This paper investigates the impact of delaying retirement on mortality among the French population. We take advantage of the 1993 pension reform in the private sector to identify the causal effect of an increase in claiming age on mortality. We use administrative data which provide detailed information on career characteristics, dates of birth and death. Our results, precisely estimated, show that an exogenous increase of one year in the claiming age has no significant impact on the probability to die, measured between age 61 and 79, even when we allow for nonlinear effects of treatment intensity. To test the power of our sample to detect statistically significant effects for rare events like death, we compute minimum detectable effects (MDEs). Our MDE estimates suggest that, if an impact of later retirement on mortality would be detectable, it would remain very small in magnitude.

We investigate the causal effect of retirement on health through literature. We explore the potential mechanisms which explain three effects: the switch from employment to retirement, later retirement, and earlier retirement. The empirical strategies used to identify the causal effects are mainly based on the observation of changes in health status at the legal age for retirement entitlement or on reforms that have led to changes in retirement incentives. Literature renders possible to make several observations on the average effect estimation. Retirement leads to better self-reported health, less depression, a decrease in healthcare consumption, a decline in cognition and an ambiguous effect on physical health. Later retirement has no effect on mortality, decreases healthcare consumption, and has a negative or non-significant impact on self-reported health. Studies on the impact of earlier retirement are scarce due to few natural experiments exploiting such a variation. Lastly, some studies find evidence of heterogeneous effects by gender and occupational status. As there are relatively few studies on this aspect, the question should be seriously explored in future research.