Education systems and the dynamics of educational inequalities in low educational attainment: a closer look at England (UK), Finland, Luxembourg, and German-speaking Switzerland

General theoretical models such as the status attainment model and related empirical evidence (e.g. Jackson et al. 2005) show education as the main mechanism of the reproduction of inequalities in modern societies (Blossfeld and Shavit 1993; Breen 2004). Educational inequalities, defined as systematic dis/advantages in the access to educational institutions and educational attainment related to certain socially constructed characteristics, such as class or gender, translate into inequalities in status attainment, labour market chances, occupational status, income, subjective well-being, and life expectancy. Though evidence (e.g. meta-analysis of Hannum and Buchmann 2005) clearly shows that educational reforms have generally stimulated an educational expansion in the sense of an increased demand for education and an increased number of graduates of upper secondary and tertiary education, it is questionable whether educational inequalities have diminished to the desired extent. While gender inequalities have greatly decreased or even reversed (‘failing boys debate’) in some realms, inequalities related to social origin are still evident. Some scholars speak of ‘persistent inequalities’ (Blossfeld and Shavit 1993; Becker 2003). New evidence suggests a weakening association between social origin and educational attainment for many countries but also remaining disadvantages (Breen et al. 2010).

Educational inequalities are strongly linked to the characteristics of the educational system. The institutional settings act as incentive and opportunity structures that also pose limitations to educational attainment. More or less significant macro-factors of education systems include the degree of stratification/external differentiation¹ (age of selection, number of tracks, etc.), vocational specificity (general versus vocation-specific training), and degree of standardisation of schooling, teacher education, and educational certificates as well as the existence of programmes to increase permeability into higher education (e.g. alternative access opportunities to universities) (e.g. Allmendinger 1989; Pfeffer 2008; Van de Werfhorst and Mijs 2010). Importantly, education systems are linked to other macro-factors of the societal context such as welfare and gender regimes and value climate (value of education, acceptance of inequality, perceptions of justice, etc.).

There are two major approaches to analysing country differences: on the one hand, multilevel modelling has the potential to receive an insight into the effects of macro-characteristics of the education system on advantages and disadvantages in educational attainment on the micro-level (cf. Bol and van de Werfhorst 2013; Griga and Hadjar 2014; Hadjar and Buchmann 2016). On the other hand, following a less-is-more strategy (Ebbinghaus 2005) and focusing on a few country cases to take a closer look at different cohorts can provide a more holistic picture of stability and change (e.g. Breen et al.2010). In this article, we follow the second strategy. The main objective of this paper is to analyse the changing patterns of educational inequalities related to social origin and gender during the course of the educational expansion (focusing on cohorts born between 1925 and 1982) in Luxembourg, Finland, Switzerland, and England (UK). These countries represent different kinds of institutional settings and education systems – as will be explored in detail later. Employing data from the European Social Survey (ESS), our analyses focus on the cohort-specific links between social origin (measured by the highest educational level of the parents), gender, and educational attainment.

This study differs from earlier comparative work on educational inequalities in secondary schooling (Blossfeld and Shavit 1993; Shavit et al.2007; Breen et al.2010) on methodological grounds. We reduce the complexity of the education variables in response to low cohort sample sizes (binary outcome variable: low educational level, below International Standard Classification of Education [ISCED] 3) and present average marginal effects (AME), following recent concerns in regard to unobserved heterogeneity. While cohort analyses regarding educational inequalities are not new, what this study adds to previous analyses is that (a) it is an update to older accounts by including 1980s cohorts, (b) it focuses on two axes of inequality (social origin and gender), (c) it links institutional settings of the educational system and changes of these settings with educational inequalities, and (d) it includes Luxembourg and Finland as still rather under-researched countries with distinct education systems.

Linking education systems and inequalities in educational attainment, our theoretical inspiration comes from Boudon's (1974) framework of primary and secondary effects. From this concept, we draw the general assumption that institutional settings of education (and social) systems affect the individual-level mechanisms behind inequalities, that is, differences in educational achievement that are linked to class-specific resources (primary effects) and class-specific educational decisions at certain points of educational transition that are based on cost–benefit evaluations against the background of resources and constraints (Breen and Goldthorpe 1997; Becker 2003; secondary effects). Gender inequalities in education are – although less explicitly – also linked to secondary effects, that is, the question of the necessity of education for status attainment and income acquisition among women (DiPrete and Buchmann 2006; Hadjar and Berger 2011; Hadjar and Buchmann 2016). As Breen et al. (2010: 33) put it: ‘With the increasing labour force participation of women investments in education become beneficial in much the same way for daughters as for sons’.

One of the major changes experienced in most European countries and the US in the second half of the twentieth century has been educational expansion, that is, an increase in educational opportunities and a greater demand for education – in particular after the educational reforms of the 1960s (Blossfeld and Shavit 1993; Hadjar and Becker 2009). The main aim of the educational reforms, shared by most of these countries, was to boost economic growth by raising the general educational level and – from a socio-political perspective – reducing the unequal distribution of educational opportunities for different social strata, genders, and geographical regions. Another major driver behind the educational expansion is the increasing skill demand of the public service sector which particularly contributed to the increasing participation of women in education and in the labour market (Blossfeld and Hakim 1997). While educational expansion (Hadjar and Becker 2009) went along with increasing educational opportunities, empirical evidence (Blossfeld and Shavit 1993; Becker 2003; Hadjar and Berger 2011) regarding educational inequalities reveals an ambiguous picture: many scholars concluded from their empirical studies of industrialised Western countries that social inequalities appear rather persistent (Blossfeld and Shavit 1993; Solga 2002; Becker 2003), educational opportunities are still distributed unequally among different social strata, with children of less-advantaged social origin still representing the most disadvantaged group within the educational system. However, new evidence by Breen et al. (2010) does suggest a reduction of inequalities regarding secondary schooling. Furthermore, the decrease in educational inequalities of social origin at earlier educational transitions (e.g. to secondary school) goes along with marked inequalities at the transition into higher education (Shavit et al.2007). In sum, educational reforms seem to have led to a decrease in social inequality in many countries but not to the expected and desired extent. One major reason behind the remaining educational inequalities relates to secondary effects, namely rather stable class-specific perceptions of costs, benefits, and success probabilities of different educational paths (e.g. Becker 2003).

In regard to gender inequalities, differences in educational attainment between male and female students have reduced for primary and lower secondary school level, less for vocational education and higher education (Hadjar 2011). As DiPrete and Buchmann (2006) and Breen et al. (2010) show, women appear to have benefitted most from educational expansion. They made good use of the new educational opportunities, motivated by the increased chances on the labour market and a lesser dependence on the marriage market, caused by changes in the professional structure and a changing image of women. All in all, disadvantages for women in the educational system have largely been eradicated in most European countries since the 1960s, with males facing some new disadvantages, whereas inequalities regarding social origin have not diminished to the desired extent, resulting in a persistent level of inequality.

The comparison of educational systems is a key issue in sociological research on educational inequalities. A macro-level perspective centres on three characteristics that are particularly relevant for the level of educational inequalities (Allmendinger 1989; Müller and Shavit 1998; Kerckhoff 2001; Pfeffer 2008; Van de Werfhorst and Mijs 2010; Hadjar and Berger 2011): stratification (number of different school types that exist in parallel, age at selection, permeability), standardisation (variability in the quality of education between schools and regions, e.g. in regard to spending, governance, teacher education), and vocational specificity (connection between educational institutions and the professional sphere). Empirical evidence shows that stratification has a comparably large impact: while highly stratified countries such as Germany, Austria, and Switzerland are characterised by a strong impact of social origin on educational attainment, Nordic countries – with less-stratified educational systems – show relatively low educational inequalities (e.g. Müller and Karle 1993; Pfeffer 2008). Gender inequalities are also lower in less-stratified countries (Breen et al.2010; Hadjar and Berger 2011).

In addition to exploring the role of the education system, we will also look at the type of welfare state system. Education systems and welfare state regimes are linked (Allmendinger and Leibfried 2003). Welfare state regimes that direct their policies very much towards abolishing inequalities like the social-democratic or Scandinavian type, at the same time, employ educational policies to equalise differences in educational attainment. In particular, we look at some features of the different welfare regimes that are conducive to female labour force participation, such as the availability of institutional childcare (Lewis and Ostner 1994; Walby 2004).

The following description of the differential macro-characteristics of the four countries focuses on characteristics that may have been relevant for the birth cohorts considered in our analyses (cohorts 1925–1982) and not on recent changes. First we describe features of the education systems, before dealing with the welfare systems.

The educational system in England contains great diversity, with local education authorities responsible for funding and admission for state schools, and the public and private sectors operating in parallel (comparably low standardisation level). Currently, compulsory schooling encompasses children from 5 to 16. Remnants of the tripartite secondary school structure (grammar, technical, secondary modern schools) of the 1944 Education Act persist alongside comprehensives and fee-paying independent schools. Although the English education system is stratified to an intermediate extent with a dominant comprehensive secondary schooling, it is prone to inequalities: the state comprehensive schools are highly socially segregated and even more socially selective than grammar schools with regard to the students’ social background (Smithers and Robinson 2010). New vocationalism of the late 1970s until the late 1990s emphasised the link between education and the economy but failed to achieve parity of esteem between the different secondary tracks. National curriculum and national assessments have been in place since the 1988 Education Reform Act. While universities in the UK in 2008/2009 enrolled on average one-third of their students from a working-class background (routine or manual occupations), the figure is below one-fifth in the top universities according to the Higher Education Statistics Agency (https://www.hesa.ac.uk).

It is precisely the concerns over social and regional equality in schooling in Finland that sparked a major reform of the educational system in the late 1960s. The resulting comprehensive school granted equal education for all children through the first nine grades (up to age 16) – making the Finnish education system one of the least stratified systems in the world. Although schooling is only compulsory between the ages of 7 and 16, almost all Finnish children attend (free) pre-schooling from the age of 6. The decentralisation trend of the late 1980s shifted decision-making power to the municipalities and granted schools greater autonomy. Curricular specialisation after the 1994 curriculum reform (Seppänen 2003) and the 1998 Basic Education Act (Rinne et al.2002) have given parents greater choice over schools, creating local ‘public school markets’ in places, particularly in big cities (Seppänen 2003). Vocational training in Finland is delayed until after compulsory schooling, and the traditional academic and vocational tracks remain separate through the upper secondary stage until tertiary education.

The education system in Luxembourg (compulsory schooling from 4 to 15 years of age) is stratified. After eight years of pre- and primary schooling, students are selected to different, clearly separated school tracks (préparatoire, l'enseignement secondaire technique and l'enseignement secondaire) at the age of 12 (European Commission 2010). Class repetition rates are rather high, more than 20% of the primary and secondary school students (MENFP/EMACS 2010). Regarding educational inequalities, working-class students and boys in particular have a lower competency level and are overrepresented in low school tracks in Luxembourg (MENFP/EMACS 2010). The reform attempts since the debates in the early 1960s have targeted the low financial expenditure on education and the curriculum (Rohstock and Lenz 2012). Although the aim of a comprehensive restructuring was not reached, compulsory schooling was prolonged from 7 to 9 years, and new curricula focusing on sciences and languages as ‘principal subjects that OECD policy aimed to specifically promote’ (Rohstock and Lenz 2012: 12) were introduced. The notion of stratification serving the needs of specific ability groups has been strong for decades. In regard to tertiary education, Luxembourg is a late bloomer: after the founding of a tertiary institution for applied sciences in 1974, the University of Luxembourg was only founded in 2003. Until then, students had to go abroad for their studies.

Instead of a Swiss education system, there are regional education systems: around 90% of educational policies are developed on the cantonal and local levels (Hega 2000), although there are attempts to harmonise the institutional settings in the German–Swiss cantons and increase the standardisation level. In the German-speaking regions, education systems are highly stratified in regard to the birth cohorts under consideration. A common feature is compulsory schooling between the ages of 6 and 15, with recent tendencies to introduce compulsory pre-schooling. Throughout the twentieth century, students were selected to three different secondary school tracks according to achievement level after four to six years of primary education. Direct transition to upper secondary education is granted only through the highest track and is almost impossible from the lowest school track (e.g. ‘Realniveau’). The intermediate school track represents varying permeability regarding the transition to upper secondary education. The education systems in the German-speaking regions are also rather vocationally specific with a dual system of vocational training. The course of the educational expansion in Switzerland has been characterised as rather slow, both in regard to the increase in participation and to educational inequalities with remaining substantial links between class respectively parental education and educational attainment (Buchmann et al. 1993, 2007), although there are also signs for a decrease in the impact of social origin on educational attainment in favour of disadvantaged social classes (see the results of Becker and Zangger 2013 based on census data). However, a rather large proportion of people attended post-secondary educational institutions below University level that imparted knowledge on the level of tertiary education (Müller et al.1997: 209).

Both Switzerland and the UK represent liberal welfare states (Esping-Andersen 1990), characterised by a dominance of the market, a central role of the private sector and a low level of state intervention, with assistance based on means testing. Given that policy measures in Switzerland are aimed at the prevention of poverty rather than reduction of inequality, the level of social inequality is high. Some scholars stress that the Swiss welfare regime is a mix of liberalism – since the welfare state is less developed, Calvinism is strong and workers’ representations and the social-democratic political representation are comparably weak – and conservatism – since there are regional health service schemes, pension, and unemployment schemes (Schröder 2013; Trampusch 2010). Despite this mixture of liberal and conservative or continental elements, in sum and comparison Switzerland can be rated as a liberal or post-liberal welfare regime (Trampusch 2010). Luxembourg is a conservative welfare state regime, with a high level of social care but also an institutional structure that supports the reproduction of inequalities (Esping-Andersen 1990). Finland, on the other hand, represents a social-democratic welfare state, characterised by a strong presence of the state, universal social rights, and fairly small inequalities in terms of income, social class, and gender.

Since women's rights were recognised very late in Switzerland, with the right to vote introduced as late as 1971, gender differences in education in favour of men began to decrease later than in other European countries. In fact, gender equality hardly featured in educational reforms in the 1950s or in the 1960s (see Hecken 2006: 124; Rieger 2001: 54f). While the proportion of females in higher and further education in Finland is clearly higher than the European average, the study subjects, and subsequently the labour market, exhibit strong gender segregation. As a result, also work contracts are increasingly gendered, with more women on temporary or short-term contracts (Lilja and Savaja 2013). The gender regimes both in Luxembourg and in the UK represent a stronger ‘male breadwinner’ model, although the Luxembourgish gender regime is more similar to the French than the German gender regime, developing towards a moderate ‘male breadwinner model’ (Lewis and Ostner 1994). Regarding childcare institutions outside the family at the age of three, which is linked to female labour force participation, Switzerland (9.6%) exhibits a much lower participation rate than the UK (82.4%), Luxembourg (69.3%), or Finland (68.5%) (OECD 2008). The female employment growth rate in Luxembourg has been one of the highest in the EU27 countries during the period 1999–2008 (Valentova 2013). The Swiss system is a typical ‘male breadwinner’ model. Although female labour force participation in Switzerland is comparable to other European countries, the rate of women working full-time is much lower. In 2005,² 60% of the female workforce aged between 20 and 64 in Switzerland were working part-time (Luxembourg 38.2%, UK 42.6%, Finland 18.6%; Eurostat 2013). Due to the extent of welfare benefits and public services, female labour participation rates for this age group in Finland, at 72.5% by 2012, are nearly as high as for males (75.5%), while in the UK and Switzerland, women's employment rate lagged men's by nearly 12%, in Luxembourg by over 14% (Eurostat 2013).

Considering the theoretical framework, we expect to see decreasing links between educational level (risk of low education) and either social origin or gender over successive birth cohorts. The changes in gender inequalities are expected to be much more prominent than the decreasing social inequalities. As for the macro-characteristics considered above, we expect lowest gender inequality in Finland with its social-democratic welfare system. Given the rapid growth in women's labour participation rates in Luxembourg, there should be a sharp decrease in low levels of education for the younger cohorts. Although women's labour participation rates in Switzerland are low, we expect to see the general trend of diminishing gender inequality in educational opportunities.

We also expect social inequality to be lowest in Finland, where the public school system is of comprehensive kind and, thus, the level of external differentiation/stratification of the education system – defined by the number of parallel school tracks, age at selection – is low. Coinciding with the comprehensive school reform, a notable decrease in social inequality in Finland should appear for the cohorts born after 1965. On the contrary, given the structured education systems in (German-speaking) Switzerland and Luxembourg, we expect to see stronger social effects and resulting inequality there. With previous empirical findings suggesting social segregation in the diverse school system in England, we expect to find strong social effects even for the younger cohorts. Given that we do not distinguish between academic and vocational post-compulsory schooling, these effects may be attenuated in our sample.

Bringing the conceptual reflections and country descriptions together, we derive the following more specific working hypotheses:

Our data are comprised of two waves (2002/2003, 2004/2005) of the ESS. The ESS is a repeated cross-sectional survey of 30 countries, with Luxembourg participating only in the first two rounds. The generated random samples are representative of all persons aged 15 and over residing in private households (European Social Survey 2012). We use the pooled dataset to ensure a maximum number of cases on the country and on the individual level. We include all people born in the respective countries.

The variables modelled have been operationalised as follows: our binary dependent variable low educational level relates to the educational level achieved up to the time of data gathering and was coded ‘1’, if the highest educational attainment of a person was below upper secondary education (ISCED 0, 1 or 2), that is, the respondent only completed compulsory education at most. Social origin is operationalised in terms of the highest parental educational level, as the ESS waves – our analyses are based on – do not include information on class positions. We used two different operationalisations: while for our graphs showing the cohort-specific inequalities, we only use a binary social origin variable (‘1’ for low social origin, parental educational level ISCED 0, 1, or 2), we use an operationalisation based on three categories in our multivariate models: tertiary educational level (ISCED 5, 6) as reference category, intermediate educational level (ISCED 3, 4), and low educational level (ISCED 0, 1, 2). The gender effect is included as a binary variable with ‘1’ for female respondents.

To reconstruct changes during the educational expansion, we focus on birth cohorts born between 1925 and 1982 and estimate cohort-specific models. Since the educational expansion started in the 1960s in most European countries (e.g. Blossfeld and Shavit 1993; Hadjar and Becker 2009; Shavit et al.2007), the 1955–1964 born people are the first cohort to have benefitted entirely from the new educational opportunities. The earlier born cohorts could be conceptualised as pre-expansion cohorts in regard to the countries under consideration. To reduce the number of interaction terms modelled in our multivariate models, we use the metric year of birth in the analysis of the dynamics of inequalities over cohort succession in Table 1.³

As we pooled two waves of the ESS, we control for ESS round in the complex analyses. The dummy variable refers to the wave 2004/2005 (reference category: 2002/2003).

First, we look at the cohort-specific proportions of people with low level of education to reconstruct the course of the educational expansion in the four country samples. Figure 1 shows the percentages of people with a low educational level (below upper secondary education, ISCED 0–2). In all four countries, the number of people with a low educational level decreases in the course of the educational expansion. Switzerland appears to be the country with the smallest overall proportion of low-educated people which may well owe to the high number of vocational tracks at the upper secondary level (Riphahn 2003). However, the strongest decrease of low-educated people –and therefore the highest increase in educational level – has taken place in Finland. Considering the 1925–1934 born cohorts and the 1975–1982 born cohorts, the low-educated group shrinks by more than 90%. The educational expansion appears to be much less intense in the UK, starting out from the same or even slightly higher percentage of low-educated people, but with about 40% of people with a low educational level in the youngest cohort under consideration (reduction by about 54% over cohort succession). In Switzerland, the proportion of low-educated people also decreased by about 65%, in Luxembourg by 78% over cohort succession.

As shown, an educational expansion with a more or less drastic reduction in the number of people with low level of education is found in all four countries. The main focus of our analyses, however, is on educational inequalities related to social origin and gender. Figure 2 shows results of binary logistic regression models regarding the effects of social origin (parental educational level) and gender (female gender) on the probability of attaining a low educational level, that is, having received education below upper secondary level. The graph of the AME – derived from separate cohort – and country-specific models – gives a visual impression of the changes.

First looking at social origin as axis of inequality, England (UK) shows rather high inequalities over cohort succession. Whereas in the other countries the link between social origin and educational attainment has reduced drastically, in England, the probability of having no post-compulsory education is still more than 30 percentage points higher for people with low-educated parents than for others among the English sub-sample born between 1975 and 1982. Finland, Luxembourg, and German-speaking Switzerland, on the other hand, started with disadvantages of more 30 or more percentage points for people of low social origin among the 1925–1934 cohorts, and ended with no significant impacts of social origin among the 1975–1982 cohorts (Luxembourg, German-speaking Switzerland) or only a very marginal effect (Finland).

Considering gender inequalities, the visual inspection reveals at least two patterns: while England and Finland were always characterised by rather low gender inequalities, Switzerland and Luxembourg experienced a strong decline in gender inequalities in favour of men over successive birth cohorts. However, all four countries exhibit distinct patterns: in the English sample, in the 1935–1944 cohorts and the youngest cohort born between 1975 and 1982, women show a significantly higher risk of leaving the education system with only low educational degrees. In Finland, the oldest cohort (1925–1934) is characterised by gender inequalities in favour of men, while gender inequalities in favour of women are noticeable regarding the 1945–1954, 1955–1964, and the 1965–1974 cohorts; interestingly, there are no significant gender differences among the youngest cohort born between 1975 and 1982. In Luxembourg, there has been an almost linear reduction of female disadvantages from the oldest cohorts with a 33% higher risk for women to end up with only a low educational level to the 1965–1974 and the 1975–1982 born cohorts with no significant gender inequalities. The same applies to German-speaking Switzerland, and gender inequalities in educational attainment have reduced rather slowly, from up to 23 percentage-points higher risk for women to be low-educated among the oldest cohort (1925–1934) to zero gender effect among the youngest cohorts.

Conclusions in regard to significant cohort and country differences can be drawn from more complex models including interaction terms in regard to changing inequalities over successive birth cohorts (Table 1).⁵ While the (significant) interaction terms indicate significant cohort differences, country differences can be identified considering the confidence intervals of the coefficients. If the 95%-confidence intervals do not overlap, the coefficients do significantly differ. However, if the intervals do overlap, ultimate conclusions cannot be drawn, as this is only a week indication that there may be no significant difference between both coefficients, but the opposite may also be true (Schenker and Gentleman 2001; Knezevic 2008). Thus, we will only draw conclusions from confidence intervals that do not overlap while being cautious with the interpretation of overlaps. Evaluating the general level of educational inequalities in the different countries (i.e. the main effects of social origin and gender), there are some differences in inequalities in regard to social origin and to gender. A low parental educational level increases the risk of low educational attainment in all four countries. In Finland, Luxembourg, and German-speaking Switzerland, originating from a family with an intermediate educational background also carries a higher risk of low educational attainment. However, this does not apply to England. Switzerland shows the lowest inequalities related to social origin in differing significantly from Luxembourg with the highest level of inequalities (if all birth cohorts are considered). Finland is very close to Switzerland in the amount of inequalities related to social origin. In regard to gender inequalities, Finland shows marginal overall inequalities in favour of women and, thus, distinctively differs from all other countries. Gender inequalities in favour of men are highest in Luxembourg – significantly differing from England and Finland, but maybe not from German-speaking Switzerland with a comparable gender inequality level. Interaction terms of birth cohort and social origin (low or intermediate parental educational level), respectively, gender (female gender) indicate how disadvantages of the respective groups compared to the reference groups (tertiary-educated parents, male gender) decreased, stayed stable, or increased. Although the direction of the effects in almost all cases indicates decreasing inequalities, that is, increasing disadvantages for people of low social origin and women, only some of the effects appear to be statistically significant. Finland is the only country where the disadvantage of the group characterised by a low parental educational level decreased significantly compared to the offspring of academics (high parental educational level). In England, the gap between the people with an intermediate parental educational background and those with a tertiary parental background even increased over successive cohorts, while the negative distinction of the low social origin people stayed the same. Gender inequalities (in favour of men) were reduced most efficiently in Luxembourg and Finland.

Limitations of the study and problems of comparisons of education systems notwithstanding, we find that in all four countries under consideration – Finland, Luxembourg, German-speaking Switzerland, and England – the number of people with compulsory schooling declined over cohort succession and therefore in the course of the educational expansion. Whereas in Finland, Luxembourg, and (German-speaking) Switzerland educational inequalities related to social origin have been diminished – with no educational inequalities among the youngest cohort born between 1975 and 1982 – in England, there has been no change in the disadvantage of people from lower backgrounds. This supports Hypothesis 1a for three countries. However, in the more complex models, a significant decline of educational inequalities related to social origin only occurs in the Finnish sample where the gap between people with a low parental educational background and people with a tertiary parental educational background is closing (significantly) over successive birth cohorts. Hypothesis 1b is also not supported by our findings: social origin-related inequalities are, across the cohorts, and on average, not significantly lowest in Finland with its lowly stratified education system. In general, the countries do not differ much in this regard. Although Finland shows at first sight lower inequality levels than England or Luxembourg (although these are not significant country differences) and is the only country where inequalities linked to social origin have been reduced, Switzerland showed the lowest on-average inequality level. All in all, Finland did not appear as distinct as expected in regard to inequalities and its change. Hypothesis 2a in regard to gender inequalities is supported for three out of the four countries: the association between gender and educational level (risk of low education) decreased significantly in Finland, Luxembourg, and German-speaking Switzerland, and only in England it remained stable. In line with Hypothesis 2b, Finland shows significantly lower gender inequalities (in favour of women) than the other countries. Furthermore, in the Finnish, in the Luxembourgish and (less pronounced) in the Swiss samples, the decline of gender inequalities – the decreasing male advantage – over successive birth cohorts appears to be statistically significant.

Our results are somewhat consistent to previous studies – although there is no study that covered the same four countries at the same time: the finding that Switzerland appears to be more equal than expected goes well with Becker's and Zangger's (2013) findings and suggestion that speaking of persistent inequalities is inadequate in regard to the Swiss case. The impression of rather persistent educational inequalities related to social origin in Britain is consistent to Breen et al. (2010). However, results of Pfeffer (2008) indicated a stronger distinction of Finland (with lowest equality levels) and Switzerland (with highest equality levels) than our results do.

How to interpret these findings in terms of the theoretical considerations regarding education system and welfare and gender regimes? First of all, the country differences provide some, but not a strong indication that macro-characteristics of the education and social systems seem to be linked with differences in educational attainment – conceptualised, although not tested, as expressions of primary and secondary effects (Boudon 1974). Macro-settings seem to at least slightly impact the magnitude of achievement differences between children of different social backgrounds and genders, and how much social classes and genders differ in their educational motivation and perceived success probability in regard to different educational pathways and finally their educational decisions. Looking at both gender and social inequalities and its changes simultaneously, Finland shows rather low levels of inequality and experienced the largest change as both inequalities related to social origin and gender inequalities profoundly decreased. This is in line with our expectations regarding the performance of a social-democratic welfare regime with a low-stratified education system and a gender regime closest to the ‘dual breadwinner model’. The Finish setting seems to be very successful in equalising achievement differences (and at the same time meet high standards) and – as the school system is a comprehensive one so that no educational decisions need to be taken at early stages – in reducing secondary effects. What we did not expect, however, was the relatively low inequality level among the younger cohorts in (German-speaking) Switzerland with its strongly stratified education system. The stigmatisation of people from lower social strata may be lower in Switzerland and the permeability of the school system higher than initially expected. Presumably this is also due to the higher supply of qualified intermediate apprenticeships and of post-secondary, non-tertiary educational paths (Müller et al.1997). The high and persistent social inequalities in the English sample were also not expected to this extent, since the English school system is stratified to an intermediate extent and the liberal welfare regime – in contrast to the conservative regime in Luxembourg – is not ‘naturally’ directed to the maintenance of hierarchy and inequality. However, our findings show that this type of welfare regime and education system (together with other factors such as the importance of private and elite education in the UK) generates persisting social inequalities. The same is true for the Luxembourgish education system and its conservative welfare regime in regard to inequalities owed to social origin. Finally, the rather slow decline in gender inequalities in Switzerland supports our assumption that a strong ‘male breadwinner’ gender regime with a lower (full-time) labour force participation also reflects in gender inequalities in education to the disadvantage of women.

In general, our findings indicate that the education system, welfare, and gender regimes jointly influence the extent and the persistence of inequalities along different axes. However, divergences also indicate that more country or even region-specific institutional characteristics need to be taken into account. There are no simple system-specific trends, but non-linear developments driven by different factors and particular combinations of these factors with variable (even region-specific) consequences.

Interpreting these findings, several limitations need to be taken into account. Firstly, our analyses focused on secondary schooling with the dependent variable of leaving the education system with no or only a compulsory school degree below the upper secondary general educational level. Focusing on other stages of the educational career (e.g. tertiary education) would have revealed other patterns. Secondly, results of comparative educational research are sensitive to classification effects, for example if an intermediate school degree like the German middle school degree is classified as high or low educational level. We tried to avoid this problem by employing the category ‘low educational level’ (only compulsory school or less, which in the four countries refers to nine years of schooling) as the dependent variable. While a dichotomous category improves comparability in regard to different countries, it is problematic for the analysis of inequalities, since it is an over-simplification of the education variable that may affect the results (Müller and Klein 2008). Thirdly, larger sample sizes in regard to the different cohorts would allow for more complex analyses and increase the validity of the results (Breen et al.2009). Fourthly, in particular among the younger cohorts, it cannot be ruled out that people obtain higher educational degrees later in their lives (Jacob and Weiss 2011); life-course data are needed. Fifthly, consequences of low educational level are not the same across countries (Müller and Shavit 1998). In countries with a high degree of correspondence between education and professional structure like in Switzerland, low education is more closely linked to low status and low income and lower life chances than in countries where the link between school and work is not as strong (e.g. the UK).

Finally, it is clear that providing ‘more education’ does not necessarily lead to a reduction of inequalities. Our results hint to the conclusion that educational inequalities related to social origin and gender are lowest if the education system is lowly stratified (with no or late selection) and the social system is based on a ‘dual-earner model’. Eventually, discovering how macro-characteristics shape educational inequalities is challenged by the availability of data and statistical methods for dealing with complexity. This is also a major challenge that undermines the obvious next step of testing whether the decrease of inequalities is due to primary or secondary effects. These challenges notwithstanding, the next step in our endeavour to analyse the impacts of education systems is to explore the effects of macro-characteristics through multilevel models.

No potential conflict of interest was reported by the authors.

Andreas Hadjar is a professor at the Institute of Education and Society, University of Luxembourg since 2010. Main research interests: Sociology of education, political sociology, international comparisons, social research methods. He studied sociology and journalism at Leipzig University (DE) and Glasgow University (UK). From 2000 to 2004, he worked at the Sociology Department, Chemnitz University of Technology (DE). From 2004 to 2010, he was a lecturer in sociology of education at the University of Bern (CH). Recent international publications: Expected and Unexpected Consequences of the Educational Expansion in Europe and the US (Haupt, 2009; co-editor), ‘Parent-child value similarity and subjective well-being in the context of migration’, Family Science 3, 2012 (co-author); ‘Migrant Background and Higher Education Participation in Europe: The Effect of the Educational Systems’, European Sociological Review 30, 2014 (co-author).

Erica Uusitalo is a postdoctoral fellow at the Institute of Education and Society, University of Luxembourg as of 2013. Main research interests: Sociology of education, migration, stratification, and social mobility. She studied sociology at the University of Aberdeen and the University of Sheffield (UK). Her current project is focused on educational achievements and labour market outcomes for migrants in Luxembourg.