FITTING AND FORECASTING MORTALITY TREND IN NIGERIA: AN APPLICATION OF THE LEE-CARTER MODEL

Lukman Abolaji  AJIJOLA; Olamide PATRICK; Prosper Toochukwu CHUKWUEMEKA

Lukman Abolaji AJIJOLA University of Lagos
Olamide PATRICK University of Lagos
Prosper Toochukwu CHUKWUEMEKA

Keywords: Lee-Carter, Prediction, Fitting, Mortality, ARIMA, SARIMAX, GBR

Abstract

ABSTRACT

In a multi-ethnic and diverse country like Nigeria, which has inherent health system challenges, there is a need for timely and accurate mortality forecasts for effective resource allocation, and policy and social program design. This research, therefore, applies the Lee-Carter model to Nigerian mortality data which is a popular mortality prediction technique. The Model is modified due to Nigeria's structure comprising both rural and urban areas with different age geographic mortality rates. The research population consists of historical and projected mortality data for Nigeria from 12/31/1950 to 12/31/2100, obtained from the United Nations (UN) data bank. The final dataset was split into two subsets: the training set (from 1952 to 2023) and the validation set (from 2024 to 2100). This split allowed for accurate performance evaluation of the models on unseen data. The study predicts mortality trends in Nigeria using the Lee-Carter model and its variations, such as ARIMA, SARIMAX, and Gradient Boosting Regressor (GBR). The Lee-Carter Model with Gradient Boosting Regressor (GBR) model significantly outperformed the other two models with an exceptionally low MSE of 4.25e-06.