This dissertation consists of three essays. The first, The Price Theory of Two-Sided Markets builds on Rochet and Tirole (2003)'s theory of two-sided markets. In it, I use the inverse hazard rate, or "vulnerability", of demand to separate the effects of interventions into those on the "price level", the sum of prices on the two sides, and the "price balance", how this total price is allocated between sides. I thereby construct a baseline positive price theory of these markets. My normative analysis characterizes socially optimal level and balance of prices and their implications for policy. Intuition for the results is supplied by examples in a tractable "linear vulnerability" class of demand functions.

My second essay, Double Marginalization in One- and Two-Sided Markets builds on the first. I use the vulnerability of demand to show that knowledge of how monopoly pass-through rates compare to unity and whether they increase or decrease in cost is necessary and sufficient for a full pair-wise comparison of mark-ups and profits across firms within and between industrial organizations (Nash, Stackelberg and integrated) in the general Spengler (1950) model of vertical monopolies. This yields new testable predictions and policy implications of the model and facilitates its extension to two-sided markets, speaking to recent debates about the industrial organization of payment cards.

My third essay, A Simple Theory of Scientific Learning is largely unrelated to the first two. It is motivated by the fact that scientists use diverse evidence to learn about the relative validity of various broad theories. Given the lack of statistical structure in this "scientific learning" problem, techniques of model selection and meta-analysis are not directly useful as quantitative guides. In the essay, I use five simplifying assumptions to make the problem tractable by standard statistical methods. Combining Bayesian and frequentist approaches, I derive simple, intuitive rules for updating beliefs. The theory incorporates trade-offs among seemingly incomparable dimensions often used to judge models: ex-ante plausibility, precision, empirical accuracy and general applicability. I establish necessary and sufficient conditions for the consistency of the learning procedure which provide easy robustness checks for applied analysis and a simple algorithm for choosing a robustly consistent trade-off between precision and accuracy. I develop the theory in the context of a motivating application to social preference data collected by Charness and Rabin (2002). In contrast to the authors' analysis, I find (for a wide range of prior beliefs and parameter values) that after taking into account its greater precision. Selfishness is the best model of choice in the simple games they consider.