Projected Malaysian Lifetable: Evaluations of The Lee-Carter and Poisson Log-Bilinear Models

There is a continuous improvement in mortality rates globally including in Malaysia. Due to systematic underestimation of mortality rates, many researchers have shown their interest towards mortality study as inaccurate mortality estimates may impact policies and interventions related to health, life insurance and pensions. In this study, two mortality forecasting models were adopted including the Lee-Carter model and its variant, the Poisson log-bilinear regression model. The out-sample forecasts of mortality rates and life expectancy at births from both models were evaluated using Mean Forecast Errors (MFEs) and Mean Absolute Percentage Forecast Errors (MAPFEs). Consequently, using the most accurate model, we forecast future mortality rates and construct Malaysian life tables for males and females. Data of Malaysian mortality rates by age and gender from year 1970 to 2017 were collected from the Department of Statistic Malaysia (DoSM). Results show that the estimation of parameters from both models are approximately similar in trends. The Lee-Carter model provides more accurate out-sample mortality rate and life expectancy at birth forecasts for Malaysian males and females compared to the Poisson log-bilinear model