Logo Logo
Switch Language to German
Kratz, Fabian; Patzina, Alexander (2020): Endogenous Selection Bias and Cumulative Inequality over the Life Course: Evidence from Educational Inequality in Subjective Well-Being. In: European Sociological Review, Vol. 36, No. 3: pp. 333-350
Full text not available from 'Open Access LMU'.


According to theories of cumulative (dis-)advantage, inequality increases over the life course. Labour market research has seized this argument to explain the increasing economic inequality as people age. However, evidence for cumulative (dis-)advantage in subjective well-being remains ambiguous, and a prominent study from the United States has reported contradictory results. Here, we reconcile research on inequality in subjective well-being with theories of cumulative (dis-)advantage. We argue that the age-specific endogenous selection of the (survey) population results in decreasing inequalities in subjective well-being means whereas individual-level changes show a pattern of cumulative (dis-)advantage. Using repeated cross-sectional data from the European Social Survey (N = 15,252) and employing hierarchical age-period-cohort models, we replicate the finding of decreasing inequality from the United States with the same research design for Germany. Using panel data from the German Socio-Economic Panel Study (persons = 47,683, person-years = 360,306) and employing growth curve models, we show that this pattern of decreasing inequality in subjective well-being means is accompanied by increasing inequality in intra-individual subjective well-being changes. This pattern arises because disadvantaged groups, such as the low educated and individuals with low subjective well-being show lower probabilities of continuing to participate in a survey and because both determinants reinforce each other. In addition to allowing individual changes and attrition processes to be examined, the employed multi-cohort panel data have further key advantages for examining inequality in subjective well-being over the life course: They require weaker assumptions to control for period and cohort effects and make it possible to control for interviewer effects that may influence the results.