Chris Bruen's math: 93 of the last 99 quarters ran supply and rent growth in opposite directions

Ninety-three of ninety-nine.

That is the score across roughly 25 years of quarterly data, per Chris Bruen, Senior Director of Research and Chief Economist at the National Multifamily Housing Council. In a 2025 NMHC research note, "Unpacking the Relationship Between Jobs and Apartment Demand," Bruen modeled rent growth as a function of two variables, net unit deliveries and employment growth, across more than 150 major metros. The headline result: in 93 of the last 99 quarters, delivery volume and rent appreciation moved in opposite directions. More deliveries, softer rents. Fewer deliveries, harder rents. Six quarters out of 99 broke the pattern.

I have stopped expecting any finding in this business to run at a 94% hit rate. This one does.

The model's structure matters as much as the score. Jobs drive apartment demand; deliveries answer it; rent growth is the residual that clears the difference. Which means rent is not really a price for apartments. It is a price for the gap between how fast a metro adds paychecks and how fast it adds keys. Employment growth is set by the economy. Deliveries are set somewhere much smaller and much stranger: hearing rooms.

So read our multifamily case files the way Bruen's regression would read them. Not as local political drama. As the delivery variable being decided, unit by unit, in public.

243 units, decided by a recall election

Fairfax, California, population about 7,500, has roughly 3,500 housing units. On May 8, 2025, Mill Creek Residential filed for 243 units in six stories on a 1.92-acre site at 95 Broadway. One project, 7% of the town's entire housing stock. Our case file calls it a 3% population event, and none of the town's politics disagreed with that arithmetic; they just disagreed about whether it should happen.

The town issued an incompleteness letter on June 4, 2025. Opposition organized around wildfire evacuation on a single egress road, around scale, around displacement of the existing commercial tenants, around traffic. By September 8, 2025, the town acknowledged that the state's housing department read the law against it and said the project would be processed ministerially, on the state-mandated path that removes discretionary hearings. The response was not an appeal. It was a recall campaign against Mayor Lisel Blash and Vice Mayor Stephanie Hellman, which failed on November 4, 2025, at 2,435 to 1,936 and 2,401 to 1,958. The officials survived by roughly ten points. The project is moving forward; our read is 52 out of 100, because ministerial eligibility made it legally viable without making one week of it politically easy.

In Bruen's frame, strip out every named person and what happened is this: 243 units stayed in the delivery pipeline of a supply-starved metro because a state statute outvoted a town. The rent consequence of that outcome is not rhetorical. It is the dependent variable.

The units nobody will ever count

Fayetteville, Arkansas, shows the other branch. LCD Acquisitions asked to rezone 1875 W. Haskell Heights from single-family RSF-4 to a planned zoning district. The planning commission raised hillside and drainage concerns at a hearing, and objected to the density. No revised plan came back. City staff recommended denial, and in March 2025 the commission denied unanimously. One commissioner's documented explanation stands for the whole record: "The developer had two hearings to show they understood this hillside. They didn't submit a revised plan either time. We had no choice but to deny."

Here is the detail I keep returning to. Our file does not state the project's unit count, because the record never firmed one up before the project died. We score the episode 15 out of 100, and the missing number is the point. Denied projects do not leave a delivery count behind. They vanish from the supply data entirely, which is why metro-level statistics feel the loss only as an absence: a delivery total lower than it would have been and a rent print higher than it should have been, with no line item explaining why.

Bruen's 93-of-99 is built on the units that showed up. The inverse relationship it documents is powered, invisibly, by the ones that did not.

One honesty note, because the standard here is citing things properly. Bruen's regressions run at metro resolution, across 150-plus markets. A town of 7,500 sits far below the model's grain, and nothing in his note prices a single Fairfax or a single Fayetteville. But the mechanism does not care about the dataset's resolution. A metro's delivery total is nothing but the sum of its hearing rooms, and every quarter, the sum of decisions like these two is what lands on the supply side of his equation while employment lands on the demand side. Jobs arrive by commute. Deliveries arrive by vote. Rent clears the difference.

That asymmetry is the quiet scandal of the whole framework. A town cannot opt out of the demand side; it imports demand every morning on the freeway. It can only opt out of the supply side, and when it does, Bruen's math says the bill lands on tenants across the metro, in a rent print nobody traces back to the hearing where it was written.

My bet, and it is checkable with public data: rank U.S. metros by units denied or stalled per 1,000 existing units over 2024 through 2026, and by 2028 that ranking predicts metro rent growth better than any demand-side variable in the standard toolkit. If deliveries and rents can run in opposite directions for 93 of 99 quarters, the places suppressing deliveries hardest are choosing the top of the next rent-growth table, on the record, one unanimous vote at a time.

Somebody should tell them. It will not be the tenants' rent statement, because that document never cites its sources.