An Extension of the Mortality Model Using State Space Model

Mortality models such as the Lee-Carter model are widely used to estimate future mortality. These mortality models assume that actual mortality rates are represented by parameters that depend on age and year of observation. For mortality models, simple models have problems in fitting to actual values, while complex models may lead to overfitting.

In this paper, we assume a time series model whose mean is a conventional mortality model and propose a state-space model in which the time series model is the state model. A simple AR(1) model is used as the time series model, and a method for estimating the parameters of the mortality and AR(1) models is formulated. Numerical experiments were conducted to compare the proposed method with existing methods for fitting to actual data and predicting future mortality rates, and the results showed that the proposed method improved on existing methods.

Find the Q&A here: Q&A on 'Mortality Models from Different Perspectives'