Abstract:Many existing mortality models follow the framework of classical factor models, such as the Lee-Carter model and its variants. Latent common factors in factor models are defined as time-related mortality indices (such as $\kappa_t$ in the Lee-Carter model). Factor loadings, which capture the linear relationship between age variables and latent common factors (such as $\beta_x$ in the Lee-Carter model), are assumed to be time-invariant in the classical framework. This assumption is usually too restrictive in reality as mortality datasets typically span a long period of time. Driving forces such as medical improvement of certain diseases, environmental changes and technological progress may significantly influence the relationship of different variables. In this paper, we first develop a factor model with time-varying factor loadings (time-varying factor model) as an extension of the classical factor model for mortality modelling. Two forecasting methods to extrapolate the factor loadings, the local regression method and the naive method, are proposed for the time-varying factor model. From the empirical data analysis, we find that the new model can capture the empirical feature of time-varying factor loadings and improve mortality forecasting over different horizons and countries. Further, we propose a novel approach based on change point analysis to estimate the optimal `boundary' between short-term and long-term forecasting, which is favoured by the local linear regression and naive method, respectively. Additionally, simulation studies are provided to show the performance of the time-varying factor model under various scenarios.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is that the assumption of factor loadings being invariant over time in the existing mortality prediction models is too strict and cannot fully reflect the dynamic changes in the actual data. Specifically, the paper points out that over a long - time span, due to the influence of factors such as the progress of medical technology and environmental changes, the relationship between the mortality rates of different age groups and the underlying common factors may change. Therefore, the paper proposes a time - varying factor model with time - varying factor loadings to more flexibly capture these changes and improve mortality predictions in different countries and for different prediction horizons. The main contributions of the paper include: 1. **Model Extension**: Propose a time - varying factor model as an extension of the classical factor model, allowing factor loadings to change over time, thus better adapting to the dynamic changes in the actual data. 2. **Prediction Methods**: Propose two methods for predicting factor loadings - the local regression method and the naive method, which are suitable for short - term and long - term predictions respectively. 3. **Optimal Boundary Estimation**: Based on change point analysis, propose a new method to estimate the optimal "boundary" between short - term and long - term predictions to optimize prediction performance. 4. **Empirical Analysis**: Conduct empirical analysis through mortality data from multiple countries and verify the superiority of the new model for different prediction horizons. Through these methods, the paper aims to provide a more flexible and accurate mortality prediction framework, especially when dealing with long - term data, being able to better capture and reflect the actual change trends.

Mortality Forecasting using Factor Models: Time-varying or Time-invariant Factor Loadings?

A Forecast-driven Hierarchical Factor Model with Application to Mortality Data

Multipopulation mortality modelling and forecasting: The multivariate functional principal component with time weightings approaches

On Time-Varying Factor Models: Estimation and Testing

A New Fourier Approach under the Lee-Carter Model for Incorporating Time-Varying Age Patterns of Structural Changes

Dynamic modelling and coherent forecasting of mortality rates: a time-varying coefficient spatial-temporal autoregressive approach

Mortality Prediction Based On Chang-Point Detection Of The Principal Component Scores

High-dimensional functional time series forecasting: An application to age-specific mortality rates

A systematic vector autoregressive framework for modeling and forecasting mortality

FORECASTING MULTIPLE FUNCTIONAL TIME SERIES IN A GROUP STRUCTURE: AN APPLICATION TO MORTALITY

Gompertz Law Revisited: Forecasting Mortality with a Multi-factor Exponential Model

DYNAMIC PRINCIPAL COMPONENT REGRESSION: APPLICATION TO AGE-SPECIFIC MORTALITY FORECASTING

Mortality Forecasting for China Based on the Long-run Equilibrium Relations

An Innovation Mortality Prediction Model with Cohort Effect

A Mortality Model for Multi-populations: A Semi-Parametric Approach

Age-coherent Extensions of the Lee–Carter Model

Functional Prediction Model for Mortality

Forecasting High-Dimensional Functional Time Series: Application to Sub-National Age-Specific Mortality

Forecasting Leading Death Causes in Australia using Extended CreditRisk$+$

Distinguishing Time-Varying Factor Models

Probabilistic mortality forecasting with varying age-specific survival improvements