Fundamental Persistence and Diagnostic Expectations

Diagnostic expectation, a behavioral framework in which agents overweight recent news based on its representativeness, typically assumes that investors overreact to information. However, this literature largely overlooks the possibility of investor underreaction. This paper generalizes the diagnostic expectations model to incorporate both over- and underreaction, demonstrating that standard empirical tests of news processing can be misleading. By building a state-space model of firm earnings, we derive closed-form expressions that link past growth, forecast errors, and stock returns. We highlight three key theoretical results. First, high fundamental persistence with overreaction is observationally equivalent to low persistence with underreaction in standard Coibion-Gorodnichenko regressions. Second, inferring the true direction of investor reaction requires joint identification, as key covariances change sign depending on the interaction between persistence and diagnostic distortion. Third, pricing implications from the equity term structure model show that return predictability mirrors forecast error dynamics identically across both regimes. Empirical evidence from the Korean stock market, an environment exhibiting low persistence and structural underreaction, supports these findings.