Three essays on Malaysian population ageing

Malaysia's ageing population is inevitable following the increase in life expectancy at birth and decrease in fertility rates over the last four decades (Mahari 2011). As a result of these demographic changes Malaysia is expected to become an aged nation by 2030 - when the proportion of elderly people reaches 15% of the total population (Goh & Lai 2013). The elderly are less healthy than the young, hence an increase in the percentage of the elderly population will result in a greater demand for long term care services (Mafauzy 2000). Currently, the public long term care system in Malaysia is not up to such a challenge, given that policy related to long term care is not developed and the entitlements for public long term care programs are limited to destitute elderly people (Ong 2002). This issue has been extensively debated by researchers and there is a growing sense of urgency for public long term care policy action (Mafauzy 2000, Ong 2007, Dahlanet al. 2010, Li & Khan 2012). Nevertheless, none of the researchers have estimated the change in the Malaysian population structure, and the likely demand for long term care in the future. Such detail is valuable for government and would be useful in policy reform. This thesis aims to contribute to filling this gap. Three research papers have been prepared, each seeking in detail the changes in the Malaysian population and its population components (mortality, fertility and net international migration) over time. The last paper also examines how changes in mortality and disability affect long term care demand and costs in the future. The first paper examines mortality forecasting methods for the Malaysian population.The comparison of mortality forecasting models is vital for identifying which of the models could provide more accurate forecasts. While the literature includes comparisons between independent and independent models (Lee &Miller 2001, Booth et al. 2005, 2006, Shang et al. 2011) and independent and coherent models (Hyndman et al. 2013), no comparisons between coherent models have been made. In this paper we include both the comparison between coherent models, and coherent models with independent models. The out-of-sample forecast errors of sub-populations' age-specific mortality rates and life expectancy at birth are evaluated and compared across five different projection methods,including coherent models - Poisson common factor (Li 2013) and product-ratio functional (Hyndman et al. 2013) - and their respective independent models - Poisson Lee & Carter (Brouhns et al. 2002) and functional time series (Hyndman & Ullah 2007) - together with the original Lee & Carter (1992) model. We extend the applications of coherent forecasting models using data from a less developed country, Malaysia, where mortality trends are different. In addition, we include two types of coherency for Malaysian sub-populations, including gender as well as ethnicity. The second paper forecasts each of the Malaysian population components - mortality, fertility and net international migration - in projecting the Malaysian age-specific sub-populations by gender and ethnicity. The stochastic cohort component projection framework from Hyndman & Booth (2008) is adopted for this purpose. Each of the population components are forecast separately and jointly between sub-populations using the functional independent model from Hyndman & Ullah (2007) and the product ratio coherent model from Hyndman et al. (2013), respectively. For the latter, in addition to gender coherence, we explored another possible type of coherency - that of developed-developing countries. The forecast of population components are then combined using the cohort component method to obtain a variety of Malaysian population forecast series. Out-of-sample sub-population forecasts are evaluated to determine the best model for long-term Malaysian population projections by age and gender. The final paper studies the impact of population change on the Malaysia's government provided long term care programs. The number of people requiring long term care in Malaysia is projected up to the year 2040 using the multiple state method from Rickayzen & Walsh (2002). The method has been applied in developed nations (Leung 2004, Hariyanto et al. 2013) and a developing country (Bueno 2013). Rickayzen & Walsh's 2002 methods are extended in this paper to allow the estimation of mortality improvement based on Malaysian historical time-series data. The future long term care costs are also estimated, providing valuable information to policy makers if the current policy continues unchanged. In addition, the cost effectiveness of current programs is estimated and compared with an Asian developed nation, Japan. We make use of WHS (2002) survey data to estimate the health utility index for elderly living in public institutional care.