Soft Reserves, Learning and Bid Adjustment in Advertising Auctions
An advertiser in a second-price auction sees what it paid, not the rival bids that supposedly set the price. Akbarpour and Li (2020) show that an auctioneer can therefore charge a winner up to its bid undetected. The Federal Trade Commission alleges that Amazon’s advertising auctions, described as second-price, used soft reserves that did this. We ask whether repeated bidding exposes such charges and what adjustment does to welfare. If each charge equal to the bid is flagged, the soft reserve is exposed quickly, however much the advertiser trusted the description. If reports show only win rates and expected charges, it is indistinguishable from a second-price auction at any fixed bid: an advertiser that trusts the description and bids its value never learns, and one that doubts it but bids for current profit needs a number of auctions at least inversely proportional to its doubt before it considers a soft reserve more likely than not. The second-price description and aggregate reporting are complements. Among informed, symmetric advertisers, the soft reserve raises no more revenue than a second-price auction, and when only some advertisers adjust, placements are misallocated. Disclosure returns surplus to advertisers; removing the surcharge also restores efficiency.
-
-
Copy CitationJoshua S. Gans, "Soft Reserves, Learning and Bid Adjustment in Advertising Auctions," NBER Working Paper 35841 (2026), https://doi.org/10.3386/w35841.Download Citation