Why Uninformed Agents (Pretend to) Know More

Why Uninformed Agents (Pretend to) Know More

Peter Schanbacher (Department of Quantitative Economics and Finance, University of Konstanz, Konstanz, Germany)
Copyright: © 2013 |Pages: 22
DOI: 10.4018/jsds.2013070102
OnDemand PDF Download:


Many social interactions (examples are market overreactions, high rates of acquisitions, strikes, wars) are the result of agents' overconfidence. Agents are in particular overconfident for difficult tasks. This paper analyzes overconfidence in the context of a statistical estimation problem. The authors find that it is rational to (i) be overconfident and (ii) to be notably overconfident if the task is difficult. The counterintuitive finding that uninformed agents which should be the least confident ones show the highest degree of overconfidence can be explained as a rational behavior.
Article Preview


Lichtenstein and Fischhoff (1977) showed that agents are overconfident in their answers. Agents overestimate the chance of stating the correct answer compared to their empirical frequency. Literature found that overconfidence is a robust and pervasive phenomenon to all kind of measures (Lichtenstein et al., 1982; McClelland & Bolger, 1994; Weinstein, 1980; Taylor & Brown, 1988). Overconfidence can be found in all kinds of human interactions (see e.g. overconfidence in driving (Svenson, 1981), mergers (Malmendier & Tate, 2005), trading (Odean, 1998), strikes (Neale & Bazerman, 1985) among others). Considering only economics overconfidence results in high rates of business failure (March & Shapira, 1987), overestimation of returns (Heaton, 2002), overinvestments in own company (Malmendier & Tate, 2005), excessively high rates of stock trading (Statman et al., 2006) and market overreaction (Daniel et al., 1998). The phenomenon we regard is called overprecision by Moore & Healy (2008) to distinct it from other similar effects such as overplacement and overestimating own performance. The most prominent method to study confidence are sets of two-choice questions, such as “Which of these nations has higher life expectancy, averaged across men and women: (A) Argentina, or (B) Canada?”. Participants choose what they believe to be the correct answer and then are directed to specify their degree of confidence that their answer is correct. After the participants answer many questions of this kind, the responses are grouped in distinct classes of confidence. The relative frequencies of correct answers in each confidence category are calculated. Plotting the empirical frequency against the level of confidence, results in the so called calibration curves. They have been found with a similar pattern - robust to the questions asked. Agents are underconfident for low confidence levels and overconfident for high level of confidence (Gigerenzer et al., 1991; Juslin et al., 1997). Broadly there are three types of explanations. Agents may be not able to solve the cognitive processing (Sniezek et al., 1990). They may suffer of a bias and underestimate counterevidence (Koriat et al., 1980; Griffin & Tversky, 1992). Finally they might overweight the supporting evidence on purpose for e.g. self-motivational reasons (Taylor & Brown, 1988).

Besides overconfidence a second effect is the hard-easy effect (von Winterfeldt & Edwards, 1986). Agents are overconfident for hard tasks but underconfident for easy tasks. It has been also argued that it is less the difficulty of the task but rather the individual performance of the agent which drives the level of confidence (Kruger & Dunning, 1999). Klayman et al. (1999) find stable individual difference in agent's confidence which indicates that the agent's background, e.g. knowledge and experience influences his confidence. The hard-easy effect is either explained by confirmation biases from the agent itself (Bjoerkman, 1994). Or by biases induced by the experimenter who creates a laboratory setting which is unrepresentative of natural ecology (Gigerenzer et al., 1991).

Complete Article List

Search this Journal:
Open Access Articles: Forthcoming
Volume 8: 4 Issues (2017)
Volume 7: 4 Issues (2016)
Volume 6: 4 Issues (2015)
Volume 5: 4 Issues (2014)
Volume 4: 4 Issues (2013)
Volume 3: 4 Issues (2012)
Volume 2: 4 Issues (2011)
Volume 1: 4 Issues (2010)
View Complete Journal Contents Listing