| Summary: | This chapter investigates the efficacy of current state-of-the-art life expectancy modeling in projecting life expectancy at “the” pension age—that is, the age at which a life annuity must be granted according to current practice in NDC (and FDC) schemes. We provide an overview of current modeling philosophy and of the genre of projection models inspired by the work of Ronald Lee and Lawrence Carter (1992). We demonstrate that models of this type, which essentially distribute an aggregate trend among the birth cohorts covered, will systematically underestimate life expectancy in an environment characterized by declining rates of mortality—which is a rather typical scenario. This has certainly happened in Japan, but also in countries such as Finland, Norway, and Sweden. The authors provide suggestions for better modeling under these circumstances but acknowledge that, regardless of the proficiency in modeling, systematic errors may continue to be part of the landscape for many decades. The authors ask whether it is possible to devise annuity models that fairly distribute the residual risk between the insured and the insurer. Simple variable annuity models, in which the annuity is reestimated at yearly and five-year intervals on the basis of continuously revised life expectancy estimates, are examined for the Scandinavian countries. The variable annuity reduces the risk for the insurer—other cohorts in a mutual insurance setting, or the government (taxpayers)—but at the relative expense of older members of the birth cohort. This is clearly not a fair outcome and suggests that we need to learn more about other models of distributing the systematic error in estimating longevity and that a more palatable safeguard is to transfer the systematic estimation error to the government (taxpayers) through an NDC bond.
|
|---|