Networks, Frictions, and Price Dispersion

This paper studies price dispersion in buyer-seller markets using networks to model frictions, where buyers are linked with a subset of sellers and sellers are linked with a subset of buyers. Our approach allows for indirect competition, where a buyer who is not directly linked with a seller affects the price obtained by that seller. Indirect competition generates the central finding of our paper: price dispersion depends on both the number of links in the network and the structure of the network (how links are distributed). Networks with very few links can have no price dispersion, while networks with many links can still support significant price dispersion. We develop a decomposition of the network that characterizes which links are redundant (i.e. have no effect on prices). We show that a particular network structure (Hamiltonian Cycle) with only two links per node has no price dispersion. We then use a theorem from Frieze (1985) to show that this network structure arises asymptotically with probability one in a randomly drawn network, even as the probability of an individual link goes to zero. We also show the finite sample properties of this relationship and find that even small sparse networks can have very little price dispersion. In an application to eBay, we show that our model reproduces the price dispersion seen in the data quite well, and that 35-45 percent of the price dispersion at eBay can be explained by the network structure alone.