Misspecification versus bubbles in the Cagan hyperinflation model
Steven N. Durlaufand Mark A. Hooker*
Most test for rational bubbles using Cagan's hyperinflation model under rational expectations assume that the model is correctly specified. Rejections of the model are consequently treated as evidence of bubbles. This paper shows that it is possible to test model specification separately from the presence of bubbles. We provide a general framework for identifying all testable implications of the null hypothesis that the model is correctly specified as well as all implications of the null hypothesis that a given price sequence is a bubble-free realisation of the model. We find that in the German hyperinflation, empirical deviations from the fundamental price solution are primarily due to misspecification rather than bubbles, conditional on the same identification restrictions used in the parametric bubble-testing literature. Our analysis potentially reconciles some contradictory results in the literature.
The Cagan hyperinflation model has provided a rich ground for the analysis of expectations-based models in macroeconomics. The reasons for interest in hyperinflations in general and the Cagan model in particular are twofold. First, hyperinflations represent periods where the role of expectations seems particularly crucial to the formation of endogenous variables. The explosive behaviour of the money supply in these episodes makes it almost certain that other observable____________________