Essays on behavioral and experimental economics

Abstract

In this dissertation of three chapters, I study individuals’ strategic sophistication in decision-making, specifically level-k reasoning and forward-looking behavior. The first chapter studies subjects’ iterative reasoning ability to derive equilibrium and effectiveness of teaching in improving their decisions (if subjects have bounded reasoning ability). In the experiment, I first measure subjects’ iterative reasoning ability using rational computer opponents to control strategic uncertainty from beliefs about opponents’ moves. I find that 90% of the subjects cannot play equilibrium. Next, I randomly treat subjects with a tutorial about iterative reasoning, and I find that 45% of the subjects learned to play equilibrium. Finally, I study subjects’ perceived ability and improvement by asking subjects to assess their performance. I find that low-ability subjects who do not learn have persistent misperception (i.e., overestimation) of their own performance and learning ability to improve performance. The findings in this chapter suggest that training on strategic thinking is a cost-effective intervention to improve decisions and to aid policy implementation, and low-ability subjects need a more intensive treatment.

The first chapter shows that subjects have limited reasoning ability, which might explain the well-documented bounded rationality in games. However, it is still unclear at the individual level whether the low-level of rationality is due to limited reasoning ability or low-order beliefs about opponents’ rationality. The second chapter reports a within-subject experiment on Amazon Mechanical Turk (Mturk), where subjects played ring games against two types of opponents simultaneously, other Mturk subjects and themselves. Lk players who are bounded by their ability would display the same reasoning depth when facing either type of the opponent (ability-bounded Lk), otherwise the players would perform higher reasoning depth when playing against themselves than other participants (belief-bounded Lk). I find that 76% of them are ability-bounded Lk players while only 10% are belief-bounded Lk players, indicating that limited reasoning ability is the primary determinant of bounded rationality.

The third chapter provides an experimental investigation of the evolutionary game model (Oyama et al., 2015) which predicts transitions among strict Nash equilibria under inexact (inaccurate but unbiased) information of opponents’ behaviors. We design a quasi-continuous-time experiment with two treatments differing in information accuracy. A group of subjects played a coordination game repeatedly in either treatment. We observe more efficiency-improving transitions among strict Nash equilibria in the more accurate information treatment than in the less accurate information treatment, contrary to the theory. We further find that more accurate information about opponents’ behaviors induces more subjects to engage in forward-looking behavior, i.e., persistent strategic deviations from the myopic best responses to the information received, which facilitates efficiency-improving equilibrium transitions. When information is less accurate, subjects are less responsive to changes in the information. The slow response to the information either blocks or delays efficiency-improving equilibrium transitions.