Mortality fitting and forecasting using Hyndman-Ullah model and Lee-Carter model: A study on Malaysia’s ethnic groups

Globally, the life expectancy of the population has shown an increasing trend over the past years due to the improvements in healthcare and socioeconomic conditions. Various countries are experiencing growth in ageing population and expecting to face a few challenges, including an increase in health care utilization, poverty after retirement and longevity risk. In order for the relevant parties to plan and minimize such risks, accurate mortality modelling and forecasting is demanded. This study will focus on evaluating the forecasting performance of Lee-Carter model and Hyndman-Ullah model for mortality rates of Malaysia’s three main ethnic groups, which are Malay, Chinese and Indian under two genders and 18 age groups. Mortality data for years 2001 to 2020 for six subpopulations, i.e. Malay male, Malay female, Chinese male, Chinese female, Indian male and Indian female, are obtained from Department of Statistics Malaysia. The data is then used for fitting and obtain the out-of-sample forecasting using Lee-Carter model and Hyndman-Ullah model. Both models have better forecasting performance for different subpopulations based on the evaluation of goodness-of-fit. Lee-Carter model is selected to forecast the mortality rates for Malay male and Indian male, while Hyndman-Ullah model is used to forecast the mortality rates for the other four subpopulations for the next 20 years (years 2021 to 2040). The forecasted mortality rates of all subpopulations at ages 60 and above are expected to decrease for years 2021 to 2040, resulting in a growth in the ageing population of elderly people in the future. The implication of this population trend is worrying as it may negatively impact Malaysia’s economic growth and productivity as well as put a strain on the public healthcare system and public funds for pensions. At the end of the study, a few suggestions have been given to overcome such challenges.