Inferring the Economics of Store Density from Closures: The Starbucks Case

This paper proposes a method that makes use of firms’ mass store closures to measure the store network effects of cannibalization and density economies. I calculate each store’s contribution to chain-level profits via one-store perturbations on the set of retained stores, and map these onto the firm’s closure choices. To separate the demand- and supply-side store network effects, I exploit the fact that the business-stealing effect intensifies with local network density, whereas the supply-side disadvantage prevails at sparse regions of the network. I apply the method to study the Starbucks chain. The average rate of cannibalization imposed by a neighbor outlet is 1.2% within one mile and 0.4% within one to three miles. For remote outlets, operation costs increase by 0.3% of revenues for each mile of distance from the network. Counterfactual analyses suggest that income level is a more important determinant of demand than population count at low levels of store penetration, whereas high-population regions can sustain denser store networks because of the softening of the cannibalization effect.