Inaccurate longevity expectations can result in suboptimal lifecycle planning with negative consequences for wellbeing in old age. We evaluate the accuracy of expectations by comparing the subjective probability of living to 75 reported in the Health and Retirement Study with survival to that age. Outcomes are positively correlated with the subjective predictions. But the prediction errors are large. Even larger than if every respondent had reported a 50–50 chance of survival. The subjective predictions are less accurate than life table probabilities based on age and sex only. On average, the respondents underestimate their chances of survival by almost 9 percentage points, with women displaying greater downward bias. Predictions are least accurate, most poorly calibrated and most noisy among the least educated and least cognitively able. By and large, using the subjective predictions to take decisions would generate less value than if everyone were to decide on the basis of the mean survival rate, or even the life table probability. Despite the predictions of the least educated being the least accurate, they are not unambiguously the least valuable, although they are most prone to generating large losses from excessive pessimism and optimism.

JEL classification:
D83; D84; I12; J14

Keywords:
Longevity; Expectations; Subjective probability; Education; Cognition