Mechanism Design Theory, introduced by 2007 Nobel laureates Hurwicz, Maskin, and Myerson, has guided economic institutions worldwide to achieve desirable goals in allocating scarce resources. However, most of the literature on Mechanism Design Theory that guides its application, in reality, assumes that people are fully rational; this omission of people’s bounded rationality raises doubt over the reliability of the theory’s empirical implications. To bridge this gap between theory and reality, we introduce new formalizations to characterize new types of boundedly rational behavior that is missing in existing models but supported by experimental evidence.
NLK, the first formalization we propose, is a new solution concept in Game Theory that connects two existing ones, Nash Equilibrium (NE) and Level-K model. Of these two, NE, introduced by 1994 Nobel Laureates John Nash has revolutionized the economics of Industrial Organization and has influenced many other branches such as the theories of monetary policy and international trade. However, there is mounting and robust evidence from laboratory experiments of substantial discrepancy between the predictions of NE and the behavior of players. Among all the alternative models that retain the individual rationality of optimization, but relax correct beliefs, Level-K model is probably the most prominent. Absent in NE, Level-K model explicitly allows players to consider their opponent as less sophisticated than themselves. But Level-K does not allow players to use an important element of strategic thinking, namely, “put yourself in the others’ shoes” and believe the opponent can think in the same way they do. Bridging NE and Level-K, NLK allows a player in a game to believe that her opponent may be either less- or as sophisticated as they—a view supported by various studies in Psychology. We compare the performance of NLK to that of NE and some versions of Level-K by applying it to data from three experimental papers published in top economics journals and to data from a field study. These studies allow us to test NLK on: (1). A static game of complete information, (2). A static game of incomplete information, (3). A dynamic game of perfect information, and (4). On field data.
NLK provides additional insights to those of NE and Level-K. Moreover, a simple version of it explains the experimental data better in many cases. As a new solution concept, NLK shares a similar foundation to NE but is also applicable to games with players of different cognitive or reasoning abilities. As an analytical tool, NLK exists and gives a sharp prediction in general, and therefore it can be applied to empirical analysis in a broad range of settings.
In the second formalization, we first propose two alternative axiomatic approaches, formalizing a distinct anomaly in hypothetical reasoning that agents fail to reason state-by-state. Our theory expands the foundation of Decision Theory and ties together a broad range of evidence documented in multiple disciplines that decision makers often choose a dominated strategy. Secondly, we extend our concept to Game Theory and Mechanism Design, where we identify a rich class of mechanisms that successfully achieve desirable goals even with boundedly rational agents and agents who mistrust the market makers. Thirdly, we test and verify our theory and its implications, by two laboratory experiments with a cross-over design that enables pooled data, within-subject, and cross-subject comparisons. Finally, we address how our approach contributes to accomplishing two goals simultaneously in modeling bounded rationality: providing a unified framework that subsumes existing ones as limiting cases and stimulating transdisciplinary conversations connecting the concepts of heuristics and emotions in Psychology, the utilization of eye-tracking technology in Neuroscience, and considerations of the moral foundation underlying a mechanism design in Ethics. The general insights of our work can be transferred to practical impacts on applications of Mechanism Design. Among these applications are the U.S. Federal Communications Commission auctions that raise more than 10 billion dollars yearly in government revenue; College Admissions that affect more than 10 million students every year around the world; and a Kidney Exchange Program with more than 1 million people waiting for kidney transplants. By formalizing bounded rationality into economic theory, our study honors the elegance of classic economic theory; at the same time, by modeling human behavior even more closely, it directs us to a new way of improving human welfare.
In the history of economic thought lies a dilemma for future economists: should we adopt simple models with unrealistic assumptions, or should we describe human behavior closely but give up elegant abstractions? In the projects above, we endeavor to create a middle way that synthesizes the merits in both directions and leave unanswered questions for future researchers.