Bounded-rationality and heterogeneous agents : long or short forecasters?

Our paper estimates and compares behavioral New-Keynesian DSGE models derived under two alternative ways to introduce heterogeneous expectations. We assume that agents may be either short-sighted or long-horizon forecasters. The difference does not matter when agents have rational expectations, but it does when a fraction of them form beliefs about the future according to some heuristics. Bayesian estimations show that a behavioral model based on short forecasters fits the data better than one based on long forecasters. Long-horizon predictors exhibit very poor predictive ability, whereas the short forecasters' model also outperforms the rational expectation framework. We show that the superiority is due to its ability to capture heterogeneous consumers' expectations. Finally, by Monte-Carlo-filtering mapping, we investigate the indeterminacy regions to complement existing literature.