How does using a third party's trademark as a keyword to trigger Google ads affect consumer behavior? In Trademarks, Triggers, and Online Search (forthcoming in the Journal of Empirical Legal Studies), Stefan Bechtold (ETH Zurich) and Catherine Tucker (MIT Sloan), present an amazing dataset that can shed light on this question due to a fortuitous natural experiment in European trademark law. For anyone who is interested in trademark policy but who hasn't seen this, it is worth a download. In short, they find that after Google changed its policy to allow companies to select third-party trademarks as advertising keywords, it had no measurable average effect on how likely searchers were to visit the trademark owner's website, but it had a noticeable effect on the behavior of different kinds of online searchers. The authors attempt to distinguish "navigational" searchers who are likely looking for the trademark owner's website, with searches like [ iPhone ] or [ Barbie ], from "non-navigational" searchers who are using the trademark in some other way, with searches like [ iPhone battery repair Paris ] or [ Barbie working conditions workers China ] (where non-navigational searches make up about 80% of all searches). Bechtold and Tucker find that after the policy change, navigational searchers are less likely to visit the trademark owner's website, while non-navigational searchers are more likely to visit.
What is the effect of the AdWords change on trademark owners? If one had described the setup of this experiment to trademark owners, I suspect many would have predicted a significant diversion of users from their websites, so the absence of any measurable net effect is itself an important finding. The trademark owners might still feel concerned about the navigational searchers who are being diverted, but the larger group of non-navigational searchers ended up spending more time on the trademark owner's site, so it would be hard to argue that there was a net harm. And click-throughs to the trademark owner's site are only a very rough proxy for maximizing profits, which is presumably what the trademark owners ultimately care about. The lost navigational searchers may have been unlikely to buy from the trademark owner anyway, or may have bought the trademarked products from authorized distributors. Of course, even if allowing third-party trademarks as advertising keywords did harm
the trademark owner, that doesn't mean trademark law should discourage
this use: trademark law aims to increase welfare by, among other things,
reducing consumer search costs, not to maximize rents for trademark
What is the effect on search engine users? As the authors note, citing Eric Goldman, "an Internet search for a particular brand does not necessarily mean that the consumer is only or even at all interested in products sold under this brand." For example, someone searching for [ iPhone ] may simply be interested in a smartphone, and may end up discovering and preferring a Samsung phone. Or she may end up buying an iPhone from an independent distributor rather than directly from Apple. But are some consumers harmed? The authors suggest that "users who search for a trademark because they are directly interested in visiting the trademark owner's website may fare better under a strict keyword advertising policy which does not allow third-party use of trademarked keywords," but that only seems true if there are users who enter [ iPhone ] with the sole goal of reaching Apple's website, and they end up getting diverted to someplace they didn't mean to be. I don't think there is evidence in this dataset or elsewhere that this is a significant problem; indeed, the Franklyn & Hyman article the authors cite provides some evidence that a number of consumers who search using a trademark are open to purchasing competing products. As the authors note, they cannot measure whether "a searcher finds
the information she is actually looking for, since this would require
stated intent data [they] do not have access to."
Why would non-navigational searchers be more likely to visit the trademark owner's website after the policy change? This finding is the most puzzling thing about this paper to me. I'd be interested in knowing more about both what results are presented and what the searchers do—much of which unfortunately seems beyond the scope of the dataset. On what results are presented, I'd like to know whether there were changes in how often the trademark owner's site shows up in both the organic results (perhaps due to a change in Google's algorithm) and the paid results (perhaps due to trademark owners buying more ads for their own trademarks). On what searchers do, I'd like to know how often searchers click on organic vs. paid results, and where they go when they don't go to the trademark owner's site: are they going to sites for competitors, or authorized distributors of the trademarked product, or unauthorized second-hand sellers, or informational review sites or blogs? Without this kind of information, it is hard to explain how the policy change is really affecting behavior.
How should these results affect trademark policy? As Eric Goldman has explained, trademark suits over keyword advertising are a losing business, and Google has harmonized its policies to allow advertisers to use trademarks as keywords worldwide; Bechtold & Tucker's study provides more evidence to support this general policy trend, which is an important contribution. But does it tell us anything beyond that? The authors suggest that their "analysis indicates that, by focusing on consumer confusion, the trademark discussion does not fully capture the multi-faceted ways in which search engine users are using trademarks today," and that the "heterogeneity in the effect of the Google AdWords policy change on consumer behavior" and the "resulting tradeoffs and the underlying allocation of property rights" should be the focus of the keyword advertising debate. But I am not convinced that any of these heterogeneous results point to the kind of welfare loss that trademark law should worry about. Still, the descriptive result is very interesting, and I think there are a lot of possibilities for using this dataset to learn even more about how trademarks affect online search.
What does this have to do with The Google Shortcut to Trademark Law? Regular readers might recall that I've written myself about Google and trademark law, and I've often been asked about how my paper relates to Bechtold & Tucker's study. The short answer: it doesn't. Both articles broadly involve "big data" approaches to trademark law, but they use data in very different ways to tackle very different problems. My article is about using online search results as evidence in particular offline trademark disputes to show whether consumers associate a mark with a certain product, and in my study, use of a trademark by a third party to refer to the trademarked product (e.g., Samsung's website saying "our phones are better than iPhones") is still evidence that the mark is associated with that product. For this reason, I excluded disputes over keyword advertising from my dataset.