Looking over many variables, alternative sample periods, different measures of realizations, and five quarterly forecast horizons, we generally find statistically insignificant differences between mean and median forecast accuracy. Of particular interest: Our results suggest generally no differences between the mean and median forecasts for the unemployment rate, real GDP, and inflation. When we find statistical significance in the other variables, the difference in forecast accuracy is often small, indicating that we may continue using the median forecast as the consensus forecast.
Battle of the Forecasts: Mean vs. Median as the Survey of Professional Forecasters’ Consensus
We use the median forecast, because we've always assumed that the mean forecast is vulnerable to extreme individual responses. For the first time, we formally test our assumption in a statistically rigorous and comprehensive manner.