Working Papers, The Johns Hopkins University, Department of Economics 594
Spatial competition among multi-store firms is ubiquitous in a wide range of retail industries. However, little is known about how those firms optimize their networks of stores after a merger due to the computational burden of solving for an equilibrium in store networks. This paper proposes an empirical framework for estimating a game of network choice by two multi-store firms, which allows us to examine the impact of a hypothetical merger on store configurations, costs, and profits. The model explicitly incorporates a fundamental determinant of location choice for multi-store firms: the trade-off between the business-stealing effect and the cost-saving effect from clustering their own stores. The method integrates the static entry game of complete information with post-entry outcome data while using simulations to correct for the selection of entrants. I use lattice-theoretical results to deal with the huge number of possible network choices. Using unique cross-sectional data on store networks and revenues from the convenience-store industry in the Okinawa Island, Japan, I estimate the firms. revenue and cost functions. Parameter estimates suggest a retailer's trade-off between cost savings and lost revenues from clustering its stores is positive across markets and negative within a market. I find an acquirer of a hypothetical horizontal merger of two multi-store firms would decrease its number of stores in suburbs but increase its number in the city center, affecting consumers in different locations differently. The trade-off from clustering plays a central role in explaining this result.