Discussion

Xavier Gabaix welcomed the broader agenda to which the paper contributes and that recognizes a larger role for behavioral features in macroeconomic models. Gabaix was curious about the ability of Woodford’s model to obtain Fisher neutrality in the long run while resolving the “forward guidance puzzle.”He followed up on a point raised by Jennifer La’O during her discussion and noted that several recent papers pursue an exercise similar to Woodford’s but consider different behavioral frictions. Gabaix asked how the predictions of the model at the aggregate level differed from the ones in these papers and suggested that these alternative models should be judged in light of these predictions. Finally, he offered a broader comment about the role of bounded rationality for decisions at the micro and macro levels. According to estimates of his, the discount factor is roughly 0.85 per quarter in a macro context, suggesting a planning horizon of 2 years. Gabaix concluded that agents can be very patient at the micro level, when making professional and familial decisions, but may have a more limited attention to macro disturbances. The author pointed out that there is indeed a formal similarity between the reduced-form equations he obtained and those associated with other behavioral frictions, including sparsity in the case of Gabaix (“A Behavioral New Keynesian Model” [Working Paper, Harvard University, 2018]). He emphasized that his model differs from other contributions, not only in terms of the foundations but also in terms of predictions at the aggregate level. An important difference, he argued, is that his model resolves the issue of multiplicity of equilibria. Another difference he mentioned is that his model allows for learning dynamics. The learning component has crucial implications for long-lasting policy experiments, such as a permanent interest rate peg. On this point, the author clarified that adaptive learning of the value functions is the very mechanism that allows one to obtain Fisher neutrality in the long run,