Are Expectations Misled by Chance? Quasi-Experimental Evidence from Financial Analysts

We examine whether finance professionals deviate from Bayes’ theorem on the processing of nondiagnostic information when forecasting quarterly earnings. Using field data from sell-side financial analysts and employing a regression discontinuity design, we find that analysts whose forecasts have barely been met become increasingly optimistic relative to when their forecasts have barely been missed. This result is consistent with an update of analysts’ expectations after observing uninformative performance signals. Our results also suggest that this behavior leads to significantly worse forecasting accuracy in the subsequent quarter. We contribute to the literature by providing important field evidence of expectation formation under uninformative signals.