Forecasting Error in the US Social Security Administration’s Economic Assumptions

Abhishek Dev, Bard College

Abstract/Artist's Statement

The accuracy of the US Social Security Administration (SSA) economic and financial forecasts is critical for the solvency projection of the Social Security Trust Funds. The SSA, however, does not provide sufficient information in the public sphere for a systematic replication and assessment of their projections. Thus no external group has been able to produce fully independent evaluation of the economic assumptions even though these forecasts have been made annually since 1941. While there is a quadrennial internal evaluation, we argue the panel’s approach is narrow, inadequate and does not address the forecast track record of the Trustees in their review.

In light of this lack of independent and comprehensive evaluation of the Trustees economic projections, we provide, to our knowledge, the first assessment of the projection performance record of the Trustees. We demonstrate that the long range projections are very uncertain and thus limit our assessment of past projections to short range period of 10 years. We use different measures to evaluate the accuracy and the biasness of the past short range forecasts. Our systematic analysis concludes that even after accounting for the cyclical pattern of forecast error with periodic over- and under-estimation that some economic variables demonstrate, the Trustees have been consistently pessimistic in the projections of these variables until the Great Recession. Moreover, we find that the arbitrarily

set band of alternative projections are too narrow and does not capture most of the projections. Following the results from this error analysis and considering` 1. Historical pattern of the economic variables, 2. Critiques of the Trustee’s forecasting methodology 3. Long range forecasts from other governmental and non-governmental agencies, we recommend set of more reasonable and unbiased values for these economic assumptions. Moreover, based on the results of our error analysis we suggest a larger range for the projection alternatives that better captures most of the short range fluctuations in these economic variable and reflects the inherent uncertainty in the future forecasts.

Based on our results, we recommend the Trustees to limit the short range forecast to 5-year and emphasize the results from these short-range projections more. And either discontinue the long range forecasts or continue them with less emphasis and more reasonable economic assumptions closer to the ones recommended by us.