A Multivariate Approach to Project Common Trends in Mortality Indices

Abstract

The improvements in mortality rates have been under investigation in many studies across the 20th century, especially in developed countries. Current literature assumes that mortality indice can be forecasted independently through the model ARIMA (Autoregressive Integrated Moving Average). However, in this paper, we apply and compare the ARIMA and the Vector Error Correction (VECM) models for mortality indices by provinces across Canada. We clearly show through the Johansen cointegration test that there is a long run equilibrium relationship between provincial mortality indices in Canada during the period from 1921 to 2009. Thus, we demonstrate that Canadian mortality indices cannot be forecasted independently, as it is stated in the current literature, but into a Vector Error Correction Model that shows a multivariate dynamic relationship between mortality indices and its improvements. Findings from this framework could help local insurers to build new financial products for hedging mortality risk purposes and also for the pricing of annuity contracts.