Applying frailty models to analyze educational differences in mortality in Turin

Virginia Zarulli, Università di Roma "La Sapienza"
Chiara Marinacci, ASL TO3, S.C. a D.U., Public Health School, Turin
Graziella Caselli, Università di Roma "La Sapienza"

According to the literature every individual has a level of unobserved frailty that influences his mortality risk. The observed population hazard is a mixture of the individual hazards. The individual hazard increases faster than the population one because frailer individuals die faster causing mortality deceleration at old ages at the population level. Unobserved heterogeneity also influences the estimated coefficients in regression models. Using 36 years individual follow up data between 1971 and 2007 for Turin, Italy, a city of 1 million inhabitants, we study the impact of selection on educational mortality patterns at old ages. We use survival analysis for right censored and left truncated data. Our aim is to assess how the estimates of the mortality differentials change due to unobserved heterogeneity. We want to test the hypothesis that decreasing effect of socio-economic variables with age is an artifact of selection at the population level as well.