Susan Wachter's model: banks lend multifamily on the comp, not the fundamental

Start with the final ledger and work backward. Beverly Hills refused to process a 19-story, 165-unit project at 125-129 South Linden Drive. It will now process a 36-story, 350-unit one.

That is the whole autopsy in two sentences, but the mechanism deserves the full table. Californians for Homeownership, a housing advocacy group, sued the city in mid-2024 over its handling of the Builder's Remedy application; the developer entity, 9300 Wilshire LLC, filed its own suit that fall. Judge Curtis Kin consolidated the cases, and on August 12, 2025, the Los Angeles Superior Court ordered Beverly Hills to process the application. Within weeks the developer amended the project: from 19 stories to 36, from 165 units to 350, from 33 affordable units to 61, the hotel component growing from 73 rooms to 80. The project remains pending, and no permit status is in our file. But the amendment is the tell. The fight was expensive, so the developer sized the entitlement at the end of it to make the fight worth having. Refusing to process left the city with more than twice the density it had refused.

Now the second body. On June 18, 2025, the Massachusetts Appeals Court affirmed, in 2025 WL 1699014, that Braintree's denial of a Chapter 40B application by 383 Washington Street, LLC died on arithmetic. Braintree's zoning board had claimed the statutory safe harbor, the General Land Area Minimum, which turns on whether a town clears a 1.5% land-area threshold. The board computed 1.65%. The state's Housing Appeals Committee computed 1.39%, because Braintree had pulled 244 conservation acres out of the denominator and the committee put them back. The Superior Court affirmed the committee's math. The Appeals Court affirmed again. Twenty-six basis points, and the denial was vacated, with the permit ordered issued and the project proceeding on the committee's terms. Our read on the case is 84 out of 100, and the file's summary line is better than anything I could write: the denominator is the defense.

Neither of those risks lives in a comp.

The model that explains the blind spot

Susan Wachter holds a chaired professorship in real estate and finance at Wharton, and a 2021 citation study ranked her the second most-cited real estate economist in the world, so when her framework describes a market failure, the failure is usually real. The framework I want is 27 years old. In "Real Estate Booms and Banking Busts: An International Perspective," written with Richard Herring in 1999, she argued that banks lend against real estate on the basis of recent market comparables rather than underlying fundamentals. And because nobody can short a property market, there is no mechanism to pull prices back toward fundamentals when the comps run hot. Rising prices validate the comps, the comps justify more credit, the credit pushes prices, and the loop feeds itself until the cycle reverses and the banking system discovers what the comps were hiding.

Herring and Wachter were writing about international banking crises, not zoning hearings. The extension to entitlement is mine, so hold me to it.

A comp is a price memory. It records what a stabilized or traded asset fetched, nearby, recently. It carries no field for whether the city will process your application, and no field for whether 244 acres of conservation land belong in a safe-harbor denominator two administrative appeals from now. Yet those were the fundamentals that decided both of these projects. Every one of them was sitting in public records before any loan on either deal would have priced: Beverly Hills' housing-element posture was a matter of public record when the Linden application landed, and Braintree's land inventory was knowable to the acre.

Run the Wachter loop on that blind spot. A hot multifamily market generates rich comps. The comps price every parcel in the submarket as if entitlement were a formality, because the parcels that made the comps got through. Credit flows against that assumption. The fights that follow, the refusals to process, the safe-harbor litigation, the three-year administrative sieges, arrive after the loan is on the books, and the lender discovers it was never holding the risk it priced. It was holding a lawsuit with a rent roll attached.

One sided note on why the loop persists. Wachter's no-shorting condition is usually read as a statement about housing. It is truer about jurisdictions. You cannot short a zoning board. There is no instrument that pays you when a city refuses to process an application, so the only market participants with an incentive to price that outcome are the ones already long the parcel, and by then the mispricing is their problem to eat rather than to correct.

What the comp said versus what the record knew

The bitter joke in both files is that comp-based pricing was wrong in opposite directions. In Braintree, the record favored the developer: the arithmetic held through the Housing Appeals Committee, the Superior Court, and the Appeals Court, and a comp-driven read of a town with a hostile board would have overpriced the risk of a project the state machinery was ultimately going to order approved. In Beverly Hills, a comp set built on the city's manicured stability would have underpriced the risk of a municipal refusal war and entirely missed the possibility that the war ends with 185 more units than the filing. The comp was not conservative in one case and aggressive in the other. It was uninformative in both, which is the actual indictment.

Fundamentals are researchable. That is the part Wachter's model leaves as an exercise, and it is the part this platform exists for. The GLAM arithmetic and the refusal-to-process pattern were public record before either term sheet.

So, the falsifiable part. My bet is that multifamily construction lending is the next place entitlement risk gets priced explicitly, and that by 2029 at least one major construction lender will quote measurably different spreads for the same product in hostile versus compliant California jurisdictions, because the loss data will force it. The market is mispricing this today; 26 basis points in a Massachusetts denominator moved a whole project's fate, while the spread between lending into Braintree and lending into Beverly Hills rounds to zero.

If your credit memo prices the comp and not the jurisdiction, you are lending on the one number in the file that carries no information about whether the building gets to exist.