A Theoretical and Empirical Analysis of Alternate Auction Policies for Search Advertisements

In the online world, publishers place ads from advertisers adjacent to internet search results for a given keyword. To sell such advertising, web publishers auction multiple ad slots using a generalized second-price auction. In this paper, we compare two auction policies that publishers can use to determine the rank and payments of bidding advertisers. The first policy, the highest bid policy, ranks ads based on the bids submitted while the second policy, the highest profit policy, ranks ads based on the expected profit generated to the publisher. Interestingly, we find that the highest profit policy may generate lower publisher profits per keyword even though it uses more information. Subsequently, we use data from a search engine and empirically establish that the correlation between valuations and click through rates are positive, an important assumption in our theoretical model. This finding provides significant support for the theoretical results.