FORECASTING LIFE EXPECTANCIES USING AR-BIVARIATE GARCH MODELS.

Abstract

The Lee-Carter methodology has been widely acknowledged for its effectiveness in forecasting mortality rates. In this paper, we examine the behavior of mortality index in Japan, UK and USA. We extend the Lee-Carter (LC) model under a constant conditional variance structure to a bivariate GARCH structure. The LC-bivariate GARCH model is able to capture the volatility dynamics of mortality index by gender and provides a sample time path of conditional correlations between male and female mortality indexes. We find evidence of volatility clustering in the time-varying mortality index and significant conditional correlations in the three developed countries.