Tackling Longevity Risk - A new model averaging approach for improved mortality rate forecasting

The accurate forecasting of mortality at retirement ages is essential to pensions and life insurance businesses. This research proposes a new method that offers improved accuracy.

With mortality rates in developed countries declining and populations aging as a result, pension funds and insurance companies face the looming challenge of longevity risk. This is the potential risk attached to the increasing life expectancy of policyholders, which can eventually result in a higher payout ratio than expected.

Concerns about longevity risk have prompted pension funds and insurance companies to investigate the modelling and forecasting of age-specific mortality rates and develop new methods with improved accuracy. Any improvement in the forecast accuracy of mortality rates will be beneficial to a) pension schemes in estimating their future financial obligations; b) insurance companies in calculating their premiums and capital requirements; and c) to governments in determining the allocation of current and future resources at both national and sub-national levels.

Different models have their advantages and disadvantages. Model averaging combines forecasts obtained from a range of models, and it often produces more accurate forecasts than a forecast from a single model.

In the paper, Model confidence sets and forecast combination: an application to age-specific mortality, a new model averaging approach is proposed. This method uses the model confidence set procedure to select a set of statistically superior models and combines the forecasts by assigning equal weights to the set of superior models.

By equally averaging the forecasts from the superior models, the research evaluates and compares point and interval forecast accuracies, as measured by the root mean square forecast error and mean interval score, respectively.

Japanese age-specific mortality rates from 1975 to 2015, obtained from the Japanese Mortality Database (2017), were examined for the purpose of this study. Ages between 60 and 100+ were considered. Japanese mortality rates were split by sex and prefecture. Mortality data at the sub-national (i.e., prefecture) level was also used, as forecasts at the prefecture level are more useful for local policy making and planning.

The study shows that potential gains in forecast accuracy can be achieved by discarding the worse performing models before combining the forecasts equally. By contrast, an existing model averaging method assigning different weights for all models underperforms the proposed method. The research finds that the proposed model averaging method offers a more robust procedure for selecting the forecasting models based on their in-sample performances, in that the model-averaged methods are protected against model misspecification.

The published version of Model confidence sets and forecast combination: an application to age-specific mortality is available for download at City Research Online. The paper has been published in Genus.