Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Justice perceptions of occupational training subsidies: findings from a factorial survey

Justice perceptions of occupational training subsidies: findings from a factorial survey Workers whose jobs are affected by structural change and digitization are required to continuously adapt their voca- tional skills to the requirements of the labor market. This adaptation is also essential for the competitiveness of their employer firms. The German legislature addressed this issue with investive measures for unemployment insurance, one of which is the Qualification Opportunities Act (Qualifizierungschancengesetz). Funds taken from unemploy- ment insurance can now be used to provide financial help for employers in a more direct way and on a broader scale than before. It became possible that not only unemployed individuals but also workers in companies receive state assistance. This paper analyses the extent to which citizens accept such public support programs for further training and which principles of justice they apply when assessing a just amount of training subsidies. We conducted two factorial surveys. First, we investigated the justice assessments of training subsidies for different types of firms. The results showed that citizens are inclined to subsidize companies by receiving social security funds for further training of their employees. However, when doing so, the principle of needs-based justice should be complied with. Second, we analyze whether citizens think it is just or unjust to provide training subsidies to different workers, as we present them with changing characteristics of workers. The findings confirmed that in addition to the principle of need, views on performance justice, as well as economic considerations are relevant in assessments of whether training subsidies co-financed by unemployment insurance are just. Keywords: Occupational training, Unemployment insurance, Justice assessments, Factorial survey, Multilevel and mixed effects model JEL Classification: C99, D63, I38, J08, J24, J65 1 Introduction been shown to be generally underrepresented among Rapid technical progress and globalization have increased those taking up occupational training (Bassanini et  al. the need for flexible adjustments of the workforce in the 2005). In particular, low-educated workers are often labor market. Investment in human capital can enable involved in undemanding routines with a high risk of workers and firms to adapt to changes in the economic substitution due to technical progress. environment and thus safeguard employment (Acemo- In view of this, the legislature in Germany enacted sev- glu and Restrepo 2017; Brynjolfsson and McAfee 2014; eral changes at the beginning of 2019 (further extended Struck 2006). Participation in training, however, is highly in 2020) with the Qualification Opportunities Act selective: workers in small and medium-sized enterprises (“Qualifizierungschancengesetz”). The Act grants com - (SMEs), older workers, and low-educated workers have panies that want to adapt the qualifications of their work - force due to structural change access to reimbursement for further training from unemployment insurance funds. *Correspondence: richard.wolff@uni-bamberg.de Hereafter also referred to as “the Act”. University of Bamberg: Otto-Friedrich-Universitat Bamberg, Bamberg, Bavaria, Germany Full list of author information is available at the end of the article © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. 6 Page 2 of 18 R. V. Wolff et al. Accordingly, funds from contribution-based unemploy- The article is structured as follows: The background ment insurance are no longer used to exclusively support of the new regulatory framework is presented in Sect.  2. unemployed individuals. This fits with the concept of The legal regulations are concordant with specific prin - a more preventive and investing social policy (Esping- ciples of justice and are partly included in the formula- Andersen et al. 2002; Morel et al. 2012). tion of the hypotheses. These are presented in Sect.  3. With the regulations of the Act, the legislature has sig- We develop a number of hypotheses on the principles nificantly expanded the group of those entitled to unem - of justice that may drive citizens’ assessments of the just ployment insurance funds. Against this backdrop, this amount of subsidies, e.g., related to needs-based justice article examines citizens’ attitudes about these changes. or efficiency-based justice (Leisering 2004). Method and Our research is situated within this framework toward data are described in Sect. 4. Using a factorial survey, we a more expansive and proactively investing “labor insur- ask participants how large training subsidies for different ance” as an expansion of unemployment insurance. Wel- firms should be and how just or unjust training subsi - fare institutions and their regulations for allocation and dies for different kinds of workers are perceived. By ran - disbursement of resources depend on certain principles domly varying either the features of firms or individuals that influence if (the level of) support payments, such described in two different vignettes, we aspire to causally as subsidized training cost reimbursements, are deemed determine the impact of these features on justice judg- appropriate. It is important for the legitimacy of the ments. Section 5 contains the empirical results. In Sect. 6 welfare state that citizens accept social policy meas- some conclusions are presented. ures and regulations and regard them as just (Rothstein 1998; Roosma et  al. 2013; Sachweh 2016). Major factors 2 Institutional background are principles of justice, which can affect the allocation In 2019, the Qualification Opportunities Act and in according to effort and compensation, (basic) needs, or 2020, the Work for Tomorrow Act (“Arbeit-von-Mor- (social) productive efficiency (Leisering 2004). An “inves - gen-Gesetz”) greatly extended the funding opportuni- tive turn” (Evers 2008) has been documented for more ties for firms undertaking occupational training for their than 10  years. This is a shift toward a more investive employees. Training subsidies for employed workers are social policy, which is also expected to yield social gains granted dependent on firm size (Klaus et  al. 2020). The (Esping-Andersen et al. 2002; Sachweh 2016, p. 309). The program aims to support employees who perform occu- Act is a step in this direction, which has been long and pational activities that can be replaced by modern tech- extensively discussed in the political sphere (German nologies, are otherwise affected by structural change, or Bundestag 2018)—but not among the general public. The plan to work in an occupation with a shortage of skilled legislature wants to support employees so that they can labor. adapt to structural change. The means of this support is Basic job training measures to be funded include— (partially) reimbursing the costs for occupational training among many others—software training courses or job- and the wage costs during training. Consequently, there specific language classes. Funding can range between are two addressees in regard to justice assessment of the partial and complete absorption of job training costs. Act: employees and firms. On the one hand, employ - Funding is conditional on the following four criteria. ers can participate. For some of them, the need may not First, the training has to provide knowledge that exceeds actually be real, which could then cause windfall gains for workplace-related short-term adaptations. Second, the such employers. On the other hand, there are employ- most recently acquired occupational degree must have ees where the need can be more direct. Furthermore, been obtained at least 4 years ago. Third, the job training respondents are able to perceive and evaluate differences measure has to be carried out by an accredited provider, between individuals and companies in terms of justifica - either outside or inside the firm. Fourth, job training tion based on performance or efficiency. Therefore, we must comprise at least 160 h (from 2020 onwards: 120 h), ask our respondents about both addressees to determine with the maximum duration not exceeding 3 years. if they deem the legal regulations for allocating train- Two types of financial support for qualification meas - ing cost reimbursements and the receipt of training cost ures are available for firms under this legislation. First, reimbursements for certain workers—both simulated by the costs for the job training itself can be subsidized. our two vignettes—as just. Employers are required to bear a partial burden of the training costs in a “reasonable manner”, which is speci- fied according to the firm size. Firms pay direct job train - Since 2007, vocational qualifications for low-skilled and older employees in German SMEs have also been subsidized through a special program, which ing costs according to firm size. Small firms with fewer has been used rarely (van den Berg et al. 2018; Klaus et al. 2020). than 10 employees can be compensated for the total costs For the purposes of this paper now simply referred to as “training subsi- of training. According to the Act, firms with 10–249 dies”. Justice perceptions of occupational training subsidies: findings from a factorial survey Page 3 of 18 6 (250–2499) [2500 and more] employees are required to as just. These principles can be aligned with the need contribute at least 50% (75%) [85%] of the costs. Costs of individual or collective actors to a greater or lesser may also be reimbursed completely for low-skilled work- degree (Kluegel et  al. 1999, p. 255; Gilliland 1993). ers participating in retraining, if the job training benefits This is connected to the principle of responsibility for a worker aged 45 or older, or if a worker is severely handi- results that can or cannot be influenced (Konow 1996, capped. Second, the wages of participating workers can 2001; Mikula 2002). However, principles of distribution be (partly) covered by unemployment insurance. Small can also be oriented toward the principle of contribu- firms of less than 10 (10–249) [250 and more employ - tion to a greater or lesser degree (Adams 1965; Young ees] may receive wage support of up to 75% (50%) [25%]. 1993), i.e., according to previously or currently being If the employee to receive job training lacks any voca- performed acts of the individual or collective (Green- tional training, labor costs can be reimbursed completely. berg 1990). For low-qualified employees participating in retraining, Moreover, it has been pointed out that justice assess- the entire wage bill might be covered by unemployment ments need to consider future effects of allocation condi - insurance during training. Training has to address needs tions (Vobruba 1996). This finding is taken into account that go beyond short-term workplace adaptations, such by “productivistic justice” (ibid. 969; cf. also Leisering as necessary adaptations due to exogenous structural 1999, p. 11). First, this means that citizens attempt to change, as smaller companies have (on average) fewer estimate a (socially) effective use of funds and consider resources at their disposal to respond to these challenges. this in their judgments. Second, this may also mean that In summary, the framework established in Germany citizens, in principle, take a favorable position toward applies the following key criteria: substitutability of jobs, an investive social policy (Sachweh et  al. 2009, p. 618), training duration, firm size, employee age, and employee which includes the Qualification Opportunities Act. The qualification. These criteria are reflected in regulations, prevention of unemployment and its associated costs which are primarily directed toward supporting com- may also save money in the mid-term (Hans et al. 2017). panies and employees who are in need (Kluegel et  al. In the following, we discuss, the potential justice 1999, p. 255; Gilliland 1993; Miller 2020), provided the assessments citizens make when determining the just company itself is not responsible for causing this need amount of subsidies for further training of different types (Konow 2001; Mikula 2002, p. 268). of firms . For this case in particular, we expect that citi- These criteria and their accompanying principles of jus - zens are mindful of neediness and a company’s capacity tice are investigated in our factorial survey. The theoreti - to act on its own behalf. It is also possible that a deci- cal background and the hypotheses that frame the survey sion is made on the grounds of effective usage of contri - are presented below. butions to unemployment insurance. Undue profits for companies might arise, as companies receive benefits from unemployment insurance for job-related training 3 Theoretical background and hypotheses that they might have carried out anyway (Kruppe et  al. Welfare state regulations impact individual life conditions 2020, p. 8). and structure social relationships. Citizens are directly Since this information is hidden, other signals, such as affected by the design of measures and regulations: as factors that influence the market position or size—SMEs recipients of transfer payments and as addressees of invest much less in training than large firms (Allaart social services, as well as contributors and taxpayers to et  al. 2009)—have been selected as indicators. A compa- finance the welfare state. It is therefore important for the ny’s capacity to act, the factor of self-responsibility and legitimacy of welfare states that citizens accept measures hereupon-derived neediness, can influence citizens’ jus - and regulations decided on a political level (van Oors- tice perceptions regarding the allocations of funds from chot et al. 2017). The acceptance of welfare programs and unemployment insurance. The Act distinguishes subsidy institutions is based on citizens’ principles about what levels according to company size but does not consider constitutes a just relationship between effort and reward, further indicators of the market position. as well as a certain living standard, which society grants its citizens in return for their contribution to society H1 Citizens provide higher training subsidies for com- (Bowles und Gintis 2000; Kaufmann 1997a; Mau 2004; panies that are economically weaker (H1a) or smaller in Roosma et al. 2013; Sachweh 2016). size (H1b) compared to companies that are economically To maintain a consensus on the condition of the strong or large in size. welfare state or its integrating functions (Kaufmann 1997b), sociopolitical norms and institutionally defined In principle, the legislature wants to provide allocation and dispensation mechanisms ought to help for employees, who are negatively affected by reflect principles of distribution that are perceived 6 Page 4 of 18 R. V. Wolff et al. technologically induced structural change. We expect However, the principle of need may also be addressed, that citizens acknowledge such need. It is presumed as employees with unstable work histories could be more that training subsidies for companies are higher if reliant on proactive, internal further training, due to their employees are affected by structural change with higher generally lesser chances on the labor market, which may probability. reduce or even reverse the effect of H5. In Germany, both unemployed and employed individu- H2 Citizens provide higher training subsidies for com- als can access training subsidies from unemployment panies, if funded occupations are strongly exposed to insurance. On the one hand, also among our respondents, technological progress. the employees themselves pay for social security insur- ance. The vast majority of our respondents and German citizens in general are under this obligation. It is possible Furthermore, we consider the extent to which train- that an expansion of beneficiaries among their own group ing subsidies are perceived as just for different types of is accepted and seen as a more appropriate compensa- employees. This also allows a comparison to the more tion for previously provided individual contributions. or less singular focus on unemployed individuals before On the other hand, unemployed individuals are probably the Act. We expect that for employees, the criteria for seen as needier than employees, not least because unem- the principle of need are also considered. ployed individuals’ benefits are also based on their past payments for unemployment insurance. Last, a focus on H3 Training subsidies for further training have a higher neediness is aligned with a more economical and effec - probability of being regarded as just for individuals who tive usage of limited resources. have characteristics that signal low labor market pros- pects (e. g., advanced age or occupations at risk of being H6 Training subsidies for further training have a higher replaced by technological progress in the future) than for probability of being regarded as just for unemployed individuals who are in a more favorable position on the individuals than for employed workers. labor market. Very long courses that may take two years and go beyond further training, such as vocational retraining, could be 4 Methods and data regarded as inefficient. That is because they are often dis - The empirical findings are based on a factorial survey; connected from current activities and qualifications, where such surveys have been proven suitable to investigate a more targeted and shorter further training would allow wide range of questions, such as social norms (Auspurg distribution of resources toward more recipients. Fund- and Hinz 2015; Rossi and Anderson 1982). We con- ing for very long vocational trainings is not possible under struct several fictitious situations (vignettes) and ask the the guidelines of the Qualification Opportunities Act that respondents to assess these situations. The situations supports companies and employees but is possible under randomly combine different characteristics along several unemployment insurance for unemployed individuals. dimensions. Three major advantages of this approach are Both derive their resources from the same source, so we that (a) respondents have to judge realistic situations; formulate the following hypothesis: (b) with the necessary caution, the causal effects of dif - ferent characteristics on assessments may be identified; H4 Training subsidies for further training have a higher and (c) the approach is relatively robust regarding biased probability of being regarded as just for shorter courses answering behavior, such as social desirability bias. than for courses with very long duration (2 years). We use two different sets of vignettes. The first relates to the firm dimension (H1 to H2), and the second focuses The principle of justice underlying social insurance on the worker dimension (H3 to H6). The vignette design schemes in Germany is the principle of equity, which creates longitudinal data with the occasion dimension for aligns contributions and benefits. From this point of each scenario, similar to panel data with the time dimen- view, those who contributed longer should also be pro- sion. For the two sets of vignettes, the respective vignette vided more benefits from insurance funds. variables are selected uniformly at random from every possible vignette set of the vignette universe and assigned H5 Training subsidies for further training have a higher to each survey participant. The four different vignettes probability of being regarded as just for employees with within each set represent a different occasion to evalu - stable work histories and therefore regular social secu- ate, so the judgment of each scenario is influenced by a rity contributions compared to individuals with irregular multitude of factors, whose intensity can vary between work histories. and within subjects. This is important, as it requires a Justice perceptions of occupational training subsidies: findings from a factorial survey Page 5 of 18 6 Table 1 Dimensions of the firm-related vignettes Dimension Characteristics Number of attributes Economic situation of the company Economically strong 2 Economically weak Company size 30 4 30,000 Number of funded employees 2 2 Degree of potential job automation in current occupation for funded Not mentioned 3 employees Already 25% of activities replaceable Already 75% of activities replaceable Duration of funding 4 weeks 2 6 months The vignette universe consists of 96 (2 × 4 × 2 × 3 × 2) combinations total, all of which are plausible (full factorial design) multilevel approach and will be addressed at the end of It can happen that the situations only differ slightly from this section. each other. For these cases, too, your judgment is impor- We briefly introduce the topic and later use different tant for us. It is not about “right” or “wrong”, we are inter- outcome variables for firm-related and worker-related ested in your assessment.” vignettes. According to the Act, firms receive subsidies Table  1 provides an overview of the dimensions of the depending on their characteristics. Therefore, we ask r fi m-related vignettes. These are the economic situation the respondents how high a percentage of the subsidies of the company, company size, number of employees tak- should be for the specific cases given in the vignette sce - ing part in training, possible degree of job automation, narios regarding companies. However, specific groups and training duration. of employees are also beneficiaries. The legislature aims The firm-related scenarios (bold letters for variable to support workers who are strongly affected by, for parameters) are presented in this text: instance, structural change and digitization. Conse- “An economically strong (weak) company with 30 quently, we also ask respondents to evaluate whether (300; 3000; 30,000) employees applies at the employ- they deem it just that a certain person be considered for ment agency for support for the occupational training funding of further training. of two (15) employees. The training lasts for 1  month First, we begin our survey with a general introduction (6 months). These employees work in professions in as presented below to make all respondents familiar with which 75% (25%; sentence not displayed) of activities the topic (German version in Appendix: Text A1): can already be replaced by computers and computer-con- “Employers and employees who are subject to social trolled machines.” security contributions are required to pay contribu- After each of the four different scenarios, respondents tions to unemployment insurance. These funds can also were asked to indicate which percentage of wages dur- be used to pay for further training in companies so that ing training and training costs should be reimbursed by employees can better adapt to new challenges at the unemployment insurance as a subsidy. Answers were workplace.” provided in 10% steps, ranging from zero to 100%. Second, we introduce the topic of the firm vignette Third, we introduced the topic of the worker vignette (underlining included) and then ask respondents to make (underlining included) and then asked respondents to a judgment about the preferred percentage of training judge how just or unjust they deem training subsidies subsidies (German version in Appendix: Text A2): (German version in Appendix: Text A3): “In the following, 4 different situations are described, in “In the following, 4 different situations are described, in which companies apply for funding of further training at which unemployed or employed people come up wanting the employment agency. further training. Please decide how much the company should receive in Please decide how just or unjust you find it, that the a certain situation for further training of its employees. respective person receives financing for further training 6 Page 6 of 18 R. V. Wolff et al. Table 2 Dimensions of the worker-related vignettes Having finished our depiction of vignettes, let us now turn to the description of our pool of respondents. The Dimension Characteristics Number of survey sample of approximately 35,000 people was sam- attributes pled uniformly at random from the Integrated Employ- Sex Male 2 ment Biographies (IEB V13.01.00-181010). The IEB Female covers all registered spells of employment subject to Age in years 34 3 social security contributions (including marginal employ- ment), unemployment, unemployment benefit receipt, job search and participation in active labor market pro- Job status Unemployed 2 grams in Germany. The sample was restricted to citizens Employed 18 years old or older at the time of data collection and to Job risk of automation Activities not replaceable in future 2 individuals of German nationality (Osiander et al. 2020). Already 75% of activities Data access further required individuals to have had replaceable an IEB spell during 2017 and at least one employment Job contribution to Continuous 2 spell during the period 2013–2017 to be included in the social insurance Intermittent and partial sample. Job training duration 4 weeks 3 Using this sample, between 11/2019 and 1/2020 24,934 6 months people were contacted via e-mail and 9551 people via 2 years post and asked to take part in an online survey. These The vignette universe consists of 144 (2 × 3 × 2 × 2 × 2 × 3) combinations total, invitations included information about the research pro- all of which are plausible (full factorial design) ject and data protection issues; a reference to the project homepage offered additional information. The e-mails from unemployment insurance funds. Likewise, it can contained an individualized link to the survey, while the happen that the situations only differ slightly from each letter included a QR code as well as a short link along other. Again, it is not about “right” or “wrong”, we are with an individual password. Almost 50% of participants interested in your assessment.” chose to answer the survey on their smartphones or tab- Table  2 provides an overview of the second type of lets, while the other half chose to use a laptop or desktop vignette, where respondents assessed whether funding PC. Overall, 1712 individuals started the survey, and after for a described individual worker was just. The dimen - accounting for missing answers to the questions, includ- sions analyzed are gender, age, risk of job loss due to ing refusal to merge their answers with the administrative automation, previous contributions to unemployment records on their labor market biographies in the IEB, a insurance, and the duration of training. balanced panel with 1010 persons remained. Worker-related scenarios (bold letters for variable In consideration of established guidelines (AAPOR parameters) are presented in this text: 2016), we calculate the net response ratios conserva- “A 34 (46; 58)-year-old employed (unemployed) tively. Approximately 3.8% of people who were contacted man (woman) works in a profession in which 75% of (2.7% for invitations via e-mail and 6.7% for invitations activities can be replaced by computers or computer- via post) on valid addresses fully answered the survey. controlled machines (which cannot be replaced by This is in line with expectations for such contact chan - computers in the future). After finishing vocational nels. The sample of our participants is not representa - training, he (she) was continuously employed and con- tive of the German labor force. However, we were able to tributed (intermittently employed and partially con- conduct selectivity analyses for the combined samples of tributed) to unemployment insurance. For this person, e-mail and post. Almost 50% of respondents in the gross unemployment insurance finances occupational training sample are female, 20% live in Eastern Germany (includ- with a duration of 4 weeks (6 months; 2 years).” ing Berlin), 75% have completed vocational training or Respondents were asked how just, from their point of university and 60% are qualified skilled workers. Our view, it is that the employment agency pays for this occu- selectivity analysis compares the gross sample to the final pational training, using funds from the unemployment sample of respondents. People from Eastern Germany insurance? (including Berlin) and people who are 65 years and above are underrepresented, while people 50–64 are overrep- Strictly analytical something can only be either just or unjust and nothing resented. Moreover, people with higher formal qualifica - in between. However, in empirical reality and in colloquial language this dis- tions, longer duration of both employment and previous tinction is not so clear-cut, so respondents may still want to select a more nuanced position. Consequently, we allowed four different outcomes (ranging unemployment receipt are also overrepresented in our from unjust and quite unjust to quite just and just). Afterward, the two sides final sample (Osiander et al. 2020). were combined into the dichotomy of unjust/just. Justice perceptions of occupational training subsidies: findings from a factorial survey Page 7 of 18 6 All participants received two sets of four vignette sce- first vignette can also benefit from this. We know that narios. Furthermore, we collect information on: gender, mixed effects with random intercept and random slopes age, qualification, political preferences and classified net can be important because the cluster-robust variance– monthly household income. We also include questions covariance estimator (Eicker–Huber–White procedure) about (a) the respondent’s attitude toward unemployed accounts only for between-person heteroscedasticity but individuals and (b) how respondents handle new tech- not for variance differences originating within clusters. In nologies at the workplace. The respondent characteris - our case, only the vignettes vary within the cluster that tics can control for selective distortions in the sample, each participant represents. such as age, income, political views and for behavior For the firm vignette, the linear mixed regression patterns, such as self-interested behavior. Depending on model includes random slopes for the attributes “com- the circumstances, someone is primarily a beneficiary or pany size”, “degree of job automation” and “duration of a payer in the unemployment insurance system. Factors funding” (cf. Table 1) to account for within-person stand- such as regional work opportunities, position in the labor ard deviation differences with random slopes. To moti - market and technical knowledge can play a role, and peo- vate the mixed model, we check three criteria: relevant ple may choose the option they believe to benefit from effects and noteworthy p-values of the correlation coeffi - the most. A summary of all variables and their operation- cients, the results of a likelihood ratio test, and improved alization is available in Appendix Table 5. information criteria (AIC/BIC). We model both the Each participant judged four vignettes related to the standard deviation for random slopes and the correlation company level and four vignettes related to the employee between random slopes, including the correlation of ran- level. These four evaluations are very likely not inde - dom slopes with the random intercept. We choose con- pendent of each other, given that they are from the same servative unstructured standard deviations because we person. cannot assume a certain standard deviation pattern. First, For both the firm vignette (size of training subsidies we confirm that all standard deviation coefficients are perceived as just) and the worker vignette (training sub- relevant in size and statistically significant (p < 0.001). For sidy is perceived as just), this is accounted for by estimat- the correlation terms, all selected random slope variables ing multilevel models such as fixed effects models, but are relevant in size and statistically significant (p < 0.001) also mixed effects models using random intercepts and as they correlate with the constant, so these variables are slopes at the individual level. The fixed aspect means that all regarded as important for inclusion in the model. Sec- only the within-variation of the vignette variables enters ond, a likelihood ratio test (p < 0.001) corroborates this the model, while mixed models can fit all coefficients. In selection of random slopes when compared to the same our mixed model, both vignette variables and individual model but without random slopes. Third, the chosen ran - characteristic variables, plus a selection of random slopes dom slopes collectively improve both information crite- (also called random coefficients), are fitted at the indi - ria, AIC and BIC (cf. Table  3), to avoid overfitting. Due vidual level. In general, mixed models can deliver more to extensive testing, we can confirm that no other possi - reliable results than standard linear panel regression with ble combination of random slopes increases the model fit random effects. Therefore, we account for the multilevel further. Please take note that the pseudo-R of Snijders/ structure in the first vignette with a linear mixed model. Bosker does not take into account random slopes, which Mixed models combine the advantages of both fixed is why its results in Model 2 and Model 3 are identical. effects and random effects (Bell et al. 2019) but are there - This measure is used here to point out a general increase fore more complex. Mixed effects with random effects in explained variance between Model 1b and Model 2 at coefficients for only intercept and residual are identi - level 1, which is our vignette level, and at level 2, which cal to random effects models. Additional slopes can be is the level of individual characteristics (Snijders and regarded as an extension of the random effects model and Bosker 1999). In summary, the selected linear mixed are the hallmark of mixed models. Analogous to the gen- effects panel model has advantages over a standard ran - eral assumption that the random intercept model follows dom effects model, as it further reduces heteroscedastic - a normal distribution, random slopes follow a multivari- ity (Bell et al. 2019). ate normal distribution. Such random slopes can capture For the worker vignette with binary coding of just or and model even more variance than the mere inclusion of unjust, a logit mixed model with random slopes deliv- a random intercept can. Therefore, unless parallel slopes ers different and more reliable results than a standard can be assumed, mixed models should be considered a panel logit with random effects could. Once again, we first choice before fitting a simple random effects model. choose to address the multilevel structure by apply- This proved particularly important for the logit models ing mixed effects, now with a logit mixed model. We used for the second vignette, but the linear models of the fit conditional logit fixed and mixed logit models by 6 Page 8 of 18 R. V. Wolff et al. Table 3 Firm vignette: size of training subsidies perceived as just (in 10% intervals) Model 1a Model 1b Model 2 Model 3 Fixed effects coefficients 1. Vignette features FE Mixed + RI Mixed + RI Mixed + RI/RS Company size (ref: 30) 300 employees − 4.163*** (0.897) − 4.028*** (0.887) − 3.993*** (0.888) − 3.644*** (0.853) 3000 employees − 9.782*** (0.956) − 9.567*** (0.937) − 9.563*** (0.937) − 9.689*** (0.885) 30,000 employees − 15.144*** (0.973) − 15.016*** (0.957) − 15.019*** (0.956) − 14.484*** (0.938) Strong company (else: weak) − 16.924*** (0.703) − 16.869*** (0.697) − 16.929*** (0.696) − 16.766*** (0.683) Training for 15 people (else: two) − 0.928 (0.668) − 0.791 (0.659) − 0.735 (0.660) − 0.934 (0.632) Job at risk of automation (ref: nothing) Already 25% replaceable − 0.569 (0.914) − 0.433 (0.881) − 0.401 (0.879) 0.092 (0.857) Already 75% replaceable − 2.588** (0.948) − 2.551** (0.924) − 2.562** (0.922) − 1.857* (0.893) Funding for 6 months (else: four weeks) − 0.792 (0.724) − 0.679 (0.690) − 0.609 (0.688) − 0.911 (0.619) 2. Personal characteristics Male (else: female) − 6.214*** (1.560) − 5.958*** (1.543) Age of respondent (ref: 50–64) Age 18–34 5.439* (2.030) 5.559* (2.002) Age 35–49 1.222 (1.793) 1.523 (1.774) Age 65–78 − 0.249 (4.030) 0.124 (4.027) German region (ref: ‘Eastern States ‘) German Southern States − 2.720 (2.389) − 3.288 (2.393) German Northern States − 4.793 (2.416) − 5.635* (2.425) German City States 0.448 (3.279) − 0.854 (3.257) Education (ref: A-levels/vocat. training) No education 6.380 (6.896) 5.986 (6.567) ( Technical) College − 3.657* (1.576) − 3.719* (1.561) Net hh-income/month (ref: 2000–2999) Less than 1000 Euro − 0.993 (4.460) − 0.932 (4.364) 1000–1499 Euro 1.018 (3.347) 0.514 (3.248) 1500–1999 Euro 6.178* (2.919) 5.475 (2.944) 3000–3999 Euro 1.023 (2.358) 0.183 (2.345) 4000–4999 Euro − 1.496 (2.615) − 1.830 (2.605) 5000 Euro or more 2.638 (2.712) 2.261 (2.681) Income not specified − 1.390 (3.153) − 2.239 (3.139) Party affiliation (ref: ‘Greens’) CDU (Christian Democratic Union) − 1.834 (2.836) − 0.760 (2.813) CSU (Christian Social Union) − 6.275 (4.633) − 5.731 (4.682) SPD (Social Democratic Party) − 0.005 (2.594) 0.251 (2.593) AfD (Alternative for Germany) − 0.812 (4.767) − 0.416 (4.758) FDP (Free Democratic Party) − 9.216* (3.373) − 8.543* (3.322) Die Linke ( The Left) − 1.059 (3.051) − 1.901 (2.994) Other party (not in parliament) 2.353 (4.574) 2.289 (4.620) Not specified 2.027 (2.029) 2.264 (2.002) Unemployment benefit (else: no) 0.921 (1.708) 0.734 (1.697) Unemployed responsibility (else: no) 1.890 (2.545) 2.412 (2.520) Difficulty with new work tech (ref: no) Difficulty with new tech 3.969 (2.489) 3.861 (2.436) Not applicable (no work) 0.264 (2.170) 0.325 (2.166) Constant 62.778*** (1.262) 66.889*** (3.693) 67.326*** (3.681) Justice perceptions of occupational training subsidies: findings from a factorial survey Page 9 of 18 6 Table 3 (continued) Model 1a Model 1b Model 2 Model 3 Random effects (RE) coefficients Sd (constant) 22.670*** (0.524) 21.774*** (0.509) 26.984*** (0.776) Sd (residual) 17.873*** (0.327) 17.873*** (0.327) 12.703*** (0.502) Sd (strong company) 14.899*** (0.937) Sd (25% replaceable) 13.116*** (1.503) Sd (75% replaceable) 13.170*** (1.487) Sd (subsidize 15) 10.051*** (1.416) Corr (strong company, 25% replaceable) 0.179 (0.108) Corr (strong company, 75% replaceable) 0.014 (0.101) Corr (strong company, subsidize 15) 0.122 (0.092) Corr (strong company, constant) − 0.431*** (0.047) Corr (25% replaceable, 75% replaceable) 0.189 (0.145) Corr (25% replaceable, subsidize 15) − 0.086 (0.124) Corr (25% replaceable, constant) − 0.263*** (0.070) Corr (75% replaceable, subsidize 15) 0.253 (0.143) Corr (75% replaceable, constant) − 0.224*** (0.067) Corr (subsidize 15, constant) − 0.390*** (0.061) Model fit criteria Information criteria: (1) AIC; (2) BIC (1) 1040; (2) 1083 (1) 36,810; (2) 36,879 (1) 36,796; (2) 37,042 (1) 36,545; (2) 36,879 R : McFadden; Lvl. 1&2 Snijders/Bosker 0.203 (MF) 0.106; 0.013 (S/B) 0.148; 0.080 (S/B) 0.148; 0.080 (S/B) 4040 vignette answers for 1010 persons FE fixed effects, RI random intercept, RS random slopes *p < 0.050; **p < 0.005; ***p < 0.001; robust standard errors in parentheses maximizing the log pseudolikelihood (cf. Appendix Table  2) to model within-person variance differences Table  6) and in postestimation predict robust aver- with random slopes. We do this because of three fac- age marginal effects (cf. Table  4). Table  6 indicates tors: the relevance of the variance coefficients and the both secular improvement with a decreasing AIC and noteworthy p-values for the covariance coefficients, an increase in McKelvey–Zavoinas’ pseudo-R from the results of a likelihood ratio test, and the improved approximately 20% to approximately 80% as a measure information criteria. We begin by modeling the vari- of variation explained by the model. This pseudo-R ances for random slopes and the covariances between is also recommended as most suitable for logit mod- random slopes, including the covariances of random els (Langer 2017). In Model 1a, we include only the slopes with the random intercept. We choose con- vignette variables for a logit conditional fixed effect servative unstructured covariances because we cannot for comparison with the following mixed models. In assume a certain covariance pattern. First, we notice Model 1b, the same vignette variables are used to fit that the results for the chosen random slopes show individual coefficient variances as random intercepts. considerable values for the individual variance coef- In Model 2, we then add personal characteristics. ficients. The covariance coefficients also have note- For reference, the results of Model 1b and Model 2 worthy effects, with (almost all) p-values smaller than are identical to those fitting a standard logit random 0.05, and are therefore regarded as relevant for inclu- effects regression, which does not consider slopes. In sion in our model. Second, further validation of the the final Model 3, we add individual random slopes to chosen random slope model is given by the results of model the variance structure. For this, we choose to a likelihood ratio test (p < 0.001) with selected random model the attributes “job status”, “job risk of automa- slopes compared to a model without random slopes. tion” and “job contribution to social insurance” (cf. Third, all chosen random slopes collectively improve both information criteria, AIC and BIC (cf. Table  6). This means that the model is not overfitted. No other Due to convention, variance and covariance is used for random coefficients here, but after applying a transformation to normalize the estimate, which possible combination of random slopes improves the does not affect p-values, the results would be analogous to standard deviation model fit further, which reassures us that it is not and correlation of the linear mixed model. 6 Page 10 of 18 R. V. Wolff et al. Table 4 Worker vignette: training subsidy is perceived as just—Avg. marginal effect Avg. predicted marginal effect Model 1a Model 1b Model 2 Model 3 1. Vignette features Logit fixed Logit mixed + RI logit mixed + RI logit mixed + RI/RS Male (else: female) − 0.058* (0.027) − 0.025** (0.011) − 0.027* (0.011) − 0.025* (0.010) Age (ref: 34 years old) 46 years old 0.074* (0.031) 0.040*** (0.012) 0.042*** (0.012) 0.035** (0.012) 58 years old − 0.034 (0.032) − 0.020 (0.014) − 0.021 (0.014) − 0.028* (0.013) Unemployed (else: no) 0.140*** (0.027) 0.078*** (0.012) 0.081*** (0.012) 0.086*** (0.011) Funding time (ref: 4 weeks) Funding for 6 months − 0.034 (0.034) − 0.027* (0.013) − 0.026* (0.013) − 0.027* (0.013) Funding for 2 years − 0.193*** (0.032) − 0.092*** (0.013) − 0.090*** (0.013) − 0.091*** (0.012) Job 75% replaceable (else: no) 0.150*** (0.029) 0.069*** (0.013) 0.067*** (0.013) 0.070*** (0.013) Continuous job (else: no) 0.267*** (0.026) 0.119*** (0.012) 0.122*** (0.012) 0.117*** (0.012) 2. Personal characteristics Male (else: female) 0.039* (0.016) 0.042* (0.016) Age (ref: 50–64) Age 18–34 0.015 (0.020) 0.027 (0.019) Age 35–49 0.011 (0.018) 0.010 (0.017) Age 65–78 − 0.030 (0.046) − 0.041 (0.046) Region (ref: ‘Eastern States’) German Southern States − 0.005 (0.024) − 0.005 (0.024) German Northern States − 0.005 (0.024) − 0.007 (0.023) German City States 0.026 (0.032) 0.017 (0.032) Education (ref: A-levels and vocational training) No education − 0.049 (0.068) − 0.068 (0.068) ( Technical) College 0.023 (0.016) 0.024 (0.016) Monthly net household Euro income (ref: 2000–3000) Less than 1000 Euro 0.002 (0.042) 0.009 (0.040) 1000–1499 Euro 0.034 (0.034) 0.030 (0.034) 1500–1999 Euro 0.016 (0.026) 0.020 (0.025) 3000–3999 Euro − 0.019 (0.022) − 0.016 (0.022) 4000–4999 Euro − 0.032 (0.025) − 0.024 (0.024) 5000 Euro or more − 0.074* (0.027) − 0.060* (0.026) Income not specified − 0.031 (0.034) − 0.028 (0.033) Parties (ref: ‘Greens’) CDU (Christian Democratic Union) − 0.060* (0.031) − 0.065* (0.031) CSU (Christian Social Union) − 0.050 (0.049) − 0.055 (0.048) SPD (Social Democratic Party) 0.031 (0.025) 0.020 (0.025) AfD (Alternative for Germany) − 0.045 (0.049) − 0.066 (0.050) FDP (Free Democratic Party) − 0.135*** (0.042) − 0.132*** (0.039) Die Linke ( The Left) 0.012 (0.031) − 0.004 (0.032) Other party (not in Parliament) 0.085* (0.031) 0.071* (0.031) Not specified − 0.004 (0.019) − 0.01 (0.019) Unemployment benefit (else: no) 0.069** (0.022) 0.075*** (0.020) Unemployed responsible for situation (else: 0.035* (0.016) 0.042* (0.016) no) Difficulty with new work technologies (ref: no) Difficulty with new tech 0.023 (0.023) 0.018 (0.022) Not applicable (no work) 0.025 (0.022) 0.027 (0.021) 4040 vignette answers; 1010 persons RI random intercept, RS random slopes *p < 0.050; **p < 0.005; ***p < 0.001; robust standard errors in parentheses Justice perceptions of occupational training subsidies: findings from a factorial survey Page 11 of 18 6 30% 26% 25% 20% 15% 15% 12% 10% 9% 10% 8% 7% 4% 4% 4% 5% 2% 0% Funding in percent Notes: 4,040 vignettes, 1,010 persons. Fig. 1 Distribution of funding percentages to subsidize firm costs of further training. necessary to include additional random slopes, as dif- (Wasserstein et  al. 2019) of the effects of firm charac - ferences in variances can already be captured by the teristics on justice assessments. Table  3 presents the chosen final model. This is also supported by dou- results from Model 1a with fixed effects and Model 1b bling the model fit criterion of pseudo-R (FE + RE) with mixed effects and random intercept. Both cover from approximately 40% to approximately 80% when only the vignette dimensions. Model 2 also controls for comparing Model 2 to Model 3 (cf. Table 6). respondent characteristics. Finally, Model 3 is the same In conclusion, a standard panel logit model with ran- as Model 2 but with random slopes. The following dis - dom effects would not be appropriate here. Such a model cussion refers to Model 3. Note that we focus our inter- would seriously violate the model assumptions of homo- pretation on point estimates (random coefficients are not scedasticity and deliver anti-conservative results that to be interpreted for this purpose)—rounded to the next introduce a bias on standard errors and point estimates ½ % or half percentage point—for coefficients with low (Bell et  al. 2019), so we prefer the final mixed model p-values of < 5%, < 0.5% and < 0.1%, as indicated by one, with the aforementioned random slopes to mitigate this two or three asterisks; standard errors can be found in problem. the respective tables to reflect this. The p-values with one asterisk between < 5% and 0.5% should be seen as merely 5 Empirical findings suggestive, while those < 0.5% may be considered signifi - In the first step, we analyze the extent to which features cant, and those < 0.1% may be considered even more so of the firm have an impact on the perceived just amount (Benjamin et al. 2018). of training subsidies. The average amount of funding First, survey respondents grant substantially higher across all vignettes is approximately 45%. As outlined subsidies to economically weak firms than to economi - above, the design of our study ultimately does not allow cally strong firms (≈ + 17 percentage points). Further- us to draw conclusions regarding the general support of more, the subsidy share increases with firm size—each training subsidies in the population. A descriptive analy- tenfold increase in the number of employees reduces sis shows that in 85% of the scenarios, the respondents the assigned subsidy (≈ − 5.0 percentage points each would financially support training subsidies at least step), particularly for very large firms (≈ − 14.5 percent- to some degree (see Fig.  1). Due to our factorial survey age points). As predicted by H1a and H1b, participants design, we can cautiously draw a causal interpretation Percent of answers per category 6 Page 12 of 18 R. V. Wolff et al. slightly more likely to be perceived as a just recipient of thus account for the basic justice principle of need when funding for further training (≈ + 3.5%), while this is less assigning the just amount of training subsidies. How- likely for an older individual of age 58 (≈ − 3%). It is plau- ever, the degree of differentiation according to firm size sible that younger workers, due to their usually more is smaller than that granted in current legislation in up-to-date training and long-term prospects for amorti- Germany. zation of their own investment in training, are perceived We find, however, no support for H2 that individuals as less needy. For older persons, respondents may assume assign more support to firms where trained workers are that they are close to retirement and that investments in employed in occupations threatened by technological training will thus not pay off. Furthermore, the respond - change. Survey participants assign even less funding if ents might presume that older learners have diminished the firm trains workers whose jobs are strongly exposed learning capabilities. Concerning age, H3 is only partly to potential job automation (≈ − 2 percentage points) confirmed, as respondents also seem to take the principle compared to the reference scenario where no information of efficiency into account. on the risk of automation was given. It is possible that the With respect to training duration, respondents per- participants consider such work sites a “lost cause” and ceive funding for training over six months (≈ − 3%) or would rather not waste resources on the company itself. two years (≈ − 9%) as less just than a shorter training of Regarding respondents characteristics, men gener- 4  weeks. While 4  weeks are already slightly preferred to ally attribute less funding to the firm (≈ − 6 percentage 6  months, the difference is quite small and not relevant points), adults younger than 35 attribute more fund- for the analysis, which focuses on long courses. Accord- ing compared to the reference group of 50–64-year-old ing to H4, such long courses will be deemed too long. respondents (≈ + 5.5 percentage points) and respond- Our results are in line with H4, so efficiency considera - ents from German “Northern States” give less (≈ − 5.5 tions may also play a role in justness assessments of train- percentage points) than respondents from the East. The ing subsidies. effect of monthly net classified household income is not H5 presumed that training subsidies for workers with relevant, while college education has a small negative continuous contributions to social insurance would more effect on the level of support (≈ − 3.5 percentage points). likely be regarded as just than subsidies to workers with Furthermore, individuals who identify themselves as vot- unstable work histories. Here, we indeed find a compara - ers of the Free Democratic Party (FDP) grant much less tively strong effect (≈ + 12%). This supports H5, which is support to firms (≈ − 8.5 percentage points), in line with based on the principle of equity, while needs-based jus- the small government approach of the party. tice may only play a minor role here. In the second step, we analyze the importance of the Finally, an important feature of training subsidies is characteristics of trained workers, who also stand to gain whether the support is directed to unemployed indi- from the Act. As beneficiaries, they are thus included in viduals or employed workers. The estimates show that the assessment of training subsidies in another vignette training subsidies for the occupational training of unem- design. The dependent variable is a binary variable indi - ployed individuals are regarded as more just (≈ + 8.5%). cating whether funding is perceived as just or unjust. According to H6 we expected that training subsidies for Approximately 20% of singular vignette answers regarded further training are more likely to be perceived as just for funding as unjust and 80% as just. In the following, we unemployed individuals than for the employed. While will use the average marginal effects of Table  4 Model 3 the respondents may consider principles of both need for further interpretation. and equity, need seems to be the dominant principle H3 posited that training subsidies are more likely to be in this context. The results provide support for H6, as perceived as just for individuals in need, in particular for unemployed individuals are clearly preferred. Neverthe- persons whose occupations are exposed to automation less, training subsidies for employees are also judged to and older workers. Indeed, the respondents judge train- be fair by the vast majority of vignette answers. This is in ing subsidies to be just more often (≈ + 7%) if the recipi- line with the legislation, as employees also have access to ent of the subsidy has been working in an occupation training paid for by the same fund that thus far has prior- where tasks could already be substituted to a high degree itized unemployed individuals. (75%) by computers. This result is in line with H3. Regarding age, the results show that compared to a To test what determines dominant principles further, interaction effects 34-year-old person, a middle-aged person (age 46) is were calculated, but they provided no further insights. Multiple tests for inter- action effects between vignettes and certain personal characteristics related to The Act grants 25 percentage points less as firm size increases from 30 the vignettes (e. g., age of respondent and age in vignette; sex of respondent to 300 and another 10 percentage points less for training costs as firm size and sex in vignette) were carried out, but none: had strong effects, improved increases from 300 to 3000—compared to our 5.0 percentage points each step. information criteria or were statistically significant/relevant enough to war - rant inclusion in the model. Justice perceptions of occupational training subsidies: findings from a factorial survey Page 13 of 18 6 Regarding respondent characteristics, we find a few of equity but also indications that the respondents appre- effects on justice considerations. Men (≈ + 4%) are ciate an economical use of funds. Public training support more inclined to show support. Respondents with a for workers is more often assessed as just, if the work- high net household income are more reluctant to con- ers are currently unemployed or if their occupation is sider assistance to be just (≈ − 6.0%). For political lean- strongly exposed to potential automation, in line with the ings, strong effects were found for a political preference principle of need. Furthermore, workers are preferred to for the Christian Democratic Union (CDU ≈ − 6.5%), for exhibit stable work histories and thus steady contribu- the Free Democratic Party (FDP ≈ − 13%) and for other tions to unemployment insurance, in line with the princi- parties not currently in parliament (≈ + 7%) when com- ple of equity. The productive use of resources is specified pared to the reference of supporting the Green party. by favoring people of middle age and lack of support for Respondents who ever received unemployment benefits long training durations, in line with economical use of through unemployment insurance show greater sup- funds. port (≈ + 7.5%). Finally, if the respondent agrees with the Fourth, when comparing both dimensions, exposure to statement in respondent characteristics that unemployed technological change is only statistically significant for individuals themselves are responsible for their situation, the worker vignette, but not for the firm vignette. While then this slightly increases the acceptance (≈ + 4%) to both vignettes are different in other ways as well, it can provide training subsidies in the worker vignette. be speculated that supporting workers with training sub- sidies is more tenable. That may be, because workers can directly and more easily benefit by switching to a more 6 Discussion of key findings suitable line of work and even industry—something firms Against the background of new funding options as part cannot easily do. of unemployment insurance regarding further training In light of principles of justice to evaluate the new for currently employed workers in Germany, this article funding possibilities of unemployment insurance, our analyzes the determinants of justice perceptions of such results show that respondents largely exhibit congruence support measures. The state provides the opportunities with neediness and thus not with the principle of equity. for firms to apply for a (partial) reimbursement of wage However, productivistic justice principles and orienta- costs during training, as well as training costs. The analy - tion toward the efficient application of social insurance sis shows that citizens generally accept the training subsi- resources can also come to the fore dies for occupational further training in the Qualification Opportunities Act. Our factorial surveys also uncover the principles of justice underlying the assessment of training subsidies for firms and for workers. Appendix First, an important technical point is that the mixed effects models highlight the need to incorporate (appro - Text A1: Original German version of general priate) random slopes, which can yield a consider- introduction able boost in explained variance to strengthen the study Sozialversicherungspflichtig Beschäftigte und Arbeitgeber results. Future theoretical research could determine, zahlen in Deutschland Beiträge zur Arbeitslosenversicherung. if the size of this boost is dependent on the number of Von diesem Geld können auch Beschäftigte in vignette variables, vignettes per respondent or other Unternehmen Weiterbildungskurse bezahlt bekommen, factors. damit Mitarbeiterinnen und Mitarbeiter sich besser Second, focusing on the firm dimension, we show that an neue Herausforderungen am Arbeitsplatz anpassen that the approach of German legislation to reimburse a können. larger share of training costs for small firms is mirrored by the assessments of training subsidies by the respond- ents. However, respondents differentiate their assess - Text A2: Original German version of the firm ments by firm size less than the recent legislation in vignette and its introduction Germany does. Respondents would grant more support Im Folgenden werden vier verschiedene Situationen to economically weak and small firms, which are more in beschrieben, in denen Unternehmen eine Förderung von need than stronger and larger firms are. We find, how - Weiterbildungen bei der Arbeitsagentur beantragen. ever, no indication that funding should increase with Bitte entscheiden Sie, wie viel das Unternehmen in training duration and with an increased degree of job einer bestimmten Situation für die Weiterbildung seiner automation to which funded workers are exposed. Beschäftigten bekommen sollte. Es kann vorkommen, Third, focusing on the worker dimension, we find sup - dass sich die Situationen nur geringfügig voneinander port for the basic justice principle of need and principle unterscheiden. Auch in diesen Fällen ist uns Ihr Urteil 6 Page 14 of 18 R. V. Wolff et al. darüber wichtig. Es geht nicht um “richtig” oder “falsch”, aus Mitteln der Arbeitslosenversicherung finanziert wir sind an Ihrer Einschätzung interessiert. bekommt. Es kann auch hier vorkommen, dass sich die Situ- ationen nur geringfügig voneinander unterscheiden. Es Firm vignette geht wieder nicht um “richtig” oder “falsch”, wir sind an Ein wirtschaftlich solides (schwaches) Unternehmen Ihrer Einschätzung interessiert. mit 30 Beschäftigten (300; 3.000; 30.000 Beschäft- igten) beantragt bei der Arbeitsagentur die Förderung einer Weiterbildungsmaßnahme für zwei Beschäftigte Worker vignette (15 Beschäftigte). Die Weiterbildung soll einen Monat Ein/e 34-jährige/r (46-jährige/r;58-jährige/r) erwerbstätige/r (6  Monate) dauern. Diese Beschäftigten arbeiten in (arbeitslose/r) Mann (Frau) arbeitet in einem Beruf, dessen Berufen, in denen bereits heute 75 Prozent (25 Prozent; Inhalte schon heute zu 75% durch Computer oder Roboter Satz nicht erwähnt) ihrer Tätigkeiten durch Computer ersetzbar sind (der auch zukünftig nicht durch Computer oder computergesteuerte Maschinen ersetzbar sind. ersetzt wird). Er (Sie) war seit der Ausbildung dauerhaft Wieviel Prozent der Kosten für den Lohn und die beschäftigt und hat Beiträge (unregelmäßig beschäftigt und Weiterbildung des Mitarbeiters soll die Arbeitsagentur hat phasenweise Beiträge) in die Arbeitslosenversicherung dem Unternehmen erstatten? eingezahlt. Die Person bekommt eine vierwöchige (sechs- monatige, zweijährige) Weiterbildung aus Mitteln der Arbe- itslosenversicherung finanziert. Text A3: Original German version of the worker Wie gerecht ist es aus Ihrer Sicht, dass die Arbeitsagen- vignette and its introduction tur die Weiterbildung aus Mitteln der Arbeitslosenversi- Im Folgenden werden vier andere Situationen beschrie- cherung bezahlt? ben, in denen Arbeitslose oder Beschäftigte vorkommen, See Tables 5, 6. die sich weiterbilden möchten. Bitte entscheiden Sie, wie gerecht oder ungerecht Sie es finden, dass die jeweilige Person eine Weiterbildung Table 5 Sample composition for respondent characteristics Variables for all items in a category add up to 100% (excl. rounding error) Percent (%) Male 53.9 Female 46.1 Age 18–34 (18 is minimum) 22.3 Age 35–49 28.7 Age 50–64 44.1 Age 65 + (78 is the realized survey maximum) 5.0 ‘Eastern States’: Brandenburg, Mecklenburg-West Pomerania, Saxony, Saxony-Anhalt, Thuringia 15.2 ‘Southern States’: Baden-Württemberg, Bavaria, Rhineland-Palatinate, Saarland 38.7 ‘Northern States’: Hesse, North Rhine-Westphalia, Schleswig–Holstein, Lower Saxony 36.8 ‘City States’: Berlin, Bremen, Hamburg 9.2 No education 1.8 A-levels and/or vocational training 54.8 ( Technical) College 43.5 Monthly net household income of less than 1000 Euro 4.4 Monthly net household income between 1000 and 1499 Euro 6.6 Monthly net household income between 1500 and 1999 Euro 9.2 Monthly net household income between 2000 and 2999 Euro 22.2 Monthly net household income between 3000 and 3999 Euro 20.7 Monthly net household income between 4000 and 4999 Euro 15.0 Monthly net household income more than 5000 euro 15.1 Monthly net household income not specified 6.7 CDU (Christian Democratic Union) 10.6 CSU (Christian Social Union) 3.2 Justice perceptions of occupational training subsidies: findings from a factorial survey Page 15 of 18 6 Table 5 (continued) Variables for all items in a category add up to 100% (excl. rounding error) Percent (%) SPD (Social Democratic Party) 9.7 AfD (Alternative for Germany) 3.6 FDP (Free Democratic Party) 4.9 Die Linke ( The Left) 7.7 Bündnis 90/Die Grünen (‘Greens’) 25.4 Other party not represented in parliament 3.8 Political factor not specified (no party affiliation, apolitical, no political affiliation) 31.2 Ever in life registered for unemployment benefits 61.8 Not ever in life registered for unemployment benefits 38.2 Agree with statement: ultimately, unemployed individuals are responsible for their situation 9.1 Agree with statement: ultimately, unemployed individuals are not responsible for their situation 90.9 Agree with statement: I will have difficulty handling new technology at work 11.2 Agree with statement: I will not have difficulty handling new technology at work 69.6 Statement not applicable due to not working 19.2 1010 persons Table 6 Worker vignette: training subsidy is perceived as just—odds ratio Fixed effects (FE) odds ratio Model 1a Model 1b Model 2 Model 3 1. Vignette features Logit conditional FE Logit mixed + RI Logit mixed + RI Logit mixed + RI/RS Male (else: female) 0.757 (0.097) 0.779 (0.085) 0.764* (0.084) 0.637* (0.120) Age (ref: 34 years old) 46 years old 1.431* (0.219) 1.514*** (0.198) 1.556*** (0.204) 1.943** (0.440) 58 years old 0.850 (0.127) 0.827 (0.109) 0.823 (0.108) 0.614* (0.138) Unemployed (else: no) 1.938*** (0.246) 2.168*** (0.253) 2.228*** (0.258) 4.747*** (1.543) Funding time (ref: 4 weeks) Funding for 6 months 0.850 (0.137) 0.748* (0.108) 0.756* (0.108) 0.590* (0.149) Funding for 2 years 0.405*** (0.062) 0.410*** (0.054) 0.418*** (0.055) 0.198*** (0.053) Job to 75% replaceable (else: no) 2.036*** (0.282) 1.985*** (0.254) 1.941*** (0.245) 2.341* (0.874) Continuous job (else: no) 3.407*** (0.432) 3.226*** (0.409) 3.321*** (0.422) 10.486*** (4.703) 2. Personal characteristics Male (else: female) 1.472* (0.227) 2.142* (0.623) Age (ref: 50–64) Age 18–34 1.165 (0.234) 1.636 (0.596) Age 35–49 1.115 (0.197) 1.198 (0.377) Age 65–78 0.757 (0.312) 0.505 (0.370) Region (ref: Eastern States) German Southern States 0.954 (0.223) 0.907 (0.403) German Northern States 0.956 (0.223) 0.880 (0.371) German City States 1.309 (0.448) 1.373 (0.826) Education (ref: A-levels & vocational training) No education 0.644 (0.369) 0.336 (0.344) ( Technical) College 1.262 (0.202) 1.546 (0.461) Monthly net household Euro income (ref: 2000–3000) Less than 1,000 Euro 1.020 (0.457) 1.201 (0.932) 6 Page 16 of 18 R. V. Wolff et al. Table 6 (continued) Fixed effects (FE) odds ratio Model 1a Model 1b Model 2 Model 3 1000–1499 Euro 1.476 (0.598) 1.841 (1.325) 1500–1999 Euro 1.194 (0.340) 1.493 (0.745) 3000–3999 Euro 0.826 (0.186) 0.744 (0.298) 4000–4999 Euro 0.731 (0.181) 0.647 (0.284) 5000 Euro or more 0.504* (0.127) 0.358* (0.163) Income not specified 0.734 (0.241) 0.607 (0.357) Party affiliation (ref: ’Greens’) CDU (Christian Democratic. Union) 0.573* (0.158) 0.327* (0.169) CSU (Christian Social Union) 0.629 (0.271) 0.385 (0.304) SPD (Social Democratic Party) 1.400 (0.388) 1.487 (0.750) AfD (Alternative for Germany) 0.655 (0.288) 0.323 (0.260) FDP (Free Democratic Party) 0.322*** (0.103) 0.124*** (0.073) Die Linke ( The Left) 1.133 (0.374) 0.924 (0.548) Other party (not in parliament) 2.945* (1.422) 5.133 (4.398) Not specified 0.956 (0.186) 0.824 (0.295) Unemployment benefit (else: no) 2.152* (0.618) 4.770** (2.468) Unemployed responsible (else: no) 1.403* (0.222) 2.102* (0.621) Difficulty with new work technologies (ref: no) Difficulty with new tech 1.264 (0.296) 1.386 (0.569) Not applicable (no work) 1.287 (0.294) 1.658 (0.664) Constant (not odds ratio) 2.235*** (0.183) 1.786*** (0.357) 3.483*** (0.706) Random effects (RE) coefficient Var (constant) 2.886*** (0.386) 2.423*** (0.333) 13,860*** (3.989) Var (unemployed ‘UE’) 5.266* (2.198) Var ( job replaceable) 14.123*** (3.658) Var (continuous job) 11.619*** (3.301) Cov (constant, unemployed) -2.721 (1.998) Cov (constant, job replaceable) − 4.646* (2.261) Cov (constant, continuous job) − 3.912* (1.952) Cov (UE, job replaceable) − 4.002* (1.496) Cov (UE, continuous job) 4.770*** (1.427) Cov ( job replaceable, Cont. job) − 5.459** (1.752) Model fit criteria Information criteria: AIC; BIC AIC:1040; BIC:1083 AIC:3369; BIC:3432 AIC:3342; BIC:3582 AIC:3207; BIC:3504 R (FE) MF; McKelvey and Zavoina 0.203 (McFadden) 0.202 (MK&Z) 0.284 (MK&Z) 0.5691 (MK&Z) R (FE + RE) McKelvey and Zavoina N/A 0.406 (MK&Z) 0.423 (MK&Z) 0.7807 (MK&Z) Intraclass correlation (ICC) N/A 0.395 Odds ratios exclude constant and random effects coefficients; robust standard errors in parentheses; secular AIC decrease in mixed 1692 (logit conditional FE) or 4040 (logit mixed) vignette answers for 1010 persons *p < 0.050; **p < 0.005; ***p < 0.001 Acknowledgements Funding We are grateful to the Support network for interdisciplinary social policy The project is funded as an FIS project (00.00117.19) of the German Federal research (FIS) and the German Federal Ministry of Labor and Social Affairs for Ministry of Labor and Social Affairs. The German Federal Ministry of Labor and funding the research project (Grant Number: 00.00117.19) and to the Data Social Affairs had no role in the design of the study and collection, analysis, and IT Management of IAB, in particular Lina-Jeanette Metzger, for support in and interpretation of data. conducting the survey. Availability of data and materials Author contributions As the data underlying our analysis are not completely proprietary, access to The cited authors contributed insights in all parts of the paper. All authors the data is restricted. The data we use are social data. They contain sensitive read and approved the final manuscript. information and are subject to confidentiality regulations. Obtaining access Justice perceptions of occupational training subsidies: findings from a factorial survey Page 17 of 18 6 to the data through the research data center of the Institute for Employ- Sozialwissenschaften, Wiesbaden (2008). https:// doi. org/ 10. 1007/ 978-3- ment Research (IAB) requires a contract with IAB. We will support researchers 531- 90929-5_ 12 interested in replicating the results with the required formalities to receive German Bundestag: Weiterentwicklung der Arbeitslosenversicherung zu einer data access. Arbeitsversicherung. Ideen und Konzepte. Wissenschaftliche Dienste 6 – 3000 – 078/18. Berlin (2018). https:// www. bunde stag. de/ resou rce/ blob/ 580578/ 0ba6a e70b4 738cd b0710 3e62d 8bd5d f7/ WD-6- 078- 18- pdf- data. Declarations pdf Gilliland, S.W.: The perceived fairness of selection systems. An organizational Ethics approval and consent to participate justice perspective. Acad. Manag. Re. 18(4), 694–734 (1993). https:// doi. Not applicable. org/ 10. 5465/ amr. 1993. 94022 10155 Greenberg, J.: Organizational justice: yesterday, today, and tomorrow. J. Manag. Consent for publication 16(2), 399–432 (1990). https:// doi. org/ 10. 1177/ 01492 06390 01600 208 Not applicable. Hans, J.P., Hofmann, S., Sesselmeier, W., Yollu-Tok, A.: Umsetzung, Kosten und Wirkungen einer Arbeitsversicherung. Friedrich-Ebert-Stiftung, Bonn Competing interests (2017) The authors declare that they have no competing interests. Kaufmann, F.-X.: Herausforderungen des Sozialstaates. Suhrkamp, Frankfurt a. M (1997a) Author details Kaufmann, F.-X.: Schwindet die integrative Funktion des Sozialstaates? Berl. J. University of Bamberg: Otto-Friedrich-Universitat Bamberg, Bamberg, Bavaria, Soziol. 7(1), 5–19 (1997b) Germany. Institute for Employment Research (IAB): The Research Institute Klaus, A., Kruppe, T., Lang, J., Roesler, K.: Geförderte Weiterbildung Beschäft- of the Federal Employment Agency, Nürnberg, Bavaria, Germany. igter: Trotz erweiterter Möglichkeiten noch ausbaufähig. IAB-Kurzbericht 24/2020 (2020) Received: 1 June 2021 Accepted: 2 May 2022 Kluegel, R., Mason, D.S., Wegener, B.: The legitimation of capitalism in the post- communist tradition. Public opinion about market justice, 1991–1996. Eur. Sociol. Rev. 15(3), 251–283 (1999). https:// doi. org/ 10. 1093/ oxfor djour nals. esr. a0182 63 Konow, J.: A positive theory of economic fairness. J. Econ. Behav. Organ. 31(1), References 13–35 (1996). https:// doi. org/ 10. 1016/ S0167- 2681(96) 00862- 1 AAPOR ( The American Association for Public Opinion Research).: Standard Konow, J.: Fair and square. The four sides of distributive justice. J. Econ. Behav. definitions: final dispositions of case codes and outcome rates for surveys Organ. 46(2), 137–164 (2001). https:// doi. org/ 10. 1016/ S0167- 2681(01) 9 (2016) 00194- 9 Acemoglu, D., Restrepo, P.: Robots and Jobs: evidence from US Labor Markets. Kruppe, T., Lang, J., Stephan, G.: Das Qualifizierungschancengesetz: Mögli- NBER Working Paper No. 23285 (2017) chkeiten und Grenzen einer Evaluation. Arbeitspapier vom 15.4.2020. Adams, J.S.: Inequity in social exchange. In: Berkowitz, L. (ed.) Advances in Institut für Arbeitsmarkt- und Berufsforschung, Nürnberg (2020) Experimental Social Psychology, vol. 2, pp. 267–299. Academic Press, New Langer, W.: How to assess the fit of multi-level-logit models with Stata? In: York (1965). https:// doi. org/ 10. 1016/ S0065- 2601(08) 60108- 2 Meeting of the German Stata User Group at the Humboldt University Allaart, P., Bellmann, L., Leber, U.: Company-provided further training in Berlin. (2017). https:// langer. sozio logie. uni- halle. de/ stata/ pdf/ Langer- Ger- Germany and the Netherlands. Empir. Res. Vocat. Educ. Train. 1, 103–121 man- Stata- Users- Group- Meeti ng- 2017. pdf (2009). https:// doi. org/ 10. 1007/ BF035 46482 Leisering, L.: Eine Frage der Gerechtigkeit. Armut und Reichtum in Auspurg, K., Hinz, T.: Factorial survey experiments. SAGE Publications, Los Deutschland. In: Aus Politik und Zeitgeschichte (Beilage zur Wochenzei- Angeles (2015) tung Das Parlament) 30.4.1999, pp. 10–17 (1999) Bassanini, A., Booth, A., Brunello, G., Paola, M.D., Leuven, E.: Workplace training Leisering, L.: Paradigmen sozialer Gerechtigkeit. In: Liebig, S., Lengfeld, H., Mau, in Europe, IZA discussion paper no. 1640 (2005) S. (eds.) Verteilungsprobleme und Gerechtigkeit in modernen Gesells- Bell, A., Fairbrother, M., Jones, K.: Fixed and random effects models: making an chaften, pp. 29–68. Campus, Frankfurt a. M (2004) informed choice. Qual. Quant. 53, 1051–1074 (2019). https:// doi. org/ 10. Mau, S.: Welfare regimes and the norms of social exchange. Curr. Sociol. 52(1), 1007/ s11135- 018- 0802- x 53–74 (2004). https:// doi. org/ 10. 1177/ 00113 92104 03931 4 Benjamin, D.J., Berger, J.O., Johannesson, M., Nosek, B.A., Wagenmakers, Mikula, G.: Gerecht und ungerecht: Eine Skizze der sozialpsychologischen E.-J., Berk, R., Bollen, K.A., Brembs, B., Brown, L., Camerer, C., Cesarini, D., Gerechtigkeitsforschung. In: Held, M., Kubon-Gilke, G., Sturn, R. (eds.) Chambers, C.D., Clyde, M., Cook, T.D., Boeck, P.D., Dienes, Z., Dreber, A., Normative und institutionelle Grundfragen der Ökonomik. Jahrbuch 1: Easwaran, K., Eerson, C., F ff ehr, E., Fidler, F., Field, A.P., Forster, M., George, Gerechtigkeit als Voraussetzung für effizientes Wirtschaften, pp. 257–278. E.I., Gonzalez, R., Goodman, S., Green, E., Green, D.P., Greenwald, A.G., Had- Marburg (2002) field, J.D., Hedges, L.V., Held, L., Ho, T.H., Hoijtink, H., Hruschka, D.J., Imai, K., Miller, D.L.: Needs-based justice. Theory and evidence. In: Bauer, A.M., Meyer- Imbens, G., Ioannidis, J.P.A., Jeon, M., Jones, J.H., Kirchler, M., Laibson, D., huber, M.I. (eds.) Empirical Research and Normative Theory: Transdiscipli- List, J., Little, R., Lupia, A., Machery, E., Maxwell, S.E., McCarthy, M., Moore, nary Perspectives on Two Methodical Traditions between Separation and D.A., Morgan, S.L., Munafó, M., Nakagawa, S., Nyhan, B., Parker, T.H., Peric- Interdependence, pp. 273–294. De Gruyter, Berlin (2020) chi, L., Perugini, M., Rouder, J., Rousseau, J., Savalei, V., Schönbrodt, F.D., Morel, N., Palier, B., Palme, J.: Towards a Social Investment Welfare State? Ideas, Sellke, T., Sinclair, B., Tingley, D., Zandt, T.V., Vazire, S., Watts, D.J., Winship, C., Policies and Challenges. Policy Press, Bristol (2012) Wolpert, R.L., Xie, Y., Young, C., Zinman, J., Johnson, V.E.: Redefine statistical Osiander, C., Senghaas, M., Stephan, G., Struck, O.; Wolff, R.V.: Acceptance significance. Nat. Hum. Behav. 2, 6–10 (2018). https:// doi. org/ 10. 1038/ of social- and labor market programs and regulations: Methodologi- s41562- 017- 0189- z cal report on the first survey wave. In: IAB-Forschungsbericht 07/2020, Bowles, S., Gintis, H.: Reciprocity, self-interest, and the welfare state. Nordic J. Nürnberg (2020) Polit. Econ. 26, 33–53 (2000) Roosma, F., Gelissen, J., van Oorschot, W.: The multidimensionality of welfare Brynjolfsson, E., McAfee, A.: The Second Machine Age: Work, Progress, and state attitudes: a European cross-national study. Soc. Indic. Res. 113(1), Prosperity in a Time of Brilliant Technologies. Norton and Company, 235–255 (2013). https:// doi. org/ 10. 1007/ s11205- 012- 0099- 4 London (2014) Rossi, P.H., Anderson, A.B.: An introduction. In: Rossi, P.H., Nock, S.L. (eds.) Meas- Esping-Andersen, G., Gallie, D., Hemerijck, A., Myles, J.: Why We Need a New uring Social Judgments: The Factorial Survey Approach, pp. 15–67. Sage Welfare State. Oxford University Press, Oxford (2002) Publications, Beverly Hills (1982) Evers, A.: Investiv und aktivierend oder ökonomistisch und bevormundend? Rothstein, B.: Just Institutions Matter: The Moral and Political Logic of the Uni- Zur Auseinandersetzung mit einer neuen Generation von Sozialpolitiken. versal Welfare State. Cambridge University Press, Cambridge (1998) In: Evers, A., Heinze, R.G. (eds.) Sozialpolitik, pp. 229–249. VS Verlag für 6 Page 18 of 18 R. V. Wolff et al. Sachweh, P.: Social justice and the welfare state: institutions, outcomes, and attitudes in comparative perspective. In: Sabbagh, C., Schmitt, M. (eds.) Handbook of Social Justice Theory and Research, pp. 293–313. Springer Nature, New York (2016) Sachweh, P., Burkhardt, C., Mau, S.: Wandel und Reform des deutschen Sozialstaatsaus Sicht der Bevölkerung. WSI-Mitteilungen 62(11), 612–618 (2009). https:// doi. org/ 10. 5771/ 0342- 300X- 2009- 11 Snijders, T., Bosker, R.: Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modeling, pp. 102–103. Sage Publications, London (1999) Struck, O.: Flexibilität und Sicherheit. Empirische Befunde, theoretische Konzepte und politische Gestaltung (in-)stabiler Beschäftigung. VS-Verlag, Wiesbaden (2006). https:// doi. org/ 10. 1007/ 978-3- 531- 90389- 7 van den Berg, G.J., Dauth, C., Homrighausen, P., Stephan, G.: Informing employees in small and medium sized firms about training—results of a randomized field experiment, IZA discussion paper no. 11963 (2018) van Oorschot, W., Roosma, F., Meuleman, B., Reeskens, T. (eds.): The Social Legitimacy of Targeted Welfare: Attitudes to Welfare Deservingness. Edward Elgar Publishing, Cheltenham (2017) Vobruba, G.: Die Faktizität der Geltung: Gerechtigkeit im sozialpolitischen Umbaudiskurs. In: Clausen, L. (ed.) Gesellschaften im Umbruch. Verhand- lungen des 27. Kongresses der Deutschen Gesellschaft für Soziologie in Halle an der Saale 1995, pp. 963–975. Campus, Frankfurt a. M (1996) Wasserstein, R.L., Schirm, A.L., Lazar, N.A.: Moving to a world Beyond “p<0.05.” Am. Stat. 73(1), 1–19 (2019). https:// doi. org/ 10. 1080/ 00031 305. 2019. 15839 13 Young, H.P.: Equity. In: Theory and Practice. Princeton University Press, Prince- ton (1993) Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal for Labour Market Research Springer Journals

Justice perceptions of occupational training subsidies: findings from a factorial survey

Loading next page...
 
/lp/springer-journals/justice-perceptions-of-occupational-training-subsidies-findings-from-a-O0F6lugHZU

References (26)

Publisher
Springer Journals
Copyright
Copyright © The Author(s) 2022
ISSN
1614-3485
eISSN
2510-5027
DOI
10.1186/s12651-022-00311-w
Publisher site
See Article on Publisher Site

Abstract

Workers whose jobs are affected by structural change and digitization are required to continuously adapt their voca- tional skills to the requirements of the labor market. This adaptation is also essential for the competitiveness of their employer firms. The German legislature addressed this issue with investive measures for unemployment insurance, one of which is the Qualification Opportunities Act (Qualifizierungschancengesetz). Funds taken from unemploy- ment insurance can now be used to provide financial help for employers in a more direct way and on a broader scale than before. It became possible that not only unemployed individuals but also workers in companies receive state assistance. This paper analyses the extent to which citizens accept such public support programs for further training and which principles of justice they apply when assessing a just amount of training subsidies. We conducted two factorial surveys. First, we investigated the justice assessments of training subsidies for different types of firms. The results showed that citizens are inclined to subsidize companies by receiving social security funds for further training of their employees. However, when doing so, the principle of needs-based justice should be complied with. Second, we analyze whether citizens think it is just or unjust to provide training subsidies to different workers, as we present them with changing characteristics of workers. The findings confirmed that in addition to the principle of need, views on performance justice, as well as economic considerations are relevant in assessments of whether training subsidies co-financed by unemployment insurance are just. Keywords: Occupational training, Unemployment insurance, Justice assessments, Factorial survey, Multilevel and mixed effects model JEL Classification: C99, D63, I38, J08, J24, J65 1 Introduction been shown to be generally underrepresented among Rapid technical progress and globalization have increased those taking up occupational training (Bassanini et  al. the need for flexible adjustments of the workforce in the 2005). In particular, low-educated workers are often labor market. Investment in human capital can enable involved in undemanding routines with a high risk of workers and firms to adapt to changes in the economic substitution due to technical progress. environment and thus safeguard employment (Acemo- In view of this, the legislature in Germany enacted sev- glu and Restrepo 2017; Brynjolfsson and McAfee 2014; eral changes at the beginning of 2019 (further extended Struck 2006). Participation in training, however, is highly in 2020) with the Qualification Opportunities Act selective: workers in small and medium-sized enterprises (“Qualifizierungschancengesetz”). The Act grants com - (SMEs), older workers, and low-educated workers have panies that want to adapt the qualifications of their work - force due to structural change access to reimbursement for further training from unemployment insurance funds. *Correspondence: richard.wolff@uni-bamberg.de Hereafter also referred to as “the Act”. University of Bamberg: Otto-Friedrich-Universitat Bamberg, Bamberg, Bavaria, Germany Full list of author information is available at the end of the article © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. 6 Page 2 of 18 R. V. Wolff et al. Accordingly, funds from contribution-based unemploy- The article is structured as follows: The background ment insurance are no longer used to exclusively support of the new regulatory framework is presented in Sect.  2. unemployed individuals. This fits with the concept of The legal regulations are concordant with specific prin - a more preventive and investing social policy (Esping- ciples of justice and are partly included in the formula- Andersen et al. 2002; Morel et al. 2012). tion of the hypotheses. These are presented in Sect.  3. With the regulations of the Act, the legislature has sig- We develop a number of hypotheses on the principles nificantly expanded the group of those entitled to unem - of justice that may drive citizens’ assessments of the just ployment insurance funds. Against this backdrop, this amount of subsidies, e.g., related to needs-based justice article examines citizens’ attitudes about these changes. or efficiency-based justice (Leisering 2004). Method and Our research is situated within this framework toward data are described in Sect. 4. Using a factorial survey, we a more expansive and proactively investing “labor insur- ask participants how large training subsidies for different ance” as an expansion of unemployment insurance. Wel- firms should be and how just or unjust training subsi - fare institutions and their regulations for allocation and dies for different kinds of workers are perceived. By ran - disbursement of resources depend on certain principles domly varying either the features of firms or individuals that influence if (the level of) support payments, such described in two different vignettes, we aspire to causally as subsidized training cost reimbursements, are deemed determine the impact of these features on justice judg- appropriate. It is important for the legitimacy of the ments. Section 5 contains the empirical results. In Sect. 6 welfare state that citizens accept social policy meas- some conclusions are presented. ures and regulations and regard them as just (Rothstein 1998; Roosma et  al. 2013; Sachweh 2016). Major factors 2 Institutional background are principles of justice, which can affect the allocation In 2019, the Qualification Opportunities Act and in according to effort and compensation, (basic) needs, or 2020, the Work for Tomorrow Act (“Arbeit-von-Mor- (social) productive efficiency (Leisering 2004). An “inves - gen-Gesetz”) greatly extended the funding opportuni- tive turn” (Evers 2008) has been documented for more ties for firms undertaking occupational training for their than 10  years. This is a shift toward a more investive employees. Training subsidies for employed workers are social policy, which is also expected to yield social gains granted dependent on firm size (Klaus et  al. 2020). The (Esping-Andersen et al. 2002; Sachweh 2016, p. 309). The program aims to support employees who perform occu- Act is a step in this direction, which has been long and pational activities that can be replaced by modern tech- extensively discussed in the political sphere (German nologies, are otherwise affected by structural change, or Bundestag 2018)—but not among the general public. The plan to work in an occupation with a shortage of skilled legislature wants to support employees so that they can labor. adapt to structural change. The means of this support is Basic job training measures to be funded include— (partially) reimbursing the costs for occupational training among many others—software training courses or job- and the wage costs during training. Consequently, there specific language classes. Funding can range between are two addressees in regard to justice assessment of the partial and complete absorption of job training costs. Act: employees and firms. On the one hand, employ - Funding is conditional on the following four criteria. ers can participate. For some of them, the need may not First, the training has to provide knowledge that exceeds actually be real, which could then cause windfall gains for workplace-related short-term adaptations. Second, the such employers. On the other hand, there are employ- most recently acquired occupational degree must have ees where the need can be more direct. Furthermore, been obtained at least 4 years ago. Third, the job training respondents are able to perceive and evaluate differences measure has to be carried out by an accredited provider, between individuals and companies in terms of justifica - either outside or inside the firm. Fourth, job training tion based on performance or efficiency. Therefore, we must comprise at least 160 h (from 2020 onwards: 120 h), ask our respondents about both addressees to determine with the maximum duration not exceeding 3 years. if they deem the legal regulations for allocating train- Two types of financial support for qualification meas - ing cost reimbursements and the receipt of training cost ures are available for firms under this legislation. First, reimbursements for certain workers—both simulated by the costs for the job training itself can be subsidized. our two vignettes—as just. Employers are required to bear a partial burden of the training costs in a “reasonable manner”, which is speci- fied according to the firm size. Firms pay direct job train - Since 2007, vocational qualifications for low-skilled and older employees in German SMEs have also been subsidized through a special program, which ing costs according to firm size. Small firms with fewer has been used rarely (van den Berg et al. 2018; Klaus et al. 2020). than 10 employees can be compensated for the total costs For the purposes of this paper now simply referred to as “training subsi- of training. According to the Act, firms with 10–249 dies”. Justice perceptions of occupational training subsidies: findings from a factorial survey Page 3 of 18 6 (250–2499) [2500 and more] employees are required to as just. These principles can be aligned with the need contribute at least 50% (75%) [85%] of the costs. Costs of individual or collective actors to a greater or lesser may also be reimbursed completely for low-skilled work- degree (Kluegel et  al. 1999, p. 255; Gilliland 1993). ers participating in retraining, if the job training benefits This is connected to the principle of responsibility for a worker aged 45 or older, or if a worker is severely handi- results that can or cannot be influenced (Konow 1996, capped. Second, the wages of participating workers can 2001; Mikula 2002). However, principles of distribution be (partly) covered by unemployment insurance. Small can also be oriented toward the principle of contribu- firms of less than 10 (10–249) [250 and more employ - tion to a greater or lesser degree (Adams 1965; Young ees] may receive wage support of up to 75% (50%) [25%]. 1993), i.e., according to previously or currently being If the employee to receive job training lacks any voca- performed acts of the individual or collective (Green- tional training, labor costs can be reimbursed completely. berg 1990). For low-qualified employees participating in retraining, Moreover, it has been pointed out that justice assess- the entire wage bill might be covered by unemployment ments need to consider future effects of allocation condi - insurance during training. Training has to address needs tions (Vobruba 1996). This finding is taken into account that go beyond short-term workplace adaptations, such by “productivistic justice” (ibid. 969; cf. also Leisering as necessary adaptations due to exogenous structural 1999, p. 11). First, this means that citizens attempt to change, as smaller companies have (on average) fewer estimate a (socially) effective use of funds and consider resources at their disposal to respond to these challenges. this in their judgments. Second, this may also mean that In summary, the framework established in Germany citizens, in principle, take a favorable position toward applies the following key criteria: substitutability of jobs, an investive social policy (Sachweh et  al. 2009, p. 618), training duration, firm size, employee age, and employee which includes the Qualification Opportunities Act. The qualification. These criteria are reflected in regulations, prevention of unemployment and its associated costs which are primarily directed toward supporting com- may also save money in the mid-term (Hans et al. 2017). panies and employees who are in need (Kluegel et  al. In the following, we discuss, the potential justice 1999, p. 255; Gilliland 1993; Miller 2020), provided the assessments citizens make when determining the just company itself is not responsible for causing this need amount of subsidies for further training of different types (Konow 2001; Mikula 2002, p. 268). of firms . For this case in particular, we expect that citi- These criteria and their accompanying principles of jus - zens are mindful of neediness and a company’s capacity tice are investigated in our factorial survey. The theoreti - to act on its own behalf. It is also possible that a deci- cal background and the hypotheses that frame the survey sion is made on the grounds of effective usage of contri - are presented below. butions to unemployment insurance. Undue profits for companies might arise, as companies receive benefits from unemployment insurance for job-related training 3 Theoretical background and hypotheses that they might have carried out anyway (Kruppe et  al. Welfare state regulations impact individual life conditions 2020, p. 8). and structure social relationships. Citizens are directly Since this information is hidden, other signals, such as affected by the design of measures and regulations: as factors that influence the market position or size—SMEs recipients of transfer payments and as addressees of invest much less in training than large firms (Allaart social services, as well as contributors and taxpayers to et  al. 2009)—have been selected as indicators. A compa- finance the welfare state. It is therefore important for the ny’s capacity to act, the factor of self-responsibility and legitimacy of welfare states that citizens accept measures hereupon-derived neediness, can influence citizens’ jus - and regulations decided on a political level (van Oors- tice perceptions regarding the allocations of funds from chot et al. 2017). The acceptance of welfare programs and unemployment insurance. The Act distinguishes subsidy institutions is based on citizens’ principles about what levels according to company size but does not consider constitutes a just relationship between effort and reward, further indicators of the market position. as well as a certain living standard, which society grants its citizens in return for their contribution to society H1 Citizens provide higher training subsidies for com- (Bowles und Gintis 2000; Kaufmann 1997a; Mau 2004; panies that are economically weaker (H1a) or smaller in Roosma et al. 2013; Sachweh 2016). size (H1b) compared to companies that are economically To maintain a consensus on the condition of the strong or large in size. welfare state or its integrating functions (Kaufmann 1997b), sociopolitical norms and institutionally defined In principle, the legislature wants to provide allocation and dispensation mechanisms ought to help for employees, who are negatively affected by reflect principles of distribution that are perceived 6 Page 4 of 18 R. V. Wolff et al. technologically induced structural change. We expect However, the principle of need may also be addressed, that citizens acknowledge such need. It is presumed as employees with unstable work histories could be more that training subsidies for companies are higher if reliant on proactive, internal further training, due to their employees are affected by structural change with higher generally lesser chances on the labor market, which may probability. reduce or even reverse the effect of H5. In Germany, both unemployed and employed individu- H2 Citizens provide higher training subsidies for com- als can access training subsidies from unemployment panies, if funded occupations are strongly exposed to insurance. On the one hand, also among our respondents, technological progress. the employees themselves pay for social security insur- ance. The vast majority of our respondents and German citizens in general are under this obligation. It is possible Furthermore, we consider the extent to which train- that an expansion of beneficiaries among their own group ing subsidies are perceived as just for different types of is accepted and seen as a more appropriate compensa- employees. This also allows a comparison to the more tion for previously provided individual contributions. or less singular focus on unemployed individuals before On the other hand, unemployed individuals are probably the Act. We expect that for employees, the criteria for seen as needier than employees, not least because unem- the principle of need are also considered. ployed individuals’ benefits are also based on their past payments for unemployment insurance. Last, a focus on H3 Training subsidies for further training have a higher neediness is aligned with a more economical and effec - probability of being regarded as just for individuals who tive usage of limited resources. have characteristics that signal low labor market pros- pects (e. g., advanced age or occupations at risk of being H6 Training subsidies for further training have a higher replaced by technological progress in the future) than for probability of being regarded as just for unemployed individuals who are in a more favorable position on the individuals than for employed workers. labor market. Very long courses that may take two years and go beyond further training, such as vocational retraining, could be 4 Methods and data regarded as inefficient. That is because they are often dis - The empirical findings are based on a factorial survey; connected from current activities and qualifications, where such surveys have been proven suitable to investigate a more targeted and shorter further training would allow wide range of questions, such as social norms (Auspurg distribution of resources toward more recipients. Fund- and Hinz 2015; Rossi and Anderson 1982). We con- ing for very long vocational trainings is not possible under struct several fictitious situations (vignettes) and ask the the guidelines of the Qualification Opportunities Act that respondents to assess these situations. The situations supports companies and employees but is possible under randomly combine different characteristics along several unemployment insurance for unemployed individuals. dimensions. Three major advantages of this approach are Both derive their resources from the same source, so we that (a) respondents have to judge realistic situations; formulate the following hypothesis: (b) with the necessary caution, the causal effects of dif - ferent characteristics on assessments may be identified; H4 Training subsidies for further training have a higher and (c) the approach is relatively robust regarding biased probability of being regarded as just for shorter courses answering behavior, such as social desirability bias. than for courses with very long duration (2 years). We use two different sets of vignettes. The first relates to the firm dimension (H1 to H2), and the second focuses The principle of justice underlying social insurance on the worker dimension (H3 to H6). The vignette design schemes in Germany is the principle of equity, which creates longitudinal data with the occasion dimension for aligns contributions and benefits. From this point of each scenario, similar to panel data with the time dimen- view, those who contributed longer should also be pro- sion. For the two sets of vignettes, the respective vignette vided more benefits from insurance funds. variables are selected uniformly at random from every possible vignette set of the vignette universe and assigned H5 Training subsidies for further training have a higher to each survey participant. The four different vignettes probability of being regarded as just for employees with within each set represent a different occasion to evalu - stable work histories and therefore regular social secu- ate, so the judgment of each scenario is influenced by a rity contributions compared to individuals with irregular multitude of factors, whose intensity can vary between work histories. and within subjects. This is important, as it requires a Justice perceptions of occupational training subsidies: findings from a factorial survey Page 5 of 18 6 Table 1 Dimensions of the firm-related vignettes Dimension Characteristics Number of attributes Economic situation of the company Economically strong 2 Economically weak Company size 30 4 30,000 Number of funded employees 2 2 Degree of potential job automation in current occupation for funded Not mentioned 3 employees Already 25% of activities replaceable Already 75% of activities replaceable Duration of funding 4 weeks 2 6 months The vignette universe consists of 96 (2 × 4 × 2 × 3 × 2) combinations total, all of which are plausible (full factorial design) multilevel approach and will be addressed at the end of It can happen that the situations only differ slightly from this section. each other. For these cases, too, your judgment is impor- We briefly introduce the topic and later use different tant for us. It is not about “right” or “wrong”, we are inter- outcome variables for firm-related and worker-related ested in your assessment.” vignettes. According to the Act, firms receive subsidies Table  1 provides an overview of the dimensions of the depending on their characteristics. Therefore, we ask r fi m-related vignettes. These are the economic situation the respondents how high a percentage of the subsidies of the company, company size, number of employees tak- should be for the specific cases given in the vignette sce - ing part in training, possible degree of job automation, narios regarding companies. However, specific groups and training duration. of employees are also beneficiaries. The legislature aims The firm-related scenarios (bold letters for variable to support workers who are strongly affected by, for parameters) are presented in this text: instance, structural change and digitization. Conse- “An economically strong (weak) company with 30 quently, we also ask respondents to evaluate whether (300; 3000; 30,000) employees applies at the employ- they deem it just that a certain person be considered for ment agency for support for the occupational training funding of further training. of two (15) employees. The training lasts for 1  month First, we begin our survey with a general introduction (6 months). These employees work in professions in as presented below to make all respondents familiar with which 75% (25%; sentence not displayed) of activities the topic (German version in Appendix: Text A1): can already be replaced by computers and computer-con- “Employers and employees who are subject to social trolled machines.” security contributions are required to pay contribu- After each of the four different scenarios, respondents tions to unemployment insurance. These funds can also were asked to indicate which percentage of wages dur- be used to pay for further training in companies so that ing training and training costs should be reimbursed by employees can better adapt to new challenges at the unemployment insurance as a subsidy. Answers were workplace.” provided in 10% steps, ranging from zero to 100%. Second, we introduce the topic of the firm vignette Third, we introduced the topic of the worker vignette (underlining included) and then ask respondents to make (underlining included) and then asked respondents to a judgment about the preferred percentage of training judge how just or unjust they deem training subsidies subsidies (German version in Appendix: Text A2): (German version in Appendix: Text A3): “In the following, 4 different situations are described, in “In the following, 4 different situations are described, in which companies apply for funding of further training at which unemployed or employed people come up wanting the employment agency. further training. Please decide how much the company should receive in Please decide how just or unjust you find it, that the a certain situation for further training of its employees. respective person receives financing for further training 6 Page 6 of 18 R. V. Wolff et al. Table 2 Dimensions of the worker-related vignettes Having finished our depiction of vignettes, let us now turn to the description of our pool of respondents. The Dimension Characteristics Number of survey sample of approximately 35,000 people was sam- attributes pled uniformly at random from the Integrated Employ- Sex Male 2 ment Biographies (IEB V13.01.00-181010). The IEB Female covers all registered spells of employment subject to Age in years 34 3 social security contributions (including marginal employ- ment), unemployment, unemployment benefit receipt, job search and participation in active labor market pro- Job status Unemployed 2 grams in Germany. The sample was restricted to citizens Employed 18 years old or older at the time of data collection and to Job risk of automation Activities not replaceable in future 2 individuals of German nationality (Osiander et al. 2020). Already 75% of activities Data access further required individuals to have had replaceable an IEB spell during 2017 and at least one employment Job contribution to Continuous 2 spell during the period 2013–2017 to be included in the social insurance Intermittent and partial sample. Job training duration 4 weeks 3 Using this sample, between 11/2019 and 1/2020 24,934 6 months people were contacted via e-mail and 9551 people via 2 years post and asked to take part in an online survey. These The vignette universe consists of 144 (2 × 3 × 2 × 2 × 2 × 3) combinations total, invitations included information about the research pro- all of which are plausible (full factorial design) ject and data protection issues; a reference to the project homepage offered additional information. The e-mails from unemployment insurance funds. Likewise, it can contained an individualized link to the survey, while the happen that the situations only differ slightly from each letter included a QR code as well as a short link along other. Again, it is not about “right” or “wrong”, we are with an individual password. Almost 50% of participants interested in your assessment.” chose to answer the survey on their smartphones or tab- Table  2 provides an overview of the second type of lets, while the other half chose to use a laptop or desktop vignette, where respondents assessed whether funding PC. Overall, 1712 individuals started the survey, and after for a described individual worker was just. The dimen - accounting for missing answers to the questions, includ- sions analyzed are gender, age, risk of job loss due to ing refusal to merge their answers with the administrative automation, previous contributions to unemployment records on their labor market biographies in the IEB, a insurance, and the duration of training. balanced panel with 1010 persons remained. Worker-related scenarios (bold letters for variable In consideration of established guidelines (AAPOR parameters) are presented in this text: 2016), we calculate the net response ratios conserva- “A 34 (46; 58)-year-old employed (unemployed) tively. Approximately 3.8% of people who were contacted man (woman) works in a profession in which 75% of (2.7% for invitations via e-mail and 6.7% for invitations activities can be replaced by computers or computer- via post) on valid addresses fully answered the survey. controlled machines (which cannot be replaced by This is in line with expectations for such contact chan - computers in the future). After finishing vocational nels. The sample of our participants is not representa - training, he (she) was continuously employed and con- tive of the German labor force. However, we were able to tributed (intermittently employed and partially con- conduct selectivity analyses for the combined samples of tributed) to unemployment insurance. For this person, e-mail and post. Almost 50% of respondents in the gross unemployment insurance finances occupational training sample are female, 20% live in Eastern Germany (includ- with a duration of 4 weeks (6 months; 2 years).” ing Berlin), 75% have completed vocational training or Respondents were asked how just, from their point of university and 60% are qualified skilled workers. Our view, it is that the employment agency pays for this occu- selectivity analysis compares the gross sample to the final pational training, using funds from the unemployment sample of respondents. People from Eastern Germany insurance? (including Berlin) and people who are 65 years and above are underrepresented, while people 50–64 are overrep- Strictly analytical something can only be either just or unjust and nothing resented. Moreover, people with higher formal qualifica - in between. However, in empirical reality and in colloquial language this dis- tions, longer duration of both employment and previous tinction is not so clear-cut, so respondents may still want to select a more nuanced position. Consequently, we allowed four different outcomes (ranging unemployment receipt are also overrepresented in our from unjust and quite unjust to quite just and just). Afterward, the two sides final sample (Osiander et al. 2020). were combined into the dichotomy of unjust/just. Justice perceptions of occupational training subsidies: findings from a factorial survey Page 7 of 18 6 All participants received two sets of four vignette sce- first vignette can also benefit from this. We know that narios. Furthermore, we collect information on: gender, mixed effects with random intercept and random slopes age, qualification, political preferences and classified net can be important because the cluster-robust variance– monthly household income. We also include questions covariance estimator (Eicker–Huber–White procedure) about (a) the respondent’s attitude toward unemployed accounts only for between-person heteroscedasticity but individuals and (b) how respondents handle new tech- not for variance differences originating within clusters. In nologies at the workplace. The respondent characteris - our case, only the vignettes vary within the cluster that tics can control for selective distortions in the sample, each participant represents. such as age, income, political views and for behavior For the firm vignette, the linear mixed regression patterns, such as self-interested behavior. Depending on model includes random slopes for the attributes “com- the circumstances, someone is primarily a beneficiary or pany size”, “degree of job automation” and “duration of a payer in the unemployment insurance system. Factors funding” (cf. Table 1) to account for within-person stand- such as regional work opportunities, position in the labor ard deviation differences with random slopes. To moti - market and technical knowledge can play a role, and peo- vate the mixed model, we check three criteria: relevant ple may choose the option they believe to benefit from effects and noteworthy p-values of the correlation coeffi - the most. A summary of all variables and their operation- cients, the results of a likelihood ratio test, and improved alization is available in Appendix Table 5. information criteria (AIC/BIC). We model both the Each participant judged four vignettes related to the standard deviation for random slopes and the correlation company level and four vignettes related to the employee between random slopes, including the correlation of ran- level. These four evaluations are very likely not inde - dom slopes with the random intercept. We choose con- pendent of each other, given that they are from the same servative unstructured standard deviations because we person. cannot assume a certain standard deviation pattern. First, For both the firm vignette (size of training subsidies we confirm that all standard deviation coefficients are perceived as just) and the worker vignette (training sub- relevant in size and statistically significant (p < 0.001). For sidy is perceived as just), this is accounted for by estimat- the correlation terms, all selected random slope variables ing multilevel models such as fixed effects models, but are relevant in size and statistically significant (p < 0.001) also mixed effects models using random intercepts and as they correlate with the constant, so these variables are slopes at the individual level. The fixed aspect means that all regarded as important for inclusion in the model. Sec- only the within-variation of the vignette variables enters ond, a likelihood ratio test (p < 0.001) corroborates this the model, while mixed models can fit all coefficients. In selection of random slopes when compared to the same our mixed model, both vignette variables and individual model but without random slopes. Third, the chosen ran - characteristic variables, plus a selection of random slopes dom slopes collectively improve both information crite- (also called random coefficients), are fitted at the indi - ria, AIC and BIC (cf. Table  3), to avoid overfitting. Due vidual level. In general, mixed models can deliver more to extensive testing, we can confirm that no other possi - reliable results than standard linear panel regression with ble combination of random slopes increases the model fit random effects. Therefore, we account for the multilevel further. Please take note that the pseudo-R of Snijders/ structure in the first vignette with a linear mixed model. Bosker does not take into account random slopes, which Mixed models combine the advantages of both fixed is why its results in Model 2 and Model 3 are identical. effects and random effects (Bell et al. 2019) but are there - This measure is used here to point out a general increase fore more complex. Mixed effects with random effects in explained variance between Model 1b and Model 2 at coefficients for only intercept and residual are identi - level 1, which is our vignette level, and at level 2, which cal to random effects models. Additional slopes can be is the level of individual characteristics (Snijders and regarded as an extension of the random effects model and Bosker 1999). In summary, the selected linear mixed are the hallmark of mixed models. Analogous to the gen- effects panel model has advantages over a standard ran - eral assumption that the random intercept model follows dom effects model, as it further reduces heteroscedastic - a normal distribution, random slopes follow a multivari- ity (Bell et al. 2019). ate normal distribution. Such random slopes can capture For the worker vignette with binary coding of just or and model even more variance than the mere inclusion of unjust, a logit mixed model with random slopes deliv- a random intercept can. Therefore, unless parallel slopes ers different and more reliable results than a standard can be assumed, mixed models should be considered a panel logit with random effects could. Once again, we first choice before fitting a simple random effects model. choose to address the multilevel structure by apply- This proved particularly important for the logit models ing mixed effects, now with a logit mixed model. We used for the second vignette, but the linear models of the fit conditional logit fixed and mixed logit models by 6 Page 8 of 18 R. V. Wolff et al. Table 3 Firm vignette: size of training subsidies perceived as just (in 10% intervals) Model 1a Model 1b Model 2 Model 3 Fixed effects coefficients 1. Vignette features FE Mixed + RI Mixed + RI Mixed + RI/RS Company size (ref: 30) 300 employees − 4.163*** (0.897) − 4.028*** (0.887) − 3.993*** (0.888) − 3.644*** (0.853) 3000 employees − 9.782*** (0.956) − 9.567*** (0.937) − 9.563*** (0.937) − 9.689*** (0.885) 30,000 employees − 15.144*** (0.973) − 15.016*** (0.957) − 15.019*** (0.956) − 14.484*** (0.938) Strong company (else: weak) − 16.924*** (0.703) − 16.869*** (0.697) − 16.929*** (0.696) − 16.766*** (0.683) Training for 15 people (else: two) − 0.928 (0.668) − 0.791 (0.659) − 0.735 (0.660) − 0.934 (0.632) Job at risk of automation (ref: nothing) Already 25% replaceable − 0.569 (0.914) − 0.433 (0.881) − 0.401 (0.879) 0.092 (0.857) Already 75% replaceable − 2.588** (0.948) − 2.551** (0.924) − 2.562** (0.922) − 1.857* (0.893) Funding for 6 months (else: four weeks) − 0.792 (0.724) − 0.679 (0.690) − 0.609 (0.688) − 0.911 (0.619) 2. Personal characteristics Male (else: female) − 6.214*** (1.560) − 5.958*** (1.543) Age of respondent (ref: 50–64) Age 18–34 5.439* (2.030) 5.559* (2.002) Age 35–49 1.222 (1.793) 1.523 (1.774) Age 65–78 − 0.249 (4.030) 0.124 (4.027) German region (ref: ‘Eastern States ‘) German Southern States − 2.720 (2.389) − 3.288 (2.393) German Northern States − 4.793 (2.416) − 5.635* (2.425) German City States 0.448 (3.279) − 0.854 (3.257) Education (ref: A-levels/vocat. training) No education 6.380 (6.896) 5.986 (6.567) ( Technical) College − 3.657* (1.576) − 3.719* (1.561) Net hh-income/month (ref: 2000–2999) Less than 1000 Euro − 0.993 (4.460) − 0.932 (4.364) 1000–1499 Euro 1.018 (3.347) 0.514 (3.248) 1500–1999 Euro 6.178* (2.919) 5.475 (2.944) 3000–3999 Euro 1.023 (2.358) 0.183 (2.345) 4000–4999 Euro − 1.496 (2.615) − 1.830 (2.605) 5000 Euro or more 2.638 (2.712) 2.261 (2.681) Income not specified − 1.390 (3.153) − 2.239 (3.139) Party affiliation (ref: ‘Greens’) CDU (Christian Democratic Union) − 1.834 (2.836) − 0.760 (2.813) CSU (Christian Social Union) − 6.275 (4.633) − 5.731 (4.682) SPD (Social Democratic Party) − 0.005 (2.594) 0.251 (2.593) AfD (Alternative for Germany) − 0.812 (4.767) − 0.416 (4.758) FDP (Free Democratic Party) − 9.216* (3.373) − 8.543* (3.322) Die Linke ( The Left) − 1.059 (3.051) − 1.901 (2.994) Other party (not in parliament) 2.353 (4.574) 2.289 (4.620) Not specified 2.027 (2.029) 2.264 (2.002) Unemployment benefit (else: no) 0.921 (1.708) 0.734 (1.697) Unemployed responsibility (else: no) 1.890 (2.545) 2.412 (2.520) Difficulty with new work tech (ref: no) Difficulty with new tech 3.969 (2.489) 3.861 (2.436) Not applicable (no work) 0.264 (2.170) 0.325 (2.166) Constant 62.778*** (1.262) 66.889*** (3.693) 67.326*** (3.681) Justice perceptions of occupational training subsidies: findings from a factorial survey Page 9 of 18 6 Table 3 (continued) Model 1a Model 1b Model 2 Model 3 Random effects (RE) coefficients Sd (constant) 22.670*** (0.524) 21.774*** (0.509) 26.984*** (0.776) Sd (residual) 17.873*** (0.327) 17.873*** (0.327) 12.703*** (0.502) Sd (strong company) 14.899*** (0.937) Sd (25% replaceable) 13.116*** (1.503) Sd (75% replaceable) 13.170*** (1.487) Sd (subsidize 15) 10.051*** (1.416) Corr (strong company, 25% replaceable) 0.179 (0.108) Corr (strong company, 75% replaceable) 0.014 (0.101) Corr (strong company, subsidize 15) 0.122 (0.092) Corr (strong company, constant) − 0.431*** (0.047) Corr (25% replaceable, 75% replaceable) 0.189 (0.145) Corr (25% replaceable, subsidize 15) − 0.086 (0.124) Corr (25% replaceable, constant) − 0.263*** (0.070) Corr (75% replaceable, subsidize 15) 0.253 (0.143) Corr (75% replaceable, constant) − 0.224*** (0.067) Corr (subsidize 15, constant) − 0.390*** (0.061) Model fit criteria Information criteria: (1) AIC; (2) BIC (1) 1040; (2) 1083 (1) 36,810; (2) 36,879 (1) 36,796; (2) 37,042 (1) 36,545; (2) 36,879 R : McFadden; Lvl. 1&2 Snijders/Bosker 0.203 (MF) 0.106; 0.013 (S/B) 0.148; 0.080 (S/B) 0.148; 0.080 (S/B) 4040 vignette answers for 1010 persons FE fixed effects, RI random intercept, RS random slopes *p < 0.050; **p < 0.005; ***p < 0.001; robust standard errors in parentheses maximizing the log pseudolikelihood (cf. Appendix Table  2) to model within-person variance differences Table  6) and in postestimation predict robust aver- with random slopes. We do this because of three fac- age marginal effects (cf. Table  4). Table  6 indicates tors: the relevance of the variance coefficients and the both secular improvement with a decreasing AIC and noteworthy p-values for the covariance coefficients, an increase in McKelvey–Zavoinas’ pseudo-R from the results of a likelihood ratio test, and the improved approximately 20% to approximately 80% as a measure information criteria. We begin by modeling the vari- of variation explained by the model. This pseudo-R ances for random slopes and the covariances between is also recommended as most suitable for logit mod- random slopes, including the covariances of random els (Langer 2017). In Model 1a, we include only the slopes with the random intercept. We choose con- vignette variables for a logit conditional fixed effect servative unstructured covariances because we cannot for comparison with the following mixed models. In assume a certain covariance pattern. First, we notice Model 1b, the same vignette variables are used to fit that the results for the chosen random slopes show individual coefficient variances as random intercepts. considerable values for the individual variance coef- In Model 2, we then add personal characteristics. ficients. The covariance coefficients also have note- For reference, the results of Model 1b and Model 2 worthy effects, with (almost all) p-values smaller than are identical to those fitting a standard logit random 0.05, and are therefore regarded as relevant for inclu- effects regression, which does not consider slopes. In sion in our model. Second, further validation of the the final Model 3, we add individual random slopes to chosen random slope model is given by the results of model the variance structure. For this, we choose to a likelihood ratio test (p < 0.001) with selected random model the attributes “job status”, “job risk of automa- slopes compared to a model without random slopes. tion” and “job contribution to social insurance” (cf. Third, all chosen random slopes collectively improve both information criteria, AIC and BIC (cf. Table  6). This means that the model is not overfitted. No other Due to convention, variance and covariance is used for random coefficients here, but after applying a transformation to normalize the estimate, which possible combination of random slopes improves the does not affect p-values, the results would be analogous to standard deviation model fit further, which reassures us that it is not and correlation of the linear mixed model. 6 Page 10 of 18 R. V. Wolff et al. Table 4 Worker vignette: training subsidy is perceived as just—Avg. marginal effect Avg. predicted marginal effect Model 1a Model 1b Model 2 Model 3 1. Vignette features Logit fixed Logit mixed + RI logit mixed + RI logit mixed + RI/RS Male (else: female) − 0.058* (0.027) − 0.025** (0.011) − 0.027* (0.011) − 0.025* (0.010) Age (ref: 34 years old) 46 years old 0.074* (0.031) 0.040*** (0.012) 0.042*** (0.012) 0.035** (0.012) 58 years old − 0.034 (0.032) − 0.020 (0.014) − 0.021 (0.014) − 0.028* (0.013) Unemployed (else: no) 0.140*** (0.027) 0.078*** (0.012) 0.081*** (0.012) 0.086*** (0.011) Funding time (ref: 4 weeks) Funding for 6 months − 0.034 (0.034) − 0.027* (0.013) − 0.026* (0.013) − 0.027* (0.013) Funding for 2 years − 0.193*** (0.032) − 0.092*** (0.013) − 0.090*** (0.013) − 0.091*** (0.012) Job 75% replaceable (else: no) 0.150*** (0.029) 0.069*** (0.013) 0.067*** (0.013) 0.070*** (0.013) Continuous job (else: no) 0.267*** (0.026) 0.119*** (0.012) 0.122*** (0.012) 0.117*** (0.012) 2. Personal characteristics Male (else: female) 0.039* (0.016) 0.042* (0.016) Age (ref: 50–64) Age 18–34 0.015 (0.020) 0.027 (0.019) Age 35–49 0.011 (0.018) 0.010 (0.017) Age 65–78 − 0.030 (0.046) − 0.041 (0.046) Region (ref: ‘Eastern States’) German Southern States − 0.005 (0.024) − 0.005 (0.024) German Northern States − 0.005 (0.024) − 0.007 (0.023) German City States 0.026 (0.032) 0.017 (0.032) Education (ref: A-levels and vocational training) No education − 0.049 (0.068) − 0.068 (0.068) ( Technical) College 0.023 (0.016) 0.024 (0.016) Monthly net household Euro income (ref: 2000–3000) Less than 1000 Euro 0.002 (0.042) 0.009 (0.040) 1000–1499 Euro 0.034 (0.034) 0.030 (0.034) 1500–1999 Euro 0.016 (0.026) 0.020 (0.025) 3000–3999 Euro − 0.019 (0.022) − 0.016 (0.022) 4000–4999 Euro − 0.032 (0.025) − 0.024 (0.024) 5000 Euro or more − 0.074* (0.027) − 0.060* (0.026) Income not specified − 0.031 (0.034) − 0.028 (0.033) Parties (ref: ‘Greens’) CDU (Christian Democratic Union) − 0.060* (0.031) − 0.065* (0.031) CSU (Christian Social Union) − 0.050 (0.049) − 0.055 (0.048) SPD (Social Democratic Party) 0.031 (0.025) 0.020 (0.025) AfD (Alternative for Germany) − 0.045 (0.049) − 0.066 (0.050) FDP (Free Democratic Party) − 0.135*** (0.042) − 0.132*** (0.039) Die Linke ( The Left) 0.012 (0.031) − 0.004 (0.032) Other party (not in Parliament) 0.085* (0.031) 0.071* (0.031) Not specified − 0.004 (0.019) − 0.01 (0.019) Unemployment benefit (else: no) 0.069** (0.022) 0.075*** (0.020) Unemployed responsible for situation (else: 0.035* (0.016) 0.042* (0.016) no) Difficulty with new work technologies (ref: no) Difficulty with new tech 0.023 (0.023) 0.018 (0.022) Not applicable (no work) 0.025 (0.022) 0.027 (0.021) 4040 vignette answers; 1010 persons RI random intercept, RS random slopes *p < 0.050; **p < 0.005; ***p < 0.001; robust standard errors in parentheses Justice perceptions of occupational training subsidies: findings from a factorial survey Page 11 of 18 6 30% 26% 25% 20% 15% 15% 12% 10% 9% 10% 8% 7% 4% 4% 4% 5% 2% 0% Funding in percent Notes: 4,040 vignettes, 1,010 persons. Fig. 1 Distribution of funding percentages to subsidize firm costs of further training. necessary to include additional random slopes, as dif- (Wasserstein et  al. 2019) of the effects of firm charac - ferences in variances can already be captured by the teristics on justice assessments. Table  3 presents the chosen final model. This is also supported by dou- results from Model 1a with fixed effects and Model 1b bling the model fit criterion of pseudo-R (FE + RE) with mixed effects and random intercept. Both cover from approximately 40% to approximately 80% when only the vignette dimensions. Model 2 also controls for comparing Model 2 to Model 3 (cf. Table 6). respondent characteristics. Finally, Model 3 is the same In conclusion, a standard panel logit model with ran- as Model 2 but with random slopes. The following dis - dom effects would not be appropriate here. Such a model cussion refers to Model 3. Note that we focus our inter- would seriously violate the model assumptions of homo- pretation on point estimates (random coefficients are not scedasticity and deliver anti-conservative results that to be interpreted for this purpose)—rounded to the next introduce a bias on standard errors and point estimates ½ % or half percentage point—for coefficients with low (Bell et  al. 2019), so we prefer the final mixed model p-values of < 5%, < 0.5% and < 0.1%, as indicated by one, with the aforementioned random slopes to mitigate this two or three asterisks; standard errors can be found in problem. the respective tables to reflect this. The p-values with one asterisk between < 5% and 0.5% should be seen as merely 5 Empirical findings suggestive, while those < 0.5% may be considered signifi - In the first step, we analyze the extent to which features cant, and those < 0.1% may be considered even more so of the firm have an impact on the perceived just amount (Benjamin et al. 2018). of training subsidies. The average amount of funding First, survey respondents grant substantially higher across all vignettes is approximately 45%. As outlined subsidies to economically weak firms than to economi - above, the design of our study ultimately does not allow cally strong firms (≈ + 17 percentage points). Further- us to draw conclusions regarding the general support of more, the subsidy share increases with firm size—each training subsidies in the population. A descriptive analy- tenfold increase in the number of employees reduces sis shows that in 85% of the scenarios, the respondents the assigned subsidy (≈ − 5.0 percentage points each would financially support training subsidies at least step), particularly for very large firms (≈ − 14.5 percent- to some degree (see Fig.  1). Due to our factorial survey age points). As predicted by H1a and H1b, participants design, we can cautiously draw a causal interpretation Percent of answers per category 6 Page 12 of 18 R. V. Wolff et al. slightly more likely to be perceived as a just recipient of thus account for the basic justice principle of need when funding for further training (≈ + 3.5%), while this is less assigning the just amount of training subsidies. How- likely for an older individual of age 58 (≈ − 3%). It is plau- ever, the degree of differentiation according to firm size sible that younger workers, due to their usually more is smaller than that granted in current legislation in up-to-date training and long-term prospects for amorti- Germany. zation of their own investment in training, are perceived We find, however, no support for H2 that individuals as less needy. For older persons, respondents may assume assign more support to firms where trained workers are that they are close to retirement and that investments in employed in occupations threatened by technological training will thus not pay off. Furthermore, the respond - change. Survey participants assign even less funding if ents might presume that older learners have diminished the firm trains workers whose jobs are strongly exposed learning capabilities. Concerning age, H3 is only partly to potential job automation (≈ − 2 percentage points) confirmed, as respondents also seem to take the principle compared to the reference scenario where no information of efficiency into account. on the risk of automation was given. It is possible that the With respect to training duration, respondents per- participants consider such work sites a “lost cause” and ceive funding for training over six months (≈ − 3%) or would rather not waste resources on the company itself. two years (≈ − 9%) as less just than a shorter training of Regarding respondents characteristics, men gener- 4  weeks. While 4  weeks are already slightly preferred to ally attribute less funding to the firm (≈ − 6 percentage 6  months, the difference is quite small and not relevant points), adults younger than 35 attribute more fund- for the analysis, which focuses on long courses. Accord- ing compared to the reference group of 50–64-year-old ing to H4, such long courses will be deemed too long. respondents (≈ + 5.5 percentage points) and respond- Our results are in line with H4, so efficiency considera - ents from German “Northern States” give less (≈ − 5.5 tions may also play a role in justness assessments of train- percentage points) than respondents from the East. The ing subsidies. effect of monthly net classified household income is not H5 presumed that training subsidies for workers with relevant, while college education has a small negative continuous contributions to social insurance would more effect on the level of support (≈ − 3.5 percentage points). likely be regarded as just than subsidies to workers with Furthermore, individuals who identify themselves as vot- unstable work histories. Here, we indeed find a compara - ers of the Free Democratic Party (FDP) grant much less tively strong effect (≈ + 12%). This supports H5, which is support to firms (≈ − 8.5 percentage points), in line with based on the principle of equity, while needs-based jus- the small government approach of the party. tice may only play a minor role here. In the second step, we analyze the importance of the Finally, an important feature of training subsidies is characteristics of trained workers, who also stand to gain whether the support is directed to unemployed indi- from the Act. As beneficiaries, they are thus included in viduals or employed workers. The estimates show that the assessment of training subsidies in another vignette training subsidies for the occupational training of unem- design. The dependent variable is a binary variable indi - ployed individuals are regarded as more just (≈ + 8.5%). cating whether funding is perceived as just or unjust. According to H6 we expected that training subsidies for Approximately 20% of singular vignette answers regarded further training are more likely to be perceived as just for funding as unjust and 80% as just. In the following, we unemployed individuals than for the employed. While will use the average marginal effects of Table  4 Model 3 the respondents may consider principles of both need for further interpretation. and equity, need seems to be the dominant principle H3 posited that training subsidies are more likely to be in this context. The results provide support for H6, as perceived as just for individuals in need, in particular for unemployed individuals are clearly preferred. Neverthe- persons whose occupations are exposed to automation less, training subsidies for employees are also judged to and older workers. Indeed, the respondents judge train- be fair by the vast majority of vignette answers. This is in ing subsidies to be just more often (≈ + 7%) if the recipi- line with the legislation, as employees also have access to ent of the subsidy has been working in an occupation training paid for by the same fund that thus far has prior- where tasks could already be substituted to a high degree itized unemployed individuals. (75%) by computers. This result is in line with H3. Regarding age, the results show that compared to a To test what determines dominant principles further, interaction effects 34-year-old person, a middle-aged person (age 46) is were calculated, but they provided no further insights. Multiple tests for inter- action effects between vignettes and certain personal characteristics related to The Act grants 25 percentage points less as firm size increases from 30 the vignettes (e. g., age of respondent and age in vignette; sex of respondent to 300 and another 10 percentage points less for training costs as firm size and sex in vignette) were carried out, but none: had strong effects, improved increases from 300 to 3000—compared to our 5.0 percentage points each step. information criteria or were statistically significant/relevant enough to war - rant inclusion in the model. Justice perceptions of occupational training subsidies: findings from a factorial survey Page 13 of 18 6 Regarding respondent characteristics, we find a few of equity but also indications that the respondents appre- effects on justice considerations. Men (≈ + 4%) are ciate an economical use of funds. Public training support more inclined to show support. Respondents with a for workers is more often assessed as just, if the work- high net household income are more reluctant to con- ers are currently unemployed or if their occupation is sider assistance to be just (≈ − 6.0%). For political lean- strongly exposed to potential automation, in line with the ings, strong effects were found for a political preference principle of need. Furthermore, workers are preferred to for the Christian Democratic Union (CDU ≈ − 6.5%), for exhibit stable work histories and thus steady contribu- the Free Democratic Party (FDP ≈ − 13%) and for other tions to unemployment insurance, in line with the princi- parties not currently in parliament (≈ + 7%) when com- ple of equity. The productive use of resources is specified pared to the reference of supporting the Green party. by favoring people of middle age and lack of support for Respondents who ever received unemployment benefits long training durations, in line with economical use of through unemployment insurance show greater sup- funds. port (≈ + 7.5%). Finally, if the respondent agrees with the Fourth, when comparing both dimensions, exposure to statement in respondent characteristics that unemployed technological change is only statistically significant for individuals themselves are responsible for their situation, the worker vignette, but not for the firm vignette. While then this slightly increases the acceptance (≈ + 4%) to both vignettes are different in other ways as well, it can provide training subsidies in the worker vignette. be speculated that supporting workers with training sub- sidies is more tenable. That may be, because workers can directly and more easily benefit by switching to a more 6 Discussion of key findings suitable line of work and even industry—something firms Against the background of new funding options as part cannot easily do. of unemployment insurance regarding further training In light of principles of justice to evaluate the new for currently employed workers in Germany, this article funding possibilities of unemployment insurance, our analyzes the determinants of justice perceptions of such results show that respondents largely exhibit congruence support measures. The state provides the opportunities with neediness and thus not with the principle of equity. for firms to apply for a (partial) reimbursement of wage However, productivistic justice principles and orienta- costs during training, as well as training costs. The analy - tion toward the efficient application of social insurance sis shows that citizens generally accept the training subsi- resources can also come to the fore dies for occupational further training in the Qualification Opportunities Act. Our factorial surveys also uncover the principles of justice underlying the assessment of training subsidies for firms and for workers. Appendix First, an important technical point is that the mixed effects models highlight the need to incorporate (appro - Text A1: Original German version of general priate) random slopes, which can yield a consider- introduction able boost in explained variance to strengthen the study Sozialversicherungspflichtig Beschäftigte und Arbeitgeber results. Future theoretical research could determine, zahlen in Deutschland Beiträge zur Arbeitslosenversicherung. if the size of this boost is dependent on the number of Von diesem Geld können auch Beschäftigte in vignette variables, vignettes per respondent or other Unternehmen Weiterbildungskurse bezahlt bekommen, factors. damit Mitarbeiterinnen und Mitarbeiter sich besser Second, focusing on the firm dimension, we show that an neue Herausforderungen am Arbeitsplatz anpassen that the approach of German legislation to reimburse a können. larger share of training costs for small firms is mirrored by the assessments of training subsidies by the respond- ents. However, respondents differentiate their assess - Text A2: Original German version of the firm ments by firm size less than the recent legislation in vignette and its introduction Germany does. Respondents would grant more support Im Folgenden werden vier verschiedene Situationen to economically weak and small firms, which are more in beschrieben, in denen Unternehmen eine Förderung von need than stronger and larger firms are. We find, how - Weiterbildungen bei der Arbeitsagentur beantragen. ever, no indication that funding should increase with Bitte entscheiden Sie, wie viel das Unternehmen in training duration and with an increased degree of job einer bestimmten Situation für die Weiterbildung seiner automation to which funded workers are exposed. Beschäftigten bekommen sollte. Es kann vorkommen, Third, focusing on the worker dimension, we find sup - dass sich die Situationen nur geringfügig voneinander port for the basic justice principle of need and principle unterscheiden. Auch in diesen Fällen ist uns Ihr Urteil 6 Page 14 of 18 R. V. Wolff et al. darüber wichtig. Es geht nicht um “richtig” oder “falsch”, aus Mitteln der Arbeitslosenversicherung finanziert wir sind an Ihrer Einschätzung interessiert. bekommt. Es kann auch hier vorkommen, dass sich die Situ- ationen nur geringfügig voneinander unterscheiden. Es Firm vignette geht wieder nicht um “richtig” oder “falsch”, wir sind an Ein wirtschaftlich solides (schwaches) Unternehmen Ihrer Einschätzung interessiert. mit 30 Beschäftigten (300; 3.000; 30.000 Beschäft- igten) beantragt bei der Arbeitsagentur die Förderung einer Weiterbildungsmaßnahme für zwei Beschäftigte Worker vignette (15 Beschäftigte). Die Weiterbildung soll einen Monat Ein/e 34-jährige/r (46-jährige/r;58-jährige/r) erwerbstätige/r (6  Monate) dauern. Diese Beschäftigten arbeiten in (arbeitslose/r) Mann (Frau) arbeitet in einem Beruf, dessen Berufen, in denen bereits heute 75 Prozent (25 Prozent; Inhalte schon heute zu 75% durch Computer oder Roboter Satz nicht erwähnt) ihrer Tätigkeiten durch Computer ersetzbar sind (der auch zukünftig nicht durch Computer oder computergesteuerte Maschinen ersetzbar sind. ersetzt wird). Er (Sie) war seit der Ausbildung dauerhaft Wieviel Prozent der Kosten für den Lohn und die beschäftigt und hat Beiträge (unregelmäßig beschäftigt und Weiterbildung des Mitarbeiters soll die Arbeitsagentur hat phasenweise Beiträge) in die Arbeitslosenversicherung dem Unternehmen erstatten? eingezahlt. Die Person bekommt eine vierwöchige (sechs- monatige, zweijährige) Weiterbildung aus Mitteln der Arbe- itslosenversicherung finanziert. Text A3: Original German version of the worker Wie gerecht ist es aus Ihrer Sicht, dass die Arbeitsagen- vignette and its introduction tur die Weiterbildung aus Mitteln der Arbeitslosenversi- Im Folgenden werden vier andere Situationen beschrie- cherung bezahlt? ben, in denen Arbeitslose oder Beschäftigte vorkommen, See Tables 5, 6. die sich weiterbilden möchten. Bitte entscheiden Sie, wie gerecht oder ungerecht Sie es finden, dass die jeweilige Person eine Weiterbildung Table 5 Sample composition for respondent characteristics Variables for all items in a category add up to 100% (excl. rounding error) Percent (%) Male 53.9 Female 46.1 Age 18–34 (18 is minimum) 22.3 Age 35–49 28.7 Age 50–64 44.1 Age 65 + (78 is the realized survey maximum) 5.0 ‘Eastern States’: Brandenburg, Mecklenburg-West Pomerania, Saxony, Saxony-Anhalt, Thuringia 15.2 ‘Southern States’: Baden-Württemberg, Bavaria, Rhineland-Palatinate, Saarland 38.7 ‘Northern States’: Hesse, North Rhine-Westphalia, Schleswig–Holstein, Lower Saxony 36.8 ‘City States’: Berlin, Bremen, Hamburg 9.2 No education 1.8 A-levels and/or vocational training 54.8 ( Technical) College 43.5 Monthly net household income of less than 1000 Euro 4.4 Monthly net household income between 1000 and 1499 Euro 6.6 Monthly net household income between 1500 and 1999 Euro 9.2 Monthly net household income between 2000 and 2999 Euro 22.2 Monthly net household income between 3000 and 3999 Euro 20.7 Monthly net household income between 4000 and 4999 Euro 15.0 Monthly net household income more than 5000 euro 15.1 Monthly net household income not specified 6.7 CDU (Christian Democratic Union) 10.6 CSU (Christian Social Union) 3.2 Justice perceptions of occupational training subsidies: findings from a factorial survey Page 15 of 18 6 Table 5 (continued) Variables for all items in a category add up to 100% (excl. rounding error) Percent (%) SPD (Social Democratic Party) 9.7 AfD (Alternative for Germany) 3.6 FDP (Free Democratic Party) 4.9 Die Linke ( The Left) 7.7 Bündnis 90/Die Grünen (‘Greens’) 25.4 Other party not represented in parliament 3.8 Political factor not specified (no party affiliation, apolitical, no political affiliation) 31.2 Ever in life registered for unemployment benefits 61.8 Not ever in life registered for unemployment benefits 38.2 Agree with statement: ultimately, unemployed individuals are responsible for their situation 9.1 Agree with statement: ultimately, unemployed individuals are not responsible for their situation 90.9 Agree with statement: I will have difficulty handling new technology at work 11.2 Agree with statement: I will not have difficulty handling new technology at work 69.6 Statement not applicable due to not working 19.2 1010 persons Table 6 Worker vignette: training subsidy is perceived as just—odds ratio Fixed effects (FE) odds ratio Model 1a Model 1b Model 2 Model 3 1. Vignette features Logit conditional FE Logit mixed + RI Logit mixed + RI Logit mixed + RI/RS Male (else: female) 0.757 (0.097) 0.779 (0.085) 0.764* (0.084) 0.637* (0.120) Age (ref: 34 years old) 46 years old 1.431* (0.219) 1.514*** (0.198) 1.556*** (0.204) 1.943** (0.440) 58 years old 0.850 (0.127) 0.827 (0.109) 0.823 (0.108) 0.614* (0.138) Unemployed (else: no) 1.938*** (0.246) 2.168*** (0.253) 2.228*** (0.258) 4.747*** (1.543) Funding time (ref: 4 weeks) Funding for 6 months 0.850 (0.137) 0.748* (0.108) 0.756* (0.108) 0.590* (0.149) Funding for 2 years 0.405*** (0.062) 0.410*** (0.054) 0.418*** (0.055) 0.198*** (0.053) Job to 75% replaceable (else: no) 2.036*** (0.282) 1.985*** (0.254) 1.941*** (0.245) 2.341* (0.874) Continuous job (else: no) 3.407*** (0.432) 3.226*** (0.409) 3.321*** (0.422) 10.486*** (4.703) 2. Personal characteristics Male (else: female) 1.472* (0.227) 2.142* (0.623) Age (ref: 50–64) Age 18–34 1.165 (0.234) 1.636 (0.596) Age 35–49 1.115 (0.197) 1.198 (0.377) Age 65–78 0.757 (0.312) 0.505 (0.370) Region (ref: Eastern States) German Southern States 0.954 (0.223) 0.907 (0.403) German Northern States 0.956 (0.223) 0.880 (0.371) German City States 1.309 (0.448) 1.373 (0.826) Education (ref: A-levels & vocational training) No education 0.644 (0.369) 0.336 (0.344) ( Technical) College 1.262 (0.202) 1.546 (0.461) Monthly net household Euro income (ref: 2000–3000) Less than 1,000 Euro 1.020 (0.457) 1.201 (0.932) 6 Page 16 of 18 R. V. Wolff et al. Table 6 (continued) Fixed effects (FE) odds ratio Model 1a Model 1b Model 2 Model 3 1000–1499 Euro 1.476 (0.598) 1.841 (1.325) 1500–1999 Euro 1.194 (0.340) 1.493 (0.745) 3000–3999 Euro 0.826 (0.186) 0.744 (0.298) 4000–4999 Euro 0.731 (0.181) 0.647 (0.284) 5000 Euro or more 0.504* (0.127) 0.358* (0.163) Income not specified 0.734 (0.241) 0.607 (0.357) Party affiliation (ref: ’Greens’) CDU (Christian Democratic. Union) 0.573* (0.158) 0.327* (0.169) CSU (Christian Social Union) 0.629 (0.271) 0.385 (0.304) SPD (Social Democratic Party) 1.400 (0.388) 1.487 (0.750) AfD (Alternative for Germany) 0.655 (0.288) 0.323 (0.260) FDP (Free Democratic Party) 0.322*** (0.103) 0.124*** (0.073) Die Linke ( The Left) 1.133 (0.374) 0.924 (0.548) Other party (not in parliament) 2.945* (1.422) 5.133 (4.398) Not specified 0.956 (0.186) 0.824 (0.295) Unemployment benefit (else: no) 2.152* (0.618) 4.770** (2.468) Unemployed responsible (else: no) 1.403* (0.222) 2.102* (0.621) Difficulty with new work technologies (ref: no) Difficulty with new tech 1.264 (0.296) 1.386 (0.569) Not applicable (no work) 1.287 (0.294) 1.658 (0.664) Constant (not odds ratio) 2.235*** (0.183) 1.786*** (0.357) 3.483*** (0.706) Random effects (RE) coefficient Var (constant) 2.886*** (0.386) 2.423*** (0.333) 13,860*** (3.989) Var (unemployed ‘UE’) 5.266* (2.198) Var ( job replaceable) 14.123*** (3.658) Var (continuous job) 11.619*** (3.301) Cov (constant, unemployed) -2.721 (1.998) Cov (constant, job replaceable) − 4.646* (2.261) Cov (constant, continuous job) − 3.912* (1.952) Cov (UE, job replaceable) − 4.002* (1.496) Cov (UE, continuous job) 4.770*** (1.427) Cov ( job replaceable, Cont. job) − 5.459** (1.752) Model fit criteria Information criteria: AIC; BIC AIC:1040; BIC:1083 AIC:3369; BIC:3432 AIC:3342; BIC:3582 AIC:3207; BIC:3504 R (FE) MF; McKelvey and Zavoina 0.203 (McFadden) 0.202 (MK&Z) 0.284 (MK&Z) 0.5691 (MK&Z) R (FE + RE) McKelvey and Zavoina N/A 0.406 (MK&Z) 0.423 (MK&Z) 0.7807 (MK&Z) Intraclass correlation (ICC) N/A 0.395 Odds ratios exclude constant and random effects coefficients; robust standard errors in parentheses; secular AIC decrease in mixed 1692 (logit conditional FE) or 4040 (logit mixed) vignette answers for 1010 persons *p < 0.050; **p < 0.005; ***p < 0.001 Acknowledgements Funding We are grateful to the Support network for interdisciplinary social policy The project is funded as an FIS project (00.00117.19) of the German Federal research (FIS) and the German Federal Ministry of Labor and Social Affairs for Ministry of Labor and Social Affairs. The German Federal Ministry of Labor and funding the research project (Grant Number: 00.00117.19) and to the Data Social Affairs had no role in the design of the study and collection, analysis, and IT Management of IAB, in particular Lina-Jeanette Metzger, for support in and interpretation of data. conducting the survey. Availability of data and materials Author contributions As the data underlying our analysis are not completely proprietary, access to The cited authors contributed insights in all parts of the paper. All authors the data is restricted. The data we use are social data. They contain sensitive read and approved the final manuscript. information and are subject to confidentiality regulations. Obtaining access Justice perceptions of occupational training subsidies: findings from a factorial survey Page 17 of 18 6 to the data through the research data center of the Institute for Employ- Sozialwissenschaften, Wiesbaden (2008). https:// doi. org/ 10. 1007/ 978-3- ment Research (IAB) requires a contract with IAB. We will support researchers 531- 90929-5_ 12 interested in replicating the results with the required formalities to receive German Bundestag: Weiterentwicklung der Arbeitslosenversicherung zu einer data access. Arbeitsversicherung. Ideen und Konzepte. Wissenschaftliche Dienste 6 – 3000 – 078/18. Berlin (2018). https:// www. bunde stag. de/ resou rce/ blob/ 580578/ 0ba6a e70b4 738cd b0710 3e62d 8bd5d f7/ WD-6- 078- 18- pdf- data. Declarations pdf Gilliland, S.W.: The perceived fairness of selection systems. An organizational Ethics approval and consent to participate justice perspective. Acad. Manag. Re. 18(4), 694–734 (1993). https:// doi. Not applicable. org/ 10. 5465/ amr. 1993. 94022 10155 Greenberg, J.: Organizational justice: yesterday, today, and tomorrow. J. Manag. Consent for publication 16(2), 399–432 (1990). https:// doi. org/ 10. 1177/ 01492 06390 01600 208 Not applicable. Hans, J.P., Hofmann, S., Sesselmeier, W., Yollu-Tok, A.: Umsetzung, Kosten und Wirkungen einer Arbeitsversicherung. Friedrich-Ebert-Stiftung, Bonn Competing interests (2017) The authors declare that they have no competing interests. Kaufmann, F.-X.: Herausforderungen des Sozialstaates. Suhrkamp, Frankfurt a. M (1997a) Author details Kaufmann, F.-X.: Schwindet die integrative Funktion des Sozialstaates? Berl. J. University of Bamberg: Otto-Friedrich-Universitat Bamberg, Bamberg, Bavaria, Soziol. 7(1), 5–19 (1997b) Germany. Institute for Employment Research (IAB): The Research Institute Klaus, A., Kruppe, T., Lang, J., Roesler, K.: Geförderte Weiterbildung Beschäft- of the Federal Employment Agency, Nürnberg, Bavaria, Germany. igter: Trotz erweiterter Möglichkeiten noch ausbaufähig. IAB-Kurzbericht 24/2020 (2020) Received: 1 June 2021 Accepted: 2 May 2022 Kluegel, R., Mason, D.S., Wegener, B.: The legitimation of capitalism in the post- communist tradition. Public opinion about market justice, 1991–1996. Eur. Sociol. Rev. 15(3), 251–283 (1999). https:// doi. org/ 10. 1093/ oxfor djour nals. esr. a0182 63 Konow, J.: A positive theory of economic fairness. J. Econ. Behav. Organ. 31(1), References 13–35 (1996). https:// doi. org/ 10. 1016/ S0167- 2681(96) 00862- 1 AAPOR ( The American Association for Public Opinion Research).: Standard Konow, J.: Fair and square. The four sides of distributive justice. J. Econ. Behav. definitions: final dispositions of case codes and outcome rates for surveys Organ. 46(2), 137–164 (2001). https:// doi. org/ 10. 1016/ S0167- 2681(01) 9 (2016) 00194- 9 Acemoglu, D., Restrepo, P.: Robots and Jobs: evidence from US Labor Markets. Kruppe, T., Lang, J., Stephan, G.: Das Qualifizierungschancengesetz: Mögli- NBER Working Paper No. 23285 (2017) chkeiten und Grenzen einer Evaluation. Arbeitspapier vom 15.4.2020. Adams, J.S.: Inequity in social exchange. In: Berkowitz, L. (ed.) Advances in Institut für Arbeitsmarkt- und Berufsforschung, Nürnberg (2020) Experimental Social Psychology, vol. 2, pp. 267–299. Academic Press, New Langer, W.: How to assess the fit of multi-level-logit models with Stata? In: York (1965). https:// doi. org/ 10. 1016/ S0065- 2601(08) 60108- 2 Meeting of the German Stata User Group at the Humboldt University Allaart, P., Bellmann, L., Leber, U.: Company-provided further training in Berlin. (2017). https:// langer. sozio logie. uni- halle. de/ stata/ pdf/ Langer- Ger- Germany and the Netherlands. Empir. Res. Vocat. Educ. Train. 1, 103–121 man- Stata- Users- Group- Meeti ng- 2017. pdf (2009). https:// doi. org/ 10. 1007/ BF035 46482 Leisering, L.: Eine Frage der Gerechtigkeit. Armut und Reichtum in Auspurg, K., Hinz, T.: Factorial survey experiments. SAGE Publications, Los Deutschland. In: Aus Politik und Zeitgeschichte (Beilage zur Wochenzei- Angeles (2015) tung Das Parlament) 30.4.1999, pp. 10–17 (1999) Bassanini, A., Booth, A., Brunello, G., Paola, M.D., Leuven, E.: Workplace training Leisering, L.: Paradigmen sozialer Gerechtigkeit. In: Liebig, S., Lengfeld, H., Mau, in Europe, IZA discussion paper no. 1640 (2005) S. (eds.) Verteilungsprobleme und Gerechtigkeit in modernen Gesells- Bell, A., Fairbrother, M., Jones, K.: Fixed and random effects models: making an chaften, pp. 29–68. Campus, Frankfurt a. M (2004) informed choice. Qual. Quant. 53, 1051–1074 (2019). https:// doi. org/ 10. Mau, S.: Welfare regimes and the norms of social exchange. Curr. Sociol. 52(1), 1007/ s11135- 018- 0802- x 53–74 (2004). https:// doi. org/ 10. 1177/ 00113 92104 03931 4 Benjamin, D.J., Berger, J.O., Johannesson, M., Nosek, B.A., Wagenmakers, Mikula, G.: Gerecht und ungerecht: Eine Skizze der sozialpsychologischen E.-J., Berk, R., Bollen, K.A., Brembs, B., Brown, L., Camerer, C., Cesarini, D., Gerechtigkeitsforschung. In: Held, M., Kubon-Gilke, G., Sturn, R. (eds.) Chambers, C.D., Clyde, M., Cook, T.D., Boeck, P.D., Dienes, Z., Dreber, A., Normative und institutionelle Grundfragen der Ökonomik. Jahrbuch 1: Easwaran, K., Eerson, C., F ff ehr, E., Fidler, F., Field, A.P., Forster, M., George, Gerechtigkeit als Voraussetzung für effizientes Wirtschaften, pp. 257–278. E.I., Gonzalez, R., Goodman, S., Green, E., Green, D.P., Greenwald, A.G., Had- Marburg (2002) field, J.D., Hedges, L.V., Held, L., Ho, T.H., Hoijtink, H., Hruschka, D.J., Imai, K., Miller, D.L.: Needs-based justice. Theory and evidence. In: Bauer, A.M., Meyer- Imbens, G., Ioannidis, J.P.A., Jeon, M., Jones, J.H., Kirchler, M., Laibson, D., huber, M.I. (eds.) Empirical Research and Normative Theory: Transdiscipli- List, J., Little, R., Lupia, A., Machery, E., Maxwell, S.E., McCarthy, M., Moore, nary Perspectives on Two Methodical Traditions between Separation and D.A., Morgan, S.L., Munafó, M., Nakagawa, S., Nyhan, B., Parker, T.H., Peric- Interdependence, pp. 273–294. De Gruyter, Berlin (2020) chi, L., Perugini, M., Rouder, J., Rousseau, J., Savalei, V., Schönbrodt, F.D., Morel, N., Palier, B., Palme, J.: Towards a Social Investment Welfare State? Ideas, Sellke, T., Sinclair, B., Tingley, D., Zandt, T.V., Vazire, S., Watts, D.J., Winship, C., Policies and Challenges. Policy Press, Bristol (2012) Wolpert, R.L., Xie, Y., Young, C., Zinman, J., Johnson, V.E.: Redefine statistical Osiander, C., Senghaas, M., Stephan, G., Struck, O.; Wolff, R.V.: Acceptance significance. Nat. Hum. Behav. 2, 6–10 (2018). https:// doi. org/ 10. 1038/ of social- and labor market programs and regulations: Methodologi- s41562- 017- 0189- z cal report on the first survey wave. In: IAB-Forschungsbericht 07/2020, Bowles, S., Gintis, H.: Reciprocity, self-interest, and the welfare state. Nordic J. Nürnberg (2020) Polit. Econ. 26, 33–53 (2000) Roosma, F., Gelissen, J., van Oorschot, W.: The multidimensionality of welfare Brynjolfsson, E., McAfee, A.: The Second Machine Age: Work, Progress, and state attitudes: a European cross-national study. Soc. Indic. Res. 113(1), Prosperity in a Time of Brilliant Technologies. Norton and Company, 235–255 (2013). https:// doi. org/ 10. 1007/ s11205- 012- 0099- 4 London (2014) Rossi, P.H., Anderson, A.B.: An introduction. In: Rossi, P.H., Nock, S.L. (eds.) Meas- Esping-Andersen, G., Gallie, D., Hemerijck, A., Myles, J.: Why We Need a New uring Social Judgments: The Factorial Survey Approach, pp. 15–67. Sage Welfare State. Oxford University Press, Oxford (2002) Publications, Beverly Hills (1982) Evers, A.: Investiv und aktivierend oder ökonomistisch und bevormundend? Rothstein, B.: Just Institutions Matter: The Moral and Political Logic of the Uni- Zur Auseinandersetzung mit einer neuen Generation von Sozialpolitiken. versal Welfare State. Cambridge University Press, Cambridge (1998) In: Evers, A., Heinze, R.G. (eds.) Sozialpolitik, pp. 229–249. VS Verlag für 6 Page 18 of 18 R. V. Wolff et al. Sachweh, P.: Social justice and the welfare state: institutions, outcomes, and attitudes in comparative perspective. In: Sabbagh, C., Schmitt, M. (eds.) Handbook of Social Justice Theory and Research, pp. 293–313. Springer Nature, New York (2016) Sachweh, P., Burkhardt, C., Mau, S.: Wandel und Reform des deutschen Sozialstaatsaus Sicht der Bevölkerung. WSI-Mitteilungen 62(11), 612–618 (2009). https:// doi. org/ 10. 5771/ 0342- 300X- 2009- 11 Snijders, T., Bosker, R.: Multilevel Analysis: An Introduction to Basic and Advanced Multilevel Modeling, pp. 102–103. Sage Publications, London (1999) Struck, O.: Flexibilität und Sicherheit. Empirische Befunde, theoretische Konzepte und politische Gestaltung (in-)stabiler Beschäftigung. VS-Verlag, Wiesbaden (2006). https:// doi. org/ 10. 1007/ 978-3- 531- 90389- 7 van den Berg, G.J., Dauth, C., Homrighausen, P., Stephan, G.: Informing employees in small and medium sized firms about training—results of a randomized field experiment, IZA discussion paper no. 11963 (2018) van Oorschot, W., Roosma, F., Meuleman, B., Reeskens, T. (eds.): The Social Legitimacy of Targeted Welfare: Attitudes to Welfare Deservingness. Edward Elgar Publishing, Cheltenham (2017) Vobruba, G.: Die Faktizität der Geltung: Gerechtigkeit im sozialpolitischen Umbaudiskurs. In: Clausen, L. (ed.) Gesellschaften im Umbruch. Verhand- lungen des 27. Kongresses der Deutschen Gesellschaft für Soziologie in Halle an der Saale 1995, pp. 963–975. Campus, Frankfurt a. M (1996) Wasserstein, R.L., Schirm, A.L., Lazar, N.A.: Moving to a world Beyond “p<0.05.” Am. Stat. 73(1), 1–19 (2019). https:// doi. org/ 10. 1080/ 00031 305. 2019. 15839 13 Young, H.P.: Equity. In: Theory and Practice. Princeton University Press, Prince- ton (1993) Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in pub- lished maps and institutional affiliations.

Journal

Journal for Labour Market ResearchSpringer Journals

Published: Dec 1, 2022

Keywords: Occupational training; Unemployment insurance; Justice assessments; Factorial survey; Multilevel and mixed effects model; C99; D63; I38; J08; J24; J65

There are no references for this article.