Overconfidence Through Group Composition

Overconfidence through Group Composition

The Effect of Group Composition on Overconfidence

Msc Behavioral Economics

Student: Arjan den Besten

Student Number: 415298

Supervisor: Chen Li

Co-Reader: Prof. Peter Wakker

Date of Completion: 17-6-2015

Erasmus School of Economics

Table of Contents

Abstract……………………………………………………………………………………….3

1. Introduction………………………………………………………………………………..5

2. Literature review…………………………………………………………………………..8

Overconfidence……………………………………………………………………...…8

Group Composition…………………………………………………………………...12

Willingness to bet……………………………………………………………………..13

Anchoring……………………………………………………………………………..13

3. Research question, conceptual model and hypotheses………………………………….15

4. Methodology……………………………………………………………………………….17

Participants and procedure design ……………………………………………………17

Dependent measures…………………………………………………………………..18

Individual differences measures………………………………………………………19

The questionnaire……………………………………………….…...... …19 5.Statistical Analysis………………………………………………………………………..21

6. Results……………………………………………………………………………………..23

7. Conclusion………………………………………………………………………………...30

Main findings…………………………………………………………………………31

Recommendations…………………………………………………………………….32

Future research………………………………………………………………………..33

Bibliography………………………………………………………………………………....35

Appendix……………………………………………………………………………………..38

Abstract

Moore and Healy (2008) present 3 distinct ways in which the research literature has definedoverconfidence: (a) overestimation of one’s actual performance, (b) over-placement of one’s performance relative to others and (c) excessive precision in one’s belief (over-precision). This thesis has attempted to find out what the influence of group composition is on the three types of overconfidence through conducting a hypothetical pub quiz among students. In the treatment group (n=30), subjects were told that they were competing with lower-skilled and less-educated participants, whereas inthe control group (n=30) their opponents were university students just like the respondents. The results have shown that overestimation occurs more when a student is hypothetical surrounded by lower-skilled and less-educated participants than themselves. On the other hand, the student is under-placing herself/himself to other participants. Further, it is shown that people, especially men, who are very confident about their own performance, are willing to bet more money in competitions.A number of possible explanations for the resultsand corresponding recommendations will be outlined in the conclusion of this study.

Chapter 1: Introduction

People are influenced by other people’s presence all the time. This could happen consciously, but unconsciously as well. For example, the presence of other people around you has an influence on the meal size you choose (De Castro & De Castro, 1989). The more people around you when you choose your meal, the larger the meal size that you choose will be.One other study reveals the social influence on people their driving speed in a car. Passengers, parents, but also unknown other drivers have an influence on the behavior of the driver(Judy & Alexia, 2010). For instance, the presence of passengers has a slowing influence on drivers, because the driver feels responsible forthe safety of people in the car. More research can still be done on the influence of the presence of others on people´s behavior.The aim of this thesis is to measure the influence of the presence of two different groupson people their overconfidence. Overconfidence is anoften occurring phenomenon in daily life, whose relation to stock investment, gender, CEO decision making, corporate investment and entrepreneurship (Barber et al., 2015; Bernardo & Welch, 1997; Malmendier & Tate, 2005) has been studied extensively in the literature. However the definition of overconfidence differs from study to study. Moore and Healy (2008) categorized the definition of overconfidence in the literature into the following three types.This first type of overconfidence is overestimation of one’s actual performance. Overestimation basically means that people believe that their own abilities are greater than they indeed are. For example, you could be overconfident in your physical strength when doing weight lifting exercises in the gym, and end up not being able to finish your weight lifting sets. The second type of overconfidence is over-placement of one’s performance relative to others. Over-placement means that people believe that they are better than others. For instance, a student has to do an exam. She estimates that her own grade will be an eight and the average grade of the class will be a seven. Then she does the exam and she only scores a seven. The actual average grade of the class is an eight. In this example the student is over-placing herself to others in class. The third type of overconfidence is over-precision, which is often defined as excessive precision in one’s beliefs. For instance, when people are ask to estimate a confidence interval of the length of a river, they have the tendency to choose a confidence interval that is too narrow.More specific,people estimate that the length of a river will fall within the 2-4 kilometers interval. They are 90% sure that this interval is correct. But, the actual length of the river is 5 kilometers. This means that people become over-precise in their estimation of the confidence interval. Healy and Moore (2008) show interesting findings about how the three types of overconfidence can interact. On difficult tasks, people overestimate their actual performances but also believe that they are worse than others (under-placement). On easy tasks, people underestimate their actual performance, but also believe that they are better than others. There is a theory that can explain these inconsistencies. After experiencing a task, people often have imperfect information about their own performances but even worse information about the performances of others. As a result, people’s post-task estimates of themselves are regressive, and their estimates of others are even more regressive. Thus, whenthe actual performance is extremelyhigh, people will underestimate their own performances, underestimate others even more so, and thus believe that they are better than others. When performance is low, people will overestimate themselves, overestimate others even more so, and therefore believe that they are worse than others.The last type of overconfidence is over-precision, which appears to be more persistent than either of the other 2 types of overconfidence, but its presence reduces the magnitude of both overestimation and over-placement. The design of this thesis is adapted from the study of Moore and Healy (2008). In this research subjects participate in a hypotheticalpub quiz competition, where the composition of participants in the quiz competition differ between the treatment and control group. In the treatment group, the participant pool consists of low-skilled workers whose education background is different from the subjects (university students), whereas in the control group, the participants are mostly university students. Next to studying the possible relationship between group composition and overconfidence, this thesis also studies the possible relationship between relative estimation and willingness to bet on their own quiz performance. It is already shownthat when people have excessive confidence in their knowledge and are ready to bet on it, they even accept bets with a negative expected value(Lichtenstein & Fischhoff, 1977).For this reason, the aimis to find a positive correlation between overconfidence and willingness to bet. To summarize, the aim of this thesisis to find an answer onthe following research question: “What is the effect of group composition on the overconfidence of people and is overconfidence positively correlated with people their willingness to bet on their own performance? The findings of this thesiscan be used by organizations, managers and policy makers when they start forming work teams within their organizations. When the findings show that people become more overconfident in a specific group composition, organizations could take this into account when they form new work teams in order to prevent overconfidence within their organization. For instance, organizations should be aware of possible overconfidence when a team is formed with only one high-skilled worker and many low-skilled workers. When the high-skilled worker has to take an important and risky decision, he might become overconfident when he is only surrounded with lower-skilled and less-educated workers around him. This problem can be solved by forming teams that consists a proper mix of people with different educational backgrounds.This thesis could also be interesting for organizations where risky decisions have to be made by people where overconfidence may easily occur. Traders, brokers, bankers and workers at financial institutions should consider to take into account the findings of this research (Dittrich, Güth, & Maciejovsky, 2005). Even primary schools could think about the formation of the groups of children in school. Putting one very intelligent child in a group with three less intelligent children might give the intelligent child too much confident. To prevent this, it is better to form groups where there is a proper mix of different levels of intelligence of children. On the other hand, not in all cases overconfidence has to play a negative role in organizations. For instance, managerial overconfidence can increase the value of the firm. So even if organizations want to stimulate overconfidence instead of preventing it, this thesis could be helpful(Gervais, Heaton, & Odean, 2002).

Chapter 2: Literature review

Overconfidence

Overconfidence can have some serious consequences. Researchers have offered overconfidence as an explanation for entrepreneurial failures, stock market bubbles and even wars (Moore & Healy, 2008). According to Johnson and Fowler (2011) humans exhibit many psychological biases, but one of the most consistent, powerful, and widespread bias is overconfidence. Overconfidence amounts to an “error” of judgment or decision-making, because it leads to: (1) overestimating one’s capabilities, and/or underestimating an opponent; (2) underestimating the difficulty of a task; (3) or underestimating possible risks. According to (Minniti, 2005) overconfidence is a perceptual distortion in the sense that one’s own abilities are systematically overestimated. Own ability means one’s own success potential. This definition is in line with the definition of overestimating one’s capabilities of Johnson and Fowler (2011).

Moore and Healy (2008) show that the research literature has defined overconfidence in three distinct ways. The first definition of overconfidence is overestimation of one’s actual ability, performance, level of control, or chance of success. People believe that their own abilities are greater than they indeed are. For example, students robustly overestimate their exam performance (Clayson, 2005). This definition is very similar to the definition of (Johnson & Fowler, 2011; Minniti, 2005). The second definition of overconfidence is over-placement of one’s performance relative to others, such as when a majority of a group of people rate themselves better than the median. In other words, people believe that they are better than others. The third definition of overconfidence is excessive precision in one’s beliefs, which is also known as over-precision. In other words, individuals believe that their estimates are more accurate than they actually are, they are overconfident about their precision (Croson & Croson, 2008). For example, when people are asked to estimate a confidence interval of the length of a river, they have the tendency to choose a confidence interval which is too narrow. In other words, people are too sure that they give the right answer and through this overconfidence they are often too precise with their estimations. Over-precision appears to be more persistent than over-placement and over-precision(Moore & Healy, 2008). Russo and Schoemaker (1992) show interesting results about over-precision in group judgments compared to individual judgments. They found that group judgments were significantly less over-precise than individual judgments. So when people discuss their confidence ranges with other people, their private judgments and the group judgments become less over-precise. When people form their confidence ranges without discussing it with other people, their private judgments becomes more over-precise.

Overestimation and over-precision are about an individual’s own ability and beliefs, while over-placement is about one’s performance relative to others. Even though, over-placement and overestimation can appear together and they often do, they are not necessarily correlated (Croson & Croson, 2008). The study of Moore & Healy (2008) demonstrates useful findings about overestimation and over-placement that both types of overconfidence are not necessarily correlated. Participants underestimated their performances on easy quizzes, overestimated them on difficult quizzes, and were accurate on medium quizzes. Participants over-placed their performances on easy quizzes, under-placed their performances on difficult quizzes and demonstrated no over- or under-placement on medium quizzes. Noteworthy is the fact that participants underestimated their performances on easy quizzes, but at the same time they over-placed their performances on these same easy quizzes. This contradiction also applies for the difficult quizzes, but then the other way around. Participants overestimated their performances on difficult quizzes, but at the same time they under-placed their performances on these same difficult quizzes. Healy and Moore (2008) relate the negative relationship between overestimation and over-placement to a theory of confidence. Namely, after experiencing a task, people often have imperfect information about their own performances but even worse information about the performances of others. As a result, people’s post-task estimates of themselves are regressive, and their estimates of others are even more regressive. Thus, when performance (difficult task) is low, people will overestimate themselves, overestimate others even more so, and therefore believe that they are worse than others. The theory of confidence assumes that people hold a subjective probability distribution(SPD) of their own score and a SPD of others’ scores over the range of possible outcomes on any task. Overestimation occurs when a person’s SPD of his or her own score has a mean that is greater than the person’s actual score on the task. Over-placementoccurs when a person’s SPD of his or her own score has a mean that is greater than the mean of the SPD of others’ scores and this belief is not justified by differences in actual scores between self and others. Strongly involved with over-placement are self-perception and our perception of others.Wojciszke(2005) demonstrates that our perception of others is different from our self-perception. With the perception of others, the perceiver is more interested in morality thancompetence of a person. The perceiver construes their behavior in moral terms and emotional responses are more based on morality than competence considerations. On the other hand, for self-perception our behavior is more construed in competence terms than in moral terms. In addition, our own competence influences self-evaluations and emotional responses more than our own morality does. In other words, when forming evaluations of others, the perceiver is more interested in the moral behavior of a person, while our own behavior is more construed in competence terms. It is reasonable to assume that self-perception is involved in overestimation, over-placement and over-precision. And others-perception is only involved in over-placement. The definitions of overconfidence defined in the paper of Moore and Healy (2008) are best suited to use in this research paper, because they demonstrate that the research literature has defined overconfidence mostly in three distinct ways: overestimation, over-placement and over-precision. These three definitions are included and tested in this thesis.According to Moore and Healy (2008) overestimation, over-placement and over-precision are conceptually and empirically distinct. But still, all three types are defined in the literature as overconfidence. Russo and Schoemaker (1992)demonstratehow overconfidence can be managed. Theyexamine the causes and remedies for overconfidence. The potential cognitive causes of overconfidence are the availability bias, anchoring, the confirmation bias and hindsight. First, the availability bias means that people have difficulties in imaging all the ways that events can unfold. Second, anchoring is the tendency to anchor on one value or idea and not adjust away from it sufficiently. This thesis will investigate the effect of anchoring on overconfidence as well. The section ‘Anchoring and adjustment’ elaborates on this effect. Third, the confirmation bias means that when we make predictions or forecasts, we often lean toward one perspective and the natural tendency is to seek support for our initial view rather than to look for disconfirming evidence. And lastly, hindsight makes us believe that the world is more predictable than it really is. Klayman and Soll (1999)give two different explanations why people are generally overconfident. The first explanation comes from the biases of people when they process information. The judge, the person who does the judgment about something in his own mind, first searches memory for relevant information and arrives at a tentative answer. Then, with this answer in mind, the judge searches for more evidence. The problem is that the judges believe that their processes are unbiased and thus perceive more support for an initial guess than is actually warranted.Motivational factors can intensify this bias. People like to think that they are intelligent and knowledgeable, and they may have reasons for wanting a particular answer to be true. This first explanation of Klayman and Soll (1999) is very similar to the confirmation bias that is given as a potential cognitive cause of overconfidence by Russo and Schoemaker (1992). The confirmation bias means that when we make predictions or forecasts, we often lean toward one perspective and the natural tendency is to seek support for our initial view rather than to look for disconfirming evidence.