Feedback Effects in Search Advertising

Like many other media services, internet search services provide two-sided platforms that coordinate interactions between media consumers and media advertisers. As with other media services, especially newspapers, the possibility has been raised that positive feedback between the consumer and advertising sides of search platforms (more consumer searchers will attract more advertisers, which will attract more consumers that attract even more advertisers, and so on) may amplify an initial advantage in either consumer usage or paid ads with the result that an initially advantaged search engine will capture a dominant share of a search market. To date, however, there has been no empirical research to establish whether two-way feedback between the consumer and paid advertising sides of a search platform exists, whether it is positive or negative if it does exist, and whether such feedback is of sufficient magnitude to significantly influence concentration in search markets. To our knowledge, this paper is the first to present evidence based on a rigorous empirical study for the existence, magnitude and directionality of feedback effects in search markets. Because the consumer/searcher and advertiser sides of a search service may influence each other, the two sides of a search service cannot be studied separately if one wants to fully understand search market dynamics. We employed three-stage least squares to estimate a simultaneous equations model that allowed for ads-consumer usage feedback to estimate a simultaneous equations model using 2007 data reported by Yahoo for over 70 of its local metropolitan markets in the United States. Data reported by Yahoo on clicks on paid search ads for four locally-supplied products were used as a measure of consumer usage and our counts of numbers of paid ads on the search pages elicited by keywords corresponding to the four local products studied were employed as the measure of paid ad sales. Use of metropolitan markets data for a single search service was necessary because, while their services are very similar, the different major independent search services in the U.S. (Google, Yahoo and Microsoft at the time our data was collected) differed sufficiently in the types of data and geographic options they offered advertisers that econometrically sound direct comparisons were not feasible. Use of local markets within a single large country as a source of cross-sectional variation was also deemed preferable to constructing an international cross-section because it would have been difficult to adequately control for country-specific differences in culture, internet use, and regulations affecting the larger advertising market. Our three-stage-least squares estimates provide statistically significant evidence for the existence of positive and quite large two-way feedback between the consumer and advertiser sides of Yahoo’s local internet search service. We use estimates for the model’s parameters to show how feedback between the consumer and advertiser sides of a search service increases the financial return to resources invested in promotional efforts to increase the number of consumers using Yahoo’s search service. Promotion through other media and investments to improve search accuracy are examples of such investments. We also discuss the importance of positive two-way feedback as a mechanism that may contribute to the observed pattern of high concentration in national internet search markets. The empirical findings presented in this paper contribute toward a more complete understanding of search service economics and thereby to the development of policies that might better address the challenges raised by highly concentrated search markets