Irrational Bias in Inflation Forecasts

This paper investigates the issue of rational expectations using inflation forecasts from the Survey of Professional Forecasters (SPF) and the Green Book. We provide an alternative test of rational expectations hypothesis by measuring the degree of persistence of potential systematic mistakes. The test is obtained by solving a signal extraction problem that distinguishes between systematic and non-systematic forecast errors. The findings indicate highly persistent systematic mistakes, which are driven by the inefficient use of available information, and reject the rational expectations hypothesis. The estimated time-varying bias can be used to improve the SPF and Green Book inflation forecast performance by at least 13.4%. This paper also documents evidence that the real interest rate plays a crucial role in explaining the level of bias that leads to under- and over predictions of actual inflation