| Classical economic models of individual and firm behavior provide valuable insights into how rational agents make decisions. However, highly generalized models, by design, omit factors that may be relevant to particular markets. Such omissions may lead to inaccurate predictions of behavior in the real world. In this dissertation, I examine three distinct settings in which the omission of relevant beliefs, preferences, and outside options may prevent traditional models from fully capturing rational decision making.;In Chapter 1, I examine why the logic of backward induction and iterated elimination of dominated strategies fail as predictors of behavior in the centipede game and similar games. I elicit players' beliefs about other players' strategies in a normal form centipede game to gain insight into players' decision making process. The experimental evidence reveals that players have heterogeneous starting beliefs about how others will play the game, which they gradually update based on past experience, leading to changes in players' chosen actions over multiple plays of the game. Players take past experience as a signal of how future opponents will play the game but anticipate that other participants may also change how they play the game as the experimental session progresses. Beliefs better predict changes in player choices than players' past experience, alone, and the pattern of belief evolution is in keeping with the convergence toward the classical theoretic prediction observed in other studies. Additionally, individuals correctly anticipate that a population of players begins the game by playing cooperatively but gradually transitions to a strategy of expected payoff maximization. These findings suggest that players' starting beliefs about others and players' experience in playing the game serve an important role in determining the evolution of strategies over multiple plays of the game.;In Chapter 2, I demonstrate that buyers engage in boycotting and "boycotting" in order to respectively punish and reward a seller's policies of which they are neither the victim nor the beneficiary. Notably, the experimental results show that third-party punishment and reward occurs even when the mechanism for executing such punishment and reward is not straightforward and when no punishment (reward) may be viewed as the neutral or default choice. This paper also finds that the average buyer does not condition her boycotting and buycotting decisions on the actions of other buyers in this setting.;In Chapter 3, I explore a setting featuring two simultaneous, private value eBay-style auctions for heterogeneous goods, in order to explain the prevalence of multiple, unique bidding strategies in eBay auctions. I find that multiple broad strategies of bidding may be supported as equilibrium strategies when goods are horizontally differentiated. I also demonstrate that eBay-style auctions may yield greater social efficiency than second-price sealed-bid auctions, under certain assumptions. Furthermore, I demonstrate that, when bidding costs are sufficiently high, the eBay auction mechanism may not yield efficient outcomes for either horizontally or vertically differentiated goods. |
