Does age or life expectancy better predict health care expenditures?

It is an unresolved issue whether age or (expected) remaining life years better predicts health care expenditures. We first estimate a set of hazard models to predict life expectancy based on individual demographic characteristics and health conditions, and then use regression analyses to compare the predictive power of age and life expectancy in explaining health care expenditures. This paper differs from previous studies in that it uses predicted life expectancy to address the censoring of death; as a result, this paper goes beyond the large health care expenditures at the end of life and the results apply to both deceased and survivors. We find that age has little additional predictive power on health care expenditures after controlling for life expectancy, but the predictive power of life expectancy itself diminishes as health status measures are introduced into the model. These results are not of esoteric interest only for their statistical properties; we show that using life expectancy rather than age results in lower projections of future health care expenditures. This result suggests that increases in longevity might be less costly than models based on the current age profile of spending would predict.