Estimating network economies in retail chains: a revealed preference approach

Abstract

We measure the effects of chain economies, business stealing, and heterogeneous firms' comparative advantages in the discount retail industry. Traditional entry models are ill suited for this high-dimensional problem of strategic interaction. Building upon recently developed profit inequality techniques, our model admits any number of potential rivals and stores per location, an endogenous distribution network, and unobserved (to the econometrician) location attributes that may cause firms to cluster their stores. In an application, we find that Wal-Mart benefits most from local chain economies, whereas Target shows a greater ability to respond to rival competition. Kmart exhibits neither of these strengths. We explore these results with counterfactual simulations highlighting these offsetting effects and find that local chain economies play an important role in securing Wal-Mart's industry leader status.169 170 / THE RAND JOURNAL OF ECONOMICS specifically, Wal-Mart, Kmart, and Target now make up three quarters of total sector sales and show few signs of slowing down. The chain store advantage is both national and local: chains take advantage of scale economies in purchasing and logistics, as well as more local "density" economies that arise from sharing local knowledge and pooling advertising resources. Despite the increasing dominance of chain superstores, there is relatively little empirical work quantifying the size and sources of the chain advantage. 1 This is the goal of our article.In part, the lack of empirical research reflects the scale and complexity of the chain store's decision problem: building a massive network of outlets and distribution centers that move products to consumers as efficiently as possible, recognizing that their rivals are doing the same. Until recently, a complex strategic game on such a high-dimensional space was thought impossible to solve, much less estimate. 2 Furthermore, many retail sectors are dominated by as few as two or three firms, begging the question of how one can recover rich systems of parameters with the observed behavior of so few players. Building on the profit inequalities approach developed by Pakes et al. (2011) and Bajari, Benkard, andLevin (2007), we propose a scalable empirical framework for analyzing network entry problems that does not require solving for an equilibrium, or even that the equilibrium be unique or played in pure strategies. Rather, it simply requires enforcing the equilibrium condition that the observed choices yield weakly higher expected payoffs than any feasible alternatives, holding rivals' actions fixed. We apply our estimator to the discount retail sector, isolating the comparative advantages of each of the three players and quantifying the relative importance of the local chain effect. Moreover, given the relative simplicity of our estimation approach, this framework could easily be applied in other multistore or multiproduct settings such as airline hub and spoke networks, automobile product line decisions, or retailer assortment pr...

The economic advantages of chain organization

“…A growing literature addresses the spread of chains. In part this reflects the success of large retailers such as Walmart (Basker, ; Foster, Haltiwanger, and Krizan, ; Holmes, ; Jia, ; Ellickson, Houghton, and Timmins, ; Foster et al., ; Zheng, ). This literature also takes advantage of newly developed methods to estimate structural models of competition from entry and location decisions .…”

Impact of Competition on Product Decisions: Movie Choices of Exhibitors

“…Interpreted at face value, this finding may seem surprising since co-owned theaters maximize joint profits and lack the incentives to steal business from one another to the extent that rival theaters do. Because of incumbents' additional incentives to compete when faced with a rival firm, the cannibalization effect is expected to be, and is often found to be, smaller than the business stealing effect (Ellickson et al 2013). Traditionally, the difference between the estimated business stealing effect and the cannibalization effect is used to infer the impact of competitive incentives on conduct (Kalnins 2004, Davis 2006b, Wilson 2015.…”

Spillovers from entry: the impact of bank branch network expansion

“…Methodologically this article is related to the literature on estimating entry models across interdependent markets (Jia, ; Holmes, ; Ellickson, Houghton, and Timmins, ). Jia () uses supermodularity to find the equilibrium of a two‐player entry game between Wal‐Mart and Kmart, where entry decisions in different markets are not independent.…”

“…He also uses a moment inequalities approach, and although his has the advantage of incorporating dynamics, it is a single‐agent model that abstracts from competition effects. Ellickson, Houghton, and Timmins () study chain advantages in a three‐firm model. Their approach also uses profit inequalities based on the necessary condition and controls for unobserved market heterogeneity.…”