The paper studies retirement behavior of wage-earners in Belgium - for the first time using rich survey data to explore retirement incentives as faced by individuals. Specifically, we use SHARE data to estimate a model à la Stock and Wise (1990). Exploring the longitudinal nature of SHARELIFE, we construct measures of financial and non-financial incentive. Our analysis explicitly takes into account the different take-up rates of the various early retirement exit paths across time and ages. The results show that financial incentives play a strong role. Health and education also matter, as does regional variation - though the latter in an unexpected way. A set of policy simulations illustrate the scope and also the limits associated with selective parametric reforms.