When an agent receives information from a source whose accuracy might be either high or low, standard theory dictates that she update as if the source has medium accuracy. In a lab experiment, subjects deviate from this benchmark by reacting less to uncertain sources, especially when the sources release good news. This pattern is validated using observational data on stock price reactions to analyst earnings forecasts, where analysts with no forecast records are classified as uncertain sources. A theory of belief updating where agents are insensitive and averse to uncertainty in information accuracy can explain these results.
When allocating scarce resources, decision makers must determine what information about potential recipients to include in their allocation criteria. In a series of experiments, we find that participants making student admission and bonus allocation decisions frequently exclude information that contains biases or noise. This exclusion reflects a tradeoff between fairness concerns and the usefulness of information. These fairness concerns are primarily focused on the allocation procedure rather than the resulting outcomes.
Evidence from the laboratory and the field has uncovered both underreaction and overreaction to new information. We provide new experimental evidence on the underlying mechanisms of under- and overreaction by comparing how people make inferences \textit{and} revise forecasts in the same information environment. Participants underreact to signals when inferring about underlying states, but overreact to the same signals when revising forecasts about future outcomes---a phenomenon we term "the inference-forecast gap." We show that this gap is largely driven by different simplifying heuristics used in the two tasks. Additional treatments suggest that the choice of heuristics is affected by the similarity between statistics in the information environment and the statistic elicited by the belief-updating problem.
Revision requested by American Economic Journal: Microeconomics
Current version: July 2023
Acquiring information about available options before making a decision is useful because it allows decision makers to switch to a superior alternative if the default option is deemed inferior. Therefore, information demand should depend on the distribution of the options’ values. In an experiment, I show that information demand increases as the default worsens, while, on average, it remains insensitive to the prior value of the alternative. These patterns reflect bounded rationality in information valuation, which stems from the difficulty of foreseeing future choices and integrating their payoffs.
This study explores how the historical price path of a stock shapes investors’ perceptions of risk. In a series of experiments, we present participants with real and fabricated stock price charts and elicit their risk perceptions. We document that a parsimonious set of three conceptually independent features, i.e., recency, clustering, and sign, can explain large proportions of the variation in risk perceptions. We find that the effects of these three features are partially mediated by their impact on perceived volatility, which is not fully explained by Bayesian inference. In addition, a significant portion of the features’ effects on risk perceptions occurs independently of perceived volatility. Using U.S. real stock and mutual fund data, we show that these three features help explain cross-sectional variations in returns, trading volume, and future volatility of individual stocks, and predict mutual fund negative flows.
New workers often compare themselves to their high-achieving senior coworkers, but they often do so without knowing how senior workers performed in the early stages of their careers. This upward social comparison under incomplete information can have adverse effects on new workers’ well-being and employee turnover. We study whether providing performance trajectory information to new workers mitigates the negative consequences of performance comparison. In a large-scale randomized control trial at a leading multinational spa chain in China, we sent workers twice-weekly messages on the performance trajectories of their high-performing senior coworkers. This information treatment reduces the attrition rate of new workers by 12%, and the effect is most pronounced for the more productive workers. The lower attrition rate is mostly driven by an improvement in new workers' stress levels and mental health due to the lowering of their beliefs about senior coworkers' past performance. Overall, this study demonstrates that showing junior workers the "Curricula Vitae" of senior workers mitigates social comparison costs within firms.
In decision problems under uncertainty, the subjective evaluation of an outcome can depend on the information content of its realization. To accommodate this dependence, we introduce and axiomatize a model of information-dependent expected utility by allowing the utility of an outcome to flexibly depend on its information content in an (Anscombe-Aumann) act. Subjective beliefs are identified in a special class of our model where the utility of an outcome can be decomposed as the sum of consumption utility and information utility. Our model allows for both information seeking and information averse preferences, as well as a comparative theory of information preferences. For information seeking preferences, we introduce a Hidden Acts representation where the value of information is as if induced from the expected utility of the optimal choice in a fictitious future decision problem given that information.
