Abstract:
This paper proposes criteria for classifying time-series forecasts of inflation as weakly, sufficiently, strongly, and strictly rational. Forecasts taken from the ASA-NBER surveys, some well-known one-step-ahead forecasting techniques, and a novel variable length autoregressive moving average model are tested against these criteria. None of the forecasts series meets the criteria for strict rationality nor, even, the less demanding criteria for strong rationality. While agents forecast as best they can, their forecasts are not likely to meet stringent rationality criteria suggested by econometricians. Copyright 1991 by MIT Press.