Content area
Full Text
INTRODUCTION
One of the main objectives of macroeconomic modeling is to understand the joint behavior of aggregate output and inflation at the business-cycle frequency. Rational expectations models with nominal rigidities, workhorses of current macroeconomics, seem to have rather weak internal propagation mechanisms and therefore face substantial difficulties in matching the persistence inherent in output and inflation data. Matching the reactions of output and inflation in response to nominal shocks has proven especially cumbersome [e.g., Chari et al. (2000) and Nelson (1998)].
The aim of this paper is to analyze what role deviations from rational forecasts might play in strengthening the internal propagation mechanisms of economic models and therefore their ability to match the data.
Presented is a simple business-cycle model with monopolistic competition, where prices are preset for one period and agents hold money to satisfy a cash-in-advance constraint. The model deviates from rational expectations by imposing the condition that agents can choose between either of two forecasting models to predict future inflation rates and that they must learn the parameters of the forecast functions.
While one of the available forecast models can correspond to a rational expectations equilibrium once learning is complete, the other available model is "inconsistent" with rational expectations in the sense that it delivers misspecified inflation forecasts for any parameterization.
Although the forecasting restriction itself does not preclude that agents acquire rational expectations (in the limit), I find that agents may learn to use the inconsistent forecast model, giving rise to an equilibrium in which forecasts are only constrained rational. Use of the inconsistent model can be optimal because it induces a law of motion for inflation that causes both available forecast models to be underparameterized. Agents then prefer the inconsistent model whenever it provides a better approximation to the law of motion it generates.
I find that the model's propagation mechanism with constrained rational expectations differs strongly from that under rational expectations. In particular, with constrained rational expectations the model is able to match important features of U.S. output and inflation data using white-noise nominal demand shocks as the unique driving forces. Output and inflation then show persistent deviations from steady state, output deviations tend to be followed by persistent inflation deviations in the same direction, and inflation is...