To Score or Not to Score? Structural Estimation of Sponsored Search Auctions

We estimate a structural model of a sponsored search auction model. The econometric model matches two pieces of data for each auction: (i) the allocation of sponsored search listing to the bidders; and (ii) the per-click prices paid by the winners. To accommodate the “position paradox”, our model relaxes the assumption of decreasing click volumes with position ranks, which is often assumed in the literature. We estimate the model using data from the Chinese online marketplace Taobao.com. We find that sellers of different qualities adopt different bidding strategies: high quality sellers bid more aggressively for informative keywords, while low quality sellers are more likely to be sorted to the top positions for vague keywords. Counterfactual evaluations show that score weighting in sponsored search auctions would dramatically increase the per-click price for the top ad position, but leave prices for lower positions largely unaffected.