Journal Article

On stochastic comparisons of population densities and life expectancies

Abstract

Cohort measures, describing the lifetime random variable are easily and unambiguously obtained using standard tools. For non-stationary populations the corresponding conventional period measures are biased. The origin of this “tempo bias” (Bongaarts and Feeney (2002)) is in the fact that unlike the cohort setting, the lifetime random variable for the period setting cannot be unambiguously defined without additional simplifying assumptions. We consider this topic using stochastic ordering of population age distributions. Some useful inequalities are obtained and applied to ordering life expectancies defined in different ways.