This paper identifies and estimates the impact of early retirement on the probability to die within five years, using administrative micro panel data covering the entire population of the Netherlands. Among the older workers we focus on, a group of civil servants became eligible for retirement earlier than expected during a short time window. This exogenous policy change is used to instrument the retirement choice in a model that explains the probability to die within five years. Exploiting the panel structure of our data, we allow for unobserved heterogeneity by way of individual fixed effects in modeling the retirement choice and the probability to die. We find for men that early retirement, induced by the temporary decrease in the age of eligibility for retirement benefits, decreased the probability to die within five years by 2.5 percentage points. This is a strong effect. We find that our results are robust to several specification changes.