This paper aims to assess empirically the relative size of incentive effects and health for the retirement decision. We specify and estimate a dynamic model for retirement behaviour that explicitly takes account of health and eligibility conditions and replacement rates of alternative exit routes from the labour force. A range of health instruments are constructed from estimates of a model for health dynamics and these are used to assess the effect of reporting errors and of endogeneity of health on the estimates of the retirement model. Our results provide evidence that health and retirement are endogenously related. Health matters but the size of the health effect depends crucially on the health measure used. We find that subjective health measures overstate the effect of health on retirement and that endogeneity of health suppresses the health effect. Incentive effects are relatively insensitive to alternative specifications for health. The incentive effects are strong for early retirement (ER) schemes. There is evidence that income streams in alternative exit routes are compared in the retirement decision and that alternative exit routes act as substitutes.