The Geographic Concentration of Retail Stores: Trends and Determinants

This dissertation mainly assesses the relative importance of two mechanisms–consumer searching and cost sharing for the spatial concentration of retail stores. I first present a simple model of firms’ location choice and consumers’ shopping destination choice. This model shows how both mechanisms promote the agglomeration of retail firms. The degree of retail agglomeration, measured by Ripley’s K function, increases with the importance of consumer searching and with the benefits from scale economy in intermediate inputs such as advertisements and delivery services. I then present stylized facts based on industry agglomeration. Using the 2012 National Establishment Time-Series database, I calculate Ripley’s K function and Ellison-Glaeser index that measure the degree of agglomeration for each CBSA-NAICS industry-year pair in the U.S. from 1992 to 2012. The estimated Ripley’s K values show that the overall degree of retail agglomeration has been decreasing since 1992. Also, I find much inconsistency between the estimated Ripley’s K and Eillson-Glaeser indices. Using the variation in the percentage of population that use Internet across regions and in the Internet sales share across retailing industries, I find that a 10 percentage points increases in the Internet use percentage and in the Internet sales share is associated with a 3.1-3.5 decrease and a 9.5-10.8 decrease in the calculated Ripley’s K value. This provides suggestive evidence of the role consumer searching plays in promoting retail agglomeration since Internet decreases the importance of consumer searching in person. A difference-in-differences analysis shows that industries easy to make comparison shopping online experience a greater decrease in the degree of agglomeration by 6.8-12.6 before and after 2000. Last, I use a triple-difference approach based on the bankruptcy of chain stores like Borders and Circuit City to distinguish the effects of consumer searching and cost sharing. The estimation results for various types of chain stores using sales, employment size, the probability of closure, and the birth of new stores as dependent variables are mixed. Consumer searching alone cannot well explain retail agglomeration; however, there is little evidence of the contribution of the cost sharing mechanism to retail agglomeration