I focus on the theoretical and empirical consequences of different types of bounded rationality assumptions and varying levels of economic rationality. The basic nature of mixed equilibria is intuitively plausible since it is often easy to recognize the advantages of being unpredictable: in sports, a pitcher tries to randomize his throws while the batter tries to outguess him; in crime and terrorism prevention, the time and place of attack are generally uncertain. And yet, we are sometimes at a loss of finding intuitive explanations for the specific probabilities in the prescribed equilibrium. I start by postulating a new behavioral bias by exploring what happens when some players use simple rules of thumb to choose among pure strategies that they perceive to lead to the highest payoff. Naïve players were conceived as the simplest possible players that did not directly contradict the fundamental assumption of utility maximization. Then I find experimental evidence that some players are indeed affected by this proposed behavioral bias. Finally, I find evidence that there exist experimental conditions that increase this bias at the individual level. These findings contribute to our general understanding of the determinants of economic choices and the (actual) nature of economic rationality.To illustrate the theory, one can imagine choosing between left or right when expected utility is constant. NE forecasts the choice depends on the payoffs of the other players. I relax typical assumptions on rationality and propose some players will flip a coin when indifferent. They can see payoffs, but they lack any strategic depth. I then prove that if their proportion is small enough, any NE of a game with no naïve players corresponds to an equilibrium of the generalized game with some naïve players where payoffs for all players are the same. The intuition is that the rest of the population compensates in the opposite direction of the distortion. Thus naïve mixers are not disadvantaged by using a behavioral rule of thumb. Nor are rational players in a position that allows them to increase their payoffs: there is no rent to be gained by this strategy restriction and no welfare loss to be fixed by a social planner.Trying to bridge the divide between theory and reality, in my second dissertation chapter, I experimentally confirm the existence of some players who consistently mix close to 50% in different settings. I first sort participants into naïve and non-naïve by letting them play variations of asymmetric matching pennies. Two weeks later, each group plays against varying proportions of automated computer players (bots) that follow changing off-equilibrium strategies, and I observe several measures of how they react to the distortions. Besides identification of naïve players, I show that the probability of being naïve can be modeled by a quantitative test: smart players today, play smart tomorrow, and smart players don't always mix uniformly. I also find evidence that the non-naïve population reacts better to off-equilibrium behavior, plays closer to NE and adjusts their behavior in the correct direction but not with the magnitude required to restore equilibria. I then employ this last result to design simple mechanisms to obtain above-equilibrium payoffs by taking advantage of naïve players.In my final paper, I test the determinants that make an individual more (or less) naïve under different conditions related to game complexity and game stakes. I find evidence that players sometimes behave relatively close to coin flipping under a distractor which consists of adding weakly dominated strategies to matching pennies games. When they face a focuser which consists of monotonically increasing payoffs, players behave relatively close to the NE. This is further evidence of methods that enable players to attain above-equilibrium payoffs by taking advantage of this behavioral bias.Some results from the first experimental paper were especially puzzling, and it was conjectured that altruistic components in players' preferences might be distorting the results. In my third paper, using computer bots, I isolate philanthropic components of players' strategies. Adding a proportion of transparent bots that (ex-ante) do not incentive any change in behavior but imply that surplus is wasted if they get any payoff, behavior gets closer to Nash equilibria. The intuition is that when there is some chance players are matched against a bot, altruistic effects (utility gained by total surplus maximizing, even if it is another player who gets it) decrease and behavior moves towards what is predicted by utility maximization. This implies that, as a researcher, one way to verify if altruism is preventing players from reaching the NE is to compare behavior with and without bots that play the NE.
Read
