Heterogenous Forecasting and Federal Reserve Information

December 2001 We provide evidence that private forecasters and the staff of the Federal Reserve use different forecasting models to predict inflation and GNP growth and heterogeneity of forecasting models is the norm in the market place. We thus argue that neither the Fed nor commercial forecasters know the "true" model of the economy. We demonstrate that their forecast errors are correlated with information available at the time when the forecasts were made and hence, by studying the systematic patterns of these forecast errors, we can deduce a great deal about their assessments of economic conditions in general and their view of monetary policy in particular. We also show that (i) although all private forecasters have the same information, their diverse forecasting models result in a distribution of forecasts which fluctuates over time. The distribution shows no tendency for convergence. The "consensus" median forecaster is a random member whose identity changes over time; (ii) the evidence shows that at any date there is a whole set of private forecasts who agree with the Fed's forecasts but the Fed forecasts are less volatile than the volatility of private forecasts measured by the variance of the cross-sectional distribution of private forecasts; (iv) diverse assessments of the impact of monetary policy on inflation and growth are important factors which contribute to the heterogeneity of forecasts. A surprising result reveals that although there is strong evidence for heterogeneity among forecasters, we also find similarity in qualitative patterns of forecast errors of all participants, including the Fed. Qualitative similarity implies similarity in the basic ideas underlying the forecasting models even when these ideas are wrong. This implies a degree of correlation among the subjective beliefs of divergent agents in the economy. (v) although all forecasts violate the standard orthogonality conditions of Rational Expectations, we argue in this paper that this should not be interpreted as irrational behavior. We provide an introduction to the theory of Rational Belief (see Kurz (1994), (1997)) and demonstrate that in contrast to the rejection of Rational Expectations, the pattern of estimated parameters is consistent with the predictions of the theory of Rational Beliefs (see Kurz (1994), (1997)). In opposition to some (e.g. De Bondt and Thaler (1985), (1990)), we argue that rejection of rational expectations should not be interpreted to imply irrational behavior. Working Papers Index