“…This implies the necessity of implementing a model with more than one degree of heterogeneity. The models of , Fehr and Uhde (2013), Fehr and Uhde (2014), Laun et al (2019), and more recently Jones and Li (2023), in which agents face idiosyncratic income risk, disability risk, and mortality risk by the skill group, and the distribution of skill groups is the same across cohorts, are potential candidates. However, in reality, education is changing across birth cohorts, which may cause that the observed increasing gap in life expectancy by educational attainment is just driven by the fact that the low-educated group becomes more negatively selected over time (Goldring et al, 2016;Hendi et al, 2021).…”