A unified approach to mortality modelling using state-space framework: characterisation, identification, estimation and forecasting

TL;DR

This paper introduces a comprehensive state-space framework for mortality modeling, integrating heteroscedasticity and stochastic volatility, and demonstrates its effectiveness using Danish mortality data.

Contribution

It develops a unified state-space approach for mortality models, proposes new identification constraints, and incorporates Bayesian methods with stochastic volatility for improved long-term mortality forecasting.

Findings

Incorporating heteroscedasticity improves model fit.

Stochastic volatility enhances long-term mortality forecasts.

The framework effectively captures period and cohort effects.

Abstract

This paper explores and develops alternative statistical representations and estimation approaches for dynamic mortality models. The framework we adopt is to reinterpret popular mortality models such as the Lee-Carter class of models in a general state-space modelling methodology, which allows modelling, estimation and forecasting of mortality under a unified framework. Furthermore, we propose an alternative class of model identification constraints which is more suited to statistical inference in filtering and parameter estimation settings based on maximization of the marginalized likelihood or in Bayesian inference. We then develop a novel class of Bayesian state-space models which incorporate apriori beliefs about the mortality model characteristics as well as for more flexible and appropriate assumptions relating to heteroscedasticity that present in observed mortality data. We show that multiple period and cohort effect can be cast under a state-space structure. To study long term mortality dynamics, we introduce stochastic volatility to the period effect. The estimation of the resulting stochastic volatility model of mortality is performed using a recent class of Monte Carlo procedure specifically designed for state and parameter estimation in Bayesian state-space models, known as the class of particle Markov chain Monte Carlo methods. We illustrate the framework we have developed using Danish male mortality data, and show that incorporating heteroscedasticity and stochastic volatility markedly improves model fit despite an increase of model complexity. Forecasting properties of the enhanced models are examined with long term and short term calibration periods on the reconstruction of life tables.