To Produce Is to Self-Preference

When you first enter antitrust from the left, you are struck by what appears to be a travesty: that a firm that monopolizes an input can get away with denying that input to downstream competitors.

One thinks to oneself: A monopoly using its power to smash a competitor. How is that not an antitrust violation? An antitrust that fails to prohibit that is a perverse, hollowed-out thing captured by the evil it was constituted to destroy. It is the equivalent of the criminal law not prohibiting killing with malice aforethought. Or the contract law not enforcing promises.

And then you get over it, because actually prohibiting monopolists from denying essential inputs to their competitors makes no sense. (I will explain momentarily.)

The trouble with antitrust today is that those setting the agenda from the left haven’t been in the field long enough to get over it.

And so we are left with the embarrassing legislation against “self-preferencing” that is currently making its way through Congress.

Self-preferencing as a concept is nothing new. It is denial of an essential input to a competitor, something that antitrust has, at various times, called a “refusal to deal,” “denial of an essential facility”, and “monopoly leveraging.”

The tech giant that “self-preferences” owns an essential input—its platform—that it denies to firms that compete with it in selling things on the platform. When Amazon uses its control over its website to prioritize advertisements for Amazon-branded products over those of third-party sellers, it denies full access to the platform input to those third-party-seller competitors.

The reason you cannot ban all denials of access to essential inputs—can’t ban self-preferencing—is that every single product produced and sold in the United States is a vast agglomeration of denials of access to essential inputs, as I point out in a recent paper. To ban all denials of essential inputs is, therefore, to ban production. Full stop.

And so the proposed legislation, which would ban self-preferencing by all firms that have websites with more than 50 million users and a market capitalization in excess of $600 billion, would simply make it illegal to have more than 50 million users and a market capitalization in excess of $600 billion.

The Make or Buy or Market Decision and Input Denial

To understand why all products are agglomerations of denials of essential inputs, it is useful to reflect that a company can go about adding a component to a product in only three possible ways.

The firm can make the product itself. The firm can buy the product. And the firm can let consumers buy the component on their own and attach it themselves. I call this the firm’s “make or buy or market” decision.

The first two options—to make or to buy—are what we conventionally mean when we say that a particular component has been incorporated into a product, and the decision whether to “make or buy” has long been a famous one in law and economics.

What has not been properly understood about these two options is that both always involve the denial of an essential input to competitors. Only the market option is consistent with an antitrust that would prohibit input denial as a blanket matter.

It follows immediately that production—the process of making or buying components and adding them together to generate a product—is impossible without input denial, and that a blanket antitrust ban on input denial would make production impossible.

Consider, for example, the act of adding an eraser tip component to a pencil.

The pencil maker can add this component in three basic ways. First, the pencil maker can manufacture the eraser itself from its basic components. It can buy the chemicals needed to make the eraser dough and then bake the dough in molds to produce the erasers. Of course, the firm would then need to solve the problem of how to acquire the precursors (and the oven, molds, and labor services required to run the operation), so making is in the end always buying. But for purposes of showing that making or buying both involve input denial, that doesn’t matter.

The second way the pencil company can add eraser tips to its pencils would be for it to buy them from an eraser supplier. The company would go out into the eraser tip supply market, find the supplier that offers the best quality at the lowest price, and do a deal.

The third and final way the pencil company can add eraser tips to its pencils—or, better put, can ensure that eraser tips are added to its pencils—would be for the company simply to sell eraser tips to consumers and leave it to consumers to affix them to pencils. The firm might sell eraser-less pencils and also separate eraser caps that consumers can buy and affix to the pencils. Or the firm might put a special ferrule on each of its pencils, perhaps like the one currently included with Palomino Blackwings, that would allow consumers to snap in a properly-shaped eraser that they could purchase separately from the pencil company.

Now, when a firm makes its own eraser tips to place on its pencils, the firm denies access to an input—its pencils—to outside manufacturers of eraser tips who would like their own eraser tips to be included on the firm’s pencils. The firm is in effect competing against those other eraser manufacturers in the market to supply eraser tips for the firm’s pencils, and in choosing to make its own eraser tips, the firm uses its control over pencils—an essential input from the perspective of eraser tip makers who have no way to get their products into the hands of consumers other than attached to pencils—to favor its own homemade product in the eraser tip market relative to those of competitors.

In the language of platforms and self-preferencing, the firm’s pencils are the platform that eraser tip manufacturers need in order to reach consumers, and in choosing to use only its own homemade eraser tips on its pencils, the firm engages in self-preferencing in the eraser tip market.

This is equally true when a firm chooses to buy rather than to make.

When the firm chooses a particular third-party supplier of eraser tips, the firm denies the pencil input to all of the other eraser-tip makers in the market. Once the firm has placed its order, none of the others can reach consumers, at least until the firm places a new order for more eraser tips.

Is Buying Really Input Denial?

There seem at first to be two problems with this argument that buying components is input denial. The first is that the third party supplier whose erasers are not picked by the pencil company at least had a chance to have them picked when the pencil company was still deciding which supplier to use. The supplier presumably lost out in some potentially competitive bidding process, which cannot necessarily be said for the case in which the firm chooses to make erasers in-house.

Be that as it may, it has nothing to do with the basic concept of input denial. A firm is no less forced from a market by denial of an essential input when the input holder thought it would not be profitable to supply the input than when the input holder thought that it would be profitable to supply the input. Indeed, one would expect that the underlying motive for input denial would always be the determination that supplying the input would be less profitable than denying it.

Where Is the Downstream Competition When You Refuse to Buy?

The second problem is that the pencil company does not seem to be competing against the eraser suppliers at all. Unlike in the first case, the pencil company does not actually manufacture any erasers of its own, much less sell them. But if the pencil company is no longer competing with eraser suppliers—just buying from them—then the pencil company cannot be said to be denying essential inputs to competitors.

But that is to miss what really bothers us about input denial. What we dislike about it is not that a product that is manufactured in-house by the denier benefits from the denial, but rather that competitors are harmed and the denier somehow benefits from that harm.

And here we can be sure that the denier does benefit.

For the denier would not choose a particular third-party supplier over others unless the denier stood to benefit from buying from that supplier rather than others. This benefit might come in the form of a discount on the price of erasers. Or it might come in the form of willingness of the supplier more faithfully to execute the instructions of the pencil company with respect to the specifications of the erasers. It might even involve a profit-sharing agreement without the pencil company formally owning the operations of the third-party supplier.

Whatever it may be, the pencil company benefits, and as a result of that benefit it is possible to say that, through the third-party supplier chosen by the pencil company the pencil company does in fact compete in the eraser market, and indeed favors its avatar at the expense of competing eraser suppliers.

In the language of platforms and self preferencing, the firm’s pencils are again the platform, and in choosing to buy from one supplier only—a supplier that necessarily offers some special benefit to the firm otherwise the firm would not buy from it—the platform self-preferences, in the sense that it puts the interests of a favored supplier above those of competitors.

Thus the only two ways in which a firm can incorporate a component into a product—making the component itself or buying the component—are both instances of input denial.

Only Selling Components Directly to Consumers Avoids Input Denial

Only the third means of adding a component to a product—that of selling the component directly to consumers and allowing them to add the component on themselves—manages not to be input denial. Indeed, only this “market” case corresponds to what we mean when we speak of the sort of freely competitive market that antitrust is meant to promote.

Only when the pencil company chooses neither to make the component itself nor purchase it and incorporate it into the pencil, but instead merely makes its pencils available for any consumer to use to affix erasers, does the pencil company make its input—its pencils—freely available for any and all eraser manufacturers to use as a vehicle for delivering their erasers to consumers. For now the question which eraser manufacturer’s erasers should appear on the pencil company’s pencils is no longer answered by the pencil company at all. It is answered only by downstream consumers, who have complete freedom to decide which erasers win out in the eraser market.

That is precisely the world antitrust seeks to create: a world in which firms that control inputs do not decide who wins in the market but rather consumers themselves decide.

But Selling Directly to Consumers Is Not Production

The trouble is: that is not a world that is consistent with the production of products that incorporate multiple components. And because all products incorporate an infinite number of components—it all depends on how you define component; if each molecule is a component of a pencil, then how many components does a pencil have?—it is therefore not consistent with the production of anything at all.

It is a world in which everything is pulled apart and atomized, and the atoms are presented to consumers in a vast menu from which no sane consumer would ever be able to choose because choice would require infinite knowledge regarding how to assemble every single product that we have today from individual atoms on up.

What those setting the antitrust agenda today miss is that banning self-preferencing by, for example, Amazon, does not merely prevent Amazon from favoring its own Amazon-branded products over those of third parties.

It also prevents Amazon from choosing the look and feel of its own website, for Amazon.com is a platform upon which web designers notionally compete in selling web design services for Amazon, and when Amazon designs its own website in-house it necessarily denies access to that input to competitors.

A rule against self-preferencing would require that Amazon enable Amazon users to buy Amazon site design from third party software developers and apply it to the Amazon website, so that every Amazon customer could, in theory, choose his preferred look and feel for the Amazon website.

It doesn’t stop there.

Amazon would also be required to allow customers to choose each and every employee who works at the company, for each of these employees is, in a sense, a labor component of Amazon, and in making hiring decisions on its own, Amazon necessarily prevents other competing workers from using the Amazon platform to supply their labor services to Amazon’s customers.

If this seems absurd, that is because it is.

The market approach to adding components cannot work as a blanket matter, and so antitrust, which is tasked with insisting upon the market approach, cannot be applied in blanket fashion to everything, or even just to everything that has more than 50 million users and $600 billion in assets. It can work only if applied sparingly, severing a component of one product here, a component of another product there, when doing so is thought to be in the best interests of consumers, whose preferences are the ultimate measure of the value of production.

That is why the actual rule antitrust applies, the rule of reason, can be summarized as the rule that input denial is illegal when it fails to improve the product that is ultimately sold to consumers.

Self-preferencing is not, in other words, something that can become the subject of a right, as in: “consumers have a right against self preferencing.” It is instead fundamentally about the best way to organize production in individual markets. The question it poses is whether the input denier can be relied upon to choose the component supplier that will make the product as good as possible for consumers, or whether consumers can instead be relied upon to make that decision.

Because consumers do not want to have to decide how everything they purchase is made, it follows that in almost every case the best answer is to let the input denier make that decision—by denying its input to downstream competitors that the denier believes will do a poor job.

Is Every Input Really Essential?

One might object that while every make-or-buy decision might well involve input denial, it does not involve denial of an essential input—it is not denial of an input by a monopolist of that input, not self-preferencing by a platform monopolist as opposed to any old platform—and so is not the sort of behavior that outrages the antitrust neophyte’s intuition that antitrust should ban input denial.

The pencil company’s effective denial of its pencil to third party eraser tip makers when the company makes its own erasers in-house is not input denial, the argument would go, unless there is only one manufacturer of pencils in the pencil market. Only then would eraser tip makers have no way of getting their wares to consumers other than through the medium of supplying eraser tips to that particular pencil company.

This argument would be good enough were all pencil companies to make the exact same pencil—same size, same wood, same graphite, same barrel color, same name, and so on. Only then would it be a matter of complete indifference to eraser tip makers whether they were to supply tips to our pencil company or another.

But, in fact, save for a few standardized commodities, like Class C crude oil, every company’s product is different from every other, even if only in brand name. The result is that every firm will prefer some inputs to others—find some more profitable to use relative to others—and so input denial will always deprive a firm of something that it cannot get anywhere else. In that sense, all input denials deny essential inputs and every maker of an input monopolizes that input.

Dixon Ticonderoga pencils are different from Palomino Blackwings are different from Faber-Castells, and so putting Acme Eraser Tips on each produces a slightly different finished eraser-tipped pencil, one that may be more or less desirable to consumers and indeed more or less profitable for Acme Eraser Tips. If Acme’s preferred pencil company—let’s say it’s Dixon—stops sourcing from Acme, Acme will end up worse off, even if Acme is able to supply Palomino or Faber-Castell instead. In this sense, Dixon is essential to Acme.

Of course, if there were only Dixon, then the consequences of rejection for Acme would be more severe—not just a hit to the bottom line but perhaps bankruptcy—but is it the magnitude of the harm that bothers us about input denial, or is it that a particular business opportunity has been put off limits?

Product differentiation makes of every input, in other words, a mini monopoly. In the “market definition” analysis that antitrust undertakes in merger and monopolization cases, antitrust has traditionally dealt with this by defining a company as a monopoly only if its products are very different from others’—in the lingo, only if they are not “close substitutes”—choosing an arbitrary cutoff between “too different” and “not different enough”.

But here’s the interesting thing: antitrust does not take this approach in deciding whether an input is essential or not. It does not do this in deciding whether a firm monopolizes an input. Instead, antitrust does this only in considering whether the input denier has a monopoly in the downstream market to which the firm is denying the input (another quirk of antitrust law that will not be considered further here because blanket bans on self-preferencing would not require such a downstream inquiry into monopoly power).

With respect to the question whether an input is monopolized, antitrust instead takes a holistic approach, sometimes asking whether the input is an “essential facility,” for example. Indeed, the proposed self-preferencing legislation eschews the market definition approach, instead prohibiting self-preferencing by those who are a “critical trading partner,” defined to mean those who have “the ability to restrict or impede . . . the access of a business user to a tool or service that it needs to effectively serve its users or customers.”

That’s pretty broad language.

The reason antitrust has always been so vague about what constitutes an essential input is that antitrust recognizes that because every input is unique, every input has a downstream firm for which it is essential.

Must an Input Be Something that You Buy?

One might also object that I have been using “input” in rather a peculiar way here, because I have called the buyer the input supplier whereas an input ought to be something that is sold, not bought. The eraser tip suppliers in my example do not buy the pencil company’s pencils. They sell eraser tips to the pencil company and are paid for doing so. In what sense does a pencil company’s refusal to buy eraser tips from some suppliers count as denial of the pencil input to those suppliers?

The answer is that the proper definition of input—the one that corresponds best to our intuition regarding the injustice of input denial—is not “a thing you buy to use in production” but rather “a thing that is necessary for you to do business.”

This broader definition is necessary to prevent clever changes in the locus of product assembly from undermining the antitrust laws. Consider eraser tips again.

I could equally have told the story of an eraser-tip company that purchases pencils, adds eraser tips to them, and then sells the bundle to consumers. In this case, it would be crystal clear that the pencil is the input and the pencil company’s decision not to supply pencils to the eraser tip company would be input denial. In this telling, the eraser tip company would be injured, just as before, by the fact that the pencil company decided either to make eraser tips in-house or to supply pencils to other eraser tip companies seeking to sell eraser-tipped pencils downstream to consumers.

Why should the pencil company be considered any less of an input denier if it were instead to decide to assemble eraser-tipped pencils itself and to start buying eraser tips from eraser tip manufacturers, though not from the eraser tip manufacturer that it targeted before for input denial? The harm to the eraser tip maker is the same, because either way that company is prevented from getting its eraser tips to consumers on the ends of pencils manufactured by that pencil company.

What matters to antitrust here is not how the eraser-tipped pencil makes it to consumers—whether by being assembled by the pencil company or the eraser tip company—but only that a decision of the pencil company not to do business with a particular eraser tip manufacturer harms the eraser tip manufacturer.

The Inevitability of Power and Suffering in Production and Life

And harm the eraser tip manufacturer it does.

But, as I have explained, that does not matter—indeed, cannot matter—unless it makes for a worse end product served up to consumers, because every act of combining two components to make a new product involves a choice regarding which components to join and which not, and so involves a decision to exclude some component makers from the business and not others. Thus the process of product design, production, and creation is always an exercise of power and the infliction of harm.

This, I think, is why it is so difficult to come to antitrust from the left.

Because progressives are uncomfortable with power and the infliction of harm. Progressives want that to go away. But it turns out that everything—everything—they have is inescapably a product of the exercise of power and the infliction of harm.

You cannot build, you cannot create, you cannot make, without choosing—rejecting some additions and embracing others—and each such decision destroys someone’s business (at least to a small extent) and leaves someone out in the cold, if not physically, then socially and mentally.

It would be nice to be able to avoid this ugly scene in which private firms make private decisions about how to make things, hurting each other along the way, and instead to commit all such decisions to the public—here the market, meaning consumers. This is the market option—to just sell all the components directly to consumers and let them decide which should be used and which not.

But the public has trouble enough selecting a President every four years. It does not want, nor does it have the intellectual capacity, to decide how everything is to be made.

And even then harm and the exercise of power could not be avoided, for the public would still have to decide. Consumers might not like Acme Eraser’s eraser tips and so not buy them to put on their pencils. And so Acme would be denied an essential input—a market—and wither as a result. We might, for the moment, think it more just that the public carry out these executions, as opposed to private firms.

But if you really are concerned about the wielding of power and the infliction of harm, then you should not much care who is doing the wielding and the inflicting, but abjure it all.

Thus input denial forces on the progressive the need to come to terms with the inevitability and pervasiveness of power and suffering in business, and, indeed, once one comes to think of it, in life. For all human behaviors involve choices regarding what to prefer and what to reject, with whom to spend time and with whom not, and so all involve input denial and the infliction of pain to a greater or lesser degree.

The really important question then appears to be not whether to condemn power and suffering but rather how to regulate their deployment and infliction. The question is: who should have the power and who should suffer?

In antitrust, the answer the law currently gives is that firms that make inferior products should suffer, and private firms should decide what combinations of components (i.e., what products) are inferior and what not, except in a relatively small set of cases in which consumers, despite their limited cognitive bandwidth, would do a better job of designing their products.

That, it seems to me, is the right approach, and one that leaves progressives plenty of scope to do good by taking authority from firms and giving it to consumers in cases in which firms produce junk.