In behavioral genetics, divorce is typically analyzed as an individual-level outcome, even though marital dissolution can only be experienced by couples. In this article, we discuss how assortative mating complicates the study of couple-shared outcomes because individual-level effects can be confounded by effects of the spouse. We then show how chain-linking affines (i.e., in-laws) provides sufficient information to estimate spousal similarity for couple-shared outcomes, which we incorporate into an extended twin model that we use to test for sex differences and assortative mating for individuals' liability to divorce. We linked the Norwegian twin register to the Norwegian population register and constructed 124,544 extended family units (1196 units with monozygotic twins) comprising 353,210 marriages entered between 1983 and 2008. We found that divorce was significantly correlated among affines, and that female relatives were more highly correlated than male relatives. The extended twin model estimated a strong correlation (r = .60, SE = .10) between female and male familial factors. Couples' liability to divorce was attributed to 18% (SE = 5%) female and 10% (SE = 3%) male familial factors, with an additional 16% (SE = 4%) accounted for by their correlation. Estimates from a classic twin model were considerably higher. These findings show that spousal similarity is an important source of variation in divorce liability and that failing to model it can inflate estimates of individual-level effects. Overall, the analytical framework offers a blueprint for dissecting any couple-shared outcome into sex-specific and assortative components.
Sunde et al. (Mon,) studied this question.