Three essays on product differentiation: computational tools for applied research, evaluating model behavior, and geographic demand

This thesis develops computational and applied tools to study differentiated products. The core of the thesis focuses on Berry, Levinsohn, and Pakes’s [1995] (BLP hereafter) model of differentiated products. First, I examine how polynomial-based methods for multi-dimensional numerical integration improve the performance of the model. Unlike Monte Carlo integration, these rules produce reliable point estimates and standard errors as well as increasing the accuracy and execution speed of the estimation software. Next, I conduct a large scale simulation study to investigate both the asymptotic and finite sample behavior of the BLP model using the traditional instruments formed from characteristics of rival goods and also supply-side cost shifters, which are necessary for asymptotic identification. The final part of the thesis evaluates the 2003 merger of Morrisons and Safeway by combining a discrete/continuous choice model of demand with census data to construct a geographic distribution of demand. I use this distribution to model the interaction between the location of consumers and stores, focusing on the welfare implications of the merger