In recent times, there have been widespread attacks on algorithmic rent pricing tools, accused of being a means to enable landlords to conspire with each other to harm renters. Not only has the DOJ brought such a suit, but numerous states and cities have jumped on the bandwagon, along with class action lawsuits. However, there is a better way to interpret such tools—that they enable rental rates to move more quickly toward what is justified by market conditions (as economists illustrate with changes in supply and demand) when those conditions change, which becomes particularly important when those conditions change rapidly.
Those algorithmic tools take information on rentals, expiration dates, likely rates of new rental applicants, and other terms for a given landlord, plus publicly available information, to recommend whether it would be more profitable to raise, lower, or leave their rental prices the same. If it recommends a change, it then also incorporates additional anonymized information from other users of the provider’s tools (to provide information about the market but prevent communicating information about any specific rivals to others) to determine the recommended rents. Subscribers can then follow, modify, or ignore that advice.
Given the very different “collusive” versus “improved competition” stories that can be used to characterize such algorithms, how can we determine whether such algorithmic tools are market-enhancers or conspiracy-enablers?
To begin, it is instructive to note when the attacks on algorithmic pricing began to arise. That was primarily in the early 2020s, when rents were skyrocketing. To illustrate, CBS News reported that rents rose 30.4 percent between 2019 and 2023. That is exactly the type of environment that would make faster adjustment to rapid changes in the market more valuable. That substantial rental inflation also meant politicians and government bureaucrats were very eager to avoid blame and, if possible, instead appear to offer solutions to problems others caused.
Unfortunately for the latter strategy, government actions that tended to boost housing rents were everywhere. Monetary policy was outrageously expansionary, fiscal policy was similarly profligate at both federal and state levels, and both pushed the word trillions into Americans’ common vocabulary. Regulatory restrictions that increased costs and delays and reduced construction were added on top of already existing zoning restrictions, permitting and impact fees, property taxes, efficiency standards, etc., etc.
Consequently, large numbers of “public servants” wanted to avoid blame for the sharp rental housing cost increases taking place. But very few such pictures could fit into a frame that excused all of them. The use of algorithmic pricing software, however, also increased over that period. That made it a popular scapegoat for those who agreed that “not me” was the answer to “Who should be blamed?”
The scapegoaters took the correlation between rental housing price increases and increasing use of algorithmic rental price software over that time period as proof of causation, ignoring one of the most basic principles of statistics: correlation does not prove causation. For example, many teenagers suffer from acne and also have growth spurts, but we do not conclude from that correlation that acne causes growth spurts. Their strategy also completely ignored the fact that, under the inflationary circumstances, algorithmic tools become more valuable in improving market adjustments. In fact, it could well be that rapid increases in rental pricing were causing increasing employment of tools especially useful in that circumstance, rather than causation running the other way.
Opponents then adorned their story with adjectives like “monopolistic” and “collusive” and nouns like “conspiracy,” sprinkled on fear and loathing of large landlords as “big business,” and pushed their shared storyline repeatedly in order to dodge their accountability. However, as Kevin Hopkins put it at R Street, “Just because monopolistic collusion and manipulation of rental markets is convincingly postulated doesn’t mean it exists.”
While the scapegoaters tried to make sure people didn’t think too carefully about their correlation-demonstrates-causation “proof” of anti-competitive actions by users of algorithmic tools, it is worth doing so, because there are many ways their arguments fail.
Landlords did not enter any direct agreements with one another by using those algorithmic tools. But that is the established standard for antitrust prosecution under the Sherman Act. In fact, algorithm opponents claimed that no such agreement should even need to be proven in this case. That suggests antitrust bureaucrats, their political overseers, and the antitrust bar may be more interested in expanding their power to win cases without having to prove the existence of a collusive agreement than they are in protecting market competition from collusion.
Far too few landlords actually use algorithmic tools to give them the kind of monopoly power they are allegedly abusing. Here, accusers suggest that someone having a large share of the market for users of algorithmic software rebuts that claim. But the relevant market for determining power to jointly raise rents is the roughly 50 million rental housing units in the country, which dwarfs the single digit percentage that uses such tools for rental housing.
When non-public information was utilized in the software’s rent-setting suggestions, it was anonymized to prevent the communication of information on specific rivals’ behavior, which undermines any ability to use it to monitor collusive agreements. Further, no software user was required to follow the software’s advice or punished for failing to do so, and deviations from recommendations were common. Both undermine scapegoaters’ arguments, since deviations from “agreed” prices is a prime Achilles’ heel that breaks down attempted collusive pricing agreements. There is also evidence that supports the improved competition scenario, but not the conspiracy scenario.
At times when underlying conditions are creating price inflation, both the increased collusive power for landlords interpretation and the better revelation of competitive market conditions interpretation point to higher prices. So finding that result does not distinguish between the two hypotheses. But the improving competition alternative would point to algorithmic tool users lowering prices more quickly than non-users in down markets, which is not true of the conspiracy argument. And as Christian Britschgi, points out,
The limited academic research on the effects of rent-recommendation algorithms suggests that’s exactly the effect rent-recommendation algorithms have: They make landlords more responsive to changing market prices, which means a greater willingness to raise and lower rents.
The conspiracy charge also implies that there should be fewer housing units available as a result, since it is the power to reduce the quantity of a good available that is the textbook source of monopoly power, and thus there should be higher vacancy rates as a result. But that is not true. RealPage—the algorithmic software market leader—pointed out, “properties using our revenue management products consistently achieve vacancy rates below the national average.” In algorithmic tools help the rental housing market work better, rather than undermining competition in it.
Further, as Kevin Hopkins found, a natural experiment further undermines the collusion charge. Real Page acquired a competitor in 2017 (and the DOJ did not at that time object to the use of algorithmic software in approving the merger), which increased its market share substantially. If that represented increasing monopoly power in the rental housing market, it should have raised rental prices relative to owner-occupied home prices. But Hopkins found that
…during the time in which RealPage most significantly expanded its RM software market share, the ratio of rental price increases to owner-occupied home price increases actually declined—by 38 percent in 2013-2016, by 10 percent in 2017-2020, and by 23 percent in 2021-2024. In other words, there has actually been a negative correlation between RealPage’s growing market share and the relative change in the cost of rental housing versus owner-occupied housing over the past 12 years.
Similarly, the conspiracy argument implies that rents should be relatively higher where a larger share of landlords use algorithmic software. But the data doesn’t show that either.
In sum, for all the dramatic rhetoric from accusers against algorithmic rental pricing software as conspiracy devices, their argument has both many logical holes and substantial contradictory evidence holes. In contrast, recognizing the desperation of those whose policies actually caused rapidly rising rents to find a common scapegoat explains their attacks far better. In fact, this episode could be well summarized by a Wall Street Journal subtitle for a story about this topic, “Biden’s antitrust cops spy conspiracy in a price discovery tool.” Algorithmic software has improved a competitive market. But, as has so often been the case in the history of antitrust, pro-competition rhetoric is again being used as a cover for attacking superior competition on behalf of parties who gain from restricting it.
