Overconfidence in Interval Estimates.

Abstract

Judges were asked to make numerical estimates (e.g., “In what year was the first flight of a hot air balloon?”). Judges provided high and low estimates such that they were X% sure that the correct answer lay between them. They exhibited substantial overconfidence: The correct answer fell inside their intervals much less than X% of the time. This contrasts with choices between 2 possible answers to a question, which showed much less overconfidence. The authors show that overconfidence in interval estimates can result from variability in setting interval widths. However, the main cause is that subjective intervals are systematically too narrow given the accuracy of one's information—sometimes only 40% as large as necessary to be well calibrated. The degree of overconfidence varies greatly depending on how intervals are elicited. There are also substantial differences among domains and between male and female judges. The authors discuss the possible psychological mechanisms underlying this pattern of findings.

Gender and the labor market: What have we learned from field and lab experiments?

“…The consistent finding from lab evidence is that men are more risk-prone than women (see surveys by Grossman, 2008, andGneezy, 2009). While some attribute these differences to the emotional reaction to uncertain situations (Lowenstein et al, 2001), others report that they are related to confidence (Soll and Klayman, 2004). Similar conclusions have been drawn from (scant) survey and field evidence (Dohmen et al, 2011;Dwyer et al, 2011).…”

The trouble with overconfidence.

“…and then have participants estimate 90% confidence intervals around their answers. Results show that these confidence intervals are too narrow, suggesting that people are too sure they know the correct answer; 90% confidence intervals contain the correct answer less than 50% of the time (Alpert & Raiffa, 1969Klayman, Soll, Gonzalez-Vallejo, & Barlas, 1999;Soll & Klayman, 2004).…”

Overconfidence and trading volume

“…The usual finding is that for all questions assigned a given probability the proportion of correct answers is lower than the assigned probability [Lichtenstein et al, 1982]. The result that people form probability distributions over uncertain quantities that are too tight seems to be robust especially when people judge difficult items [see Klayman et al, 1999 or Soll andKlayman, 2004].…”

“…See the CEPR version of this paper for details (Glaser and Weber [2003]). 22 See Section 3.1.1 and, for example, Cesarini et al [2006], Klayman et al [1999], Biais et al [2005], Soll and Klayman [2004]. 23 Seven investors answered one question, three investors answered two questions, four investors answered three questions, and nine investors answered four questions.…”