Longevity Risk Modeling with the Consumer Price Index

Abstract

Economic growth has been shown to be an important factor that explains changes in mortality probabilities. Economic growth is commonly measured via the gross national product per capita (GDP), but this article argues that the Consumer Price Index (CPI) is a more natural factor to explain mortality dynamics. It is the CPI that approximates the affordability of health care, food and housing. We augment the well-known Lee-Carter model with the observable CPI factor and test this model using data from the United States, Canada, Australia, and France. We show that the in-sample model fit of our proposed model improves compared with the Lee-Carter model (either augmented with the gross national product factor or not). We also show that the out-of-sample forecasting performance of our proposed model, as measured by the mean squared forecast error, is a considerable improvement. Also, the Lee-Carter model augmented with both the gross national product and CPI factors performs even better in sample and out of sample.