Educational inequalities in general health: Does the curricular tracking system matter?

Abstract

Educational tracking amplifies social inequalities in a wide range of outcomes. From an institutional perspective, the current study examines whether this holds for educational disparities in general health. To investigate this question, we use information from the European Social Survey (Rounds 1–8: 2002–2016) for individuals between the ages of 18 and 45 years (N = 99,771) in 22 European countries. The estimated three-level hierarchical models indicate that tracking is indeed associated with larger educational inequalities in overall health. Individuals who have attained vocational education fare worse in terms of general health than do individuals who have pursued academic qualifications. However, the strength of association is much higher in countries with highly tracked systems (e.g. Germany and Czech Republic) than it is in countries with more comprehensive systems (e.g. the United Kingdom and Scandinavian countries). This result suggests that health inequalities between educational groups can be reduced by reorganizing secondary educational systems.

Keywords

Education self-rated health curricular tracking cross-national differences multilevel

Introduction

Educational inequalities in health are among the most consistent findings in the social–epidemiological literature. In general, people with higher qualifications report higher levels of perceived health, less psychological distress, and a lower incidence of disease than is the case for their counterparts with less education (Groot and van den Brink, 2006; Knesebeck et al., 2003).

Comparative research also indicates that the size of the relationship between educational attainment and health varies from one country to another (e.g. Mackenbach et al., 2008; von dem Knesebeck et al., 2006, 2011). To date, these between-country differences have been attributed to differences in welfare regimes (Beckfield et al., 2015; Eikemo et al., 2008), healthcare systems (Gesthuizen et al., 2012; Präg et al., 2017), and labour market policies (Bracke et al., 2014; Dudal et al., 2017). Nevertheless, previous studies have systematically ignored the possibility that differences in various aspects of institutional arrangements in education themselves may play a role. This lack of knowledge is regrettable, as it hinders a complete understanding of the mechanisms that reinforce health inequalities between those with less and more schooling. It is all the more unfortunate, given the recent emphasis on the relevance of educational policies for improving population health (Low et al., 2005; Marmot, 2005).

In this paper, we focus on the role of curricular tracking – a form of sorting students into secondary education according to individual ability levels (Gamoran and Mare, 1989). Previous studies have revealed that rigid tracking systems have negative effects in terms of social inequality. For example, curricular differentiation has been found to increase inequalities in student achievement (Brunello and Checchi, 2007; Montt, 2011), political engagement (Janmaat et al., 2014; van de Werfhorst, 2017), and labour market opportunities (van de Werfhorst, 2011; van der Velden and Wolbers, 2003). In line with this research, the current article investigates whether educational disparities in health are reproduced through differences in the national context of secondary school differentiation. More specifically, we set out to answer the following research question: Are educational inequalities in health between people educated in vocational tracks and people who have pursued academic qualifications greater in countries with strict educational tracking, as compared to countries with more comprehensive educational systems?

This research question moves beyond the quantitative definition of educational attainment that dominates the social–epidemiological field. That is to say, a person’s level of schooling is not defined by years of education, but rather by the type of education obtained. However, perhaps a more important contribution of this study is its effort to introduce an institutional perspective as a broad conceptual framework for studying the relationship between education and health. Although such an approach is largely absent from comparative health sociology research, it may nevertheless be very helpful for explaining country-based differences in the ability of schooling to shape health benefits.

Background

The micro context: Health inequalities between people from different tracks

In past decades, an extensive body of literature has focused on the relationship between years of education and health. Based on a human capital perspective, health sociologists stress that higher levels of education are associated with better health through active coping, healthy life styles, and better access to material and social resources. However, less attention has been given to the idea that health inequalities may also be a function of qualitative differences between educational types (i.e. vocational and academic forms of education). Therefore, our study first explores the relationship between educational track and health on the individual level.

Several reasons make it plausible to expect health inequalities between people educated in vocational tracks and people educated in academic tracks. Perhaps the most direct explanation is that graduates from vocational and academic forms of secondary education are socialized into different health behaviours that are carried into adulthood (de Looze et al., 2013; Rathmann et al., 2016). In previous studies, vocational secondary school students have been found to be more likely to be daily smokers (de Looze et al., 2013), to drink at higher-risk levels (Berten et al., 2012), and to display more sexual risk behaviour (van Rossem et al., 2010) when compared to academic secondary school students. A critical reason for this social patterning could be that students enrolled in vocational education attempt to compensate lack of recognition by creating a ‘school-alienated’ peer climate that can be considered as detrimental to health (Nutbeam et al., 1993).

Further, more indirect explanations can also be invoked. A first argument in this regard concerns differences in curriculum covering. Ten Dam and Volman (2003) and van der Werfhorst (2017) argue that vocational tracks generally focus more on practical activities and work-relevant skills, whereas academic tracks give greater emphasis on cognitive and critical thinking skills. According to the theory of learned effectiveness (Mirowsky and Ross, 2005), the latter competences are tools needed to develop a sense of self-direction that protects against ill health. A second argument is that health-risk behaviours are not the only consequence of the low status of vocational education. People educated in vocational tracks may also suffer from damaged self-esteem (van Houtte, 2005) and display high levels of fatalism or sense of futility (Schafer and Olexa, 1971; van Houtte, 2016) as a result of feelings of relative deprivation (van Praag et al., 2017). Given that these negative feelings signal a lack of self-efficacy (Bandura, 1977) – which is a well-known determinant of poor health – it can be expected that people graduated from lower tracks experience worse health outcomes than do people graduated from higher tracks. A third argument is that it may be that the relationship between educational track and health points to a selection effect: Children in vocational tracks are different from children in academic tracks, because of selection factors that influence both track placement and health (van de Werfhorst, 2017). The most important selection factor in this regard is social background. It is widely recognized that children from parents with little education are less likely to be assigned to academic tracks than are children from parents with higher degrees (Boone and Van Houtte, 2013), as well as that they tend to have poorer health in adulthood (Kestilä et al., 2006).

Hypothesis 1: Based on these arguments, we hypothesize that individuals graduated from vocational tracks report worse health than do individuals graduated from academic tracks.

The macro context: The impact of the degree of tracking

All European secondary school systems maintain a certain distinction between vocational and academic education. However, considerable cross-national variation exists in the degree to which this distinction is institutionalized (Bol and van de Werfhorst, 2013a). That is, countries differ in the age of first selection and in the number of tracks available to students in (upper) secondary schooling. For instance, whereas countries such as Germany, Austria and Belgium currently have highly differentiated educational institutions, countries such as the United Kingdom and the Scandinavian countries are characterized by less rigid tracking systems (Rathmann et al., 2016). The main objective of the current study is therefore to explore how far these differences in the macro-institutional context of educational tracking can be linked to between-country differences in health inequalities between educational types. Are countries with strongly differentiated educational systems associated with larger health inequalities between people educated in vocational tracks and people educated in academic tracks? There are several reasons to expect that this may be the case.

The first group of arguments claims that the individual-level mechanisms described above are stronger in highly tracked settings. For instance, strong institutional tracking can be argued to intensify curriculum differences between vocational and academic tracks (van de Werfhorst, 2017). In countries with higher levels of educational differentiation, vocational tracks may pay even less attention to skills that are relevant to health promotion. Also, the stronger the experience of being grouped in separate tracks, the more children are socialized into different school cultures and related health behaviours (van Houtte and Stevens, 2009). Another argument that focuses on the strengthened importance of individual-level mechanisms points out that relative deprivation is more pronounced in countries with early and strong tracking. In such systems, public preference for academic education that leads to university degrees is all the more evident whereas lower secondary tracks that prepare students for manual occupations are generally undervalued (Bol and van de Werfhorst, 2013a). The macro context of intensive tracking therefore makes people educated in vocational tracks probably more vulnerable to relative deprivation (Wilkinson, 2002) and to its health consequences (such as chronic stress, and feelings of fatalism) (Pampel, 2002).

The last argument that stresses the increased influence of individual-level factors focuses on the intensified importance of social background – as a selection factor – for track placement in strongly differentiated educational systems, when compared to more comprehensive ones (Brunello and Checchi, 2007; Montt, 2011). Two different explanations dominate the field. Cultural capital theorists have stressed that one property of strongly stratified educational systems is that they are unclear and hard to navigate (Boone and van Houtte, 2013; Lareau, 2000). In particular, parents with less education tend to be less informed about the nearly irreversible nature of choices made at the onset of secondary school, which subsequently makes them less ambitious with regard to the track placement of their children. In contrast, scholars adhering to the theory of rational action suggest that parents from higher social backgrounds base their educational decisions on information about potential labour market returns of various degrees (Holm and Jæger, 2008). Given that the latter are more obvious in countries with highly stratified educational systems (Allmendinger, 1989), well-educated parents in such countries characteristically assign their children to higher tracks. Regardless of the underlying reasons, it can be concluded that the selection effect of social background in the relationship between educational track and health becomes more pronounced with higher degrees of curricular tracking.

The second group of arguments considers mechanisms that are unique to societies that more rigidly select children into separate educational types. It looks at the institutional effect of early and strong tracking on the labour- and marriage-market opportunities of the different educational types. Previous research has shown intensive tracking to influence the association between schooling and occupation because it frames the institutional context in which hiring decisions are made (Allmendinger, 1989; Levels et al., 2014). In contrast to more comprehensive systems, in which career advancements are less defined within secondary school curriculum, educational systems of intensive sorting induce employers to rely predominantly on credentials when selecting graduates to work in their practices, as they are taken as an accurate reflection of an applicant’s level of skills and ability. In this context, education functions as the major channel for labour market allocation. Individuals graduating from higher tracks thus commonly achieve more occupational status than their peers from lower tracks (Bol and van de Werfhorst, 2013a). This situation would not be problematic if no health advantages were associated with position on the labour market, but such is not the case (Ross and Mirowsky, 1995).

In addition, tracking can be equally argued to affect the marriage market opportunities of the different educational types. Central to this argument is that schools constitute one of the core arenas of partner selection (Kalmijn, 1998). When schools are characterized by a higher degree of educational differentiation, future partners have fewer opportunities to meet each other outside their own educational peer groups (Blossfeld and Timm, 2003). In other words, educational homogamy is the highest in countries with strongly stratified educational institutions. Combining this observation with the social–epidemiological knowledge that marriage constitutes a critical context in which the education-based resources of the spouses spill over to affect each other’s health (Huijts et al., 2010), we expect schooling-related baseline differences in health to be reinforced in countries with higher levels of educational differentiation.

Hypothesis 2: In sum, all these arguments lead to the expectation that health inequalities between people graduated from vocational tracks and people who have pursued academic qualifications are larger in countries with rigidly tracked educational systems than in countries with more comprehensive ones.

Relevant macro-level control variables

The institutional structure of national tracking systems cannot be understood in isolation from other social arrangements that are likely to impact population health. It is thus necessary to control for these confounding institutional factors. First, the degree of tracking is related to the vocational orientation of the educational sector, in that it would be highly likely for strongly stratified educational systems to contain many vocationally oriented tracks in their secondary schools (van de Werfhorst, 2017). It would be feasible to assume that higher levels of vocational enrolment induce smaller health inequalities between individuals educated in vocational and academic tracks. That is to say, in vocationally oriented educational systems, vocational qualifications are less stigmatized and face good employment opportunities (e.g. higher incomes). Second, the degree of educational tracking is known to be related to certain welfare state characteristics. Comparative-historical sociologists point out that education and social policies are part of a broader government strategy, some of which are more beneficial to the health of individuals at the lower tail of the educational spectrum than others (Lavrijsen and Nicaise, 2016). Third, we control for gross domestic product (GDP) per capita and the Gini coefficient to elicit the role of economic development and income inequality in determining health inequalities.

Data and methods

Data

The current study draws on data from eight rounds (2002–2016) of the European Social Survey (ESS), a biennial survey that is representative of the non-institutionalized population aged 15 years or older living in more than 30 European countries. In each country, the ESS sample is drawn according to a strict randomized probability procedure, and data are collected by means of standardized face-to-face interviews. National response rates differ across countries and across time, ranging from 30.5% for Germany in 2010 to 78.1% for Israel in 2014. We restrict our analytical sample to individuals aged 18–45 years.¹ This limitation is based upon a recent study by van de Werfhorst (2017), which indicates that all individuals within this age range can be regarded as having experienced quite similar levels of tracking within their countries. Likewise, a broader age coverage would correspond to a lesser degree of stability in educational institutional settings. In addition, we include only those countries for which the Tracking Index and important macro-control measures have been defined. Other ESS countries/country-years are removed from the sample because their national educational coding scheme does not allow the separation of vocational and academic tracks. These restrictions, along with the deletion of missing cases on the variables of interest (N = 3729, 3.6%), resulted in a final dataset of 99,771 cases living in 22 different countries.

Measurements

Dependent variables. ‘Self-rated health’ is derived from the survey question, ‘How good is your health in general?’ The five-item response scale has been coded such that a score of 0 denotes very poor health and a score of 4 represents excellent health. As a general assessment of one’s health status, self-rated health has been shown to cover physical, mental and social aspects of health (Idler and Benyamini, 1997). In addition, it has proven a strong predictor of mortality and morbidity (Ferraro and Farmer, 1999).

Independent variables: individual level predictors. ‘Educational type’ is defined with reference to the highest achieved level, as derived from the European Survey version of ISCED 1997 (ES-ISCED). This measure was designed to enhance the harmonization of country-specific educational codes (Schneider, 2010), and it is thus especially suitable for purposes of cross-country comparison. The ES-ISCED originally consists of seven categories with an evident classification of vocational and academic types in upper-secondary education (I = less than lower secondary education, II = lower secondary education, IIIa = upper-secondary academic education, IIIb = upper-secondary vocational education, IV = advanced vocational education below Bachelor’s degree level, VI = lower tertiary education, VII = higher tertiary education). For Sweden, France and Israel, we use the national coding schemes for survey years that do not contain information on the harmonized measure. Ultimately, our analysis compares those who attained full upper-secondary vocational education (ES-ISCED IIIb, IV; reference category) with those who attained at most lower secondary education (ISCED I, II), those who attained full upper-secondary academic education (ES-ISCED IIIa), and those who attained tertiary education (ES-ISCED VI, VII).

Furthermore, we include parental education to account for social-origin effects. It indicates the highest level of education completed by the respondent’s father or mother, and it is operationalized according to six categories: primary education (reference category); lower secondary education; upper secondary education; post-secondary non-tertiary education; tertiary education; and an additional category for missing values.

We also add employment status, marital status, and household income to the models. It is important to test the cross-level interaction term with and without these variables as they may increase the health gap between people educated in vocational tracks and people educated in academic tracks in countries with rigid tracking systems. Employment status is measured according to a seven-group categorization: employed (reference category); unemployed, looking for work; unemployed, not looking for work; permanently sick or disabled; housework; student; and a remaining category for community or military services, retired, and “other.” Household income is used as an indicator of economic hardship, and it is assessed using the modified Organisation for Economic Co-operation and Development (OECD) scale (OECD, 2013). This scale assigns a weight of 1 to the first adult in the household, 0.5 to all other adults older than 14 years, with a weight of 0.3 for children up to 14 years of age. To ensure the cross-national equivalence of this variable, we categorize the scale into five groups: lowest income (<50% of the median income), modest income (>50% and <80% of the median), high income (>80% and <120% of the median), highest income (>120% of the median, which is the reference category), and missing values. Marital status distinguishes among respondents who were married or living in civil partnerships (reference category), divorced or separated, widowed, and single.

At the individual level, we control for gender (men = reference category), respondent’s age (measured in years), and survey year.

Independent variables: country-level predictors. To define the level of educational tracking, we use the Tracking Index as developed by Bol and Van de Werfhorst (2013a, 2013b). This index was derived from a factor analysis on three country-level variables measuring curricular tracking equally well: (a) age of first selection (reverse coded), (b) number of different school tracks available to a typical 15-year-old student, and (c) length of differentiated curriculum, expressed as a proportion of total length of primary and secondary education. All countries are given a relative score on the index, with a mean of 0 (signaling average tracked educational systems) and a SD of 1. The values of the index range from −1.08 (UK, and Norway, i.e. comprehensive systems) to 1.79 (Germany, i.e. strongly stratified system).²

The level of vocational orientation is operationalized by the Vocational Enrollment Index (Bol and van de Werfhorst, 2013a, 2013b), which is the result of factor analysis of data from the OECD (OECD, 2006) and United Nations Educational, Scientific and Cultural Organization. It measures the number of students enrolled in vocational education as a proportion of the total number of students in upper-secondary education. Next, two indicators refer to welfare institutional designs. At the country-level, the Healthcare Access and Quality index (HAQ)—an indicator developed by the research group involved in the Global Burden of Disease Study 2015 (Barber et al., 2017)—takes values between 0 and 100, with higher values indicating better personal health-care access and quality. At the country-year level, gross public social expenditure (as percentage of GDP) is included, which is obtained from the SOCX database (Adema et al., 2001). GDP per capita at current prices (in US dollars) and the Gini coefficient of post-tax post-transfer household income are defined for each country and each year separately, using data from the World Bank and the SWIID database (Solt, 2009), respectively.

Statistical analyses

We use three-level hierarchical models to estimate the following equation

\begin{matrix} Y_{i j k} = β_{0 i j k} + Educational {type}_{i j k} β_{1} + X_{1 i j k} β_{2} + X_{2 i j k} β_{3} + β_{4} {Tracking}_{κ} + Z_{1 k} β_{5} \\ + Z_{2 j k} β_{6} + Educational type * {Tracking}_{ijk} β_{7} + v_{0 k} + μ_{0 j k} + e_{0 i j k} \end{matrix}

In this equation, Y_ijk stands for the self-rated health score for individual (i) in country-year (j) in country (k). β _1j is a vector that represents the effects of educational types and allows for random slope parameters; the term β ₂ is a vector of individual-level control variables (age, gender, and survey year); β ₃ refers to other individual-level covariates that are important to include (parental education, employment status, marital status, and household income); β ₄ estimates the country-level effect of tracking; β ₅ represents country-level control variables (vocational orientation, and HAQ index); β ₆ is a vector of country-year level control variables (social expenditure, GDP, and Gini coefficient); β ₇ refers to the cross-level interaction between educational type and a country’s degree of curricular tracking (this term is our prime focus); and v _0k, μ _0jk, and e _0ijk are the error terms.

The analysis is carried out in MLwiN (version 2.35) with the Markov Chain Monte Carlo estimation procedure. The Markov Chain Monte Carlo approach is applied as it has been shown to increase the reliability of cross-level coefficients in nesting structures containing a quite low number of country units (Stegmueller, 2013). Self-rated health is treated as a metric variable following a normal distribution, even though additional analyses indicate that this is not entirely accurate. More specifically, Kolmogorov–Smirnov tests confirm the deviation from normality. Nevertheless, sensitivity checks (not shown, but available upon request) using ordered logit models lead to similar conclusions. We present the original models, as they are more straightforward to interpret. With the exception of the index measures, all continuous variables are centered on the grand mean, in order to facilitate the interpretation of the unstandardized coefficients. We use design weights to account for differences in sampling probability between respondents, but we do not include population weights.

The models are built stepwise and all control for gender (with men reporting better self-rated overall health), age (with younger people reporting better self-rated overall health), and period differences. The baseline model (Model 1) assesses the most important micro and macro effects: the effect of educational type (micro) and the effect of educational tracking (macro). It also includes a random slope for the education dummies at the country-level. Furthermore, it contains the effect of parental education (being positively associated with reporting good health). This can be included from the start as the in- or exclusion of this variable does not alter the results, suggesting that social background is not an important confounding factor when measuring the individual relationship between educational type and health. In Model 2, the cross-level interaction term is added. In Model 3, we include the other individual-level covariates. Model 4 includes all macro-level control variables. It also tests for the interactions between the latter and educational type. However, for reasons of parsimony, non-significant cross-level interactions are excluded. Finally, the Deviance Information Criterion provides an indicator of the overall goodness of the model fit, with lower scores indicating a better model (Spiegelhalter et al., 2002).

Results

Descriptive statistics

Before turning to the actual testing of our hypotheses, some descriptive results are discussed. Appendix A reports the mean health values for each educational type and each country separately. In addition, Figure 1 depicts, for each country in our sample, the difference in self-rated health between people graduated from vocational tracks on the one hand and people graduated from other types of education on the other. As can be seen, in almost all countries, individuals who have pursued vocational qualifications generally report worse health than individuals who have obtained academic degrees. In addition, this figure suggests that health inequalities between the latter two educational types may be more pronounced in countries with higher levels of curricular differentiation. Finally, descriptive statistics for all individual and macro-level variables are given in Table 1 and Appendix B, respectively.

Figure 1.

Observed differences in self-rated health between respondents with vocational qualifications and the other educational types (for each country separately).

Table 1.

Individual-level descriptive statistics (N = 99,771).

	Range	Mean	SD
Health	0–4	3.13	0.8
Age (in years)	25–40	33	7.9
		%	N
Gender (men)		48.7%	48,593
Educational type
Up to lower secondary education		17.3%	17,296
Full upper-secondary education vocational		32.4%	32,362
Full upper-secondary education academic		24.4%	24,322
Tertiary education		25.9%	25,791
Parental education
Primary education		13.8%	13,794
Lower secondary education		15.1%	15,028
Upper secondary education		35.7%	35,598
Post-secondary non-tertiary education		8.1%	8113
Tertiary education		22.1%	22,025
Missing values		5.2%	5213
Employment
Employed		67.2%	67,011
Unemployed, looking for work		6.0%	6012
Unemployed, not looking for work		1.9%	1940
Permanently sick or disabled		1.4%	1432
Housework		8.9%	8857
Student		12.5%	12,513
Others		2.0%	2006
Household income
<50% of median income		10.9%	10,861
50–80% of median income		16.1%	16,092
80–120% of median income		24.3%	24,211
>120% of median income		29.4%	29,324
Missing values household income		19.3%	19283
Marital status
Married or civil partnership		44.4%	44,270
Divorced or separated		6.7%	6658
Widowed		0.5%	468
Single		48.5%	48,375

SD: standard deviation.

Note that almost one-third of the respondents had attained vocational upper-secondary education (32.4%), whereas another fourth had completed tertiary education (25.9%). Fewer respondents had attained an academic upper-secondary degree (24.4%), with those having no more than lower secondary education constituting the smallest group (17.3%). The macro-level descriptive statistics reveal considerable variability between countries with regard to tracking, with Germany and Czech Republic reporting the highest levels of tracking and Sweden, Norway and the UK reporting the lowest levels.

Multilevel results

The results of the multilevel analyses are presented in Table 2. First, Model 1 indicates that people educated in vocational tracks report significantly worse health than do those who have pursued academic qualifications (b = 0.039, p < 0.05). The estimated signs for the other education dummies are also in line with what can be expected. On average, we see that those with vocational qualifications have better health than those with the lowest education reported (b = −0.155, p < 0.001), but worse health than those who have completed tertiary education (b = 0.181, p < 0.001). The model further reveals that curricular tracking is overall not statistically associated with people’s health.

Table 2.

Self-rated health regressed on educational type, and tracking index.

	Self-rated health
	Model 1	Model 2	Model 3	Model 4
Intercept	3.038***	3.041***	3.175***	3.107***
Intercept	(0.045)	(0.044)	(0.042)	(0.067)
Gender	−0.085***	−0.085***	−0.073***	−0.073***
Gender	(0.005)	(0.005)	(0.005)	(0.005)
Age	−0.015***	−0.015***	−0.014***	−0.014***
Age	(0.001)	(0.001)	(0.001)	(0.001)
Parental education (primary education = reference)
Lower secondary education	0.046***	0.046***	0.028**	0.028**
Lower secondary education	(0.010)	(0.010)	(0.009)	(0.009)
Upper secondary education	0.065***	0.065***	0.036***	0.036***
Upper secondary education	(0.009)	(0.009)	(0.009)	(0.009)
Post-secondary non-tertiary education	0.055***	0.055***	0.023*	0.023*
Post-secondary non-tertiary education	(0.012)	(0.012)	(0.011)	(0.011)
Tertiary education	0.080***	0.080***	0.048***	0.049***
Tertiary education	(0.010)	(0.010)	(0.009)	(0.010)
Educational level unknown	−0.060***	−0.059***	−0.052***	−0.051***
Educational level unknown	(0.015)	(0.015)	(0.014)	(0.014)
Educational type (full upper-secondary vocational = reference)
Up to lower secondary	−0.155***	−0.160***	−0.103***	−0.140***
Up to lower secondary	(0.024)	(0.023)	(0.020)	(0.025)
Full upper-secondary academic	0.039*	0.033*	0.029*	0.018
Full upper-secondary academic	(0.019)	(0.015)	(0.012)	(0.018)
Tertiary	0.181***	0.178***	0.129***	0.120***
Tertiary	(0.019)	(0.018)	(0.016)	(0.021)
Tracking	−0.029	−0.082*	−0.074	−0.085
Tracking	(0.039)	(0.040)	(0.044)	(0.053)
Education attainment (full upper-secondary vocational = reference)*tracking
Up to lower secondary		0.035	0.034	0.021
Up to lower secondary		(0.024)	(0.022)	(0.020)
Full upper-secondary academic		0.053**	0.042***	0.040**
Full upper-secondary academic		(0.016)	(0.012)	(0.013)
Tertiary		0.026	0.022	0.015
Tertiary		(0.019)	(0.017)	(0.016)
INDIVIDUAL-LEVEL COVARIATES
Employment status (employed = reference)
Unemployed, looking for work			−0.113***	−0.113***
Unemployed, looking for work			(0.010)	(0.021)
Unemployed, not looking for work			−0.202***	−0.202***
Unemployed, not looking for work			(0.017)	(0.017)
Permanently sick or disabled			−1.349***	−1.349***
Permanently sick or disabled			(0.020)	(0.020)
Housework			−0.031***	−0.032***
Housework			(0.009)	(0.019)
Student			0.031***	0.031***
Student			(0.008)	(0.009)
Others			−0.309***	−0.310***
Others			(0.017)	(0.017)
Household income (>120% of median income = reference)
<50% of median income			−0.189***	−0.189***
<50% of median income			(0.009)	(0.009)
50%–80% of median income			−0.120***	−0.119***
50%–80% of median income			(0.007)	(0.007)
80%–120% of median income			−0.059***	−0.059***
80%–120% of median income			(0.006)	(0.006)
Missing values			−0.030***	0.031***
Missing values			(0.007)	(0.007)
Marital status (married = reference)
Divorced or separated			−0.080***	−0.080***
Divorced or separated			(0.010)	(0.010)
Widowed			−0.170***	−0.169***
Widowed			(0.033)	(0.034)
Single – never married/civil partnership			−0.048***	−0.048***
Single – never married/civil partnership			(0.006)	(0.006)
MACRO-LEVEL COVARIATES
Vocational enrollment index				−0.004
Vocational enrollment index				(0.024)
HAQ index				0.016
HAQ index				(0.010)
Gross public social expenditure (% of GDP)				−0.015***
Gross public social expenditure (% of GDP)				(0.004)
GDP (/1000)				−0.004***
GDP (/1000)				(0.001)
Gini coefficient				−0.003
Gini coefficient				(0.009)
Educational attainment (full upper-secondary vocational = reference)*vocational enrolment
Up to lower secondary				0.071*
Up to lower secondary				(0.033)
Full upper-secondary academic				0.022
Full upper-secondary academic				(0.023)
Tertiary				0.016
Tertiary				(0.027)
Educational attainment (full upper-secondary vocational = reference)*Gini
Up to lower secondary				0.009*
Up to lower secondary				(0.004)
Full upper-secondary academic				0.004
Full upper-secondary academic				(0.003)
Tertiary				−0.006
Tertiary				(0.004)
VARIANCE
Random intercept
Country	0.036	0.033	0.032	0.040
Country	(0.013)	(0.012)	(0.012)	(0.016)
Country-year	0.005	0.005	0.004	0.003
Country-year	(0.001)	(0.001)	(0.001)	(0.001)
Individual	0.554	0.554	0.520	0.520
Individual	(0.002)	(0.002)	(0.002)	(0.002)
Random slope educational type
Up to lower secondary	0.011	0.009	0.007	0.005
Up to lower secondary	(0.004)	(0.004)	(0.003)	(0.002)
Full upper-secondary academic	0.007	0.004	0.002	0.002
Full upper-secondary academic	(0.003)	(0.002)	(0.001)	(0.001)
Tertiary	0.006	0.006	0.004	0.003
Tertiary	(0.002)	(0.002)	(0.002)	(0.001)
DIC	224,219.87	224,219.96	217,876.26	217,874.71

All continuous variables are centered on their grand mean. Total population is aged 18–45 years (N _countries = 22; N _{country-years} = 129; N _individuals = 99,771, weighted sample). All models are additionally controlled for period differences.

*p < 0.05; **p < 0.01; ***p < 0.001 (two-tailed test).

HAQ: Healthcare Access and Quality; GDP: gross domestic product; DIC: Deviance Information Criterion.

More important to our hypothesis is Model 2, which adds the interaction term between educational type and the tracking index. This model provides evidence to support our expectation that educational differentiation is related to greater inequalities in self-rated health between vocational and academic forms of upper-secondary education. For each SD increase on the scale of tracking, the negative association between having pursued vocational qualifications relative to having completed full upper-secondary academic education and health increases (b = 0.053, p < 0.01). Also note that part of the variance in the slope of vocational education is explained by adding this cross-level interaction term.

Model 3 introduces other relevant individual-level covariates into the equation. All of the latter predictors are in accordance with common findings of the social-epidemiological literature. Higher self-rated overall scores are observed among employed people (except when compared to the group of students), those with higher household incomes, and married people. The addition of these variables does not affect our main finding that health inequalities between people educated in vocational tracks and people graduated from academic tracks increase with higher levels of educational stratification. The only noticeable change is that the coefficient is reduced from 0.053 to 0.042, although it remains clearly significant. This reduction suggests that one or more of the added variables partly accounts for the stronger association between type of secondary education and health in countries with rigid tracking systems. Additional analyses of the data at hand may provide deeper understanding. Using multinomial multilevel modelling and adjusting for relevant individual-level and macro-level control variables, we regress, respectively, employment status, marital status, and household income on the main effects and the interaction effects of educational type and tracking. As can be seen in Appendix D, these analyses reveal that (a) vocational education is associated with higher levels of unemployment and lower levels of household income than is academic education, and that (b) this pattern seems to be more pronounced in countries with higher degrees of educational tracking.

The final model serves as a robustness check to determine whether the differential effect of institutional tracking persists over and above the impact of other macro-level variables. Although they are not of primary interest to this study, we provide a brief discussion of the latter. Counterintuitive as it may be, we see that population health is negatively related to social spending and GDP.³ Furthermore, it appears that countries with higher levels of vocational orientation on the one hand and country-years with higher degrees of income inequality on the other are associated with a stronger gradient in self-rated health between people who attained full upper-secondary vocational education and the lowest educated. However, more importantly, this model shows that the heterogeneous relationship between education track and health according to the degree of school differentiation remains unchanged, and it can thus be regarded as quite robust.

To better understand the impact of the cross-level interaction term, we turn to Figure 2. Based on Model 4,⁴ this figure demonstrates the estimated gap (cf. Figure 1 depicts the observed gap) between the self-rated health of respondents with vocational qualifications and respondents with academic qualifications (estimated health differences to the other educational groups are presented for ease of comparison). It can immediately be seen that health inequalities between vocational and academic education are substantially larger in countries with strongly differentiated educational systems, as compared to countries with more comprehensive systems. Figure 2 also shows that the gap in self-rated health between people with upper-secondary vocational education and people with upper-secondary academic education increases with 0.11 point scores when moving from a Tracking Index of −1.08 to 1.79, that is when moving from the UK to Germany. This increase might seem to be small, but as a comparison, the health gap between graduates from vocational and academic forms of secondary education is almost as large as the health gap between people educated in vocational tracks and people who have pursued tertiary education in countries with intensive tracking systems. This implies that in heavily tracked contexts, inequalities emerge even within groups with the same number of years of schooling and that these inequalities are almost as substantial as inequalities between groups with different years of schooling.

Figure 2.

Estimated differences in self-rated health between respondents with vocational qualifications and respondents with other educational qualifications by level of tracking.

Sensitivity check

As mentioned in the background section, curricular tracking does not arise in a vacuum, but in a social and cultural context. In the analysis, we have controlled for some confounding institutional factors, but not for all of them. For instance, five of the six countries with the lowest scores on the tracking measure are Nordic countries, which are highly likely characterized by relatively egalitarian values. Unfortunately, we were not able to control for this potential macro-level confounder as well as for others given that there are not enough degrees of freedom available. To take account of this problem, we conducted a sensitivity analysis that uses country fixed effects. In this way, we avoid the omitted variable bias through controlling for country heterogeneity. A drawback of this fixed-effect approach is that no main effects of country variables can be included since all the variance at this level is already explained (i.e. this approach does not allow us to assess the overall effect of curricular tracking). Nonetheless, it is possible to include a cross-level interaction term (the main focus of our paper), since this coefficient focuses only on potential non-linearities in the impact of individual-level variables (Bennett and Moehring, 2015; Bol et al., 2014). As can be seen in Appendix E, the coefficients of the cross-level interaction terms are very similar to those presented in the main analysis, suggesting that unobserved factors are not confounding the impact of tracking systems.

Discussion

Most European secondary school systems apply some form of educational tracking, notwithstanding growing evidence for its harmful effect on social disparities in a wide range of outcomes (Brunello and Checchi, 2007; Janmaat et al., 2014; van de Werfhorst, 2017; van der Velden and Wolbers, 2003). In this article, we consider the role of tracked education in shaping health inequalities. Drawing on three-level hierarchical analysis and using information from the ESS (2002–2016) on 99,771 individuals in 22 countries, the results indicate two interesting findings.

The first finding is that we see that individuals who have completed upper-secondary vocational education are generally more vulnerable to poor health than are individuals who have pursued academic qualification. Second, and more importantly, we find that this gap in overall health is strengthened in societies in which students are more rigidly tracked (e.g. Germany or Czech Republic), as compared to countries with more comprehensive education systems (e.g. the UK and Scandinavian countries). Although it is not possible to determine from our data the reason for the first finding, several explanations are possible. Some of them assume a causal sequence in which education track affects health directly (i.e. differences in health socialization) (de Looze et al., 2013; Rathmann et al., 2016; van Rossem et al., 2010) or indirectly through multiple pathways such as differences in curriculum covering (ten Dam and Volman, 2003; van de Werfhorst, 2017), and the role of relative deprivation (van Houtte, 2005, 2016). Another explanation postulates the importance of social selection: Respondents with highly educated parents are more likely to graduate from higher tracks than are respondents from parents with lower degrees (Boone and van Houtte, 2013), as well as that they tend to obtain better health (Kestilä et al., 2006). We believe, however, that the latter explanation is less likely since the coefficients of the education dummies barely change after controlling for parental education.

The second finding indicates that health inequalities between educational types is context-specific. Indeed, the association between educational type and health becomes stronger with increasing levels of institutional differentiation. Ancillary analyses demonstrate that at least part of this cross-level interaction can be attributed to more pronounced differences in economic returns in countries in which students are more rigidly stratified. Other arguments—which cannot be empirically tested with the data at hand—can also account for the observed finding, such as the stronger degrees of social segregation (van de Werfhorst, 2017), the higher levels of educational heterogamy that relate to intensive tracking (Blossfeld and Timm, 2003), and the intensified importance of social selection processes (Brunello and Checchi, 2007; Montt, 2011).⁵ However, regardless of which explanation prevails eventually, the cross-level interaction term can be regarded as quite robust, given its persistence when controlling for relevant macro-level variables.

Two limitations of our study should be mentioned. First, as already briefly alluded to above, most explanations cannot be evaluated on the basis of the present data. Unfortunately, the ESS does not provide enough detailed information to model mediating pathways. We are also unaware of any other available dataset that contains information so finely grained and that covers so many countries. Future studies should therefore focus on one single country (or a few countries) to identify the mechanisms that drive our results. However, it must be noted that it was not our main intention to detect the underlying processes. It was our aim to introduce an institutional thinking about educational systems and health inequalities. Second, the cross-sectional nature of the ESS restricts our capacity to rule out selection effects. At the individual level, controlling for parental education only partially resolves this problem. Other unobserved factors remain a concern. If health and education track are both related to additional confounders that cannot be observed (e.g. future orientation, childhood health, and genetic endowments), social selection may still be an important force. More complex experimental designs are needed to address this issue in detail. At the macro level, the cross-level interaction term may reflect differences in the strength of selection. That is, in countries with rigid tracking systems, people graduated from lower tracks may constitute an even more negatively selected group on covariates relevant for health than do their peers in countries with comprehensive systems. However, whether through intensified processes of causation or selection, tracking as a feature of educational systems renders an institutional context that strengthens the association between education track and health. The sensitivity analysis that we performed further supports this claim. That is, models including country dummies to account for all unobserved heterogeneity at the country level, obtain similar results to the results presented in the main models of the paper. This renders further evidence that it is indeed tracking as an institutional feature that strengthens the relationship between educational type and health.

Following from this observation, it should be clear that health inequalities can also be targeted beyond the traditional realm of healthcare systems. Indeed, it is shown that characteristics of educational institutions also have an impact on the (re)production of health inequalities. Health policies should therefore be coordinated with educational policies. However, this is not a new insight. In 2005, Low and colleagues came to a similar conclusion when theoretically discussing clear pathways for health intervention (Low et al., 2005).

In sum, our study offers a complementary perspective to recent studies that have revealed the detrimental effects of curricular tracking on a wide range of outcomes. Perhaps even more importantly, however, our results open a new direction for social–epidemiological research on the relationship between education and health. To the best of our knowledge, this study is the first to establish a structural-institutional perspective on the role of schooling systems in reproducing educational inequalities in health within various national contexts. Given the importance and relevance of the findings, we can only welcome future studies exploring the impact of other educational institutional features (e.g. pursuit of quality, level of standardization, and compulsory attendance laws). Understanding the root causes of social reproduction ultimately constitutes an important step towards improving population health.

Supplemental material

Supplemental Material, Appendix_supplemental_material - Educational inequalities in general health: Does the curricular tracking system matter?

Supplemental Material, Appendix_supplemental_material for Educational inequalities in general health: Does the curricular tracking system matter? by Katrijn Delaruelle, Mieke Van Houtte and Piet Bracke in Acta Sociologica

Footnotes

Acknowledgements

We would like to thank Prof. Dr. Tim Huijts and the reviewers for their helpful suggestions on earlier versions of this article.

Funding

The author disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Special Research Fund (BOF) of Ghent University under grant no. 01D20316.

Supplemental material

Supplemental material for this article is available online.

Notes

References

Adema

Fron

Ladaique

(2011) Is the European Welfare State Really More Expensive? Indicators on Social Spending, 1980-2012; and a Manual to the OECD Social Expenditure Database (SOCX). OECD Social, Employment and Migration Working Papers. Paris: OECD Publishing. DOI: https://doi.org/10.1787/1815199X.

Allmendinger

(1989) Educational systems and labor market outcomes. European Sociological Review 5(3): 231–250.

Bandura

(1977) Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review 84(2): 191–215.

Barber

Fullman

Sorensen

, et al. (2017) Healthcare Access and Quality Index based on mortality from causes amenable to personal health care in 195 countries and territories, 1990–2015: A novel analysis from the Global Burden of Disease Study 2015. Lancet 390(10091): 231–266.

Beckfield

Bambra

Eikemo

, et al. (2015) An institutional theory of welfare state effects on the distribution of population health. Social Theory & Health 13(3–4): 227–244.

Bennett

Moehring

(2015) Cumulative (dis)advantage? The impact of labour market policies on late career employment from a life course perspective. Journal of Social Policy 44(2): 213–233.

Berten

Cardoen

Brondeel

, et al. (2012) Alcohol and cannabis use among adolescents in Flemish secondary school in Brussels: Effects of type of education. BMC Public Health 12(1): 215–223.

Blossfeld

H-P

Timm

(2003) Educational systems as marriage markets in modern societies: A conceptual framework. In: Blossfeld

H-P

Timm

(eds) Who Marries Whom? Educational Systems as Marriage Markets in Modern Societies. Dordrecht: Kluwer Academic, pp.1–18.

Bol

van de Werfhorst

(2013a) Educational systems and the trade-off between labor market allocation and equality of educational opportunity. Comparative Education Review 57(2): 285–308.

10.

Bol

van de Werfhorst

(2013b) The measurement of tracking, vocational orientation, and standardization of educational systems: A comparative approach. Gini Discussion Paper 81. Available at: https://ideas.repec.org/p/aia/ginidp/81.html (accessed 1 June 2017).

11.

Bol

Witschge

van de Werfhorst

, et al. (2014) Curricular tracking and central examinations: Counterbalancing the impact of social background on student achievement in 36 countries. Social Forces 92(4): 1545–1572.

12.

Boone

Van Houtte

(2013) In search of the mechanisms conducive to class differentials in educational choice: A mixed method research. The Sociological Review 61(3): 549–572.

13.

Bracke

Van de Straat

Missine

(2014) Education, mental health, and education-labor market misfit. Journal of Health and Social Behavior 55(4): 442–459.

14.

Brunello

Checchi

(2007) Does school tracking affect equality of opportunity? New international evidence. Economic Policy 22(52): 782–861.

15.

de Looze

ter Bogt

Hublet

, et al. (2013) Trends in educational differences in adolescent daily smoking across Europe, 2002–2010. The European Journal of Public Health 23(5): 846–852.

16.

Dudal

Verhaest

Bracke

(2017) Educational inequalities in depression: Do labor markets matter? Society and Mental Health 8(2): 93–107.

17.

Eikemo

Bambra

Judge

, et al. (2008) Welfare state regimes and differences in self-perceived health in Europe: A multilevel analysis. Social Science & Medicine 66(11): 2281–2295.

18.

Ferraro

Farmer

(1999) Utility of health data from social surveys: Is there a gold standard for measuring morbidity? American Sociological Review 64(2): 303–315.

19.

Gamoran

Mare

(1989) Secondary school tracking and educational inequality: Compensation, reinforcement, or neutrality? American Journal of Sociology 94(5): 1146–1183.

20.

Gesthuizen

Huijts

Kraaykamp

(2012) Explaining health marginalisation of the lower educated: The role of cross-national variations in health expenditure and labour market conditions. Sociology of Health & Illness 34(4): 591–607.

21.

Groot

van den Brink

(2006) What does education do to our health? In: Measuring the effects of education on health and civic engagement: Proceedings of the Copenhagen Symposium. Amsterdam: OECD. Available at: http://www.oecd.org/education/innovation-education/37437718.pdf (accessed 30 November 2018).

22.

Holm

Jæger

(2008) Does relative risk aversion explain educational inequality? A dynamic choice approach. Research in Social Stratification and Mobility 26(3): 199–219.

23.

Huijts

Monden

CWS

Kraaykamp

(2010) Education, educational heterogamy, and self-assessed health in Europe: A multilevel study of spousal effects in 29 European countries. European Sociological Review 26(3): 261–276.

24.

Idler

Benyamini

(1997) Self-rated health and mortality: A review of twenty-seven community studies. Journal of Health and Social Behavior 38(1): 21–37.

25.

Janmaat

Mostafa

Hoskins

(2014) Widening the participation gap: The effect of educational track on reported voting in England. Journal of Adolescence 37(4): 473–482.

26.

Kalmijn

(1998) Intermarriage and homogamy: Causes, patterns, trends. Annual Review of Sociology 24: 395–421.

27.

Kestilä

Koskinen

Martelin

, et al. (2006) Determinants of health in early adulthood: What is the role of parental education, childhood adversities and own education? The European Journal of Public Health 16(3): 305–314.

28.

Knesebeck

Ovd

Lüschen

Cockerham

, et al. (2003) Socioeconomic status and health among the aged in the United States and Germany: A comparative cross-sectional study. Social Science & Medicine 57(9): 1643–1652.

29.

Lareau

(2000) Home Advantage: Social Class and Parental Intervention in Elementary Education. Lanham: Rowman & Littlefield.

30.

Lavrijsen

Nicaise

(2016) Ascription, achievement, and perceived equity of educational regimes: An empirical investigation. Social Sciences 5(4): 64.

31.

Levels

Van der Velden

Di Stasio

(2014) From school to fitting work: How education-to-job matching of European school leavers is related to educational system characteristics. Acta Sociologica 57(4): 341–361.

32.

Low

Baumler

, et al. (2005) Can education policy be health policy? Implications of research on the social determinants of health. Journal of Health Politics, Policy and Law 30(6): 1131–1162.

33.

Mackenbach

Stirbu

Roskam

A-JR

, et al. (2008) Socioeconomic inequalities in health in 22 European countries. New England Journal of Medicine 358(23): 2468–2481.

34.

Marmot

(2005) Social determinants of health inequalities. The Lancet 365(9464): 1099–1104.

35.

Mirowsky

Ross

(2005) Education, learned effectiveness and health. London Review of Education 3(3): 205–220.

36.

Montt

(2011) Cross-national differences in educational achievement inequality. Sociology of Education 84(1): 49–68.

37.

Nutbeam

Smith

Moore

, et al. (1993) Warning! Schools can damage your health: Alienation from school and its impact on health behaviour. Journal of Paediatrics and Child Health 29(s1): 25–30.

38.

Organisation for Economic Co-operation and Development (OECD) (2006) Education at a Glance OECD Indicators 2006. Paris: OECD Publishing.

39.

Organisation for Economic Co-operation and Development (OECD) (2013) Guidelines for Micro Statistics on Household Wealth. Paris: OECD Publishing.

40.

Pampel

(2002) Inequality, diffusion, and the status gradient in smoking. Social Problems 49(1): 35–57.

41.

Präg

Wittek

Mills

(2017) The educational gradient in self-rated health in Europe: Does the doctor–patient relationship make a difference? Acta Sociologica 60(14): 325–341.

42.

Rathmann

Moor

Kunst

, et al. (2016) Is educational differentiation associated with smoking and smoking inequalities in adolescence? A multilevel analysis across 27 European and North American countries. Sociology of Health & Illness 38(7): 1005–1025.

43.

Ross

Mirowsky

(1995) Does employment affect health? Journal of Health and Social Behavior 36(3): 230–243.

44.

Schafer

Olexa

(1971) Tracking and Opportunity: The Locking-Out Process and Beyond. Scranton: Chandler.

45.

Schneider

(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(3): 343–357.

46.

Solt

(2009) Standardizing the world income inequality database. Social Science Quarterly 90(2): 231–242.

47.

Spiegelhalter

Best

Carlin

, et al. (2002) Bayesian measures of model complexity and fit. Journal of the Royal Statistical Society: Series B (Statistical Methodology) 64(4): 583–639.

48.

Stegmueller

(2013) How many countries for multilevel modeling? A comparison of frequentist and Bayesian approaches. American Journal of Political Science 57(3): 748–761.

49.

ten Dam

Volman

(2003) A life jacket or an art of living: Inequality in social competence education. Curriculum Inquiry 33(2): 117–137.

50.

van de Werfhorst

(2011) Skill and education effects on earnings in 18 countries: The role of national educational institutions. Social Science Research 40(4): 1078–1090.

51.

van de Werfhorst

(2017) Vocational and academic education and political engagement: The importance of the educational institutional structure. Comparative Education Review 61(1): 111–140.

52.

Van der Velden

Wolbers

(2003) The integration of young people into the labour market: the role of training systems and labour market regulation. In: Müller

Gangl

(eds) Transitions from Education to Work in Europe. The Integration of Youth into EU Labour Markets. Oxford: Oxford University Press, pp.186–211.

53.

Van Houtte

(2005) Global self-esteem in technical/vocational versus general secondary school tracks: A matter of gender? Sex Roles 53(9–10): 753–761.

54.

Van Houtte

(2016) Lower-track students’ sense of academic futility: Selection or effect? Journal of Sociology 52(4): 874–889.

55.

Van Houtte

Stevens

(2009) Study involvement of academic and vocational students: Does between-school tracking sharpen the difference? American Educational Research Journal 46(4): 943–973.

56.

Van Praag

Demanet

Stevens

, et al. (2017) Everyone has their own qualities: Tracking and academic self-appraisal in Flemish secondary education. Social Psychology of Education 20(3): 601–618.

57.

Van Rossem

Berten

Van Tuyckom

(2010) AIDS knowledge and sexual activity among Flemish secondary school students: A multilevel analysis of the effects of type of education. BMC Public Health 10(1): 30.

58.

von dem Knesebeck

Pattyn

Bracke

(2011) Education and depressive symptoms in 22 European countries. International Journal of Public Health 56(1): 107–110.

59.

von dem Knesebeck

Verde

Dragano

(2006) Education and health in 22 European countries. Social Science & Medicine 63(5): 1344–1351.

60.

Wilkinson

(2002) Unhealthy Societies: The Afflictions of Inequality. London: Routledge.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.44 MB