Don Moore of Berkeley’s Haas School of Business and his colleagues gave a math test to a group of volunteers after manipulating their confidence. Half were told that the results of their intake surveys predicted that they would get most answers right; the other half were told that they’d bomb the test. Going in, members of the first group were optimistic about how they would perform, while those in the second group had real doubts. Their expectations matched those of observers who were asked to guess which subjects would do better. But the two groups of test takers scored nearly the same. The conclusion: Confidence doesn’t always boost performance.

A version of this article appeared in the November–December 2020 issue of Harvard Business Review.