The age of artificial intelligence, long predicted by futurists and science fiction writers, is finally here. Consumers and producers now have access to AI technology such as ChatGPT that radically transforms their abilities and makes them more efficient at managing a wide range of tasks. AI can now help us drive our cars, shop for groceries, buy and sell products and services online, manage our homes and our offices, plan our travels, prepare our homework, do our jobs, and even care for the sick and the elderly. AI's efficiency saves us resources, makes us better at what we do, and helps us to accomplish new things that without AI we could never do before.
Real estate property management software is one area, like so many others, where AI-based technological innovation is making a positive difference. Property managers now have access to technologies (such as RealPage's YieldStar and other AI-based revenue management systems) that can assist them to manage their properties more efficiently. Rather than have to expend substantial resources researching and analyzing how to manage the pricing and marketing of their properties, the timing of leases, and the ratio of occupations to vacancies, real estate management firms can now receive assistance with all of that from non-binding, advisory software applying AI-technology algorithms to public, industry, and user-provided data for a fraction of the cost that lessors would need to expend to operate without such data-driven software. Best of all, this technology can benefit not only landlords (by enhancing their profitability), but also their tenants (through lower costs and more accurate, market-reflective pricing that better matches them with desired amenities).
Unfortunately, this nascent innovation in real estate management software is now under threat from antitrust litigation. Attorneys representing renters of units in multifamily dwellings have brought multiple class action antitrust lawsuits against real estate management software developers, alleging that such software facilitates the unlawful sharing of sensitive pricing information and thereby reduces competition in the multifamily real estate rental market to the detriment of consumers. And antitrust regulators have recently compounded this threat by participating in this litigation and, at the urging of some members of Congress, by launching an investigation of their own.
As a longtime professor of antitrust law, my message to antitrust regulators and to the judges who must decide these cases is simple: proceed with caution. AI technology provides many benefits to markets that regulators and even market participants themselves may not yet fully appreciate. While it may be tempting to view this new, algorithmic pricing technology with suspicion or as inherently harmful and to condemn it as ipso facto unlawful, such a knee jerk reaction is ill-advised. For as Dan Akroyd's character famously once put it in the movie Spies Like Us, it is all too easy to "mock what we don't understand."
Maureen Ohlhausen, a distinguished former Federal Trade Commission Chair, has mocked AI-based algorithms by analogizing them to "a guy named Bob." In an oft-cited 2017 speech, she reasoned that if it is unlawful for horizontal competitors to share pricing data and thereby coordinate pricing decisions through "a guy named Bob," then "it probably isn't OK for an algorithm to do it either." To Ohlhausen, the mere sharing of pricing information is evidence of an unlawful price fixing conspiracy. But Ohlhausen’s analysis insufficiently appreciates the benefits to consumers that such AI-based data sharing can generally entail, as well the specific benefits of such data sharing in real estate rental markets.
The most obvious benefit to consumers from AI-based data sharing in real estate rental markets is that pricing will more accurately reflect market values than it would without such information sharing. Just as Amazon helps consumers to shop more efficiently by bringing market pressures to bear on product sellers, so too does this rental software benefit renters by enabling lessors to incorporate relevant market information into rental transactions more quickly and accurately than they would if such AI-based software were unavailable. As with pricing on Amazon, such data sharing can certainly cause prices to rise at times. But it can just as easily cause prices to fall. And importantly, even rising prices can help consumers by more accurately matching them with desired product amenities.
Consider the case of a lessor who owns two similar apartments in the same city except that one is near a train station, and one is far away from it. The train station has been closed for a year and it is uncertain when it will reopen. Suppose that because the train is not currently operating, a lessor using traditional real estate management practices would price the apartments equally and rent them out indiscriminately to renters who may value train access differentially. Suppose, however, that rental software recommends charging more rent for the apartment that is closer to the train – perhaps because data in the algorithm is indicating that the market will value this apartment more because of its proximity to the train, or perhaps because of some other known or unknown factors.
The actual reasons for the software’s pricing recommendation are unimportant. Regardless, the apartment that prices higher on the software’s recommendation are now more likely to end up renting to tenants who idiosyncratically value the apartment relatively more highly, whether due to eventual access to the train or for any other reason, and who is therefore willing to pay higher rent. This is a more efficient and better outcome for consumers than having a tenant who does not value the apartment as highly ending up renting it and paying less for the apartment.
As a group, tenants are better off when apartments end up in the hands of those who value them the most, which is more likely to happen with algorithmic pricing than without it. Multiplying the effect of RealPage's data-driven pricing recommendations across a myriad of known and unknown and even hidden factors that affect rental values, one can readily see the potentially enormous benefit to consumers of using RealPage as a sorting and pricing mechanism. And needless to say, these benefits to consumers from algorithmic pricing are not something that “a guy named Bob” could ever provide.
Four senators -- Ed Markey, Bernie Sanders, Tina Smith, and Elizabeth Warren -- would have us believe otherwise. In a letter last March to the head of the Antitrust Division at the Department of Justice, these senators urged the Department to investigate RealPage because its algorithm “has effectively created a ‘cartel’ of landlords that is able to take advantage of non-public information to fix rents.” The senators do not offer any evidence in support of this assertion, and there is reason to doubt the claim.
First, there are many landlords that do not avail themselves of algorithmic pricing, which is used by no more than 35 percent of rented, multifamily properties in even the markets where landlord use of algorithmic pricing is the most widespread. If 65 percent or more of the properties that are available for rent in any given real estate market do not use the algorithm, the algorithm cannot effectively control prices – since any coordinated effort to raise price anticompetitively by the 35 percent would be competed away by the 65 percent who do not avail themselves of the algorithm.
Moreover, the RealPage software does not require landlords to follow its price recommendations. Landlords are free to depart from those recommendations (both upwardly and downwardly) when they see reason to do so, which often they do, because ultimately each piece of real estate is unique and has its own reasons for pricing higher or lower. So any ostensible commitment by landlords to adopt AI-generated price recommendations is at most illusory.
For an alleged price fixer to violate Section 1 of the Sherman Act, there must be an “agreement” in restraint of trade. Pricing real estate using a system that generates non-binding price recommendations hardly constitutes an agreement for purposes of Section 1. And it is even doubtful whether AI, which lacks legal capacity or agency, is itself capable of making such an agreement in any case. Yale scholars Ian Ayres and Jack Balkin have argued persuasively in a recent working paper that the law will need to change in many areas in order for regulators to reign in AI, given its lack of agency and legal capacity. But such legal changes have yet to come to antitrust. AI therefore lacks as yet the capacity to fix prices in violation of the antitrust laws.
To be sure, antitrust law does not and should not allow algorithmic pricing systems to serve as a cover for price fixing conspiracies among legal actors (whether human or corporate) who are horizontal competitors. But so long as there is not any evidence of such a conspiracy, regulators and courts should keep their hands off algorithmic pricing mechanisms and allow them to flourish.
Comment
Show comments
Hide Comments
