| The thesis consists of three essays that cover major research topics in the field of Industrial Organization.;Chapter 1, "Diversification to Capture R&D Spillovers. Do Conglomerates Really Destroy Value?", explores the idea that conglomerate mergers have an economic rationale grounded in the attempt to capture R&D spillovers. In the dynamic general equilibrium model presented in the chapter, buyers are endogenously matched with target firms. An implication of the model is that the best managers will be the first to buy and the worst managers will be the first to sell. As time passes the quality of the buyers falls and the quality of the sellers rises. The model generates a merger wave and predicts that buyers would be willing to pay a premium in order to merge early. This premium will fall along the wave.;In Chapter 2, "Estimation of Auctions with Incomplete Bidding Data", we assess the validity of nonparametric estimators in auction models where only a subset of the bids is observable. A finding of this chapter is that, even though nonparametric identification holds, the corresponding estimation problem is irregular in the sense that the inverse of the mapping from the parent distribution to that of the observable bids is not Lipschitz continuous. As a consequence, in many relevant scenarios it is not possible to achieve the familiar root-n rate. A finding of this chapter is that the problem becomes more severe as the number of bidders increases. The chapter attempts to describe implications for functionals of practical interest of the parental distribution, such as the expected revenue of the auction and the optimal reserve price, and to propose regularizations that yields estimators that are asymptotically Gaussian uniformly in valuations at an adaptive pointwise rate.;Chapter 3, "Estimating Auction Models From Transaction Prices with Extreme Value Theory", I expand on the previous chapter, proposing a simple alternative parametric approach to estimate functionals of practical interests in environments where only one of the bid is observed. This method is allowed by extreme value theory and doesn't require strong assumptions on the unknown parental distribution. The theory predicts that the approximation becomes tighter when the number of bidders increases. I present results from Montecarlo simulations and compare them to results obtained through standard nonparametric estimators: I shows that the approximation allowed by extreme value theory produces better results, especially when the number of bidders is high. |
