What is it about?
When mortality patterns shift abruptly, combining forecasting models can produce more robust life table forecasts than relying on a single method. Life tables and mortality forecasts are important for public health planning, pensions, life insurance, and longevity-linked financial products. But unusual periods such as the COVID-19 pandemic can create structural changes that make mortality harder to forecast accurately with a single model. This paper investigates whether an ensemble of traditional and machine-learning time-series methods can improve forecasts of age-specific mortality rates for groups of countries that share common longevity trends. The study uses Generalized Age-Period-Cohort stochastic mortality models to capture age and period effects, applies K-means clustering to group countries with similar longevity patterns, and then uses ensemble learning to forecast future longevity and annuity price markers. The models were calibrated using data for 14 European countries from 1960 to 2018. The results show that the ensemble approach produced the most robust overall performance, with the lowest RMSE, especially in the presence of structural changes in the shape of the time series around COVID-19.
Featured Image
Photo by Martin Sanchez on Unsplash
Why is it important?
Mortality forecasting matters because it supports decisions in public pensions, population projections, public health, life insurance, and risk management. In periods of disruption, inaccurate mortality forecasts can affect both policy decisions and financial valuations. This study is important because it shows that no single forecasting model is guaranteed to work best in all situations, especially during unusual shocks such as the COVID-19 pandemic. By combining multiple forecasting approaches, the ensemble method can provide more robust predictions when time-series patterns change unexpectedly. The paper is also practically relevant because it links mortality forecasting to future longevity and annuity price markers, which are directly relevant for actuarial and financial applications.
Perspectives
The main lesson from this paper is that mortality forecasting should be treated as an adaptive problem rather than as a search for one universally best model. Pandemic conditions can alter the shape of mortality time series, so flexible forecasting systems are especially valuable. Our approach combines three ideas: stochastic mortality modelling, clustering countries with similar longevity patterns, and ensemble learning. Together, these allow forecasts to use both demographic structure and the predictive strengths of multiple time-series methods. This perspective is especially useful for researchers and practitioners working with cross-country mortality data, life expectancy forecasting, annuity pricing, and risk management under uncertainty.
Prof. Afshin Ashofteh
Universidade Nova de Lisboa
Read the Original
This page is a summary of: Life Table Forecasting in COVID-19 Times: An Ensemble Learning Approach, June 2021, Institute of Electrical & Electronics Engineers (IEEE),
DOI: 10.23919/cisti52073.2021.9476583.
You can read the full text:
Contributors
The following have contributed to this page
