| Empirical and theoretical topics in applied microeconomics are discussed in this dissertation. The first essay identifies and measures managerial advantages from access to high-quality deals in venture capital investments. The underlying social network of Harvard Business School MBA venture capitalists and entrepreneurs is used to proxy availability of deal access. Random section assignment of HBS MBA graduates provides a key exogenous variation for identification. Being socially connected to peer venture capital firms and private equity seeking startups leads to more deal flow, larger asset under management and better performance in the inaugural funds of HBS-executive run venture capital firms. The second essay presents a two-stage model of competing ad auctions. Search engines attract users via Cournot-style competition. Meanwhile, each advertiser must pay a participation cost to use each ad platform. Advertiser entry strategies using symmetric Bayes-Nash equilibrium that lead to the Vickrey-Clarke-Groves outcome of the ad auctions are derived. Consistent with the model predictions, empirical evidence shows that multi-homing advertisers are larger than single-homing advertisers. Comparative statics on consumer choice parameters, quality, and user welfare are used to analyze the prospect of joining auctions to mitigate participation costs. The analysis provides conditions when such joins do and do not increase welfare. The third essay develops and computes a dynamic model of search in internet advertising. Micro-level browsing data from Microsoft's Bing.com (formerly known as Live.com) is used for structural estimations. The model predicts that users do not click on any ad with weak signals due to accumulating search cost and monotonicity of the value function. Rational search reveals a cascading pattern: the user clicks on a sufficiently high, highest-signal ad first, then moves on to the ad with the next highest conditionally expected probability of match once his assessment on the current ad degrades over time. The user exits when maximum assessment of likelihood of match over all ads is below a threshold value. The essay provides a novel approach to understanding rational herding behavior when product quality is only partially unraveled. |
