Dissertations

Dissertation Committee:
Andreas Ortmann (chair)
Ralph Hertwig
Dmitry Ryvkin
Peter Katuščák

Abstract:

The first chapter offers a theoretical model that suggests an alternative explanation to the so-called unskilled-and-unaware problem—the unskilled overestimate their skills while the skilled underestimate (but less than the unskilled). The unskilled-and-unaware problem was experimentally identified about a decade ago and numerous authors have elaborated on this problem since. We propose that the alleged unskilled-and-unaware problem, rather than being one of biased judgments, is a signal extraction problem that differs for the skilled and the unskilled. The model is based on two assumptions. First, we assume that skills are distributed according to a J-distribution, which can be regarded as an approximation of the very right tail of the IQ distribution. This assumption is reasonable given the typical subject pool used in the experimental studies of overconfidence – students from prominent US universities. Second, we assume an error term in own-ability perception, which is a common assumption in psychology models. Our simple model generates, by means of analytical computations, patterns similar to those identified in the previous experimental literature. We also discuss conditions under which the unskilled-and-unaware problem should disappear.

The second chapter reports the results of three experiments (one field, two laboratory) through which we tested the theoretical model and some informal extensions. Specifically, we examine the impact of general information and specific information (feedback) on the quality of absolute and relative self-assessment (“calibration”) in various tasks (microeconomics exam, skill-oriented task, and general-knowledge oriented task). In our experiments, we used a specific subject pool—CERGE-EI preparatory semester students who are competitively selected students from their home universities around Central and Eastern Europe. Overconfidence behavior initially prevails in almost all settings. We find a strong positive effect of general information on calibration. We also show that calibration improves more when feedback is provided. Moreover, our results suggest that the absolute self-assessment is more responsive to information. In our experiments we also show that it is the unskilled who improve their calibration the most. Based on the results, we conclude that information plays an important role in the absolute as well as the relative self-assessment and that the unskilled-and-unaware problem arises mostly due to the lack of information.

The third chapter reviews, categorizes, and evaluates experimental studies on overconfidence and self-assessment in business, economics, and finance. First, we review the main results of experimental research in psychology and highlight the main issues in psychology as well as current issues in economics. Then we create a non-opportunistic set of experimental studies from business, economics, and finance concerning overconfidence or self-assessment. We identify nine paradigms (General-knowledge questions, Confidence intervals, Forecasting, Market-entry games, Auctions, Willingness to sell/buy, Information, Assessment of others, Self-awareness questions) and categorize the experimental studies according to those paradigms. For each paradigm we then review the corresponding studies and point out the shortcomings of each study, paying attention to issues identified in psychology as well as to issues already known in economics. Finally, we discuss the existing research for each paradigm and, based on the review, make suggestions for further research.

Full Text: “Overconfidence in Business, Economics, Finance, and Psychology: How Much of a Problem is It?” by Marian Krajč