Endogenous Sunk Costs and the Geographic Distribution of Brand Shares in Consumer Package Goods Industries∗

This paper describes industrial market structure in consumer package goods (CPG) industries using a unique database spanning 31 industries and the 50 largest US metropolitan markets. In these data, the geographic cross-section of markets accounts for a much larger component of a brand’s total share variation than the time-series. Most industries exhibit striking asymmetries across markets in both a given brand’s market share and the industry’s rank-order of shares. Following Sutton’s theory of endogenous sunk costs, a robust set of gametheoretic predictions is generated regarding industrial market structure and the role of brand advertising. For non-advertising intensive industries, the lower bound on concentration falls to zero in larger markets. However, concentration is bounded below in advertising-intensive industries even as market size grows large. Connected to this finding, we observe an escalation in the level of advertising for larger markets. In contrast, the number of advertised brands does not vary with market size and, consistent with the theory, there does not appear to be an escalation in entry. Nevertheless, we do observe a proliferation in the number of non-advertised brands in larger markets. Finally, we collect historic entry dates for two of our industries and find that order of entry has a strong impact on the rank-order of shares in a market. The geographic distribution of entry in these two industries also accounts for spatial clustering, whereby geographically “close” markets exhibit similar market structures. The geographic variation in historic entry accounts for spatial dependence both in a brand’s share and in its advertising intensity. Several alternative explanations for the geographic patterns in our data are ruled out. In general, the consistency between the patterns in our data and endogenous sunk costs theory suggest a role for advertising in the formation of long-run industrial market structure. JEL classification: L11, L66, M30, M37, R12 ∗We are very grateful to Tim Conley, Matt Gentzkow, Jonathan Levin, Sanjay Sood, Raphael Thomadsen and Florian Zettelmeyer for insightful comments and discussions. We also thank seminar participants at Cornell University, Harvard, MIT, New York University, the University of Chicago, the Yale SOM, the 2004 Choice Symposium in Estes Park, Colorado, the 2004 Marketing Camp at Leuven, and the 2005 NBER Winter I.O. meetings for valuable feedback and helpful suggestions. We are also grateful to Ed Lebar of Young & Rubicam Brands for providing us with the BAV data, to Jeff Hermann of Nielsen Media Research for the historical advertising data and to the librarians of the National Museum of American History for the archival data on coffee. We acknowledge the research assistance of Avtar Bhatoey with the collection of the local brand entry data. The authors are also grateful for research support from MSI, grant #4-1301. The second author thanks the Neubauer faculty fund for research support. The third author acknowledges the support of the Kilts Center for Marketing. †UCLA Anderson Graduate School of Management, University of California, Los Angeles ‡Graduate School of Business, University of Chicago