Forecasting Mortality Rates in Malaysia using ARIMA and Lee-Carter Models

Abstract

Forecasting mortality rates is crucial for understanding demographic trends and describing Malaysian public health planning, such as optimising resource allocation and guiding evidence-based policy development to meet future healthcare demands. It is also an important technique that provides professionals with valuable insights, enabling them to predict trends and remain vigilant. The objectives in this study is applying ARIMA and Lee-Carter models to mortality rate in Malaysia, then comparing the accuracy between ARIMA and Lee-Carter models to identify the most reliable forecasting approach by analysing RMSE, MAE and MAPE and lastly is using the most suitable model to forecast mortality rate of Malaysia. The goal of the study is to determine the best model to forecast the mortality rate in Malaysia between the ARIMA model and the Lee-Carter model. This forecasted mortality could help researchers and policymakers understand the trend of mortality rates and enhance health initiatives in Malaysia.  The methodology involves using ARIMA model by using Box-cox transformation and differencing for stabilising data meanwhile Lee-Carter model uses Singular Value Decomposition to capture trend of future rates. The mortality rate data in Malaysia for this study was obtained from the United Nations World Population Prospects and this study compares the ARIMA and Lee-Carter models using mortality data from 1950 to 2018, categorized into seven age groups which is infant, toddler, childhood, teenager, young adult, adult and elderly. In result, ARIMA (1, 1, 0) performs better than Lee-Carter model for most of the age groups, especially for the younger age groups.