Empirical Tests of the Rationality of Economic Forecasters: A Fixed Horizons Approach

This paper examines the behavior of 4 major forecasters and the forecast consensus. We employ a new technique of "Fixed Horizon" models. This technique analyzes the sequence of adjustments of a series of forecasts of the same event. We first demonstrate that these forecast adjustment sequences should fluctuate randomly under rationality. We then examine approximately 1200 forecast adjustments over the 1978-1982 period to examine the statistical properties of forecast adjustments. The evidence clearly demonstrates that there are marked and significant elements of statistical rationality for these major forecasters. Information shocks are processed slowly. The pattern of adjustments is consistent with forecasters being adverse to "Inconsistency," i.e., large rapid changes in forecasts. There also may be evidence that forecasters move towards a consensus in hero-like fashion.