The determinants of marital instability is an important area of research for demography, sociology and economics, with a host of public policy outcomes being significantly affected by family breakdown. This paper improves our understanding of the issue through the use of rich longitudinal data and the application of advanced research approaches. In both method and data terms our approach represents a significant advance in this research area. Using data from waves 1–7 of HILDA, 2,482 married couples—where both partners are respondents in the first wave—are traced over six years to identify factors associated with marital instability. The data are analysed dyadically; that is, the characteristics of both partners in each couple are considered in tandem. This allows assessment of whether marriages between partners with similar characteristics (homogamy) are more likely to last than are marriages between dissimilar partners, or whether particular characteristics of wives or husbands—independent of their partners’—are more strongly associated with marital stability. A Cox proportional hazards model with time-varying covariates is used to assess the association of characteristics with marital separation. We find the following factors are associated with higher risk of marital separation: spousal differences in age, education, preference for a(nother) child, and drinking and smoking behaviours; dissatisfaction with the relationship; low household income; husband’s unemployment and perceived financial stress; young age at marriage; separation of parents; second-plus marriage; and resident children born before marriage.