Our search neutrality paper has received some recent attention. While the initial response from Gordon Crovitz in the Wall Street Journal was favorable, critics are now voicing their responses. Although we appreciate FairSearch’s attempt to engage with our paper’s central claims, its response is really little more than an extended non-sequitur and fails to contribute to the debate meaningfully.
Unfortunately, FairSearch grossly misstates our arguments and, in the process, basic principles of antitrust law and economics. Accordingly, we offer a brief reply to correct a few of the most critical flaws, point out several quotes in our paper that FairSearch must have overlooked when they were characterizing our argument, and set straight FairSearch’s various economic and legal misunderstandings.
We want to begin by restating the simple claims that our paper does—and does not—make.
Our fundamental argument is that claims that search discrimination is anticompetitive are properly treated skeptically because: (1) discrimination (that is, presenting or ranking a search engine’s own or affiliated content more prevalently than its rivals’ in response to search queries) arises from vertical integration in the search engine market (i.e., Google responds to a query by providing not only “10 blue links” but also perhaps a map or video created Google or previously organized on a Google-affiliated site (YouTube, e.g.)); (2) both economic theory and evidence demonstrate that such integration is generally pro-competitive; and (3) in Google’s particular market, evidence of intense competition and constant innovation abounds, while evidence of harm to consumers is entirely absent. In other words, it is much more likely than not that search discrimination is pro-competitive rather than anticompetitive, and doctrinal error cost concerns accordingly counsel great hesitation in any antitrust intervention, administrative or judicial. As we will discuss, these are claims that FairSearch’s lawyers are quite familiar with.
FairSearch, however, grossly mischaracterizes these basic points, asserting instead that we claim
“that even if Google does [manipulate its search results], this should be immune from antitrust enforcement due to the difficulty of identifying ‘bias’ and the risks of regulating benign conduct.”
This statement is either intentionally deceptive or betrays a shocking misunderstanding of our central claim for at least two reasons: (1) we never advocate for complete antitrust immunity, and (2) it trivializes the very real—and universally-accepted–difficulty of distinguishing between pro- and anticompetitive conduct.
First, we acknowledge the obvious point that as a theoretical matter discrimination can amount to an antitrust violation in some cases under certain specific circumstances—not the least important of which is proof of actual competitive harm. To quote ourselves:
The key question is whether such a bias benefits consumers or inflicts competitive harm. Economic theory has long understood the competitive benefits of such vertical integration; modern economic theory also teaches that, under some conditions, vertical integration and contractual arrangements can create a potential for competitive harm that must be weighed against those benefits . . . . From a policy perspective, the issue is whether some sort of ex ante blanket prohibition or restriction on vertical integration is appropriate instead of an ex post, fact-intensive evaluation on a case-by-case basis, such as under antitrust law. (Manne and Wright, 2011) (emphasis added).
This is not much of a concession. While FairSearch tries to move the goalposts by focusing on a straw man proposition that search bias is categorically immune from antitrust scrutiny, this sleight of hand doesn’t accomplish much and reveals what FairSearch is missing. After all, consider that almost every single form of business conduct can be an antitrust violation under some set of conditions! The antitrust laws apply in principle to (that is, do not categorically make immune) horizontal mergers, vertical mergers, long-term contracts, short-term contracts, exclusive dealing, partial exclusive dealing, burning down a rival’s factory, dealing with rivals, refusing to dealing with rivals, boycotts, tying contracts, overlapping boards, and all manner of pricing practices. Indeed, it is hard to find categories of business conduct that are outright immune from the antitrust laws. So—we agree: “Search bias” can conceivably be anticompetitive. Unfortunately for FairSearch, we never said otherwise and it’s not a very interesting point to discuss.
With that point firmly established, one can return focus to the topic FairSearch painstakingly avoids throughout its response and on which we think the issue really does (and should) turn: Where’s the proof of consumer harm?
We make the rather common sense point that when engaging in a case-by-case factual analysis of the competitive effects of business conduct, one should take advantage of what we already know about the class of business practice generally. Along those lines, we noted that the economic literature has extensively analyzed the competitive effects of such discrimination and concluded that in most cases it yields significant efficiencies and is therefore unlikely ex ante to amount to an antitrust violation. Perhaps FairSearch was confused by this argument, though we explicitly frame the choice as one between ex ante categorical regulations and ex post case-by-case analyses, not as one between ex ante prohibitions and “allowing Google free rein to discriminate against competitors,” as FairSearch bizarrely claims. This is something akin to claiming a defense attorney demands “immunity” for a client with an alibi rather than the correct outcome under a faithful application of the law. (Although we hasten to add that it’s not clear that would be, in fact, anything at all wrong with Google “discriminating against competitors.” We’re sure its shareholders would be apoplectic if it didn’t, in fact).
At any rate, even before we get to the question of whether there is any evidence of consumer harm, we do indeed think it is relevant to any analysis that bias is hard to identify, that one user’s “bias” is another user’s “relevant search result,” that a remedy would be difficult to design and harder to enforce, and that the costs of being wrong are significant. These are not dispositive, nor do we claim they are. But they do underscore another point that FairSearch misses: That harmful search bias would be exceptionally difficult to detect even if it were assumed to exist.
This problem can’t be brushed off, and it mirrors the well-known uncertainty at the core of the antitrust enterprise: That a given course of conduct may prove pro-competitive or anticompetitive under differing situations. Former AAG Tom Barnett and current Google critic (representing Expedia) has himself echoed this point, observing that
No institution that enforces the antitrust laws is omniscient or perfect. Among other things, antitrust enforcement agencies and courts lack perfect information about pertinent facts, including the impact of particular conduct on consumer welfare . . . . We face the risk of condemning conduct that is not harmful to competition . . . and the risk of failing to condemn conduct that does harm competition . . .
Barnett further notes that “discerning whether a monopolist’s actions have hurt or helped competition can be extremely difficult.”
At the same time, FairSearch does not dispute that significant efficiencies can arise from vertical integration in the search engine market. Rather, FairSearch distracts itself from this substantial problem in its response by instead quibbling over the factual details of our shelf space analogy and arguing incoherently that Google’s conduct violates antitrust laws merely because it deprives rivals of scale. But here FairSearch misses the mark in telling ways.
First, on the shelf space analogy: As we’ve discussed numerous times (e.g., here and here) the shelf space analogy illustrates that promotional arrangements and vertical integration can have obvious benefits; for example, such arrangements can align incentives for product promotion between up- and downstream actors, thereby increasing output and consumer welfare. FairSearch confuses the analogy by focusing upon the fact that supermarkets, unlike Google, certainly do not have monopoly power. To borrow a quote, we do not think that means what FairSearch thinks it means. That competitors without market power in highly competitive markets engage in a challenged conduct just as competitors with it do might suggest to the astute antitrust observer that the conduct has real competitive merit unrelated to any exclusionary effect. Moreover, the shelf space analogy is designed to highlight the essential fact that because a firm is large (or has a large market share) does not mean its conduct does not also generate consumer benefits! As the court in Berkey Photo noted:
[A] large firm does not violate § 2 simply by reaping the competitive rewards attributable to its efficient size, nor does an integrated business offend the Sherman Act whenever one of its departments benefits from association with a division possessing a monopoly in its own market. So long as we allow a firm to compete in several fields, we must expect it to seek the competitive advantages of its broad-based activity – more efficient production, greater ability to develop complementary products, reduced transaction costs, and so forth. These are gains that accrue to any integrated firm, regardless of its market share, and they cannot by themselves be considered uses of monopoly power.
The point of the analogy, which FairSearch entirely misses yet does not deny, is that these arrangements yield pro-competitive efficiencies – and that these efficiencies are not a function of monopoly power but rather of efficient—even innovative—forms of business organization.
Similarly, FairSearch’s arguments about scale in the search engine market are unpersuasive and represent a serious misunderstanding of the market dynamics. FairSearch seems to argue that increased access to ad space leads to higher profits, and that Google, as the recipient of this windfall, can accordingly out-buy its rivals, implying that, eventually, they will simply wither on the vine as Google ends up possessing all the money in the world. Or something like that.
Of course this argument is outlandish on its face. Microsoft, which has overwhelming resources, is one of Google’s closest competitors and certainly cannot be easily out-bought. Moreover, the initial claim itself is weak because (1) many advertisers multi-home, dissipating any efficiency that increased access would yield, and (2) advertisers pay per click – because search engines with fewer users necessarily initiate fewer clicks, advertising may be cheaper on smaller platforms. So, while the value on two different size platforms may diverge, so do prices. Merely asserting that one has more value and therefore higher prices is not sufficient to demonstrate divergence on the relevant dimension; rather these factors could – and likely do – increase commensurately.
The most glaring flaw in FairSearch’s argument, however, is its failure to present any evidence whatsoever of competitive harm. In demonstrating a Section 2 violation, the burden is on the plaintiff in the first instance to make a showing of harm to competition. Rather than engage in this effort itself, FairSearch summarily dismisses our argument because we “do not seek to marshal evidence that Google does not manipulate search results to harm competition.” However, this is not our burden to bear. We repeat: FairSearch fails to present any evidence of its own case-in-chief; it relies on an economically incoherent and premature dismissal of claims we did not make (categorical immunity for “bias”) in making its own naked assertions.
Note FairSearch’s remarkable sleight of hand here – it (1) attempts to shift the burden of proof to us while (2) simultaneously equating bias with harm to consumer welfare, in an attempt to further simplify its burden. Bias, however, does not implicate harm to consumers, as FairSearch suggests–nor has harmful bias ever even been convincingly demonstrated.
FairSearch asserts that research has shown that Google often ranks its own results above those of rivals “without any apparent relationship to the quality of these Google sites as compared to competing sites.” However, Professor Ben Edelman’s study, upon which FairSearch relies for this assertion, is a far cry from the sort of rigorous analysis Mr. Barnett called for in Section 2 cases as the Attorney General (“there seems to be consensus that we should prohibit unilateral conduct only where it is demonstrated through rigorous economic analysis to harm competition and thereby to harm consumer welfare”). Indeed, comparing 32 hand-picked search queries across search engines is hardly an adequate sample size or methodology for these purposes – and certainly does not suggest that Google is entirely unconcerned with the quality of its results. In any event, so-called “bias,” even if proven, may at most represent harm to rivals – and this is not the relevant metric of antitrust injury. Furthermore, as noted above, bias offers significant benefits and more often than not enhances consumer welfare.
Overall, FairSearch’s critique of our paper is weak and crucially flawed. FairSearch relies upon a bald assertion that bias exists to equate that bias to consumer harm while conceding (but ignoring) that vertical integration may introduce significant efficiencies in the search engine market. It tries to wish these efficiencies out of existence by erroneously claiming that monopoly power in the downstream market forecloses that possibility. The basic economic analysis of vertical integration says otherwise. But most essentially, FairSearch simply offers no evidence at all of consumer harm. By alleging antitrust injury, FairSearch has the burden of demonstrating competitive harm in the first instance. This burden is especially high in the search engine market, where, in contrast to the dearth of evidence of consumer harm, evidence of innovation abounds.
All of which is to say, our original argument bears repeating: Claims of anticompetitive conduct should be viewed with serious skepticism when there is abundant evidence of consumer benefits, in the form of innovation and competition, and zero evidence of consumer harm. If FairSearch wishes to mount a credible challenge to our analysis they have a lot more work to do.