Advanced Techniques in Mortality Trend Estimation: Integrating Generalized Additive Models and Machine Learning to Evaluate the COVID-19 Impact

Abstract

The last two centuries have seen a significant increase in life expectancy. Although past trends suggest that mortality will continue to decline in the future, uncertainty and instability about the development is greatly increased due to the ongoing COVID-19 pandemic. It is therefore of essential interest, particularly to annuity and life insurers, to predict the mortality of their members or policyholders with reliable accuracy. The goal of this study is to improve the state-of-the-art stochastic mortality models using machine learning techniques and generalize them to a multi-population model. Detailed cross-country results conducted for Finland, Germany, Italy, the Netherlands, and the United States show that the best forecasting performance is achieved by a generalized additive model that uses the framework of APC analysis. Based on this finding, trend forecasts of mortality rates as a measure of longevity are fulfilled for the future, given a range of COVID-19 scenarios, from mild to severe. Discussing and evaluating the plausibility of these scenarios, this study is useful for preparation, planning, and informed decision-making.