Description:
This paper studies the behaviors of uncertainty through the lens of several popular models of expectation formation. The full-information rational expectations model (FIRE) predicts that both the ex ante uncertainty and the variance of ex post forecast errors are equal to the conditional volatility of shocks to the fundamentals. Incomplete-information models such as Sticky Expectation (SE) and Noisy Information (NI) and non-rational models such as Diagnostic Expectations (DE) predict distinctive rankings of these moments. The paper also shows that uncertainty provides additional parametric restrictions to favor SE over NI as a model of information rigidity, although both predict similar aggregate patterns of forecast errors and disagreement.