Mortality Modeling Using Covariates with Ghana Census Data

Abstract

Over the years, actuaries as well as demographers have devoted much time to the construction of mortality models that fit the age pattern of mortality in many countries. However, the emphasis has been on data-rich developed countries with little attention to the special characteristics and challenges associated with rapidly growing populations in developing countries. These include a relative lack of mortality data as well as the need to account for other factors, such as the availability of health care, reporting rate, and type of religion in the area (e.g., Christian or Muslim). In this work, the authors construct a mortality model that directly addresses cases with limited mortality histories and factors that affect mortality reporting. Using a covariate (v) to represent the effect of additional factor permit the authors to compare different mortality categories such as gender and region. When this model is fitted to the 2010 population and housing census mortality data from Ghana, a key finding is that male and female mortality rates for ages above 50 years diverge, indicating that male mortality above the age of 50 years is higher than female mortality. The authors also use the model to show that the Greater Accra region has lower overall mortality rates than the Ashanti region. Finally, by borrowing mortality data from South Africa the authors were able to introduce a time dimension to forecast mortality changes in Ghana during a period of six years. Using this kind of approach will allow researchers and policymakers to get a better picture of developing countries’ demographics and to improve the design of health, retirement, business, and other programs calibrated to each country’s evolving population characteristics. TOPICS: Retirement, statistical methods