Disagreement as a Measure of Uncertainty

This paper presents evidence from the Livingston survey of inflation forecasts that forecaster disagreement provides a useful measure of forecast uncertainty. The evidence is analogous to the evidence for ARCH effects. Disagreement at the time of the forecast has a large positive effect on the conditional variance of the subsequent forecast error. As a conditioning variable, forecaster disagreement dominates ARCH for both survey errors and the error terms in Engle's quarterly model of inflation. As measured by the resulting conditional variances, disagreement indicates larger and more variable levels of uncertainty for the 1946-94 period