The goal of this study was to conduct a joint analysis of parental alcoholism and parental separation on offspring early alcohol, tobacco, and illicit drug involvement in a population-representative U.S. sample of European and African ancestry female twins. Using a survival-analytic framework, we compared offspring from nonalcoholic intact families to offspring from (i) alcoholic separated, (ii) alcoholic intact, and (iii) nonalcoholic separated families. We also employed propensity score methods to infer, within a counterfactual framework, whether observed risks are unique to parental separation or due to unmeasured confounders, including parental alcoholism. The importance of considering counterfactuals has long been recognized (e.g., Rosenbaum and Rubin, 1983), but rarely implemented in addictions research (Heath et al., in press). By successfully matching offspring from separated and intact families across a range of predicted probabilities, based on predictors such as parental alcoholism, our confidence in the specificity of risks associated with parental separation is greatly increased; to the extent that we are unable to match across the spectrum of risk of separation, our confidence in such effects is undermined.
