How can we measure Google’s market power? - Comments of Mark Patterson

Posted by Mark R. Patterson (Fordham)

Surprisingly, given the amount of attention that has been given to whether Google has acted anticompetitively in the search market, much less attention has been paid to whether Google has market power. Those who favor antitrust scrutiny of Google generally cite its large market share, from which they infer or assume its dominance. Those who are skeptical of competition law’s role in regulating search, on the other hand, cite Google’s “competition is only a click away” mantra to suggest that Google’s market position is precarious. In fact, the issue of Google’s power is more interesting than either of these approaches suggests.

The Limited Relevance of Market Share

Competition law uses market share as a proxy for power because it often reflects the ability of a firm to act without regard to competition. Where the product is information, however, competing firms may be able to respond quickly and easily. For example, if Google were to act anticompetitively, a competing search engine would easily be able to “produce” products to meet the demand of those who were unhappy with Google’s products. The products at issue are search results, so the only obstacle to a competing search engine producing more of them would be the installation of more server capacity to deliver the results to customers. Although expanding server capacity imposes some costs and takes some time, those limitations are small compared with, say, expansion of capacity in the production of the archetypal widget. Hence, market share is a relatively poor proxy for power when the product at issue is information.

There is another, perhaps more important element related to the volume of search results delivered, though, and that is the advantage a search engine gains from the information gathered from searches. A search engine delivering a larger volume of search results gains valuable information from its users’ searches and thus is likely to deliver better search results. This effect is not, however, so much one of current market share as of the cumulative number of searches delivered, so that it does not support using current market share as a measure of power. It is better treated, perhaps, as exclusive access to a valuable input, in the sense that prior searches are the raw material from which search results are in part derived. Although this advantage may be important, the focus of the comments here is on another possible source of power: the difficulty of evaluating the quality of search results.

Knowing When to “Click Away” from Google

As Google argues, users can easily switch to other search engines. But the ease of clicking to another search site does not mean that Google has no power. For Google to be constrained, it must also be the case that users can determine when it is advantageous to click away. Because search information is provided for free, users will make that determination by comparing search engines on the basis of quality. Users may find it quite difficult to determine whether the quality of the results they are receiving from a particular search engine justify switching. As Kenneth Arrow described, “[information’s] value for the purchaser is not known until he knows the information, but then he has in effect acquired it without cost.” As a result, in many instances of search, a consumer will be seeking information in circumstances in which she will be unable to evaluate the quality of the information she receives.

Although one is sometimes confident before performing a search that its quality will be high, as for example when one searches for a particular institution, like the “Antitrust and Competition Policy Blog,” more often one is searching for something less specific, like “New York hotels.” In that case, one cannot be sure before performing the search, and perhaps even before going to the resulting web pages or even the hotels themselves, whether one has received useful results. Even if a user ran the search on another search engine and obtained different search results, it would not be clear which results were better. In that sense, a search result can be a credence good, a good whose quality is difficult or impossible to assess. This lack of transparency in quality can give an information provider market power, just as can an absence of transparency in price for other products.

Significant Power to Distort Search Results?

The real question, though, particularly for Sherman Act § 2 and Article 102 TFEU, is whether Google has significant power. One possible way to make this assessment would be to consider whether moving a site up or down several places in search results would show significant market power. As is discussed in the fuller version of these comments, the existence and apparent success of search engine optimization are evidence that there is little user response to relatively minor differences in search results. If there is no significant user response to moving a site down several spots in the results, then we can ask if that constitutes a significant lowering of search-result quality, from which we could infer the presence of significant market power.

One way to quantify the loss in quality of moving sites in search results, which is offered tentatively here, is to use the prices paid for placement in Google’s AdWords results. Because the value of search positions in the so-called “organic” results is not quantified, but AdWords positions are, we can use AdWords as a proxy for the organic results. Indeed, if the price difference between positions in the AdWords results differs significantly, one would expect the difference in value between similar positions in the organic results to be comparable, or even greater. Although actual prices for AdWords are not easy to obtain, one can use Google’s own “Traffic Estimator” to estimate some figures. For example, using the keyword phrase “kitchen faucet,” the Traffic Estimator provides the following numbers for different specified maximum costs per click (“CPC”):

maximum CPC (specified)	estimated average CPC	estimated ad position	estimated daily clicks	estimated daily cost
$2.50	$1.26	1.49	404.08	$510.90
$2.00	$1.11	1.71	379.86	$421.63
$1.50	$0.92	2.14	338.75	$312.35
$1.00	$0.70	3.00	264.30	$184.10
$0.50	$0.44	5.83	119.69	$ 52.16

As can be seen in the table, the price difference between ad position 2 and 3 in these estimates is greater than ($0.92 – $0.70) / $0.92 = 23.9%. The difference between positions 1 and 2 is greater than ($1.26 – $0.92) / $1.26 = 27.0%. For this keyword phrase, then, if Google could move a site from position 3 to position 2 or from position 2 to position 1, it would be decreasing the value of its placement by approximately 25%. The ability to lower quality by that amount would, following the U.S. Merger Guidelines test, show the existence of market power and, because 25% is significantly greater than the Guidelines’ 10% threshold, perhaps also monopoly power.

Admittedly, this approach is not unproblematic. Among other things, it uses prices paid by advertisers as a proxy for value to consumers. That is perhaps not entirely unreasonable in a two-sided market like Google’s, in that AdWords advertising serves not only an informational role but a signaling one. An investment in AdWords is an indication that the advertiser expects to receive a return on that investment, and that return will come from consumers’ purchases from that advertiser. Hence, the willingness to spend on advertising is likely related to the value placed by consumers on positioning in search results.

Conclusion

In any case, these comments are offered here not so much as presenting a precise means of assessing market power as an example of the kind of approach that could be used for an product like Google’s. As information products come to constitute a larger proportion of the market, and to be involved in a larger proportion of allegations regarding anticompetitive conduct, it seems likely that competition law will have to develop entirely new techniques for addressing the special problems posed by information.

This post is derived from a working paper available here.