Varieties of Capitalism and labour market opportunities for the youth

In this study, we examine the extent to which socio-economic institutions shape young people’s perceptions of labour market opportunity structures and their employment attitudes (i.e. skills and retraining). Building on the varieties of capitalism approach, we expect young people (aged 18–35) in coordinated market economies (CMEs) with encompassing welfare states to regard firmand industry-specific skills as more important than their peers in liberal market economies (LMEs). To assess this proposition, we draw on original survey data and compare young people’s employment attitudes in five European countries: the United Kingdom (UK), which represents a typical liberal market economy, and Austria, Denmark, Germany and Switzerland as representatives of coordinated market economies. To what extent do different training regimes in CMEs and LMEs shape individual attitudes towards skill formation? The empirical analysis shows that young people’s attitudes with regard to the specificity of skills and the willingness to undertake retraining differ systematically between CME and LME countries and supports our argument that the specific socio-economic institutions matter.


Introduction
The varieties of capitalism (VoC) literature as established by Hall and Soskice (2001) stresses that different training systems and skill formation structures are key distinguishing characteristics of discrete economic systems (see also Estevez-Abe et al., 2001;Amable, 2003;Busemeyer, 2009;Busemeyer andTrampusch, 2011, 2012). This literature differentiates between liberal market economies (LMEs), such as the United States and the United Kingdom (UK), and coordinated market economies (CMEs), such as Germany and Austria. These two types of capitalism differ according to their training systems, which in turn condition a range of political choices (Culpepper, 2007, 611). CMEs tend to stress firm-based training, and the skills developed are at a much higher level of specialisation than the comparatively more general skills inculcated in LMEs. In CMEs, the existence of unemployment insurance and high levels of replacement rates for unemployment make it more attractive for employees to invest in firm-and industry-specific skills (Rhodes, 2005;Schröder, 2013, 70-74). Hence this fundamental difference. VoC is a firm-level approach, which at first glance does not seem to lend itself to the microlevel analysis of individuals' attitudes and perceptions (but see, e.g., Busemeyer and Jensen, 2012). However, the institutional contexts in which individuals are embedded is central to both their perception of risks and opportunities and plays a key role in shaping their attitudes (e.g. Vráblíková, 2014;Walter, 2017). Accordingly, we pose the following question: To what extent do different training regimes in CMEs and LMEs shape individual attitudes towards skill formation?
Educational decisions and skill formation are processes that mainly take place during adolescence and early adulthood. As Estévez-Abe et al. (2001, 145) argue, 'young people are less likely to invest in specific skills if the risk of loss of employment opportunities that require those specific skills is high'. Accordingly, to answer our research question, we draw on an original and bespoke survey dataset explicitly designed to capture attitudes and work values of young people (18-35 years) in Europe gathered in the context of the large-scale collaborative CUPESSE project (Tosun et al., 2019). We compare the attitudes of young people towards skill formation in the UK, an archetypical liberal market economy, with their counterparts in four coordinated market economies: Austria, Denmark, Germany and Switzerland.
In the remainder of this article, we outline our main theoretical argument, methodological approach and research design. We then report and interpret our empirical results, which reveal that attitudes towards skill formation differ systematically between the UK and the four CMEsmeaning they align with our theoretical argument. Lastly, we offer some concluding remarks.

Theoretical Reasoning and Hypotheses
In its most basic form, the VoC approach claims that the success of a political economy depends on firms and companies solving their coordination problems in various economic subsystems such as industrial relations or education and vocational training (Hall and Soskice 2001).
Following this main line of reasoning, LMEs and CMEs represent the extremes of a spectrum between pure market powers and the strong coordination of market forces. It is also worth noting that more recent contributions to the VoC literature have identified mixed-market economies (Molina and Rhodes, 2007) and dependent market economies (Nölke and Vliegenthart, 2009).
As their respective names suggest, in LMEs strong market forces dominate, while in CMEs coordination mechanisms between the various market participants play a major role. In LMEs, companies coordinate their activities in all subsystems of the economy primarily by focusing on price signals, free competition in the market and marginal cost. In CMEs, non-market based relations play a crucial role in the development of corporate core competencies. Hall and Soskice (2001) find that CMEs produce goods that stem from incrementally innovative production, which requires especially industry-and firm-specific human capital, more efficiently than LMEs. One of the reasons for this is the nature of intercompany relations giving rise to joint research, product differentiation and niche production, rather than direct product competition.
Another reason is based on the human capital argument. In CMEs, employees are provided with sophisticated training schemes and apprenticeships that rely on sufficiently specialised knowledge to propose and implement changes in the production line (see, e.g., Estevez-Abe et al., 2001;Busemeyer, 2009;Trampusch, 2010;Busemeyer and Trampusch, 2012). This requires a 'mix of company-specific and more general technical skills' in the workforce (Hall and Soskice 2001, 36). Furthermore, the likelihood of becoming unemployed is lower in CMEs, because firms are more diversified, which makes investments in firm-specific skills more attractive. The differences between CMEs and LMEs has a direct effect on the policy-making process and ultimately on policy design (see, Estevez-Abe, 2005;Bonoli and Reber, 2010;Jensen, 2011;Hörisch and Weber, 2014;Marques and Hörisch, 2019).
Several contributions to the political economy literature claim that the VoC approach 'brought the question of skill formation from the periphery to the centre of comparative welfare state research and political economy' (Busemeyer and Trampusch, 2011, 424;Dobbins and Busemeyer, 2015). Regarding the systems of skill creation, there are several reasons to assume that different institutional contexts in varying types of capitalism shape individuals' attitudes towards skill formation. Along these lines, we argue that the existing socioeconomic institutions shape the attitudes of young people in two ways. First, they alter the strategic behaviour of young people by heightening their (rational) perception of risk; and second, they encourage normative adaption to the existing institutions through processes of socialisation.
The skill creation systems of LMEs and CMEs generate different types of skills. Substantial parts of these differences stem from the rational adaptation of variations in welfare state arrangements by (young) adults. For example, Estévez-Abe argues that 'welfare states … affect the decisions of workers and employers about the skills in which to invest ' (2005, 191; see also Estévez-Abe et al., 2001). In CMEs, employers invest in the formation of industry-and firmspecific skills 1 by means of strong vocational training and education systems, while enjoying the support of generous welfare state policies that incentivise (young) employees to invest in those specific skills (Estévez-Abe et al., 2001;Iversen and Soskice, 2001;Mares 2003;Busemeyer and Trampusch, 2011;Schröder, 2013;Busemeyer and Schlicht-Schmälzle, 2014). Furthermore, the higher degree of diversification of firms between different sectors in CMEs makes it more attractive for (young) adults to invest in firm-specific skills, as the risk of becoming unemployed is lower in case of sector-specific downturns. In contrast, the more competitive environment of LMEs severely disincentivises skill investments by the firms. Instead, skill formation primarily takes place in formalised education institutions as well as on-the-job training, resulting in general rather than firm-specific skills (Busemeyer and Trampusch, 2011;Coulter, 2018). This institutional arrangement might go hand-in-hand with the skill formation preferences rationally developed by (young) adults.
1 Several skill typologies can be found in the literature. For example, Estévez-Abe et al. (2001) distinguish between firmspecific skills, industry-specific skills and general skills. Even though some strands of the literature criticise the differentiation between general and specific skills as too broad categories (Streeck 2011), we rely on the difference between general and specific skills here, subsuming firm-specific and industry-specific skills into the latter category, because we are interested in general differences as well as broad patterns between varieties of capitalism and skill formation preferences.
Besides these more rational adaptation procedures, socialisation processes and social norms also play a major role in the development of attitudes towards skill formation. Political economies are socially constructed entities comprising both formalised structures (such as rules) and informal codes and expectations (see Craig, 2015). The process that leads to the shared understanding of formal and informal constructions in political economies is socialisation (see Mendoza, 2007). We argue that socialisation processes in political economies shape and structure individuals' attitudes towards skill formation and will engender attitudinal differences in different varieties of capitalism, while controlling for an extensive battery of alternative explanations. This argument aligns with newer contributions to the literature on political economy that demonstrate how macro-level phenomena affect individual-level perceptions (e.g. Walter, 2017).
Herein we argue that young people will have clear attitudinal positions on skill formation for two main reasons. First, they are either currently in education and training or have relatively recently made decisions concerning education and training programmes. As a result, issues around education, training and skill formation are at the forefront of their decision-making in this area. Second, in many European countries, young people do not experience linear transitions into the labour market (Côté, 2014). Their experience is more likely to be elongated by delays, difficulties and periods of unemployment that make the issue of skill formation salient for them.
Given our theoretical proposition that the institutional context -i.e. growing up and living in an LME or a CME -is crucial to the pattern of attitudes and beliefs, we expect young people socialised in these different milieus to have varying views.
Hypothesis 1: Young people in LMEs will consider a general education to be more important than job-specific skills compared to their counterparts in CMEs.
Firms in LMEs do not require the same type of specialist skills as CMEs and tend to oppose initiatives to introduce far-reaching insurance mechanisms. The consequence is that there are a less-specialist training and a more flexible workforce in LME production lines. This enables companies to produce their goods relatively cheaply and to innovate quickly through the firing of current and hiring of new employees. Along this line, Streeck (2011) argued that it is important to note that general skills do not go hand-in-hand with high skills, and specific skills do not automatically accompany low skills. From his viewpoint, the difference between general and specific skills relates to labour market mobility and the connectivity to the firm. The latter is usually substantially higher in CMEs because of their strong employment protection, codetermination rules and collective bargaining agreements. This also leads workers to rationally develop more general skill sets that are transferable between sectors and companies to remain employable and competitive in the labour market. As Lloyd and Payne (2002) contend, in LMEs such as the United Kingdom, '… a viable high skills project in the UK is fundamentally incompatible with the model of British capitalism as currently configured' (see also Coulter, 2018). Furthermore, Finegold and Soskice (1988) have shown that the failure of several attempts by British governments to establish a dual apprenticeship training system can be explained by the fundamental differences in the socioeconomic institutional framework, especially when compared to CMEs such as Germany.
Accordingly, we include a second question in our analysis, assessing the extent to which young people would be willing to retrain to get a new job. We expect strategic behaviour as well as the socialisation processes in different types of capitalism to affect young people's willingness to consider retraining. In particular, we argue that long-term employment and the focus on industryspecific skills that is typical of CMEs provide young people with a stronger incentive to learn new skills that are also beneficial to their current occupation, while at the same time disincentivising them from retraining completely. In contrast, the fluid labour market and the focus on general skills we observe in LMEs should make retraining more attractive: Hypothesis 2: Young people in LMEs are more willing to retrain compared to young people in CMEs.
While the overall differences between LMEs and CMEs concerning skill formation are clear from a theoretical viewpoint, the ordering within the group of CMEs is less distinct. Within the VoC literature there is an ongoing debate related to the hybridisation of CMEs that sees several CMEs moving closer to LMEs. This process of hybridisation has been discussed in the German (e.g. Lütz and Eberle, 2008), Danish (Campbell and Pedersen, 2007;Dobbins and Busemeyer, 2015), Swiss (Armingeon et al., 2004;Börsch, 2007;Afonso and Mach, 2011) and Austrian contexts (Afonso and Mach, 2011;Schröder, 2013). Table 1 provides an overview of the VoC coordination index scores for the five countries included in our analyses based on four well-established VoC coordination indices. The four indices vary across time (1998 to 2016), and while there are slight differences, the overall pattern is similar. According to all four indices, Austria is the most coordinated economy, followed by Germany and Denmark and then Switzerland. Our selection of CMEs is close to optimal, since all four selected CMEs rely on collective skill formation systems (cf. Busemeyer and Trampusch, 2012) and there are no ambiguous cases -e.g. the Swedish statist skill formation system. In contrast, the UK is the closest to the ideal-type LME with an exceptionally low degree of coordination. In our analyses, we will assess the extent to which these coordination differences are reflected in the skill formation attitudes of the young adults in our sample.

Clarifications on Data and Methods
In this study, we draw on a unique dataset that was produced in the CUPESSE project (Tosun et al., 2019 can be characterised as a social democratic coordinated one (Schröder, 2013).
The country surveys were conducted online with the assistance of leading commercial polling companies. Consistent sampling frames were used and the survey companies provided probability samples of individuals between 18 and 35 years old, giving their employment status (e.g. employed; self-employed; unemployed; in education/training), NUTS-2 region, age, education, and migration background/minority group membership. The dataset consists of 1,648 respondents in Austria, 1,142 in Denmark, 3,279 in Germany, 1,002 in Switzerland and 3,004 in the UK. The variation in the number of individuals surveyed was mostly due to population size, though practical issues related to survey costs and the different procedures followed by the respective polling firms also played a role. The data were collected throughout 2016.
The CUPESSE survey contains two questions that are central to our analysis. The first one addresses respondents' attitudinal positions on the importance of a good general education, as opposed to job-specific skills. The second and broader question asks respondents about their willingness to learn completely new skills or to retrain to secure a new and different job, as opposed to learning new skills that may help secure a job in their current industry. These questions investigate differences related to the skill sets that firms tend to support in LMEs and CMEs. In LMEs, firms usually focus on general skills that are transferable between sectors and companies, i.e. a good general education, whereas in CMEs the focus is on occupational or sector-specific skills. In addition, they address differences in the employment conditions of LMEs and CMEs: employment in LMEs is more fluid, making retraining and general skills more attractive, while employment in CMEs is more long-term, increasing the benefits of occupationspecific skills and of learning new skills that will also be beneficial to the current occupation. The exact wording of the question and the corresponding response categories are: b) Occupation-specific or job-specific skills.
Question 2: What changes would you be willing to make to get a new job? a) I would be willing to learn new skills such as a new language, computer programmes.
b) I would be willing to learn completely new skills or retrain to get a job.
Response: 1 (No), 2 (Maybe), 3 (Yes). Numbers only include respondents for which all control variables were available: a) 1: less than lower secondary; 2: lower secondary; 3: lower tier upper secondary; 4: upper tier upper secondary; 5: advanced vocational; 6: lower tertiary; 7: higher tertiary; b) 0: female; 1: male; c) measured on a left-right scale from 0 (left) to 10 (right); d) personal monthly total net income measured in income groups from 1 (lowest) to 10 (highest); e) dummy for whether the respondent has ever had a job for one year or more; f) dummy for whether the respondent has ever been unemployed for a period longer than six months; g) dummy for whether the respondent or their parents were born in a different country; h) 1: independent income; 2: partially dependent income; 3: dependent income; i) due to missings, the number of respondents in Switzerland is only 582 for the comparison of skill specificity and 575 for the comparison of willingness to retrain.
We test our hypotheses through seven multinomial regression models: a base model with only the country dummies, five-country models with control variables and a full model with both the country dummies and control variables. Control variables include the age and sex of the respondent, their employment status, level of educational attainment, income, political affiliation, previous employment and unemployment experiences, migration background, and income dependency (see, e.g., Baartman and De Bruijn, 2011;Chen, 2011). The summary statistics of the dependent and independent variables are shown in Table 2.

Empirical Findings
We begin the discussion of the empirical findings with the question on skill specificity. As we see in Table 3, 95 per cent of our respondents consider a good general education to be either rather or very important -the same goes for job-specific skills. The response patterns show that there is indicative support for Hypothesis 1, because the percentage of respondents in CMEs who consider job-specific skills as important is higher than in the UK (96.2 per cent vs. 94.4 per cent; the difference is significant at the 1 per cent level). At the same time, more respondents in the UK consider a good general education to be important (96.0 per cent vs 95.3 per cent; however, the difference is not significant). Since the response patterns vary across these two items and are unevenly distributed, we directly compare the responses to the two questions. We do this by examining whether the respondents consider a good general education and job-specific skills to be equally important to finding a good job or whether they prioritise one over the other. This approach aligns with the empirical strategy adopted by Busemeyer et al. (2017) in their study on attitudes towards education policy. The authors argue that surveys on preferences regarding education policies and skills that do not force participants to prioritise the relative importance of the individual education goals and skills are likely to blur general attitudes and specific preferences. When not forcing respondents to choose, surveys are likely to capture 'cheap talk' rather than an expression of preferences (Busemeyer et al., 2017, 38). From this perspective, the survey approach we followed in the CUPESSE project corresponds to the latest developments in the relevant literature on survey methodology.
Our dependent variable includes the following possible response categories: 'a good general education is more important than job-specific skills to find a job', 'job-specific skills are more important than a good general education to find a job', and 'both a good general education and job-specific skills are equally important to find a job (ties)'. Of the respondents who rated one statement higher than the other, a majority considers job-specific skills to be more important than a good general education in Switzerland, Denmark and Austria. Although a majority considers a good general education more important in both the UK and Germany, the margin is substantially higher in the UK. The UK has the highest share of respondents judging a good education to be more important (28.9 per cent vs 15.0 per cent -24.3 per cent) and the lowest share of respondents judging job-specific skills to be more important (20.8 per cent vs 22.7 per cent -26.5 per cent). The overall pattern thus supports Hypothesis 1. Note. Odds ratios, robust standard errors in parenthesis, significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001.
Given the measurement level of the dependent variable, ordered logit regression would initially appear the most appropriate method. However, since the assumption of parallel slopes, on which ordered logit models are based, does not hold, this method is not suitable for the analyses we conduct. Therefore, we analyse the data via multinomial logistic models. The conclusions we arrive at, however, would be the same for ordered logit regression models. Table 4 presents the results of our multinomial logistic regression models. The coefficients are displayed as odd-ratios with robust standard errors. Table 4 shows the coefficients for comparing the base outcome (job-specific skills are more important than a good general education) with the two other outcomes (a good general education is more important than job-specific skills, and both are equally important). Our main interest is in the comparison between the perception of job-specific skills or a good general education being more important. Consequently, we omitted the comparison with both being equally important. The full table can be seen in Table 4 in the appendix. Both the baseline model and the full model show that the differences between the UK and the CMEs are significant at a 1 per cent or a 0.1 per cent significance level. In substantive terms, the predicted probabilities of considering a good general education more important are 29.0 per cent in the UK, but they range from 14.8 per cent (Austria) to 24.8 per cent (Denmark) in the four CMEs. In contrast, the predicted probabilities of considering  (Austria) is also the country in which respondents more strongly favour job-specific skills over a good general education (cf. Afonso and Mach, 2011). When turning to the control variables, we observe that significantly more men in the full model than women consider job-specific skills to be more important, while people with right-leaning ideological views consider a good general education to be more important than job-specific skills. Turning to our second dependent variable, the willingness to retrain, we observe that less than 4 per cent of respondents indicate that they are unwilling to learn new skills in all of the countries in our sample (see Table 5). The percentage of respondents who are unwilling to retrain varies between the countries but is consistently below 10 per cent. Looking at the data, there is again some indication that the theoretical argument holds. Compared to the CMEs, the willingness to learn new skills is lower in the UK (75.4 per cent vs. 81.7 per cent; the difference is significant at the 0.1 per cent-level), whereas the willingness to retrain is higher (65.6 per cent vs. 62.3 per cent; the difference is significant at the 5 per cent-level). In contrast to our expectation, however, the willingness to retrain is actually higher in Switzerland (75.8 per cent) than in the UK (65.6 per cent).
A chi-squared test indicates a significant difference between the LME and the CMEs at the 5%-level. Our data present us with an uneven distribution of responses, so we directly compare the responses of each individual as we did with skill specificity. Our response categories are as follows: 'more willing to learn new skills than to retrain to get a new job', 'more willing to retrain than to learn new skills to get a new job', and 'equally willing to both learn new skills and retrain (tied)'. The frequency of the tied response is markedly higher than it was for skill specificity.
Furthermore, there is also greater variance in the frequency of the tied response. While only 61.3 per cent of Danish respondents gave a tied answer, 74.8 per cent of the Swiss respondents are equally willing to learn new skills and retrain. In all countries, respondents are on average more willing to learn new skills than to retrain. Nevertheless, we can still observe patterns as predicted by the VoC approach. In the UK, the percentage of respondents more willing to retrain is higher than in the four CMEs (8.5 per cent vs 2.4 per cent -7.3 per cent). As for the willingness to learn new skills, respondents in the UK are less willing to learn new skills than respondents in three of the CMEs (Germany, Austria and Denmark; 19.1 per cent vs 25.7 per cent -36.3 per cent). The results for Switzerland, however, contradict expectations, as the percentage of Swiss respondents more willing to learn new skills is below that of the UK (17.9 per cent).
As with skill specificity, we conduct multinomial logistic regression models to test whether the difference is significant and whether it holds when adding control variables. Table 6 presents the results. Respondents who are more willing to learn new skills are compared to respondents who are more willing to retrain, as well as to respondents who are equally willing to do both. As before, we omitted those equally willing to do both because the former comparison is key to our theoretical implications. The proportion of respondents who are more willing to retrain is significantly higher in the UK than in Denmark, Germany and Austria, though not in Switzerland.
The predicted probability of being more willing to retrain is significantly higher in the UK (8.3 per cent) than in Denmark and Austria (2.2 per cent and 4.1 per cent), but only marginally and insignificantly lower in Switzerland (8.0 per cent) and Germany (7.1 per cent). At the same time, the predicted probability of being more willing to learn new skills is lower in Switzerland than in the UK (17.8 per cent vs. 19.3 per cent), but it is significantly higher in Austria, Germany and Denmark (25.8 per cent -36.2 per cent). As already indicated by our descriptive results, the regression models confirm our expectations for only three of the four CMEs. With respect to the control variables, respondents who are older, have lower educational attainment levels, are in employment or are ideologically right-leaning are more willing to retrain.
Following the hybridisation hypothesis, it is plausible that the country with the strongest degree of hybridisation, namely Switzerland, shows the weakest effects. Although it is typically classified as a CME, Switzerland is often seen as incorporating a mixture of elements from both liberal and coordinated market economies. As a less regulated labour market than other CMEs (Schröder, 2013, 124-126), Switzerland incentivises retraining and employment attitudes more closely follow the pattern of LMEs. Rather surprisingly, however, the share of respondents who are willing to retrain is actually lowest in Denmark, a country known for its flexicurity approach (Emmenegger, 2010). Despite its flexible labour market with low levels of job security regulations and pronounced active labour market policies, these institutions do not translate to corresponding individual attitudes. Table 6. Multinomial logistic models for the comparison of willingness to retrain vs. willingness to learn new skills.

Conclusions
VoC theory contends that LMEs and CMEs differ by the degree of coordination between firms and other actors. The different coordination mechanisms produce distinct advantages in innovation and affect policymaking. In this article we have sought to advance the VoC literature by illustrating how it can be combined with the microlevel, thereby offering insights into how individuals living and working in the different political economies internalise their structural characteristics and adjust their attitudes and arguably their behavioural patterns accordingly. We argued that the institutional differences in the structure of the labour market are reflected in young people's attitudes towards skill formation. Using the unique CUPESSE survey data, we compared young people's attitudinal positions on skills and retraining in Austria, Denmark, Germany and Switzerland, and the UK.
We found statistically significant differences between the respondents in the individual countries for both types of employment attitudes. Respondents in the UK considered general skills as opposed to job-specific skills to be more important than respondents in Austria, Denmark, Germany and Switzerland. Young British citizens are also more willing to retrain in order to get a new job rather than simply learn new skills when compared to all of the CMEs except for Switzerland. These findings are in line with our expectations of how the respective skill creation systems in political economies shape the employment attitudes of the young. Our findings are robust and are reinforced by the inclusion of various control variables. Finally, drawing on hybridisation arguments within the VoC literature, we were also able to make some sense of the ordering of the countries along the CME-LME spectrum. Nevertheless, some of the results for the within-CME variation are more mixed. Further research could thus investigate in greater depth the interplay between hybridisation tendencies and the development of attitudes towards skill formation.
The findings reported in this study are important for increasing the academic impact of political economy research, as they connect it to subdisciplines in political science ( Europe demonstrates that making poorly informed choices regarding education and training can lead to a mismatch between the qualifications of jobseekers (supply-side) and the qualifications employers seek (demand-side) (e.g. Pompei and Selezneva, 2017). In this respect, our study presents positive news: it shows that the attitudes of employees towards education, training and retraining generally align well with the institutional context they live in. LMEs and CMEs tend to reproduce themselves on the microlevel: institutional configurations that rely more heavily on general skills and a flexible workforce tend to produce employees that are more willing to learn general skills and undergo complete retraining -and vice versa. A good match between the institutional context and the attitudes of young people should improve their chances of finding a suitable job in the labour market.
We fully recognise that our research is open to challenge. Most importantly, it rests on the assumption that the differences we find in the employment attitudes of the younger generation in the five countries can be traced back to the different skill creation systems of the two varieties of capitalism. An alternative perspective might argue that the differences are not inherent to the varieties of capitalism approach, and there may be other explanatory factors at play. While we accept that other factors will have an influence on the outcomes, the response patterns within the countries we examined are consistent enough to suggest that they are the outcome of established differences between these countries (see Iversen and Stephens, 2008). From a theoretical viewpoint, it is plausible and credible to argue that the different skill creation regimes engender different employment attitudes towards the process of skill formation. Our empirical findings support this.
Further research could build on the results presented above by testing whether they also hold for the older workforce, in a larger set of countries with additional LMEs, and for the other subsystems of the economy defined by varieties of capitalism. Moving forward, an additional avenue for future research would be to take into account the more recently identified types of mixed-market economies (Molina and Rhodes, 2007) and dependent market economies (Nölke and Vliegenthart, 2009). While our objective was to empirically test the classic differentiation between LMEs and CMEs, we believe it would be worth replicating our analysis using these more refined categorisations of VoCs. Note. Odds ratios, robust standard errors in parenthesis, significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001. Note. Odds ratios, robust standard errors in parenthesis, significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001.