Structural breaks in mortality trends with an application to Dutch and Belgian data

In this paper we consider the modelling of mortality rates. Many mortality models exist, and for several countries these models are estimated and back tested. We focus on a collection of mortality models (Cairns et al. (2006), Renshaw and Haberman (2006), Plat (2009), OHare and Li (2011)), applied to two age buckets (5-89 and 60-89) of Dutch and Belgian mortality data. The mortality models are estimated and for the models that fit the data best (based on BIC), we investigate the predictive power by performing back tests. In the mortality modelling literature projections are often made using simple time series models like a random walk with drift, of which the projections are highly sensitive to the calibration period. Our main contribution to the literature is that we introduce a modelling strategy for the time-dependent parameters that allows for objective, statistical detection of structural breaks in the time series. By comparing projections based on different calibration periods, we show that the proposed methodology leads to more robust mortality projections.