Managing uncertainty: The bias and efficiency of federal macroeconomic forecasts

Abstract

Several federal agencies generate forecasts of future economic performance. How accurate are these projections? The article compares actual and projected outcomes for four macroeconomic indicators across three federal agencies. The bias and efficiency of the forecasts are estimated using tests drawn from the macroeconomic literature on private forecast performance, and some alternative explanations for observed bias are considered. The tests indicate that many of the short-term forecasts are unbiased while long-term forecasts are severely biased. The absence of short-term bias is somewhat surprising, given that private sector forecasts of the same variables can be biased and that we expect that government forecasters may be subjected to pressures to distort forecasts. The presence of systematic bias in the long-run forecast confirms that public sector forecasters face incentives to deviate from squared-error minimization strategies, especially if forecasts have long time horizons.