When it comes to self-preferencing, right wing economists are right: incentives matter.
Last week, the House Judiciary Committee spent 23 hours debating a historically unprecedented package of tech competition bills, surprising observers by passing all six of the legislative proposals under consideration, with bipartisan support.
Each of the six bills is interesting in its own ways — for example, the ACCESS Act (HR 6487) uses interoperability and standards to reduce the costs we bear when we leave monopoly platforms behind, by letting us stay in touch with the friends who stay.
But while ACCESS uses a digital solution to tackle digital problems, another bill, Pramila Jayapal’s Ending Platform Monopolies Act (HR 3825), hearkens back to the golden age of trustbusting, by breaking up large platform companies and forcing them to divest of subsidiaries that compete with the companies that rely on their platforms.
The philosophy here is simple: a platform makes a marketplace, sets its rules, and mediates between buyers and sellers to keep everyone honest. If the platform also competes with the sellers who rely on it, they will be tempted to cheat, to engage in “self-preferencing” that pads the company’s bottom line by steering its users to buy or use its own products instead of offerings from rivals that are cheaper or better (or both).
This is called “structural separation,” and it was once a bedrock of antitrust law. We once banned railroads from operating freight companies that competed with their own customers and prohibited banks from owning businesses that competed with the businesses it issued loans to. After all, a railroad’s decision about what to charge for freight determined what could be profitably bought and sold, while a bank’s decisions about interest rates, as well as its willingness to work with borrowers who come up short some months, determined what businesses could operate in its community and how fat their margins had to be in order to stay in business.
Structural separation wasn’t a measure of first resort; before it was tried, lawmakers and regulators tried to keep the system honest through rules that required large, structurally important companies to operate fairly and not tilt their practices towards their own shareholders’ interests and against the public interest.
This was a failure on two counts. First, it‘s too easy to cheat — in a big, complex platform business, there are lots of ways to nominally stay within the rules while still screwing over your suppliers. A good example can be found in Ida Tarbell’s seminal History of the Standard Oil Company, a piece of astounding, long-form muckracking journalism that brought down John D Rockefeller’s global energy monopoly (I’ve written about this book here before).
Breaking up Standard Oil — a conglomerate that controlled oil refining, shipping and marketing, as well as pipelines and railroads — was not Plan A for American regulators. Instead, they imposed rules of fair play upon Rockefeller, such as a rule that said that the railroads he controlled had to charge the same price to ship oil from refineries to ports, no matter whether the refinery was owned by Rockefeller or his competitors.
Rockefeller, an inveterate and shrewd cheater, figured out how to obey this rule while still tilting the playing field in his favor. You see, Rockefeller controlled all the oil that was refined in Chicago, while his competition was clustered in Cleveland. He also controlled the railroads that linked these two cities to the port of New Orleans, a center of global oil shipment.
On Rockefeller’s railroads, it cost about the same to ship lumber, steel, flour or other goods from Chicago to New Orleans as it did to ship them from Cleveland to New Orleans. But when it came to oil, the freight rates diverged sharply: it cost about 50 percent more to ship oil from Cleveland (where Rockefeller’s competitors had their refineries) than it did to ship oil from Chicago (whose refineries were controlled by Rockefeller):
Not only are outside refiners at just as great disadvantage in securing crude supply to-day as before the Interstate Commerce Commission was formed; they still suffer severe discrimination on the railroads in marketing their product. There are many ways of doing things. What but discrimination is the situation which exists in the comparative rates for oil freight between Chicago and New Orleans, and Cleveland and New Orleans? All, or nearly all, of the refined oil sold by the Standard Oil Company through the Mississippi Valley and the West is manufactured at Whiting, Indiana, close to Chicago, and is shipped on Chicago rates. There are no important independent oil works at Chicago. Now at Cleveland, Ohio, there are independent refiners and jobbers contending for the market of the Mississippi Valley. See how prettily it is managed. The rates between the two Northern cities and New Orleans in the case of nearly all commodities is about two cents per hundred 2278pounds in favour of Chicago. For example, the rate on flour from Chicago is 23 cents per 100 pounds; from Cleveland, 25 cents per 100 pounds; on canned goods the rates are 33 and 35; on lumber, 31 and 33; on meats, 51 and 54; on all sorts of iron and steel, 26 and 29; but on petroleum and its products they are 23 and 33!
In other words, the US government told Rockefeller that his railroads couldn’t charge his competitors’ refineries more than they charged his own, and technically, they didn’t. Anyone who wanted to ship oil from Chicago would enjoy the Rockefeller discount, while anyone who shipped from Cleveland had to pay the competitors’ penalty. The fact that every refinery in Chicago was run by Rockefeller (and the fact that Rockefeller didn’t have any refineries in Cleveland) allowed Rockefeller to offer a “universal” rate while still making it impossible for his rivals to turn a buck.
This was just one example among many of the technical loopholes that lurk within any set of rules meant to allow a company honest while it operates a platform while competing with its own business customers.
But even if it were possible to craft a bulletproof and airtight set of rules against self-preferencing, there’d still be a reason to demand structural separation: credibility.
We don’t let referees make bets on the outcomes of the games they officiate. We don’t let lawyers represent both the plaintiffs and the defendants in cases. We don’t let judges weigh in on cases involving their family members.
It’s conceivable that we could create rules that would keep these parties honest, but if the ref calls a foul for your team and also has bet his life’s savings on the other team, could you every truly believe that the call was fair? When a judge rules against you and in favor of the guy who’s suing you — who is also the judge’s husband —could you ever trust the outcome?
The legitimacy of a system is inextricably bound up with its perceived fairness, and that perception of fairness is fundamentally incompatible with conflicts of interest.
This is especially true of systems that require human judgment. For example, bank officers have always made case-by-case determinations about when a missed loan payment should trigger a demand for full payment, and when that missed payment should be tolerated as the result of a temporary blip in the borrower’s finances.
But when the bank forecloses on your business, and spares a rival business that the bank also owns a large share in, would you ever really believe that outcome was driven by the manager’s sincere belief that your financial woes are insurmountable and your competitors are merely temporary?
In the recent debate over the difficulty in finding low-waged employees willing to do service work, you may have heard an economist blame the whole thing on “moral hazard,” a favorite idea from right-wing economic theory.
A moral hazard occurs where a system pays cheaters more than it does rule-followers. The classic example is an insurance policy that means that your house is worth more if you burn it down than if you continue to live in it. This commonsensical notion is weaponized to demand limits on social safety nets: paying sub-starvation welfare and unemployment benefits in the name of fighting “laziness,” or charging a co-pay every time you visit the doctor to prevent “frivolous” medical concerns.
This cruel doctrine’s rallying cry is “incentives matter” — if you “incentivize” a person to stay home and play video-games by setting unemployment benefits higher than the minimum wage, then the right fears that no one will work.
But the doctrine of moral hazard has always been highly selective. Capitalists who fret about workers choosing time with their families over flipping burgers or driving Ubers focus entirely on creating disincentives to opt out of low-waged work (threatening refuseniks with eviction and starvation, say), they are seemingly immune to the blindingly obvious idea that if you want workers to choose jobs over relaxing at home, you could pay more.
A moral hazard occurs where a system pays cheaters more than it does rule-followers. The classic example is an insurance policy that means that your house is worth more if you burn it down than if you continue to live in it.
Incentives matter when it comes to CEO pay — handing already wealthy executives hundreds of millions in stock options that rise with the value of the company they helm — but when it comes to workers, only disincentives matter.
The same neoliberal economists who demand millions to lure CEOs off the golf-course and into the board-room (and who demand punishments to get workers out of their family homes and behind the wheel of an Uber) have a similar double-standard for monopolists, who, in neoliberal economics, are deemed to be “efficient” (in neoliberal thought, markets destroy inefficient businesses, so any large business must be efficient or it wouldn’t be so large, just as any pile of manure large enough must have a pony underneath it).
They assert that no platform monopolist would unjustly preference its own offerings: if Amazon recommends its house-brands over the original products it cloned, if the App Store recommends Apple Music over Spotify; if Google sells an advertiser space on YouTube instead of a rival service; then this must be fair. These platform operators have weighed their own offerings against their rivals and found them to be objectively better, and that’s why their algorithms have promoted them to you over competing products.
But incentives do, in fact, matter. The idea that the leaders of large businesses are so far-sighted and prudent that they would never cheat has been disproven time and again: when Equifax’s executives discovered that they had incurred billions in liability by leaking the financial details of every person in America (and millions abroad), they didn’t leap into action to repair the breach. Instead, they covered up the news, left the vulnerability open for continued exploitation, and sold off their stock. When banks learned that the mortgage-backed securities they were selling were garbage, they didn’t stop selling them — instead, they doubled down on selling them to their customers while betting against them.
It’s not just CEOs who can’t be trusted not to cheat; we also can’t trust mid-level employees who are incentivized to cheat by systems that dangle rewards for rule-breakers. It really seems like Jeff Bezos thought he was telling the truth when he told Congress that Amazon doesn’t spy on the businesses who use its service to decide when to knock off someone’s product and sell it cheaper — but he was wrong. The bonuses and promotions of product managers who are in charge of Amazon’s house brands depend on that kind of data, and simply banning the practice is no guarantee that it won’t take place. Incentives matter.
Likewise, the banks whose traders cost cities and states around the world trillions by rigging the Libor rate to pocket a few millions for themselves: these traders were “rogue,” not acting on executive orders, but the fact that executives built a system that rewarded corrupt behavior with vast fortunes makes them culpable in the outcome.
Between the incentive to cheat and the complexity of catching cheaters, it’s impossible for a large, powerful business to both referee the games it creates and participate in it.
Even seemingly straightforward questions grow impossibly gnarly when examined in detail. Take Google’s ad-tech stack, in which advertisers bid to place ads; publishers bid to attract ads, and Google acts as broker, market-maker, and customer for ads. There are some ways in which Google is definitely cheating in this system, but even if the overtly dirty self-dealing was purged from the Google ad-tech exchanges, it would be hard to know whether the system was playing fair.
You see, when Google runs an ad-placement auction on behalf of a publisher (where different advertisers bid to place an ad in front of a specific user) the system doesn’t necessarily take the highest bid. Some bids are algorithmically discarded or given lower priority based on the system’s predictions about whether they are scams, or merely poorly targeted. If you’re a publisher relying on Google to place ads on your pages, will you ever truly believe that the reason Google placed one of its own ads on your site and threw out an ad that would have paid you twice as much (but made much less for Google) was because Google objectively determined that it was a better ad?
These subjective judgments crop up all the time in self-preferencing debates: if a Google search puts a YouTube link at the top of the screen instead of a link to a video hosted on a rival service (Google owns YouTube and makes money when you watch videos there), how can you be certain whether the YouTube result was there because it’s the best match for you, and not merely the most profitable match for Google?
The more you think about this, the harder it becomes to know whether the system is credible and legitimate. When Google shows you its own weather results rather than Weather.com’s, how can you tell if that was a fair ranking? Given that both services have the same raw weather data, how can you definitively say which one is the “best weather result?”
You can’t, just as you can’t ever truly trust that a judge can fairly hear a dispute between you and her husband, or that a referee can fairly call a match that he’s bet his life’s savings on. It’s entirely possible that a judge or a ref could render fair judgments under those circumstances, but any time they rule in a way that serves their own interests, it’s impossible to tell whether they’re rigging the game or calling it fair.
This is why Congress hit on structural separation to begin with: for the system to run, it must be legitimate. To be legitimate, the large companies who operate its critical infrastructure have to choose between being a market and selling in that market, between being a ref and having a stake in the game’s outcome.
There are serious challenges to structural separation. The Big Tech companies will argue that it’s impossible to say what part of their business is the “platform” and what part is the “platform user” — and to be fair, these questions are often thorny ones.
But the alternative question — how do you trust a system whose scorekeeper has a stake in the outcome? — is much thornier. Creating a legitimate system requires honest refs, and as hard as it will be to figure out how to carve up the tech platforms, it’s impossible to figure out how to keep them intact and still have a legitimate system.