SOEPpapers on Multidisciplinary Panel Data Research 491
Background: This study investigates possible mechanisms that can explain the association between unemployment and smoking, that is a) unemployment increases smoking probability (causation), b) smoking increases the probability to become unemployed (selection), and c) differences in both smoking and unemployment probabilities trace back to differences in socio-economic position (common cause). Methods: Longitudinal data from the German Socio-Economic Panel (SOEP) from the years 1998, 1999, 2001, 2002, 2004, 2006, and 2008 were used to examine the effect of unemployment on smoking probability and vice versa (65,823 observations from 18,735 respondents, aged 18-60 years). Effects were estimated by using random and fixed effects logistic panel regressions. Results: Results from the random effects logistic regression model suggest that unemployed have a higher probability to smoke and that smokers have a higher probability to become unemployed. However, the fixed effects models indicate that the observed associations are driven by unobserved factors. Results indicate that both smoking and unemployment probability co-vary systematically with (childhood) socio-economic position. Conclusion: In contrast to previous studies, the present investigation suggests that there is neither a direct causal effect of unemployment on smoking behaviour nor a direct effect of smoking on unemployment probability. Rather, smoking and unemployment seem to be related through a common cause, with people from low socio-economic backgrounds being more likely to smoke as well as to become unemployed. These findings are interpreted in the frame of a life course perspective on the development of socially unequal health behaviours.
health behaviour smoking unemployment longitudinal analysis life course health inequality fixed effects random effects