Comparing Stochastic Models for Forecasting Mortality in Malaysia: Lee-Carter, Booth-Maindonald-Smith, and Age-Period-Cohort

Abstract

Problem:  The Human Life Table Database (HLD) provided the death rate data utilised in this study. With an emphasis on gender disparities and applying data from 2001 to 2020, this study analyses the problem of selecting the most accurate stochastic model for forecasting mortality rates in Malaysia.

Aims/Objectives: This study compares the Lee-Carter, Booth-Maindonald-Smith, and Age-Period-Cohort models to determine which stochastic model is more effective in predicting gender-specific death rates in Malaysia between 2021 and 2026.

Methodology/approach: The Human Life Table Database (HLD) included mortality rates for male and female participants aged 0 to 80. The data were divided into 5-year age intervals from 2001 to 2020. The Lee-Carter model, the Booth-Maindonald-Smith model, and the Age-Period-Cohort variant of the Lee-Carter model are the three stochastic models that the study applies to this dataset. The residuals and error data were analysed to determine how well each model predicted mortality rates. The model producing the lowest error values was determined to be the most suited model for forecasting mortality trends. Comparisons were conducted across the models to assess the accuracy and robustness of the models in representing the mortality dynamics for both genders in Malaysia.

Results/finding:  The Lee-Carter model was identified as the most accurate stochastic model for forecasting mortality rates in Malaysia from 2001 to 2020, as it produced the lowest error values compared to the other models.