Improving Lee-Carter forecasting: methodology and some results

This paper is meant to contribute to the research addressing the forecast of longevity. To this aim, we review our idea of correcting the predictive accuracy of the existing mortality projection models, by modelling a measure of their fitting errors as a Cox-Ingersoll-Ross process, originally tested on the CBD (or M5 model). In the context of this paper, we apply such a corrective methodology to the Lee- Carter (LC or M1) model. We carry out the backtesting procedure within a static framework. The resulting model, we call “mLC”, proves itself to outperform the LC model, when forecasted over a fixed prediction span and with respect to the mortality data relating to the Italian females aged 18. The research presented in this paper leaves scope for futher investigations and numerical applications.