UC San Diego

Theoretical models in economics rely on the assumption that goal-oriented agents optimize with respect to statistically correct beliefs about the environment. Behavioral evidence casts doubt on the assumption of accurate beliefs. There is widespread evidence that agents' expectations are not just inaccurate, but systematically biased. This bias has potentially large implications in economic settings since most of economic models assume that the economic agents are free of these biases. This dissertation investigates the existence of the biased self image - individuals' systematically biased view on own behavior or choices compared to others - and attempts to provide rational frameworks on some of them. I examine a different bias in each chapter. In the first chapter, I study the extent to which agents systematically overestimate (or underestimate) their own attributes compared to others. I run a lab experiment which is designed to test whether agents possess this bias over objective and unambiguous attributes. In the second chapter, I collect survey data from two field tournaments where the participants make forecasts of their relative performance. I study the relationship between the accuracy of their forecasts and the quality of performance. In the third chapter, I propose a simple model of the false consensus bias that can explain, under mild assumptions, the bias as rational behavior. In the first essay, "(Biased) Self Image in Objective Qualities", I attempt to confirm the existence of a prominent systematic bias, identify the contexts in which it is likely to arise, and come up with behavioral assumptions that describe belief formation more accurately than conventional economic assumptions. While a large number of studies confirm the existence of biased self image, the results can often be traced to the ambiguities in the definition of the quality investigated. I run an experiment to find out whether individuals' beliefs about objectively defined qualities are biased. In particular, I describe data obtained by asking university students about information on their transcripts. Students answered questions about the number of high and low grades they received and the number of classes they took in different departments. They were also asked to rank their own characteristics relative to others. I attempt to test which behavioral model of biased self image explains the data the best. I found subjects accurately report their characteristics in their transcripts. Their relative assessments, however, are noisy. The fact that I could not find a compelling evidence of systematic bias can raise a question of the existence of biased beliefs in objective qualities. On the other hand, I found a strong evidence of people overestimating the proportion of the group with the same qualities. Also, using within subject analysis, I can categorize individual-level bias. With 5% significance level, almost 60% of the subjects are classified as possessing bias that can be explained by behavioral models of biased self image. The second essay, "Forecasts of Relative Performance in Tournaments; Evidence from the Field" is coauthored with Luis Santos-Pinto. We use a field survey to investigate the quality of individuals' beliefs of relative performance in tournaments. We use data obtained from two field settings, poker and chess, which differ in the degree to which luck is a factor and also in the information that players have about the ability of the competition. The main finding of the paper is that players' forecasts in both types of tournaments are biased towards overestimation of relative performance. However, the accuracy of the participants' beliefs differ between the two settings. We find that poker players' forecasts of relative performance are random guesses with an overestimation bias. Chess players also overestimate their relative performance but make informed guesses. We find support for the "unskilled and unaware hypothesis" in chess: high skilled chess players make better forecasts than low skilled chess players. Finally, we find that chess players' forecasts of relative performance are not efficient in the sense that they could enhance their accuracy of the forecast if they used the information about other participants' skill. One of the main findings of the first essay is that subjects do possess the false consensus bias - tendency for individuals to overestimate the prevalence of own behavior or choices. Psychological evidence suggests that the bias is frequently observed in various cases. But the existing analysis of the behavior is abstract. In the third essay "A Simple Model of the False Consensus Effect", I propose a simple model of information acquisition that describes this false consensus effect and several of its empirically identified qualitative properties. I show that, under mild assumptions, the tendency to believe one's own choice to be more informative about the behavior of a group than that of another's can be explained as a rational reaction. This 'generalized' false consensus effect has kept being considered irrational since the observation of Dawes (1989) that provides rational explanation for the false consensus effect. The paper shows that what is called 'truly false' by Dawes (1989) does not need to be false