This dissertation contains three essays, each of which is pertinent to topics in empirical Industrial Organization. The second and third chapters are coauthored with Matthew H. Shapiro and Amil Petrin, respectively. In the first chapter, any empirical evidence presented in this chapter is derived based on data from the Nielsen Company (US), LLC and marketing databases provided by the Kilts Center for Marketing Data Center at The University of Chicago Booth School of Business. I quantify the consumer benefit of introducing one-stop shopping as a new shopping choice and estimates the benefit to firms in turn. Firm scope benefits consumers by allowing them to purchase multiple goods at one location like supercenters, malls, or department stores. I exploit an exogenous change in the allowable firm scope in Washington State, which recently deregulated the retail liquor industry to allow liquor sales in grocery stores. After deregulation, the number of liquor-selling stores is increased fourfold, and 75% of the liquor shopping has been done by one-stop shopping with groceries. Moreover, the liquor quantity sold has increased despite the increased after-tax price of liquor, implying that the choice set of shopping trips has improved. To disentangle the value of one-stop shopping from the value of reduced shopping distance due to more liquor-selling stores, I build a structural demand model of choices for shopping trips. I use household panel and retailer sales data from both before and after deregulation and extend the standard method to allow for endogenous prices to the setting where a store can have two separate qualities of grocery and liquor sections. The estimated consumer benefit of one-stop shopping is $2.52 per trip per household, which is 8% of the household's expenditure on liquor. Selling liquor inside of a grocery store increases its grocery sales by 4.5%, and liquor sales are increased by 30% compared to being sold outside of the grocery store. In the second chapter, we describe a model and estimation strategy to assess the efficacy of several US federal investment projects amounting to $130 million to build out the foundation of an plug-in electric vehicle (PEV) charging network and to encourage purchase of these vehicles. Using a new micro-level data set of electric vehicle purchases in California to estimate a rich discrete choice model of automobile demand, we analyze whether these charging stations have had a significant role in the adoption of electric vehicles in California over the past several years and weigh them against several other policy alternatives. In the last chapter, we suggest an alternative approach to identify demand and supply in discrete choice demand set up. A major contribution of Berry et al. (1995) (BLP) is to show how to transform the market shares in a non-linear discrete choice demand system into product-quality indices that are linear in price, product characteristics, and the demand error, so standard IV techniques can once again be used to estimate demand. They treat price as endogenous but assume that the product characteristics observed by the researcher are uncorrelated with the demand error, that is, the characteristics observed by consumers and producers but not by the researcher. As in Spence (1976) (e.g.) if firms set observed and unobserved characteristics at the same time then this assumption may not hold. We mimic BLP exactly but instead of using their mean independence assumption we learn about demand and supply by assuming firms are maximizing expected profits given their beliefs about preferences, costs, and competitors' actions when they choose observed and "unobserved" product characteristics. We allow firms' information sets at the time they choose characteristics to potentially include other firms' product characteristics, demand and cost shocks, signals on all of these, or no information at all on them in the setting of Hansen (1982). Ex-post firms may wish they had made different decisions and our identification is based on the assumption that firms are correct in their choices on average. Using the same automobile data from BLP, we find that some of the slightly puzzling parameter estimates of BLP go away as all of our parameter estimates are of the correct sign. We find significantly more precise estimates given the same exact data equivalent to approximately a sixteen-fold increase in the number of observations. We strongly reject the standard identification assumption; the conditional correlations of the demand and cost unobservables with observed characteristics equal 0.85 and 0.7 respectively.