Nonparametric Estimation of Sponsored Search Auctions and Impact of Ad Quality on Search Revenue

This paper presents an empirical model of sponsored search auctions where advertisers are ranked by bid and ad quality. Our model is developed under the “incomplete information” setting with a general quality scoring rule. We establish nonparametric identification of the advertiser’s valuation and its distribution given observed bids and introduce novel nonparametric estimators. Using Yahoo! search auction data, we estimate value distributions and study the bidding behavior across product categories. We also conduct a counterfactual analysis to evaluate the impact of different quality scoring rules on the auctioneer’s revenue. Product-specific scoring rules can enhance auctioneer revenue by at most 24.3% at the expense of advertiser profit (−28.3%) and consumer welfare (−30.2%). The revenue-maximizing scoring rule depends on market competitiveness. This paper was accepted by Jean-Pierre Dube, marketing. Funding: This work was supported by the Social Sciences and Humanities Research Council of Canada under the Insight Development Grant [Grant 430-2022-00841] to D. Kim. Supplemental Material: The online appendix and data files are available at https://doi.org/10.1287/mnsc.2023.02052 .