Our analysis of the results for each quarterly Survey of Professional Forecasters has emphasized the growth rates of the median forecast for levels. Instead, we could first compute each panelist’s growth forecast and report a consensus projection as the median of the growth rates, but we have always assumed, without formal analysis, that our method produces a better consensus forecast.
In this Research Brief we formally study which method produces a more accurate consensus projection for growth.
Our analysis shows very few statistically significant differences between the two methods for computing consensus growth-rate projections. Even when we find a statistically significant difference between the methods, the measured difference in forecast accuracy is so small that we regard the result as not very meaningful. Our main conclusion is that either method for computing a consensus forecast for growth works reasonably fine.