Individual and institutional influences on EU labour market returns to education: a comparison of the effect of the 2008 economic crisis on eight EU countries Open Access

A body of research indicates the significance of the impact of the level of educational attainment on individual labour market outcomes: people with a high level of education have higher occupational positions, earn more and have lower unemployment risks (Allmendinger 1989; Carbonaro 2006; Andersen and van de Werfhorst 2010). Previous studies have also revealed that the impact of education on labour market prospects varied largely according to the institutional context of the countries (Estevez-Abe et al.2001; Carbonaro 2006; Werfhorst 2011). However, research on this topic rarely explores Central and Eastern European (CEE) countries (Münich et al. 2005; Andersen and Van de Werfhorst 2010; Večernik 2013). Some authors have studied the impact of education on the labour market entry process in CEE countries (Kogan and Unt 2005; Saar et al. 2008; Kogan et al. 2011) but these studies concentrate on specific employment career phase. Consequently, little is known about whether or not previous findings about the effect of education on labour market outcomes are consistent for these countries and this is why this paper places great emphasis on CEE countries when selecting the country cases for comparison.

Temporal changes are particularly relevant in periods when an economic crisis affects labour market opportunities. Previous analyses have indicated considerable differences in the experiences of countries, both in the depth and duration of the initial recession and in the extent of recovery from the economic crisis (Gallie 2013a). Tåhlin (2013) expects that the effect of the 2008 economic crisis will be relatively small in countries with an institutional structure favouring low inequality in labour market rewards, for example, wage bargaining coordination and centralisation, strong trade unions and active labour market policies. When comparing returns to education in different countries, the impact of macro-structural trends might also be dissimilar across countries. Broader institutional frameworks provide varying degrees of institutional shelter from the impact of macro-structural changes to educational groups (Gangl 2003b). Kalleberg (2011) argues that an economic crisis might accelerate the process of labour market polarisation in advanced societies. Research indicates that the 2008 economic crisis influenced, in the unemployment context, some educational groups more than others, with the low educated being the most affected (European Commission 2010). This finding means that the absolute difference between educational groups has increased. However, labour market institutions seem to modify the association between an increase in the unemployment rate and the distribution of this rate.

This article deals with cross-national variation in the impact of education on labour market outcomes – specifically occupational status and the risk of unemployment – before, during and after the crisis. In this regard, our approach provides an alternate to the regular analysis by economists predicting wage returns to human capital investments. Instead, we are interested in labour market returns in terms of finding a job at all i.e. avoiding unemployment as well as attaining a better job with higher occupational status. We used data from existing studies to compare countries with different institutional frameworks since the impact of education should vary according to this institutional context. The comparison includes three CEE countries where little research of this type has been carried out: Estonia, Hungary, and Slovenia. The other countries studied were Denmark, Germany, Spain, Sweden and the UK. These countries were chosen as typical representatives (‘ideal types’) of the different institutional frameworks – we elaborate on this in a subsequent section. Our approach was to compare countries at an individual level to assess the empirical pattern of differences and the potential relevance of the different institutional context. We also explore trends over time of the impact of education in three distinct periods: immediately before (2007), during (2009) and after (2014) the economic crisis. The role of institutions was expected to be particularly crucial in these periods. The main research questions were:

• How has the increase in unemployment during the crisis affected different educational groups? How do the differences by educational groups vary by country?

• Has the crisis changed the variation of distributive occupational outcomes? Are these changes country specific?

• Does the pattern in the variation in distributive outcomes in countries belonging to different ideal types differ and how?

The European Union Labour Force Survey (EULFS) offers good comparative data on the periods before the recession (EULFS 2007), during the recession (EULFS 2009) and afterwards (EULFS 2014). Based on the research questions, the study also concentrated on three CEE countries to assess the empirical pattern of difference in these countries and to explore whether existing typologies could be useful in classifying these countries.

Conceptualising how to match workers to jobs plays a central role for most sociological theories about labour markets. These resulting models describe the outcomes of allocation processes resulting from the individual actions of employers and workers (Sørensen and Kalleberg 1981; Logan 1996). Research on educational differences in labour market outcomes also distinguish between supply-focused (job applicant) and demand-focused (employer) theoretical perspectives. Human capital theory (Becker 1964) argues that differences in labour market outcomes of different educational groups arise from the practical utility of relevant skills, and from the balance of supply and demand. Workers prefer positions which maximise their perceived utility and promise adequate returns to their investment in education. Otherwise, they prefer unemployment if this maximises their perceived utility. Signalling theory (Spence 1973) contends that educational credentials give advantages in the labour market through education signalling higher individual productivity, competency, etc. irrespective of whether an individual actually possesses these attributes. However, the decisions and actions of applicants and potential employers depend on the institutional and structural conditions under which the decisions are made (Müller and Gangl 2003).

Level of education is expected to moderate the relationship between macro-level changes and micro-level outcomes. Previous results indicate that education is strongly related to occupational position, employment status and earnings at the individual level (Breen and Buchmann 2002; Müller and Gangl 2003; Hout 2012). However, there are significant differences in the size of the returns to education between countries and time periods. Institutional characteristics of societies, such as educational systems and labour market institutions, explain country variations. The impact of education on unemployment risk depends on the specificity of vocational skills provided by the educational system (Shavit and Müller 2000). Countries with extensive occupational tracking in educational systems, strong vocational orientation and limited tertiary enrolment also tend to be characterised by a strong relationship between education and occupational status (Van der Velden and Wolbers 2003; Breen 2005; Andersen and Van de Werfhorst 2010). In addition, labour market institutions seem to generate particular differences in returns to education. The signalling value of education also plays an important role in bridging the educational system and the labour market (Spence 1973, 1981). Educational expansion in general and the economic crisis, in particular, may have an influence on this process by reducing the signalling value of educational credentials and this may lessen returns to education. In addition, an extensive coordination will reduce the effect of education on occupational status, as well as on unemployment risk, because coordination strongly influences inclusion processes and reduces inequalities resulting from educational attainment (Estevez-Abe et al.2001). It is noticeable that the current analysis deals with level of education and type of education is not considered.

Changes in macrostructural context govern job competition and job-matching activity in the labour market. The skill-based technological change hypothesis addresses the variation in educational returns (Acemoglu 2002). This hypothesis suggests that technological change allows employers to substitute low-skilled jobs with machines, reducing the market value of the less educated. At the same time, it gives more value to the highly educated. On the demand side, the 2008 economic crisis increased the speed of occupational upgrading causing a significant decrease in the proportion of low-skilled jobs but especially skilled manual jobs (Cedefop 2010). Also, there is the contrasting trend of an increase in the demand for elementary occupations. The full process can be described as labour market polarisation when employment in both high-skilled and elementary occupations has grown during the twenty-first century (Cedefop 2011).

On the supply side, the recession has led to a rise in educational enrolment as individuals continue their education as an alternative to employment when jobs are scarce. Tighter competition, stemming from the changing balance of supply and demand in the market, is likely to affect the least competitive individuals most heavily (Coleman 1991). Generally, falling recruitment and a relatively large decrease in the availability of low-skilled jobs might lead to downward replacement: the difficulty for the well-educated to find a high-skilled job due to an economic recession might lead them to take a low-skilled job. Consequently, less educated workers have difficulty in finding and keeping even low-skilled jobs and they experience the greatest difficulties in coping with tighter job competition. Hence, while the recession is likely to have a general adverse effect in terms of unemployment risk, this effect should apply particularly to those with lower levels of education.

Based on these considerations we formulated the following hypotheses:

Considerable differences have been found between country patterns and the relatively high stability of those patterns over time. Research about the role of institutions in mediating processes of structural change has developed different typologies for classifying countries into regime types. We use the typology based on employment regimes because this typology places a great emphasis on the degree of inclusiveness of the regulative system (Gallie 2007, 2013a,b). This dimension is very important to study the impact of the economic crisis on employees with different level of education. The typology which is rooted in welfare regime theory¹ (Esping-Andersen 1990) differentiates Continental (corporatist) countries, Nordic (social-democratic) countries, Southern European countries, Anglo-Saxon liberal countries and Transition countries.

The Continental countries have strong employment protection, which is traditionally expected to protect workers from cyclical fluctuations in the labour market (Gangl 2003a). In other words, these countries belong to the insider labour market fraction in accordance with the insider-outsider divide where incumbent workers are in favourable working conditions (Lindbeck and Snower 1988). However, the Continental countries tend to have occupational systems characterised by rigidities in the linkage between qualifications and the types of jobs, which should limit the downward replacement tendencies. The Continental countries have high labour market policy spending. Tåhlin (2013) considers that Continental countries have a mid-level of institutionalised equality.

The Southern countries have relatively high employment protection (even higher than in Continental countries). But as opposed to strictly regulated permanent contracts, temporary workers form a very flexible tier in the labour market of Southern European countries. These countries are considered to have a form of paternalist dualism where labour market divisions are more strongly defined by age than by qualifications as in the Continental countries with a form of corporatist dualism. Consequently, the differences between educational groups should be less pronounced in the Southern European countries which can also be characterised by certain peculiarities of familial kind in terms of their welfare techniques (Ferrera 1996). However, Southern European countries are rather close to the equality level of Continental Europe (Tåhlin 2013).

The Anglo-Saxon Liberal countries have two common features: a relatively high level of internal inequality (in terms of qualification, class, etc.) and greater flexibility in changing economic circumstances (Gallie 2013c). A quite flexible labour market and a smaller welfare state imply a high vulnerability of the labour market. On the degree on institutionalised equality the Anglo-Saxon countries are the least equal (Tåhlin 2013). Occupational polarisation might have first of all had an impact on the labour market opportunities of low educated people.

The Nordic countries are characterised by smaller differences between educational groups and lower levels of inequality in the labour market. In contrast to the Continental countries with high employment protection, Nordic countries tend to combine high labour market policy spending with lower employment protection (Auer 2006). Another typical feature of these countries is the mixture of flexibility and security (= ‘flexicurity’) (Muffels et al. 2014).

A critical observation on this typology is that the allocation of Transition countries has remained controversial. Gallie himself argues that the Central and Eastern European (CEE) countries seem to be the closest to the Liberal countries in institutional terms, with their low level of bargaining coordination and bargaining coverage. However, Bohle and Greskovits (2012) distinguish between the embedded neoliberal model of Visegrad countries (the Czech Republic, Hungary, Poland and Slovakia) and the neoliberal model of the Baltic countries. The Baltic countries have rather high labour market flexibility, low trade union density and coverage of collective agreements, modest expenditures on labour market policies and flexible wages. These institutional features make them similar to Liberal countries. In the embedded neoliberal economies, the state makes a significantly bigger contribution to labour market measures. Slovenia is an exception with an institutional framework being closer to the Continental countries (see also Gallie 2013b).

The temporal change in the experience of unemployment related to the economic crisis varied significantly across European countries.

Previous analyses indicated that an initial rise in unemployment levelled off quickly and remained at a relatively low level in the Nordic and Continental countries (Gallie 2013b). These countries appeared to have highly stable institutional patterns that have changed little with the economic crisis compared to countries with more deregulated employment institutions. At the same time, in Southern European countries there was a sharp and continuous rise in unemployment until 2010 (European Commission 2011). There was a high degree of diversity in the Anglo-Saxon Liberal countries. As Gallie (2013b) mentioned Ireland in some ways fitted better than the UK the assumptions about the impact of economic crisis on employment prospects (the unemployment increased sharply in Ireland but in the UK the rise was quite modest). However, the UK was particularly unequal with a bigger difference in employment decline between workers with primary and secondary education (Tåhlin 2013). Previous research indicated that occupational polarisation during the crisis was most intense in countries with weaker equality-promoting labour market institutions, i.e. in the Anglo-Saxon countries.

Previous analyses indicated important variations in the impact of the crisis on labour opportunities within CEE countries (Gallie 2013c, Muffels et al. 2014). Previous research showed that the economic crisis hit the labour market of the Baltic states more severely than most other European countries (see Eamets and Krillo 2011). The increased premium of education during the crisis might have polarised the labour market opportunities of educational groups. Visegrad countries, with the embedded neoliberal model, have been found to have fairly high levels of mobility and flexibility but similarly low transition income and employment security levels when compared to Southern European countries, indicating risk management practices skewed especially towards the young and low-skilled workers (Muffels et al. 2014). Due to weak institutionalisation, there was a diversity of employer strategies during the economic crisis in the CEE countries of the Liberal type.

In exploring country variations and the potential effects of institutional differences we compared eight countries with varying institutional frameworks. Following previous typologies, Germany represents the Continental (corporatist) countries, Denmark flexicurity, Sweden is a Nordic (social-democratic) country, Spain embodies the type of the Southern European countries, the UK is an Anglo-Saxon liberal country, Estonia can be labelled a neoliberal transition country, Hungary belongs to the transition countries with an embedded neoliberal model and Slovenia is a neo-corporatist transition country being close to the Continental model.

We formulated the following hypotheses based on previous findings and institutional variations:

The analysis used data from EU Labour Force Survey (EULFS), which provides standardised, cross-sectional information on individuals compiled from national labour force surveys. The EULFS provides core data on large samples of national labour force participation and unemployment as well as core demographic information and educational attainment. Based on large-scale and annually repeated national surveys, the EULFS offers a rich time-series of cross-sectional labour market data. We used data from three surveys: 2007, 2009 and 2014.

The first dependent variable in our analyses was unemployment according to the definition of the International Labour Organisation (ILO). The second dependent variable was occupational status, more specifically the International Socio-Economic Index (ISEI) developed by Ganzeboom and Treiman (1996). Respondents being in labour force are analysed. The main predictor was education with three categories: (1) less than a secondary school diploma (ISCED 0-2); (2) vocational or general training at the upper secondary level and post-secondary, non-tertiary education (ISCED 3-4), and (3) tertiary education (ISCED 5-6).² The statistical models also controlled for age, gender and immigrant status. Gender and immigrant status were dummy-coded whereas age was grouped in ten-year intervals (15-24, 25-34, 35-44, 45-54, older than 54).

In order to examine the influence of education, the period effects and the variation by countries on unemployment risks and occupational status, logit models (for the odds of unemployment) and OLS regression (for estimating the ISEI scores) were used. Model 1 provided a test on the standard hypothesis: the variation by country and time has an impact on unemployment risks and occupational status. Model 2 extended the model further and included education. The three interaction terms were added to the models separately. Model 3 builds on the previous model by adding interactions between education and year. The interaction term between country and education comes in Model 4 and Model 5 extended it further by including the interaction between country and year. The final Model 6 made the estimations complete by adding all three interactions mentioned before. This final model allowed us to assess the time- and country-specific impact of education on the risk of unemployment and on occupational status.³ Furthermore, analysis was carried out by allowing for the clustering of standard errors. The final models are presented in Appendix 1 (logit estimates) and Appendix 2 (linear regressions); while the predicted probabilities calculated from the last model are presented in graphical form.

This section provides general descriptive information on the eight countries investigated. The comprehensive indicator of GDP uncovers a clear cleavage between Denmark, Sweden, Germany, the UK, on the one hand and Estonia, Spain, Hungary, Slovenia, on the other hand. A temporal change of GDP reveals some economic recovery for 2014 after the crisis in most of the countries except Slovenia and Spain. Changes in the unemployment rate between 2007 and 2014 reflect the impact of the economic crisis in most countries: unemployment peaking in 2009 is followed by a decline. This is not the case in Spain and Slovenia, again. Germany is another exception where the unemployment rate has dropped during the era studied.

Despite the brevity of the period of analysis, 2007–2014, some educational upgrading can be observed: the percentage of uneducated tends to decline and the proportions of people with tertiary education increase in almost all countries. At the same time, the levels of education fluctuate considerably between these countries. The proportion of graduates is the highest in the UK (2014: 38.1%) and the lowest in Hungary (2014: 20.5%). The proportion of low educated people is the biggest in Spain. For job status, the mean of the ISEI score varies slightly between 40 and 46 for the majority of the countries. The exception is Hungary where the mean job status is somewhat lower. There are only minor modifications in job status over time, though a slight drop occurs in Hungary and a slight increase appears in Slovenia (Table 1).

The key aspect of the analysis was the effect of the level of education (ISCED) on the dependent variables (unemployment and ISEI score) focussing on changes over the period 2007–2014. Starting with the risk of unemployment, the results indicated substantial differences between countries in the probabilities of being unemployed. Time also had an impact on unemployment in the expected direction: the probability of unemployment was higher in 2009 and 2014 than in 2007 (Appendix Table A1, Model 1). The inclusion of education in the model revealed the particular harmful role of low education in increasing (positive estimates) the risk of unemployment (Appendix Table A1, Model 2). Most interaction terms between education and year were significant. The probability of unemployment increased significantly for respondents with a level of education lower than ISCED 5–6 for the years 2009 and 2014 as compared to the period before the economic crisis in 2007. In fact, a smaller increase of unemployment occurred to those with higher level of education. However, the differences in the unemployment probabilities between high and medium educated individuals were smaller in 2014 than in 2007 (Appendix Table A1, Model 3).

The impact of education on unemployment also differed between countries. A lower level of education increased the risk of unemployment in Hungary substantially more than in the UK (the reference country). As shown by negative estimates, respondents with less education were in a relatively better situation in Denmark and Slovenia, both countries with a low level of inequality (Appendix Table A1, Model 4). As for country differences in temporal changes, estimates showed a lower rate of increase in unemployment incidents in Germany for 2009 and 2014. On the contrary, there was a big increase in 2009 and a smaller one in 2014 for Estonia compared to 2007, but an even bigger increase in the unemployment risks appeared for Spain in particular in 2014. The odds of unemployment were lower for 2009 but higher for 2014 in Sweden and Slovenia while the pattern in Denmark did not differ significantly from the UK. (Appendix Table A1, Model 5). Model 6 included all three sets of interaction terms and the estimates turned out to be almost the same.

Turning to the investigation of occupational status, the main effects displayed significant country variations as well as important changes in the occupational status over time. The uncontrolled pattern displays a decline in the first phase of the analysis between 2007 and 2009, followed by a recovery between 2009 and 2014 (Appendix Table A2, Model 1). However, when controlling for education the trend turns to linear downward (Appendix Table A2, Model 2 and onwards). Nevertheless, the U-shape pattern is present again when all interaction terms are added to the equation (Appendix Table A2, Model 6). The expected lower status for those with lower levels of education showed up in the next model. Low education (ISCED 0-2) led to 32 point loss in the ISEI score; secondary level (ISCED 3-4) led to 24 point loss on average, compared to a tertiary level of qualification (Appendix Table A2, Model 2).

The interaction terms on the temporal change in the effect of education revealed an increasing deficit for the low educated (ISCED 0-2); they secured jobs with lower ISEIs, partly in consequence of the crisis, and the tendency persisted even for 2014. The difference in ISEI scores between the medium and the highly educated did not change (Appendix Table A2, Model 3). The second interaction terms between education and country showed the nation-specific job losses with lower level education than tertiary. In the case of the lowest level (ISCED 0-2), the biggest deficits appeared for Slovenia and Hungary (around 17 and 15 points in the ISEI score, respectively). Similar occupational status deficits were smaller (below 5 points) for the other countries. For the secondary educated, the estimates showed smaller losses in the occupational status with the exception of Hungary and Slovenia where the deficits were around 9–10 points in the ISEI score (Appendix Table A2, Model 4).

The third set of interactions on temporal differences in country variation in occupational status revealed varying patterns. Compared to the UK in 2007, occupational ISEI scores have deteriorated the most in Hungary and Spain, and more for 2014 than for 2009, partly due to the economic crisis but also to the existing rigidities in the labour market in these countries. Germany, Sweden and Estonia showed no significant deviation from the UK in 2014; while Denmark and Slovenia, both with a low level of inequality, were notable for positive estimates (Appendix Table A2, Model 5).

In order to provide a check of the results, sensitivity analyses were carried out.⁴ In addition to having the control variables in the models, the effects were re-estimated for men and women as well as for the youngest cohort vs. the older ones. Models were also run by countries separately. All in all, the differences in the impact of education for different socio-demographic groups are on the small or medium level but few results are noteworthy.

Regarding the risk of unemployment, there is not much gender difference on the average, though lower education harms women slightly stronger. In particular, there is no gender difference for the UK, Estonia, Hungary, when countries are analysed separately. The stronger effect of education holds for Denmark and Sweden, the two Nordic countries but education has a stronger impact on men in Germany and Slovenia, two countries classified as corporatist. For predicting occupational status, the loss in terms of ISEI is bigger for women with lower education. This holds for most analysed countries, except Slovenia.

In terms of cohort differences, additional analyses underline that the young cohort is more vulnerable than the older ones, both for the risk of unemployment and for the loss in ISEI. This seems to hold better under the circumstances of recession as temporal variation is bigger for the young cohort, though not in all countries. In Germany, Estonia, Spain and the UK the impact of the crisis on unemployment was very similar for young and older age groups. In Denmark and Hungary this effect was more pronounced for the young age group. But the crisis led to bigger loss in ISEI for the young people in most studied countries. The less educated in the young cohorts were the most vulnerable, especially in Germany and Sweden.

The aim of the paper was to explore the link between education and labour market outcomes in a comparative perspective. In parallel with a cross-national comparison, the paper compared three different time periods: 2007, 2009 and 2014 in order to investigate how the association between education and employment has been affected by the economic crisis. The analysis included eight European countries with differing institutional frameworks, both Western European and CEE nations. Our results confirm but also extend earlier findings regarding the impact of education on labour market outcomes.

First, the findings on the impact of education are consistent with the assumptions that an absence of pertinent qualifications increases the risk of unemployment and leads to lower occupational status. Clearly, both the institutional context of the labour markets and the educational systems in the countries affect the relationship. In order to provide a comprehensive picture of this national variation, an overview is presented in Table 2.

We found the impact of education on unemployment to be the strongest in Hungary and Estonia. These countries have different levels of inequality and flexibility. However, both are small post-socialist countries with so-called dependent market economies (Nölke and Vliegenthart, 2009). Both countries depend on the economic situations of other more developed European countries. At the other extreme, where the effect of education on the probability of unemployment is the lowest, are Denmark and Slovenia. Both Denmark and Slovenia have low levels of inequality, though they differ in the degree of flexibility. Moreover, the low level of inequality in Denmark, the well-known ‘flexicurity’ character of the country (Muffels and Luijkx 2008), has probably contributed to weakening the relationship between the deficit in education and the risks of being unemployed. In this regard, the analysis also uncovers the existing variation within the post-socialist countries: Slovenia vs. Hungary and Estonia.

Nevertheless, our results were not entirely in line with expectations. We expected that the effect of education on unemployment risk would be weaker in countries with low levels of inequality and high levels of flexibility. This is not the case for Sweden. Our results indicate that the level of inequality seems to have more important impact, especially in post-socialist countries. The analysis also shows that during the crisis the impact of education on unemployment increased in countries with high inequality scores, like Spain. Perhaps flexibility primarily moderates the effect of education on the long-term unemployment rate.

For the effect of education on occupational status, our findings suggest the strongest association to be in Hungary and Slovenia, and the weakest in the UK and Spain. This distinction accords with the position of the UK on the scale of labour market flexibility and the impact of education on the occupation. Low flexibility increases the impact, while high flexibility decreases it. Previous studies have also indicated a weak impact of education on labour market opportunities in Southern European countries (Müller and Gangl 2003). For Hungary and Slovenia, an additional explanation could be that education has a greater impact in countries with a vocational school system.

Germany, however, though having a similar educational system, does not fit this pattern since losses in occupational status caused by a lower level of education are smaller. The explanation of this contradiction may be that our analysis could not differentiate between the general and vocational tracks within secondary education. This led to the outcome that statistically Germany is in the same group as the two Nordic countries and Estonia, where the impact of education on occupation was at the medium level. This is not a surprising outcome for Sweden and Denmark, where a relatively lower effect of education was expected in accordance with these countries’ low levels of inequalities. Estonia, however, with its liberal institutional features and high level of flexibility, could be closer to the UK.

In summary, the results confirm our hypotheses about the weaker effect of education on labour market outcomes in countries with high levels of flexibility. On the other hand, the level of inequality seems to shape the effect of education on labour market outcomes with more variation: a stronger impact on unemployment and more negligible influence on occupational status.

Concerning the second topic of the paper, the comparison of the three time periods (2007, 2009 and 2014) confirms the hypothesis that the low educated respondents were hit harder by the economic crisis, in terms of unemployment risks and job status loss. As regards country differences, two groups can clearly be distinguished. The low educated suffered the biggest impact in two of the CEE countries, Hungary and Estonia, where the unemployment rate increased the most and the decline in job status was also substantial. As a further variation within the post-socialist countries, Slovenia coped with the recession better than the other two former socialist countries, at least in the initial phase of the crisis. The situation for most of the educational groups worsened between 2009 and 2014 in Slovenia, Hungary and Spain; while an improvement had already started in Estonia after 2009. At the same time, the magnitude of the economic crisis was relatively small for most educational groups in the other group of countries, in Denmark, Sweden and Germany and even in the UK.

Differences in both the degree of flexibility and the extent of inequality contribute to this division. High flexibility and a lower level of inequalities could provide some protection to people in the Nordic countries; while high flexibility could be helpful in the UK despite its liberal institutional and welfare system. In Germany, the corporatist features and stronger employment protection may explain why workers were less affected by the fluctuations in the labour market, in particular for the higher risk of unemployment. The same mechanism could, to some extent, have an effect in Slovenia but the country’s economy is much weaker than the German economy and the low level of flexibility proved to be a disadvantage. In Hungary and Spain, neither the flexibility of the labour market nor the extent of social safety nets could help to minimise the negative consequences of the crisis. Finally, Estonia was greatly affected by the recession in the beginning, probably because the liberal regime did not provide any protection. However, the flexibility in the labour market allowed Estonia to begin the economic recovery faster than in Hungary, Slovenia or Spain. (See also the GDP data in Table 1.) In conceptual terms, these country cases provided important insights into how labour market flexibility and social inequality interact in various institutional contexts.

In conclusion, we would like to underline two merits of this paper. First, we placed great emphasis on the theoretical approach of the institutional context of the relationship between education and two labour market outcomes, unemployment and occupational status, in times of economic recession and recovery in Europe. We elaborated on two mechanisms in this regard: flexibility in the labour market and the degree of inequality and proved their role in influencing the impact of education. From a theoretical perspective, we would like to underline that the institutional contexts play a particularly important role during periods of ‘irregular’ economic activity, such as those covered in this study. The labour market measures as well as the equality in society helped to counterbalance the negative consequences of the recession and to speed up the economic recovery at individual level.

Second, we placed great emphasis on investigating the research topic in the post-socialist context. We selected three CEE country cases in order to explore variation in these societies. Indeed, the analysis proved Estonia, Hungary and Slovenia to be significantly different and this final outcome could deepen the knowledge on how education and the labour market relate to each other in international context. In terms of theory, our results also contribute to the discussion on the forms of capitalism, and support the typology developed by Bohle and Greskovits (2012) where the distinction among these three countries is emphasised. Moreover, the empirical evidence also provides a refinement to the theory, particularly for the temporal variation in the links between education and the labour market, which were observed for Estonia and Slovenia, making them markedly different both from each other and from their Liberal (UK) and Corporate (Germany) counterparts.

Certain limitations of the study should also be mentioned. We compared only eight nations but focused on the CEE countries, and consequently had a reduced number of countries representing other regimes. An important goal was to analyse countries where we have more knowledge of how to interpret what the data reveal in statistical terms. We selected only two dependent variables, although there are other relevant labour market features. In the next research phase, we are planning to use Qualitative Comparative Analysis. This form of analysis preserves and uses the data of a case study as a configuration of characteristics as the basis of analysis rather than analysing isolated variables and their effects on a particular outcome extraneous to other influences.

No potential conflict of interest was reported by the authors.

Péter Róbert, Prof. has graduated in sociology at Eötvös Lóránd University (ELTE), received his PhD from the Hungarian Academy of Sciences (HAS) and has made his habilitation at ELTE University. He is fulltime Professor of Sociology at Széchenyi University, Győr, at the Department of Social Work and Sociology. He has been working for TÁRKI Social Research Institute as senior researcher since 1986. His current project titled Children in school: Well-being and beyond - The International Survey of Children’s Well-Being (ISCWeB). His research interests involve social stratification, educational inequalities, school-to-work transition, the perception of social inequalities and the subjective well-being. He has published in edited volumes by Oxford University Press, Princeton University Press, Edward Elgar Publishing Ltd., Routledge, Springer or Stanford University Press as well as in journals like RSSM, ESR, European Societies, ERE.

Ellu Saar, Dr, is a Professor of sociology at the Institute of International and Social Studies and School of Governance, Law and Society, Tallinn University, Estonia. She coordinated the EU Sixth Framework Project ‘Towards a Lifelong Learning Society in Europe: The Contribution of the Education System’ (LLL2010). She is now leading the project ‘Cumulative processes in the interplay of educational path and work career: explaining inequalities in the context of neoliberalization’ and participates in the EU Horizon 2020 projects ‘Encouraging Lifelong Learning for an Inclusive and Vibrant Europe’ (ENLIVEN) and 'Technological inequality - understanding the relations between recent technological innovations and social inequalities'. Her research areas are social stratification and mobility, educational inequalities and life course studies in comparative perspective.

Margarita Kazjulja (PhD), is a researcher at the Institute of International and Social Studies, School of Governance, Law and Society at Tallinn University. Her research interests include education, different aspects of social stratification in labour market, and individual strategies in coping with changes in a transition society.