An Evaluation of Recent Macroeconomic Forecast Errors

D espite a significant decline in the pace of economic growth in the second half of 2000, macroeconomic forecasters underpredicted real GDP growth and overpredicted the unemployment rate by a significant amount, for the fifth consecutive year. On average, real GDP forecasts were about 2 percentage points below the actual data for the 1996-2000 period, and unemployment rate forecasts about 0.5 percentage point above. On a more positive note, forecasters ended their chronic overprediction of inflation during much of this period. Nevertheless, surprisingly large and persistent errors in recent forecasts of GDP, inflation, and unemployment have perplexed macroeconomists and policymakers for quite some time, and they merit closer examination. 1 To begin, we ask whether the large, persistent, and one-sided errors observed recently violate the principal objectives of economic forecasting in a statistically significant way. And if they do, what caused these significant errors and are they likely to continue? Such violations undercut the credibility of forecasting models and complicate the already difficult task of setting appropriate monetary policy. Efforts to discern the causes of recent errors and to improve macroeconomic forecasting models may help improve the conduct of policy. This article evaluates recent forecast errors made by private forecast-ers in an attempt to understand why forecasts have gone so far awry. The investigation centers on errors in forecasts of real GDP growth, inflation, the unemployment rate, and nominal and real short-term interest rates since 1969. The focus is on one-year-ahead forecasts because lags in the effects of monetary policy require the Federal Reserve to forecast economic activity well in advance when setting its current interest rate target. Average forecasts, which tend to yield smaller errors, and individual forecasts are examined.