Abstract This article focuses on the relationship between equality of educational opportunity and equality of income in different institutional contexts. By combining insights from the literature on Varieties of Capitalism and education sociology, the study investigates how the educational system and the political economy jointly affect the social stratification of educational choices and condition income differentials between graduates of vocational and general education programmes. The empirical analysis contributes to the literature by contrasting the two conceptions of equality and applying a richer institutional approach than previous studies within the fields of education sociology and Varieties of Capitalism. The results reveal that tracking hinders equality of educational opportunity but is also related to higher incomes for vocational education graduates in certain contexts. Wage bargaining coordination reinforces the more equal opportunities of weakly tracked contexts and improves the relative income of vocational graduates in these contexts. 1. Introduction Most people agree that equality is desirable but disagree regarding in what respect we should be equal. Should we be equal in welfare, income or opportunities? This prototypical ethical dilemma has received little attention in the area of comparative political economy and in the influential Varieties of Capitalism (VoC) approach. Although no other single country characteristic has been the object of as much study in the VoC literature as equality has, it is typically exclusively studied in terms of equality of income (e.g., Pontusson, 2005; Iversen and Soskice, 2009; Busemeyer and Iversen, 2011; Thelen, 2014). On one side of the continuum, we have the equal and ‘gentle’ Coordinated Market Economy (CME) version of capitalism, and on the other, we have the unequal Liberal Market Economy (LME) ‘cut-throat’ variety (Bohle and Greskovits, 2009; Acemoglu et al., 2012). Income differentials between high- and low-income earners may be of great concern, but for social scientists, a question of equal weight must be how people end up in these different positions. Most would agree that in this process of stratification, matters such as merit and effort should be rewarded rather than matters that are beyond the influence of the individual. How do the institutions of CMEs and LMEs fare in this regard? Do the institutions that promote equality of income also promote equality of opportunity? In the VoC literature, the comparatively high level of equality within CMEs is argued to be the result of a complex interaction of the institutions of education, the labour market and the political economy (Hall and Soskice, 2001). Vocational education is argued to play a decisive role (Estevez-Abe et al., 2001; Iversen and Stephens, 2008; Busemeyer and Iversen, 2011). Strong vocational education systems provide students with specific skills that offer opportunities for reasonably well-paid jobs without an academic education. However, education sociologists have long argued that the same type of educational systems that clearly differentiate between vocational and academic tracks are associated with inequality of educational opportunity and increased social stratification (e.g., Shavit and Blossfeld, 1993; Pfeffer, 2008; Bol and van de Werfhorst, 2013). In this literature, vocational education is typically considered a ‘double-edged sword’ (Shavit and Müller, 2000) that may facilitate labour market prospects for the academically unskilled while reducing equality of educational opportunity. The VoC literature has devoted little attention to this trade-off; thus, the equality-promoting qualities of vocational education are ambiguous. VoC scholars may respond to this challenge from education sociology by arguing that the trade-off is dependent upon the broader institutional context and that it might collapse if we—in contrast to education sociologists—also consider the institutions of political economy. That is, we must take into account that institutions interact and jointly form outcomes (cf. institutional complementarities, Hall and Soskice, 2001). Busemeyer and Jensen (2012) have argued along these lines. They examined how educational preferences are conditioned by the institutional context and found that coordinated wage bargaining may mitigate the stratifying effect of a vocationally oriented educational system. In addition, Busemeyer (2014) recently made a major effort to incorporate education into the comparative study of welfare states and socio-economic inequality. This study shares Busemeyer’s objective of understanding the relationship between socio-economic inequality and inequality of educational opportunity but adopts a different approach. Whereas Busemeyer focuses on the overall effects of educational institutions, the current study seeks to determine how institutions condition the effects of individual-level characteristics, thereby gaining a better understanding of individual-level mechanisms. The aim of this article is to reveal how the institutions of political economy and education jointly condition equality of educational opportunity and socio-economic inequality. In other words, how sharp are the two sides of the double-edged sword of vocational education? This study makes three contributions. First, institutional precision in empirical analysis permits us to distinguish the effects of different institutions and how they interact. Second, to truly be able to study a trade-off, we must study both outcomes jointly. Only then can we understand how institutions shape the balance between equality of income and equality of opportunity. Finally, the empirical approach focuses on actual educational choices as these are what shape educational stratification. The results show that vocationally oriented systems may have diverging effects on equality of income and equality of opportunity, but the character of these effects is highly dependent upon other institutions rather than vocational enrolment per se. This finding suggests that we need to be clear about what institutional setup we refer to when discussing the effects of vocational education. Tracking not only reduces equality of educational opportunity but is also related to higher incomes for vocational graduates in certain contexts. Coordinated wage bargaining contributes to more equal educational opportunities and to smaller income differentials, but only in weakly tracked contexts. The results imply that equality of income does not necessarily go hand in hand with equality of opportunity; consequently, scholars should separate these different conceptions of equality to nuance the way that institutions affect equality. This introduction is followed by an overview of the literature and a theoretical discussion. The third section explains the methodology and data. The empirical results follow in the fourth section. The paper concludes with a summary and discussion of the results. 2. Education and different dimensions of equality 2.1 Equality of educational opportunity In empirical research, equality of educational opportunity is typically understood as the relationship between students’ socio-economic backgrounds and their educational performance (e.g., Breen and Jonsson, 2005; Brunello and Checchi, 2007; Bol and van de Werfhorst, 2013). A simplification of the theoretical concept of equality of opportunity (see, e.g., Dworkin 1981a,b), this operationalization opens up avenues for empirical research. It suggests that we cannot be responsible for the socio-economic status of our parents and that their status should therefore not affect our opportunities. Nevertheless, worldwide, the social and educational background of a person’s parents is a strong predictor of his or her own educational trajectory (e.g., Shavit and Blossfeld, 1993; Breen and Goldthorpe, 1997; for an overview see Breen and Jonsson, 2005). The strength of this relationship, however, varies substantially across countries, and these differences may provide us with meaningful information on the degree of (in)equality of educational opportunity (Pfeffer, 2008). The literature also commonly assumes that this variation is attributable to different country characteristics, primarily institutions (Breen and Jonsson, 2005). In the comparative study of educational systems and equality of educational opportunity, two institutional features tend to attract most of the scholarly attention: tracking and the degree of vocational orientation (e.g., Shavit and Müller, 2000; Pfeffer, 2008; Bol and van de Werfhorst, 2013). Tracking refers to students being separated into different tracks according to ability or future plans for further education, whether academic or vocational (Shavit and Müller, 2000). The age at which tracking is introduced is often understood as the prime factor that determines the extent of tracking. There is robust evidence that more extensive tracking has a negative effect on equality of educational opportunity (Brunello and Checchi, 2007; Schütz et al., 2008; Bol and van de Werfhorst, 2013; for an overview, see Van de Werfhorst and Mijs, 2010). The German educational system is a well-known example of a heavily tracked system. At the age of 10 or 12 years, children are typically selected into lower secondary schooling (Hauptschule), middle secondary schooling (Realschule) or upper secondary schooling (Gymnasium). These tracks convey different social statuses and tend to predetermine later educational choices (Powell and Solga, 2011). The degree of vocational orientation of an educational system is determined by the extent to which the system provides students with occupation-specific skills that are readily applicable in the labour market rather than general skills (Shavit and Müller, 2000; Bol and van de Werfhorst, 2013). Vocational education and tracking are different phenomena but are arguably related as the existence of vocational programmes implies some form of tracking. However, the age at which tracking is introduced may vary independent of the degree to which upper-secondary education provides students with occupation-specific skills. Even a weakly tracked system may have rather extensive vocational education. Sweden is one such example, where at the age of 15 years children choose between vocational or general programmes at the upper secondary level. The comparably high age of selection, together with the organization of the different programmes within an integrated upper secondary school (Gymnasieskola), indicate a low level of tracking. Nevertheless, an internationally relatively large share of approximately half of each cohort enrols in vocational programmes (Eurydice, 2012). The most prominent framework within education sociology for explaining educational choices and the intergenerational persistence of educational inequality is arguably sociological rational choice theory (SRC) (Goldthorpe, 1996; Breen and Goldthorpe, 1997; Jaeger, 2007). SRC assumes that educational choices are made by rational individuals on the basis of expected costs and benefits as well as the probability of success. Utility is primarily calculated in terms of social status. A central point is the concept of relative risk aversion, which states that individuals from different social classes reason similarly in being risk averse regarding downward social mobility (Breen and Goldthorpe, 1997, p. 283). In essence, individuals from both the upper and working classes will go to great lengths to avoid (the risk of) downward mobility. This may explain why systems with more tracking are associated with a stronger effect of socio-economic background on educational choices. The different tracks signal different class destinations; from a status maintenance perspective, upper-class children would therefore be strongly inclined to choose the general track, whereas working-class children would be ‘free’ to choose either track. Despite recent developments in the comparative institutional study of education (e.g., Bol and van de Werfhorst, 2013; Braga et al., 2013), there are scarcely any studies of how the broader institutional context of the welfare state and political economy interacts with education and affects equality of educational opportunity. The separate research fields of education and the welfare state have seen little integration until recently (Busemeyer, 2014). There are noteworthy exceptions (e.g. Allmendinger and Leibfried, 2003; Beller and Hout, 2006), but those works study the effect of regimes, which hinders an understanding of how distinct institutions affect outcomes. This lack of integration is unfortunate as skill formation—and vocational education in particular—plays such a central role in the VoC literature. 2.2 Varieties of skill regimes Estevez-Abe et al. (2001) were the first within VoC to argue that vocational education in CMEs is conducive to social equality as it offers a route to reasonably well-paid jobs without requiring academic education. The argument has been questioned by Bradley et al. (2003) but reinforced by Busemeyer and Iversen (2011) and Busemeyer (2014). Central to this line of reasoning is how the institutional context conditions the choice of vocational education. Individuals are expected to make rational choices when investing in education by calculating the expected utility of pursuing a certain type of education. By assuming that individuals seek to maximize their economic return on education based on the degree of risk involved, Estevez-Abe et al. argue that the degree of social protection and wage-bargaining institutions are crucial to the formation of different skill regimes. Compared with obtaining a general education, opting for a vocational education and learning specific skills implies greater risk as the range of alternative jobs is more constrained in the event of unemployment. However, generous unemployment insurance and wage coordination may offset these risks and make vocational education an attractive choice. Whereas unemployment insurance protects the incomes of workers in the event of unemployment, wage coordination compresses the wage distribution and entails both higher and more secure wages for vocational graduates. Estevez-Abe et al.’s account of the equal CMEs relies on a mechanism of more equal opportunities in terms of income across classes, whereas inequality of educational opportunity might be unaltered. In other words, CMEs provide opportunities for good incomes (and possibly income mobility) without educational mobility. This pathway is presumed to be relatively rare in LMEs as there are few well-paying jobs available for those lacking higher education. 2.3 Bridging education sociology and political economy Busemeyer and Jensen (2012) took the first step in bridging the gap between education sociology and VoC by studying how educational preferences are affected by the institutional context. To combine the literature on VoC and education sociology, we need to go further and incorporate political economy institutions into the theoretical underpinnings of educational stratification. Moreover, to fully study the effects of institutions, we must study them as part of the broader institutional context because we expect that institutions will interact with one another (Hall and Soskice, 2001). Thus, theoretical and empirical clarity are necessary when scrutinising institutions to make sure that we actually capture the relevant institutions (Korpi and Palme, 2003). It is unfortunate that the VoC literature often relies on a single measure of enrolment in vocational education at the secondary or postsecondary level to determine the character of the educational system (Busemeyer and Jensen, 2012; Iversen and Stephens, 2008). I argue that we need to differentiate between vocational orientation and the degree of tracking. These institutions may be regarded as a clarification of the concept of skill specificity (Busemeyer, 2009); an earlier separation of vocational and general programmes is expected to make the learned skills more specific. Tracking has seen little attention within VoC but is an institution of considerable relevance for equality of educational opportunity and socio-economic equality alike. In line with VoC, in the discussion below, I will concentrate on the choice between vocational and general education as well as the wage differentials between graduates from these different programmes. Starting with the way that political economy institutions affect educational stratification, the focus will be on the degree of coordination of wage bargaining and the generosity of unemployment insurance, which are central to Estevez-Abe et al.’s (2001) argument. How would we expect these institutions to affect the role of socio-economic background in educational choices? This is effectively a matter of whether coordinated wage bargaining and generous unemployment insurance make vocational education a relatively more attractive choice for working-class or upper-class youths. Estevez-Abe et al. argue that these institutions raise the economic utility of vocational education as they imply a higher and more secure income for vocational graduates. This clearly makes vocational education a more attractive alternative compared to general education for both working-class and upper-class youths. However, equality of educational opportunity (in empirical research) is about the strength of the relationship between socio-economic background and educational choices. The way this relationship is affected by the increased economic utility of vocational education hinges upon the relative differences in attraction between youths of different backgrounds. If, on the one hand, working-class youths would be more attracted than upper-class youths by the increased economic utility and consequently would choose a vocational education to a larger extent, the overrepresentation of working-class youths in vocational tracks would be reinforced, which in turn would deepen class-based educational stratification. If, on the other hand, upper-class youths would be more attracted than working-class youths, educational stratification would be weakened because this would result in a larger share of upper-class youths in vocational education. Put another way, the question is whether the increased economic utility of vocational education would primarily divert working-class youths from choosing general education or, rather, would attract upper-class youths to choose a vocational track instead of a general track. SRC and the assumption of relative risk aversion appear to predict that working-class youths would be relatively more attracted by higher incomes because choosing a vocational education would imply status degradation for someone with an upper-class background, regardless of income. If so, the increased economic utility of vocational education in CMEs would lead to stronger educational stratification (but see Busemeyer and Jensen, 2012, for the opposite argument). However, it could be argued that vocational education in the CME setting actually implies a higher social status than in LMEs (Bol and van de Werfhorst, 2011). Because workers with specific vocational skills are presumed to be essential for CME production, those jobs could confer higher social status, even when controlling for wage levels. For example, this could be argued to be the case in Germany, with the strong cultural commitment to vocationalism (Beruflichkeit; Powell and Solga, 2011). This difference in social status would make vocational education in CMEs a more attractive alternative for upper-class youths as it would represent a less clear case of downward social mobility. When comparing the effects of collective wage bargaining and generous unemployment insurance, the latter differs in its explicit functioning as insurance and thus more directly refers to the issue of risk aversion. Youths with a lower-class background who cannot fall back on the economic resources of their parents may be expected to be more risk averse and may thus perceive greater value in unemployment insurance. This line of reasoning would suggest that generous unemployment insurance has a larger effect on the attractiveness of vocational education for working-class youths than for those from a higher-class background. Moreover, the status argument above does not appear to play out as much for unemployment insurance as for coordination. Consequently, we would expect increased educational stratification in the presence of generous unemployment benefits. To summarize how the institutions of political economy may affect educational stratification, in the case of wage coordination, considering the countervailing effects discussed above, I argue that the theoretical expectations are ambiguous. Thus, this is an empirically open question. For generous unemployment insurance, however, I hypothesize a reinforcing effect on the degree of educational stratification. As discussed in section 2.1, tracking is expected to also increase educational stratification. With regard to equality of income, how would we expect the institutional context to condition the effects on income inequalities between vocational and general education graduates? I concentrate on wage coordination and tracking and leave unemployment insurance aside as my focus is on wages for the employed. The fact that wage coordination compresses the wage distribution is common knowledge in the comparative study of political economy (e.g., Rueda and Pontusson, 2000; Busemeyer and Iversen, 2011). However, the way the relative wage of vocational education graduates is affected by wage bargaining institutions and the structure of the education system is less studied. It is possible that wage coordination only compresses the wage distribution without altering the relative positions of different types of occupations. Yet, I would argue that the VoC argument, as represented by scholars such as Estevez-Abe et al., also implies that the relative wages of vocational education graduates should be higher in coordinated economies with specific skill formation. This is the case because specific vocational skills are assumed to be in high demand by employers in these countries and because wage coordination improves the bargaining power of vocational education graduates. I thus hypothesize that the relative wages of vocational education graduates should be higher in countries with a high degree of wage coordination. Understanding tracking as a measure of skill specificity leads me to the hypothesis that tracking should also be related to higher relative wages of vocational graduates. Furthermore, we would have reason to expect these institutions to interact and to further improve the wages of vocational graduates in relation to general education graduates. Consequently, I also hypothesize an interaction between wage coordination and tracking that would further improve the wages of vocational graduates. All hypotheses are summarized in Table 1. Table 1. Summary of hypotheses Hypothesis The stratifying effect of parental education on choice of educational programme 1.1: Wage coordination +/− 1.2: Generous unemployment insurance + 1.3: Tracking + The relative wages of vocational graduates 2.1: Wage coordination + 2.2: Tracking + 2.3: Wage coordination × Tracking + Hypothesis The stratifying effect of parental education on choice of educational programme 1.1: Wage coordination +/− 1.2: Generous unemployment insurance + 1.3: Tracking + The relative wages of vocational graduates 2.1: Wage coordination + 2.2: Tracking + 2.3: Wage coordination × Tracking + 3 Methodology 3.1 Model My focus is on studying how institutions condition the effects of individual characteristics, such as socio-economic background. This interaction occurs between the macro- and micro-levels. I therefore choose to explore the cross-level interaction and apply country-fixed effects in all models. This approach implies that the country dummies absorb all of the average differences in the dependent variable across countries. Only the variation that is conditioned by individual characteristics is used to estimate how institutions influence outcomes. This type of model has been applied in several studies that examine how educational institutions condition the effect of individual characteristics (Brunello and Checchi, 2007; Schütz et al., 2008; Braga et al., 2013). It is considered a more rigid way of addressing country-level heterogeneity than including a country-level random intercept. However, it should be noted that the approach using cross-level interactions and country-fixed effects remains vulnerable to country-level variables that affect the relationship between micro and macro factors and that are correlated with the macro factors in the interaction (Brunello and Checchi, 2007). Such variables would bias the coefficient of the cross-level interaction. To some extent, this bias might be counteracted by including ‘control cross-level interactions’ for these factors if they can be identified and observed. A disadvantage of this type of fixed effects model is that it makes it impossible to study the direct effect of country-level institutions as they are absorbed by the country dummies. The model only allows us to make observations concerning how institutions condition the effect of individual factors. However, this aim is the stated goal of this article: to study how the institutional context conditions individual-level effects of socio-economic background and education. The model can be formally described as follows: Yik=α+β1Xik+ΓXik×Λk+ΩZik+δk+εik, where Yik is the outcome of interest (for each respondent i in each country k). Xik denotes an individual characteristic such as the respondents’ parents’ educational background, and Λk denotes a vector of institutional characteristics at the country level k (such as wage coordination). Zik is a vector of individual control variables, and δk represents the country dummies. The coefficients of interest are those for the cross-level interactions, the vector Γ, as these represent how the institutional context conditions the effect of an individual characteristic. Depending on whether the outcome is a binary or approximately continuous variable, logit or OLS regression is used. All models are estimated with country-fixed effects and country-clustered robust standard errors. The latter relax the unrealistic assumption of independence of observations across countries and allow the error terms to be correlated within countries. 3.2 Two applications of the model To study the two conceptions of inequality, in educational opportunity and in income, I apply the model in two ways. For equality of educational opportunity, the dependent variable is whether an individual has completed a vocational or general programme at the upper-secondary level. At this level, most countries implement some type of separation between vocational and general education. Although the upper-secondary level is rarely compulsory, in practice, it currently represents the minimum level of education for the absolute majority of youths in Europe. Focusing on this level is also consistent with the lion’s share of the literature on vocational education (e.g., Shavit and Müller, 2000; Busemeyer, 2009; Busemeyer and Iversen, 2011; Bol and van de Werfhorst, 2013; but see Culpepper, 2007, for a discussion on tertiary vocational training). Making the choice of education a dichotomous one between vocational and general programmes at the upper-secondary level is arguably a simplification, but it opens up the potential for international comparison of more than length of education, and it corresponds with Estevez-Abe et al.’s (2001) argument that stresses the choice between general and specific skills. Moreover, this approach highlights a critical choice in one’s educational career that, in most educational systems, has considerable consequences for further educational opportunities and, ultimately, for social class (Bol and van de Werfhorst 2013; Jaeger, 2007; Shavit and Müller, 2000). To study years of educational attainment instead would fail to identify different types of education of the same length, yielding a poor operationalization of the VoC concept of specific skills. Although several countries have produced reforms to open up the educational system by allowing vocational graduates to upgrade their qualifications or to allow vocational graduates access to academic tertiary education, these reforms have largely been unable to equalize educational opportunities (e.g. Hall, 2012; Buchholz and Schier, 2015). The choice of secondary education is still where social stratification is reproduced. In the models for equality of educational opportunity, the individual factor that is interacted with the institutional measures is parental education (Xik). The coefficients for these interactions indicate how institutions alter the effect of parental education on educational choices. For the study of income inequality, the dependent variable is relative income measured in country-specific deciles. The individual factor (Xik) refers to whether a respondent has completed a general or vocational programme at the upper-secondary level. The same variable is used as the dependent variable in the first equality of educational opportunity application of the model. When interacting this variable with the institutional measures, we may discern how the income differences between vocational and general education graduates are conditioned by the institutional context. Considering how VoC scholars emphasize the importance of vocational education for the high degree of equality within CMEs, this approach constitutes a valid test of the VoC proposition. In addition, as these income differences approximate the wage differentials between working-class and upper-class occupations, they have a clear bearing on socio-economic inequality. 3.3 Data 3.3.1 Individual level For the individual level, I use cumulative European Social Survey (ESS) data. The six currently available ESS rounds (2002–2012) are combined to form one dataset. This is possible because the ESS rounds were conducted during a limited time window and with a largely comparative set of variables. The ESS is among the few surveys that offer relatively good variables for both the respondents’ level of education and the respondents’ parents. The cumulative ESS dataset includes 30 countries in total, of which 14 participated in all six rounds. The data are weighted using ESS design weights to take into account differences in sampling probability between respondents. The variables reporting educational level in the ESS have recently been developed to improve their quality and comparability (Schneider, 2009, 2010; ESS, 2011). This development includes an incorporation of the new European Survey version of ISCED (ES-ISCED). The main objective of ES-ISCED is to allow for differentiation between types of educational programmes within levels of education (Schneider, 2010). ES-ISCED shares this objective with the widely applied CASMIN scheme (Brauns et al., 2003) but has the advantage of making it possible to use country-level information already collected in surveys (Schneider, 2010). This variable for educational classification thus allows the researcher to differentiate between vocational and general education at the upper-secondary level. This capability has rarely been found in earlier survey-based research as most surveys have relied on classifications of education in terms of levels alone (Schneider, 2009). The respondents’ choice of education at the upper-secondary level—vocational or general—is thus derived from the ES-ISCED indicator and summarized as a binary variable. A vocational education is associated with ES-ISCED IIIb and a general education with ES-ISCED IIIa. The primary difference between these programmes is that the first category prepares students for the labour market through the teaching of occupation-specific skills, whereas the second category of programmes focuses on preparation for further education. A defining aspect of the indicator is that a general programme provides access to tertiary education, whereas a vocational programme does not (Schneider, 2010). I only have data on the highest level of education that a respondent has completed. The choice of programme is thus operationalized as the programme that the respondent chooses to complete. I code those who have completed a higher level beyond upper-secondary education as having completed a general programme at the upper-secondary level. In most countries, only a general programme would provide access to higher education. Those who have not completed upper-secondary education are excluded from the sample. For the socio-economic background of the respondents, I use a variable referring to the education of the respondents’ parents. This five-step variable is based on the ISCED-97 framework (UNESCO, 2006). The highest level of education of either the mother or father is summarized in one variable. To facilitate analysis, this five-step variable is recoded to the expected years of education for different educational degrees on a per-country basis using Schneider’s (2009) framework.1 1 Schneider’s (2009) framework is based on ISCED-97. In cases in which the framework differentiates between more ISCED levels than what is reported in the ESS, an average of the different levels is used. This transformed variable may then be interpreted as years of education for the most highly educated parent and is used in the models as an approximately continuous variable. For the study of wage differentials, I use the only available income measure in the ESS, which is relative household income expressed in deciles. I therefore need to assume that household income is an approximate measure of individual wage. To make this assumption more plausible, I control for the number of household members (using dummies) and study only respondents who currently are working and living in households where the main source of income is paid work. Unfortunately, the income measure was changed between the ESS rounds which adds some uncertainty.2 2 In the three most recent ESS rounds (2008–2012), household income is measured in country-relative deciles. The first three rounds (2002–2006) employed a measure with fixed income intervals across countries. The average income level in a country thus affects the measure, making it incomparable with the latter decile measure. To address this issue, I transformed the earlier measure with 12 fixed intervals into approximate income deciles relative to the country and ESS round. 3.3.2 Institutions The individual-level data are complemented with institutional indicators from different sources. For the institutional setup of the educational system, I use two continuous measures. First, the percentage of vocational enrolment at the upper-secondary level is used to generally account for the degree to which the system is vocationally oriented. Data are derived from Eurydice (2005, 3 For a few cases in which Eurydice information is unavailable, the measure is complemented with UNESCO data., reference year 2001/02).3 This measure is common in comparative research on educational systems within the fields of educational sociology and VoC (e.g. Busemeyer and Jensen, 2012; Bol and van de Werfhorst, 2013). Second, I consider the proportion of the total curriculum of primary and secondary education that is tracked as a measure of the degree of tracking (the reference year is 2002).4 4 The indicator is calculated as follows: (agef − ages)/(agef − agep), where agef is the age when secondary school is completed, ages is the age of first selection, and agep is the age at the beginning of primary school. See Brunello and Checchi (2007). The variation is primarily attributed to differences in the age of first selection. The indicator effectively—as a percentage—expresses the proportion of the total years of primary and secondary education that are tracked. This indicator comes from Brunello and Checchi (2007), and I reuse their data. The institutions of the political economy are represented by two measures: the degree of coordination of wage bargaining and the unemployment insurance replacement rate. I use the ICTWWS wage coordination measure (Visser, 2013, based on Kenworthy, 2001) that exclusively focuses on wage coordination and covers a broader set of countries than the Hall and Gingerich (2004, 2009) measures that are often applied in the VoC literature. Because the theoretical argument is grounded in wage bargaining coordination specifically rather than coordination as a more general feature of a political economy, the ICTWWS measure is also preferable from a theoretical perspective. I use the average of this measure for the 2002–2010 period. Most countries experienced little change during this period. The measure originally ranges from 1 to 5 but has been rescaled to fall between 0 and 1. The unemployment insurance replacement rate is obtained from the OECD and represents the net replacement rate for an average single worker (earning 100% of an average worker’s wage without any additional benefits during the initial phase of unemployment, with 2002 as the reference year). I choose to restrict the sample to countries in Europe, thus excluding Israel. Not all institutional measures are available for all ESS countries. The number of countries included in the models thus varies between 18 and 25. The sample includes a comparatively broad set of European countries, such as most EU members, including many of the members from Eastern Europe as well as some non-EU members, such as Norway and Switzerland. In Figure 1, the country-level institutions are plotted against one another to provide an overview of the relationship between them and their distributions. Tracking, coordination and unemployment insurance show little relationship with vocational enrolment (Pearson’s r = 0.05, 0.03 and 0.01). Tracking and wage coordination are weakly related (r = 0.17), as are tracking and unemployment insurance (r = 0.10, data not shown). The generosity of unemployment insurance is more strongly correlated with wage coordination (r = 0.43, data not shown). Figure 1. View largeDownload slide The country-level relationship between vocational education enrolment, tracking, wage coordination and unemployment insurance. Figure 1. View largeDownload slide The country-level relationship between vocational education enrolment, tracking, wage coordination and unemployment insurance. 3.3.3 Limitations One issue is that I use cross-sectional data to study outcomes that naturally reflect developments over time. I use institutional data from the early 2000s because earlier comparable data are difficult to obtain. The respondents of the ESS surveys have gone through the educational system and experienced these institutions at quite different times depending on their age and the ESS round. This is hardly optimal, but the problem might be mitigated. Educational institutions have been particularly sticky. Most reforms that decreased tracking and made schools more comprehensive were implemented in the 1950s to 1970s. As an example, in the European countries for which I have data, the age of selection has only substantially changed (by more than one year) in three countries since the mid-1980s and has not changed in any countries from the mid-1990s to the early 2000s (Brunello and Checchi, 2007). Even the changes in vocational enrolment are relatively small. Thus, the institutional differences between the younger cohorts included in the ESS data are small. However, it may be argued that although the institutions as such are unaltered, the effects of the institutions are contingent upon the broader context, which has arguably changed considerably over time. I therefore choose to restrict the sample to respondents aged between 22 and 40 years to avoid the possibility that older cohorts’ experiences may bias the study. The lower limit is applied to exclude respondents who have not yet completed their secondary education (cf. Bol and van de Werfhorst, 2013). In addition, birth year and ESS-round dummies are added to absorb some of the potential differences between different cohorts and survey rounds. To take into account possible differences between cohorts within countries, some models also include the interaction of birth year and country dummies (cohort-by-country fixed effects). A control for gender is included in all models. Immigrants could potentially bias the results as they may not have gone through the education system of their current country of residence. I therefore exclude immigrants who immigrated after the typical age for starting school. 4. Empirical results 4.1 Equality of educational opportunity The estimates from the first set of models are presented in Table 2. The results are presented as logit coefficients rather than odds ratios as the latter are difficult to interpret for models involving interaction terms (Norton et al., 2004). A positive coefficient implies an increased probability of having completed a general programme rather than a vocational programme at the upper-secondary level. For example, from Model (1), we observe that the probability of choosing a general programme is higher for females than for males. Table 2. Influence of parental education on completing a general programme rather than a vocational programme at the upper-secondary level in different institutional contexts (vocational program = 0, general program = 1, logit coefficients) (1) (2) (3) (4) (5) (6) Parental education 0.18 0.01 0.23∗∗ 0.07 −0.02 −0.07 (0.12) (0.08) (0.11) (0.07) (0.10) (0.13) Par.edu. × Voc. enr. 0.11 0.14 0.10 0.14 0.13 0.22* (0.19) (0.14) (0.18) (0.11) (0.12) (0.12) Par.edu. × Tracking 0.40∗∗∗ 0.42∗∗∗ 0.46∗∗∗ 0.67∗∗∗ (0.10) (0.08) (0.09) (0.20) Par.edu. × Wage c. −0.08 −0.12∗∗∗ −0.14∗∗∗ 0.04 (0.06) (0.04) (0.04) (0.12) Par.edu. × UI 0.16* (0.09) Par.edu. × Tr. × WC −0.41 (0.27) Female 0.37∗∗∗ 0.35∗∗∗ 0.37∗∗∗ 0.36∗∗∗ 0.32∗∗ 0.36∗∗∗ (0.13) (0.13) (0.13) (0.13) (0.14) (0.13) Pseudo-R2 0.18 0.17 0.17 0.17 0.16 0.17 Countries 23 21 22 20 18 20 Observations 42 193 39 555 39 401 36 763 33 791 36 763 (1) (2) (3) (4) (5) (6) Parental education 0.18 0.01 0.23∗∗ 0.07 −0.02 −0.07 (0.12) (0.08) (0.11) (0.07) (0.10) (0.13) Par.edu. × Voc. enr. 0.11 0.14 0.10 0.14 0.13 0.22* (0.19) (0.14) (0.18) (0.11) (0.12) (0.12) Par.edu. × Tracking 0.40∗∗∗ 0.42∗∗∗ 0.46∗∗∗ 0.67∗∗∗ (0.10) (0.08) (0.09) (0.20) Par.edu. × Wage c. −0.08 −0.12∗∗∗ −0.14∗∗∗ 0.04 (0.06) (0.04) (0.04) (0.12) Par.edu. × UI 0.16* (0.09) Par.edu. × Tr. × WC −0.41 (0.27) Female 0.37∗∗∗ 0.35∗∗∗ 0.37∗∗∗ 0.36∗∗∗ 0.32∗∗ 0.36∗∗∗ (0.13) (0.13) (0.13) (0.13) (0.14) (0.13) Pseudo-R2 0.18 0.17 0.17 0.17 0.16 0.17 Countries 23 21 22 20 18 20 Observations 42 193 39 555 39 401 36 763 33 791 36 763 ∗ P < 0.10, ∗∗P < 0.05, ∗∗∗P < 0.01. Country-clustered robust standard errors in parentheses. Country-fixed effects; birth cohort-fixed effects; ESS round dummies. Voc. enr: Vocational enrolment, UI: Unemployment insurance, WC: Wage coordination. When interpreting coefficients in an interaction model, one must consider the joint effect of both the interaction and its constitutive elements (Brambor et al., 2006). In this case, there are no coefficients estimated for the institutional elements of the interactions because the country dummies absorb these effects. The interpretation of interactions in logit models is more complicated because of non-linearity. The fact that most of the interacted institutional measures never take on the value of zero also adds to the complexity of interpretation. The regression tables are therefore complemented with graphical presentations. The coefficients of interest are those for the cross-level interactions as they indicate how institutions condition the effect of parental education. A positive coefficient implies that an institution strengthens the relationship between parental education and educational choices (i.e. that students with well-educated parents choose a general programme). In contrast, a negative coefficient mitigates this relationship. In Model (1), I begin by examining how vocational enrolment conditions the effect of parental education. We observe that increased vocational enrolment is related to a stronger effect of parental education, but the interaction is insignificant. Nevertheless, I choose to include this measure in all models as it is the most widely applied measure in the literature. It assumes the role of a cross-level control. In Model (2), tracking is added. As expected, tracking clearly magnifies the importance of parental education in the choice of educational programme. The degree of tracking ranges from 0.15 (Norway) to 0.69 (Germany). Moving between these endpoints implies a substantial change in the stratifying effect of parental education on the choice of educational track. Model (3) reflects the conditioning effects of vocational enrolment and wage coordination (hypothesis 1.1). Although wage coordination appears to be related to a decrease in the effect of parental education on educational choice, the effect is insignificant. Model (4) adds tracking again. Tracking retains its strong stratifying effect; furthermore, the mitigating effect of wage coordination now becomes stronger and significant. The fact that the mitigating effect is reinforced may be explained by the positive correlation between tracking and wage coordination (Figure 1), although they have countervailing effects on equality of educational opportunity. This finding underlines the importance of separating different institutional effects. The mitigating effect of wage coordination may be interpreted as indicating that vocational education conveys a higher social status in countries with high levels of coordination, which makes it a more attractive choice for upper-class youths. This status effect consequently outweighs the potential opposite effect of that working-class youths are relatively more attracted by higher incomes than are youths with an upper-class background. In Model (5), the generosity of unemployment insurance is added. The effect is positive and would thus increase stratification, in line with the theoretical argument above (Hypothesis 1.2). The coefficient, however, only reaches statistical significance at the 90% confidence level. Finally, in Model (6), a three-way interaction between parental education, tracking and wage coordination is tested in response to the VoC proposition of institutional complementarities. This test determines whether the degree of coordination interacts with skill specificity (measured as tracking) and jointly affects the importance of parental education for educational choices. As the interaction becomes insignificant, however, there is no support for such an interaction. In addition to the models reported in Table 2, I tested an expansion of Model (4) to determine whether employment protection legislation and union density could alter the effect of parental education. However, these coefficients were far from achieving statistical significance and did not substantially affect the other coefficients.5 5 All results from the robustness checks are available in the Supplementary Appendix. Furthermore, the results of Model (4) are robust against adding cohort-by-country fixed effects and excluding any single country. As might be expected, the weaker significance of unemployment insurance in Model (5) is more sensitive to such alterations. I consequently regard the effect of unemployment insurance as a preliminary finding and focus on Model (4) in the further interpretation of the results. These models with multiple interactions are difficult to interpret simply by examining Table 2. The conditional effects of the main institutional variables of tracking and wage coordination are therefore portrayed in Figure 2. The plot is based on Model (4) and depicts how the marginal effect of parental education on the predicted probability of choosing a general education is altered by different levels of tracking and wage coordination.6 6 The non-linearity of logit models means that the effect of a variable depends on the values of all of the covariates. I therefore calculate the predicted probabilities for each observation for the different combinations of fixed values for parental education, tracking and coordination. I then use the average of these predictions for each combination (Gelman and Pardoe, 2007). The two levels of tracking represent the mean (between countries) plus or minus one standard deviation. The marginal effect of an additional year of parental education is calculated at 12 years of parental education, typically signifying completion of the upper-secondary level and equal to the median in the sample. Figure 2. View largeDownload slide The marginal effect of parental education conditional on tracking and wage coordination (95% confidence intervals, at 12 years of parental education), based on Model (4) in Table 2. Figure 2. View largeDownload slide The marginal effect of parental education conditional on tracking and wage coordination (95% confidence intervals, at 12 years of parental education), based on Model (4) in Table 2. The marginal effect of parental education may be interpreted as a measure of inequality of educational opportunity. Comparing the left and right panels of Figure 2 indicates that tracking amplifies the effect of parental education. More surprisingly, the figure reveals that the mitigating effect of wage coordination is only maintained in weakly tracked contexts. In heavily tracked contexts, the degree of wage coordination makes little difference. Formal testing (two-tailed tests) reveals that the marginal effect is significantly different in weakly tracked contexts compared to heavily tracked contexts (as specified in the plot) when wage coordination is above 0.6 (90% confidence) or 0.7 (95%). Rather than mitigating the effect of heavy tracking, wage coordination reinforces the more equal opportunities of weakly tracked contexts. One interpretation of this result is that the different tracks in a system with early tracking more clearly signal differences in social status, an effect that would eliminate the higher social status of vocational occupations in coordinated economies. The most equal educational opportunities are consequently observed in countries with little tracking and a high degree of wage coordination. In these countries, the postponed tracking means that parents have less influence on the choice of education, whereas coordinated wage bargaining may decrease the status difference between vocational and general education, further mitigating the social stratification of upper secondary education. These results imply that in a country such as Norway, which has little tracking (0.15) and a high degree of wage coordination (0.75), the increase in the probability of choosing a general track equals 1.9% for each additional year of parental education (at 12 years). In the UK, which has an equally low level of tracking (0.15) but little wage coordination (0), the marginal effect would be substantially higher at 3.1%. In a heavily tracked country such as Germany (0.69), with substantial wage coordination (0.75), the corresponding figure equals 3.8%; that is, the influence of parental education is approximately twice as large as that in Norway. Thus far, we have discussed equality of educational opportunity; next, we explore how the institutional context conditions income inequality. 4.2 Equality of income The relationship between income and the type of education is tested using the same type of models as above, with country-fixed effects and a focus on the cross-level interactions. The dependent variable is now country-relative household income in deciles, and the individual-level characteristic that is interacted with the institutional measures is whether the respondent has completed a vocational or general programme at the upper-secondary level. This variable was used as the dependent variable in previous models but is now reverse coded as our focus is on the wages of vocational graduates (general education = 0 and vocational education = 1). The number of years of full-time education is added as a control variable to ensure that we measure the effect of the type of education rather than the length of education. To account for possible country-specific income differences between birth cohorts, the interaction of birth year and country dummies is added to all models. Because the dependent variable is approximately continuous, OLS regression is used. The results are presented in Table 3. Model (1) includes only the individual-level variables and demonstrates that the average difference between vocational and general education graduates equals approximately 0.8 deciles when controlling for years of education and gender. Table 3. The effect of completing a vocational programme compared to a general programme at the upper secondary level on relative income in deciles in different institutional contexts (OLS) (1) (2) (3) (4) (5) Vocational education −0.81∗∗∗ −0.40 −0.75∗∗∗ −0.60∗ −1.47∗∗∗ (0.06) (0.24) (0.23) (0.34) (0.39) Voc. edu. × Voc. enr. −0.44** −0.51* −0.49 −0.11 (0.19) (0.25) (0.32) (0.16) Voc. edu. × Tracking −0.35 −0.49 1.07 (0.35) (0.31) (0.70) Voc. edu. × Wage coord. 0.44 (0.29) 0.50∗ (0.27) 1.65∗∗∗ (0.56) Voc. edu. × Tracking × WC −2.70∗∗ (1.24) Years of education 0.13∗∗∗ 0.13∗∗∗ 0.13∗∗∗ 0.12∗∗∗ 0.12∗∗∗ (0.02) (0.02) (0.02) (0.02) (0.02) Female −0.36∗∗∗ −0.34∗∗∗ −0.33∗∗∗ −0.32∗∗∗ −0.32∗∗∗ (0.04) (0.04) (0.04) (0.04) (0.04) Adj. R2 0.23 0.23 0.25 0.24 0.24 Countries 28 22 23 21 21 Observations 27 748 24 924 24 192 23 041 23 041 (1) (2) (3) (4) (5) Vocational education −0.81∗∗∗ −0.40 −0.75∗∗∗ −0.60∗ −1.47∗∗∗ (0.06) (0.24) (0.23) (0.34) (0.39) Voc. edu. × Voc. enr. −0.44** −0.51* −0.49 −0.11 (0.19) (0.25) (0.32) (0.16) Voc. edu. × Tracking −0.35 −0.49 1.07 (0.35) (0.31) (0.70) Voc. edu. × Wage coord. 0.44 (0.29) 0.50∗ (0.27) 1.65∗∗∗ (0.56) Voc. edu. × Tracking × WC −2.70∗∗ (1.24) Years of education 0.13∗∗∗ 0.13∗∗∗ 0.13∗∗∗ 0.12∗∗∗ 0.12∗∗∗ (0.02) (0.02) (0.02) (0.02) (0.02) Female −0.36∗∗∗ −0.34∗∗∗ −0.33∗∗∗ −0.32∗∗∗ −0.32∗∗∗ (0.04) (0.04) (0.04) (0.04) (0.04) Adj. R2 0.23 0.23 0.25 0.24 0.24 Countries 28 22 23 21 21 Observations 27 748 24 924 24 192 23 041 23 041 ∗ P < 0.10, ∗∗P < 0.05, ∗∗∗P < 0.01. Country-clustered robust standard errors in parentheses. Country and birth cohort-fixed effects and their interaction; dummies for the number of household members; ESS round dummies; age control. Only currently working respondents that live in households where the main income is paid work. Voc. enr.: Vocational enrolment, WC: Wage coordination. Models (2)–(5) indicate how the relative income of vocational education graduates is affected by the institutional context. I focus on the central theoretical variables, vocational enrolment, wage coordination and tracking. In Models (2) and (3), the conditioning effects of tracking and wage coordination are tested separately together with vocational enrolment. Although vocational enrolment has a negative effect on the incomes of vocational graduates, the coefficients for tracking and coordination are insignificant. In Model (4), all of these interactions are included, and wage coordination becomes significant at the 90% level. As wage coordination varies between 0 and 1, this effect implies that vocational graduates in countries with the highest level of wage coordination have an income that is 0.50 deciles higher than the incomes of graduates in the least coordinated countries, thus supporting Hypothesis 2.1. This effect almost eliminates the income difference between vocational and general education graduates. However, we would expect tracking, understood as a measure of the specificity of skills, and wage coordination to jointly affect the wages of vocational graduates (Hypothesis 2.3). In Model (5), such a three-way interaction is added to the model, and we obtain some interesting and clearly significant results. These results are also largely robust to excluding any single country and controlling for the number of hours worked. The coefficients tell us that both tracking and wage coordination may be related to smaller income differentials, but these effects diminish when combined. This is the opposite of the theoretical expectation of mutually reinforcing effects. In addition, the effect of vocational enrolment almost disappears and is far from significant. The earlier significant effects of enrolment in Model (2) and (3) do not hold for a more flexible institutional model. As it is challenging to make sense of a three-way interaction, I continue with graphical interpretation. In Figure 3, the marginal effect on relative income of completing a vocational programme at the upper-secondary level, compared with a general programme, is presented for different levels of wage coordination on the x-axis and for two levels of tracking. Figure 3. View largeDownload slide The marginal effect of completing a vocational programme rather than a general programme at the upper-secondary level on relative income in deciles (95% confidence intervals, based on Model (5) in Table 3). Figure 3. View largeDownload slide The marginal effect of completing a vocational programme rather than a general programme at the upper-secondary level on relative income in deciles (95% confidence intervals, based on Model (5) in Table 3). The income differences between vocational and general education graduates are most pronounced in countries with modest tracking and little wage coordination. In a country such as the UK, with little tracking (0.15) and no wage coordination (0), the income difference would equal 1.5 deciles. This finding is consistent with the VoC argument that individuals in LMEs who lack higher education enjoy few alternatives to low-qualified and poorly paid jobs. However, for higher levels of wage coordination in weakly tracked contexts, the income differences quickly diminish and even become insignificant. This effect most likely reflects the location of the comparatively equal Nordic countries near the end of the continuum. In a country such as Sweden, with little tracking (0.17) and a relatively high degree of wage coordination (0.75), the income difference would be equivalent to 0.48 deciles. The right side of Figure 3 tells a different story. In countries with a tracked education system, the degree of wage coordination makes little difference for the relative incomes of vocational graduates. However, the incomes of vocational graduates in these countries are approximately 0.4 to 0.6 deciles higher overall than those in countries with low levels of both tracking and wage coordination. This can be illustrated by the fact that in a country such as Germany, with both a high degree of tracking (0.69) and wage coordination (0.75), the income difference would account to 0.90 deciles. Yet, in a country such as Hungary, with a similar high level of tracking but a low level of wage coordination, the income differences would be largely the same at 0.85. This result suggests that tracking, to some extent, may substitute for wage coordination in reducing income differences. This finding gives support to the proposition that vocational education in tracked educational systems may function as a safety net (Shavit and Müller, 2000). 5. Conclusion and discussion The empirical results are summarized in Table 4. Beginning with equality of educational opportunity, tracking clearly magnifies the effect of socio-economic background on the choice between vocational and general education. These results are consistent with the findings of other empirical studies (e.g., Brunello and Checchi, 2007; Schütz et al., 2008; Bol and van de Werfhorst, 2013). What is new here is the exploration of the institutions of political economy. Wage coordination contributes to lessening the importance of socio-economic background, but only in weakly tracked contexts. This finding supports the view that vocational education confers higher social status in coordinated economies. However, it also implies that wage coordination does not have the potential to mitigate the stratifying effect of tracking, as Busemeyer and Jensen (2012) argue. Their conclusion appears to be premature based on the average effects of these institutions. The stratifying effect of tracking dominates the countervailing effect of wage coordination in heavily tracked contexts. This may be because early tracking enlarges the social status differences between tracks, thereby making upper-class youths more reluctant to choose the lower tracks and risk downward social mobility. More generous unemployment insurance is related to stronger educational stratification, even though the effect barely reaches statistical significance. Although preliminary, this result points to the dilemma that institutions that increase the economic utility of vocational education also risk increasing educational stratification, at least insofar as these institutions do not also increase the social status of vocational education. Table 4. Summary of results Degree of wage coordination Effect of parental education Low High Low Moderate Small High Large Large Degree of tracking Wage differential Low Large Small High Moderate Moderate Degree of wage coordination Effect of parental education Low High Low Moderate Small High Large Large Degree of tracking Wage differential Low Large Small High Moderate Moderate The stratifying effect of parental background on choosing a vocational or general track and wage differentials between vocational and general education graduates, conditional on the institutional context. With regard to equality of income, wage coordination in weakly tracked contexts improves the relative wage of vocational education graduates. Thus, the well-known smaller wage differences related to wage coordination reflect not only a general compression of the wage distribution but also the relative positions of different occupations. Furthermore, in certain contexts, tracking appears to have the potential to improve the wages of vocational education graduates. The wage differentials between vocational and general education graduates are the least marked in countries with little tracking and a high degree of wage coordination (e.g., the Nordic countries), whereas the wage differentials in countries with tracked educational systems are restrained regardless of the level of wage coordination (e.g. Germany and Hungary). These results complement the country-level empirics presented by Busemeyer (2014) regarding the relationship between enrolment in vocational education and socio-economic inequality through the focus on wages for vocational education graduates. Ultimately, if we wish to study the effect of vocational education on income inequality, this effect should arguably be clearest at the individual level. My results differ to some extent. I find that it is not the degree of enrolment in vocational education as such but rather wage bargaining institutions and tracking that are relevant to the wages of vocational education graduates. To return to the main question of this article, what have we learned about the double-edged sword of vocational education? First, it is not vocational education per se as much as tracking that reduces equality of educational opportunity. However, to the extent that extensive vocational education is associated with tracking, this ‘negative’ side of the sword is supported in the present study. In heavily tracked contexts, this connection applies regardless of the level of wage coordination. Furthermore, because tracking is related to confined income differentials, there is some support for the ‘positive’ side of the sword. However, it should be noted that this ‘positive’ side is not as sharp as the other as the smallest income differences are observed in weakly tracked countries with a high degree of wage coordination. These results suggest that we do not face a genuine trade-off regarding the effects of vocational education as there are countries that combine extensive vocational education with high levels of equality of opportunity and equality of income. The overarching theme here is the complex relationship between equality of outcome and equality of opportunity. I have claimed that the equalising promise of specific skills offered in the VoC literature is a promise of relative economic equality without class mobility. However, it is also important to consider what the alternative to vocational education would be. The criticism that tracking in conjunction with vocational education reduces equality of educational opportunity tends to assume that the alternative is higher education and the high-status jobs to which it often affords access. However, higher education cannot be an alternative for everyone. For those concerned by labour market stratification, identifying institutions that improve the prospects of reasonably good jobs even for those who lack academic skills should be of great concern. The guiding principle of this article is that by studying how institutions interact with each other and with individual-level factors, we may learn how institutions condition outcomes. Thus, this article explored beyond the conclusions of more limited comparative institutional studies. Ultimately, however, there naturally exists a trade-off between to what extent we may distinguish between different institutions while simultaneously studying the broader context. The above discussion suggests that future studies should provide a more in-depth understanding of the relationship between tracking and vocational education. Another crucial point that requires further development is how to incorporate the study of alternatives to vocational education. In particular, the dropout option arguably represents the worst alternative from a labour market perspective. The differences between countries in the social status connected to different types of education, especially vocational education, require further study. Social status plays a central role in educational choices. Understanding how the status of jobs and education is conditioned by the institutional context could shed new light on the mechanisms of educational stratification. Supplementary material Supplementary material is available at Socio-Economic Review online. Acknowledgements The author is grateful for valuable comments from Per Adman, Karl-Oskar Lindgren, Joakim Palme, two anonymous reviewers and seminar participants at the Department of Government and the Uppsala Center for Labor Studies. Funding This work was supported by Department of Government at Uppsala University and the Uppsala Center for Labor Studies. References Acemoglu D., Robinson J., Verdier T. ( 2012) Can’t we all be more like Scandinavians? Asymmetric Growth and Institutions in an Interdependent World, NBER Working Paper no. 18441. Allmendinger J., Leibfried S. ( 2003) ‘ Education and the Welfare State: the Four Worlds of Competence Production’. Journal of European Social Policy , 13, 63– 81. Google Scholar CrossRef Search ADS Beller E., Hout M. ( 2006) ‘ Welfare States and Social Mobility: How Educational and Social Policy may Affect Cross-National Differences in the Association Between Occupational Origins and Destinations’, Research in Social Stratification and Mobility , 24, 353– 365. Google Scholar CrossRef Search ADS Bohle D., Greskovits B. ( 2009) ‘ Varieties of Capitalism and Capitalism «tout court»’, European Journal of Sociology , 50, 355– 386. Google Scholar CrossRef Search ADS Bol T., van de Werfhorst H. ( 2013) ‘ Educational Systems and the Trade-off Between Labor Market Allocation and Equality of Educational Opportunity’, Comparative Education Review , 57, 285– 308. Google Scholar CrossRef Search ADS Bol T., van de Werfhorst H. ( 2011) ‘ Signals and Closure by Degrees: The Education Effect Across 15 European Countries’, Research in Social Stratification and Mobility , 29, 119– 132. Google Scholar CrossRef Search ADS Bradley D., Huber E., Moller S., Nielsen F., Stephens J. D. ( 2003) ‘ Distribution and Redistribution in Postindustrial Democracies’, World Politics , 55, 193– 228. Google Scholar CrossRef Search ADS Braga M., Checchi D., Meschi E. ( 2013), ‘ Educational Policies in a Longrun Perspective’, Economic Policy , 28, 45– 100. Google Scholar CrossRef Search ADS Brambor T., Clark W. R., Golder M. ( 2006) ‘ Understanding Interaction Models: Improving Empirical Analyses’, Political analysis , 14, 63– 82. Google Scholar CrossRef Search ADS Brauns H., Scherer S., Steinmann S. ( 2003) ‘The Casmin Educational Classification in International Comparative Research’. In Hoffmeyer-Zlotnik J., Wolf C. (eds) Advances in Crossnational Comparison: A European Working Book for Demographic and Socio-economic Variables , New York, NY: Springer, pp. 221– 244. Breen R., Goldthorpe J. H. ( 1997) ‘ Explaining Educational Differentials Towards a Formal Rational Action Theory’, Rationality and Society , 9, 275– 305. Google Scholar CrossRef Search ADS Breen R., Jonsson J. O. ( 2005) ‘ Inequality of Opportunity in Comparative Perspective: Recent Research on Educational Attainment and Social Mobility’, Annual Review of Sociology , 31, 223– 243. Google Scholar CrossRef Search ADS Brunello G., Checchi D. ( 2007) ‘ Does School Tracking Affect Equality of Opportunity? New International Evidence’, Economic Policy , 22, 781– 861. Google Scholar CrossRef Search ADS Buchholz S., Schier A.. ( 2015) ‘ New Game, New Chance? Social Inequalities and Upgrading Secondary School Qualifications in West Germany’, European Sociological Review , 31, 603– 615. Google Scholar CrossRef Search ADS Busemeyer M. R. ( 2014) Skills and Inequality , Cambridge, UK: Cambridge University Press. Google Scholar CrossRef Search ADS Busemeyer M. R. ( 2009) ‘ Asset Specificity, Institutional Complementarities and the Variety of Skill Regimes in Coordinated Market Economies’, Socio-Economic Review , 7, 375– 406. Google Scholar CrossRef Search ADS Busemeyer M. R., Iversen T. ( 2011). ‘Collective Skill Systems, Wage Bargaining, and Labor Market Stratification’. In Busemeyer M.R., Trampusch C. (eds) The Political Economy of Collective Skill Formation , Oxford, Oxford University Press, pp. 205– 233. Google Scholar CrossRef Search ADS Busemeyer M. R., Jensen C. ( 2012) ‘ The Impact of Economic Coordination and Educational Institutions on Individual-Level Preferences for Academic and Vocational Education’, Socio-Economic Review , 10, 525– 547. Google Scholar CrossRef Search ADS Culpepper P.D. ( 2007) ‘ Small States and Skill Specificity Austria, Switzerland, and Interemployer Cleavages in Coordinated Capitalism’, Comparative Political Studies , 40, 611– 637. Google Scholar CrossRef Search ADS Dworkin R. ( 1981a) ‘ What is equality? Part 1: equality of welfare’, Philosophy & Public Affairs , 10, 185– 246. Dworkin R. ( 1981b) ‘ What is Equality? Part 2: Equality of Resources’, Philosophy & Public Affairs , 10, 283– 345. ESS, ( 2011) Education Upgrade ESS1-ESS4 – Documentation Report, The ESS data archive. Estevez-Abe M., Iversen T., Soskice D. ( 2001) ‘Social Protection and the Formation of Skills: A Reinterpretation of the Welfare State’. In Hall P. A., Soskice D. (eds) Varieties of Capitalism: The Institutional Foundations of Comparative Advantage , Oxford, Oxford University Press, pp. 145– 183. Google Scholar CrossRef Search ADS Eurydice ( 2005) Key Data on Education in Europe 2005 , Luxembourg, Office for Official Publications of the European Communities. Eurydice ( 2012) Key Data on Education in Europe 2012 , Brussels, Belgium: Education, Audiovisual and Culture Executive Agency. Gelman A., Pardoe I. ( 2007) ‘ Average Predictive Comparisons for Models with Nonlinearity, Interactions, and Variance Components’, Sociological Methodology , 37, 23– 51. Google Scholar CrossRef Search ADS Goldthorpe J. H. ( 1996) ‘ Class Analysis and the Reorientation of Class Theory: The Case of Persisting Differentials in Educational Attainment’, The British journal of sociology , 47, 481– 505. Google Scholar CrossRef Search ADS Hall C. ( 2012) ‘ The Effects of Reducing Tracking in Upper Secondary School Evidence from a Large-Scale Pilot Scheme’, Journal of Human Resources , 47, 237– 269. Google Scholar CrossRef Search ADS Hall P. A., Gingerich D. W. ( 2004) Varieties of Capitalism and Institutional Complementarities in the Political Economy: An Empirical Analysis , MPIfG Discussion Paper 05/5. Max Planck Institute for the Study of Societies. Hall P. A., Soskice D. W. ( 2001) ‘An Introduction to Varieties of Capitalism’. In Hall P. A., Soskice D. (eds) Varieties of Capitalism: The Institutional Foundations of Comparative Advantage , Oxford, Oxford University Press, pp. 1– 68. Google Scholar CrossRef Search ADS Hall P. A., Gingerich D. W. ( 2009) ‘ Varieties of Capitalism and Institutional Complementarities in the Political Economy: An Empirical Analysis’, British Journal of Political Science , 39, 449– 482. Google Scholar CrossRef Search ADS Iversen T., Soskice D. ( 2009) ‘ Distribution and Redistribution: The Shadow of the Nineteenth Century’, World Politics , 61, 438– 486. Google Scholar CrossRef Search ADS Iversen T., Stephens J. D. ( 2008) ‘ Partisan Politics, the Welfare State, and Three Worlds of Human Capital Formation’, Comparative Political Studies , 41, 600– 637. Google Scholar CrossRef Search ADS Jæger M. M. ( 2007) ‘ Economic and Social Returns to Educational Choices Extending the Utility Function’, Rationality and Society , 19, 451– 483. Google Scholar CrossRef Search ADS Kenworthy L. ( 2001), ‘ Wage-Setting Measures’. World Politics , 54, 57– 98. Google Scholar CrossRef Search ADS Korpi W., Palme J. ( 2003) ‘ New Politics and Class Politics in the Context of Austerity and Globalization: Welfare State Regress in 18 countries, 1975-95’, American Political Science Review , 97 ( 3), 425– 446. Google Scholar CrossRef Search ADS Norton E. C., Wang H., Ai C. ( 2004) ‘ Computing Interaction Effects and Standard Errors in Logit and Probit Models’, Stata Journal , 4, 154– 167. Pfeffer F. T. ( 2008) ‘ Persistent Inequality in Educational Attainment and its Institutional Context’, European Sociological Review , 24, 543– 565. Google Scholar CrossRef Search ADS Pontusson J. ( 2005) Inequality and Prosperity: Social Europe vs. Liberal America , New York, NY: Cornell University Press. Powell J. W., Solga H. ( 2011) ‘ Why are Higher Education Participation Rates in Germany so Low? Institutional Barriers to Higher Education Expansion’, Journal of Education and Work , 24, 49– 68. Google Scholar CrossRef Search ADS Rueda D., Pontusson J. ( 2000) ‘ Wage Inequality and Varieties of Capitalism’, World Politics , 52, 350– 383. Google Scholar CrossRef Search ADS Schneider S. L. ( 2009) Confusing Credentials: The Cross-Nationally Comparable Measurement of Education Attainment, PhD thesis , Oxford, UK: Nuffield College, University of Oxford. Schneider S. L. ( 2010) ‘ Nominal Comparability is not enough: (in-)Equivalence of Construct Validity of Cross-National Measures of Educational Attainment in the European Social Survey’, Research in Social Stratification and Mobility , 28, 343– 357. Google Scholar CrossRef Search ADS Schütz G., Ursprung H. W., Wößmann L. ( 2008) ‘ Education Policy and Equality of Opportunity’, Kyklos , 61, 279– 308. Google Scholar CrossRef Search ADS Shavit Y., Blossfeld H.-P. (eds) ( 1993) Persistent Inequality: Changing Educational Attainment in Thirteen Countries . Boulder, CO: Westview. Shavit Y., Müller W. ( 2000) ‘Vocational Secondary Education, Tracking, and Social Stratification’. In Hallinan M. T. (ed.) Handbook of the Sociology of Education , New York, Kluwer, pp. 437– 452. Thelen K. ( 2014) Varieties of Liberalization and the New Politics of Social Solidarity , New York, NY: Cambridge University Press. Google Scholar CrossRef Search ADS UNESCO. ( 2006), International Standard Classification of Education – ISCED 1997 , (May 2006, Re-edition), UNESCO Institute for Statistics. Van de Werfhorst H., Mijs J. ( 2010) ‘ Achievement Inequality and the Institutional Structure of Educational Systems: A Comparative Perspective’, Annual Review of Sociology , 36, 407– 428. Google Scholar CrossRef Search ADS Visser, J. (2013) ICTWSS Data base, version 4.0, Amsterdam, The Netherlands: AIAS, University of Amsterdam. © The Author 2017. Published by Oxford University Press and the Society for the Advancement of Socio-Economics. All rights reserved. For Permissions, please email: firstname.lastname@example.org
Socio-Economic Review – Oxford University Press
Published: Jan 1, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
All the latest content is available, no embargo periods.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud