Antitrust Monopolization

When the Food Section Gets Bigness, but the Business Section Doesn’t

It’s a good thing that The New York Times’ Food department hasn’t gotten the small-is-beautiful memo.

On the same day that the paper ran another flawed installment in its crusade against bigness, this time targeting Google for bringing competition to wireless-speaker-maker Sonos, the Times’ food section ran a paean to behemoths of the restaurant business–chains like IHOP and Applebee’s–that highlights many of the reasons why size is often a good thing, both for workers and consumers.


Let’s start with the Times’s wrongheaded defense of Sonos.

As the paper did in an earlier defense of cloud-computing startup Elastic, the Times here continues to confuse harm to competitors with harm to competition. Google, the paper suggests, is competing unfairly with Sonos, by “flooding the market with cheap speakers that [Google] subsidize[s] because [the speakers] are not merely conduits for music, like Sonos’s devices, but rather another way to sell goods, show ads and collect data.”

The Times is talking about Home, Google’s answer to Amazon’s Echo, which includes a high-definition speaker that plays music, but also provides access to Assistant, Google’s AI-powered virtual assistant, which allows users to run internet searches, buy products, order food, and do much more by conversing with the speaker system.

The Times weeps that Sonos can’t turn a profit selling its speakers–which only play music–for less than $200, whereas Google sells Home for $50. The implication is that Google is engaged in predatory pricing–sales of products below their cost of production–for purposes of driving competitors from the market. That’s always possible in an abstract sense, and would be an antitrust violation if some other criteria were also met.

But there’s an obvious alternative explanation staring the Times in the face, one that doesn’t involve anticompetitive conduct: that Google isn’t in the market to sell speakers, Google is in the market to sell virtual assistants that also happen to play music.

And when you count up all the different ways Google is able to generate revenue from its product, including commissions Google earns on goods purchased through Home, revenues Google generates from selling ad-targeting services using the data generated by Home, and, of course the $50 purchase price of a Home unit itself, those revenues probably cover Google’s costs, including the cost of making the speakers that go into Home.

We don’t say that a hotel that offers guests free breakfast is engaged in predatory pricing designed to drive the local Starbucks out of business, even though a breakfast price of $0 is definitely below the cost of making the breakfast. Because the hotel is not selling breakfasts. The hotel is selling a package, and the hotel includes the cost of the breakfast in the price the hotel charges for the room. If the local Starbucks wants to compete, then either Howard Schultz needs to get into the hospitality business, or Starbucks needs to offer better coffee than the hotel. (Have you ever skipped out on a free breakfast to go somewhere better? I have.)

The same goes for Sonos. To beat Google, Sonos can try to field its own virtual assistant. Admittedly unlikely, but not a reason to condemn Google, for reasons to be discussed in a moment. Or Sonos can build better speakers than the ones Google bundles with Home, speakers that are enough better to make music lovers willing to choose them over, or in addition to, Home.

The Times makes much of the fact that Google may have used information Google collected from its partnership with Sonos to copy Sonos’s speakers. But as I have emphasized in relation to other reporting by the Times, copying is good for competition, not bad, and is certainly no antitrust violation. If Google copies Sonos’s speakers, making Google’s own as good as Sonos’s, that will have the competitive result of preventing Sonos from leveraging the superiority of its products to charge monopoly prices for them.

Of course, we want innovators to reap some rewards for innovating, which is why the patent laws prevent copying for a limited period of time. Sonos is suing Google for patent infringement, and if Google has infringed, then the court will award Sonos lost profits, as it should. But such patent litigation is about enabling firms to preserve the legislatively-sanctioned monopoly that is a patent on a desirable product. Patent litigation is not about preserving competition.

The Times also makes much of the fact that Sonos’s CEO confessed to being frightened about suing Google, because Google might respond by terminating a partnership with Sonos that allows Sonos owners to use their Sonos speakers, in lieu of Home, to communicate with Google’s Assistant.

But that’s just business. If Sonos postponed suing Google for patent infringement because Sonos wanted to continue to be able to have access to consumers wishing to buy virtual assistants, rather than just speakers, then Sonos was effectively licensing its speaker patents to Google at a price equal to the extra profits that Sonos was generating from the virtual assistant business. If Sonos is asserting its patents now, that means that Sonos thinks it can make more from court-ordered licensing than from the informal exchange of access to its technologies for access to Google’s Assistant.

Standing behind the Times’s article is the unspoken assumption that without the ability to offer access to virtual assistants through its speakers, Sonos is doomed, regardless how good its speakers may be, because consumers don’t care enough about great speakers to be willing to buy them in lieu of, or in addition to, speakers bundled with a virtual assistant, such as Google Home. That may be true, and sad for Sonos, but the ultimate cause must be that Sonos is simply less technically savvy than Google.

Google invested in the search and AI it needed to produce a virtual assistant. Sonos didn’t. True, Sonos may have pioneered wireless speaker technology that Google was not able to match without licensing that technology (informally so far, perhaps formally, under court order, in future) from Sonos. But Sonos could have taken the same tack against Google, reverse-engineering Google’s search algorithms and Assistant AI to create its own virtual assistant. If Sonos wasn’t able to do that, because it would have required too much time and money, then that’s evidence that what Google has achieved in search and AI is much more of a technological advance than are Sonos’s speakers.

Which takes us back to the basic point that to the extent that Sonos is failing to compete effectively against Google it’s because Google is doing a better job than Sonos at giving consumers what they want, not because Google is restraining competition. Once again, the Times has mistaken a textbook case of effective competition for an example of monopoly.

It’s also worth noting that Google has not actually yet retaliated by cutting Sonos off from access to Assistant, no doubt because Google recognizes that Sonos is better at making speakers than is Google, and Google can build its virtual assistant market share by reaching consumers who care about getting great speakers through Sonos.

That, too, is how markets are supposed to work. If Google can make its product better by combining it with rival technology, Google will do that. The fact that Sonos might not be able to survive without Google but Google can survive without Sonos means that Google can drive a hard bargain with Sonos and absorb most of the gains from trade. But Google can’t drive such a hard bargain as to make Sonos unwilling to go on, because then Google will lose the customers it can only get through Sonos.

That means that Sonos will not turn into the next tech giant. But with 1,500 employees and a billion dollars in annual sales (which the Times rather humorously tries to downplay as “a nice little business”), Sonos is doing just fine, even with the short end of the stick. We don’t all get to be the next tech fairy tale. (And if Google does pull the plug on its partnership with Sonos, the company can always compete to supply its speaker technology to Google for incorporation into Home. Indeed, Sonos’s patent suit may be a prelude to a transition into that new business model.)

The Times’ piece on Sonos is also a sobering reminder of the extent to which the paper’s business pages have become a mouthpiece for writers’ self-interested war on Google, Facebook, and Amazon, three companies that writers see as having tanked their earnings in recent years, as I have argued in depth elsewhere.

It’s not just the rhetoric that belongs more comfortably in a polemic than a news feature (Sonos is “under the thumb of Big Tech,” according to the Times). It’s also the sourcing.

The Times tells us that “congressional staff members have discussed [Sonos CEO Patrick Spence’s] testifying to the House antitrust subcommittee soon about his company’s issues with them,” but fails to mention that those hearings have been convened by a Congressman who is simultaneously sponsoring legislation pushed by the News Media Alliance, an industry trade group, that would give newspapers an exemption from the antitrust laws. The Times also quotes an employee of the Open Markets Institute describing Sonos’s fear of Google as “real,” without revealing that Open Markets is run by a journalist with ties to an organization that advocates on behalf of writers. More on both connections here.

But do you think that the Times would care to ask an actual antitrust law scholar whether Google’s conduct is anticompetitive? Nuh-uh. The article couldn’t have been written more critically of Google if Open Markets, or the House antitrust subcommittee, had authored the article itself and issued it as a press release.


Thank goodness neither Google, Facebook, nor Amazon is in the restaurant business. Because in that case it would be hard to imagine the Times publishing Priya Krishna’s recent love letter to massive chain restaurants, “Current Job: Award-Winning Chef. Education: University of IHOP.”

According to Krishna’s piece in the Times:

Chain restaurants are often accused of a sterile uniformity and a lack of attention to quality ingredients, nutrition and the environment. But for anyone trying to enter the restaurant business, they have particular attractions: formalized training, efficient operations, predictable schedules and corporate policies that claim to discourage the kind of abuses that have come to light in the #MeToo era. The pay is sometimes better than at independent restaurants, and the Affordable Care Act requires companies with 50 or more full-time employees to provide health insurance.

The article highlights several “acclaimed chefs [at independent restaurants] who prize the lessons they learned . . . in the scaled-up, streamlined world of chain restaurants,” from the influential chef who eats at Waffle House to Jacques Pepin, who spent ten years working at Howard Johnson’s.

Chain restaurants provide workplaces that are, it turns out, less heirarchical than independent restaurants. Because egalitarianism is more efficient. At Applebee’s, for example, there isn’t “a strict heirarchy . . . because the kitchen [isn’t] centered on a chef, as in many independent restaurants. ‘There is this understanding that every person is important to making the restaurant run smoothly . . . Nobody thought the dishwasher was a lower status than them.'”

According to the article, “[s]everal chefs point to rigorous customer-service standards of the chains where they worked. ‘It was pretty much that the customer is always right,'” said one chef, who observed to the Times that “[i]t’s a level of hospitality he doesn’t always see in fine-dining restaurants.”

Another chef reported having had to “make sacrifices: lower pay, or forgoing pay while training” when she moved to working at independent restaurants.

She also had to put up with abuse. The article quotes her as recalling that when it took her too long to run food to a table at an independent restaurant, “‘the chef threw a potato and it hit me in the head. . . . That kind of stuff doesn’t happen in a chain restaurants [sic] because of corporate structure. You tend to be treated more fairly.'”

Shortly after reading this article, I went to a small family-run butcher’s shop to get a thinly-sliced cut of meat that my wife needed for a dish she was preparing. The slicing machine was in a back room into which a small internal window had been cut. I could just make out through the glare that the butcher was handling the meat with his bare hands.

I didn’t complain, but I did make my next stop a Kroger’s, the largest grocery store chain in the world. Economics teaches that if this firm were a monopoly, it should have lower quality standards than firms on the competitive fringe, like the family-owned butcher shop I had just left. I went to the meat section and asked for the same cut. The slicing machines were all directly behind the counter, in full view of customers. And the first thing the butcher did was to put on some gloves. True, he wasn’t as friendly as the folks in the family store. But when I got home, I gave only the cuts from Kroger’s to my wife. Big is not always bad.

Small businesses are a good thing, in my view, but only when they are actually better than big businesses. Thousands of independent restaurants survive, particularly in the luxury space, despite treating their labor less well than do the chains, because they provide a shot at top-chef fame for employees and a unique dining experience for customers that chains haven’t yet been able to match. The success of independent coffee shops in resisting Starbucks by taking coffee connoisseurship to another level is also a great example.

But when a smaller firm fields a product that isn’t better than what its rival has to offer, when a firm tries to sell speakers to consumers who would rather buy speakers-plus-virtual-assistants, the solution is not to try to use the antitrust laws to shelter the smaller firm.

The solution is to let the company up its game, or clear out.


The Times’ Elastic Conception of Monopoly

Pressure groups, too, used ideological symbols for selfish ends, sometimes to mask operations that were completely at variance with the professed ideals. [That] made it difficult to follow the struggle, define positions, and identify the participants.

Ellis W. Hawley, The New Deal and the Problem of Monopoly 36 (Fordham Univ. Press 1995) (1966).

As I have been arguing for some time now, the press’s antitrust crusade against Amazon, Google, and Facebook is about protecting competitors, not competition. The press is only really interested in protecting two groups of competitors in particular, newspapers and publishers, the firms that give a livelihood to the writers who have created this crusade.

But movements grow through the building of coalitions, and writers have worked feverishly to sell their crusade as protection for all competitors of Amazon, Google, and Facebook, and not just those that hire writers in particular. The result has been a series of articles about small businesses and startups that have felt competitive pressures from the three tech giants.

The most recent of these illustrates beautifully writers’ failure to understand that harm to competitors, whether writers or anyone else, is not the same thing as harm to competition.

Prime Leverage: How Amazon Wields Power in the Technology World,” which appeared in The New York Times, tells the story of Elastic, a cloud-computing startup that introduced an innovative search tool that Amazon at the time did not offer. What happened next is exactly what you would expect of a healthy, competitive market. But the Times finds it all very sinister.

When Amazon saw that Elastic had created a superior product, Amazon did what competitors are supposed to do in competitive markets. Amazon copied the product as best it could, and introduced its own version into the market.

Copying, so far from being evidence of monopolization, is actually the market economy’s principle defense against monopolization. When a firm introduces a superior product, the firm in effect takes consumers hostage, because consumers face only inferior products as alternative purchase options. That allows the innovative firm to charge consumers a monopoly price for the superior product.

Copying naturally limits the innovator’s monopoly power. As competitors copy or even improve upon the innovator’s product, consumers start to enjoy an increasing number of alternatives of equal or superior quality, and the prices the innovator is able to charge for the product fall.

Amazon’s move to copy Elastic’s product limited Elastic’s pricing power, preventing the company from charging monopoly prices for the tool. No doubt that made Elastic’s founders very sad. Tech startup culture has accustomed startup founders to the expectation that they will be able to generate monopoly profits from their innovations, and become fabulously wealthy as a result.

But as progressives have been arguing for a long time now, such out-sized profits are neither necessary incentives to induce entrepreneurs to innovate, nor fair. The presence of large incumbents in startup markets who are able to compete away monopoly profits quickly is good for the economy, and the distribution of wealth. Note that to win the competition with Elastic, Amazon can’t charge monopoly prices for its version of Elastic’s search tool, so neither Amazon nor Elastic generate monopoly profits from this technology. Consumers are the principal beneficiaries of Amazon’s ability to copy competitors’ products quickly and effectively.

Now, one might worry that Amazon’s copying of Elastic’s product might ultimately harm competition by driving prices so low that Elastic is not able to recoup its costs of innovating, effectively sending a signal to the market that innovation no longer pays because Amazon will steal the fruits.

But this concern isn’t about the presence of excessive monopoly power–Amazon’s or anyone else’s–but rather a concern about excessive competition. Indeed, it’s a concern about the ability of innovators to capture the benefits they confer on the economy despite competitive pressures.

If an innovation is too easy to copy, then the innovator will enjoy only a very short period of exclusivity in the market, and may not be able to recoup its development costs before competition sets in and prices fall to modest levels. The law handles this problem by creating intellectual property rights, which allow innovators to enjoy legally-guaranteed exclusivity for a set period of time or under certain circumstances. Elastic clearly was not able to patent its search tool, and Amazon was clearly able to invent around any trade secret or copyright protections enjoyed by Elastic.

What this all means is that the intellectual property laws don’t view Elastic’s tool as the kind of innovation that needs extra legal protection from copying. From the law’s perspective, Elastic will do just fine on its own, thanks to the low cost of developing new software tools and the advantages in brand recognition and follow-on innovation that Elastic enjoys from being a first mover with respect to this technology.

That is just what seems to have actually happened in Elastic’s case. Although the first paragraphs of the Times article lead the reader to expect news of Elastic’s bankruptcy by the end of the story, the paper is forced eventually to admit that “Elastic . . . went public last year and now has 1,600 employees,” up from 100 when Amazon first copied its tool. So Elastic didn’t need intellectual property protection to earn enough from its innovations to thrive, despite competition from Amazon. (The article informs us that Amazon did violate Elastic’s trademark by giving its tool the same name as Elastic’s. Trademark law will provide Elastic with compensation, in the form of lost profits, for that infringement.)

This is exactly how you want competition to function. New firms enter markets by introducing innovative products, and competition from incumbents struggling to catch up prevents the new firms from growing into monopolies themselves, but at the same time does not drive them out of the market.

The story of Elastic is a story of healthy competition that benefits society as a whole, but not, of course, Elastic, which would much rather have faced no competition from Amazon at all. The Times notes that startups refer to Amazon as engaging in “strip mining” when the firm copies the products of competitors. Of course, every firm would prefer that competitors never be allowed to copy their products, because that would give every firm a permanent monopoly in the technologies that the firm innovates. To the startup entrepreneur who wants an easy path to tech riches, all competition is “strip mining.”

But preventing copying is not good for America. That the Times would seem to be promoting this sort of innovation-based monopolization, despite the paper’s advocacy of antitrust enforcement against Amazon, reflects the contradictions inherent in any attempt to use the antitrust laws to protect competitors rather than competition. Protecting competitors means giving them monopolies. Protecting competition, by contrast, means leaving competitors to sink or swim.

There are more contradictions. The Times seems, for example, not to realize that the fact that Amazon dominates cloud computing must surely have resulted in more protection, and less competition, for Elastic, than Elastic would have enjoyed in a less concentrated market. The less concentrated the market, the greater the number of firms in a position to copy an innovation, and the more severe the resulting price competition is likely to be.

Because Elastic’s only competitor was Amazon, the market was a duopoly, and duopoly market pricing can be much closer to the highs of monopoly pricing than to the lows of competitive pricing. That might explain why Elastic was able to recoup its development costs, grow exponentially, and successfully go public despite Amazon’s sale of an essentially identical product to that offered by Elastic.

The Times also suggests that the fact that Amazon both owns the cloud servers used by most internet firms, including Elastic, and also competes in the market to provide software services for use on those servers, is inherently anticompetitive. The notion that firms should not be allowed to compete on their own platforms is probably the most embarrassing and superficial effusion of the press’s crusade against Amazon, Google, and Facebook.

For all businesses are nothing more than collections of platforms, from the vast majority of which every business excludes competitors (think of office space). To prohibit competition on one’s own platform is to prohibit productive activity entirely. The fact that Amazon has decided to throw open its cloud servers–which the firm initially developed exclusively for its own use–to competitors is an act of great pro-competitiveness, not the opposite.

That Amazon may have used information about Elastic’s tool that Amazon could only have generated from its administration of the servers that Elastic used to run the tool is hardly anticompetitive. To the contrary, that allowed Amazon to accelerate the copying process and introduce competition into the market for that tool more quickly than the company might otherwise have been able to do. If that is too much competition for Elastic, then the intellectual property laws can be changed to provide software tools with more protection than they receive today. The fact that Elastic did not succumb to this competition, but thrived, suggests, however, that more intellectual property protection–more monopoly for Elastic–would not be appropriate.

One problem with the press’s crusade against Amazon, Google, and Facebook has been a failure to recognize that these companies’ size makes them more efficient, and that breaking them up would therefore result in a net loss for society. From this perspective, the press is calling for more competition where competition would not be a good thing. But in the Times’s defense of Elastic, there is something quite different: the promotion of monopoly–saving Elastic from competition from Amazon–where competition would be more appropriate.

The only way to understand the press’s adherence to these inconsistent positions–promoting competition here, monopoly there–is as a desire to have the antitrust laws pick winners and losers in the market according to the press’s own particular set of special interests and political preferences.

Is the press aware that it is ultimately trying to do that?

You bet.

Civilization Miscellany

An Internet of Steles

kairinfo4u, Wadi Hammamat: Dynasty 11, Mentuhotep IV, CC BY-NC-SA 2.0. kairinfo4u writes: “In the ‘Middle Kingdom’ Nebtawyre Mentuhotep IV sent an large expedition in Year 2 of his reign to quarry bekhen-stone for the King’s sarcophagus. We are told by Vizier Amenemhat how a gazelle gave birth on the block they had chosen, an auspicious omen which greatly encouraged the workforce of 10,000 men. Another of Vizier Amenemhat’s texts records the wonder of rain, a flash flood that produced a well of clear water.”

[H]e engraved on a stone the whole story.

The Epic of Gilgamesh 61 (N. K. Sanders trans., Penguin Books 1960).

Bronze age rulers erected steles so that their words would endure. We do that today with blockchain.

Almost everything we know about ancient Egypt, for example, even the name “Ramses,” comes to us from steles and other inscribed stones dug out of the sand as many as 5000 years after pharaohs ordered them carved. Inscriptions in stone endure because stone is difficult to work. Hard to destroy; harder to recarve in ways that do not betray the fact that recarving has taken place. When the pharaohs made a record in monumental stone, they made public records the authenticity of which could be verified, even by scholars working millennia hence.

But that’s just what blockchain does for the internet. Blockchain inscribes information onto computer memory in a manner that, like a stone carving, is very difficult to change.

Making the information stored in computer memory permanent is not easy, because computer memory is engineered for rapid change. Computers record data by rearranging the electrons adhering to the physical material of a disk, tape, or chip. Changing the data therefore requires no more than an application of electricity.

The ease with which data can be changed in computer memory is the source of computing’s power, driving the cost of communication almost to zero. In the millennia following the carving of the first steles, which are very costly to create, civilization succeeded at finding increasingly inexpensive methods of recording information. But even the most inexpensive methods devised, such as paper, still required costly manipulation of matter on a macro scale–the application of inks–to be useful. Computer memory outdid all alternatives by requiring only manipulation of the utterly insubstantial electron.

But with the reduction in costs came impermanence. You, or a hacker, could change your data without leaving a trace of what came before. Indeed, without anybody being able to say for sure whether your data had been changed at all.

Blockchain tries to solve the problem of data impermanence, while preserving all of the advantages of electronic computing and communication, by storing data in an encrypted format. Changes to the data not made using the proper format can immediately be detected by readers. So merely changing the data electronically, while just as easy as it has always been, won’t fool readers, who can see that the changes don’t conform to the standard.

Indeed, the fact that blockchain solves the impermanence problem without changing the basic ease of storing data with electrons means that blockchain allows computers to continue to communicate quickly and cheaply. Data endures because it has been tied to encryption cyphers, not because it has been tied to the physical world, as in the case of a pharaoh’s steles.

A different approach to the problem of internet permanence would be to rig up a computer system in which robots would store data by automatically carving the data onto stone tablets. That too would solve the permanence problem. Anyone who wanted to verify the data could inspect the stone tablets to ensure that they had not been altered, just as archeologists inspect ancient steles today. But having computers write data to steles would make it difficult to copy and transfer that data even when the data has not been altered. Blockchain captures the unalterability of stone inscriptions without suffering from limits on communicability associated with the use of stone as a medium.

But why exactly does encryption breed permanence? Can’t you just crack the code and change the data in a way that respects the encryption format and therefore is not detectable by others? The answer is no because cracking codes is hard, requiring powerful computers, lots of electricity to run them, and time. Just as effectively rechiseling a stone inscription requires expertise, energy, and time. So blockchain uses encryption to restore the permanence in data that the information age destroyed.

With one important difference. Blockchain is an effective check on the undetected rewriting of data, as are steles, but, unlike steles, blockchain is no check on destruction of data, in the sense that blockchain makes deleting data from computer systems no harder to do than before. That is the price blockchain pays for allowing users to continue to communicate quickly with each other. Blockchain sits on top of the electron-based storage systems of computers, making it very hard to change the data undetected, but no harder to destroy the data on those systems. An electric shock will still suffice for that.

So the pharaohs still have something on computers, at least with respect to preventing data destruction, rather than just the alteration of data. (Of course, unlike data stored on steles, internet data is stored in multiple locations, forcing the destroyer to travel to be effective.)

Blockchain is so much an artifact of information technology that it could not be useful without that technology. The basic blockchain concept of using encryption to prevent alteration of data has been around forever. People wrote in code in the 16th century as much to keep their words secret as to ensure that what they did write could not be altered imperceptibly. But encoding and decoding are expensive and time intensive, even when you have the key to the code, and are not trying to crack it. Blockchain is feasible on a large scale only because users can rely on computers to determine whether data conforms to the required format. Thus information technology, despite feeding on impermanence, also enables a new kind of permanence.

The tearing down of constraints, so feted in our technological age, is not always a good thing. Nature is constraint. Technology, in mastering nature, removes constraint. But a lack of constraint is chaos, the opposite of civilization. One way to retain constraint is through law, but that has proven a feeble method. The only alternative is therefore to use technology to build artificial constraints back into nature, albeit in ways that are more suitable to our needs than natural constraints once were. Blockchain is an installment in that enterprise.

(I thank Thibault Schrepel and Sam Weinstein for piquing my interest in blockchain.)


Power in Profiles

One strategy that the press has deployed in its war on Amazon, Google, and Facebook is the profile. Because profiles legitimate, and what the press’s war on these companies lacks is legitimate intellectual support. Readers assume that if The New York Times or The Wall Street Journal chooses to write about you, then you must know something, and that causes bureaucrats, legislators, other journalists, and even some academics, all of whom read the papers, to start treating you like you do.

The Times’s profile two years ago of a recent law school graduate who had written a student note attacking Amazon appears to have had this effect.

The Wall Street Journal’s profile of Dina Srinivasan today seems to be trying to do the same thing.

The profile pitches Srinivasan as having made a surprisingly successful academic case for antitrust intervention against Facebook. But one strains to find anything in the profile that distinguishes Srinivasan other than the fact that she titled a law review article she wrote as “The Antitrust Case against Facebook.”

Is Srinivasan an antitrust expert? No. As the article beguilingly informs us, when she wrote her paper Srinivasan had “neither any institutional affiliation or a law license,” though she did have a law degree from Yale that she had “never put to use” (until now, apparently). Indeed, we learn that until recently she’s been unemployed, having “quit her job as a digital advertising executive two years ago.”

Was the article in question published in a top law review? No. It was published in a specialty law review, rather than the general interest law reviews that make up the first tier of outlets for legal academic work. The Journal tries, appallingly, to make this placement sound like a coup, by saying that the law review in question published her “unsolicited article.” But in law virtually all articles appearing in academic law reviews are unsolicited.

Is this pioneering work? No. Srinivasan’s argument that Facebook charges users a price denominated in data rather than dollars is nothing new. In 2017, when Srinivasan wrote her piece, the concept of the data price was literally everywhere one turned in antitrust circles. Two years earlier, in fact, John Newman, a genuine antitrust scholar, made that argument in the University of Pennsylvania Law Review, a top-tier academic journal. Ariel Ezrachi and Maurice Stucke, antitrust scholars at Oxford and the University of Tennessee, respectively, had just published a book saying the same thing. Germany’s antitrust enforcers had undertaken a major investigation of Facebook based on exactly that premise. And a graduate student, Elias Deutscher, was presenting a paper advancing exactly this argument with a great deal more sophistication. No doubt more examples can be found.

But only Srinivasan wrote the idea up as “The Antitrust Case against Facebook.”

Has Srinivasan’s paper in particular met with an unusual level of academic acclaim? No. We are told that she has presented her paper at the “American Antitrust Institute’s annual conference.” AAI is a politically moderate advocacy organization, not an academic forum, and presenting at one of its meetings, while nice, is hardly a high-prestige affair within antitrust circles, let alone a reason to stop the presses. But if you’re not impressed by that (which you shouldn’t be), news flash: the Journal informs readers that Srinivasan “is presenting her work at an international antitrust conference in Brussels this week.” I wish every time I attended an international conference, which I and dozens, if not hundreds, of antitrust scholars do every year, the Journal would write me up.

The paper does little to hide the fact that Srinivasan is no authority on antitrust because the point of this profile is not to report. But to create. And for that all you really need is publicity.

If you aren’t convinced yet that this profile is about advancing the press’s narrow competitive interest in the demise of Facebook, just read the article through to the end. Srinivasan, we learn, is employed again. She is “currently working with The Wall Street Journal’s parent company, News Corp.”


Getting Asabiya into Chess

Ibn Khaldun famously observed that asabiya (social cohesion) explains the arc of history, and Peter Turchin has done a marvelous job of showing how modern statistical analysis supports this view. Peoples caught at the focal points of conflict develop strong social bonds that eventually propel them to dominance, but dominance and the associated lack of threats erodes cohesion, and over time these groups are replaced by new groups forged in the cauldron of conflict.

The game of chess does a great job of modeling conflict. No other game captures the way circumstance (the juxtaposition of pieces on the board) can create positions of great power and then wipe them away in the blink of an eye.

But while individual chess players learn over time from playing the game, improving as their strategies are tested under conflict, the rules of the game themselves do not take asabiya into account.

One small tweak that would take the game in the right direction would be to upgrade pieces based on the amount of pressure to which they are subject from other pieces on the board. In particular, I would suggest that any piece that could be taken by any one of more than three other opposing pieces on the following move be automatically upgraded one level in value.

So, for example, if white’s pawn could be captured on the next move by black’s bishop, rook, pawn, and queen, then white’s pawn could be replaced with a knight or bishop (the two pieces traditionally considered to be immediately higher in value relative to pawns). If the piece subject to attack from those four pieces is a bishop, then white could substitute a rook for the bishop, because the rook is the next level up in power relative to the bishop, and so on.

The idea behind this substitution rule would be to model the way subjecting a particular group to pressure and conflict–placing it at the center of battle–makes the group stronger. The rule might also help solve the problem of boredom in modern chess, by discouraging the buildup of pressure on particular pieces and encouraging capturing and sacrifices for positional advantage, a la the 19th century era of romantic chess.

There are admittedly some shortcoming to this rule as a step toward modeling asabiya. For one thing, the rule doesn’t really strengthen cohesion between the pieces, unless you think of more powerful pieces as being better able to coordinate with others because more powerful pieces have more freedom of movement under chess rules. The rule seems more to model increases in individual strength. But the rule does do a good job of modeling the power-increasing character of nexuses of conflict.

Perhaps a better rule from the standpoint of modeling asabiya would be to allow any piece subject to attacks from four or more adversaries to swap positions with any laterally or diagonally adjacent piece, as a reflection of the way people in proximity to each other work together to repel threats. But that is too complex to be a good rule for chess.


Qualcomm in One Diagram

The district court’s opinion in the Federal Trade Commission’s case against Qualcomm is 233 pages long, but it really comes down to a single diagram. Because the case is really just this: that Qualcomm excluded rival chip makers from the market by refusing to license patents either to rivals for bundling with their chips or to chip buyers so that the buyers could assemble the bundle on their own.

The licenses are for “standard-essential patents,” meaning patents necessary for chip buyers, who manufacture cell phones, to make phones that comply with industry standards. Chip buyers must, therefore, acquire the licenses to function. Qualcomm sold licenses in a bundle along with its chips, but refused to sell them any other way, either directly to cell phone makers who might want to combine them with chips purchased from Qualcomm’s chip-selling rivals (the “no license no chips” policy), or to rival chip makers who might want to bundle them with their own chips and sell them on to cell phone makers. That made the chips supplied by rival makers undesirable to cell phone makers, because the rival chips did not come with patent licenses and the cell phone makers could not obtain those licenses independently.

The diagram describes a supply chain, with patent licenses, which amount to an essential input into the production a usable chip product for cell phone makers, at the top of the chain. Cell phone makers must acquire both this input and chips themselves to function. The second, middle, level in the supply chain is the physical chip itself. Qualcomm competes in this market, and uses its control over the essential patent input to exert power over rivals in this market. Cell phone makers are just as happy buying licenses and chips separately and assembling them on their own, or buying licenses bundled with chips. Qualcomm excludes rival chip makers by making it impossible for either rival chip makers or cell phone makers to acquire the license and bundle it with rivals’ chips.

The diagram reflects this by putting Xs through supply arrows leading from patent licenses directly to cell phone (“equipment”) makers and from chip patents to rival chip makers. Because Qualcomm nixed these supply routes for patent licenses, the only way for cell phone makers to obtain licenses was through purchases of Qualcomm chips. Because cell phone makers couldn’t get the licenses independently or through rival chip makers, they had no reason to buy chips from rival chip makers, wiping out those rivals of Qualcomm in the chip market.

Analogy to Pencils and Erasers

It’s as if a pencil manufacturer that also happened to be the world’s exclusive producer of erasers were to refuse to supply erasers to competing pencil makers and also refuse to sell them directly to consumers. The only way to obtain an eraser for use with a pencil would be to purchase an eraser-tipped pencil from the pencil manufacturer. Competing pencil makers would be unable to compete, because most pencil users want to be able to erase.

Actually, the case against Qualcomm is a lot clearer than would be the case against this hypothetical pencil conduct. What ultimately drives antitrust liability in single firm conduct cases is whether the input supplier’s decisions about how to route supply results in an end product that is ultimately better for buyers than alternatives. The inputs here are the patent licenses, and the question is therefore whether Qualcomm’s insistence on only supplying licenses alongside its own chips made those chips better.

The answer has got to be no: a patent license is just a piece of paper– really, just an idea–the guarantee that the licensor won’t sue for infringement. Combining this bit of ephemera with the chips creates no value greater than the sum of its parts. It doesn’t make the chips run faster, or sip less power, or process more data. It adds nothing at all to the product sold by Qualcomm. Cell phone makers can do just as good a job combining the license with Qualcomm’s chips as can Qualcomm, and rival chip makers can do just as good a job combining the license with their chips as can Qualcomm.

By contrast, the pencil maker might be better at affixing erasers to pencils than any other firm in the business, or indeed than pencil consumers, in which case the pencil maker’s insistence on reserving its entire eraser supply to itself might actually make consumers better off. The pencil maker can defend its decision not to treat erasers as a standalone component and supply it to others as necessary for it to improve its end product. Qualcomm just can’t make that kind of argument.

So Qualcomm should lose on the economic merits. Buyers suffer the harmful consequences, in the form of higher chip prices, that come from Qualcomm’s freezing-out of rival chip makers. But buyers enjoy none of the advantages in the form of improvements in the quality of the product offered by Qualcomm, because no such improvements follow from Qualcomm’s bundling of licenses with chips.

At least, the economic case is clear before you take into account that licenses are not normal production inputs, but rather government-granted rights of exclusivity the purpose of which is to aid the grantee in excluding competitors from markets. Patents fail by design to add anything to the products with which firms combine them. Their whole purpose is to exclude, nothing more. Treating a firm’s decision to license only its own products as monopolization in violation of the antitrust laws simply because the license does nothing to enhance the value of the product would make all exercises of patent rights illegal monopolization.

The Importance of the Standards Context

If a refusal to license patents were all there was to this case, the FTC would surely not have brought it. What makes this case special is that Qualcomm’s patents are essential inputs into chip production only because standard-setting bodies chose to incorporate Qualcomm’s patented technologies into industry standards. This matters, because we respect the monopoly power created by a patent only because we assume that it derives from the fact that the patented technology represents an innovation that competitors have failed to match. We assume that what makes a patent necessary for competitors to be in the market is that the patent covers some invention that makes the product better. The patent, and the monopoly power that flows from it, then serves to reward innovative activity, creating incentives for all firms to innovate and improve their products.

But when a patent is a necessary input because a standards body has incorporated it into an industry standard, we cannot be sure that the patent’s essentiality ultimately derives from the fact that the patented invention is better than alternatives. Standard setting bodies often have a menu of technologies that they may incorporate into a standard, each of which is equally innovative and equally suited to accomplishing a given task. That is to say, standard-setting bodies often choose from a menu of competing technologies, none of which is essential precisely because other technologies on the menu accomplish the same task. The standards body must choose only one technology, however, because that’s the point of establishing a technical standard, and it is that step of adopting the technology for the standard, and not the technology’s superiority or uniqueness, that makes the standard essential. All the firms in the industry must adhere to the standard, and the decision of the standards body therefore eliminates all competing, equally innovative technologies, from the field.

The source of the technology’s essentiality, of its ability to confer monopoly power, is therefore the decision of the standard setting body to use the technology, not the fact that the technology is superior to other existing technologies. That eliminates the basic rationale for which the antitrust laws usually allow patent holders to refuse licenses to competitors. The monopoly power that results cannot, in the case of standards-essential patents, be assumed to reward superior performance. It follows that in this unique context, it is proper to proceed to the next step of asking whether a patentee’s decision to lock up access to its patents does anything to improve the products it sells, and to conclude from the fact that it cannot that consumers are harmed.

The Prior Dealings Wrinkle

While the case against Qualcomm is therefore clear as an economic matter, it is less clear as a legal matter. The trouble is that Chicago School influence over the antitrust laws has restricted the ability of the courts to decide antitrust cases on the economic merits. The Supreme Court’s Aspen Skiing and Trinko cases suggest that a firm’s decision to lock others out of its supply chain (as Qualcomm has done in effectively supplying patent licenses only to itself in its production of bundles of chips and licenses for sale to cell phone makers) can violate the law only when the decision represents a termination of a prior profitable course of dealing.

That is, unless the firm has a history of voluntarily supplying the input (here patent licenses) to competitors, the firm’s decision not to supply the input cannot violate the antitrust laws.

Thus when a termination of a prior profitable course of dealing does exist, the courts are free to get to the economic merits and decide whether the termination made the product better. But when a termination does not exist, the courts must kill the case, even if the termination does nothing for the product, and therefore must harm consumers. So much is true, at least, if the Supreme Court really meant this to be a hard and fast rule of law. Courts see the existence of a termination as a signal that the firm’s refusal to deal might be motivated by anticompetitive intent, and that is in turn some evidence that the move does not improve the product. Rather than parse through all cases looking for product improvements or the lack thereof, courts prefer to devote their attention only to those in which the existence of a termination suggests that a lack of product improvement is likely. The courts assume that the rest of the cases involve benign, product-improving, conduct.

The FTC has tried to get around this problem by arguing that Qualcomm’s contractual commitment to license its patents to rival chip makers, which Qualcomm made as part of its participation in the organizations that set cell phone standards, amounted to a prior course of dealing. The FTC advances this argument even though at the time that Qualcomm made this commitment, Qualcomm never actually licensed its patents to competing chip makers. Indeed, Qualcomm never did license these patents to competitors, the company only ever promised to do so. If that promise counts as prior dealing, then Qualcomm’s subsequent refusal to carry out its promise and license its patents would constitute a termination of a prior dealing. And, argues the FTC, that prior dealing was notionally profitable because Qualcomm’s commitment included a promise only to license at fair, reasonable and nondiscriminatory (FRAND) terms, which means profitable terms.

The FTC’s argument is obviously an abuse of the concept of a termination of a prior profitable course of dealing. Because a promise to deal profitably in the future is not in fact a prior profitable course of dealing. They are different things. But that does not mean that the FTC should lose, only that the rule toward which the Supreme Court gestured in Aspen Skiing and Trinko is not a great rule, and should not be treated as hard and fast by the courts.

Where, as in the Qualcomm case, the essentiality of a patent to competitors results from the actions of a standards body and not from the innovative superiority of the patented invention, and the firm’s refusal to license did not improve the product as an economic matter, and therefore must have harmed consumers, the courts must hold the defendant liable for monopolization in violation of the Sherman Act.

The courts can continue to treat a termination of a prior profitable course of dealing as suggestive of the potential for consumer harm in a refusal to deal. But when the courts encounter a case in which the essentiality of a patent arises by administrative fiat, rather than market forces, they shouldn’t kill the case on account of the absence of a termination of a prior profitable course of dealing.

Which is why the district court got this case right, and the Ninth Circuit should affirm.

(The FTC has also tried to get around the termination of a prior profitable course of dealing requirement by arguing that Qualcomm’s refusal to license directly to cell phone makers is a separate antitrust violation, distinct from Qualcomm’s refusal to license to rival chip makers. While there is some basis in law for treating this conduct (which amounts to tying) as an independent basis for liability for monopolization, I see it as just an indirect way of refusing to sell to rival chip makers, one to which the termination of a prior profitable course of dealing requirement could apply. Licensing directly to cell phone makers is equivalent to having Qualcomm license to chip makers and then having the chip makers ask cell phone makers to pay a portion of the purchase price of their chips directly to Qualcomm. A willingness on the part of Qualcomm to license either directly to cell phone makers or to rival chip makers would amount to a willingness to supply an input essential to rival chip makers’ success, and Qualcomm’s refusal to license through both channels therefore amounts to a refusal to supply an essential input to competitors. This behavior should count, either way, as a refusal to deal with competitors, and be considered under the legal rules governing such refusals.)


Is Go the Ultimate Antitrust Boardgame?

Scsc, Gofin, (Ramsi Woodcock added arrows and text to the image.) CC BY-SA 3.0

In the ancient game of Go, players compete by placing stones on a grid board with the goal of enclosing the greatest amount of territory with their stones. Think of the board as a two-dimension product space. For breakfast cereals, for example, one axis might be sweetness and the other crunchiness. The goal of a firm is to field a product that has few close substitutes, that faces competition from products that are quite unlike it, and therefore unlikely to lure away consumers of the firm’s product, allowing the firm to raise prices without worrying about triggering a reduction in demand. The goal of a firm, therefore, is to reserve the largest possible extent of product space to itself. But that’s just what a Go player seeks to do: to exclude his opponent from the largest possible expanse of the Go board.

We can think of the stones that a Go player places on the board as differentiated products. A stone’s location on the board corresponds to a particular combination of the two product attributes represented by the Go board as two-dimensional product space. In the cereal example, a stone placed at the center of the board would represent a breakfast cereal that is half sweet and half crunchy. A stone in one corner might represent a cereal that is very sweet and not crunchy at all.

At the start of a game of Go, players place their stones in separate areas, trying to stake out control of whole quadrants of the board, just as firms strive to introduce products that have no close substitutes, ensuring that consumers will be forced to buy the firm’s product at high prices (an animated picture of how a game of Go develops may be found here). We can think of each node in the Go board grid as representing a consumer with a preference for a product with attributes that correspond to that node. If no product has precisely those attributes, the consumer will purchase the product with the closest set of attributes, meaning the product represented by the Go stone on the nearest node to the consumer’s. Placing a stone in an empty quadrant of the board means that the player will get the business of all of the consumers in that quadrant. (Those familiar with the Hotelling location model of differentiated product competition will recognize that this account of Go is just the Hotelling model expanded from one dimension to two. We’re not talking about pizza places along the line that is the Coney Island boardwalk, but pizza places on the grid that is midtown Manhattan.)

Now, firms pioneering a new technology may decide to stop innovating, stop introducing new product flavors, and just coexist in oligopolistic harmony. One of the extraordinary things about Go is that, like firms, the players can stop the game at any time. Whenever both players agree to stop, they count up the territory enclosed by each and the player with the largest territory wins. Just so, when firms collude, the firm with the most desirable product–the product that controls the greatest share of consumer demand–will profit most from the oligopoly.

When to keep fighting and contesting competitors’ markets, and when instead to try to respect competitors’ markets and seek accommodation, is an important question in business strategy. Just as it is in Go. At the start of the game, both players usually believe they can achieve advantage through continued play, and so the game goes on. Competition at the inception of an innovative technology, when a new product space opens up, is often fierce. Think of the competition in streaming video today.

Firms take the battle to each other by striving to make very close substitutes of their competitors’ offerings, cutting into demand for competitors’ products and increasing demand for their own. So it is in Go, too. The players invade each others’ quadrants by placing stones right next to their opponents’, just like firms trying to siphon off rivals’ demand. At the extreme, a firm may create a set of products that so well captures all of the attributes that consumers value in a competitor’s product that demand for the competitor’s product eventually withers and the product is driven from the market. In Go, that happens when one player has placed stones that fully encircle a stone belonging to his opponent. When that happens, the opponent’s stone is removed from the board and the player gains a point. We can think of that point as representing profits enjoyed from having permanently locked up the demand of the consumer represented by the strangled node.

Firms sometimes succeed by driving competing products from the market. But they also often succeed by building up a wall of differentiated products that competitors cannot penetrate. Indeed, the FTC famously alleged in the 1970s that Kellogg did precisely that with breakfast cereals by proliferating the number of cereal flavors that the firm offered. In Go, this is reflected in prolonged struggles between the players to wall off territory (text in Wikipedia accompanying the featured image of this post provides a nice overview of how players control space through these struggles).

Differentiated product competition is the only kind of competition that we have in the real world. If Go is a game of differentiated product competition, does it tell us anything about how to regulate real world competition through antitrust?

Play it and find out.


Ethno-Nationalist When We Want to Be

John Broich’s explainer on the Kurdish question is a good example of the contradictions of contemporary American Kurdophilia. He seems to lament the failure of the Kurds to construct what he admits would be an ethno-nationalist homeland out of the ashes of the Ottoman Empire, even though America today is built upon a rejection of ethno-nationalism of all kinds. As I have observed before, it’s easy to advocate self-determination for peoples abroad, but a lot harder to do it at home, because self-determination weakens and fragments. That makes it good foreign policy and bad domestic policy, at least in the short run, but that also means that advocates of Kurdish statehood don’t have principle on their side.

Broich seems to think that it follows naturally from the fact that the Kurds are “a group of around 40 million who identify with a regional homeland and common historical background, but are now divided between four countries,” that they ought to have their own country.

But I rather doubt that he would support calls by white nationalists to carve an independent white homeland out of the northwestern United States, calls by black nationalists to carve an independent black homeland out of the United States, or calls by Native Americans to carve an independent Native American homeland out of the United States. Or indeed calls by blue staters to secede. Carving up the United States would surely eliminate the region’s current global military and economic dominance.

The fact is that if we believe in democratic pluralism at home, then we can’t try to protect oppressed groups abroad by supporting their calls for statehood, either diplomatically or militarily. The best we can do is support their calls for democracy and equal treatment within whatever countries they happen already to belong. At least, that’s the best we can do if we want to act toward them in a way that is consistent with the way we treat ethno-nationalist aspirations here at home. (Of course, we might not want to run our foreign policy based on consistency and principle, but that’s not how America’s advocates of Kurdish statehood have been making their case.)

Broich observes that the failure of the allies actually to create an independent Kurdistan after World War I resulted largely from European self interest. The British and French were themselves worried that hacking Arabia into too many pieces would make it difficult for both to maintain their spheres of influence in the region, so they scrapped plans for Kurdish self-determination. But the fact that the Kurds lost their chance at statehood because of European self interest doesn’t mean giving them a state would have been good for the region, or consistent with the principles according to which we organize our own country today.

Broich’s unreflective observation that Woodrow Wilson “himself was explicit in calling for a new, broadly encompassing Kurdistan,” sums up the contradictions in contemporary American advocacy of Kurdish statehood. For Wilson, of course, surely believed in white ethno-nationalism for America, and famously segregated the federal government.

At least he was consistent.


Rewards Not Ours to Give

What I do not understand about all the criticism of President Trump’s abandonment of the Kurds is why, exactly, the Kurds should be entitled to a state in northern Syria. I had always thought that carving up sovereigns and doling out territory to favored groups was by general agreement consigned to the dustbin of imperialist history after World War Two. Especially by us Americans, with our relatively anti-colonial past.

But that’s exactly what anyone lamenting President Trump’s withdrawal from Syria seems to be calling for: that we back, militarily, the attempt of a particular group to carve its own ethnic homeland out of an existing UN-recognized country.

In other words, while other Syrians were fighting the Assad regime to create a more democratic, tolerant Syria, the Kurds were fighting the Assad regime to grab land for themselves. That’s hardly the sort of democratic behavior we normally think of ourselves as supporting. Yes, the Kurds helped us fight the Islamic State, but Syria isn’t ours to carve up and dole out to our allies like so many slices of reward cake. Yes, the socialist Kurdistan Workers’ Party has a women’s movement, but again, Syria isn’t ours to carve up and dole out to socialist women’s movements like so many slices of reward cake, especially when we’re not (yet) voting socialist here at home.

And while I’m on the subject of the beatification of the Kurds, I wish to note the irony of the House’s recent rebuke of Turkey for attacking the Kurds in Syria by voting to recognize the Armenian mass killing as genocide. For, as any Armenian will tell you, the Kurds played an important role in carrying out that genocide a century ago.

President Trump made the right call on Syria.


The Antitrust Laws Give Chicago Cabbies a Remedy

The New York Times is reporting that New York taxi companies bought hundreds of Chicago taxi medallions–enough to jack up medallion prices–and then turned around and sold them to cabbies, earning monopoly profits. Many cabbies borrowed to buy the medallions and have gone bankrupt.

If the New York companies’ purchase of large numbers of medallions really did give them the power to raise prices, then those companies achieved monopoly power for purposes of the antitrust laws, which define it as the power profitably to raise prices.

It’s not illegal to charge monopoly prices. But it is a violation of Section 7 of the Clayton Act to assemble a monopoly through asset acquisitions, such as the purchase of taxi medallions.

The U.S. Department of Justice and the U.S. Federal Trade Commission have the power to challenge anticompetitive acquisitions after the fact. And they can seek disgorgement of profits, which here would mean an order requiring the New York companies to compensate cabbies for the inflated prices that they paid. The article suggests that in some cases those prices may have been eight times the competitive prices, or $350,000 in monopoly excess per medallion.

The Times suggests that much of the alleged conduct took place six or more years ago. But for civil antitrust actions brought by the government, there is no statute of limitations.

It may also be possible to challenge this conduct as fraud, or market manipulation. But this is a case of the raw acquisition of monopoly power, without any semblance of an efficiency justification, placing it squarely within the core of the antitrust laws. It would be nice to see those laws earn their keep here.

And to see antitrust enforcers return to policing local and regional monopoly power.