On May 12, 2026, Zillow filed a federal antitrust lawsuit in the Northern District of Illinois against Compass International Holdings and MRED — the multiple listing service that covers Chicago, the nation’s third-largest real estate market.
The complaint accused the two companies of conspiring to withhold listing data from Zillow unless the portal agreed to carry Compass’s private, off-market inventory nationwide. The threat was explicit: comply, or lose access to the Chicago feed entirely.
The map every buyer is using was built by parties with a stake in how it looks. That’s what the lawsuit is actually about.
This is not a story about Zillow.
What the consensus gets right
The standard housing market thesis is not wrong on its own terms.
As of late 2024, roughly 82 percent of homeowners with mortgages held interest rates below 6 percent. Many locked in at 3 percent during the 2020–2021 refinancing window. Listing your home means surrendering that rate and taking on a new mortgage at nearly double the cost. Of course inventory is low. The math says stay put.
The delinquency picture, by historical standards, looks manageable. The New York Federal Reserve’s data shows borrowers in high-income zip codes are posting historically low default rates. The Mortgage Bankers Association puts the total delinquency rate at 4.26 percent of all loans outstanding as of Q4 2025 — elevated from post-pandemic lows but nowhere near the 9.3 percent peak of the 2008 financial crisis. Tight underwriting since 2009 means the median credit score on newly originated mortgages has stayed persistently above 750.
The people who say prices don’t fall from here have real data behind them.
The problem isn’t what they can see.
The working hypothesis: The headline data that most buyers rely on is structurally incomplete in two places at once — the stress side and the supply side — and both gaps are widest in the same markets.
The first blind spot: two delinquency surveys, one headline
The MBA’s National Delinquency Survey is what most people mean when they cite the delinquency rate. It covers roughly 40 million first-lien loans reported by participating servicers: banks, credit unions, independent mortgage companies. That is the agency universe — loans backed by Fannie Mae, Freddie Mac, or Ginnie Mae, where servicers have a standing obligation to report into a common data stream.
There is a second universe. The non-agency residential mortgage-backed securities market — loans that don’t meet the standardized underwriting guidelines required for a government guarantee — carries $1.7 trillion in outstanding securities. It is growing fast. Private-label RMBS new issuance rebounded to $132 billion in 2024 and is projected at $160 billion in 2026, which would make it the highest year since the financial crisis. This market reports not to the MBA but through servicer remittance data to rating agencies — primarily Fitch, Kroll, and Moody’s — who publish performance monitors on a quarterly lag.
Those reports show a different picture than the MBA headline. Fitch’s Q1 2026 RMBS Performance Monitor puts the Prime Jumbo 30-day delinquency rate at 1.09 percent, up 22 basis points year-over-year. The Non-QM sector — loans made to self-employed borrowers, investors, and others who don’t qualify under standard guidelines — sits at 7.26 percent, up 118 basis points year-over-year. These numbers don’t appear in the MBA figure the market quotes.
The GSE non-performing loan sale program illustrates the same structural gap in smaller form. Since 2014, Fannie Mae and Freddie Mac have sold 173,571 deeply delinquent loans — carrying $31.8 billion in unpaid principal balance — to private investors whose servicers report outcomes to the GSEs but not to the MBA’s survey. The program was designed to transfer credit risk to the private sector, and that includes the reporting. Sale volumes have slowed since their 2021 peak, so this pool is not the primary data gap. It’s the secondary illustration of how the agency and non-agency universes operate on different reporting rails.
Now add the VantageScore signal. In July 2025, VantageScore’s CreditGauge reported that 90-plus-day delinquencies in its Superprime tier — borrowers with scores between 781 and 850, the cohort lenders treat as the safest — were up 109 percent year-over-year. The following month, that figure accelerated to more than 300 percent year-over-year. VantageScore’s own commentary was explicit: absolute levels remain extremely low. The signal is in the rate of change and the identity of the borrowers moving. When the tier calibrated to be the last to fail starts moving first, in consecutive monthly reports, that is worth naming.
This is not the 2008 story. The 2008 stress was concentrated in subprime borrowers — thin credit histories, minimal down payments, adjustable-rate mortgages that couldn’t survive a rate reset. The current signal is appearing in the prime cohort: people who bought in 2021 and 2022 with excellent credit and strong incomes, at the top of a market inflated by remote-work demand and historically low rates.
The geography is specific. From Q4 2019 to the 2022 peak, Austin median home prices rose approximately 90 percent. Phoenix rose approximately 63 percent. Dallas rose approximately 48 percent. These were the destination markets for the remote-work migration wave. Austin’s tech sector employment subsequently declined 1.6 percent in 2024, with startup employment falling 4.9 percent. Texas recorded near-zero net job growth overall in 2025. Rents in Austin have declined roughly 20 to 22 percent from their peak as new construction from the boom years floods supply faster than demand absorbs it.
The 2021-era remote worker who moved to Austin on a $250,000 salary, purchased at $750,000 with a 3.2 percent mortgage, got called back to a San Francisco office, and is now navigating a compensation environment restructured around AI productivity — that person is not in the subprime bucket. They are in the Superprime bucket. The data says the Superprime bucket is moving.
The stress data built to catch the 2008 crisis is calibrated for the wrong borrower, in the wrong geography, with the wrong mechanism.
The second blind spot: the door, and who walked through it
The supply data has a different problem. It has two mechanisms — and the piece requires naming them separately, because they are not the same thing.
The first mechanism is a formal policy. On March 25, 2025, the National Association of Realtors adopted the Multiple Listing Options for Sellers policy — MLOS. The rule gives sellers the option to instruct their listing agent to delay syndication, the process by which a listing flows from the local MLS to Zillow, Redfin, and Realtor.com, for a period determined by each local MLS. During the delay, the listing remains visible inside the MLS platform to agents. It is not visible to the buyer running a search on their phone.
In practice, MLOS alone creates limited distortion. Bright MLS estimated that by April 2025 delayed listings represented about 8 percent of its market — and 90 percent of those listings transitioned to standard active status before they were sold. Most sellers use the delay window to prepare, not to hide. MLOS is the policy that opened the door. It is not the primary mechanism of the inventory data distortion this piece is arguing.
The second mechanism is a business strategy. Compass’s “3PM” program — Private Exclusive, Coming Soon, MLS — moves listings through three distinct phases. In phase one, a property is marketed exclusively within the Compass agent network. In phase two, it appears on Compass’s website only. In phase three, it enters the MLS and public syndication. There is no mandated maximum time in phases one or two. A listing can sit in the private network indefinitely before going public — or sell there, never appearing in MLS data at all.
This is not a delay. It is a parallel routing system. When Compass launched seven Private Exclusives in Washington state in April 2025, the Northwest MLS cut off Compass’s IDX data access entirely, arguing the strategy violated the one-business-day public marketing rule. Compass sued. MRED went the opposite direction — expanded its Private Listing Network nationally and partnered with Compass to route its full inventory, including off-market listings, through the network. The Zillow antitrust complaint alleges that MRED then threatened to cut off Zillow’s data access entirely unless Zillow displayed those listings.
The inventory data distortion argument rests on Compass 3PM and the MRED partnership, not MLOS alone. MLOS created the legal framework that made phase-one-and-two marketing defensible. The Compass structure is the practice that made that marketing indefinite and institutionalized it at national scale.
The “inventory recovery” narrative rests on MLS-sourced figures: Zillow active listing counts, NAR existing-home reports, the FRED series economists quote. None of those sources captures listings currently sitting in phase one or phase two of a 3PM strategy. How many listings that is, nationally, is — by design — not publicly known. In a market that remains roughly 25 percent below pre-pandemic inventory levels, even a fraction of supply routed through a private channel changes what a buyer can see, what price they bid, and what they conclude about scarcity.
Two gaps, same geography
The two distortions are structurally independent. The non-agency RMBS market predates MLOS by two decades. Neither gap was designed to work in concert with the other.
But they converge on the same outcome in the same markets. The buyers most exposed to undercounted delinquency stress are concentrated in the Sunbelt remote-work migration metros — Austin, Phoenix, Tampa, Dallas — where the 2021-era prime cohort took on peak-price mortgages and employment conditions are now reversing. Those are also the markets where Compass has deepest brokerage penetration and where MRED’s national Private Listing Network is most actively expanding. A buyer in Austin or Phoenix is navigating both gaps simultaneously: the stress data doesn’t fully see their neighbors’ delinquencies, and the listing data doesn’t fully see the supply competing with the home they’re bidding on.
The national aggregate obscures this. High-income zip codes nationally are posting historically low delinquency rates — the NY Fed data is accurate. But national aggregate data averaging stable Midwest and Northeast markets with stressed Sunbelt markets is the wrong lens. The thesis is not that the national housing market is about to collapse. The thesis is that the buyers most likely to overpay in the most stressed markets are doing so with the least complete data.
The same structure appears in Who Sold You This?: a product that looks one way from the public data layer and materially different to participants with private access. The retail BDC investor was buying 2022-vintage private credit calibrated for the prior cycle. Here, the buyer stretching their budget at a Zillow-listed asking price is making a decision on data the seller’s agent may already know is incomplete.
The Intervention That Isn’t Getting Built showed why the political economy selects against supply reform — existing homeowners vote at higher rates on zoning questions than first-time buyers do. This piece shows the data infrastructure that would allow the market to price housing stress accurately being eroded through market-structure decisions at exactly the moment that stress is building. The political economy and the data architecture are moving in the same direction.
Neither piece says prices are about to fall. Both say the evidence that would tell you whether prices are about to fall is harder to see than it was two years ago — and hardest to see in the markets where the question matters most.
What would change my mind
The thesis is about data architecture. The falsification conditions are data conditions.
1. Non-agency RMBS prime delinquency stabilizes. Fitch’s quarterly U.S. RMBS Performance Monitor is the primary public source for the non-agency universe. If the Prime Jumbo 30-day delinquency rate stabilizes below 1.25 percent in the Q3 2026 report — reversing the current upward trend without a lag into 90-plus-day readings — the prime cohort stress signal is dissipating and the non-agency data gap is not amplifying anything worth worrying about. Score this as refuted if Q3 2026 Fitch data shows stabilization.
2. Zillow v. Compass/MRED restores public syndication and inventory responds. If the court issues an injunction restoring Compass’s private listings to public IDX feeds, and MLS-reported active inventory in MRED-served markets rises 5 percent or more within six months of reinstatement, the shadow inventory layer was real, it was measurable, and litigation reversed it. Score as partially refuted — the gap existed but was correctable.
3. The Sunbelt delinquency signal reverses without foreclosures. If Superprime 90-plus-day delinquency rates return to their mid-2024 baseline in VantageScore’s Q3 2026 CreditGauge without a corresponding uptick in foreclosure filings in Austin, Phoenix, or Tampa metro areas, the spike was a measurement artifact — possibly related to the return of student loan reporting into credit bureau data, which produced similar statistical anomalies in 2024. Score as refuted if normalization holds through Q4 2026.
All three conditions are trackable with public data on defined timelines. None has been met.
Related: Who Sold You This? — the same two-tier information gap appears in private credit: public NAV figures one story, private-access data another. The Intervention That Isn’t Getting Built — why the political economy that produces supply constraints and the data architecture that obscures them are moving in the same direction. Paying BNPL First — the same Sunbelt prime cohort likely carrying hidden BNPL balances that no mortgage underwriter can see.
The investor companion to this piece — what a prime-cohort housing stress event looks like for portfolios, and why the policy tools designed to catch it are calibrated for the wrong crisis — is available to subscribers. Read it here.
If you found this useful, the best thing you can do is forward it to one person who would push back on it. I’d rather be wrong in public than right in private.