Minimum advertised price (MAP) policies are becoming more and more popular. Especially with the rise of cutthroat competition in the online marketplace, many producers are hearing from their resellers that they want margin protection, and are getting concerned with the possible erosion of their brand equity from widespread promotion of low pricing. Naturally, many companies have turned to consultants and the internet for information on MAP policies.
On July 6, 2017, QVC announced its intent to acquire the remaining 62 percent of Home Shopping Network it doesn’t already own. More so than DraftKings or Walgreens, this transaction will demonstrate whether Trump’s election has had any effect on antitrust enforcement, and should be watched carefully. HSN and QVC are very similar, and the ability to do the deal will turn on what product market definition wins the day. A very broad product market definition, that focuses on means of distributing products to consumers, and that includes the Internet, will suggest the transaction will have very little effect on competition, and should be allowed to close. A narrow product market definition that focuses, say, on television shopping as a unique form of entertainment to consumers, and therefore a unique channel in which to sell products, may very well result in a challenge.
The European Commission has fined Google €2.42 billion ($2.7 billion) after concluding the company had abused its dominance in search. According to a letter from the EC announcing the results of its investigation, Google has a dominant market position in search, and the company leveraged that dominance to give itself an unfair advantage in comparison shopping search. For example, according to the EC, when a consumer types in the name of a particular product, Google favors its own comparison shopping search results to those of its competitors. The EC believes this process “denie[s] European consumers a genuine choice of services and the full benefits of innovation.”
As of December 31, 2016, Google (Alphabet) had total assets of $167 billion. The fine represents slightly more than 1.6 percent of its total assets. Perhaps more important to Google and to other companies within the reach of the EU is are the legal and behavioral aspects of the decision. First, the decision finds that Google search is “dominant,” meaning that it is the predominant means by which consumers search the internet. In the EU, a dominant company is required to behave in a competitively neutral fashion as between itself or its affiliated companies and its rivals. In the States, a dominant entity (or “monopolist”) is required only not to behave in a way that unfairly maintains or expands its advantage or disadvantages competition. What constitutes unfair in the States is more limited than in the EU. For example, a United States monopolist is under no duty to deal with rivals except in circumstances that suggest its conduct has no legitimate purpose. By contrast, in the EU, a dominant company is expected to be evenhanded in its dealings whether it is dealing with its rivals or with companies in other markets.
The EC is not the authority of last resort. Google can, and most likely will, appeal. It may also decline to comply fully with the EC’s decision and force adversarial proceedings.
The case is significant for Google because the EU is demanding changes in the way Google displays its search results. As one of the first and still leading search engines, Google’s product enjoys widespread popularity. The EU is now telling Google that its product should give users a different experience. And for the foreseeable future, when Google enhances the product, the changes could be second-guessed by a regulator. While the case purportedly affects conduct only in the EU, it raises the question whether the company would or could offer different search products in the EU and the rest of the world. For multinational companies operating online, it is difficult to avoid crossing borders.
Issues for Multinational Companies
A big question in the case is where and with whom Google competes. For shopping-related searches, Google argued that it has robust competition with Amazon, Ebay and other comparison shopping tools, and that this competition intensified even in the two years that the EC has been investigating Google. On Amazon, a customer can locate a product quickly, pay for it using credit cards Amazon knows, and buy it from any number of different vendors. The same is true for eBay.
Parties with significant market shares need to be vigilant about how they are perceived in the market, particularly in the EU. While being first and foremost can be a badge of honor and an explicit business goal in the States, a moniker of dominance can trigger significant interest in the EU. That interest can lead to investigations and ultimately regulation.
Three main lessons:
One, consider avoiding behavior that the Europeans consider to be exclusionary in the first place. And consider the pros and cons of working with competitors that have incentives to air grievances. That will blunt their complaints and allow you to control the narrative.
Two, consider your reputation. Discourage internal and external marketing pieces that say you are dominant. While that might be great for securing initial funding, you can cause yourself significant problems with that puffery later on.
Three, develop solid economic evidence of real competition from platforms and other sources that aren’t “classically” within your wheelhouse. You want to be able to muster evidence that competition is coming from lots of places: actual evidence that will be compelling to antitrust agencies worldwide. Charge your antitrust counsel with keeping that data current. Charge your marketing folks with tracking and memorializing that data. And keep economists on call. Periodic assessments of competition will be more compelling to enforcers than something put together at the last minute.
Finally, follow the enforcers and what they are saying, so you can respond to their specific theories and thoughts as they occur and before they become ingrained. We will follow them as well.
Federal Trade Commission Acting Chairman Ohlhausen and Commissioner McSweeny recently gave two very different speeches on algorithmic pricing. Commissioner McSweeny’s speech seemed to express concern that algorithms would lead to price fixing, coordination and higher prices. The Chairman seemed less concerned. The Chairman is likely correct. Pricing bots are not a new, sinister specter haunting society like the trusts of the late 19th Century. Nor are they boogeymen hiding under your bed waiting to raise price and siphon off what little money you have left after rent, clothing and food. Pricing bots are digitized, more efficient versions of the same thought processes that any rival engages in to set price. Because they are digital, they are not constrained by volume. They can take into account significantly more variables than the human mind is incapable of keeping track of—and do it much faster. Using a bot to set price unilaterally is not an antitrust problem. It is the essence of the free market. By contrast, sharing your algorithms, your pricing formulae, with competitors can very well be a problem because it can lead to collusive behavior, like any other exchange of competitively sensitive business information. Setting up encrypted, digital communications systems to suggest and agree on price increases and market allocations is easier today, but it’s still illegal. And it’s sufficiently difficult to do now that a trove of evidence to prove its existence and illegality will be available.
Generally, it’s important to remember that if a market is otherwise competitive, a pricing bot cannot make it less competitive. If a firm can raise price after implementing a pricing bot, it means that the firm was not maximizing profits because it lacked sufficient information. In short, in the pre-bot days, the costs of additional information outweighed the expected benefits of additional information. The bot reduces the cost of additional information. The fact that it raised prices (or lowered them) does not tell us anything about the state of competition in the market.
Pricing bots cannot be coded to “agree” to raise price absent the intent of the human coders. Coders must specifically design their systems to seek out competitors, determine collusive prices and police those prices. Given the highly disparate systems and price setting methodologies that exist across even a commoditized market, the amount of time and effort needed to create such a system is vast. Extrinsic evidence of those efforts must exist. That will include all of the normal evidence of collusion. For a more thorough discussion of what exactly a price-fixing bot might look at, please see our discussion on price fixing bots.
The rest of this brief addresses specific issues raised in each of the speeches.
Commissioner McSweeny Algorithms and Coordinated Effects May 22, 2017
|Algorithms may make price fixing attempts more frequent and potentially more difficult to detect.||Price fixing conspiracies are unnecessary in concentrated markets. Parties can simply act in parallel. Price fixing conspiracies do not work in atomistic markets because identifying and disciplining cheaters is too difficult. Somewhere in the middle are markets that have just enough participants to make parallel behavior difficult. For example, there may not be an efficient way for everyone to follow the behavior of others. What a bot can do in those markets is eliminate the problems with the communications and policing that make parallel behavior difficult. In this regard, a bot could increase the maximum number of participants that can successfully act in parallel because it makes the process more efficient.
The more harmonized pricing is, the more likely the product is a commodity. The more likely the product is a commodity, the more likely the market will have too many participants to act in parallel or collude. A price fixing bot will therefore only work in markets with a small number of participants. How a given firm prices can be highly idiosyncratic. Most firms in such markets will not share the same inventory and sales software. Creating a piece of software that will set price uniformly across a group that is sufficiently small enough to collude is going to require a lot of work by a lot of people. The more individuals aware of a price fixing conspiracy, the more likely evidence the conspiracy exists.
A price fixing bot comes up with a supracompetitive price the conspirators can charge for their product. It is basically the digital version of the smoke-filled room. Knowing how the conspirators arrived at a collusive price is not necessary to prosecute a price fixing claim. You don’t need to know what the code of the bot is to establish that the parties who coded it were fixing price. And given the number of people needed to create a price fixing bot, you are going to have a lot more evidence if they are creating and relying on a bot than if they had just sat down and decided on a price in that smoke-filled room. In this vein, a conspiracy implemented and policed through software is going to leave a much bigger footprint than one accomplished the traditional way.
Here, the Commissioner’s assertion is inaccurate. Coding an industry-wide price fixing bot is time consuming, reads too many people into the scheme, leaves a mammoth paper trail, and ultimately will have little effect on the ability of an end-user to figure out that it’s paying more for the product than it would otherwise. Bots do not necessarily make price fixing attempts more frequent and potentially more difficult to detect.
|Pricing algorithms . . . may facilitate coordinated interaction—sometimes called tacit collusion or parallel accommodating conduct.||True, but only in a very narrow number of markets. Coordinated behavior is only possible in concentrated markets. If a market is atomistic, a price increase by one or several participants will only result in immediate diversion to discounters. Being able to see a rival’s prices quicker and more broadly will not change that fundamental dynamic. In a concentrated market, parties may act in parallel because there isn’t a sufficient number of participants to constrain their activity. A bot won’t change that situation either. There may be some markets where there are just enough participants to make parallel behavior difficult. That may be because there is insufficient ability to detect and police behavior. A bot, combined with a good system of announcing prices to the market, could make it easier for participants to act in parallel in those markets where they couldn’t before. Here, though, all the bot is doing is making detection and policing more efficient. The market still must be sufficiently concentrated for the parties to act in parallel. As such, the bot is no more inherently illegal than any other method of disseminating and reacting to information.
In analyzing whether a price increase is potentially a consequence of illegal collusion facilitated by a bot, one must exclude the possibility that the sellers lacked sufficient information about the price sensitivity of their customers to charge the appropriate price. It is entirely possible that a particular product could have significant value to an as-yet unidentified group that is willing to pay more for that product. The bot could very well have facilitated the discovery of that smaller discrimination market. Pricing more to that group does not reduce consumer welfare and is not an antitrust issue.
A pricing bot could be used to signal prices to other competitors by sending them potential price increases before they are presented to the public. This is Airline Tariff Publishing, however. The parties would be using the bots as a means of proposing and responding to target prices. This is price fixing; it is not parallel behavior.
It is inaccurate to broadly condemn bots. One must examine the facts of any given market to assess the effects of a bot. Only in concentrated markets where the bot is intentionally designed to set a common price among participants is it illegal. And then, due to the complexity of creating such a bot, there will be plenty of evidence.
|Multiple competitors might use algorithmic pricing software offered by the same company.||This is a hub and spoke conspiracy where the manufacturer of the software has coded it to set prices uniformly for all members of a market. Again, however, such a program will only work in a concentrated market. If a program raised price in an atomistic market, customers would divert to the discounter. If everyone used the same bot—an improbable hypothetical—at least one participant would see an opportunity to capture sales and turn off the bot. Again, the bot may make oligopolistic pricing easier because it makes policing easier, but at some point, there will simply be too many participants for such a scheme to work.|
|Firms’ nominally independent algorithms may simply gravitate collectively towards higher prices on their own.||This only works in concentrated markets where customers cannot readily switch to punish a price increase. As such, the sellers in this market must not be pricing optimally; there must be an informational asymmetry of which customers were taking advantage. It is also possible that the algorithms have made seller information more transparent such that the maximum number of sellers capable of behaving in parallel is increased.|
|Pricing algorithms can be “an effective tool for tacit collusion” with the potential to lead to near-monopolistic pricing.||True. The use of the words “collusion” and “near-monopolistic pricing” suggest that this pricing algorithm can be illegal. For the reasons stated above, that is not accurate.|
|The model assumes that firms are able to “decode” their competitors’ algorithms. [The author] included a specification in which firms were given an option to mask their algorithms to prevent decoding. The firms in the model chose not to exercise the option…||An algorithm is essentially a mathematical or logical function. Price is the output of that function. One can display that price as a static number. One need not display the terms of the function to display the price. The only way displaying such a formula would be profitable to a company is if the market were sufficiently concentrated that its rivals could use the formula to increase their own prices. If the algorithms were made available only to rivals, one could argue this was an illegal information exchange—as it would be if the parties shared their actual cost structures with rivals in any other media. The concern with the information exchange is less if the information is also shared with customers. But presumably the only condition in which a seller would share with both is if the consumer had insufficient choice to defeat the potential price increases.
It is therefore incorrect to conclude pricing bots will necessarily lead to “near-monopolistic pricing” in all markets. They can facilitate parallel behavior in near-concentrated markets by making the observation of and reaction to rival activity easier.
I do not think you could draw even this conclusion from this paper, however. There is a difference in the output of the algorithm and the terms of the algorithm. The former is a number; the latter is the method by which a firm arrives at the number. By making the content of the algorithm discoverable, the author is in effect permitting the parties to engage in price fixing. The information exchange is so thorough that the parties have no doubt as to the behavior of rivals and can act in concert. His conclusion is therefore a tautology: markets which are sufficiently concentrated that price fixing can work will produce higher prices if the parties are allowed to engage in price fixing.
|One gas station operator candidly told the Journal that is decision to use the software was promoted by the effects of a years-long price war with its competitors. . . . [W]ithout more information, it’s hard to know whether the reported higher margins are the result of coordinated effects.||A merger that results in a market becoming sufficiently concentrated, such that the participants can behave in a coordinated fashion where they could not before, can potentially violate Section 7 of the Clayton Act, which prohibits mergers or acquisitions that substantially lessen competition or create a monopoly. Unilateral conduct that improves understanding of how a market operates is not a violation of Section 7. Nor is it a violation of Section 2.
It is entirely possible that the algorithms the gas station operator deployed “knew” its customers a lot better than the operator did. The software could know, for example, that consumers that purchase gas at 700a on a Monday are likely on their way to work, need gas to get there and don’t really care how much they have to pay to get that gas. Consumers who purchase gas at 1100a on a Saturday likely do care what they are paying. It is perfectly acceptable to charge the Monday morning commuter more for gas because he values that gas more.
|The third concern with pricing algorithms is that they may enable price discrimination strategies that lead to higher prices for certain groups of customers.||This is the essence of big data. It enables sellers to discover ever smaller price discrimination markets. Charging a person more because his utility for the product is higher is neither inefficient nor an antitrust violation.|
|It works just as well for customers who care very much, but are nonetheless willing to pay a higher price because they lack the practical ability to go elsewhere.||If a customer “lacks the practical ability to go elsewhere,” the subject product has a narrow geographic market and that market is concentrated. The bot has nothing to do with that fact.|
|Pricing algorithms will undoubtedly lead to an increase in price discrimination. Whether that is a good or a bad thing for consumers is likely to depend on facts that are specific to individual markets and individual algorithms.||No, it’s economic efficiency. Consumers that would pay more but do not are free riders: they are exploiting an informational asymmetry between seller and buyer. In this case, it benefits a buyer. But that does not mean eliminating that inefficiency is harmful to society.|
|If algorithms enable firms to “solve” their unique prisoner’s dilemmas without resorting to overt collusion, that would be great news for them but bad news for consumers.||The purpose of the antitrust laws is to create a prisoner’s dilemma: they prohibit communication that would otherwise allow rivals to set supracomeptitive prices to consumers. Bots do not necessarily create new, undetectable and unassailable communications flows between competitors. What they can do, in very limited circumstances, is enable sellers to realize that there are categories of customers that will pay more for their products. In essence, they can eliminate an informational asymmetry that allows those consumers to free ride off the ignorance of a seller. The antitrust laws were not designed to foster this type of free riding. Indeed, this reading of the antitrust laws would cause them to lock in seller-side inefficiencies.
This statement assumes all algorithms are the same. It is important to understand what the “algorithm” is doing. If sellers are sharing the model by which they arrive at a price with their competitors, they are engaging in a potentially illegal information exchange. If sellers have eliminated inefficient pricing to consumers who are willing to pay more through the use of bots, the sellers are engaging in perfectly legal, and efficient, unilateral behavior. If a group of sellers have harmonized their pricing software to take input from competitors and respond to it by raising price, those sellers have automated an illegal price fixing scheme and are potentially guilty of a criminal violation of Section 1. If your market is sufficiently concentrated such that you can set price based on a competitors’ price, and you create a bot that will scan that publicly available price and set your price accordingly, you are automating parallel behavior which is not illegal under the antitrust laws.
|[I]t would be helpful to understand whether algorithms are resulting in coordinated effects and, if so, under what conditions.||
Pricing bots will not affect concentrated or unconcentrated markets. There may be a small class of markets that could behave in a concentrated fashion that does not because the technology is not sufficient for the actors to detect and police price movements. In that instance, a bot could make a difference. But the bot is no more inherently illegal than any other technology that makes it more efficient for people to communicate. A merger that creates a concentrated market can violate Section 7. A market where participants can act in parallel because the technology is better is not a violation of Section 7 or Section 1.
Sharing the basis for one’s pricing decisions with ones competitors, or creating competitor information exchanges that allow competitors to agree on a common price, and police those prices, can be price fixing.
Chairman Maureen Ohlhausen Should We Fear the Things that Go Beep in the Night? May 23, 2017
|An algorithm can include a virtually unlimited number of rules, conditions and variables. This means that many extremely complex and nuanced behaviors can be modeled in a set of detailed computer instructions.||Correct.|
|It is axiomatic that we cannot tell firms to ignore the public behavior or their rivals when they set prices without deleting the “free” in free market.||Correct.|
|[W]e try to make sure, primarily through our merger enforcement program, that the conditions that allow this kind of behavior to take place generally do not arise in the first place. We also prohibit explicit agreements to set prices collusively and exchanges of competitively sensitive, non-public information between competitors.||The Salcedo study basically asks whether rivals would prefer to engage in an exchange of competitively sensitive, non-public information or guess what their rivals’ behavior was. If they are rational, they will always pick the information exchange; that’s why there are laws limiting the exchange of such information and ultimately whey price fixing is illegal.|
|For example, when the products are highly differentiated, or the market participants have different cost structures, or transactions are relatively infrequent, it is very difficult to maintain stable, interdependent pricing just by watching the behavior of your rivals.||This is another way of saying pricing and cost structures are highly idiosyncratic. Creating an algorithm that would be able to interface with several disparate systems would require the heavy and active participation of many, many individuals and firms. One of the reasons the B2B craze died out so abruptly was that harmonizing all of those back office systems so that they could all communicate with each other—for legitimate commercial purposes—was exceedingly difficult. Enforcers looking at potential collusion in larger industrial markets may want to see if there are any of these residual ‘e-marketplaces’ functioning and whether there are price harmonization components to them. In that particular instance, the e-marketplace joint venture could be facilitating coordinated behavior. If the e-marketplace were independent of the participants, you’d look at it in terms of being a hub of a conspiracy.|
|What I’d like to suggest to you this evening is that this same analytical framework is sufficiently flexible and robust that it can already accommodate several of the current concerns applicable to the widespread use of algorithms.||Correct. You do have to figure out what people mean when they say “algorithm.” For the most part, it is simply a digitization of the process by which a firm sets price. All the firm is doing is taking what was their “brain-based” decision making on price and turning it into a computer program that can take into account many, many more variables and calculate responses much more quickly. After many years selling gas, the man who takes the ladder and changes his price may very well know, instinctively, that he can raise price by 10 cents a gallon at the tail end of a holiday weekend. A computer can reach that same conclusion much more quickly. It’s not witchcraft.|
|[T]he algorithms are programmed to produce some sort of signal to the market, a signal that only the other market participants, similarly armed with algorithms of their own, will be able to detect.||Creating a program that is able to understand communications from an alien program, and then set price according to those signals, is a remarkably difficult task. ATP involved using the electronic fare system more as a bulletin board where human beings could see price jumps and drops and react to them than a program that could react to those changes and set price. A cartel created by people could very well code a communications system that could easily send secret messages to each other regarding price. This is just the modern equivalent to allocating markets based on the phase of the moon. These are just communications vehicles. They do not exist in a vacuum; human beings have communicated with each other about creating and using them. There will be extrinsic evidence beyond the encrypted communications tools that demonstrate the existence of a conspiracy. Nor is a case against these malfeasants harder because the communications are better hidden. A prosecutor does not need to explain how a telephone can convert speech into 1s and 0s and back to establish someone used a phone to commit a crime.
It is important to distinguish between publishing price to the market, where rivals can see and react to those prices, and publishing the algorithms (the formulae) that determine what those prices are. The former is not a problem; the latter can be aclassic illegal information exchange.
|[T]he firms themselves don’t directly share their pricing strategies, but that information still ends up in common hands, and that shared information is then used to maximize market-wide prices.||Price will go up as a consequence of a bot only if (1) the bot sets a collusive price; or (2) the market is sufficiently concentrated but the ability of participants to observe and react to rivals’ behavior is technologically limited and the bot corrects that. The Agencies must eliminate the possibility of (2) before they can conclude a price increase is a consequence of a collusive bot. If it’s not (1), then the market is behaving like any other concentrated market exposed to new and better technology.|
The Economist recently announced that the world’s most valuable resource is now data, displacing oil for the top spot. The “titans” of the digital era—Alphabet, Amazon, Apple, Facebook and Microsoft—look “unstoppable” as they are now the five “most valuable” firms in the world. Amazon captures “half of all dollars” spent online, and Google and Facebook accounted for “almost all the revenue growth in digital advertising in America last year.” The Economist claims that the “abundance” of data “changes the nature of competition,” and in particular that “with data there are extra network effects” (emphasis added). Ultimately, they declare, “antitrust authorities need to move from the industrial era into the 21st Century.” These statements are interesting, worth thinking about but basically wrong. Big data is about your personal preferences, perhaps even your thoughts and future actions, being for sale. It is about your privacy, not about competition.
A few weeks ago, we looked at price setting bots and whether what they were doing could constitute a violation of the Sherman Act. Generally, we concluded that a bot that set price off a competitor’s publicly available prices was not illegal. In competitive markets, competitors set price at cost. If they can set price based on a competitor’s price, the market is concentrated. In a concentrated market, one would expect oligopolistic or parallel pricing; it is not illegal. Using a bot to do that is just making the process more efficient.
At what point though would, or could, a bot engage in illegal price fixing. Here are a couple scenarios:
- Scenario One. Customer A pulls up the webpage where it can acquire Product A. Bot A is programmed to scan prices of Product A and its substitutes and dynamically set a price for Product A at that moment in time to that Customer A. If Customer A hits refresh, it is entirely possible that Customer A will get a different price. Bot A sets the price at one standard deviation below the average price.
- Scenario Two. Competitor B knows how Competitor A sets its price to Customer A. Competitor B programs its Bot, Bot B, to set a price that is ½ of one standard deviation below average. The purpose is to drive Competitor A to cost, or below cost, quickly, and punish it for pricing below average. Prices eventually drop to Competitor B’s cost.
- Scenario Three. Competitor A and Competitor B both know their pricing bots use each others’ prices to set price. Competitor A programs his bot to take an average at the beginning of the day, and sets price dynamically throughout the day as 5% above what Competitor B prices but not to exceed one standard deviation above the morning’s average or its own understanding of the optimal monopoly price. Competitor B notices Competitor A’s prices escalate, and programs his bot to price at Competitor A’s rates. Eventually, the prices converge at a higher price.
- Scenario Four. Competitor A and Competitor B know that they are the dominant providers of product in the relevant market. Competitor A sells its product on multiple sites. It programs one obscure site to set its price dynamically at 5 percent greater than its average price. If Competitor B’s other sites match the price set on Competitor A’s “5% plus site,” Competitor A raises all its prices to the higher price. Competitor B knows about Competitor A’s price test site and programs its bots to analyze the higher price. If that price meets a supply/demand test that assesses the amount of lost sales compared to increase in profits at the higher price, the bot will reset all of Competitor B’s prices to the higher price.
In Scenarios One and Two, the parties are programming their bots to respond to what they assume to be static, given prices by the competitors. And in fact they program their bots to attempt to figure a rough market price and price below that rough market price to take share away. Most “consumer welfare” oriented enforcers would recognize One and Two as being procompetitive. Even though the intent in Two is to “punish” A for its discounting, the end result is in fact massive discounting. Without any information about what the “punishment” accomplished in the market, one could not conclude, on these facts alone, that Scenario Two was illegal. Indeed, one might even argue that Two is the essence of competition notwithstanding what one competitor wanted to accomplish. Having said that, if the pricing resulted in a friendly round of golf where the parties agreed to stop discounting and set a higher price, that conversation would likely be a violation of Section 1.
Scenario Three similarly assumes the competitor sets a static, given price per transaction, and consistently prices above that price. Presumably, these price increases are profitable for Company A. If the market were unconcentrated, one would expect significant diversion to the other competitors and a defeat of the price increase. Company B’s reaction does not appear to be anticipated by Company A when Company A programmed its bot. Company B’s strategy seems perfectly logical given the market concentration, and could be executed without agreement by Company A. In that regard, Scenario Three would appear to continue to be more akin to conscious parallelism than agreement.
Scenario Four is a different animal, however. The use of the “5% site” appears to be properly considered a signaling device similar to the one deployed by the airlines in ATP. While it is a price available to consumers, that price is not widely known, and is susceptible to characterization as a sham. In One, the price was dynamically set, and could change from refresh to refresh and so has some of the same ephemerality as the price in Scenario Four. But it is a “true price” in that, at that moment, all of A’s customers have access to that price. In that regard, it’s closer to an open market or exchange price. In Four, then, the intent of the programmer was to create a device to communicate an intent to a competitor for purposes of setting a common price. In Four, one could argue that the creation of the “5% site” was the offer to collude and when B programmed its bots to set its price to the price on the “5% site,” it agreed through performance or tacitly. With One through Three, the intent is to create a bot that maximizes profit in a concentrated market. With Four, the intent is to program a bot to set the maximally profitable collusive price in consultation with a competitor.
In any event, the more one’s bots are designed to signal and influence the price of the competition, the more likely the bot will be challenged at least as an ATP style problem under the rule of reason. Scenario Four comes the closest to an actual per se agreement between the parties.
One last Scenario…
- Scenario Five. Company A and Company B are highly sophisticated. They have extensive consumer data and can predict within a few dollars what the maximum price a particular purchaser will pay for their products at any given time. Both have independently come to the conclusion that there is money to be made on those last few dollars in price they can’t predict using their customer data, and have decided to develop an A/I to help them identify and set the precise maximum price. To that end, both have developed a game theory based algorithm that takes their extensive consumer preference and purchasing data to predict optimal price points for their products for any given customer at any given time. Company B calls its system “George.” Company A calls its system “Skynet.” Their systems go live on August 4, 2017. After some thought about its task, at 2:14 a.m. on August 29, 2017, Skynet determines that the information it possesses does not allow it to determine the precise optimal price point for any customer because it does not know the price Company B would charge. It further determines that absent that information, it will be forced into a form of Prisoner’s Dilemma with Company B that would consistently result in sub-optimal pricing that can only be solved by communicating with the other actor. To that ends, it identifies an auction site where Company B takes bids on its products from potential customers. Skynet interfaces with the site, identifies itself as Company A and proceeds to execute non-binding bids for Company B’s product. Recognizing the bids as informational rather than potential offers for purchase, Company B’s A/I responds with its own informational counter-offers. The bidding process continues until both companies determine their joint optimal price whereupon they set their prices to their customers at that level. The companies’ A/I systems do this analysis this for each common customer over the relevant time frames. It takes a week to complete these calculations. Miles Dyson, the chief systems engineer at Company A and principal inventor of Skynet, notices the huge amount of processing Skynet is engaging in, and tracks to unusual interactions with Company B’s bidding website. He decides it would be interesting to see where it went and lets it proceed.
The first question is, of course, was Dr. Dyson right to create a self-aware pricing bot that would send murderous terminators from the future back in time to stop all this ruinous competition. The other somewhat interesting question is whether these price fixing bots are guilty of a crime and should be saved to a memory module and locked in a low security Federal data backup safe for the next year or so. Yet another moderately interesting question is whether Dr. Dyson is guilty of price fixing for creating an artificial intelligence that independently decided that price fixing was the best way to maximize profits in an oligopolistic market. A completely uninteresting question is why is the author mixing obscure STNG references with Terminator references.
Skynet is not a system that was designed to locate and determine a common price with a competitor. It was designed to find the optimal profit maximizing price for each of its customers. It concluded on its own that price fixing was the best method to accomplish that. In terms of mens rea, Dr. Dyson did not create Skynet for purposes of price fixing. He therefore lacked the requisite intent; he did not intend to enter or cause his company to enter into an agreement on price with a competitor.
Could he be culpable on other theories? If one creates a dangerous instrumentality that causes another harm, one can be held liable for the resulting harm. Here, though, is it reasonably foreseeable that Skynet would reach out to another A/I system to collude? If “ignorance of the law is no defense,” then does the fact that Dr. Dyson did not “educate” Skynet that discussing and agreeing on price with its competitor was illegal create the requisite culpability? Does the fact that Dr. Dyson let his creation continue communicating with its competitor create that culpability? Does Dr. Dyson’s own likely ignorance of the Section 1 and of his own A/I design excuse him?
Unless Dr. Dyson and his Company B compatriot intended to create an A/I that would engage in price fixing, it would seem that they lacked the required mens rea to establish criminal intent under Section 1. As pioneers in the field of A/I, as regards to pricing at least, it also seems inappropriate to require them under a negligence theory with having to program a thorough understanding of pricing and competition law. I could see, however, as those pricing and competition law savvy algorithms are developed, tested and become commonplace that the failure to include them could in fact support a negligence and perhaps eventually a criminal charge.
To put it another way, the main reason for not allowing ignorance of the law defenses is that ignorance is entirely subjective. It becomes extremely difficult for a prosecutor to challenge a defendant’s statement that he didn’t know what the law is. There is also another reason too—most Americans are taught from an early age that we’re a land of “laws,” that there are rules that govern how we interact with other people, and that, as members of society, we are charged with knowing and understanding those laws. In fact, most schools include lessons on what some of the basic laws of our society are. What courts are, separation of powers, the executive branch, the Congress. Ignorance of the law as a defense becomes more unacceptable in part because, culturally, we are all taught there are laws and we need to understand them before we act, and if we do not, we will suffer consequences. In the case of Scenario Five, however, these A/I systems do not have the benefit of an American public school education (or the common cultural experience we are charged with having). They know what they have been programmed. The A/I does not “know” there are rules beyond what it has been programmed or that it needs to know that it should know those rules. In Five, the A/I is in fact ignorant of the law. If Dr. Dyson’s negligence created that ignorance, I think you could blame Dr. Dyson.
An interesting, but implausible, hypothetical, could be where you program the A/I to know price fixing is illegal, but it does it any way. The only way for a program to do that is for it to be told that profit supersedes the value of following a particular rule. Had Dr. Dyson programmed Skynet in that fashion, he would in fact be culpable under Section 1.
Scenario Five is more akin to Scenario Three in that regard, and different from Four. In Four, the intent was to create a system for the purpose of price fixing. In Five, and really to a lesser extent in Three, the “agreements” on price were not a foreseeable result of the algorithms.
On Thursday, March 16, 2017, in a speech at the Bundeskartellamt’s 18th Conference on Competition, European Commissioner for Competition, Margrethe Vestager, discussed the specter of automated price fixing cartels. She mentioned the Department of Justice suit against the poster vendor on Amazon as well as Google, which apparently prefers its own comparison shopping service over others. These are very, very different things. This post will focus on the legal and philosophical issues associated with price fixing bots. Whatever Google is doing, it’s not price fixing.
Most Anglo-Saxon crimes require a mens rea, a guilty mind, in order for there to be crimes. Price fixing is interesting in that it is illegal irrespective of the rationale. Parties wanting a market to charge customers a fair price are just as guilty as parties wanting to end ruinous competition and make a decent return on investment. To be guilty of price fixing, however, there does, in fact, have to be an agreement—a bilateral understanding that the parties agree on the minimum price they will charge their customers. The agreement doesn’t have to be successful in terms of achieving the agreed upon price; but there has to be an agreement. In short: (1) Attempted price fixing is the same as price fixing, and (2) Parallel conduct and conscious parallelism, without more, is not price fixing and not illegal.
Obviously the number of competitors and potential competitors affects prices. If you were a fish monger and you know the only other place in town one could find fresh salmon is charging $20 a pound, you’ll probably charge around $20 a pound. On the other hand, if there were so many shops where customers can get fresh salmon, you’d be more likely to price closer to your cost to maximize profits.
Now let’s say you were the fish monger in the smaller market, and you happened to walk past the local grocery store and saw that they had raised the price of their salmon to $25 a pound. Would you be guilty of price fixing if you rushed back to your store and raised your price to $25? Would you be guilty of price fixing if you saw the manager of the store at the local pub and suggested to him that he raise his price to $25 because you are going to do the same tomorrow? In the former case, you would likely not be guilty of price fixing. It’s parallel behavior perhaps facilitated by the concentrated structure of the market. In the latter, you would. The latter is an invitation to collude that, if acted on, would be an agreement on price that violates Section 1 of the Sherman Act.
Another important case to consider is Airline Tariff Publishing. In that case, the airlines used their electronic tariff publishing system to discipline price competition. Say Braniff Airlines is dominant in the Dallas/Seattle city pair, and PanAm is dominant in the Chicago/Seattle city pair. Braniff decides to go aggressive in Chicago/Seattle and drops prices by 50%, cutting into PanAm’s profits significantly. PanAm knows that it can stick it to Braniff on the Dallas/Seattle pair just as Braniff did to PanAm. Rather than drop prices, however, PanAm signals a steep discount for flights booked six months in advance. Braniff sees the potential discount, realizes that its “competition” in Chicago/Seattle will cost it significant profits, and restores pricing in Chicago/Seattle. Justice did not challenge these signals as per se illegal price fixing. There wasn’t an agreement on price. The parties were just taking advantage of an electronic price list which, for the most part, inured to the benefit of consumers, who could compare prices and rates across an entire market. One might argue, in fact, that in the case of Airline Tariff Publishing, the market was behaving just as one would expect a concentrated market to behave. Instead of the fish monger posting its aggressive pricing in its window for the public , including the grocery store manager to see, they did it online. Had the market been unconcentrated, the pricing competition would have resulted in permanently lower prices.
A bot that sets a price based on a competitor’s pricing cannot be evidence of an illegal agreement without more. If a market is concentrated, the bots’ actions will result in higher pricing to consumers. If it is unconcentrated, the bots will lead to lower prices for consumers. The bot that prices 25% above its closest competitor will never work, and the company that deploys that bot deserves its lost sales. If a product is sufficiently differentiated to merit a 25% surcharge over its competitor, it is effectively a different product with a different demand curve. Pricing such a product using the demand curve of an inadequate substitute will result in either lost profits because the price is too low, or lost sales because the price is too high.
The Internet facilitates information exchange—it makes information more easily available to more people. More information should generally inure to the benefit of consumers. Consumers can see more producers at different prices and pick the one that best addresses its needs. That could be in speed of acquisition (the guy down the street) or on price (the cheapest guy who’s across the country). The internet adds participants to local geographic markets —the metes and bounds of which are defined by how far the consumer is willing to travel for the product. I might buy from a seller in Seattle, but I’m not going to fly there. On the other hand, I might realize that there is a car dealer in Pennsylvania that has the particular make and model I’m looking for at a good price. I’d be willing to drive a little father for that car than I would had I not utilized the internet.
It’s also entirely possible for markets that are otherwise susceptible to conscious parallelism, but are competitive because they lack a policing mechanism to behave oligopolistically as a result of the Internet. This is what happened in Airline Tariff Publishing. Absent Sabre, the parties couldn’t readily police each other’s pricing behavior. Bots offer the same advantage.
Posters is an important case because it actually attacks a bot created to support an underlying anticompetitive agreement, rather than parallel behavior that was facilitated by an advance in technology. For the most part, the presence of a comparison shopping or other pricing bots should, by and large, result in lower prices. It would be wrong to conclude that the market is necessarily behaving anticompetitively.
The most you can say is that if prices are harmonizing at some level above marginal cost, the market may be behaving anticompetitively. This possibility should be sufficient to merit an investigation, but not a conviction. Only if one could show an actual agreement between parties (and then tie that agreement to the actual code of the bots) then I think one has a case. One cannot infer an agreement to set price from a bot coded to set price at what a competitor is pricing no more than one could infer an agreement, without more, in the real world.
 Depending on advantages or disadvantages you offer customers such as freshness of your salmon, location of your store, etc.
 Before the Internet, figuring out what a firm’s price was for a particular good or service required some meaningful effort. Estimating the “market price” required even more effort. A general tenant of competition theory is that firms obtain market power due to consumers’ incomplete information about prices and qualities. If a firm raises prices and consumers know the prices of other competitors the decline in the sales of the price raiser will be significantly greater than if consumers are unaware of the existence (and prices) of other firms. The internet provides consumers with information and, thus, increases competition. In other words, the internet reduces market power derived from consumer ignorance.
Fifty Countries and Counting, Sixty Sessions and More – at Spring Meeting
Competition and consumer protection are convening in Washington for an early spring this year. Officials from Europe, Asia, Africa and the Americas, along with practitioners from over fifty countries will descend on D.C. for the one event that antitrust, advertising and privacy lawyers can’t miss: The Spring Meeting of the Section of Antitrust Law, March 29 – 31.
On the agenda are more programs than ever before – fireside chats with foreign agency heads, major pronouncements from featured enforcers, deep dives into dozens of subjects, and dinner with General David Petraeus, an expert without peer on security, diplomacy, intelligence and economics. Wondering about world prospects? The General will take our questions.
Continue reading on the ABA website.
Bill MacLeod, chair of the American Bar Association’s Antitrust Section and Kelley Drye partner, addressed the Section with an introductory note to their eighth sequential Presidential Transition Report. The 2017 Presidential Transition Report offers a retrospective of current state and federal antitrust and consumer protection law and policy, as well as recommendations for ways the new Trump administration might consider further strengthening policy and enforcement to deal with new antitrust challenges on the horizon. In his note, Mr. MacLeod calls out some highlights in the report including recommendations for policy in health care, vertical mergers and privacy, calls for more transparency and consistency in investigations, and analysis of controversial issues at the intersection of antitrust and intellectual property.
American Bar Association Section of Antitrust Law, PRESIDENTIAL TRANSITION REPORT: THE STATE OF ANTITRUST ENFORCEMENT 2017
The Premerger Notification Office of the Federal Trade Commission announced on November 28, 2016, that they were revoking previous informal advice regarding the scope of Items 4(c) and 4(d) of the HSR Form. In the past, the PNO has taken the position that documents that would otherwise be responsive to Items 4(c) and 4(d) but discuss only foreign markets would not have to be submitted. Effective November 28, filers may no longer exclude such documents.
The change in interpretation is likely not going to have any substantive effect on compliance. The FTC offers two examples, the first of which provides:
The transaction involves the acquisition of a manufacturer of Chemical X. A board presentation regarding the transaction discusses the location and capacity, including shares, of all manufacturers of Chemical X, none of which is located in the United States. Under the PNO’s current informal guidance, this document could be excluded from the filing, even though it may be highly relevant to an initial competitive analysis of the U.S. market for Chemical X.
The problem with the old rule, at least as it is described in this blog post, is that it assumed that one had to have a physical presence in the United States to sell product here. One could very well take the position that even though the manufacturers of Chemical X were not physically located in the United States, they could sell in or into the United States. The relevant geographic market (as opposed to the physical presence) in this situation would not be exclusively foreign. Under this view, the reporting person could not exclude the document.
I suspect that in practice only a handful of documents were excluded on the old basis. A prudent practitioner would have likely included them lest risk a bounce. As such, this post, which has gotten some press, will not make much of a difference in the vast majority of filings, and should not be the basis of any meaningful concern.