Consensus and accuracy of Japanese GDP forecasts

Abstract

This article investigates the real gross domestic product forecasts of Japanese institutional forecasters for 25 years. It finds that a consensus forecast does not exist in 9/8 years for the current-year/year-ahead forecasts. The variance of the forecast distribution is positively correlated with the absolute forecast error of its mean forecast, but the correlation is significant for the current-year forecasts only. The economy tends to hit the peak or the bottom when forecast dispersion is large, but nonparametric analysis shows that the correlation is statistically insignificant.