Comment

The authors build a quantitative framework, broad enough to nest different types of ideas proposed in behavioral macroeconomics, to help make sense of the survey evidence on expectations. The goal is not so much to understand survey evidence for its own sake, though that might be interesting. Nor is it to decide between a class of competing models, all of which work fairly well. Rather, the goal is to inform future directions for research in a field that at the moment seems a bit lost in a “wilderness.” In the end, using their approach, the authors are able to say that theweight of the evidence favors some types of models rather than others. The focus on survey evidence comes at a time of uncertainty about future directions inmacroeconomics.Many, including some of the authors in prior work, have noted serious failings with benchmark models. These failings have a solution, perhaps, within a different class of models, ones that do not impose rational expectations. But rational expectations are a powerful disciplining device, and this broader class seems impossibly large. The literature has, for several years, recognized the potential of survey evidence to itself be a disciplining device, filling some of the role that rational expectations used to, and do still, play. Barberis et al. (2015), for example, argue for using survey evidence to decide between competing models in financial economics. In macroeconomics, Coibion and Gorodnichenko (2015) argue that one need not have a model to interpret survey evidence: a positive sign in regressions of forecast errors on forecast revisions is