Comparison of the lee-carter and arch in modelling and forecasting mortality in Zimbabwe.

Abstract

This paper aim to determine which model best fit the mortality profile of Zimbabwe. The model that were being compared were the LEE-CARTER and ARCH and the period under review was from 1983-2004 using mortality rates of five years age interval. The parameter of the LEE-CARTER model were estimated using singular value decomposition. The box Jenkins approach was used to determine the order of lags for the ARIMA. An ARIMA (1, 1, 1) was used to forecast the overall mortality levels. To determine which model best fit the mortality profile of Zimbabwe goodness of fit using the Kolmogorov-Simonov and root mean square error was used. The result suggested that the LEE-CARTER model provided a better fit as compared to ARCH. However both models failed to fit for ages between 50-59 years. Both models were used to forecast but selecting the one that had the best fit of the two model on a particular age group interval for the period 2005-2014.This paper concludes that generally the LEE-CARTER model provided a better fit for mortality profile of Zimbabwe.