Did LA's mansion tax freeze the high-end market?

In November 2022, Los Angeles voters passed Measure ULA, billed as a mansion tax: 4% of the full sale price on property sales above $5 million and 5.5% above $10 million, on top of the small transfer tax every sale already paid. The money funds affordable housing and eviction defense. It covers everything - mansions, yes, but also apartment buildings, offices, and warehouses.

Because the tax switched on at a known date with a known price threshold, it created a natural experiment. This page measures what it did to the market, using only public records, with the analysis plan locked in advance. Below: what the tax looks like, how the fight over it unfolded, and what the data says.

You cannot rerun Los Angeles without the tax, so the next best thing is a comparison group: 16 cities in LA County - Beverly Hills, Pasadena, Long Beach, and others - that never changed their transfer taxes. They experienced the same interest rates, the same market cycles, the same regional shocks. If expensive sales fell everywhere, that is the market. If they fell only in LA City, starting exactly when the tax started, that points at the tax.

Two safeguards. The analysis plan - outcomes, comparison cities, time windows, statistical tests - was written and locked before computing any result, so the findings could not be steered after the fact. And no sale-price dataset is published for free anywhere, so prices were rebuilt from county property-tax records (California reassesses a property at its purchase price when it sells) and cross-checked against the City's own tax-collection records. Every number on this page traces to public data.

This analysis measures what the tax cost the market. It says nothing about what the tax bought, and a policy verdict needs both sides. Through April 2026 ULA raised about $1.2 billion from 1,633 transactions. The City's housing department reports 1,409 affordable units funded through its main housing program and more than 14,000 households provided eviction defense. Set the roughly 199 forgone or delayed sales against that, and the debate becomes a real tradeoff rather than a slogan on either side.

The map above is also a history lesson. The neighborhoods paying most of this tax concentrated their wealth over a century shaped by racially restrictive covenants, exclusionary zoning, and freeway routing that pushed cost and displacement elsewhere. A transfer tax on $5M sales lands on the present-day endpoint of that history. Measuring its market effects honestly, as this page tries to do, is one input into that larger question, not an answer to it.

With a single treated city, standard errors from a textbook regression overstate certainty, so this project leans on two stricter designs. The first, in the placebo-test tradition, pretends each comparison city got the tax and asks whether the real LA stands out from the sixteen: on raw estimates LA ranks fifth, because tiny cities produce wildly noisy fakes; corrected for city size, LA ranks first of seventeen (p = 0.059, the smallest value seventeen cities can produce). The second permutes over time instead of cities - conformal inference, in the trade - and asks whether any other stretch of months since 2018 looks like the post-tax period. Almost none does: p = 0.028, with a 95% interval on the effect of roughly −58% to −8%. The two tests rest on different assumptions and agree. Every number is reported; none was picked after seeing the results.

The most persuasive evidence needs no statistics. Sales in the $3-4M and $4-5M ranges, just below the tax's reach, barely moved, while sales above $5M fell 38% and above $10M fell 35%. A market-wide slowdown would not stop at exactly the dollar amount written into the law. A final check ran the whole design against a made-up start date of April 2021, before the tax existed: it finds nothing.

Property-tax records cannot see sales disguised as company transfers, where the seller hands over the LLC that owns a building instead of the building itself. The tax gave LA sellers a new reason to do exactly that, so part of the measured decline could be relabeling rather than disappearance. The records only run through June 2024. 2025 is excluded outright: the Palisades fire burned part of the taxed area, and no method can separate a tax effect from a disaster in the same months. The project repository logs all 32 analytical decisions in order, including the review round that cut the headline down.