Annuitisation and Mortality: Disentangling Selection from Behaviour

We re-examine the widely documented positive correlation between annuitisation and longevity by cleanly separating selection from causal effects—an empirical distinction that the literature has acknowledged but has not been able to identify credibly. Using administrative micro–data from Chile’s centralized pension quote system, we exploit monthly variation in the relative generosity of annuities versus programmed withdrawals generated by regulated PW formulas and market-based annuity pricing. These shocks serve as plausibly exogenous instruments for annuitisation in an IV bivariate probit model of post-retirement survival. Across horizons up to fifteen years, we find no statistically significant causal effect of annuitisation on survival for either men or women. The results imply that the observed longevity advantage of annuitants in Chile reflects selection rather than behavioral responses. By isolating the causal channel, the paper contributes to ongoing discussions on decumulation design by showing that moral-hazard-driven survival effects are unlikely to be a relevant policy concern in this setting