Estimating the Effect of Premarital Cohabitation on Timing of Marital Disruption

Abstract

In this article, we extend the propensity score method by matching on multiple groups. Using data from first wave (1987–1988) and third wave (2001–2003) of National Survey of Families and Households (NSFH), we match married individuals with no premarital cohabitation, single premarital cohabitation with the spouse, and serial premarital cohabitations, and apply Cox proportional hazards models to explore how premarital cohabitation history affects marital disruption. Our results indicate that both selection and causation help explain the relationship. The selection effect played a large role in 1987–88 when cohabitation was uncommon but disappeared in 2001–03 when cohabitation became prevalent. Postmatching results demonstrate that the causal effect of cohabitation on marital disruption was strong among serial cohabitors and weak among one-time cohabitors with the spouse. The imputation-based sensitivity analysis shows that our conclusion is robust even with the presence of unobserved characteristics that have a moderate association with cohabitation and marital disruption.