Two Estimates. One Algorithm Still Running.

Introduction

Two branches of the federal government measured the same problem and got different numbers: the FTC's economists said $53 a month per unit, the White House's said $70. RealPage said none of it was happening. A federal judge signed the settlement anyway — zero dollars, no admission of wrongdoing — and the algorithm kept running. Now states are passing their own bans. RealPage is suing them for that too.

What Two Sets of Government Economists Found

Start with the part that didn't exist a few months ago. A number.

On February 24, 2026, economists Sophie Calder-Wang of Wharton and Gi Heung Kim of Boston College presented a study at the FTC's 18th Annual Microeconomics Conference. They built a structural model of the multifamily rental market using Moody's REIS data covering more than 37,000 buildings and 7.2 million units across the top 50 metro markets, roughly half of all market-rate apartments in the country. Then they ran a conduct test designed to distinguish two explanations for the prices they saw. Were buildings using the same pricing software competing with each other, or coordinating? Their finding, in the conference paper, is that the data "favor coordination (i.e., joint-profit-maximization) over competition among users of the same algorithm." The price tag they attached was an average markup of $53 per month per unit from algorithmic collusion, across more than 4.2 million adopted units by 2024.

Here's the part that makes it stick. RealPage's algorithm was never proven in court to fix rents, and the DOJ never got a verdict. But the government's own economists went ahead and quantified the cost anyway, and the company's answer to every state now trying to ban the software is to sue them on a theory that would put algorithmic price coordination beyond the reach of any government.

One study is easy to wave off, so look at the second number. The White House Council of Economic Advisers, in separate work cited by the National Low Income Housing Coalition, put the markup in algorithm-using buildings at roughly $70 a month. Two government bodies, working from different datasets and different models, landed on the same conclusion. That gap between the two estimates isn't a contradiction. It's the range you'd expect from two teams measuring the same overcharge with different tools. And the Wharton authors are explicit that their figure is "likely a lower bound," because rental prices are strategic complements: when one landlord raises rent, neighbors can follow, which means the model probably understates the spillover. The honest caveats are in the paper too. The authors call it a "back-of-envelope" calculation and state plainly that it is "not forensic evidence of collusion or price fixing," and the data runs through 2019, before the full post-COVID rent spike. None of that changes the direction. It just means the actual cost is probably higher than what the model found.

What the Settlement Actually Lets RealPage Keep Doing

Here's where the timeline turns strange. The DOJ settled with RealPage in November 2025, with no fine and no admission of wrongdoing. On March 26, 2026, the U.S. District Court for the Middle District of North Carolina entered the consent decree, with the DOJ's response to public comments published May 8. The deal is signed and fully operative.

What does the operative deal require? Behavioral fixes. RealPage has to stop using competitors' nonpublic data in real-time pricing, remove features that block price decreases, and accept a three-year court-appointed monitor. Going forward, any nonpublic data used to train the models has to be at least 12 months old. What the deal does not require: that RealPage stop selling the product, that landlords give renters a dime back, or that anyone disgorge a single dollar of the markup the economists just measured.

So the algorithm keeps running. New CEO Dirk Wakeham said the settlement ensures the company's revenue management solutions remain "fully available, legally compliant, and aligned with evolving laws and policies." In a January 29 statement, RealPage offered customers the option to remove nonpublic competitor data from rent calculations, but only as an option, not a default. The company says it has "no plans to sunset any of its solutions," and it kept every customer through the entire litigation without a single one walking. The product page still markets the software as "proven, cycle-tested, disciplined analytics that balance supply and demand to maximize multifamily revenue growth." Read that as a renter and "maximize revenue growth" is doing a lot of quiet work.

The Cities Moving Faster Than Washington

If the federal settlement left a hole, local governments noticed. While the DOJ was negotiating its zero-dollar deal, cities and states were writing their own rules, most of them stricter than anything the consent decree contains.

Portland's City Council voted 8-2 on November 19, 2025 to ban the use and sale of algorithmic rent-pricing software, with the ordinance taking effect roughly February 17, 2026. The text names RealPage directly, notes the company controls over 80% of the commercial revenue management software market, and sets civil penalties of up to $1,000 per violation, with each unit and each month counting separately. New York went statewide: a prohibition under General Business Law Section 340-b, signed October 16, 2025 and effective December 15, 2025. Maryland's Protection from Predatory Pricing Act was signed April 28, 2026. San Francisco got there first back in 2024, and Philadelphia, Minneapolis, San Diego, Seattle, Jersey City, Hoboken, and Providence have all passed versions since.

The state attorneys general didn't fold either. Nine states that co-sued with the DOJ kept their own cases alive, and a separate group (North Carolina, Arizona, Maryland, New Jersey, Kentucky, and DC among them) is pursuing independent litigation. North Carolina AG Jeff Jackson put it bluntly in February: "If these owners met in a back room and agreed to set prices, that would be price fixing. This is no different, it's just using software instead of a back room." A bipartisan coalition of nine states also reached a separate $7 million settlement with Greystar, the country's largest landlord, which manages nearly 950,000 units. Even there, the algorithm survives the deal. Greystar just has to stop feeding it nonpublic competitor data.

The Move That Could Make All of It Moot

Now the counterattack, which is the real story. RealPage isn't just defending itself against these bans. It's suing the governments that pass them, on a theory that, if a court buys it, ends the whole debate.

In late November 2025, RealPage sued New York Attorney General Letitia James in the Southern District of New York (RealPage v. James, 1:25-cv-09847). The argument, led by Gibson Dunn's Theodore Boutrous Jr., is that New York's ban is "a sweeping and unconstitutional ban on lawful speech" that "prohibits the use of math and publicly available information." RealPage filed a near-identical suit against Berkeley in April 2025. The claim is the First Amendment: a pricing recommendation, the company says, is speech.

Sit with what that means. If a court agrees that algorithmic pricing recommendations are protected speech, then any law restricting them has to clear strict scrutiny, the hardest standard in constitutional law to satisfy. A city ordinance capping penalties at $1,000 a unit doesn't survive that, and neither does New York's statute. The entire coast-to-coast patchwork of bans becomes unenforceable in one ruling. RealPage doesn't need to win on the rent-fixing merits. It needs one judge to agree that software telling a landlord to charge more is the same kind of protected expression as a newspaper editorial. AG James, in her January 9 motion to dismiss, called RealPage's promotion of the software "unabashed zeal in promoting its software to stifle competition." The motion is pending. Berkeley already postponed its ban to avoid the litigation cost, citing a $27 million budget shortfall, which is the strategy working before a single ruling lands: make the legal risk expensive enough that cities blink.

Who Benefits

Follow it to the money, and the money runs through Chicago. Thoma Bravo, the private equity firm, bought RealPage in April 2021 for $10.2 billion in an all-cash deal, paying $88.75 a share, a 30.8% premium. The revenue management software is only about 7% of RealPage's revenue, but it's the moat: it's the product that pulls landlords' lease data into RealPage's systems, which makes those landlords stickier customers and feeds the data advantage that justifies the >80% market share. Kill the algorithm and the data relationships that lock landlords in go with it, which is worth far more than the 7% line item suggests.

A zero-dollar, no-admission settlement is exactly what a private equity owner with a valuation to protect would want. An admission of price-fixing on the core product is a permanent stain on the asset; a behavioral consent decree is a footnote you can market around. And the market believed it. In June 2025, at the peak of the litigation, Thoma Bravo closed its largest fund ever, $34.4 billion across three vehicles, bringing its assets under management to roughly $181 billion. Bloomberg reported in April 2026 that RealPage's earnings actually improved at the end of 2025, settlement and all. Thoma Bravo partner Scott Crabill sits on RealPage's board. So the First Amendment lawsuits are moat maintenance: every city ban that holds closes off a market, and every one a court strikes down reopens it. The litigation is protecting the asset Thoma Bravo paid $10.2 billion for.

And the renters? There's no mechanism in the DOJ settlement for any of them to get paid. The separate landlord class-action settlements have now crossed $218 million across more than two dozen defendants. Run the math: 4.2 million units paying that monthly markup comes to roughly $2.67 billion a year in above-market rent. The landlords settled for less than a tenth of a single year's estimated overcharge, and RealPage itself paid nothing in those cases.

Two Arms of the Same Government, Two Answers

The pattern here is cleaner than the usual corporate-accountability story. Normally the company says one thing and the documents say another, and the job is to find the gap. This time the gap runs between two arms of the same government: the DOJ closed the case as if the question were settled, and then the FTC's conference economists and the White House CEA published numbers showing the cost was real and ongoing. One branch walked away while the other quantified exactly what walking away cost.

The Question No Court Has Answered Yet

For the DOJ, November 2025 was the end. For RealPage, it was the opening of a much bigger fight, where the goal isn't to prove the algorithm is legal but to make the question of legality something no city council or state legislature is allowed to ask.

So the real test isn't the consent decree that's already signed. It's whether a single federal judge decides that a number on a screen telling a landlord to raise the rent is the kind of speech the First Amendment was written to protect. If one does, every ban from San Francisco to New York becomes a dead letter, and the $53 you may already be paying becomes the price of a market nobody is allowed to close.