Where Should We Build a Mall? The Formation of Market Structure and Its Effect on Sales

by Doug J. Chung, Kyoungwon Seo, and Reo Song

Overview — In spite of the recent surge in e-commerce, brick-and-mortar retail, specifically in the form of large-scale shopping malls, is still the dominant venue for consumer purchases in the developed world. The construction of mass-scale shopping malls has also experienced tremendous growth in newly industrialized countries such as China. This research provides a rigorous, yet practical, framework to understand and evaluate why retail stores join a shopping mall and how their decisions affect mall revenue. The model can be extended and applied to a number of settings where a decision maker must choose among alternative sites to construct a market, for example, for transportation hubs such as airports or train stations.

Author Abstract

We estimate a structural model that takes into account the entry decisions of retail stores and their corollary effects on total shopping mall sales. By understanding the endogenous behavior of individual store entry, we provide guidance on location choice for mall developers. We find negative competition effects to dominate within store categories—especially among discount and midscale stores—but positive agglomeration effects to exist across store categories. Although varying by store brand, our results suggest that upscale stores are likely to enter malls in more populated, affluent areas, whereas midscale stores enter less populated, lower-income areas. We find positive causal brand effects for specific upscale and midscale stores, above and beyond market effects, but find negative causal brand effects for all discount stores on mall sales. This paper also introduces three main methodological innovations to the marketing literature. First, we correct for endogeneity with regard to both store entry and mall sales to identify the causal effect of store entry on mall sales. Second, we address multiple equilibria by estimating equilibrium selection from the observed data. Finally, we overcome the computational burden of solving games of complete information with multiple equilibria by utilizing the GPGPU technology, using multiple processing cores in a graphics processing unit to noticeably increase computational speed.