Subjective survival probabilities and life tables: an empirical analysis of cohort effects

Abstract

This paper studies potential bias and heterogeneity in subjective survival probabilities. To analyze potential bias, average subjective survival probabilities from a representative survey of the Italian working-age population are compared to longitudinal predictions from cohort life tables constructed using a long sequence of cross-sectional life tables. Subjective beliefs for men are reasonably close to the longitudinal predictions from the life tables, whereas subjective probabilities for women are always lower than the life-table predictions. As for heterogeneity across individuals, subjective survival probabilities depend negatively on self-reported health problems, not being married, holding a risky job, and being unemployed, and depend positively on age and satisfaction with job security and stability. Somewhat surprisingly, education does not seem to matter, even after controlling for sample selection due to item nonresponse.