Enforcement began September 2023. The central pitch was that returning short-stay apartments to the long-term market would ease rents. We measured it, me and my AI agents: 414,061 rental listings from an 80-million-record market atlas, one quarter at a time. Median asking rent is up 9% since enforcement began, and 14% above the last full pre-ban year. It never dipped once.
Then we ran the sharper test nobody had run: if the mechanism were real, the neighborhoods that lost the most short-stay listings should have gotten the most relief. We ranked every zip by pre-ban listing density and checked. The neighborhoods with nine times the exposure got exactly nothing extra.
To be fair about what this is: asking rents, not signed leases, and correlation, not causation. Rents move for many reasons. But the ban's promise was specific, public, and measurable, and this is the measurement. Whatever the ban accomplished, cheaper rent isn't it. The parties visibly better off are the hotels, which hit record rates (that one is from the newspapers, not our atlas), while ordinary hosts lost income and visitors pay more.
If short-stay conversions genuinely threaten housing supply, there are honest tools for that: caps, taxes, per-host limits. A blanket ban that delivered none of its promised relief is not housing policy. It's a favor wearing one.
Before asking whether the ban worked, ask what it could possibly have done. Here is the entire whole-home short-stay market the ban abolished, next to the city it was supposed to fix.
If every one of those 22,254 homes had converted to a long-term rental on enforcement day, New York's housing supply would have grown by six-tenths of one percent, about two percent in the densest quartile of zips. That is the ceiling: the largest rent effect the ban could have produced under its own most generous assumptions, and it sits at the edge of what's even measurable in a market this size.
And the market being abolished wasn't the cartel of the imagination: 81% of those listings belonged to hosts with exactly one listing, and 91% to hosts with two or fewer. Mostly spare homes, not hoarded portfolios. The rest of this page measures what actually happened; this section is why nothing else on it should surprise you.
The ban's logic is local: a converted short-stay unit becomes a home on its own block. So relief should concentrate where listings were dense. Each dot below is a NYC zip code: how many entire-home short-stay listings it had per 1,000 homes before the ban, against what happened to its median asking rent from the last pre-ban year to the last twelve months.
Read those tiles left to right, because they are the whole story. Over the full period, the neighborhoods with roughly nine times the short-stay exposure saw the same rent growth as the neighborhoods with almost none: +12.8% against +12.3%. No relief. And in the first year after enforcement, the year twenty-plus thousand delisted units were supposedly flooding home, the most-exposed neighborhoods actually rose nearly twice as fast: +5.2% against +2.8%. Not only did the promised local relief never arrive, year one delivered the opposite.
To be precise about the claim: we are not saying more airbnbs caused higher rents. We are saying the ban's mechanism predicts the dense zips should have cooled relative to everywhere else, and they did not, in any window, by any cut: correlation r = −0.17 across all 80 zips, statistically indistinguishable from zero, flat within boroughs (r = −0.18), flat weighted by listing volume (r = −0.09), flat excluding outliers (r = −0.05). Williamsburg's 11211, the densest zip in the city at 45 listings per 1,000 homes, rose 10.7%, right at the citywide norm.
This is the test the citywide average can't fake. If the mechanism worked at all, it would show up here first. There is nothing here.
Every cut below is a different way the ban's rent promise could have shown up in the data. None of them found it.
This is the chart an economist would ask for first. The two groups tracked each other for six straight quarters before the ban, which is what makes the comparison fair, and then kept tracking for twelve quarters after it, which is what makes the ban look like nothing. If removing short-stays eased rents, the gold line falls away from the blue one after the gray marker. It crosses it instead, five times.
Before rents can fall, the banned apartments have to actually show up as rental listings. They didn't. The most-airbnb neighborhoods' share of the city's long-term listings went down after enforcement, not up. Whatever happened to those twenty-two thousand entire homes, becoming your next apartment wasn't it.
The ban never made the homes come back. It made them change one field. In 2019, 12% of entire-home listings had 30-night minimums; today it's 90% of 16,808 homes, sitting on the same platform as furnished monthlies, above the 30-day line the law draws. Not tourists' homes returned to renters: renters' homes converted to a product most renters can't use.
If the ban could work anywhere, it was here: one home in twenty-two was an entire-home short-stay. Three years later Williamsburg's rent is up 7% since enforcement against the city's 9%, a difference inside the noise. The single most treated neighborhood in America looks exactly like everywhere else.
The spirit of this page: a one-day, for-fun exploration of public data by me and my AI agents. Every "we" on this page means exactly that. It is not journalism, not legal advice, not investment advice, and it is not meant to be quoted, cited, or republished. If you need numbers you can stand behind, pull the primary sources linked above yourself.
The rent series is asking rents from broker and landlord feeds, not signed leases; listing mix (unit sizes, neighborhoods, feed coverage) shifts over time and is not controlled for. Covid-era dynamics are entangled in the 2021–2022 baseline. Correlation is not causation.
The exposure measure is the 2019 vintage of the public listings snapshot, the last one freely mirrored; the mid-2023 snapshots are no longer publicly downloadable. Airbnb's NYC geography was highly persistent from 2015 through the ban, and the test needs neighborhoods ranked correctly rather than counted exactly, but a 2023-vintage dose would be strictly better and we'd rerun with it gladly. Absence of a local effect is evidence against the claimed mechanism, not proof no unit anywhere returned.