The Curse of Knowledge in Economic Settings: An Experimental Analysis

The question

Economics assumes a better-informed person can work out what a less-informed person will think. Can they?

The method

Laboratory market experiments rather than a psychology questionnaire, deliberately built to answer the objections economists raise to psychology findings: “we test arguments 1-3 by using market experiments to see whether financial incentives, learning from feedback, and market forces make the curse of knowledge disappear” [2].

The findings

The bias, stated plainly: “in predicting the judgments of others, agents are unable to ignore the additional information they possess” [2]. The word doing the work is unable. “Better-informed agents are unable to ignore private information even when it is in their interest to do so; more information is not always better” [1]. It is not a failure of care or of effort, and paying people to overcome it does not overcome it. That directly contradicts the standing assumption: “the conventional assumption in such analyses of asymmetric information is that better-informed agents can accurately anticipate the judgments of less-informed agents” [2]. Formally, the curse violates the law of iterated expectations — “better-informed agents should ignore their additional information when forecasting the forecasts of less-informed agents” [4], and they do not. (The term itself is not theirs: a footnote credits Robin Hogarth with suggesting it [2].)

The corrections did not correct it. “We find that feedback alone has little effect, while market forces reduce the magnitude of the curse by approximately 50 percent” [2]. Halved, in the most disciplined setting the authors could build; “subjects in markets show about half as much bias” [3].

One consequence the authors dwell on is that the bias can cost the person who has it: “better-informed agents may suffer losses. More information can actually hurt” [2].

The limits

The setting is stylised — laboratory markets with student traders forecasting other subjects’ judgments about earnings data, not experts describing a job. What transfers is the mechanism and its stubbornness, not a rate. The paper is also about predicting judgments, which is narrower than the everyday sense in which people use “curse of knowledge” to mean bad explanations.

• An expert on a Calibration Call tells you what a competent person “obviously” knows. Without this paper you treat that as a report. This says their model of the less-informed person is systematically contaminated by what they know, and that neither incentives nor experience fixes it.
• You are asking someone to predict how a candidate will do. The same machinery is running. The person judging cannot subtract their own knowledge from the model of the person they are judging.
• You are writing the job description after doing the job yourself. You are now the better-informed agent, and the paper’s finding is about you. Get someone at the level you are hiring for to read it back to you.