Modeling and Projecting Mortality. A New Model of Heterogeneity and Selection in Survivorship

The demographic and epidemiological literature offers abundant examples of a range of shortcomings of statistical modeling to describe mortality by sex, age, time/cohort, and cause-of-death. Statistical modeling of mortality operating with implicitly homogenous sub-groupings exposed to mortality risk fails to consider latent biological heterogeneity at the level of individuals, and thereby important biological and social selection of survivorship. Defined on the state space of the simple life model, this study presents a proportional hazard model that makes up for such drawbacks as far as latent biological heterogeneity is concerned. The model describes heterogeneity and selection in individual survivorship by iterative stochastic micro simulation using cohort-based population mortality as an empirical benchmark. The model offers efficient linkage between past assorted mortality, on one hand, and informed anticipation of future heterogeneous survivorship, on the other hand. The combination of stochastic micro-simulation and log-linear modeling of the period effect or trend uncovered under the model makes the new Heterogeneity and Selection Model a powerful analytic and predictive tool of survivorship. Postulating a trend independent of age makes the popular Lee-Carter model (1992) unfit for professional demographic and actuarial use. Moreover, by sweeping latent biological heterogeneity under the rug, mortality analysis and projection based on central rates such as the Lee-Carter model (1992) underrates mortality in the mature and elderly ages. This is demonstrated by comparing current official mortality projections of Sweden, Denmark, and England & Wales to a set of alternative mortality projections under the Heterogeneity and Selection Model