Forecasting mortality: when academia meets practice

Abstract

Techniques used in practice often differ from tools developed in academia. The lack of communication that may exist between academia and practice can then have important consequences for many insurance companies or pension funds. This issue is illustrated with what is currently happening in Switzerland. Swiss pension funds use mortality tables that are regularly updated with new observations. A new version of these tables has been recently published and includes a procedure to forecast mortality until 2150. The method applied for these projections is very different from the several forecasting models that have been developed in academia over the last decades. In this paper, we compare mortality forecasts used by practitioners in Switzerland and the forecasts resulting from two simple approaches well-known in academia, the Lee–Carter model and the Heligman–Pollard function. These two approaches have the advantage of simplicity and thus, all insurance companies and pension funds may implement them without any difficulties. The analysis demonstrates that both academic methods forecast a more important decrease in mortality than the approach applied by pension funds, especially in the long-run and for females. Impacts on pension liabilities are then evaluated, enlightening the future challenges many institutions will face. Finally, a few points which insurance companies or pension funds need to be cautious with, when using mortality forecasts, are summarized.