Estimating a Model of Strategic Store-Network Choice

Competition among multi-store chains is common in retail industries.This paper proposes a method for eliminating a model of strategicstore-network choices by two chains. In contrast to previous studies, Iallow chains to not only choose which markets to enter but also how manystores to open in each of those markets. I use lattice-theoreticalresults to deal with the huge number of possible network choices. I showthat a chain's net trade-off between costs and benefits from clusteringtheir stores in a market can be either positive or negative while stillenduring the existence of an equilibrium. By doing so, the modelprovides a way to freely estimate this within-market effect from thedata. Incorporating revenue data allows us to interpret parameters inmonetary units and to decompose the within-market effect into costsavings from clustering stores (economics of density) and lost revenuesfrom competition with one's own stores (own-chain business-stealingeffect). I apply the technique to a new data set from theconvenience-store industry in Okinawa, Japan. Parameter estimatesconfirm that own chain business-stealing is an important considerationfor a chain. I then use the estimated structural model to perform twocounterfactual analyses. First, I consider a hypothetical merger of twochains and find that the merger would decrease the number of stores andtotal sales, and raise the acquirer's profits thereby reallocatingsurplus from consumers to the acquirer. Second, I examine howeliminating the zoning regulation introduced in Japan in 1968, which hasbeen at the forefront of urban policy debates, affects store-network choices.