Robust Predictions for DSGE Models with Incomplete Information

We study the quantitative potential of DSGE models with incomplete information. In contrast to existing literature, we offer predictions that are robust across all possible private information structures that agents may have. Our approach maps DSGE models with information-frictions into a parallel economy where deviations from fullinformation are captured by time-varying wedges. We derive exact conditions that ensure the consistency of these wedges with some information structure. We apply our approach to an otherwise frictionless business cycle model where firms and households have incomplete information. We show how assumptions about information interact with the presence of idiosyncratic shocks to shape the potential for confidence-driven fluctuations. For a realistic calibration, we find that correlated confidence regarding idiosyncratic shocks (aka “sentiment shocks”) can account for up to 51 percent of U.S. business cycle fluctuations. By contrast, confidence about aggregate productivity can account for at most 3 percent