasks to apprentices, such as exercises. While training regulations are quite prescriptive in terms of the content and the amount of training a firm has to provide its apprentices, the firms have a considerable degree of freedom regarding the work allocated to apprentices.

Labour market regulation and training strategies

In contrast to the similarities in the apprenticeship systems, labour market regulations differ considerably between Germany and Switzerland. Overall labour market strictness, as measured by the Organisation for Economic Co-operation and Development (OECD), was 2.3 (out of a potential maximum of 6) for Germany in the year 2000 and 1.1 for Switzerland (for a detailed description of this indicator, see Venn, 2009). When ranking the 29 OECD countries in the year 2000 according to their labour market strictness from the least (USA) to the most regulated country (Turkey), Switzerland ranked 5th and Germany 21st.

In accordance with the hypothesis formulated in the previous section, an empirical analysis of the net costs of apprenticeship training (gross costs minus productive contribution) in Germany and Switzerland for the year 2000 (Dionisius et al., 2009) showed that, for an average 3-year apprenticeship, a training establishment in Germany had to bear net costs, whereas a comparable training establishment in Switzerland recorded a net benefit at the end of the training period. The difference in net costs between comparable German and Swiss establishments in 2000 amounted to €25,000. Using matching models, Dionisius et al. (2009) showed that the main factor explaining this difference was the allocation of productive tasks to the apprentices. Swiss apprentices created greater benefits for the training establishment because they spent more of their time in the production process, whereas their German counterparts spent more time on non-productive exercises.³ Also in line with theoretical predictions, the majority of Swiss apprentices (two-thirds) left the training company within 1 year of qualification, whereas exactly the opposite could be observed in Germany (one-third left the training company within 1 year). The higher degree of employment protection resulting in low labour market mobility enabled German establishments to recoup their investments in the post-training period, whereas their Swiss counterparts, operating in a competitive labour market, had to protect themselves against a likely loss of their investments by recouping their investments during the training period.

In another study, Mühlemann et al. (2010) analysed not only the differences between the training strategies of German and Swiss establishments but also the differences between training and non-training establishments in both countries. According to these analyses and consistent with the hypothesis formulated, non-training establishments in Switzerland generally refrain from training because of the expected high net investment costs during the training period; whereas in Germany, non-training establishments are generally those that, despite the tighter labour market regulations, do not expect substantial post-training benefits.

Changes to German labour market regulations and the expected consequences

In response to stagnating economic growth and high unemployment rates, the German government of Chancellor Schröder introduced a comprehensive economic reform, also known as Agenda 2010, in 2003. The reforms aimed to foster economic growth and increase employment by adopting three primary measures: reducing non-wage labour costs, increasing labour market flexibility and reforming the social welfare system. Agenda 2010 consisted of various legislative changes, which successively came into force between 2003 and 2005; among them were also the four law packages Hartz 1 to 4 (Gesetze zur Reform des Arbeitsmarktes 1 to 4).⁴ These changes affected a wide array of political and economic areas, such as the social welfare system, the health and pension insurance system, labour market regulation and family and education policy.

In the present analysis, we concentrate on the components of the reform that targeted labour market flexibility, and only present changes that could be expected to have an effect on the companies’ training decisions and behaviour. To increase labour market flexibility, protection for regular employment and laws on temporary employment were changed. Two major changes affected regular employment. First, the threshold at which employment protection law applied was raised from 5 to 10 employees per firm; thus, firms with fewer than 10 employees were no longer constrained in their dismissal decisions. Second, since 2004, in selection for dismissal the employer has to take account of only four clearly defined criteria: tenure, age, maintenance obligations and severe disability. Before 2004, all different aspects of the employees’ current situation could enter the selection criteria, which gave employees much more possibilities to contest dismissal; thus, the employer faced a higher degree of legal uncertainty in firing decisions.

In relation to temporary work contracts, the 2004 reform of the Arbeit-nehmerberlassungsgesetz (AÜG), a component of Hartz 1, eased constraints on temporary work contracts and made those types of contracts more popular. The changes provided firms with more flexibility to employ temporary workers whenever they are in need of them. As a result of the legislative changes, the data from the Institut für Arbeitsmarkt- und Berufsforschung (IAB) establishment panel show that of all new work contracts, the share of temporary contracts increased by more than a third from 32 percent in 2001 to 45 percent in 2011 (IAB, 2012). The proportion of all contracts that were temporary increased from 5 percent in 2001 to 7 percent in 2007 (Hohendanner, 2014).

The changes to the German employment protection system are also represented in the overall OECD employment protection index, which decreased from 2.3 in 2000 to 2.1 in 2007. In particular, the indicator for temporary employment showed a substantial decline of 0.7 units.⁵ In contrast, Swiss labour market regulation did not change during this period, as demonstrated by the constant OECD indicators (Table 1).

Table 1.

Strictness of employment protection.

OECD index	Germany		Switzerland
	2000	2007	2000	2007
Overall	2.3	2.1	1.1	1.1
Regular employment	2.7	3.0	1.2	1.2
Temporary employment	2.0	1.3	1.1	1.1

Source: OECD Employment and Labour Market Statistics.

OECD: Organisation for Economic Co-operation and Development.

The German labour market reform, which limited the strictness of protections for temporary employment, alleviated the problem of evaluating the competence of new employees and thereby reduced the expected rents that firms can extract from trainees after training. As a result, some firms could be expected to end apprenticeship training and those remaining active in training could be expected to change their training behaviour in order to reduce the net costs of training.⁶ As there is no evidence for a reduction in training activity, we expect firms to reduce their up-front investment in training in response to the reduced strictness in employment protection.

Theoretically, the net costs of training can be reduced in several ways. Firms could attempt to cut gross costs by employing fewer trainers or by reducing apprentices’ pay. However, decreasing training hours would result in lower training quality, and even if the firm is unwilling to employ a former apprentice as a skilled employee, reputational considerations and legal constraints limit the possibility of significantly reducing training quality. Apprentice wages, however, are determined collectively in Germany and are therefore not completely at the firm’s discretion (see Ryan et al., 2013 on this issue, comparing the German situation with Switzerland and the United Kingdom). In consequence, both potential strategies to reduce the gross costs of training are rather limited in the German context.

Therefore, a more promising strategy for reducing net training costs is to increase the benefits. Because regulations concerning the work allocated to apprentices are less strict and it is unclear whether performing productive tasks reduces or even increases the quality of training, this strategy seems more likely.

Data

The concept for the cost–benefit surveys used in this study was developed by the 1974 report of the Sachverständigenkomission Kosten und Finanzierung der beruflichen Bildung (Expert Commission on Costs and Financing of Vocational Education and Training), also known as the Edding Commission. In our analysis, we use four cross-sectional establishment-level surveys, two conducted in each country at different points in time, which follow the same procedure. Two nearly identical surveys were conducted simultaneously for 2000 (see Beicht et al., 2004 for Germany and Schweri et al., 2003 for Switzerland). The second pair of surveys was conducted for 2007 in Germany (Schönfeld et al., 2010) and 2009 in Switzerland (Strupler and Wolter, 2012). It is unfortunate that the second pair of surveys was not conducted in exactly the same year. However, we argue that this does not affect their comparability because we expect the Swiss results to have been very similar even if the data had been collected in 2007. The Swiss cost–benefit data were remarkably stable in varying economic circumstances throughout the decade, as shown by comparing the 2009 data to an additional survey conducted in Switzerland in 2004 (see also Mühlemann et al., 2010). All three Swiss surveys (2000, 2004 and 2009) produced very stable results and showed no significant differences in costs and benefits of apprenticeship training.⁷

The data contain detailed information on the costs and benefits of apprenticeship training and establishment characteristics for German and Swiss establishments.⁸ In terms of methodology, the surveys are similar and comparable in nearly all respects. Nevertheless, there are some minor differences: one related to the questions eliciting the training hours. The questions were changed in the Swiss and the German questionnaires relative to the 2000 survey, but in different ways. As a consequence of these changes, we will not be able to compare changes over time in the gross costs of training. However, as the major initial difference in the net costs of training between the two countries stems from differences in the benefits from training (Dionisius et al., 2009), this should not substantially affect our analyses. As in previous comparative studies, the analyses are restricted to 3-year apprenticeship programmes, as longer programmes are not fully comparable across the two countries, lasting 3.5 years in Germany and 4 years in Switzerland. The final sample consists of 1471 Swiss and 1738 German training establishments in 2000, and 1842 Swiss and 2161 German training establishments in 2009 and 2007, respectively.

Empirical strategy

To analyse the influence of the German labour market reforms on training behaviour, we combine a difference-in-differences approach with a matching strategy, similar to Heckman et al. (1998). In contrast to a matching estimator which only compares the outcomes after the reform, the difference-in-differences estimator controls for time-invariant differences in the outcomes between German and Swiss establishments.

Our aim is to show how the German labour market reforms affected training behaviour. Therefore, we estimate an average treatment effect on the treated (ATT), where the treatment is the reform and the treated establishments are German establishments. The fundamental identification problem is that for a particular German establishment and time, we never observe both potential outcomes, with and without the reform, simultaneously. German establishments were all exposed to the reform, and therefore, the counterfactual outcome in the absence of the reform cannot be observed. Thus, we cannot directly observe the effect of the reform but instead estimate an ATT by applying a difference-in-differences matching strategy and therefore use Swiss establishments as the (no reform) counterfactual.

The crucial identifying restriction in difference-in-differences models is the common trend assumption, which means that in the absence of the treatment, treated and non-treated establishments would have followed a parallel path in terms of their training behaviour. Unfortunately, there are no data available on pre-treatment trends to support this assumption. Previous surveys in Germany and Switzerland on the costs and benefits of apprenticeship training are not comparable to the surveys in 2000, because their methodology differs strongly from the later studies. For example, the previous studies did not differentiate across occupations and only covered some industries in Germany (Beicht et al., 2004).

To ensure that similar establishments are compared, we apply a matching strategy proposed by Abadie et al. (2004). In a first step, we match each Swiss establishment in the 2009 sample to one establishment in 2000 with the same characteristics (establishment size, job categories, industry and region). If more than one establishment with exactly the same characteristics exists, the outcome is averaged over these establishments. The same strategy is applied for German establishments. As a result, each establishment in the post-treatment period receives a (potential) outcome from the pre-treatment period, and as a consequence, a within-country difference over time can be calculated. In a second step, we estimate the difference of the within-country differences. To ensure a balanced comparison group, a matching strategy is again applied: we match each German establishment in the post-treatment period with one (or more) similar Swiss establishments (from the post-treatment period). This is equivalent to estimating an ATT.

To ensure that the matching estimator identifies and consistently estimates the treatment effect on the treated, two assumptions have to be satisfied: the first assumption holds if assignment to treatment (establishments located in Germany) is independent of the outcomes conditional on covariates. Independence conditional on covariates means that there are no unobservables that simultaneously affect the outcomes and the residence choice of the establishment (unconfoundedness). Although the location of establishments is not random, apprenticeship training is never the core business of an establishment, so we can safely assume that establishments choose their country locations independent of factors influencing the benefits and costs of apprenticeship training. Therefore, the unconfoundedness assumption holds. The second assumption holds when the probability of assignment is restricted between 0 and 1 (Abadie et al., 2004). This identification assumption holds, as large numbers of establishments with similar establishment characteristics are available in both samples.

Results

Apprentice benefits and benefit components

Our results indicate that the benefits apprentices generate for the establishment during training increased substantially in Germany, whereas no significant change could be identified in Switzerland. Table 2 shows the average treatment effects on the treated and the difference-in-differences estimates for several dependent variables. The increase in benefits for an average training year was more than €2000 higher for the average German establishment compared to a similar Swiss establishment. For the entire training period, this sums to over €6000 and represents an increase of nearly 25 percent.⁹

Table 2.

Average treatment effects on the treated for Switzerland, Germany and difference-in-differences (DID) estimates.

	ATT (ΔY⁰)		ATT (ΔY¹)		ATT (Δ^DID)
	Switzerland		Germany		Germany – Switzerland
Variable	Coefficient	Std error	Coefficient	Std error	Coefficient	Std error
Benefit per year per apprentice (in €)	−499.16	324.95	1639.93	200.54***	2137.20	275.17 ***
Share of productive tasks (unskilled, 1st year)	3.02	1.59*	18.01	1.39***	14.52	1.54***
Share of productive tasks (skilled, 1st year)	−3.11	1.85*	11.22	1.40***	14.71	1.46***
Share of tasks with no direct value to establishment (1st year)	0.09	1.23	−29.43	1.45***	−29.44	1.41***
Share of productive tasks (unskilled, 2nd year)	2.90	1.64*	5.31	1.21***	2.43	1.56
Share of productive tasks (skilled, 2nd year)	−2.42	1.72	16.66	1.38***	19.97	1.65***
Share of tasks with no direct value to establishment (2nd year)	−0.48	4.50	−22.18	1.12***	−22.62	1.06***
Share of productive tasks (unskilled, 3rd year)	4.50	1.26***	1.48	1.25	−0.88	1.53
Share of productive tasks (skilled, 3rd year)	−4.31	1.41***	12.34	1.54***	15.97	1.83***
Share of tasks with no direct value to establishment (3rd year)	−0.19	0.87	−14.12	1.05***	−15.39	1.12***
Number of observations, 1st year	1308		1133		1133
Number of observations, 2nd year	1257		1326		1326
Number of observations, 3rd year	1237		1195		1195
Number of observations total	1842		2161		2161

ATT: average treatment effect on the treated.

Matching variables are establishment size (exact), job categories (exact) and industry (+region for within-country matching). Standard errors are robust.

p < 0.1; **p < 0.05; ***p < 0.01.

The results of a more detailed analysis (also shown in Table 2) indicate that German establishments increased the benefits of training by changing the tasks allocated to apprentices, whereas no substantial change in task allocation can be found for Switzerland. German establishments increased the shares of and the days spent in productive unskilled and skilled tasks compared to Swiss companies. These changes were all at the expense of tasks with no direct value to the establishment. The difference-in-differences results reveal that the share of tasks without direct value to the establishment decreased by nearly 30 percentage points in the first year of training, whereas the shares of skilled and unskilled productive tasks each increased by approximately 15 percentage points. In terms of working days, these results represent increases in unskilled and skilled tasks of 26 and 25 days, respectively, with a corresponding reduction in the number of days of ‘non-productive’ practice. For the second and third years of training, the share of skilled tasks increased in German establishments, whereas unskilled tasks did not change relative to Swiss establishments. For the entire period of an apprenticeship, the relative change was +83 days in skilled and +33 days in unskilled tasks for German establishments compared to Swiss establishments.

The extensive use of apprentices in non-productive instead of productive work has primarily been defended by those fearing that the use of apprentices as ‘cheap labour’ would contradict the purposes of training and the qualification goals of apprenticeship. Although measuring the quality of training is difficult, the relative productivity of an apprentice in skilled tasks compared to a fully trained skilled worker can be used as a valuable proxy for the quality of training.¹⁰ Additional analyses show that the relative productivity, compared to that of an average skilled worker – as reported by the companies – increased in Germany in all training years in absolute terms and also relative to Swiss establishments. In 2007/2009, relative productivity in the average German and Swiss training establishments reached comparable levels.¹¹ The evidence presented also suggests that learning and (skilled) work are joint products and not substitutes, and that the involvement of apprentices in the production process can have a positive effect on the competencies they acquire. The positive impact of increased work exposure on the productivity of German apprentices is also credible when considering that non-productive tasks were mostly substituted by work requiring skills and not by unskilled activities.

Robustness checks

The most crucial assumption for our analysis is the common trend assumption. Trends in the outcomes in Germany and Switzerland are assumed to be comparable. Unfortunately, no comparable data exist for the pre-treatment periods; therefore, pre-treatment trends cannot be analysed. As a result, this article cannot entirely rule out that other differences than the changes in labour market regulation explain part of our results. However, we provide some additional empirical analyses to show how and to what extent other explanations affected apprenticeship training, and to strengthen our view that deregulating the labour market affected training behaviour in Germany.

Germany and Switzerland both have a considerable degree of within-country heterogeneity in economic activity. While the primary divides in Germany are East–West and North–South, in Switzerland, the differences in economic activity are more pronounced between the linguistic regions (French-, Italian- and German-speaking regions). To determine whether the difference-in-differences estimates are affected by these divisions, we run all estimations comparing only the German-speaking regions of Switzerland with the two regions (Bayern and Baden-Württemberg) in Germany that border Switzerland. Not only are the cross-border economic activities in these regions very intense, but also company activities are heavily intertwined; hence, the two areas can almost be considered a single economic area. This implies that with the exception of labour market regulations and other laws, there would not be major differences for other factors such as the business cycle or technological progress. When conducting the analyses with the reduced sample, we obtain qualitatively identical results to those obtained for the full sample, which supports the interpretation that the differences in the developments in Germany relative to Switzerland cannot have been caused by unobserved economic differences.

Germany – unlike Switzerland, where unions have always played a minor role – faced a period of declining unionization (Fitzenberger et al., 2011), which could also have affected training behaviour. Weaker union bargaining power could affect apprentices’ wages, as earlier research has shown that unionized firms pay higher wages for unskilled workers (and therefore most likely also for apprentices) than non-unionized firms (Dustmann and Schönberg, 2009). Moreover, firms could use de-unionization to engage apprentices more in productive work. In unionized firms, unions are likely to resist the use of apprentices for productive work, for fear that skilled and unskilled workers might be substituted by apprentices. To analyse whether de-unionization drives our results, we made several empirical checks. Some evidence against the importance of the unionization argument is already given by our reduced sample taking into account only Bayern and Baden-Württemberg. The decline in union power was much more pronounced in East Germany than in the West (Fitzenberger et al., 2011). Therefore, if de-unionization would be the reason for the increased involvement of apprentices in the production process, the increase of training benefits would have to be considerably lower for this subsample, but this is not the case.

In a next step, we estimate the full sample with controls for collective wage agreement. If de-unionization had been the driving force for the increased engagement of apprentices in productive work, then controlling for the unionized establishments should change our results significantly. The results are, however, comparable to those in Table 2 without such controls. Furthermore, we repeated our estimations excluding all establishments with a collective wage agreement from our analysis and obtained the same results. Finally, we also ran a difference-in-differences estimation only for German establishments, comparing establishments with a collective agreement with establishments that had never been covered. The results show that training benefits did not increase significantly more in establishments without collective agreements than in establishments covered by those agreements. All results are available upon request.¹²

In summary, we do not see that the de-unionization could explain the changes in the benefits for apprenticeship training for German establishments over time. However, de-unionization could affect the net costs by lowering gross costs of training. For methodical reasons, not all components of the gross costs of training can be compared. Nevertheless, we conducted additional analyses and compared changes in the relative wages of apprentices and skilled workers. Our results¹³ show that while wages of skilled workers in the trained occupation did not change significantly in Germany relative to Switzerland, real wage costs of apprentices in Germany decreased significantly in absolute terms and relative to Switzerland. In line with the analyses of Dustmann and Schönberg (2009), it is likely that the decline in union power reduced apprentice pay and thereby contributed to a further reduction of the net costs of apprenticeship training in Germany.

Conclusion

Labour markets that have frictions rather than being competitive have played a prominent role in explaining why firms in some countries are willing to finance large up-front investments in general skills, while firms in other countries do not support such schemes. In this training literature, Germany has served as an exemplary case, demonstrating why labour market regulations (as an important source of labour market frictions) were a prerequisite for a functional large-scale apprenticeship training system. Although the case of Switzerland, where labour market regulations were considerably less strict than in Germany, indicated that it should also be possible to have an apprenticeship training system in a deregulated labour market, this did not prove that the German apprenticeship training system would not be damaged by the deregulation of the German labour market planned and implemented by the government from 2003.

Our empirical analysis shows that German firms, realizing that post-training benefits decrease in a more flexible labour market, managed successfully to reduce their up-front investments in general skills, instead of leaving the apprenticeship training system as anticipated by the modern training literature. Firms increased the benefits of the apprenticeship training during the training period as a result of a change in the allocation of productive tasks to apprentices. The lack of experimental data and the complexity of firm’s training behaviour do not allow us to draw strictly causal conclusions from our difference-in-differences analysis. However, the results demonstrate that in contrast to the predictions in the literature, training participation remained high in Germany after the labour market reform.

In conclusion, this article shows that labour market regulations and frictions are indeed a prerequisite for net investments in general skills by firms, as stipulated by the modern training literature, but that a net investment is not a precondition for a functional apprenticeship training system.

Footnotes

Acknowledgements

The authors thank the editor, two anonymous referees as well as Samuel Mühlemann, Harald Pfeifer, Andries de Grip and Ben Kriechel for comments and valuable suggestions.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/ or publication of this article: The Swiss part of this research project benefited from generous funding by the Staatssekretariat für Bildung, Forschung und Innovation (SBFI) through the Leading House on the Economics of Education at the Universities of Zurich and Bern.

Notes

Author biographies

Anika Jansen is a researcher at the German Federal Institute for Vocational Education and Training (BIBB) in Bonn.

Mirjam Strupler Leiser is a researcher at the Centre for Research in Economics of Education at the department of economics, University of Bern.

Felix Wenzelmann is a researcher at the German Federal Institute for Vocational Education and Training (BIBB) in Bonn.

Stefan C Wolter is Professor and head of the Centre for Research in Economics of Education at the department of economics, University of Bern. He is also a research fellow of CESifo (Munich) and IZA Forschungsinstitut zur Zukunft der Arbeit (Bonn), and Chair of the OECD Group of National Experts on Vocational Training and Education.

References

Abadie

Drukker

Leber Herr

Imbens

(2004) Implementing matching estimators for average treatment effects in Stata. Stata Journal 4(3): 290–311.

Acemoglu

Pischke

J-S

(1998) Why do firms train? Theory and evidence. The Quarterly Journal of Economics 113(1): 79–119.

Acemoglu

Pischke

J-S

(1999a) Beyond Becker: Training in imperfect labour markets. The Economic Journal 109: 112–142.

Acemoglu

Pischke

J-S

(1999b) The structure of wages and investment in general training. Journal of Political Economy 107(3): 539–572.

Becker

(1962) Investment in human capital: A theoretical analysis. The Journal of Political Economy 70(5): 9–49.

Beicht

Walden

Herget

(2004) Kosten und Nutzen der betrieblichen Berufsausbildung in Deutschland. Bielefeld: Bertelsmann.

Dionisius

Mühlemann

Pfeifer

Walden

Wenzelmann

Wolter

(2009) Costs and benefits of apprenticeship training: A comparison of Germany and Switzerland. Applied Economics Quarterly 55(1): 7–37.

Dustmann

Schönberg

(2009) Training and union wages. The Review of Economics and Statistics 91(2): 363–376.

Dustmann

Schönberg

(2012) What makes firm-based vocational training schemes successful? The role of commitment. American Economic Journal: Applied Economics 4(2): 36–61.

10.

Fitzenberger

Kohn

Wang

(2011) The erosion of union membership in Germany: Determinants, densities, decompositions. Journal of Population Economics 24(1): 141–165.

11.

Heckman

Ichimura

Todd

(1998) Matching as an econometric evaluation estimator. Review of Economic Studies 65(2): 261–294.

12.

Hohendanner

(2014) Befristete Beschäftigung Mögliche Auswirkungen der Abschaffung sachgrundloser Befristungen. Nuremberg: IAB Stellungnahme.

13.

IAB (2012) Befristete Neueinstellungen. IAB Aktuell, 23 February. Available at: http://doku.iab.de/aktuell/2012/befristete_beschaeftigung_uebernahme.pdf

14.

Ichino

Riphahn

(2005) The effect of employment protection on worker effort: Absenteeism during and after probation. Journal of the European Economic Association 3(1): 120–143.

15.

Kriechel

Mühlemann

Pfeifer

Schütte

(2014) Works councils, collective bargaining, and apprenticeship training – Evidence from German firms. Industrial Relations 53(2): 199–222.

16.

Leuven

(2005) The economics of private sector training: A survey of the literature. Journal of Economic Surveys 19(1): 91–111.

17.

Matthes

Ulrich

Flemming

Granath

R-O

(2014) Ausbildungsplatzangebot- und Nachfrage. Datenreport zum Berufsbildungsbericht, Bundesinstitut für Berufsbildung, Bonn, pp. 14–29.

18.

Mühlemann

Pfeifer

Walden

Wenzelmann

Wolter

(2010) The financing of apprenticeship training in the light of labor market regulations. Labour Economics 17(5): 799–809.

19.

Riphahn

Thalmaier

(2001) Behavioural effects of probation periods: An analysis of worker absenteeism. Journal of Economics and Statistics 221(2): 179–201.

20.

Ryan

Backes-Gellner

Teuber

Wagner

(2013) Apprentice pay in Britain, Germany and Switzerland: Institutions, market forces and market power. European Journal of Industrial Relations 19(3): 201–220.

21.

Schönfeld

Wenzelmann

Dionisius

Pfeifer

Walden

(2010) Kosten und Nutzen der dualen Ausbildung aus Sicht der Betriebe. Bielefeld: Bertelsmann.

22.

Schweri

Mühlemann

Pescio

Walther

Wolter

Zürcher

(2003) Kosten und Nutzen der Lehrlingsausbildung aus der Sicht Schweizer Betriebe. Chur, Zürich: Rüegger Verlag.

23.

Strupler

Wolter

(2012) Die duale Lehre: Eine Erfolgsgeschichte – Auch für die Betriebe. Chur, Zürich: Rüegger Verlag.

24.

Venn

(2009) Legislation, Collective Bargaining and Enforcement: Updating the OECD Employment Protection Indicators. Paris: OECD Social, Employment and Migration Working Papers.

25.

Von Bardeleben

Beicht

Fehér

(1995) Betriebliche Kosten und Nutzen der Ausbildung: Repräsentative Resultate aus Industrie, Handel und Handwerk. Gütersloh: Bertelsmann.

26.

Wolter

Ryan

(2011) Apprenticeship. In: Hanushek

Machin

Woessmann

(eds) Handbook of the Economics of Education, vol. 3. Amsterdam: North-Holland, pp. 521–576.