What Longevity Predictors Should Be Allowed for When Valuing Pension Scheme Liabilities?

Data held within administration records of occupational pension schemes yield a rich source of information on mortality rates and the statistical predictors (covariates) of longevity. In this paper we provide, for the first time, a multivariable analysis of post-retirement mortality using the detailed information held within occupational pension scheme records. Using the extensive dataset of over one million living pensioners and dependants and 530,000 historic deaths collected by Club Vita, we investigate the importance of factors including gender, affluence and lifestyle on the observed period life expectancy of individuals. We describe one approach to constructing a multivariable model for pensioner baseline mortality, showing how such factors explain a variation in observed period life expectancy in excess of ten years. The relative importance of each factor on mortality is analysed and we describe the interactions between these factors and age, answering questions such as whether the impact of a healthy lifestyle or affluence attenuates with age. Further, we highlight the importance of the choice of affluence measure in analysing mortality, and show that the salary at retirement is a better predictor of longevity than the pension amount for male pensioners. The results of this paper are directly relevant to any pensions actuary advising on an appropriate baseline assumption (i.e. current mortality rates) for use with occupational pension schemes.