Contests for Revenue Share∗

We analyze a model where budget-constrained buyers need to partition their budgets among multiple capacity-constrained sellers. Each seller designs his allocation rule endogenously, as part of a two-stage game: First, sellers simultaneously design allocation rules. Second, given these rules, buyers simultaneously decide how to partition their budgets among the different sellers. The goal of each seller is to obtain the largest possible revenue share, i.e., to maximize the sum of budgets she receives. Our main interest is in understanding what types of allocation rules maximize their designer’s revenue share. This setting is related to literature on capacity-constrained competition and the Bertrand paradox, search with costs, and the recently expanding literature on antitrust economics, on top of the obvious connection to the two frameworks of contest design and of Blotto games. More details are given below. Our particular interest in this setting stems from observations regarding the market for Internet search advertising. This is now a multi-billion dollar market, where thousands of advertisers post hundreds of millions of ads every day. Quite interestingly, market shares differ significantly from revenue shares in this market.1 In 2012, for example, Google had a market share of about 65% and a revenue share of about 75%, while Microsoft (Bing) had a market share of about 20% but a revenue share of less than 10% which is less than half of its market share.2 There could be a variety of reasons for differences in revenue shares. For example, it might be that the “items” are valued differently, and this can be checked empirically.3 Another reason could be that sellers can somehow leverage their market share to obtain a disproportionally higher revenue share. In this paper we focus on this latter hypothesis, and study the connection between market shares and revenue shares in cases where all items are in fact identical. ∗The full paper can be obtained from http://ie.technion.ac.il/ ̃ ronlavi/papers/FLT-new-version.pdf †Weizmann Institute of Science. Email: [email protected]. ‡Technion – Israel Institute of Technology. Email: [email protected]. §Technion – Israel Institute of Technology. Email: [email protected]. The “items” in this market are ad slots on users’ search results pages. The “market share” of a certain search engine is its number of available ad slots divided by the total number of available ad slots of all search engines. Its “revenue share” is the budget spent on ads in this search engine relative to the budgets spent on all search ads. In other industries, what we refer to as a market share is also commonly termed unit market share or volume market share, and what we refer to as revenue share is also commonly termed dollar market share or value market share. Source: eMarketer, June 2013. Search advertising revenue share of leading search sites in the United States from 2011 to 2015. Available from http://www.statista.com/statistics/255863/search-ad-revenue-share-at-leading-searchsites-in-the-us/. For example, an ad slot for a search “pizza in NYC” on Google might provide a different value than the same ad slot for the same search on Bing. We are not aware of any empirical investigation of this.