A Functional Data Approach to the Estimation of Mortality and Life Expectancy at Birth in Developing Countries

The functional data model has recently received increasing attention particularly in their application to mortality forecasting. The advantage of this method over the well-known Lee-Carter model is the ability to treat the underlying process as functional, and provide estimations that are robust to outliers. This research investigates the accuracy of functional data approach in estimating the mortality rates and life expectancy at births in developing countries including Malaysia, Indonesia, Thailand and Singapore. The functional data method was applied to these countries' mortality data, and the out-sample forecast errors showed that, in terms of overall, the functional data model was more accurate than that of the original Lee-Carter model for males and females. The results provide evidence that the functional model is accurate to forecast the life expectancy at births for developing countries.