Optimal Keyword Auctions with Shadow Costs

Search engines incur shadow costs whenever an advertisement negatively affects user experiences, which damages the search engine’s long-run revenue. We characterize the optimal mechanism for keyword advertising in a general framework that accommodates clickthrough rates and shadow costs that can differ across advertisers and positions. We show that shadow costs significantly affect how advertisers are allocated and the optimal payment rule takes a form of “progressive second price”. Under special cases, the optimal mechanism can be implemented as scoring auctions. Through an analysis of special cases, we obtain insights on how shadow costs impact scoring rules and minimum bid policies and bridge the gap between keyword auction theory and practices.