Russel Sage Foundation

Rational endogenous information acquisition plays a key role in many economic models. This assumption may be violated, however, if agents are overconfident about their knowledge. I design a novel experiment to directly test whether people pay too much or too little for information, and to quantitatively assess the role played by overconfidence. The subjects - internet businesspeople experts - make guesses about the prices of actual domain names and the quality of actual websites, mimicking behaviors they engage in in their professional lives. They then have the opportunity to purchase information to improve their guesses. I study the demand for both direct information of known accuracy and for subjective information from other experts. I find that subjects underpay for both types of information. Underpayment is widespread, with over 80% of subjects systematically underpaying for highly accurate signals. Both across subjects and across tasks, overconfidence is associated with a lower demand for information, but overconfidence is unlikely to provide a full explanation for suboptimal information investment.
Even if the value of information is adjusted to account for subjects’ overconfidence and subjects’ tendency to sometimes misuse information, subjects still underpay for information.