“…Independent variables captured (a) the impact of the baseline (i.e., the duration of the union over the years of the BHPS); (b) the effects of a time‐varying variable that is a continuous function of the duration of the BHPS (e.g., the age of the partners, number of previous migrations/moves, time since the last migration/move, the length of the union in months from the time it started, and the age of the couple's children); (c) the values of a time‐constant variable (e.g., gender, race, religion age when union started, age difference in the couple, and attitudes to gender roles); and (d) the effects of time‐varying variables whose values can change only at discrete times (e.g., level of education, employment status and occupational status and the changes in those). To test the proportional hazard assumption, we fitted models where some covariates (e.g., distance of migration, reasons for migration, and changes in both partners' employment and occupational characteristics) have both time‐invariant and time‐variant components (i.e., the main effect and the interaction with the time variable; Statacorp 11 2009; Longhi & Nandi, ; Boyle, Feng, & Gayle, ). To control for the clustering of events within individuals and possible unobserved determinants of union dissolution, we fitted our models with robust standard errors.…”