Input Price Dispersion Across Buyers

We use a comprehensive dataset of electronic invoices from Chilean firms to document new facts about price dispersion across buyers of the same manufactured intermediate input and assess its implications for aggregate productivity. Price dispersion is pervasive: more than half of manufacturing sales involve products sold to multiple buyers within the same month, with an average interdecile range of prices across buyers of 0.16 log points. The dominant source of this dispersion is match-specific: buyer-seller pair identity absorbs more than 40 percent of the variation in price gaps, with observable buyer-seller characteristics — including size, relationship length, and transaction volume — jointly explaining up to 10 percent. Price gaps vary little with geographic distance or payment method (cash vs. credit), suggesting that they primarily reflect markup differences rather than marginal cost differences. Guided by these facts, we use a production network model calibrated to the Chilean data to quantify the aggregate efficiency gains from eliminating markup dispersion in manufactured intermediate inputs. We find that markup heterogeneity across buyers of the same product accounts for 30 to 64 percent of the total productivity gains from fully equalizing markups across both products and buyers.