Endogeneity is a major concern when modelling the relationship between health and labor market decisions. Health might be endogenous to labor due to justification bias or due to simultaneity. We employ a simultaneous equations model to analyze the causal effects of health and working in old age, allowing for endogeneity. In addition, we exploit the time dimension of the panel data to control for unobserved heterogeneity. Unlike the previous studies, we distinguish between working full-time and part-time to analyze the potentially heterogenous effects of working different numbers of hours. The results confirm the finding in the literature that better self-reported health has a positive effect on the probability of working. As for the reverse effect, we find that working full-time has a negative effect while working part-time has a positive effect on the general health of the elderly.