Abstract:
We show that in the formalization of representativeness (Kahneman and Tversky (1972)) developed by Gennaioli and Shleifer (2010), overreaction and confidence are affected by uncertainty, as a news effect interacts with an uncertainty effect. In the time series domain, this interaction emerges in a smooth version of Diagnostic Expectations (DE). Under smooth diagnosticity, agents overreact to new information. Since new information typically changes not just the conditional mean, but also the conditional uncertainty, changes in uncertainty surrounding current and past beliefs affect the severity of the DE distortion and confidence. Smooth DE implies a joint and parsimonious micro-foundation for key properties of survey data: (1) overreaction of conditional mean to news, (2) stronger overreaction for weaker signals and longer forecast horizons, and (3) overconfidence in subjective uncertainty. An analytical RBC model featuring Smooth DE accounts for overreaction and overconfidence in surveys, as well as three salient properties of the business cycle: (1) asymmetry, (2) countercyclical micro volatility, and (3) countercyclical macro volatility.