Modelling longevity risk. A practical study of the effect of statistical pre-adjustments on mortality trend forecasts

An important risk in the actuarial industry is the longevity risk, therefore the as accurate as possible prediction of mortality rates is very crucial. Such predictions are performed by modelling the mortality rates using mortality models and predicting the future mortality trends. Aiming at possible improvements of such forecasts, we examine the effect of data transformation-'linearisation' on the quality of time series forecasts of mortality, using data resulted from mortality models for England-Wales. By time series 'linearisation' is meant the treatment of causes that disrupt the underlying stochastic process. Results indicate a clear improvement for interval forecasts of mortality. However, the result for point forecasts is not as clear. The documented improvement in interval forecasts can significantly affect the solvency capital requirement, rendering some pension providers at a competitive advantage. It was also confirmed that the transformed-linearised series satisfy better the need for normality as compared to the original series.