Position Competition in Sponsored Search Advertising



We model how advertisers compete for advertising positions in sponsored search advertising. Similar to location competition in economics and marketing, one of the biggest challenges in estimating these models is the existence of multiple equilibria. Instead of imposing assumptions to fully describe and estimate the position competition among advertisers, we allow the existence of multiple equilibria without requiring researchers to take stance on how an equilibrium is selected. The basic idea of this approach is to infer model parameters from incomplete econometric models, that is, competitive interactions of players in a game are not fully specified in models. Instead, we adopt inequality conditions to avoid imposing restrictive assumptions of equilibrium generating processes in the advertiser competition. The proposed model is general, and our estimation results are quite robust to various behavioral specifications and assumptions about equilibrium-generating processes in the advertiser competition.
We use the database with bids and clicks of the sponsored search advertising in a particular brand-name keyword provided by a leading search engine site in Korea. The dataset contains individual-IP-level impressions and clicks on the complete set of competing advertisers in the sponsored search list. The search engine offers multiple advertising positions to potential advertisers; each advertising position is sold independently through daily auctions. Similar to online consumer auctions (e.g., eBay), each position auction is designed with a buy-it-now (BIN) price. Our results show that the existence of multiple equilibria is very common; indeed unique equilibrium rarely exists in any time of our data period. Based on the mean profit per click for each advertiser, we quantify the impact of BIN prices on the expected profit of the search engine. We also conduct some “what-if” experiments to investigate the optimality of the pricing policy of the search engine.
Contact information:
Dr. S. Puntoni