Behavioral New Keynesian models : learning vs. cognitive discounting

This paper estimates a New Keynesian model with new and old behavioral elements. Agents in the model exhibit cognitive discounting, or myopia: they discount variables far into the future at higher rates than typically implied in the benchmark model. We investigate the model under different expectational assumptions: rational expectations, subjective expectations with infinite-horizon learning, and subjective expectations with Euler-equation learning. Under rational expectations, the model necessitates of large, possibly unrealistically so, degrees of myopia. The same result persists under infinite-horizon learning, given that agents are still remarkably far-sighted. But, under Euler-equation learning, the model can fit the data with only minimal estimated degrees of myopia. The results indicate that the empirical evidence for cognitive discounting may be sensitive to the modeling of expectations, and they highlight learning as a key behavioral feature to understand macroeconomic fluctuations.