This paper introduces novel log-linear fixed-effect latent-trait Markov-chain models for the panel- data analysis of transitions among dichotomous states and applies them to an analysis of the effects of divorce/widowhood on changes in personal-efficacy levels. Sufficient statistics of individual latent traits for conditional likelihood estimation are derived for each model. These models are developed to distinguish the covariate effects on upward mobility from those on downward mobility in dichotomous states of the dependent variable, controlling for unobserved heterogeneity caused by latent individual traits.

This distinction is theoretically important in the application presented in this paper, which separates the effect of divorce/widowhood on increasing the vulnerability of personal efficacy (i.e., on increasing the probability of downward mobility in personal efficacy) from its effect on increasing the suppression of personal efficacy (i.e., on decreasing the probability of upward mobility in personal efficacy). The application demonstrates that divorce/widowhood increases the vulnerability of personal efficacy but decreases the suppression of personal efficacy.

* University of Chicago