Accuracy of Long Range Actuarial Projections of Health Care Costs

Abstract

The Office of the Actuary, mandated to provide projections of future medical spending for use by the U.S. Medicare and Medicaid programs, publishes forecasts that have been widely used by private firms and government budget officials as a baseline for expected long run premium trends and to estimate liabilities for retiree health benefits. Although these projections have been made publicly available since 1986, they have not yet been subject to systematic evaluation by an external reviewer. This paper develops a method for assessment of both short and long-run accuracy, and applies it to the 17 sets of projections made public over the last 25 years. The more recent set of projections (1998-2010) incorporating lagged macroeconomic effects appear to be more accurate than the older (1986-1995) projections that relied more heavily on demographic cost of illness trends. The average annualized error of the forecasts is approximately 0.5% - 1% per year, whether assessed over a span of one, two or ten years. Projecting “excess” growth in health spending (the rise in the share of wages or GDP) tends to be more accurate than forecasting nominal or real spending per capita.