Financial Analysts Are Overconfident: Quantifying Cognitive Biases in Analyst Earnings Forecasts
Warren Buffett would be proud of this….the following research was conducted at the University of Nebraska (& Iowa)
“We find strong evidence to suggest that analysts place too much weight on their 30 private information, consistent with the model of investor overconfidence”
“past forecast errors influence current forecasts in a manner predicted by the theory of cognitive dissonance.”
Abstract (Via SSRN)
This paper develops a formal model of analyst earnings forecasts that discriminates between rational behavior and that induced by cognitive biases. In the model, analysts are Bayesians who issue sequential forecasts that combine new information with the information contained in past forecasts. The model offers a number of testable implications that allow us to detect cognitive biases, and also to quantify their magnitude. We estimate the model and find strong evidence that analysts are overconfident about the precision of their own information and also subject to cognitive dissonance bias. We examine the influence of the relative amount of private information as a measure of ambiguity on the magnitude of the biases. The variation in overconfidence between the low- and high-ambiguity groups is consistent with the well-established variations documented in the psychological literature. We also demonstrate a relationship between book-to-market ratios and cognitive bias.
Additional Excerpts(Via SSRN)
Our study is aimed at throwing light on several questions. Perhaps most fundamentally, we are interested in providing evidence relevant for assessing the validity of the rational expectations assumption that underpins modern asset pricing theory. Investors are assumed to be able to form rational expectations of relevant economic variables such as dividend or earnings streams and discount rates. If analysts’ earningsforecasts are good proxies for the expectations of investors, then any evidence of divergence from rationality casts doubt on the rational expectations assumption.
More specifically, we present evidence relevant for assessing recent asset pricingmodels that have relaxed the assumption of full rationality in order to provide explanations for certain persistent features of the data that are not adequately explained by the standard theory (Daniel, Hirshleifer and Subrahmanyam, 1998, 2001; Odean, 1998; Barberis, Shleifer and Vishny, 1998). Daniel, Hirshleifer and Subrahmanyam (1998) develop a behavioral model based on the assumption that investors display overconfidence and self-attribution bias with respect to their private information about stock returns. Overconfidence causes them to attach excessive weight to private relative to public information. Self-attribution bias refers to the tendency to attribute success to skill and failure to misfortune and accentuates the effect of overconfidence in the short run. The model generates asset prices that display short-horizon momentum and longhorizon mean reversion. It also predicts that the magnitude of the short-horizon momentum effect will depend upon the severity of the investor’s overconfidence and self-attribution bias.
Our model of security analyst earnings forecasts contrasts rational and biased forecasts. We consider two cognitive biases that are both common and relevant in aneconomic context: overconfidence and cognitive dissonance. An individual who is overconfident overestimates the precision of his private information. Cognitive dissonance can be characterized for our purposes by the proposition that individuals tend to acquire or perceive information to conform with a set of desired beliefs.1 Thus if an analyst issues an optimistic earnings forecast on the basis of favorable private information, she will have a tendency to interpret subsequent information in such a way as to support or conform to the prior belief2. We develop a series of tests that identify and measure both the information content and cognitive biases in earnings forecasts. Notably, both of these measures are independent of asset prices and the momentum effect3. Using data on individual analyst forecasts, we estimate the model and find strong evidence of both overconfidence and cognitive dissonance in analyst earnings forecasts. We then show that the cross- variation in these measures is consistent with theirinterpretation as indicators of cognitive bias.
So What? What Does All This Mean? (Via SSRN)
We find strong evidence to suggest that analysts place too much weight on their 30 private information, consistent with the model of investor overconfidence in Daniel, Hirshleifer and Subrahmanyam (1998). We also show that past forecast errors influence current forecasts in a manner predicted by the theory of cognitive dissonance. These results are perhaps not surprising. Indeed, as Hirshleifer (2001) points out in his review of the subject, experts in many fields systematically suffer from these biases. But it is certainly still a matter of dispute whether the biases are sufficiently large to be of economic significance. Our results suggest that they are. The impact of overconfidence is substantial, in that analysts place twice as much weight on their private information as is justified by rational Bayesian updating.