Partial identification via inequality restrictions: Inference and applications in industrial organization

Applied economists have at their disposal a vast tool kit of econometric methods to conduct empirical research. Until recently, however, empirical research has relied almost entirely on econometric models in which parameters are point identified. The recent literature on partial identification enables researchers to conduct empirical work when their economic models do not yield point identification. This dissertation develops a new way to perform inference on partially identified parameters in econometric models that are comprised of moment inequalities, and uses such models to address questions in the empirical study of oligopoly markets.; Chapter 1 develops a new method to perform statistical inference in econometric models with parameters that are partially identified by moment inequalities. The conventional method for conducting inference is to compute standard errors and construct confidence sets for the parameters of interest. The theorems that justify these techniques typically pre-suppose point identification. The inferential methodology of this chapter allows researchers to perform inference despite a possible lack of point identification.; Chapter 2 examines identification and estimation of the parameters of an oligopoly model, without relying on a potentially fallible equilibrium assumption. Rather, I consider inference on model parameters when the researcher doesn't know precisely what decision rules firms use, but is willing to consider a set of possibilities consistent with many different equilibrium models of firm behavior. The identifying power of the imposed restrictions is discussed, and a strategy for the estimation of model parameters is given. Using the inferential methodology of the first chapter, estimation and inference is performed on the parameters of such a model with data on a nineteenth century railway cartel.; Chapter 3, coauthored with Aviv Nevo, considers identification of the parameters of a linear model with an endogenous regressor. Standard practice is to use an instrumental variable that satisfies two assumptions: relevance and exogeneity. The latter assumption has often been criticized in empirical work, particularly in the estimation of the demand for differentiated products. This chapter investigates the identifying power of relaxing the usual instrument exogeneity restriction in a way that is consistent with these criticisms.