| Abstract: | This paper introduces a novel extension of mover-stayer models for duration data that allows time-dependent covariates to be used for both a pair of regression equations, one that identifies the determinants of event timing and one that identifies the determinants of the probability of ultimate event nonoccurrence. Existing models intended to distinguish covariate effects on event timing from those on event nonoccurrence cannot use time-dependent covariates in the equation for the probability of ultimate event nonoccurrence. This paper applies the new model to an analysis of remarriage among American women. The analysis generally demonstrates that some covariates effect remarriage timing while others affect the probability of ultimate remarriage nonoccurrence. Some differences in patterns of remarriage between black women and white women are also reported. Theoretical implications of these findings are discussed. |
