SURVEY vs ARCH MEASURES OF INFLATION UNCERTAINTY.

This article examines the performance of conditional heteroscedasticity model of inflation uncertainty in the U.S. It has often been argued that high inflation causes high inflation uncertainty, and that uncertainty in turn causes inefficiencies in the real economy. This argument involves conjectures about the causes and consequences of inflation uncertainty which can only be substantiated by empirical analysis. For such analysis to be convincing, inflation uncertainty must be measured in a convincing way. This paper examines the performance of one type of uncertainty measure which has become popular in recent years, the conditional heteroscedasticity model developed by Engle. This measure is indirect in the sense that it relies on some maintained hypothesis about how uncertainty is generated. For example, in the autoregressive conditional heteroscedasticity model, the variance surrounding a forecast is assumed to be related to the size of past forecast errors.