The paper, which can be found here, seeks to describe the precise nature of the various anticompetitive claims against Google, and to develop an economic framework and empirical test to assess these claims. In particular, the paper seeks to develop a conceptual framework to assess claims of leveraging monopoly power and foreclosure of vertical search competitors that could be empirically tested and applied in other jurisdictions or future investigations in platform-settings with related allegations.
The paper is structured as follows:
Section II describes the precise nature of the antitrust claims against Google.
Google operates a multisided platform that offers users free access to its content. In turn, it sells access to those users to advertisers who wish to convert those users to purchasers. Advertising platforms must compete for both users and advertisers. These platforms attract users through quality content, more relevant and less intrusive ads, and lower prices (typically, at free or highly subsidised prices). Ad platforms attract advertisers by having users, offering better targeting and conversion rates, and lower prices to advertise.
In 2007, Google introduced the concept of “universal search.” Universal search blends various “vertical” or specialised search results, that is, results within a narrow category (for example, shopping, finance, local, images, news), with its customary “horizontal” search results, that is, plain blue links, to create a composite results page. The content of specialised results, such as shopping options, can overlap with content provided by comparison shopping sites. Google also prominently features search ads for some results. With universal search, specialised results were placed on similar footing with horizontal search results, and ranked for relevancy within the page.
Importantly, Google’s specialised results are often populated with its own content. Consequently, Google started to compete, in terms of content overlaps, with third-party websites, when previously the relationship between Google and other websites was more complementary.
Section III develops an economic framework to consider the antitrust claims against Google.
A key question is whether Google has substantial market power in search advertising. This assessment is not straightforward and is a highly fact and data intensive endeavour. In any event, three important considerations potentially mitigate the extent of Google’s market power in search advertising. Firstly, to what extent do specialised, or vertical, search engines compete with Google for users? This is an empirical question. Google and a specialised search engine may overlap in some content areas but not in others. However, if users perceive specialised and general search to be substitutes, then, in turn, advertisers will also consider the two types of search platforms to be substitutes. Secondly, do a large number of search users multihome on substitute general search engine platforms? There are indications that multihoming is common, which would further mitigate Google’s market power over users and advertisers. Thirdly, there is empirical evidence that Internet advertising and offline advertising are substitutes.
A second question is whether Google’s conduct is anticompetitive. The core antitrust allegation against Google’s search engine is that it engages in search bias—that is, Google prominently and undeservingly displays its own specialised search services to the detriment of not only rival specialised search engines but also to competition and consumers. The idea is that rival sites are experiencing a drop in user traffic that affects their ability to compete. With weakened rivals, Google extends its alleged market power, which leaves search advertisers with fewer alternatives to reach users.
Economists will recognise the allegation as one of partial foreclosure and raising rivals’ costs. At the heart of the search bias theory is Google’s use of vertical search results. Given that virtually all other general search engines use specialised search results, its use – in and of itself – cannot be a cause for anticompetitive concern. Rather, the widespread adoption of specialised search services suggests that it is a procompetitive and welfare enhancing innovation. The question is whether there are conditions when the use of specialised results by a search engine with significant market power.
Section IV suggests some comparative metrics to help discern whether Google’s conduct is pro or anticompetitive.
The introduction of universal search blending horizontal and vertical search results could have had two potential effects. First, it either increased or decreased search quality. Second, unless the specialised result is located at the bottom of the page, it will also push other links down the page, some of which might include competing vertical sites. These two effects are said to have changed the amount of user traffic that Google provides to competing vertical search sites. At the heart of the antitrust claim is the notion that Google not only competes, to a degree, with vertical search sites for both users and advertisers but is a significant source of traffic to these sites as well. Thus, the question is whether Google is effectively foreclosing vertical competitors or raising rivals’ costs in a manner that also results in harm to competition.
In order to determine the effects of a switch to universal search, three key questions must be asked:
(1) How dependent are vertical competitors on Google’s traffic? While the answer is likely to vary depending on the specific vertical site, what is relevant for antitrust purposes is how dependent the entire vertical search category, or market that Google is allegedly foreclosing, is on Google’s traffic.
(2) How does universal search affect that traffic? It is not clear, ex ante, whether the loss in traffic from a lower position is greater than the gain in traffic from the total traffic effect of a change to universal search, given that more overall pages are being served by Google that contain these vertical links. In other words, to properly measure the impact of adopting universal search or a similar change on vertical sites, it is critical to use the right counterfactual, which is traffic relative to no-universal search. The wrong counterfactual compares whether universal search results in less traffic relative to a hypothetical change to universal search that more prominently features the content of competing vertical sites.
(3) Are users benefiting from the design change? This is, potentially, the most challenging variable to measure. One approach is to measure ad-click rates. These rates should decline if user quality has increased. The intuition is that all links on a search results page are substitutes, and if the use of a specific Google search result provides a more relevant search result page for users, then this will tend to draw clicks away from the other “options” available on the page. In sum, if users are benefiting from the design change, then, even if there is a net loss of traffic to vertical competitors within a specific search category, we cannot simply conclude that the practice has resulted in anticompetitive harm.
This is a valiant attempt at developing a methodological framework applicable to platforms that favour their own services. I particularly enjoyed the economic framing in terms of raising rival costs and foreclosure.
While I am not knowledgeable enough to comment on the rest of the analysis, I will point out two elements that struck me as having the potential to lead an analysis pursued according to this methodology astray.
Firstly, I do not know of any case where the switch to universal search was problematic. Instead, the basis of the European case was that, in the context of universal search, Google was favouring its own products. As such, the point the author makes about the counterfactual seems to be misplaced – as the counterfactual will need to be established by reference to the purportedly anticompetitive act, i.e. the favouring of one’s own service.
Secondly, and relatedly, the author seems to consider that all the results arising from a Google search are perfectly substitutable. However, the placement of the results may well have significant impact in ad click-rates. As such, the proposed metric for determining whether users are benefitting from a design change towards universal search strikes me as problematic.