Background Recent research has emphasised that the challenge in researching socioeconomic differences in adolescent health cross-nationally lies in providing valid and comparable measures of socioeconomic position (SEP) across regions. This study aims to examine measures of occupational status derived from the International Standard Classiﬁcation of Occupations (ISCO), alongside commonly used affluence measures in association with adolescent self-rated health (SRH).
Methods Data were from the 2005/2006 ‘Health Behaviour in School-aged Children study’ (HBSC); 27 649 individuals aged 11, 13 and 15 years from Germany, Macedonia, Norway, Turkey, Wales and Scotland. Three occupational scales were compared: the International Socioeconomic Index of Occupational Status (ISEI), the Standard International Occupational Prestige Scale (SIOPS) and the Erikson–Goldthorpe–Portocarero class categories (EGP). Correlation analyses compared these occupational scales with the family affluence scale (FAS) and a family well-off measure, while logistic regression assessed the association between occupational scales and poor SRH. Multiple imputation techniques investigated possible bias arising from parental occupation missingness.
Results Moderate correlations existed between occupational scales and FAS and family well-off. Socioeconomic inequalities in poor SRH were found for ISEI, SIOPS and EGP in all regions, independent of FAS and family well-off. Models of imputed data sets did not alter the results. The relationship between SEP and SRH was therefore not biased by high levels of missing values for ISCO.
Conclusions ISCO-based indicators of occupational status in cross-national self-administered adolescent health surveys were found to be robust measures of SEP in adolescence. These measure different aspects of SEP independent of FAS and family well-off.
- ADOLESCENTS CG
- Health inequalities
Statistics from Altmetric.com
The socioeconomic position (SEP) refers to the relative position of an individual or family within a stratified social structure, according to their access to, or control over, wealth, prestige and power.1 ,2 Many studies, particularly from high-income countries, have documented that as with other stages in the life course, adolescents with a higher SEP have better health.3 The impact of SEP in adolescence on health during adolescence and beyond is an important topic for public health research.4 However, the measurement of SEP in self-administered adolescent questionnaires is challenging as adolescents themselves have little economic power; are still at school and therefore do not have a SEP of their own.5 ,6 More recently, underlying mechanisms such as material factors, psychosocial determinants and health-related behaviours have been proposed linking adolescents’ SEP and health.7 ,8 Although these are often found to work simultaneously, they are dependent on the measures of SEP understudy.9
The association between family assets, often measured by indicators of family affluence, and adolescent health is predominantly explained by material mechanisms,6 ,10 while perceived SEP affects health through psychological pathways.11 Parental occupation is believed to influence adolescent health through material conditions and psychosocial resources that directly stem from occupational prestige and job conditions, as well as indirectly from parental earnings.12 ,13 Koivusilta et al14 found that indicators of SEP (eg, parental occupational and educational status, family material wealth, adolescents’ school achievement) were differently associated with health, and could therefore not be used interchangeably as they measured independent constructs of SEP. These findings emphasise the need to incorporate multiple measures of SEP in order to monitor and understand the mechanisms underlying inequalities in adolescent health.9
Furthermore, recent research has emphasised that the challenge in researching socioeconomic differences using adolescent surveys lies in providing valid and comparable measures of SEP across countries.15 Among other issues, respondents in cross-national adolescent surveys live in diverse cultural and economic settings, speak different languages and might understand concepts and ideas of SEP in varying ways.16 For instance, it has been shown that the ‘International Standard Classification of Education (ISCED 97)’ and years of education inadequately measure the educational attainment in cross-national surveys as they do not capture the distinct value of vocational training in some countries.17 In addition, Schnohr et al18 found country variation in the item functioning of the family affluence scale (FAS), advising caution in its use for international comparisons of inequalities. However, the cross-country comparability of measures of SEP in adolescent health surveys is a neglected area and further indicators are needed.
Several indicators of SEP have been developed for adult surveys with cross-country comparability, such as the International Standard Classiﬁcation of Occupations (ISCO).19 ISCO was developed by the International Labour Office (ILO) for national agencies to generate classifications that allowed for comparison of similar jobs across nations with different vocational training and labour market systems.19 ,20 Based on a set of work tasks and duties, and the skill level in the job, the ILO organises jobs in a hierarchical four digit-system of occupational groupings. The main types of traditional occupational scales are21: socioeconomic scales, prestige scales and nominal class categories. Socioeconomic scales, for example, the International Socioeconomic Index of occupational status (ISEI), arrange occupations according to the average level of education and average earnings of job holders.22 Prestige scales are based on a popular evaluation of occupational prestige (ie, the social standing, respect, attractiveness and prestige of occupations), known as Standard International Occupational Prestige Scale (SIOPS).23 Nominal class categories, such as the Erikson–Goldthorpe–Portocarero (EGP), are categorical class schemes, founded on the contractual relationship between employer and employee, skill levels and sectorial differences of occupations.24
ISCO-based measures have been used in international research on inequalities in adult health.13 ,25 However, they rarely have been used in adolescent health surveys. Research found high prevalence of missing data on parental occupation in adolescent health surveys.26 However, as well as the issue of cross-national comparability, commonly used measures of material affluence in adolescent samples may not be capturing an important construct of SEP; one which could be measured using ISCO. The main objective of this study is to examine the cross-national applicability of ISCO-based measures in adolescence. Specifically, the aims of the present study are (1) to assess the relative importance of three cross-nationally standardised measures of occupational status (ISEI, SIOPS, EGP) in measuring socioeconomic inequalities in adolescent health, (2) to analyse their association with other commonly used SEP indicators (FAS, family well-off) and (3) to compare the results of analyses based on listwise-deleted and multiply imputed data considering the health differences of adolescents with missing occupational data.
The data analysed were collected by the ‘Health Behaviour in School-aged Children (HBSC) Study’, a WHO collaborative cross-national study collecting nationally representative data on adolescents every 4 years, since 1982, in a growing number of countries in Europe and North America and in adherence to a protocol updated at every round.27 The survey is a self-complete questionnaire administered in schools by teachers. Participation is voluntary and consent is obtained from school administrators, parents and children. Participating countries employ a multistage sample procedure with the school or class as the sampling unit. In 2005/2006 the response rate for schools and pupils exceeded 70% in the majority of the 38 participating countries.28 Ethical approval is obtained for each national survey according to national guidance and regulations at the time of data collection.
The present analysis is based on data from six regions (Germany, Macedonia, Norway, Turkey, Wales and Scotland) which applied the ISCO-coding in 2005/06, a sample of 33 504 young people aged 11–15 years. As the survey only asked for the current occupation of parents, we had to exclude those cases where both parents were unemployed or not economically active (n=1715; %=5.1), resulting in a likely underestimate of socioeconomic differentials.29 All missing values were excluded from the analyses (n=1810), except for the ISCO. Therefore, the net sample, excluding missing values for ISCO (n=5978), had 24 001 individuals aged 11, 13 or 15 years, and the gross sample, imputing missing values for ISCO, had 29 979 individuals (table 1).
Measures of SEP
Parental occupation and ISCO-based measures
Two open-ended questions assessed parents’ occupational status. Students were asked to indicate where their father and mother worked, and to describe what kind of job they did. Survey regions were required to categorise the answers using the International Standard Classification of Occupation (ISCO-88).19 These are considered to be good proxy reports of parental occupation.30–32 Three traditional measures of occupational status based on ISCO were examined: ISEI, SIOPS and EGP (table 1).20 ,21 The details for generating the occupational status measures are described in Ganzeboom and Treiman.21 Classification was by occupation of the ‘head of the household’, defined as whoever had the dominant occupational position.31 This method minimised the number of missing cases, as many mothers were not economically active (27.8%).
Additional measures of SEP
We considered two additional measures of SEP which have been used widely in the HBSC study and other adolescent health surveys: The FAS and subjective family well-off (table 1). The FAS is an index of material assets and expenditures of family wealth, based on four self-report items: family car, own bedroom, family holidays and family computer.6 Family well-off is a subjective indicator of families’ economic position, measured using the question ‘How well-off do you think your family is?’. The five response options are ‘very well-off (1)’ ‘quite well-off (2)’ ‘average (3)’ ‘not so well-off (4)’ and ‘not at all well-off (5)’.
Self-rated health (SRH) was measured by asking students to assess their general health with the question: “Would you say your health is excellent, good, fair or poor?” From this item, we created a dichotomous outcome measure (excellent/good vs fair/poor).8 SRH has been widely used as an indicator of subjective health in public health research.33 International studies showed that SRH is relatively stable through adolescence and a valid measure of physical and emotional dimensions of adolescent well-being.33 ,34
All analyses were carried out with STATA V.13.0 (StataCorp, Texas, USA). For each measure of SEP, the distribution of the variable and the level of missing data were examined overall and by region. To test the hypothesis that mean of measures of SEP were equal across regions multivariate Wald test of means was used with a p value of <0.05 regarded as statistically significant. Rejection of the null hypothesis implies that the mean level of SEP differed significantly between regions.
Pearson product-moment correlations were conducted for each region and for the total sample to explore the strength of the association between SEP measures, within and across regions.
The relationship between SEP and SRH was examined using logistic regression modelling procedures. All measures of SEP were standardised to a mean of zero and a SD of one. As, in general, the effects of SEP indicators did not statistically differ by gender and age, the analyses were not separated by subgroups. ‘Univariate’ models (models I–V) were used to calculate ORs for ‘poor SRH’ using each measure of SEP and adjusting for gender and age. Multivariable models (models VI–VII) were then created for each ISCO-based measure adjusting for gender, age, family affluence and family well-off to assess the separate effect of SEP measures. Finally, multivariable logistic regression models with imputed missing values for each ISCO-based measure were examined to assess the impact of missing values in the relationship between ISCO-based measures and SRH. A dummy variable for missing ISCO values and an interaction between imputed ISCO values and the dummy were included in the model, to evaluate whether the association between SRH and each ISCO-based measure differs between imputed and observed observations.
On the basis of missing at random assumptions, we used multivariate imputation by chained equations (MICE), in STATA V.13 (StataCorp. 2013).35 The MICE procedure replaces all missing values with randomly selected observed values. Missing ISCO values (ISEI/SIOPS/EGP) were imputed sequentially ten times, and for each region. The imputation model contained all variables of the full estimation model (age, gender, SRH). Owing to the scoring of the ISCO-measures, missing ISEI (16–90) and SIOPS (7–78) values were imputed using linear regression and missing EGP-scores (1–11) using ordered logistic regression.
To assess the cross-national generalisability of the relationship of SEP measures with SRH, regression models were examined for each region separately and for the entire sample combined. We used the seemingly unrelated estimation (SUEST) approach in STATA in order to determinate statistically significant differences in the effect of SEP measures across regions.36 The SUEST technically applied an Eicker-Huber-White-sandwich covariance technique and compared the results obtained (coefficients and corresponding SEs) across regions. Wald test with a p value of <0.05 indicated an significant difference in the effect of SEP measure across regions.
Average scores of the three ISCO-based measures, FAS and well-off varied significantly across regions and were not consistent in terms of their regional ranking (table 1). For instance, Norway showed the highest level of family affluence and family well-off, with moderate scores for ISCO-based measures. Higher rates of missing information (19.2–22.2%) were found for the occupational measures than for family affluence (0.2–5.5%) and family well-off (0.9–6.0%). Missing information for ISCO was found among less wealthy groups measured by family well-off and family affluence but did not differ by gender (results not shown).
A high correlation was observed between the ISCO-based measures with small differences across regions (table 2). In contrast, the correlation between ISCO-based measures and FAS/family well-off was relatively low in each region and in the total sample, for example, the highest correlation observed between FAS and ISEI was 0.44 in Turkey. A relatively low correlation was also found for the association between FAS and family well-off indicating that the ISCO-based measures, FAS and family well-off capture different aspects of adolescents’ SEP.
Table 3 shows the ‘univariate’ logistic regression models for poor SRH for each SEP measure by regions and for the pooled sample. In the pooled sample, higher levels of SEP were associated with lowered risk of poor SRH for all SEP measures. For instance, an increase in the ISEI-score by one SD was associated with an 18% decrease in risk of poor SRH. The strongest effects were observed for family well-off and FAS. Cross-regional differences in SRH risks existed in relation to all SEP measures. For example, while FAS was significantly associated with SRH in Turkey (OR: 0.67; p<0.001), this relationship was insignificant in Norway (OR: 0.94; p≥0.05). Similar variations were observed for all ISCO-based measures. However, the cross-national test of the SUEST indicated that the effects of FAS and family well-off differed significantly. There was no statistically significant cross-national difference in the effect of the ISCO-based measures.
When the models for the ISCO-based measures were additionally adjusted for FAS and family well-off, in the pooled sample, the effects of ISEI, SIOPS and EGP decreased, compared with the unadjusted models I–III in table 3, but continued to be significantly associated with SRH (table 4). The same pattern was observed for FAS and family well-off, which still showed the strongest association with SRH. For instance, an increase of family well-of by one SD was associated with a decrease in risk of poor SRH of 26% (models VI–VIII in table 4). The ISCO-based measures became insignificant in Macedonia, Norway and Turkey after adjusting for FAS and family well-off. The latter was the strongest predictor of adolescent SRH among all SEP measures, particularly in Turkey. The cross-national differences in the effects of ISCO-based measures increased after adjusting for FAS and family well-off, supported by the test on equality of coefficients.
Imputing data did not alter the findings (table 5). A socioeconomic gradient in SRH by ISEI, SIOPS and EGP remained. However, the effects of the imputed ISCO-based measures slightly decreased in comparison to models VI–VIII in table 3, where the dummy variable for missing ISCO values was significant, for example, in Germany and Norway. Furthermore, a positive interaction effect between the ISCO-based measures and the missing dummy was observed in Scotland and in the pooled sample. Therefore, the negative association between occupational status and poor SRH was marginally, but significantly weaker for adolescents with missing ISCO values.
This study is the first to assess the application of the ISCO classification and standardised measures of occupational status (ISEI, SIOPS, EGP) in a large cross-national health survey among young people. The results showed a significant social gradient in adolescent SRH for all three ISCO-based measures of parental occupation. Differences in the effects of ISEI, SIOPS and EGP were small. The associations between ISCO-based measures and SRH were not altered by FAS and family well-off. There was no apparent bias in the association between ISCO-based measures and SRH caused by ISCO missingness. Imputing missing values proved to be an adequate strategy to overcome problems caused by the high prevalence of missing parental occupation data, such as underpowering, in self-administered adolescent questionnaires.
Socioeconomic inequalities in adolescent SRH presented in the current study are consistent with findings of previous studies.8 ,37 We conclude that irrespective of the SEP measure, adolescents with lower SEP are more likely to have poor SRH. However, correlations between SEP measures were considerable low, suggesting status incongruence.9 The weak associations imply that the ISCO-based measures, the FAS and family well-off might capture different underlying characteristics of SEP. Accordingly, the ISCO-based measures seem to be more relevant for adolescent health in wealthier regions such as in Germany, Scotland or Wales, while family affluence seem to have more relevance in less wealthy regions such as in Turkey. This could be due to the fact that ISCO-based measures incorporate processes of job prestige and psychosocial factors,12 ,13 while family affluence is related to material conditions.6 ,10 In accordance with Wilkinson's argument, the psychosocial effects of relative position in social strata rather than absolute material standards are the major determinant of health in richer countries.38 The relevance of the ISCO-based measure and family affluence in adolescent health may therefore differ by context as these capture different dimensions of SEP.
The measure of family well-off combines experiences of absolute and relative socioeconomic conditions. It has been argued that subjective indicators of SEP reflect the internalised SEP identity, which incorporates past and current social circumstances along with future prospects.39 ,40 Therefore, subjective evaluations of SEP act as mediator between objective dimensions of SEP (such as the ISCO-based measures and family affluence) and health.41 Accordingly, family well-off was the strongest predictor of SRH in all model specifications and in each region and seemed to suppress the effect of the ISCO-based measures and family affluence. However, subjective evaluations are more sensitive towards personal influences such as affect or personality.40 ,42 Therefore, results for the association between SRH and family well-off might be biased due to confounding. In addition, the cross-national comparability of family well-off has not been analysed comprehensively to date. Although recent research indicates the same functioning of family well-off across countries, others found, for adults, that individuals tend to see themselves as being in the middle of the social hierarchy and that these self-perceptions are strongly influenced by countries’ economic conditions.18 ,43
The ISCO-based measures indicated a stronger cross-country comparability than family affluence and family well-off. ISEI, SIOPS and EGP were explicitly developed for cross-national studies of socioeconomic inequalities.21 While the ISCO-based measures use a standardised classification scheme, there is still some debate concerning the application of the FAS in cross-national studies.15 ,18 The SEP measures included in our study reflect different dimensions, and most likely different mechanisms through which SEP affects health.9 We therefore endorse the use of a range SEP indicators in health research, and in particular the use of ISCO measures validated for cross-national comparisons.
Although research has validated adolescents’ reports of parental occupation, collected by self-complete surveys,30 ,32 ,44 ,45 most studies report high levels of missing data.26 ,30 ,46 We also observed high levels of missing data on parental occupation, which were higher for less well-off adolescents.9 ,47 Analyses using multiple imputations methods for missing values in the ISCO scheme found no substantial difference in results. Therefore, we argue that multiple imputations might be a strategy to overcome problems caused by missing data in self-administered adolescent questionnaires on parental occupation.
The study has several strengths. It was conducted in several countries and regions in Europe and involved a large sample size of adolescents aged 11–15 years. There were a sufficient number of SEP measures included in the study to allow comparisons to be made for ISCO-based measures.
However, some limitations need to be acknowledged. Interpretation of the results assumed that all of the effects of FAS and family well-off had been removed from the relationship between the occupational-based measures of SEP and adolescent SRH. In line with previous research, we found a low correlation between the ISCO-measures, FAS and family well-off.9 However, from a conceptual point of view, parental occupation, FAS and family well-off must surely be related, though this relationship may not be linear, as occupation is the primary source of individuals’ income, and is necessary to establish a certain standard of living. Therefore, results should be carefully interpreted with regard to the separate effects of each SEP measure. Owing to the cross-sectional design of the study the causal direction between SEP and SRH could not be assessed. In accordance with previous studies, we assume that SEP predominantly predicts health in adolescence.48 However, the causal direction could be reversed, particularly in regions that do not meet the economic and educational needs of unhealthy individuals and particularly for subjective measures such as family well-off. Moreover, regional diversity was also lacking as the sample predominantly consists of high-income regions. Furthermore, we were not able to analyse the reasons for missing ISCO data as the HBSC did not provide sufficient information. As economically inactive people cannot be classified into the ISCO-based classification schemes, we had to exclude young people reporting both parents economically inactive from the analyses, which may have resulted in an underestimation of socioeconomic differences in adolescent SRH.29 However, results indicated that missingness in ISCO was only weakly associated with poor SRH in two out of six regions. Finally, we were not able to consider alternative measures of SEP such as parental income and education or adolescents’ school achievement.49 Therefore, we were not able to quantify the overall discriminant power of ISCO-based measures.
The ISCO-classification and ISCO-based measures of occupational status are valid measures of SEP in self-administered adolescent health surveys. These measure dimensions of SEP not captured by family affluence or family well-off. ISCO-based measures such as ISEI, SIOPS or EGP minimise problems of cross-national comparability, and confounding which subjective measures of SEP are prone to. They also have the advantage of being directly comparable with other surveys and routine data. Multiple imputations can be utilised to handle problems of missing data for parental occupation. Analyses which examine indicators reflecting different dimensions of SEP could illuminate the mechanisms behind social inequalities in adolescent health and well-being. Further research should aim to further examine the interrelation between subjective and objective measures of SEP in adolescent health surveys, and to evaluate the relevance of different ISCO-coding measures for adolescent health and well-being. In particular, research is needed to establish next steps in addressing socioeconomic causes of health problems in adolescence. Future research should extend the theoretical debate around, and empirically test, mechanisms linking different factors of SEP to adolescent health.7 ,8 Additionally, evaluations of policy interventions aiming to reduce inequalities in adolescent health may help to identify the key modifiable mechanisms.50 For example, the school setting as a central context of adolescents’ development in many countries gives the opportunity to implement and test direct interventions aiming at groups at risks.51 Learning from complex interventions enables researchers to increase their understanding of the existence and development of inequalities in adolescent health. The first step, however, in addressing inequalities should be to describe and to monitor inequalities as comprehensive as possible using a broad range of SEP measures as our study suggests.
What is already known on this subject
International studies showed a strong association between socioeconomic position (SEP) and adolescent health and health behaviours.
Measures of SEP were differently associated with adolescent health.
However, the cross-country comparability of measures of SEP in adolescent health surveys is a neglected area and further indicators are needed.
What this paper adds
Our study provides first evidence on the applicability the International Standard Classiﬁcation of Occupations (ISCO) classification and standardised measures of occupational status (International Socioeconomic Index of Occupational Status (ISEI), Standard International Occupational Prestige Scale (SIOPS), Erikson–Goldthorpe–Portocarero class categories (EGP)) in a large cross-national health survey among young people.
ISCO-based measures capture different aspects of socioeconomic position SEP compared to family affluence and family well-off.
ISCO-based measures were significantly associated with self-rated health independent from family affluence and family well-off.
Multiple imputation was an adequate strategy in overcoming problems of high missing values.
The Health Behaviour in School-aged Children (HBSC) study is an international survey conducted in collaboration with the WHO Regional Office for Europe. The current International Coordinator of the study is Candace Currie, CAHRU, University of St Andrews, Scotland. The data bank manager is Oddrun Samdal, University of Bergen, Norway. The data collection in each country was funded at the national level. We are grateful for the financial support offered by the various government ministries, research foundations and other funding bodies in the participating countries and regions.
Contributors TKP designed and wrote the first draft of the manuscript. SG analysed the data. MR initiated the idea and supervised the study. KAL and TT were involved in the analysis and interpretation of data. All authors critically reviewed the manuscript.
Competing interests None.
Ethics approval Ethical approval was obtained for each national survey according to the national guidance and regulations at the time of data collection.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement All data from this study are available to request through http://www.hbsc.org.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.