Dynamic Games, Social Preferences, and Assignment Markets: Theory and Experiments

Artur Dolgopolov

Advisor: Cesar A Martinelli, PhD, Department of Economics

Committee Members: Daniel E Houser, Igor Griva

Online Location, Online
July 29, 2020, 02:30 PM to 04:30 PM


In this dissertation, I discuss robust theoretical approaches to three dynamic environments with progressively larger scope: social preferences, strategic games, and markets for indivisible goods. A significant part of human interaction is inherently dynamic, but the study of such dynamic systems is hindered by the multitude of competing explanations. The complexity of these environments also suggests that agents may depart from theoretical predictions due to decision errors and heuristics. Each chapter contains empirical tests addressing these issues, comparisons of theories, and measures of their power and predictive success. Finally, I show how to leverage mixed-integer programming to take these models to the data, bringing the problems together methodologically.

The first chapter proposes an algorithm to learn the strategies of players from the observed sequence of play in repeated games. The algorithm accounts for the possibility of measurement and decision making errors and stays agnostic about equilibrium restrictions. I provide conclusive evidence that players use strategies of memory no more than one or two periods and recover new strategic patterns in games of Prisoner's dilemma and asymmetric coordination games.

In the second chapter, I construct non-parametric tests for two theories of social preferences - inequality aversion and increasing benevolence, both having important implications for social welfare choices and charitable giving. A laboratory experiment shows significant heterogeneity among the subjects, and that inequality aversion, increasing benevolence, and general other-regarding preferences explain the behavior of comparable shares of the sample.

In the third chapter, I study theoretically and experimentally assignment markets: two-sided markets where indivisible heterogeneous items with unit demand and unit supply are traded for money. I explore how different trading institutions help agents discover and reach optimal assignments in a noncooperative environment. Especially in view of the complexity of the market, experimental results are perhaps surprisingly well-predicted by theory. Market outcomes are close to Nash equilibrium predictions under auction-like institutions and close to generalized bargaining for institutions that feature decentralized communication.