THE COMPARISON OF RANDOM WALK WITH DRIFT AND SES IN FORECASTING INDONESIAN MORTALITY RATE WITH LEE-CARTER MODEL

Abstract

Nowadays, the demand of the mortality rate data in the future is needed by the insurance companies to determine the premium which has been paid by the insured. To do so, the insurance companies require a mathematical model which is able to represent problems in forecasting mortality rate. One of the models which has been acknowledged by the Society of Actuaries of Indonesia in forecasting mortality rate is Lee-Carter (LC) model. This study used LC model to forecast the Indonesian mortality rate based on total population. The mortality rate data from 1950 to 2015 is considered. The first stage is to estimate all parameters in the LC model using least square method and singular value decomposition method. The second stage is to project the value of the time-dependent parameter in LC model by considering two methods: (1) Random walk with drift and (2) SES. Afterwards the result of parameter projection from these two methods is compared by considering the Mean Absolute Error (MAE). The final result shows that the random walk with drift is more accurate than the SES in projecting the time-dependent parameter with MAE of 0.168730974. The projected parameter obtained by random walk with drift is then used to calculate the Indonesian mortality rate for one period ahead.