Mortality modelling

Picture from Antonio et al., 2017

Back in 2014 I served as a member of the Projection Tables Working Group of the Royal Dutch Actuarial Association. The AG 2014 mortality projection model for the Dutch population is a stochastic mortality forecasting model of type Li-Lee. Using maximum likelihood estimation with Poisson assumption for the number of deaths, this multi-population model is calibrated on a set of European countries with similar socio-economic status. Since the publication in 2014, the Royal Dutch Actuarial Association biyearly publishes an update of the projection model, subject to methodological changes (if deemed necessary).

As principal investigator, I lead a similar project for the Institute of Actuaries in Belgium (IA|BE) with projection models published in 2015 (see IABE 2015) and 2020 (see IABE 2020). This IA|BE mortality projection model delivers a best estimate for future mortality at the level of the Belgian population, but can also be used as a scenario generator for future mortality. The IA|BE 2020 was presented in November 2020, including a first attempt to assess the (possible) impact of COVID-19 on the mortality forecasts, see IABE 2020 Covid impact analysis. Presentation slides are available here and here.

Our work on the impact of COVID-19 on a stochastic multi-population mortality model of type Li & Lee was published in 2022 in Risks. The paper is available in open access via this link.

A GitHub repo with the source code (in R) for our multipopulation mortality models is maintained by Jens Robben, see the MultiMoMo package.

This is joint work with Sander Devriendt and Jens Robben.

Katrien Antonio
Katrien Antonio
professor in actuarial science and insurance analytics

I’m a professor in actuarial science who loves data science, programming and teaching.