| The dissertation consists of three essays on bounded rationality and behavioral heterogeneity using structural estimation with experimental data. The first essay investigates social learning in networks with three-person, connected networks experimentally. Learning in networks is challenging when individuals have different observational neighborhoods because they must draw higher-order inferences in order to process correctly the information revealed by their neighbors' behavior. Due to the inevitable noise in the data, we modify the Bayesian theory by combining it with the notion of Quantal Response Equilibrium (QRE). The estimated decision rules of the Bayesian QRE model at the aggregate level appear to be qualitatively similar to the data. Nonetheless, we observe heterogeneous behavior that is in contrast to Bayesian rationality.; The second essay provides an alternative model of learning in networks, a cognitive hierarchy (CH) model, and uses the same experimental data to estimate the model and perform a goodness-of-fit test. Motivated by a hierarchy of cognitive tasks in the games, We define a cognitive hierarchy that comprises a set of cognitive types whose behavior ranges from random to substantively rational. The CH model is not only flexible enough to detect various levels of rationality but also allows a comparative-static comparison of equilibrium type distributions across different environments. The structural estimation results reveal that the majority of the subjects in the laboratory exhibit Bayesian rationality and that the CH model succeeds in replicating the empirical patterns of learning dynamics in the laboratory.; The third essay studies experimentally the substantive and procedural rationality of decision making under uncertainty. By using novel graphical representations of budget sets over bundles of state-contingent commodities, we generate a very rich data set at the individual level. We find that, using revealed preference theory, most subjects behave consistently with the utility-maximization hypothesis. Given the prevalence of heterogeneity, we detect prototypical heuristics and propose a type-mixture model (TMM) based on Expected Utility Theory employing three distinctive heuristics. The TMM explains the individual-level behavior well and suggests a linkage between the substantive and procedural rationality of decision making under uncertainty. |
