Cohabitation, divorce, and the trial marriage hypothesis

Previous research on the link between cohabitation and divorce in the United States unanimously rejects the trial marriage hypothesis (TMH), which posits that premarital cohabitation decreases the risk of divorce. This dissertation reconsiders the rejection of the TMH by arguing that previous work has confused selection with causation both on the theoretical and on the empirical level. The theoretical part of this dissertation parses a standard version of the TMH from an explicitly causal perspective and offers three novel conclusions. (1) The TMH implies that cohabitation reduces the risk of divorce among couples that cohabit first and then marry, but this effect is entirely due to selection . (2) The TMH does not imply that cohabitation causally reduces the risk of divorce among couples that cohabit first and then marry compared to identical couples that marry right away. (3) The TMH does imply that cohabitation causally reduces the risk of divorce by preventing the marriages of some couples that would have divorced, had they married right away in lieu of cohabitation. In short, this dissertation argues that the causal mechanism by which cohabitation is predicted to reduce the risk of divorce relies on marriage prevention rather than marriage improvement. The methodological part of this dissertation argues that previous empirical work by design cannot shed light on either the selection implication or the causal implication of the TMH, and then proposes a series of more suitable statistical tests. The empirical part of this dissertation implements these new statistical tests on the 1995 National Survey of Family Growth and reports three main empirical results: (1) Cohabitation does not appear causally to reduce the risk of divorce among couples that cohabit first and then marry. This result is consistent with the TMH. (2) Cohabitation does not appear to reduce the risk of divorce via selection. This result is inconsistent with the TMH. However, this test suffers from severe data limitations. (3) Cohabitation does appear causally to reduce the risk of divorce by preventing the formation of some doomed marriages. This result confirms the central causal implication of the TMH.