The Dirty Faces game requires players to perform iterative reasoning in order to arrive at equilibrium play. The game is dominance solvable with a unique equilibrium when it is correctly specified. The particular payoff structure has significant implication on whether the reasoning process leads to equilibrium play. This paper illustrates that the traditional specification - as used by Weber (2001) - leads to multiple equilibria and the game loses its dominance solvability. We modify the payoff structure and restore uniqueness. The resulting game, which is dominance solvable, is implemented in an experiment to test the depth of iterative reasoning in humans. Our data analysis suggests that some deviation from equilibrium play is due to limited depth of iteration. Additionally, we find evidence that the lack of confidence in other players’ iterative abilities also induces deviations from equilibrium play. / Thesis (M.Ec.) -- School of Economics, 2007

The Dirty Faces game requires players to perform iterative reasoning in order to arrive at equilibrium play. The game is dominance solvable with a unique equilibrium when it is correctly specified. The particular payoff structure has significant implication on whether the reasoning process leads to equilibrium play. This paper illustrates that the traditional specification - as used by Weber (2001) - leads to multiple equilibria and the game loses its dominance solvability. We modify the payoff structure and restore uniqueness. The resulting game, which is dominance solvable, is implemented in an experiment to test the depth of iterative reasoning in humans. Our data analysis suggests that some deviation from equilibrium play is due to limited depth of iteration. Additionally, we find evidence that the lack of confidence in other players’ iterative abilities also induces deviations from equilibrium play. / Thesis (M.Ec.) -- School of Economics, 2007

This dissertation studies econometric analysis of game-theoretic models. I develop novel empirical models and methodologies to facilitate robust and computationally tractable econometric analysis. In Chapter 1, I develop an empirical model for analyzing stable outcomes in the presence of incomplete information. Empirically, many strategic settings are characterized by stable outcomes in which players’ decisions are publicly observed, yet no player takes the opportunity to deviate. To analyze such situations, I build an empirical framework by introducing a novel solution concept that I call Bayes stable equilibrium. The framework allows the researcher to be agnostic about players’ information and the equilibrium selection rule. Furthermore, I show that the Bayes stable equilibrium identified set is always weakly tighter than the Bayes correlated equilibrium identified set; numerical examples show that the shrinkage can be substantial. I propose computationally tractable approaches for estimation and inference and apply the framework to study the strategic entry decisions of McDonald’s and Burger King in the US. In Chapter 2, I study identification and estimation of a class of dynamic games when the underlying information structure is unknown to the researcher. I introduce Markov correlated equilibrium, a dynamic analog of Bayes correlated equilibrium studied in Bergemann and Morris (2016), and show that the set of Markov correlated equilibrium predictions coincides with the set of Markov perfect equilibrium predictions that can arise when the players might observe more signals than assumed by the analyst. I propose an econometric approach for estimating dynamic games with weak assumption on players’ information using Markov correlated equilibrium. I also propose multiple computational strategies to deal with the non-convexities that arise in dynamic environments. In Chapter 3, I propose an extremely fast and simple approach to estimating static discrete games of complete information under pure strategy Nash equilibrium and no assumptions on the equilibrium selection rule. I characterize an identified set of parameters using a set of inequalities that are expressed in terms of closed-form multinomial logit probabilities. The key simplifications arise from using a subset of all identifying restrictions that are particularly easy to handle. Under standard assumptions, the identified set is convex and its projections can be obtained via convex programs. Numerical examples show that the identified set is quite tight. I also propose a simple approach to construct confidence sets whose projections can be obtained via convex programs. I demonstrate the usefulness of the approach using real-world data.

This dissertation studies information economics in strategic and decision settings. In Chapter 1, I introduce flexible endogenous monitoring into dynamic moral hazard. A principal can commit to not only an employment plan but also the monitoring technology to incentivize dynamic effort from an agent. Optimal monitoring follows a Poisson process that produces rare informative signals, and the optimal employment plan features decreasing turnover. To incentivize persistent effort, the Poisson monitoring takes the form of "bad news'' that leads to immediate termination. Monitoring is non-stationary: the bad news becomes more precise and less frequent. In Chapter 2, which is joint work with Qingmin Liu, we analyze a model of strategic exploration in which competing players independently explore a set of alternatives. The model features a multiple-player multiple-armed bandit problem and captures a strategic trade-off between preemption---covert exploration of alternatives that the opponent will explore in the future---and prioritization---exploration of the most promising alternatives. Our results explain how the strategic trade-off shapes equilibrium behaviors and outcomes, e.g., in technology races between superpowers and R&D competitions between firms. We show that players compete on the same set of alternatives, leading to duplicated exploration from start to finish, and they explore alternatives that are a priori less promising before more promising ones are exhausted. In Chapter 3, I study how a forward-looking decision maker experiments on unknown alternatives of spatially correlated utilities, modeled by a Brownian motion so that similar alternatives yield similar utilities. For example, a firm experiments on its size that yields unknown, spatially correlated profitability. Experimentation trades off the opportunity cost of exploitation for the indirect inference from the explored alternatives to unknown ones. The optimal strategy is to explore unknown alternatives and then exploit the best known alternative when the explored becomes sufficiently worse than the best. The decision maker explores more quickly as the explored alternative worsens. My model predicts the conditional Gibrat's law and linear relation between firm size and profitability.

The general topic of our thesis is forward looking behavior in strategic situations. Mixing theoretical and experimental analysis, we document how strategic thinking is affected by the specific features of a dynamic interaction. The overarching result is that the information regarding decisions that are close to the current one, receive a qualitatively different consideration, with respect to distant ones. That is, the actual decisions are based on reasoning over a limited number of steps, close to actual decison node. We capture this feature of behavior both in a strategic (limited backward induction) and in a non-strategic (limited farsightedness) set up, and we identify relevant consequences on the outcome of the interaction, which powerfullly explain many observed experimental regularities.<p>In the first essay, we present a general out-of-equilibrium framework for strategic thinking in sequential games. It assumes the agents to take decisions on restricted game trees, according to their (limited) foresight level, following backward induction. Therefore we talk of limited backward induction (LBI). We test for LBI using a variant of the race game. Our design allows to identify restricted game trees and backward reasoning, thus properly disentangling LBI behavior. The results provide strong support in favor of LBI. Most players solve intermediate tasks - i.e. restricted games - without reasoning on the terminal histories. Only a small fraction of subjects play close to equilibrium, and (slow) convergence toward it appears, though only in the base game. An intermediate task keeps the subjects off the equilibrium path longer than in the base game. The results cannot be rationalized using the most popular models of strategic reasoning, let alone equilibrium analysis.<p>In the second essay, a subtle implication of the model is investigated: the sensitivity of the players’ foresight to the accessibility and completeness of the information they have, using a Centipede game. By manipulating the way in which information is provided to subjects, we show that reduced availability of information is sufficient to shift the distribution of take-nodes further from the equilibrium prediction. On the other hand, similar results are obtained in a treatment where reduced availability of information is combined with an attempt to elicit preferences for reciprocity, through the presentation of the centipede as a repeated trust game. Our results could be interpreted as cognitive limitations being more effective than preferences in determining (shifts in) behavior in our experimental centipede. Furthermore our results are at odds with the recent ones in Cox [2012], suggesting caution in generalizing their results. Reducing the availability of information may hamper backward induction or induce myopic behavior, depending on the strategic environment.<p>The third essay consists of an experimental investigation of farsighted versus myopic behavior in network formation. Pairwise stability Jackson and Wolinsky [1996] is the standard stability concept in network formation. It assumes myopic behavior of the agents in the sense that they do not forecast how others might react to their actions. Assuming that agents are perfectly farsighted, related stability concepts have been proposed. We design a simple network formation experiment to test these extreme theories, but find evidence against both of them: the subjects are consistent with an intermediate rule of behavior, which we interpret as a form of limited farsightedness. On aggregate, the selection among multiple pairwise stable networks (and the performance of farsighted stability) crucially depends on the level of farsightedness needed to sustain them, and not on efficiency or cooperative considerations. Individual behavior analysis corroborates this interpretation, and suggests, in general, a low level of farsightedness (around two steps) on the part of the agents. / Doctorat en Sciences économiques et de gestion / info:eu-repo/semantics/nonPublished