Despite its clear relevance and policy significance, there is still sparse evidence on the effects of ill-health on the dynamics of labour state transitions among older individuals. We provide novel evidence by considering retirement as mobility among full-time work, part-time work, self-employment and inactivity, using a dynamic multinomial choice model that simultaneously accounts for state dependence, individual-level and state-specific unobserved heterogeneity, captivity and correlations between labour market states. We also simulate the dynamic paths for the four labour states from both transitory and permanent health shocks. We find strong state dependence for all four labour states even after accounting for individual effects. Both ill-health and health shocks are found to greatly increase the probability of leaving full-time employment into inactivity, and we find some evidence of part-time and self-employment paths. Significant evidence is found for “captivity” effects for the “inactive” state, and correlations across labour states. We also show that the degree of state dependence is over-estimated and, for men, the effects of ill- health under-estimated, if unobserved individual effects are not controlled for in dynamic models.

With Mark N. Harris, Curtin University, Perth, Australia and Xueyan Zhao, Monash University, Melbourne, Australia