Model fitting for Malaysian mortality rate: Comparison of Heligman-Pollard and P-splines smoothing

Abstract

Abstract Malaysia has been experiencing longevity risk since the last decade due to improvements of mortality rates. Longevity risk refers to the probability of a person living longer than expected. According to the Department of Statistics Malaysia, a baby born in the year 2018 is predicted to live an average life of 75 years. Since the minimum retirement age policy of 60 years had come into force in 2012, the 2018 baby would live approximately 15 more years after retirement. Therefore, this study aims to compare the Heligman-Pollard and P-splines smoothing for fitting the Malaysian mortality rate. This model fitting will give a clear picture of the mortality pattern in predicting the mortality rate accurately, especially for the baby boomer generation. The data obtained from the Department of Statistics Malaysia are split into groups of five years, from 0 to 75 years old, and time ranges from 1995 to 2018. The data set from 1995 to 2010, known as the training set is used to fit the mortality rate. After fitting the mortality rate for both methods, this study will measure the performance in the testing set from 2011 until 2018. This study uses the mean absolute percentage error (MAPE) to identify the better method to fit the Malaysian mortality rate. Based on the MAPE values, P-splines smoothing gives a relatively smaller value compared to the Heligman-Pollard. For overall performance from 1995 to 2018, P-spline smoothing has proven to fit the Malaysian data well.