Mediation and Decomposition Analysis: Why We Cannot Do What We Think We Can Do, and How Causal Mediation Analysis Can Help Us

Research often aims to explain the share of a total effect that arises from intervening mediating mechanisms. This research frequently aims to identify the most important mechanism by way of either a conventional version of the Kitagawa–Oaxaca–Blinder decomposition approach or the Karlson-Holm-Breen method. As these approaches treat different mediators as causally independent, they yield biased estimates if systematic interrelations between mediators are present. I argue that interdependencies between mediators are the rule rather than the exception, and outline how novel methods of causal mediation analysis can be used to identify the most promising targets for intervention when mechanisms causally affect each other. I use gender, race, and origin-specific disparities to develop the argument. Building on this discussion, I use data from the Survey of Health, Ageing, and Retirement in Europe (SHARE) to delve into a nuanced empirical analysis of origin-specific health disparities at old age. Overall, the study demonstrates how we can assess which intervention targeting which mechanism is most promising to reduce a social disparity if mechanisms affect each other.