Dimensions of Social Stratification and Their Relation to Mortality: A Comparison Across Gender and Life Course Periods in Finland

Hoffmann, Rasmus; Kröger, Hannes; Tarkiainen, Lasse; Martikainen, Pekka Article — Published Version Dimensions of Social Stratification and Their Relation to Mortality: A Comparison Across Gender and Life Course Periods in Finland Social Indicators Research Provided in Cooperation with: German Institute for Economic Research (DIW Berlin) Suggested Citation: Hoffmann, Rasmus; Kröger, Hannes; Tarkiainen, Lasse; Martikainen, Pekka (2019) : Dimensions of Social Stratification and Their Relation to Mortality: A Comparison Across Gender and Life Course Periods in Finland, Social Indicators Research, ISSN 1573-0921, Springer, Heidelberg, Vol. 145, Iss. 1, pp. 349-365, https://doi.org/10.1007/s11205-019-02078-z This Version is available at: https://hdl.handle.net/10419/205809 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. http://creativecommons.org/licenses/by/4.0/ Vol.:(0123456789) Social Indicators Research (2019) 145:349–365 https://doi.org/10.1007/s11205-019-02078-z 1 3 Dimensions ofSocial Stratification andTheir Relation toMortality: AComparison Across Gender andLife Course Periods inFinland RasmusHoffmann1,2 · HannesKröger1,3· LasseTarkiainen4· PekkaMartikainen2,4 Accepted: 21 January 2019 / Published online: 18 March 2019 © The Author(s) 2019 Abstract Differences in mortality between groups with different socioeconomic positions (SEP) are well-established, but the relative contribution of different SEP measures is unclear. This study compares the correlation between three SEP dimensions and mortality, and investigates differences between gender and age groups (35–59 vs. 60–84). We use an 11% random sample with an 80% oversample of deaths from the Finnish population with information on education, occupational class, individual income, and mortality (n = 496,658; 274,316 deaths between 1995 and 2007). We estimate bivariate and multivariate Cox proportional hazard models and population attributable fractions. The total effects of education are substantially mediated by occupation and income, and the effects of occupation is mediated by income. All dimensions have their own net effect on mortality, but income shows the steepest mortality gradient (HR 1.78, lowest vs. highest quintile). Income is more important for men and occupational class more important among elderly women. Mortality inequalities are generally smaller in older ages, but the relative importance of income increases. In health inequality studies, the use of only one SEP indicator functions well as a broad marker of SEP. However, only analyses of multiple dimensions allow insights into social mechanisms and how they differ between population subgroups. Keywords Social inequality· Mortality· Health inequality· Register data· Education· Income· Occupation· Socioeconomic position 1 Introduction Differences in health and mortality between groups with different socioeconomic positions (SEP) have been found in numerous studies and across gender, periods, ages, and countries (Elo 2009; Mackenbach etal. 2015). Measuring SEP is fundamental to such research; it Electronic supplementary material The online version of this article (https ://doi.org/10.1007/s1120 5-019-02078 -z) contains supplementary material, which is available to authorized users. * Rasmus Hoffmann [email protected] Extended author information available on the last page of the article 350 R.Hoffmann et al. 1 3 can be influenced by data availability, cross-country comparability, or simply by the interest in a specific dimension of SEP and its relation to health. Many authors have addressed the question as to which indicators of SEP might be most appropriate for studying health inequalities (Hoffmann etal. 2019; Krieger etal. 1997; Lynch etal. 2000), and also for studying other consequences of SEP such as the educational attainment of an individual’s children (Bukodi etal. 2014). The core argument is that SEP is inherently multidimensional, and consequently should not be measured with only one indicator; furthermore, different measures should not be treated as interchangeable (Braveman etal. 2005; Geyer etal. 2006; Goldthorpe 2010), but should be studied both separately and in conjunction (Bukodi etal. 2014). Our study contributes to this discussion by empirically examining the multivariate relations and associations between mortality and the three variables education, occupational class, and income in Finland, and assessing differences by gender and age. The concept of SEP as a multi-dimensional construct implies that all dimensions are correlated, but that each dimension has a unique relationship with health, which can be revealed by multivariate analyses, resulting in net-effects of variables that may be attributed to specific mechanisms which relate them to health. The effects of multiple SEP dimensions include the underlying temporal and causal relations between them (Galobardes etal. 2007). For example, education can have a direct effect on health, through knowledge about risk factors, risk behaviour, etc., but also an indirect one, via resources provided by occupational class and income (Lahelma etal. 2004). Thus, the net effect of a SEP variable is only part of its total effect. If SEP variables are treated as individual characteristics, it is important to “keep in mind that they are derived from larger social and economic processes that shape the distribution of education, occupation and income across the population” (Lynch and Kaplan 2000: 22). Social determinants of health derive their social significance and their relevance to health from social processes of attribution and distribution that take place on a non-individual level (household, neighbourhood, country), although a link to the individual level must persist to influence health. Our study has three aims: First, we establish the relative importance of three commonlyused SEP indicators and their related mechanisms for mortality using a high-quality dataset from Finland. Second, we study how these findings differ by gender and age. Third, we illustrate the magnitude of the error if only one SEP indicator is used and all health inequalities are interpreted as the result of one dimension, if results based on different single SEP indicators are compared, or if one SEP indicator is applied to different life cycle stages and gender. 2 Comparative Framework 2.1 Pathways fromDifferent Dimensions ofSEP toHealth Education imparts knowledge regarding health risks and healthy behaviour, and provides cognitive skills like self-efficacy for dealing with complex information, such as the effect of behaviour on health and dealing with healthcare institutions. Better education promotes reduced stress, as well as better coping and preventive behaviour (Hummer etal. 1998). Research on human capital has shown that education promotes cognitive and non-cognitive skill formation. These skills in turn facilitate the accumulation of health capital through self-regulation and choices (Cunha and Heckman 2007). 351 Dimensions ofSocial Stratification andTheir Relation to… 1 3 The positive health effect of education as ‘learned effectiveness’ is cumulative and selfamplifying, because education increases the sense of control, which shapes healthy life styles. The perceived success in controlling one’s health determinants (e.g. weight control) creates incentives for further investments in health and other life domains (e.g. sporting activity) (Mirowsky and Ross 2003). The observed overall association between education and health can be partly explained by material factors and behaviour, because higher education means higher income and more healthy behaviour (van Oort etal. 2005). Better educated people also have more rewarding and healthier jobs, which is another indirect effect of education on health (Mirowsky and Ross 2003). The use of education as an indicator for SEP is widespread, because of its simplicity, availability, and comparability, especially in internationally comparative studies (Eikemo etal. 2008; Mackenbach etal. 2015). However, its significance as an indicator of SEP and its benefit for health is disputed, partly due to the unclear differentiation between direct and indirect effects. Some authors argue that education is not a measure of SEP, but rather a mechanism by which individuals gain more dominant (or less dominated) positions in society, and that it has little direct effect on health (Bartley 2003; Blane 2006). Others claim that education is an important marker of SEP (Lynch and Kaplan 2000) and, if one also considers indirect effects via occupation and income, is the most important SEP dimension for health (Mirowsky and Ross 2003). Analyses of school reforms as natural experiments have demonstrated a small positive causal effect of education on health (Gathmann etal. 2015). Occupational class influences health through the social advantages that a job can provide, and through physical and mental health risks at the workplace. In most cases these two dimensions are congruent: jobs involving substantial health risks are also those with a lower occupational class position. An important argument for occupational class as indicator for SEP is that it is a relational variable reflecting superiority, equality, and inferiority in employment conditions, which is of major importance in modern societies (Goldthorpe 2010). It is a complementary measure to the attributional variables education and income (Goldthorpe 2012). Occupation is partly determined by education and determines income, and several studies suggest that occupation does not have much effect on health net of education and income (Bassuk etal. 2002; Warren and Kuo 2003). However, there is a long research tradition demonstrating specific causal pathways from occupations with a negative effort-reward balance to stress and heart disease (Siegrist etal. 1990). A related argument is that jobs that involve more productive self-expression, rather than self-suppression, favour health, and that better educated people are more likely to find such jobs (Mirowsky and Ross 2003). The practical implementation of occupation as an SEP indicator is limited by the fact that not all people work, be they homemakers or retired people no longer exposed to current work conditions. Occupational class is thus considered less important than education and income for retired people (Hoffmann 2008; Huisman etal. 2003), but it is unknown to what extent the association between past occupational class and health decreases after retirement. Income has been shown to be strongly associated with mortality (Martikainen etal. 2014; Tarkiainen etal. 2012). It influences health and mortality through the affordability of health care, environmental hazards, consumption (diet, housing), insecurity, and the psychological burden of being poor. Besides material explanations of the benefits of income, it also enhances effective capabilities, control, freedom, and the general ability to achieve goals (Mirowsky and Ross 2003; Robeyns 2011; Sen 1999). The effect of income on mortality has been found to be large compared to education and occupation (Duncan etal. 2002; Hoffmann 2011b). Nevertheless, there is also a strong association between education 352 R.Hoffmann et al. 1 3 and income, partly because higher education provides better opportunities on the labour market (Autor 2014). Consequently, income (or material conditions more generally) has indeed been found to partly mediate the effect of education and occupational class (Hoffmann 2011b; Mirowsky and Ross 2003; van Oort etal. 2005). The observed association between health and income has also been partly attributed to reverse causality from health to income. The relative strength of this pathway is debated (Galama and van Kippersluis 2010; Kröger etal. 2015; Martikainen etal. 2009). It is also argued that health inequalities that manifest via processes described above are also dependent on the distribution and investment of social and material resources at the societal level. These ‘neo-material’ factors include investments in public infrastructure such as education, health and welfare services (Lynch etal. 2000). In this regard, it is noteworthy that in international comparisons health inequalities are not systematically smaller in Finland and other Nordic welfare states than in other European regions and welfare regimes. For example, although Finland is characterized by tax-financed public provision of various social services such as child care, basic and advanced education, hospital care and health services for the elderly, average to high health inequalities are still observed (Andersen etal. 2007; Mackenbach 2012). In our empirical analysis, the SEP-specific effects and mechanisms correspond to the net effects, i.e. controlled for the other two SEP dimensions. In this regard, the three dimensions can be compared alongside each other. Conversely, the three dimensions are deeply intertwined, complementing and operating via each other, i.e. education operating via occupation and income, and occupation operating via income. The definition of net effects and total effects depends on the variables included in a specific study. Our design can partition the effects of education into net and total effects, the latter being partly mediated by occupation and income, but we cannot separate the effects of income into net effects of material factors and behaviour. 2.2 Gender Differences intheDeterminants ofHealth Health inequalities are usually larger among men than among women. This has been explained by (1) men’s greater involvement in spheres that create unequal and unhealthy living conditions, such as the working environment (Goldthorpe 1983), (2) a more unequal distribution of social resources, for example job status or income, and (3) these resources’ stronger impact on men, e.g. via unhealthy behaviour. McDonough etal. (1999) claim that the higher educational mortality gradient ensues from men receiving greater financial rewards from education. They show that the educational gradient is the same for both gender if income is controlled for. Although mortality differentials are generally larger for men we expect this gender difference to differ between SEP variables: mortality differences between educational levels should not differ much between men and women, because the direct effects of education on health (e.g. health knowledge) should be similar across gender, as suggested by McDonough etal. (1999). The occupational gradient might be smaller for women because, as explained above, unhealthy working conditions are more common and more unequally distributed for men, or it might be smaller for men, because superiority, equality, and inferiority in employment and the effort-reward balance can be more important for women, while men define their status and success more by income. This relates to our hypothesis on the income gradient that we expect to be larger among men, because material reward and success is possibly more important for men than women, and because more women can partly 353 Dimensions ofSocial Stratification andTheir Relation to… 1 3 rely on their partner’s contribution to the wealth of the household, which logically would result in a weaker association between women’s individual income and mortality. 2.3 Age Differences intheDeterminants ofHealth We simplify the perspective on age by differentiating between working ages (35–59) and retirement ages (60–84). Most available SEP indicators are related to the labour market. Education can be understood as input to, occupational class as the position in, and income as output of the labour market. In modern societies, the labour market is the most important system for allocating persons to socioeconomic positions. The question arises as to whether the association between different SEP dimensions and health changes after retirement, as has been suggested in the literature (Avlund etal. 2003; Grundy and Holt 2001). This change can be motivated by a life course perspective in which education, occupational class, and material resources not only appear in a rough temporal and causal sequence across the life course, but might also have specific life course phases in which they influence health. For example, material factors may be especially important for older people, because material factors take over the role of occupational class and status after retirement (Avlund etal. 2003). We hypothesise that the educational and occupational gradients in mortality are smaller in retirement than at working age, because the direct effects of education and occupation fade over time, and that mortality differences between income groups decrease much less and thus have increasing relative importance. When hazard ratios (HRs) for different SEP dimensions are compared between age groups, we are faced with at least two major underlying and opposing processes that hinder conclusions on the change of effects over age: First, accumulation of social and health advantages and disadvantages, which increases intra-cohort health inequalities and, second, ‘age-as-leveller’, a process which implies that health inequality decreases over time, either because of mortality selection, or because poor health and the biological processes of ageing dominate social influences (Hoffmann 2011a). The net effect seems to be that health inequality decreases at older ages, as most studies corroborate (Hoffmann 2005). We will examine this commonly found decrease for three dimensions of SEP to reveal changes in the importance of specific SEP dimensions during the life course. 3 Data Our data comes from an 11% sample from Finnish population registers aged 35 to 84years in 1987–2007, augmented by an 80% oversampling of those who died between 1987 and 2007. The oversampling is addressed by sampling weights, reflecting unequal sampling probability. From this sample we selected those who lived in Finland in 1995, with a 13-year follow-up period until 2007. Education and occupational class are measured in the baseline year, and income is measured before the baseline year as the average of annual income data from 1987 to 1995. Further variables are age, gender, date of death, employment status, language, and region. For details about the three SEP variables, see Table1. For details about all other variables, see Online Resource Table1. We exclude 1.6% of the sample who have no information on occupational class, of whom 87% are retired. We further exclude 1% students, 6.7% self-employed people, and 13.2% farmers, because, first, we could not define their hierarchical occupational class position and, second, because the income data of employed and self-employed persons is not directly comparable. Online 354 R.Hoffmann et al. 1 3 Resource Table2 shows the remaining sample size after each of these exclusions, which result in a total sample size of 496,658 individuals with 274,316 deaths. 3.1 Variables forSEP Education was coded as four categories of highest achieved education according to the 2007 classification of Statistics Finland, which is comparable to the International Standard Classification of Education (ISCED). Occupational class is coded in five categories, according to the classification of socioeconomic groups in the Finnish registers, which is similar to the Erikson-Goldthorpe-Portocarero classification (Erikson and Goldthorpe 1992). For retired, other non-employed, and unemployed persons, this variable refers to the occupation at the last quinquennial census for which occupational data was available. For the category labels of education and occupational class, see Table1. Income data is derived from the registers of the Finnish tax administration and social insurance institution, defined as individual income subject to state taxation (corrected for inflation, reference year 2000). It includes wages, capital income, and taxable income transfers, for example sickness benefits, but does not include tax-free benefits, such as child benefits (not means-tested), housing allowances, and social assistance. Income quintiles are Table 1 Summary statistics for all samples ISCED International Standard Classification of Education Percentages and death rates weighted for oversampling of deaths. For the frequencies of the income-quin- tiles, employment status, age, native language, region, and gender, see Online Resource Table1 Total sample Women Men Age 35–59 Age 60–84 Age 35–59 Age 60–84 Education (%, baseline year 1995) Primary, low secondary (ISCED 0–2) 46.9 34.5 75.1 35.6 69.6 Upper secondary (ISCED 3–4) 29.4 35.9 15.1 36.4 13.3 Lowest tertiary (ISCED 5) 12.2 17.5 4.6 12.5 7.7 Higher tertiary (ISCED 6–8) 11.6 12.2 5.2 15.4 9.5 100 100 100 100 100 Occupational class (%, in 1995) Non-specialized manual 22.9 18.8 33.4 20.9 24.5 Specialized manual 25.3 12.1 21.8 35.6 40.0 Lower white collar, non-managerial 13.7 26.9 11.7 5.0 3.2 Lower white collar, managerial 21.2 25.4 23.3 17.1 17.1 Upper white collar 16.8 16.8 9.7 21.4 15.3 100 100 100 100 100 Income In €, average between 1987–1995 18,266 17,516 11,588 23,197 17,154 Observations 496,658 105,293 148,396 127,036 115,933 Deaths between 1995–2007 (total: 274,316) 296,793 24,022 119,363 53,289 100,119 Weighted death rate 0.170 0.039 0.361 0.090 0.465 355 Dimensions ofSocial Stratification andTheir Relation to… 1 3 calculated from gender-specific income distributions. We used average income for 8years before the baseline year to ameliorate the problem of reverse causation from health to income. 4 Methods A Cox proportional hazard model (Cox 1972) is used to estimate hazard ratios (HR) that show the risk of dying in a certain SEP group relative to the reference group. The process time is calendar time. Individuals who are alive at the end of the observation period or who emigrate are censored. Age at baseline (equivalent to birth cohorts) is accounted for by controlling for five-year age groups. We calculate bivariate models, and multivariate models including education, occupational class, and income. After the main models applied to the whole sample, we run separate analyses by gender and by age group (35–59 and 60–84), comparing the results along these two dimensions. The age differentiation roughly separates working ages and retirement ages. In all models we control for three potential confounders: (1) 21 regions in Finland, that mainly take into account mortality differences between southwest and northeast Finland and the regional differences in health care provision, (2) language groups that reflect the Finnish speaking majority, the Swedish speaking minority (with a higher average SEP) and other languages, and (3) employment status, which is not part of our definition of SEP but highly correlated with all SEP dimensions and mortality. Since we are interested in the gross association between SEP and mortality, we do not control for other factors such as marital status, because they are more part of the mechanisms by which SEP influences health than a confounder. To investigate and to better compare the relative importance of the three SEP dimensions for health, we calculate the population attributable fraction (PAF), which takes the sizes of the categories of each variable into account. It expresses the total impact of a variable across its entire distribution and thus allows a better comparison of the three SEP variables. It does not assume a linear relationship between an SEP dimensions and mortality, as other commonly used summary measures for health inequality do (Relative Index of Inequality). Its calculation is based on hazard ratios for each risk category of an SEP variable, weighted by the relative size of the category (see formula below) (Miettinen 1974; Rockhill etal. 1998). It can be interpreted as the proportional reduction in overall mortality that would occur were everyone to hypothetically experience the rates of the highest socioeconomic group. Formula for the population attributable fraction n = number of exposure categories. Pi = proportion of population currently in the ith exposure category. HRi = Hazard Ratio for mortality in the ith exposure category. PAF = n ∑ i=1 Pi(HRi−1 HRi ) 356 R.Hoffmann et al. 1 3 5 Results Table1 shows the distribution of the variables in the total sample and by gender and age. Overall, the largest educational group are those with primary or low secondary education, but in the age group 35–59 the largest group are those with upper secondary education, which demonstrates the expansion of education across time and cohorts. A difference between age groups is also visible in the distribution of occupational class; higher classes are more common in the younger age group. Among men the largest group is specialized manual workers and only up to 5% are lower white collar non-managerial occupations, which represents a share of 26.9% among women aged 35–59. Among older women the largest occupational group is non-specialized manual workers. Income is higher among men and lower for older ages. In Table2 we show the correlation between the three SEP variables, for the total sample and by gender and age. All correlations are positive, are consistently higher among men and range between 0.33 and 0.60, showing that our analysis is not limited by too high collinearity. Table3 shows three bivariate models, one for each SEP variable, and then sequential models adding occupation, then income (for the HRs of all variables in the model and goodness-of-fit measures, see Online Resource Table3). In bivariate models, each of the three SEP variables show the expected association with mortality. The hazard ratio of 1.59 for the lowest educated group in the bivariate model (and also the HRs for the other educational categories) decreases by about one half when occupation is added to the model (HR 1.31) and decreases even more when income is added (HR 1.10). This shows that most of the total effect of education on mortality occurs via occupation and income. The share of the indirect effect (IND) from the total effect can be quantified with the formula IND = (ln(1.59) − ln(1.10))/ln(1.59). This results in 79% of the difference in the mortality hazard between primary education and higher tertiary education that can be accounted for by the inclusion of education and income. The respective numbers for upper secondary and lowest tertiary education are 93% and 90%. A similar situation can be found for lower occupational classes, where the HRs decrease by about half when including income. This suggests that almost half of the negative effect of low occupational class on mortality can be explained by low income. In particular, 43% of the difference in the mortality hazard between non-specialized manual workers and upper white collar workers can be accounted for by the inclusion of income. This is not true for lower white collar non-managerial jobs, which display zero health disadvantage, and lower white collar managerial jobs, where the disadvantage is not apparently related to income. Table 2 Spearman rank correlations between the SEP-variables education, occupational class and income Calculated on the 11% random sample, without oversampling Total sample Women Men Age 35–59 Age 60–84 Age 35–59 Age 60–84 Edu Occ Inc Edu Occ Inc Edu Occ Inc Edu Occ Inc Edu Occ Inc Education 1.00 1.00 1.00 1.00 1.00 Occupation 0.53 1.00 0.48 1.00 0.39 1.00 0.60 1.00 0.56 1.00 Income 0.47 0.45 1.00 0.34 0.37 1.00 0.36 0.33 1.00 0.42 0.49 1.00 0.51 0.53 1.00 363 Dimensions ofSocial Stratification andTheir Relation to… 1 3 References Andersen, T. M., Holmström, B., Honkapohja, S., Korkman, S., Tson, S. H., & Vartiainen, J. (2007). The Nordic model. Embracing globalization and sharing risks, ETLA Series B. Helsinki: The Research Institute of the Finnish Economy. Autor, D. H. (2014). Skills, education, and the rise of earnings inequality among the “other 99 percent”. Science, 344(6186), 843–851. Avlund, K., Holstein, B. E., Osler, M., Damsgaard, M. T., Holm-Pedersen, P., & Rasmussen, N. K. (2003). Social position and health in old age: The relevance of different indicators of social position. Scandinavian Journal of Public Health, 31(2), 126–136. Azimi, S. S., Khalili, D., Hadaegh, F., Yavari, P., Mehrabi, Y., & Azizi, F. (2014). Calculating population attributable fraction for cardiovascular risk factors using different methods in a population based Cohort study. Journal of Research in Health Sciences, 15(1), 22–27. Bartley, M. (2003). Commentary: Relating social structure and health. International Journal of Epidemiology, 32, 958–960. Bassuk, S. S., Berkman, L. F., & Amick, B. C. (2002). Socioeconomic status and mortality among the elderly: Findings from four US communities. American Journal of Epidemiology, 155(6), 520–533. Blane, D. (2006). Commentary: The place in life course research of validated measures of socioeconomic position. International Journal of Epidemiology, 35, 139–140. Braveman, P., Cubbin, C., Egerter, S., etal. (2005). Socioeconomic status in health research: One size does not fit all. JAMA, 294, 2879–2888. Breen, R., Karlson, K. B., & Holm, A. (2018). Interpreting and understanding logits, probits, and other nonlinear probability models. Annual Review of Sociology, 44(1), 39–54. Bruzzi, P., Green, S. B., Byar, D. P., Brinton, L. A., & Schairer, C. (1985). Estimating the population attributable risk for multiple risk factors using case–control data. American Journal of Epidemiology, 122(5), 904–914. Bukodi, E., Erikson, R., & Goldthorpe, J. H. (2014). The effects of social origins and cognitive ability on educational attainment: Evidence from Britain and Sweden. Acta Sociologica, 57(4), 293–310. Cole, S. R., & Hernán, M. A. (2004). Adjusted survival curves with inverse probability weights. Computer Methods and Programs in Biomedicine, 75(1), 45–49. Cox, D. R. (1972). Regression models and life tables. Journal of the Royal Statistical Society: Series B, 34, 187–202. Cunha, F., & Heckman, J. (2007). The technology of skill formation. American Economic Review, 97(2), 31–47. Duncan, G. J., Daly, M. C., McDonough, P., & Williams, D. R. (2002). Optimal indicators of socioeconomic status for health research. American Journal of Public Health, 92, 1151–1157. Eide, G. E., & Gefeller, O. (1995). Sequential and average attributable fractions as aids in the selection of preventive strategies. Journal of Clinical Epidemiology, 48(5), 645–655. Eikemo, T. A., Huisman, M., Bambra, C., & Kunst, A. E. (2008). Health inequalities according to educational level in different welfare regimes: A comparison of 23 European countries. Sociology of Health & Illness, 30(4), 565–582. Elo, I. T. (2009). Social class differentials in health and mortality: Patterns and explanations in comparative perspective. Annual Review of Sociology, 35, 553–572. Erikson, R., & Goldthorpe, J. H. (1992). The constant flux. A study of class mobility in industrial societies. Oxford: Clarendon Press. Galama, T., & van Kippersluis, H. (2010). A theory of socioeconomic disparities in health over the life cycle: RAND labor and population. Working Paper Series WR-773. Galobardes, B., Lynch, J., & Smith, G. D. (2007). Measuring socioeconomic position in health research. British Medical Bulletin, 81–82, 21–37. Gathmann, C., Jürges, H., & Reinhold, S. (2015). Compulsory schooling reforms, education and mortality in twentieth century Europe. Social Science and Medicine, 127, 74–82. Geyer, S., Hemström, Ö., Peter, R., & Vågerö, D. (2006). Education, income, and occupational class cannot be used interchangeably in social epidemiology. Empirical evidence against a common practice. Journal of Epidemiology and Community Health, 60(9), 804–810. Goldthorpe, J. H. (1983). Women and class analysis: In defence of the conventional view. Sociology, 17(4), 465–488. Goldthorpe, J. H. (2010). Analysing social inequality: A critique of two recent contributions from economics and epidemiology. European Sociological Review, 26(6), 731–744. Goldthorpe, J. H. (2012). Back to class and status: Or why a sociological view of social inequality should be reasserted. Revista Espanola de Investigaciones Sociologicas, 137(1), 201–216. Grundy, E., & Holt, G. (2001). The socio-economic status of older adults: How should we measure it in studies of health inequalities? Journal of Epidemiology and Community Health, 55, 895–904. 364 R.Hoffmann et al. 1 3 Haataja, A. (2006). Nordic breadwinner–caretaker models: Comparison of Finland and Sweden. Helsinki: Reports of the Ministry of Social Affairs and Health. Hoffmann, R. (2005). Do socioeconomic mortality differences decrease with rising age? Demographic Research, 13(2), 35–62. Hoffmann, R. (2008). Socioeconomic differences in old age mortality. Dordrecht: Springer. Hoffmann, R. (2011a). Illness, not age, is the leveler of social mortality differences in old age. The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 66(3), 374–379. Hoffmann, R. (2011b). Socioeconomic inequalities in old-age mortality: A comparison of Denmark and the USA. Social Science and Medicine, 72(12), 1986–1992. Hoffmann, R., Eikemo, TA., Kulhánová, I., Dahl, E., Deboosere, P., Dzúrová, D., van Oyen, H., Rychtaríková, J., Strand, BH., & Mackenbach, JP. (2013). The potential impact of a social redistribution of specific risk factors on socioeconomic inequalities in mortality - illustration of a method based on Population Attributable Fractions. Journal of Epidemiology and Community Health, 67(1), 56–62. Hoffmann, R., Kröger, H., & Pakpahan, E. (2018). The reciprocal relationship between material factors and health in the life course: Evidence from SHARE and ELSA. European Journal of Ageing, 15(4), 379–391. Hoffmann, R., Kröger, H., & Geyer, S. (2019). Social causation versus health selection in the life course: Does their relative importance differ by dimension of SES? Social Indicators Research, 141(3), 1341–1367. Huisman, M., Kunst, A. E., & Mackenbach, J. P. (2003). Socioeconomic inequalities in morbidity among the elderly; A European overview. Social Science and Medicine, 57, 861–873. Hummer, R. A., Rogers, R. G., & Eberstein, I. W. (1998). Sociodemographic differentials in adult mortality: A review of analytic approaches. Population and Development Review, 24(3), 553–578. Koskinen, S., & Martelin, T. (1994). Why are socioeconomic mortality differences smaller among women than among men? Social Science and Medicine, 38(10), 1385–1396. Krieger, N., Williams, D. R., & Moss, N. E. (1997). Measuring social class in US Public Health Research: concepts. Methodologies, and Guidelines, Annual Review of Public Health, 18, 341–378. Kröger, H., Pakpahan, E., & Hoffmann, R. (2015). What causes health inequality? A systematic review on the relative importance of social causation and health selection. European Journal of Public Health, 25(6), 951–960. Kröger, H., & Skopek, J. (2017). Logistic confusion: An extended treatment on cross-group comparability of findings obtained from logistic regression. Working Paper. https ://doi.org/10.13140 /RG.2.2.28652 .16006 . Lahelma, E., Martikainen, P., Laaksonen, M., & Aittomäki, A. (2004). Pathways between socioeconomic determinants of health. Journal of Epidemiology and Community Health, 58(4), 327–332. Lynch, J. W., Davey Smith, G., Kaplan, G. A., & House, J. S. (2000). Income inequality and mortality: Importance to health of individual income, psychosocial environment, or material conditions. British Medical Journal, 320(7243), 1200–1204. Lynch, J., & Kaplan, G. (2000). Socioeconomic position. In L. F. Berkman & I. Kawachi (Eds.), Social epidemiology (pp. 13–35). Oxford: Oxford University Press. Mackenbach, J. P. (2012). The persistence of health inequalities in modern welfare states: The explanation of a paradox. Social Science and Medicine, 75(4), 761–769. Mackenbach, J. P., Kulhánová, I., Artnik, B., Bopp, M., Borrell, C., Clemens, T., etal. (2016). Changes in mortality inequalities over two decades: Register based study of European countries. British Medical Journal, 353, i1732. Mackenbach, J. P., Kulhánová, I., Menvielle, G., Bopp, M., Borrell, C., Costa, G., etal. (2015). Trends in inequalities in premature mortality: A study of 3.2 million deaths in 13 European countries. Journal of Epidemiology and Community Health, 69(3), 207–217. Martikainen, P., Makela, P., Peltonen, R., & Myrskyla, M. (2014). Income differences in life expectancy: The changing contribution of harmful consumption of alcohol and smoking. Epidemiology, 25(2), 182–190. Martikainen, P., Valkonen, T., & Moustgaard, H. (2009). The effects of individual taxable income, household taxable income, and household disposable income on mortality in Finland, 1998–2004. Population Studies, 63(2), 147–162. McDonough, P., Williams, D. R., House, J. S., & Duncan, G. J. (1999). Gender and the socioeconomic gradient in mortality. Journal of Health and Social Behavior, 40(1), 17–31. McGovern, P., & Nazroo, J. Y. (2015). Patterns and causes of health inequalities in later life: A Bourdieusian approach. Sociology of Health & Illness, 37(1), 143–160. Miettinen, O. S. (1974). Proportion of disease caused or prevented by a given exposure, trait or intervention. American Journal of Epidemiology, 99(5), 325–332. 365 Dimensions ofSocial Stratification andTheir Relation to… 1 3 Mirowsky, J., & Ross, C. E. (2003). Education, social status, and health. New York: De Gruyter. Montez, J. K., Hayward, M. D., Brown, D. C., & Hummer, R. A. (2009). Why is the educational gradient of mortality steeper for men? The Journals of Gerontology, Series B: Psychological Sciences and Social Sciences, 64(5), 625–634. Mortensen, L., Rehnberg, J., Dahl, E., Diderichsen, F., Elstad, J. I., Martikainen, P., etal. (2016). Shape of the association between income and mortality. A cohort study of Denmark, Finland, Norway and Sweden in 1995 and 2003. BMJ Open, 6(12), e010974. Robeyns, I. (2011). Capability approach. In J. Pell & I. van Staveren (Eds.), Handbook of economics and ethics (pp. 39–46). Cheltenham: Edward Elgar. Rockhill, B., Newman, B., & Weinberg, C. (1998). The use and misuse of population attributable fractions. American Journal of Public Health, 88(1), 15–19. Ross, C. E., & Wu, C.-L. (1996). Education, age, and the cumulative advantage in health. Journal of Health and Social Behavior, 37, 104–120. Sen, A. (1999). Health in development. Bulletin of the World Health Organization, 77(8), 619. Siegrist, J., Peter, R., Junge, A., Cremer, P., & Seidel, D. (1990). Low status control, high effort at work and ischemic heart disease: Prospective evidence from blue-collar men. Social Science and Medicine, 31(10), 1127–1134. Tarkiainen, L., Martikainen, P., & Laaksonen, M. (2016). The contribution of education, social class and economic activity to the income–mortality association in alcohol-related and other mortality in Finland in 1988–2012. Addiction, 111(3), 456–464. Tarkiainen, L., Martikainen, P., Laaksonen, M., & Valkonen, T. (2012). Trends in life expectancy by income from 1988 to 2007: Decomposition by age and cause of death. Journal of Epidemiology and Community Health, 66(7), 573–578. Torssander, J., & Erikson, R. (2009). Marital partner and mortality: The effects of the social positions of both spouses. Journal of Epidemiology and Community Health, 63(12), 992–998. Torssander, J., & Erikson, R. (2010). Stratification and mortality: A comparison of education. Class, Status, and Income, European Sociological Review, 26(4), 465–474. Van der Weele, T. J. (2009). Marginal structural models for the estimation of direct and indirect effects. Epidemiology, 20(1), 18–26. van Oort, F. V. A., van Lenthe, F. J., & Mackenbach, J. P. (2005). Material, psychosocial, and behavioural factors in the explanation of educational inequalities in mortality in the Netherlands. Journal of Epidemiology and Community Health, 59, 214–220. Warren, J. R., & Kuo, H.-H. (2003). How to Measure “What People do for a Living” in Research on the Socioeconomic Correlates of Health. Annals of Epidemiology, 13, 325–334. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Affiliations RasmusHoffmann1,2 · HannesKröger1,3· LasseTarkiainen4· PekkaMartikainen2,4 Hannes Kröger [email protected] Lasse Tarkiainen [email protected] Pekka Martikainen pekka.mar[email protected] 1 European University Institute, Via Dei Roccettini 9, 50014SanDomenicodiFiesole, Italy 2 Max Planck Institute forDemographic Research, Konrad-Zuse-Straße 1, 18057Rostock, Germany 3 German Institute forEconomic Research, Mohrenstraße 58, 10117Berlin, Germany 4 Department ofSocial Research, Sociology, Center forPopulation, Health andSociety (CPHS), University ofHelsinki, PL 18, Unioninkatu 35, Helsinki, Finland