Clearinghouse models of online pricing---such as Varian (1980), Rosenthal (1980), Narasimhan (1988), and Baye-Morgan (2001)---view a price comparison site as an "information clearinghouse" where shoppers and loyals obtain price and product information to make online purchases. These models predict that the responsiveness of a firm's demand to a change in its price depends on the number of sellers and whether the price change results in the firm charging the lowest price in the market. Using a unique firm-level dataset from Kelkoo.com (Yahoo!'s European price comparison site), we examine these predictions by providing estimates of the demand for PDAs. Our results indicate that the number of competing sellers and both the firm's location on the screen and relative ranking in the list of prices are important determinants of an online retailer's demand. We find that an online monopolist faces an elasticity of demand of about -2, while sellers competing against 10 other sellers face an elasticity of about -6. We also find empirical evidence of a discontinuous jump in a firm's demand as its price declines from the second-lowest to the lowest price. Our estimates suggest that about 13% of the consumers at Kelkoo are "shoppers" who purchase from the seller offering the lowest price.

The market values of online platforms, such as Yahoo, stem from their ability to monetize the clicks they generate for firms advertising on their sites. We exploit a unique dataset on clicks from one of Yahoo's price comparison sites to estimate the determinants of clicks received by online retailers. We find that a firm enjoys a 60% jump in its clicks when it offers the lowest price at the site. This discontinuity is consistent with a variety of models that have been used to rationalize the price dispersion observed in online markets. We also show that one may use estimates of the determinants of a firm's clicks to obtain bounds on its underlying demand parameters, including own- and cross-price elasticities. Our results have potentially significant ramifications for online retailers, platforms, and policymakers: Failure to account for discontinuities distorts parameter estimates by 50 to 100 percent.