A quantitative analysis of mortality index linearity in Lee-Carter model and its implications : observations from 37 countries

Abstract

Lee-Carter’s proposed model established the foundation of fitting and forecasting mortality. Howver, several studies, which lead to the improvement of Lee-Carter’s proposed model, questioned the validitty of the method’s assumptions. An unresolved doubt in their analysis is the assumption of an almost linear kt series. This uncertainty has formed the basis of our investigation. Our research seeks to test this assumption on a wider span of population; 37 countries in order to gain a broader perspective. We first seek explain the reasons behind the non-linearity and the appearance of kinks in the kt series. To further distinguish the kinks in the series, we have classified data into different intervaled fitting periods and identified its position. We will then delve further into analysing the kinks in the series by excluding the kink in forecasting the mortality index and life expectancy of each country. Our analysis reveals that in general excluding the kinks in kt series improves the linearity however, may not improve the forecasting ability.