Subjective life expectancy: differences by smoking, education and gender

Sergei Scherbov, Wittgenstein Centre (IIASA, VID/ÖAW, WU) and International Institute for Applied Systems Analysis (IIASA)
Bruno Arpino, Universitat Pompeu Fabra
Valeria Bordone, University of Southampton

Despite the well-known higher mortality rates among smokers than non-smokers, little investigation has focused on subjective survival probabilities (SSP) by smoking behaviour. We give attention to sub-group differences in subjective survival probabilities, comparing subjective predictions to objective ones (SP) and accounting for the role of education. We use biannual data from the Health and Retirement Study (HRS) from 2000 to 2012 carried out in the USA. Based on a sample of 23,895 respondents aged 50-89, we calculate, for each respondent, the “gap” between SSP and the estimated survival probability (SP) from the HRS data. We find that people currently smoking report lower survival probabilities especially if they are low educated. This is consistent with real mortality data that show higher mortality among these groups. When comparing subjective and objective survival probabilities we find that irrespectively of the smoking status, high educated people are more likely to correctly predict their survival probabilities than their low educated counterparts. Within education groups, people who smoked in the past are the best at predicting their mortality. Interestingly, those who currently smoke show the highest probability to incorrectly overestimate their survival probability (i.e., to underestimate the negative effect of smoking on mortality).

See extended abstract

 Presented in Session 100. Analyzing life expectancy and lifespan: factors and methods