Three Essays in Industrial Organization

This thesis includes three chapters studying topics related to industrial organization. In the first two chapters, I study auctions with discrete bidding increments and bidding costs (henceforth pay-per-bid auctions). In chapter one, I develop a general theoretical framework for pay-per-bid auctions. In this model, bidders have private valuations and arrive exogenously throughout the duration of the auction. Participants are uncertain about the number of other bidders in the auction and I allow the bidders to have non-equilibrium beliefs about this state variable. I then propose a simulated method of moments estimator which I use to recover the underlying structural parameters even when bidders have biased beliefs. Using data from an online auction website called Swoopo, I show that the revenues we observe cannot be generated by a model where bidders have rational expectations but is consistent with a model where bidders have biased beliefs. In chapter two of the thesis, I show that while pay-per-bid auctions generate excess revenues, there is also significant variation in these revenues and markups across different products and categories. Using the theoretical framework and methodology developed in the first chapter, I separately estimate the structural parameters for different products. Finally, given these estimates, I use counterfactual analysis to decompose the sources of markup heterogeneity in these auctions In chapter three of the paper, I study the identification and estimation of dynamic models of market entry when firms can choose to buy or lease fixed inputs. I am particularly interested in the identification of counterfactuals which change the stochastic process of the state variables. I first show that without any additional restrictions, the structural parameters of the model are not point identified. Second, I propose plausible exclusion restrictions that provide point identification of the counterfactuals of interest. Finally, I propose a simple simulation based method to construct confidence regions for counterfactual choice probabilities when they are not point identified. I show the validity of the method using Monte Carlo experiments.Ph.D