Navigating Industrial Transformation: The Impact of Continuing Vocational Education and Training on Unemployment Risk in Germany’s Industrial Sector (1986

Abstract

The late 1980s and 1990s marked a transformative period in Germany, characterised by rapid structural change and labour market disruptions in the industrial sector. In this context, continuing vocational education and training (CVET) was widely promoted as a policy tool to support workers in adapting and to mitigate risks of unemployment. Despite these expectations, the effectiveness of CVET in reducing unemployment during this turbulent period remains contested. This study examines the impact of CVET on the risk of unemployment among industrial workers in Germany between 1986 and 1999. Drawing on longitudinal data from the German Socio-Economic Panel, the analysis applies a propensity score matching approach to estimate the Average Treatment Effect on the Treated (persons who participated in CVET). This method helps to account for observable differences between workers who participated in CVET and those who did not. The findings suggest that participation in CVET had no statistically significant effect on reducing future unemployment among industrial workers after the training. These results are consistent with other empirical studies. The findings highlight the limitations of CVET as a short-term labour market policy tool during periods of rapid economic transformation, while underscoring the importance of considering sector-specific dynamics such as technological changes when evaluating training measures.

Keywords

continuing vocational education and training unemployment industrial sector SOEP adult education deindustrialisation

Introduction

Deindustrialisation and the associated structural changes in society, as well as the fear of job losses, were dominant themes in the 1980s and 1990s. The industrial sector in Western industrial societies was particularly affected by major transformations. Structural unemployment was widespread, and the notion of the ‘crisis of the working society’ was invented (Dahrendorf, 1978; Schlutz, 1985; Struck, 2016). Adult education and training have often been seen as a protection against such transformations. However, workers in industrial sectors, particularly those performing manual tasks, tended to participate less in continuing education than workers in other sectors (Figure 1; Fey, 2025). At the same time, they were often the group most affected by these upheavals. Similar patterns can be observed today, which is why a deeper understanding of the role of continuing education in this period could yield insights relevant to current transformations.

Figure 1.

Percentage of Participants in CVET, by Sector.

In the 1990s, after reunification, the process of deindustrialisation accelerated in East Germany, leading to mass plant closures and mass unemployment. Only about one-third of industrial jobs were retained (Hoffmann, 2021; Ragnitz, 2020). At the same time, the pressure to continuously invest in one’s human capital also increased at the individual level (lifelong learning; English & Mayo, 2021; Fey, in press). The presumed positive and protective effects of such training courses are a recurring topic of discussion. Empirical historical studies of industrial workers in Germany are even rarer than those of other sectors of the economy.

Table 1.

Descriptive Statistics (Period 1)

	N	%	Mean	SD
Continuing education in 1986–1989
No	1,353	84.14
Yes	255	15.86
Unemployed 1990–1999
No	1,092	73.39
Yes	396	26.61
Job loss 1974–1984
No	1,742	82.64
Yes	366	17.36
Education
Elementary, primary, secondary level I	758	35.21
Secondary level II, intermediate level	1,159	53.83
Tertiary level	236	10.96
Industry sector
Mining and quarrying	41	1.89
Manufacture of food products, beverages, and tobacco	186	8.58
Manufacture of textiles and textile products and manufacture of leather and leather products	189	8.71
Manufacture of wood and wood products	36	1.66
Manufacture of pulp paper and paper products, publishing, and printing	122	5.62
Manufacture of coke refined petroleum products and nuclear fuel	6	0.28
Manufacture of chemicals chemical products and man-made fibres	160	7.38
Manufacture of rubber and plastic products	86	3.96
Manufacture of other non-metallic mineral products	58	2.67
Manufacture of basic metals and fabricated metal products	492	22.68
Manufacture of machinery and equipment n.e.c.	230	10.60
Manufacture of electrical and optical equipment	255	11.76
Manufacture of transport equipment	225	10.37
Manufacturing n.e.c.	83	3.83
Sex
Male	1,538	70.91
Female	631	29.09
Company size
<20	270	12.63
20–199	587	27.46
200–1999	671	31.38
>2000	610	28.53
Employment status
Full-time	2,031	93.64
Part-time	138	6.36
Fixed-term contract
No	1,895	94.23
Yes	116	5.77
Migration background
No	1,215	56.02
Direct	941	43.38
Indirect	13	0.60
Job at risk
No	1,966	92.39
Yes	162	7.61
Personal economic concerns
Major concerns	388	18.19
Some concerns	1,076	50.45
No concerns	669	31.36
General economic concerns
Major concerns	619	28.90
Some concerns	1,201	56.07
No concerns	322	15.03
Share of unemployed time 1990–1999	1,488		6.59	15.75
Birth year	2,169		1,945.81	11.56
Occupational status (ISEI88)	2,115		37.64	12.27
Income (Netto)	2,007		982.78	481.20

Note. Sample of people who worked in the industrial sector in 1985. All variables refer to 1985, except job loss 1974-1984 (1984) and continuing education between 1986–1989 (1989).

Table 2.

Descriptive Statistics (Period 2)

	N	%	Mean	SD
Continuing education in 1990–1993
No	1,942	80.82
Yes	461	19.18
Unemployed 1994–1999
No	1,582	72.97
Yes	586	27.03
Education
Elementary, primary, secondary level I	728	25.61
Secondary level II, intermediate level	1,675	58.92
Tertiary level	440	15.48
Industry sector
Mining and quarrying	106	3.69
Manufacture of food products, beverages, and tobacco	223	7.76
Manufacture of textiles and textile products and manufacture of leather and leather products	230	8.00
Manufacture of wood and wood products	39	1.36
Manufacture of pulp paper and paper products, publishing, and printing	138	4.80
Manufacture of coke refined petroleum products and nuclear fuel	7	0.24
Manufacture of chemicals chemical products and man-made fibres	242	8.42
Manufacture of rubber and plastic products	104	3.62
Manufacture of other non-metallic mineral products	90	3.13
Manufacture of basic metals and fabricated metal products	623	21.68
Manufacture of machinery and equipment n.e.c.	314	10.93
Manufacture of electrical and optical equipment	389	13.54
Manufacture of transport equipment	278	9.67
Manufacturing n.e.c.	91	3.17
Sex
Male	1,949	67.81
Female	925	32.19
Company size
<20	289	10.14
20 – 199	687	24.11
200 – 1999	979	34.36
>2000	894	31.38
Employment status
Full-time	2,649	92.17
Part-time	225	7.83
Migration background
No	1,911	66.49
Direct	906	31.52
Indirect	57	1.98
Personal economic concerns
Major concerns	588	20.60
Some concerns	1,472	51.58
No concerns	794	27.82
General economic concerns
Major concerns	705	24.68
Some concerns	1,533	53.68
No concerns	618	21.64
Find an equivalent job
Easy	667	23.29
Difficult	1,714	59.85
Almost impossible	483	16.86
East German
No	2,079	72.34
Yes	795	27.66
Share of unemployed time 1994-1999	2,168		9.65	21.38
Birth year	2,874		1,951.14	11.57
Occupational status (ISEI88)	2,858		38.68	13.13
Income (Netto)	2,749		998.03	614.74

Note. Sample of people who worked in the industrial sector in 1990. All variables refer to 1990, except continuing education between 1990-1993 (1993).

Therefore, this research examines the extent to which continuing vocational education and training (CVET) could protect workers in industrial sectors from future unemployment, following the research question: What was the effect of Continuing Vocational Education and Training on Unemployment in the Industrial Sector? Two time periods are selected from the German Socio-Economic Panel (SOEP) to estimate this effect. The aim is to include both East and West Germany, and to cover a wide period. A radius-matching method is used to assign statistical twins who did not participate in CVET to each treated person who did. The aim is to simulate a counterfactual state to isolate the effect of CVET between 1986 and 1989 on future unemployment in West Germany (n = 2,169) and the effect of participation in CVET between 1990 and 1993 on future unemployment in Germany as a whole (n = 2,874).

The study is structured as follows: First, it provides a broader overview of the context of deindustrialisation, examines industrial workers’ participation in continuing education, and reviews existing research on the topic. This is followed by the theoretical background from which the research hypotheses are derived. Then, the data and methods used to investigate the research question are presented. Finally, the results are reported and discussed.

Background

Deindustrialisation and Continuing Education

Technological change and the globalisation of the economy have shifted the production of labour-intensive intermediate goods to countries with lower personnel costs (Raphael, 2019). This often threatened industrial workers whose jobs were at risk of disappearing, and began a process often referred to as deindustrialisation (Raphael, 2019). At the same time, the structure of the potential labour force shifted toward much higher skill levels as a result of greater access to education and expanded educational opportunities. Rapid technological change was devaluing qualifications (Müller, 1978, p. 10), and there were concerns that entire occupational fields would become obsolete. There were also debates about the decreasing importance of Germany’s dual vocational training system, which was considered an economic pillar of the German industrial model (Baethge & Wolter, 2015). More education was seen as necessary in a ‘knowledge society’ in a ‘post-industrial’ era (Bell, 1976). Thus, there were supply- and demand-side-induced changes in the qualification structure (Falk & Koebel, 1998, p. 362). This increasingly called into question the value of initial education and increased individual pressure on workers to continue their education to remain employable (Hübler, 1998, p. 99).

Jensen (2011) has shown that in more coordinated market economies, such as Germany, deindustrialisation led to an increase in public spending on education between 1980 and 2000 relative to liberal market economies. Continuing education became an important location factor for regional competitiveness and a central instrument for adapting to global economic and social change.

In Germany, a socio-politically and socially integrative motivated phase began at the start of the 1990s, in which the FuU¹ (Further training and retraining measures) funds were also expanded considerably for a short time (Lechner & Wunsch, 2009, p. 661; Schrader, 2011, pp. 202, 204). The focus of ‘career improvement measures, […] has shifted toward removal of the skill deficits and skill mismatch of the unemployed’ (Lechner & Wunsch, 2009, p. 661). More than 50% of the working population in East Germany participated in qualification measures between 1990 and 1994 (Hübler, 1998, p. 97). This increase in (re)training was directly related to the reunification that characterised East Germany in the 1990s (Lechner, 1995). Unemployment, which was temporarily cushioned by involuntary short-time work (‘Kurzarbeit’), rose rapidly. To mitigate this transformation, the state had spent DM 26 billion on training and retraining by 1993 (Lechner, 1995). The political actors of state labour market policy have pursued the goal of the bridging and structural function of active labour market policy. This led to extensive financing of continuing training and retraining measures, job creation measures with qualification components, and short-time work combined with FuU (Further training and retraining measures) activities (Pannenberg, 1995, pp. 142–143).

Participation of Industrial Workers in CVET

This increase in participation in CVET was socially selective, favouring those who already had higher levels of education and social status, and the benefits were probably not the same across all occupational groups (Büchel & Pannenberg, 2004). In particular, the type of employment and the industry sector affected the likelihood of participation. The content of this training was rarely in-depth, but mostly supplementary. To this day, the group of (industrial) workers is stigmatised as being more ‘educationally distant’ (Bremer et al., 2015) and having less access to educational opportunities than ‘white-collar’ workers. This is particularly true for low-skilled industrial workers, as companies are often less willing to support their participation in training. Using current data, Heß and Leber (2025) presented evidence that company support in the form of payment of course fees or paid leave reduces differences in participation. At the same time, structural change has often had the greatest impact on workers in the industrial sector, particularly low-skilled industrial workers. Two popular publications from the 1960s and 1970s, which were in the spirit of educational expansion and aimed at making education accessible to a wider population, describe this problem in an exemplary way: ‘Warum kommen nur wenige Industrie-Arbeiter in die Volkshochschulen?’ (Why do few industrial workers come to adult education centres?) by Tietgens (1978) ²and ‘Bedingungen politischer Lernbereitschaft bei Industriearbeitern’ (Conditions of political willingness to learn among industrial workers) by Müller (1978). Tietgens provides a deeper analysis of the reasons for the low participation of industrial workers in educational programmes. He shows that participation in adult education events increased with the duration and type of schooling (Tietgens, 1978, p. 145). Children from the middle classes started school with a considerable educational advantage, meaning that children from lower classes had to work harder, contributing to a class-differentiated mobilisation of ‘educational energies’ (Tietgens, 1978, p. 146). Therefore, it is very much about the impact of social background on children’s and adolescents’ development. The theoretical proximity of Tietgen’s study to Bourdieu’s concept of Habitus is apparent. Bourdieu’s major work La Distinction (1982) was published in German translation 4 years after Tietgen’s text.

Müller (1978), on the other hand, presents empirical results showing that participation in CVET among industrial workers was not as low as previously thought. More than half of all industrial workers participated in continuing education during their working lives, mostly of a vocational nature. These courses were often closely related to occupational status and tasks. This did not necessarily imply a promotion – rather, it served to preserve employability (Müller, 1978, p. 29). About a quarter of industrial workers also occasionally or frequently participated in training in areas unrelated to their work, often in trade union training courses (Müller, 1978, p. 29). In the case of CVET, companies dominated as institutions and providers (Müller, 1978, p. 30). Müller (1978) observed that industrial workers who had received initial training but had not completed a formal apprenticeship and were no longer working in their profession adopted a more defensive orientation aimed at securing their social and material existence. In contrast, industrial workers who had received systematic training and were working in their chosen profession tended to take a more proactive approach to further training, focusing on professional development and skill enhancement. Low time budgets, e.g. due to shift work and heavy workloads, severely limited active continuing education behaviour and plans (Müller, 1978, p. 33). Workers’ assessment of their professional future was also an important factor in their willingness to engage in continuing education. Dissatisfaction with one’s situation may indicate a greater willingness to participate in continuing education (Müller, 1978, p. 114). Several other studies have examined the social factors associated with participation in continuing education, not only among industrial workers (e.g., Behringer, 1995; Erler, 2018; Fey, 2025; Lee & Desjardins, 2019; Schiersmann, 2007).

Effects of CVET on Labour Market Outcomes

The previous section examined the disadvantages that industrial workers often face in participating in continuing education. This section focuses more specifically on research in Germany related to the paper’s main question: the effect of CVET on future unemployment. Particular focus is given to older historical studies.

The literature distinguishes between training for people in employment and training for re-entry into employment (on-the-job training and off-the-job training; Haelermans & Borghans, (2012), p. 504). CVET in the current study is categorised as non-formal on-the-job training, which does not necessarily mean that the training took place at the workplace, but rather that the respondents were employed. Off-the-job training has been extensively researched, with mixed results regarding its effect on re-employment (e.g., Fitzenberger & Speckesser, 2007; Hofbauer & Dadzio, 1982; Lechner, 1995; Lechner & Wunsch, 2009; Pannenberg, 1995). Using more recent data from 2009 to 2016, Ebner and Ehlert (2018) show that non-formal education reduces both upward and downward job mobility. In addition to employment history, income is another common outcome of continuing education that has been examined in many studies concerning on-the-job training. A meta-analysis of this topic from 1981 to 2010 is provided by Haelermans and Borghans (2012). In particular, the study by Ehlert (2017) provides interesting insights that support the assumption that monetary returns differ across labour market contexts. Several studies also investigate the non-pecuniary returns of work-related training (e.g., Hartmann & Kuwan, 2011; Ruhose et al., 2019; Schrader et al., 2020).

Previous research has often examined the effect of unemployment on different groups. Pannenberg (2001) estimates the effect of on-the-job CVET on the future risk of unemployment in West Germany from 1986 to 1997, focusing on gender differences. He reports a lower risk of unemployment 3 years after completing a CVET. This effect persists for men after 6 years, but for women it appears only after 3 years. Büchel and Pannenberg (2004) examine the effect of on-the-job CVET on unemployment after its provision between 1984 and 2001. The study focuses on heterogeneity in age and East/West Germany. They find that younger workers could earn higher incomes and reduce their future risk of unemployment through CVET. They do not find this effect for older workers. No differences in age groups were found between East and West Germany (Büchel & Pannenberg, 2004). Dieckhoff (2007) analyses different labour market outcomes in a cross-national study. Using the European Community Household Panel (1994–1996), she finds no statistically significant effect of continuing education on the transition from employment to unemployment. Fitzenberger and Prey (1998) provide a review and critique of methods for evaluating the impact of training measures in East Germany, along with an overview of empirical studies investigating this issue. Lechner’s work (1999, 2000) on East Germany is particularly interesting because he uses a matching approach to strengthen causal inference. In his 1999 study, he examines the effects of CVET in East Germany between 1990 and 1994. He finds no effect on the probability of unemployment but a large effect on earnings (Lechner, 1999). He attributes this missing effect on unemployment to the special post-reunification labour market situation in East Germany. In the 2000 article, Lechner examines the effects of public-sector-sponsored training and retraining on East German workers, concluding that publicly funded training had no impact on individual labour market opportunities in East Germany. A recent study by Rupieper and Thomsen (2024) also finds no evidence of an effect of ‘voluntary adult education’ in adult education centres on unemployment in East Germany after reunification. A more recent study supports the finding that a German training voucher programme for work-related training of employed workers in 2010 had no impact on wages and employment (Görlitz & Tamm, 2016). The results for the years 2007 to 2010 suggest a different outcome. Dauth (2020) finds that further training subsidies for low-skilled workers improved labour market outcomes (employment and earnings). Similarly, Christensen (2001) finds a positive effect of continuing vocational education and training on workers’ risk of job loss in West Germany.

While the effectiveness of training on employment trajectories across the years has been widely studied, there is a gap in research on the specific group of industrial workers. A group that was undoubtedly particularly important in the context of deindustrialisation.

Theory

This section introduces theoretical perspectives that could explain how continuing vocational education and training (CVET) may affect unemployment. In light of the competitive nature of the German labour market during the 1990s and the mounting pressure on workers, the focus is on how employers evaluate workers’ training investments.

According to human capital theory (Becker, 1964; Mincer, 1974), education enhances productivity, which benefits employers by raising output and increasing the cost of replacing trained workers. From the worker’s perspective, such investments yield returns in the form of higher wages and greater employment stability. In particular, firm-specific human capital constitutes sunk costs that are lost to the employer if the employment relationship ends, providing an additional incentive to retain trained workers (Pannenberg, 2001, p. 277).

A related perspective is provided by job-matching models (Jovanovic, 1979a; 1979b), which emphasise that employers and workers learn over time whether they are a good match. Training can improve this matching process, making continued employment more likely. In addition, Spence (1973) signalling theory highlights that training may serve as a signal of productivity, influencing employers’ hiring and retention decisions even beyond its direct productivity effects.

Finally, from a psychological perspective, the Theory of Work Adjustment (Dawis & Lofquist, 1968) posits that training enhances workers’ skills and job satisfaction, which can in turn contribute to better performance and stronger employer valuation.

Taken together, these approaches suggest multiple mechanisms through which CVET may reduce the risk of unemployment: by increasing productivity, strengthening job matches, signalling commitment and ability, and improving job satisfaction. These theoretical insights provide the basis for examining the impact of CVET on unemployment in Germany between 1989 and 1999.

Study Design

Hypotheses

Combining these perspectives, this study understands participation in continuing vocational training as a central factor for future unemployment risk. Thus, the following hypotheses are formulated:

There was a negative association between participation in CVET and the probability of becoming unemployed in the years following the training for industrial workers.

Because of the particularly risky period of unemployment after reunification and the increasing competition in East Germany’s industrial sector, the effect size for workers in the industrial sector in East Germany during the second treatment period (participation in CVET between 1990 and 1993) should be higher.

The negative association between participation in CVET and the probability of becoming unemployed in the years following the training was higher for industrial workers in East Germany than for industrial workers in West Germany in the second period.

Data

This study relies on the German Socio-Economic Panel (SOEP; Liebig et al., 2022). The SOEP is provided by the German Institute for Economic Research (DIW) and has been conducted annually since 1984. It is one of the best-known panel surveys in Germany. In the 1980s and 1990s, the survey focused on continuing education in 1989 and 1993. For the analysis, only persons employed in the industrial sector in 1985 (first treatment period; n = 2169) or 1990 (second treatment period; n = 2874) are included in the sample. This selection ensures that the variables used to define the sample are measured prior to the treatment (i.e., participation in CVET), thereby avoiding potential reverse causality. The sample includes all individuals employed in the industrial sector, making it very diverse. It also includes workers who work primarily in the company’s administration or sales, as well as highly skilled workers. The industrial sector is chosen because it was heavily affected by structural change during the 1980s and 1990s, making it particularly relevant for studying the effects of CVET participation. Occupations in the industrial sector are identified using the Statistical Classification of Economic Activities in the European Community (NACE Rev. 1.1; two-digit level; European Commission, 1996).³

Participation in CVET has been surveyed retrospectively over the previous 3 years for any job-related training or courses, including those still ongoing. This retrospective survey results in two treatment periods: 1986–1989 and 1990–1993.

Unfortunately, it is difficult to include additional information about the courses in the matching models because they refer to different courses (the latest three or the most important one professionally) and are hard to distinguish from one another. After the matching model calculations, a closer look is taken at the courses the participants attended to gain at least a descriptive impression of the types of courses to which the effects relate (Table 6).

The variables used in the matching procedure should predict participation in training. Therefore, they should precede the treatment. For the first period, the information from 1985 is used. This avoids any overlap with CVET courses that were asked about retrospectively back to 1996.

Since East Germany was included in the SOEP in 1990, information on the matching variables from 1990 is used for the second treatment period. Therefore, it is conceivable that in some cases, during the second treatment period, continuing education had already occurred a few months before the matching information was collected.

Description

The following variables are used for matching: The birthyear, sex (binary), level of education (ISCED 97; in three categories), size of the company (<20; 20–199; 200–1,999; >2,000), occupational status (ISEI-88), industry-sector (NACE Rev. 1.1; two-digit), which is further summarised for the analysis based on the European Commission (1996, pp. 35–43), employment status (full- or part-time), income (net), and employment contract (fixed-term or indefinite; not collected in East Germany in 1990, so it is only used in the first treatment period; Table 1), migration background (direct: not born in Germany; or indirect: father or mother have a migration background), future expectations regarding job loss in the next 2 years (four categories recoded to binary; not collected for West Germany in 1990, so it is only used in the first treatment period), general and personal economic concerns (three categories), and assessment of the likelihood of finding an equivalent job (only available from 1987 onwards and therefore only used in the second period; Table 2). The first treatment period also included a variable indicating whether there was any unemployment between 1974 and 1984. For the second treatment period, the distinction between East and West Germany is added.

The central outcome variable indicates whether respondents were registered as unemployed in the last year and, if so, for how many months. This is used in different ways in different models. First, as unemployment for each year separately, after the treatment period. Second, the proportion of time spent unemployed after the treatment period. Moreover, third, being unemployed at all after the treatment period.

Methods

To improve the validity of causal inference, a radius-matching procedure is used to assign statistical twins who did not participate in CVET (untreated/control group) to each treated person (who participated in CVET). The aim is to simulate a counterfactual state to isolate the effect of CVET on future unemployment. A counterfactual state describes the hypothetical situation in which a subject exposed to an intervention would be under the same conditions if he or she had not been exposed to the intervention. In the hypothetical existence of the counterfactual state, the subject in the two states (factual vs. counterfactual) would differ only in terms of participation or non-participation in CVET and would otherwise be identical (Müller, 2012, p. 2). As selection into continuing education varies considerably across sociodemographic groups, which in turn affects unemployment, it is to be expected that those who have participated in continuing education are already more likely to have a better career path. It is therefore necessary to control for these selection effects using radius-matching.

For this purpose, the presented matching variables (Chapter 4.3.) are used, which are assumed to influence both the decision to participate in training and the outcome of being unemployed. First, a logistic regression is calculated to estimate the probability of receiving treatment given the observed covariates (propensity score):

P (T_{i} = 1 ∣ X_{i}) = \frac{e^{{X_{i} β}}}{1 + e^{{X_{i} β}}}

T_i = 1 for treated units and Ti = 0 for control units. X_i is the set of covariates (matching variables). β is the estimated coefficient from the logistic regression.

For the matching procedure, the Stata ado radiusmatch is used. It implements a one-to-many calliper matching algorithm that uses all comparison observations within a specified distance of the treated unit’s propensity score (e.g., Huber et al., 2012).

The matching calculation is set to use the maximum allowable distance as the 90th percentile of the distance distribution, so that only 90% of potential matches are considered. The matching radius is set to 300% of the distance to the largest one-to-one match. This prevents bad matches and balances match quality and sample size (Huber et al., 2015). For each treated unit, all control units within the calliper range are considered as potential matches, and multiple matches are allowed.

After matching, the outcomes of treated and matched control units are compared to estimate the Average Treatment Effect on the Treated (ATET):

ATET = E [Y_{1} - Y_{0} ∣ T = 1]

Where Y₁ is the potential outcome in case of treatment (CVET), and Y₀ is the potential outcome in the absence of treatment.

Results

The results are separated into three parts. First, the logistic regression on which the propensity score is based is presented – this score is used to match the treated group to the untreated group. Then the number of matched cases and the propensity distribution are shown, and finally, the ATET of the models are presented.

Logistic Regression

Table 3 presents the results of the logistic regressions used to estimate propensity scores for the matching procedure. In the first period (1986–1989), having secondary II or intermediate education increased the probability of participating in CVET significantly by 11.5% compared to individuals with elementary, primary, or secondary I education, while tertiary education increased the probability significantly by 15.6%. A one-point increase in occupational status (ISEI) was significantly associated with a 0.5% higher probability of participation, and an increase of 1,000 in income raised the probability significantly by 4.3%. Fixed-term contracts increased the probability of participation significantly by 9.9% relative to permanent contracts, whereas each additional year of age reduced participation significantly by 0.6%. Individuals reporting no concerns about the general economic situation had a 7.2% significantly lower probability of participating than those with major concerns.

Table 3.

Logistic Regressions of Matching Variables on CVET Participation

	Period 1 AME (SE)	Period 2 AME (SE)
Unemployed 1974-1984	−0.009
Unemployed 1974-1984	(0.023)
Age	−0.006***	−0.005***
Age	(0.001)	(0.001)
Educ. Secondary II/Intermed.	0.115***	0.102***
Educ. Secondary II/Intermed.	(0.020)	(0.021)
Education tertiary level	0.156***	0.148***
Education tertiary level	(0.040)	(0.034)
Female	−0.006	−0.029
Female	(0.024)	(0.020)
Company size 20–199	−0.030	0.000
Company size 20–199	(0.033)	(0.028)
Company size 200–1999	−0.023	0.011
Company size 200–1999	(0.033)	(0.027)
Company size >2000	0.042	0.060*
Company size >2000	(0.034)	(0.029)
Part-time	−0.035	0.010
Part-time	(0.051)	(0.036)
Occupational status (ISEI88)	0.005***	0.005***
Occupational status (ISEI88)	(0.001)	(0.001)
Income (in 1,000)	0.043*	0.027
Income (in 1,000)	(0.021)	(0.015)
Fixed-term contract	0.099**
Fixed-term contract	(0.037)
Migration background - direct	−0.031	−0.090***
Migration background - direct	(0.022)	(0.020)
Migration background - indirect	−0.079	−0.116**
Migration background - indirect	(0.071)	(0.041)
Job at risk	−0.026
Job at risk	(0.037)
Pers. Econom. - Some concerns	0.043	−0.020
Pers. Econom. - Some concerns	(0.025)	(0.023)
Pers. Econom. - No concerns	0.056	−0.017
Pers. Econom. - No concerns	(0.029)	(0.028)
Gen. Econom. - Some concerns	−0.028	−0.015
Gen. Econom. - Some concerns	(0.021)	(0.020)
Gen. Econom. - No concerns	−0.072*	0.019
Gen. Econom. - No concerns	(0.030)	(0.028)
East German		0.040
East German		(0.023)
Equiv. Job - difficult		−0.009
Equiv. Job - difficult		(0.019)
Equiv. Job - almost impossible		−0.036
Equiv. Job - almost impossible		(0.028)
N	1,308	2,237
Pseudo-R²	0.27	0.19
AIC	885	1,836
BIC	1,056	2,019

Note. Logistic regression models, Persons working part or full time in industry sector; NACE sectors not shown (no significant effects); average marginal effects, standard error in parenthesis.

*p < 0.05; **p < 0.01; ***p < 0.001.

In the second period (1990–1993), the significant effects of age, education and occupational status remained similar, while additional significant effects emerged. Employment in companies with more than 2,000 workers increased the probability of participation by 6.0% relative to firms with fewer than 20 workers. Having a direct migration background reduced the probability significantly by 9.0%, and an indirect migration background by 11.6%. Notably, the effects of income and general economic concerns were no longer significant in this period. All reported changes represent average marginal effects (AMEs) derived from the logistic regression models.

Matching

Table 4 presents the numbers of treated and untreated (control) observations across the different models. Models m1 to m10 each use unemployment in a specific year between 1990 and 1999 (i.e. 1990, 1991, etc.) as the outcome variable. The same applies to models m13 to m18, which use unemployment in individual years between 1994 and 1999 for the second period. The number of cases should provide sufficient statistical power for further analysis. m11 and m12, like m19 and m20, use the same matching algorithm because their outcomes (share of time unemployed after the treatment period and being unemployed at all after the treatment period) are based on the same variable and therefore have the same missing cases.

Table 4.

Number of Observations in Different Models

Model	Treated	Control	Total
Period 1
m1: 1990	179	1,020	1,199
m2: 1991	173	983	1,156
m3: 1992	167	957	1,124
m4: 1993	160	899	1,059
m5: 1994	153	856	1,009
m6: 1995	153	809	962
m7: 1996	149	789	938
m8: 1997	143	734	877
m9: 1998	134	701	835
m10: 1999	130	675	805
m11: Share	181	1,034	1,215
m12: Total	181	1,034	1,215
Period 2
m13: 1994	383	1,610	1,993
m14: 1995	360	1,536	1,896
m15: 1996	359	1,471	1,830
m16: 1997	336	1,382	1,718
m17: 1998	323	1,317	1,640
m18: 1999	318	1,260	1,578
m19: Share	388	1,628	2,016
m20: Total	388	1,628	2,016

Figure 2 shows the propensity score distribution, as an example, for models m11/12 in the first period. Almost all treated persons could be matched to untreated cases. With higher propensity scores, fewer control cases are available for matching. The same holds for the second period, as shown in Figure 3, which displays the propensity score distribution for model m19/20. As a robustness check, the models were also estimated using a matching approach that excluded cases with a propensity score greater than 0.6 (not reported). This did not change the treatment effects. The other models in both periods have similar distributions. This provides a matched sample for the following calculations.

Figure 2.

Propensity score distribution period 1.

Figure 3.

Propensity score distribution period 2.

Table 5 shows the results of the matching balance tests. The pseudo-R² is close to zero after matching, and the LR-Chi²-test is no longer significant, i.e. none of the covariates is suitable for predicting group membership (e.g., Müller, 2012, p. 29). In addition, the average relative bias in variable values between the treatment and control groups has decreased significantly. For the first period before matching, the average bias across all variables is 21.9%. After matching, it is reduced to 5.0%. For the second period, it drops from almost 18.6% to 3.1%.

Table 5.

Covariate Balance Test Indicators

Model	Sample	Pseudo-R²	LR-Chi²	p>chi²	Mean bias
m11/12	Unmatched	0.263	269.95	0.000	21.9
m11/12	Matched	0.024	12.19	0.999	5.0
m19/20	Unmatched	0.198	400.37	0.000	18.6
m19/20	Matched	0.014	15.34	0.992	3.1

Average Treatment Effect on the Treated (ATET)

Before moving to the calculated ATET, Figures 4 and 5 show the trajectories of mean unemployment in the treated and control groups per year descriptively. A small drift apart can be observed after, or even during, the treatment period. This indicates that, on average, industrial workers who participated in CVET seem to have a lower future unemployment rate than those who did not. This supports the assumed hypothesis (H1: There was a negative association between participation in CVET and the probability of becoming unemployed in the years following the training for industrial workers.), while the unspecified timing of treatment may explain why this trend begins slightly during the 3-year period of potential CVET.

Figure 4.

Unemployment rate over time (Period 1).

Figure 5.

Unemployment rate over time (Period 2).

A stronger drift or greater differences cannot be observed in the exclusively East German sample (Figure 6), which does not support the second hypothesis (H2: The negative association between participation in CVET and the probability of becoming unemployed in the years following the training was higher for industrial workers in East Germany than for industrial workers in West Germany in the second period.).

Figure 6.

Unemployment rate over time (Period 2, east Germany only).

Figures 7 and 8 show the ATET of all calculated matching models. A negative effect refers to a lower risk of unemployment for treated cases after the treatment. There is a small trend toward more negative treatment effects in the years 1990–1993, which were the years closest to the time of treatment (m1: −0.008; m2: −0.013; m3: −0.013; m4: −0.028). And similarly for the years of the second period 1994–1997 (m13: −0.048; m14: −0.042; m15: −0.007; m16: −0.007). In the later years 1994–1998, after the treatment for the first period, quite positive effects are observable (m5: 0.042; m6: −0.007; m7: 0.024; m8: 0.027; m9: 0.021). The same is true for the later years of the second period, 1998–1999 (m17: 0.026; m18: 0.013). However, this trend is not entirely consistent and not statistically significant. The estimated ATETs show no statistically significant difference between the treated and control groups with respect to the risk of unemployment in each year after the treatment, neither in the models of period one nor in the models of period two. The models testing the effect on the share of time a person was unemployed after the treatment show a small negative effect for the first period (m11: −0.008) and a small positive effect for the second period (m19: 0.012). They also show no statistical significance. The models testing whether there was any effect on unemployment after the treatment are both negative and not statistically significant (m12: −0.026 for the first period and m20: −0.014 for the second period). They do not differ in the direction of their effects between the two periods.

Figure 7.

Average treatment effect on the treated in treatment period 1.

Figure 8.

Average treatment effect on the treated in treatment period 2 (total/east Germany).

Furthermore, the effects for the East Germany-only models (grey dots in Figure 8) show no clear pattern: some are positive, others negative, and none are statistically significant. In some models, the direction of the effect changes compared to the models for all of Germany (m13 and m20), but these differences are not statistically significant.

So far, the calculated models have not distinguished between different vocational courses. To give an idea of the types of courses participants attended, information about a few of them is presented below (Table 6).

Table 6.

Description of Visited Courses of Treated

Category	Period 1 (n)	Period 1 (%)	Period 2 (n)	Period 2 (%)
Treated people	181		388
Reported course length
One day	6	3.3	38	9.8
Up to 1 week	46	25.4	160	41.2
Up to 1 month	16	8.8	46	11.9
Up to 3 months	4	2.2	50	12.9
Up to 1 year or more	9	5.1	94	24.3
Aim of the course
Adapting to new job requirements	67	37.0	276	71.1
Retraining	0	0.0	31	8.0
Training at new workplace	5	2.8	53	13.7
Qualify for promotion	35	19.3	145	37.4
Participation without financial support
Yes, sure!	25	13.8	119	30.7
Maybe	22	12.2	80	20.6
No	17	9.4	88	22.7
Initiative
Own	34	18.8	177	46.0
Company	18	9.9	107	27.6
Half-half	29	16.0	101	26.0
Financial support
From employer	60	33.2	220	56.7
No	17	9.4	99	25.5

Note. Sample of people who worked in the industrial sector and visited CVET before the treatment periods (treatment groups of models m11/m12 and m19/m20). Percent includes missing values. ‘Reported course length’ refers to the longest of the last three courses attended. ‘Aim of the course’ also refers to the three most recent courses; multiple answers were allowed, and here it counts if at least one of the three courses had this aim. ‘Participation without financial support’, ‘Initiative’ and ‘Financial support’ refer to the course that was considered as the most important from a professional perspective.

A total of 181 people were in the treatment group in the first period, compared to 388 in the second. In most cases, the longest course among the last three visited was up to 1 week (25.4% in the first period and 41.2% in the second).⁴ Participants primarily aimed to adapt to new job requirements (37.0% and 71.1%, respectively) and qualify for promotion (19.3% and 37.4%). No one reported retraining in the first period, whereas 8.0% did so in the second. Of the three possible courses, the one considered most professionally important was then specified. A total of 13.8% in the first period and 30.7% in the second indicated that they would also have attended this course if they had not received financial support from their employer, the Employment Office, or another agency. 12.2% and 20.6% reported ‘maybe’, while 9.4% and 22.7% respectively reported that they would not have attended the course. When asked about their motivation for attending the course, 18.8% (first period) and 46.0% (second period) reported that it was their own initiative. 9.9% and 27.6% reported that their company initiated it, and 16.0% and 26.0% reported a joint initiative. Regarding the most important course, 33.2% and 56.7% respectively stated that they had received financial support from their employer. Meanwhile, 9.4% and 25.5% said that they had not received any support.

Discussion

The group of workers in the industrial sector in the 1990s is often reported to be rather underrepresented in adult education and particularly affected by, and vulnerable to, structural change. However, this group is relatively under-researched, which is problematic given that structural changes and deindustrialisation, as well as changes in employment structures and work tasks, have increased the importance of (adult) education over the life course (lifelong learning). This raises the question of whether continuing education and, respectively, CVET could actually help to prevent the consequences of these structural transformations, such as unemployment, for workers in the industrial sector. The results of this study do not confirm this hypothesis (H1) and show no significant differences in the effects of participation in a CVET between East and West Germany between 1990 and 1993, thus rejecting the second hypothesis (H2). None of the presented models showed that the group of people who participated in CVET between 1986 and 1989 or 1990 and 1993 statistically differed in their risk of becoming unemployed after CVET from the corresponding matched control group who did not participate in CVET. This finding is consistent with other studies examining different groups and educational offerings, which have not found a causal effect of continuing education on future unemployment (e.g., Dieckhoff, 2007; Lechner, 1999; Rupieper & Thomsen, 2024). Previous research provided several potential explanations for this. For example, Rupieper and Thomsen stated that the realisation of ‘education effects depends on macroeconomic conditions’ (2024, p. 29), meaning that the difficult labour market and high unemployment rates in the late 1980s and 1990s may not have provided sufficient opportunities to realise potential benefits of continuing education. Companies were able to choose already-skilled labour from the pool of the unemployed. This says nothing about the content of the training programmes. Similarly, the study presented was unable to make any statements on that matter.

Limitations

While this study provides new insights into the lack of effect of continuing vocational education and training (CVET) on unemployment among industrial workers in Germany between 1989 and 1999, several limitations must be acknowledged.

The method and study design of the present study impose clear restrictions: The question about training in the SOEP, which was used for the analyses, is rather broad and general. It does not allow for distinguishing in the models between different types of training, i.e., when and where it took place, who paid for it, how long it lasted, and what the content was. These are plausible factors that could affect returns (Ehlert, 2017). Ebner and Ehlert (2018) showed that the effects of CVET on job mobility differ depending on whether it is organised and paid for by the employer or the worker. Some additional information could be added to the description of the attended courses (Table 6) retrospectively, but it was not possible to distinguish the effects of the different courses. It can be assumed that the employment effects of the different courses vary significantly. Short-term courses, safety training, or familiarisation with new machines and computer programmes, which are ordered and paid for by the company, probably have less impact on employment decisions and further job opportunities than long courses related to new tasks or retraining.

In addition, the retrospective survey of participation in training over the previous 3 years is rather imprecise because the exact dates of the courses are unknown. It blurs temporal differences in the extent of the effect. It is conceivable that the effect varies in strength over time, which is what the different models attempt to capture by using different years for their results. However, this may not be adequately captured by the models, as participation may have occurred one or 3 years earlier. In addition to the weaker coverage of the treatment, the sample of industry workers is rather small. This could reduce statistical power. Furthermore, the sample is diverse within the industrial sector, covering various activities and positions across different areas of the ‘secondary sector’. This was taken into account in the matching process so that a treated person was matched with a similar person, but in the average effect, this was not differentiated again. It is conceivable that the effects differed across groups within the sample.

Conclusion

Despite this study’s limitations, the findings provided contribute to the growing literature on the effectiveness of CVET during periods of economic transition. The historical findings may have implications for policymakers and practitioners designing training initiatives in response to contemporary labour market challenges. Governance in this area is complicated, and the real impact of the training provided is often not guaranteed. Any policy aimed at increasing participation in continuing training must also be aware of the unintended side effects. Displacement, substitution, and deadweight effects may weaken or neutralise the positive effects, or even lead to an overall negative outcome (Büchel & Pannenberg, 2004, p. 123).

The results do not, per se, call into question the usefulness of continuing education as a means of mitigating the negative consequences of structural change. Continuing education is more than just vocational training with an assumed impact on unemployment. It is also about engaging people politically and empowering them to shape their lives independently and self-determinedly. As economies continue to undergo rapid technological and structural transformations, understanding the role of skill enhancement in fostering labour market resilience is more important than ever. Thus, this study underscores the need for targeted, context-sensitive interventions that address both the immediate risks and long-term opportunities associated with structural change.

Footnotes

Acknowledgments

The project Confronting Decline: Challenges of Deindustrialization in Western Societies since the 1970s (CONDE), in cooperation between the German Institute for Adult Education (DIE), the Institute for Contemporary History (IfZ), and the Luxembourg Centre for Contemporary and Digital History (C²DH), is funded by the Leibniz Competition Cooperative Excellence. Thanks to Andreas Martin and Max P. Jansen for proofreading and/or helpful comments.

ORCID iD

Jonas Fey

Funding

The author disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was supported by Leibniz Competition Cooperative Excellence.

Declaration of Conflicting Interests

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

Data Availability Statement

For the analyses and data preparation, Stata version 18.0 was used. The syntax is available on OSF: .

Notes

References

Baethge

Wolter

(2015). The German skill formation model in transition: From dual system of VET to higher education? Journal for Labour Market Research, 48(2), 97–112. https://doi.org/10.1007/s12651-015-0181-x

Becker

G. S.

(1964). Human capital: A theoretical and empirical analysis, with special reference to education. NBER.

Behringer

(1995). Weiterbildung in Ost-und Westdeutschland. Vierteljahreshefte zur Wirtschaftsforschung, 64(1), 26–51.

Bell

(1976). The coming of post-industrial society. Basic Books.

Bourdieu

(1982). Die feinen Unterschiede: Kritik der gesellschaftlichen Urteilskraft. Suhrkamp-Taschenbuch-Verl. Suhrkamp Verlag KG.

Bremer

Kleemann-Göhring

Wagner

(2015). Weiterbildung und Weiterbildungsberatung für “Bildungsferne”: Ergebnisse, Erfahrungen und theoretische Einordnungen aus der wissenschaftlichen Begleitung von Praxisobjekten in NRW. W. Bertelsmann Verlag GmbH & Co. KG.

Büchel

Pannenberg

(2004). Berufliche Weiterbildung in West-und Ostdeutschland: Teilnehmer, Struktur und individueller Ertrag. Journal for Labour Market Research, 37(2), 73–126.

Christensen

(2001) Berufliche Weiterbildung und Arbeitsplatzrisiko: ein Matching-Ansatz (1033). Kiel Working Paper.

Dahrendorf

(1978). Wenn uns die Arbeit ausgeht. Zeit.

10.

Dauth

(2020). Regional discontinuities and the effectiveness of further training subsidies for low-skilled employees. ILR Review, 73(5), 1147–1184. https://doi.org/10.1177/0019793919885109

11.

Dawis

R. V.

Lofquist

L. H.

(1968). A psychological theory of work adjustment. University of Minnesota Press. https://vpr.psych.umn.edu/sites/vpr.umn.edu/files/2022-05/Dawis_Lofquist/201984/20A/20Psychological/20TWA.pdf

12.

Dieckhoff

(2007). Does it work? The effect of continuing training on labour market outcomes: A comparative study of Germany, Denmark, and the United Kingdom. European Sociological Review, 23(3), 295–308. https://doi.org/10.1093/esr/jcm002

13.

Ebner

Ehlert

(2018). Weiterbilden und Weiterkommen? Non-formale berufliche Weiterbildung und Arbeitsmarktmobilität in Deutschland. KZfSS Cologne Journal of Sociology and Social Psychology, 70(2), 213–235. https://doi.org/10.1007/s11577-018-0518-x

14.

Ehlert

(2017). Who benefits from training courses in Germany? Monetary returns to non-formal further education on a segmented labour market. European Sociological Review, 33(3), 436–448. https://doi.org/10.1093/esr/jcx042

15.

English

L. M.

Mayo

(2021). Lifelong learning, global social justice, and sustainability. Springer International Publishing. https://doi.org/10.1007/978-3-030-65778-9

16.

Erler

(2018). Auf den Spuren der Teilnahmeforschung. Historische Entwicklungen und inhaltliche Schwerpunkte. Magazin Erwachsenenbildung, 34. https://doi.org/10.25656/01:15727

17.

European Commission . (1996) In NACE Rev.1: Statistical classification of economic activities in the european community. European Commission.

18.

Falk

Koebel

(1998). Determinanten der qualifikatorischen Arbeitsnachfrage in der westdeutschen Industrie 1978-90: FuE-intensive versus nicht FuE-intensive Industrien. In Pfeiffer

Pohlmeier

für Europäische Wirtschaftsforschung

(Eds.), Qualifikation, Weiterbildung und Arbeitsmarkterfolg. Nomos-Verl.-Ges.

19.

Fey

(2025). It’s the routine, stupid! (or is it?). Journal for Research on Adult Education, 48(2), 151–174. https://doi.org/10.1007/s40955-025-00324-1

20.

Fey

(in press). Von (Aus-)Bildung zum Lebenslangen Lernen Die neoliberale Wende im politischen Diskurs zur Erwachsenenbildung im Kontext der Deindustrialisierung (1970-2000). In Marx

Krebs

(Eds.), Deindustrialisierung westeuropäischer Gesellschaften seit den 1970er Jahren. Wahrnehmung – Reaktionen – Erinnerung (Zeitgeschichte im Gespräch; 38). Metropol.

21.

Fitzenberger

Prey

(1998). Beschäftigungs-und Verdienstwirkungen von Weiterbildungsmaßnahmen im ostdeutschen Transformationsprozeß: Eine Methodenkritik. In Pfeiffer

Pohlmeier

Europäische Wirtschaftsforschung

(Eds.), Qualifikation, Weiterbildung und Arbeitsmarkterfolg (1, pp. 39–95): Nomos-Verl.-Ges.

22.

Fitzenberger

Speckesser

(2007). Employment effects of the provision of specific professional skills and techniques in Germany. Empirical Economics, 32(2–3), 529–573. https://doi.org/10.1007/s00181-006-0088-z

23.

Görlitz

Tamm

(2016). The returns to voucher-financed training on wages, employment and job tasks. Economics of Education Review, 52, 51–62. https://doi.org/10.1016/j.econedurev.2016.01.004

24.

Haelermans

Borghans

(2012). Wage effects of on-the-job training: A meta-analysis. British Journal of Industrial Relations, 50(3), 502–528. https://doi.org/10.1111/j.1467-8543.2012.00890.x

25.

Hartmann

Kuwan

(2011). Gesellschaftliche und individuelle Funktionen von Weiterbildung. In Weiterbildungsbeteiligung 2010. Trends und Analysen auf Basis des deutschen AES: DIE spezial (pp. 17–19). Bertelsmann: Bielefeld. https://doi.org/10.25656/01:6060

26.

Heß

Leber

(2025). Betriebe unterstützen die Teilnahme von Höherqualifizierten an Weiterbildungen häufiger als bei Geringqualifizierten. IAB-Forum. https://www.iab-forum.de/betriebe-unterstuetzen-die-teilnahme-von-hoeherqualifizierten-an-weiterbildungen-haeufiger-als-bei-geringqualifizierten/

27.

Hofbauer

Dadzio

(1982). Die Wirksamkeit von Maßnahmen zur Verbesserung der Vermittlungsaussichten für Arbeitslose nach § 41 a AFG. Mitteilungen aus der Arbeitsmarkt- und Berufsforschung, Institut für Arbeitsmarkt- und Berufsforschung (IAB), 15(4), 426–433.

28.

Hoffmann

(2021). Im Hochgeschwindigkeitszug durch den Strukturwandel: Deindustrialisierung in Ostdeutschland. In Krzyk

(Ed.), Abschied von der Kohle: Struktur-und Kulturwandel im Ruhrgebiet und in der Lausitz: Vol. Band (10751, pp. 46–57). Bpb Bundeszentrale für politische Bildung.

29.

Huber

Lechner

Steinmayr

(2012). Radius matching on the propensity score with bias adjustment: Finite sample behaviour, tuning parameters and software implementation. University of St.Gallen, School of Economics and Political Science, Economics Working. Paper Series No, 1226.

30.

Huber

Lechner

Steinmayr

(2015). Radius matching on the propensity score with bias adjustment: Tuning parameters and finite sample behaviour. Empirical Economics, 49(1), 1–31. https://doi.org/10.1007/s00181-014-0847-1

31.

Hübler

(1998). Berufliche Weiterbildung und Umschulung in Ostdeutschland—Erfahrungen und Perspektiven. In Pfeiffer

Pohlmeier

für Europäische Wirtschaftsforschung

(Eds.), Qualifikation, Weiterbildung und Arbeitsmarkterfolg (pp. 97–132). Nomos-Verl.-Ges.

32.

Jensen

(2011). Capitalist systems, deindustrialization, and the politics of public education. Comparative Political Studies, 44(4), 412–435. https://doi.org/10.1177/0010414010393475

33.

Jovanovic

(1979a). Firm-specific capital and turnover. Journal of Political Economy, 87(6), 1246–1260. https://doi.org/10.1086/260834

34.

Jovanovic

(1979b). Job matching and the theory of turnover. Journal of Political Economy, 87(5 Part 1), 972–990. https://doi.org/10.1086/260808

35.

Lechner

(1995). Effects of continuous off-the-job training in east Germany after unification. ZEW Discussion Papers, 95.

36.

Lechner

(1999). The effects of enterprise-related training in east Germany on individual employment and earnings. Annales d’Économie et de Statistique, 55(56), 97. https://doi.org/10.2307/20076193

37.

Lechner

(2000). An evaluation of public-sector-sponsored continuous vocational training programs in east Germany. The Journal of Human Resources, 35(2), 347–375. https://doi.org/10.2307/146329

38.

Lechner

Wunsch

(2009). Are training programs more effective when unemployment is high? Journal of Labor Economics, 27(4), 653–692. https://doi.org/10.1086/644976

39.

Lee

Desjardins

(2019). Inequality in adult learning and education participation: The effects of social origins and social inequality. International Journal of Lifelong Education, 38(3), 339–359. https://doi.org/10.1080/02601370.2019.1618402

40.

Liebig

Goebel

Grabka

Schröder

Zinn

Bartels

Franken

Gerike

Geschke

S.-C.

Griese

Kara

König

Krause

Kröger

Liebau

Nebelin

Petrenz

Richter

Schupp

German Institute for Economic Research (DIW Berlin) . (2022). Socio-Economic Panel, data from 1984-2020, (SOEP-Core, v37, EU Edition)Sozio-oekonomisches Panel, Daten der Jahre 1984-2020 (SOEP-Core, v37, EU Edition). https://doi.org/10.5684/SOEP.CORE.V37EU

41.

Mincer

(1974). Schooling, experience, and earnings. National Bureau of Economic Research.

42.

Müller

C. E.

(2012). Quasiexperimentelle Wirkungsevaluation mit Propensity Score Matching: Ein Leitfaden für die Umsetzung mit Stata (19). CEval-Arbeitspapier. https://www.ssoar.info/ssoar/bitstream/handle/document/42894/ssoar-2012-muller-Quasiexperimentelle/Wirkungsevaluation/mit/Propensity/Score.pdf/sequence/1&isAllowed/y&lnkname/ssoar-2012-muller-Quasiexperimentelle/Wirkungsevaluation/mit/Propensity/Score.pdf

43.

Müller

H.-F.

(1978). Bedingungen politischer Lernbereitschaft bei Industriearbeitern: Eine empirische Studie zum Zusammenhang zwischen gesellschaftlichem Bewußtsein u. Bildungsinteresse. Metzler. UNESCO.

44.

Pannenberg

(1995). Weiterbildungsaktivitäten und Erwerbsbiographie: Eine empirische Analyse für Deutschland (8). Campus-Verl.

45.

Pannenberg

(2001). Schützt Weiterbildung on-the-job vor Arbeitslosigkeit? In von Weizsäcker

RK.

(Ed.), Bildung und Beschäftigung (pp. 275–291). Duncker and Humblot.

46.

Ragnitz

(2020). Strukturwandel nach Entindustrialisierung. Bundeszentrale Für Politische Bildung. https://www.bpb.de/themen/deutsche-einheit/lange-wege-der-deutschen-einheit/47137/strukturwandel-nach-entindustrialisierung/

47.

Raphael

(2019). Jenseits von Kohle und Stahl: Eine Gesellschaftsgeschichte Westeuropas nach dem Boom. Suhrkamp.

48.

Ruhose

Thomsen

S. L.

Weilage

(2019). The benefits of adult learning: Work-related training, social capital, and earnings. Economics of Education Review, 72(2), 166–186. https://doi.org/10.1016/j.econedurev.2019.05.010

49.

Rupieper

L. K. K.

Thomsen

S. L.

(2024). Can voluntary adult education reduce unemployment? Causal evidence from east Germany after reunification. SSRN Electronic Journal, 16779. https://doi.org/10.2139/ssrn.4717692

50.

Schiersmann

(2007). Weiterbildungsbeteiligung – Stand der Forschung und Vergewisserung des Gegenstandes. REPORT, 30, 33–43.

51.

Schlutz

(Ed.), (1985). Krise der Arbeitsgesellschaft – Zukunft der Weiterbildung. Diesterweg.

52.

Schrader

(2011). Struktur und Wandel der Weiterbildung. Bertelsmann. https://www.die-bonn.de/doks/2011-strukturwandel-01.pdf

53.

Schrader

Ioannidou

Blossfeld

H.-P.

(2020) (Eds) With springer fachmedien wiesbaden. Monetäre und nicht monetäre Erträge von Weiterbildung: = Monetary and non-monetary effects of adult education and training, Vol 7, Springer. https://doi.org/10.1007/978-3-658-25513-8

54.

Spence

(1973). Job market signaling. The Quarterly Journal of Economics, 87(3), 355–374. https://doi.org/10.2307/1882010

55.

Struck

(2016). Krise der Arbeitsgesellschaft? Der Bamberger Soziologiekongress 1982 im Rückblick und Ausblick im Kontext von Digitalisierung, Arbeitsmarkt-und Qualifikationsentwicklung. Working Paper, Issue 17. https://www.uni-bamberg.de/fileadmin/arbeitswiss/Arbeitspapiere/Soziologie-Arbeitswissenschaft-WP-17-Krise_der_Arbeitsgesellschaft-endfassung.pdf

56.

Tietgens

(1978). Warum kommen wenig Industrie-Arbeiter in die Volkshochschulen? In Schulenberg

(Ed.), Erwachsenenbildung (pp. 98–174).

Navigating Industrial Transformation: The Impact of Continuing Vocational Education and Training on Unemployment Risk in Germany’s Industrial Sector (1986–1999)

Abstract

Keywords

Introduction

Background

Deindustrialisation and Continuing Education

Participation of Industrial Workers in CVET

Effects of CVET on Labour Market Outcomes

Theory

Study Design

Hypotheses

Data

Description

Methods

Results

Logistic Regression

Matching

Average Treatment Effect on the Treated (ATET)

Discussion

Limitations

Conclusion

Footnotes

Acknowledgments

ORCID iD

Funding

Declaration of Conflicting Interests

Data Availability Statement

Notes

References