In November 1989, choking on ozone, Mexico City ordered most of its cars off the road one weekday a week, picked by the last digit of the plate. Thirty-seven years on, the ban is still running — five ozone alarms fired in the first four months of 2025 alone. Four peer-reviewed studies have examined what it achieved. Not one of them found what it was built to deliver.
The rule is simple enough to explain in a sentence, which is part of why it has survived thirty-seven years. Look at the last digit of your number plate; it tells you which weekday you may not drive. Mexico City started keeping cars off the road one weekday a week by the last digit of their plate in November 1989, after an ozone emergency, and made the rule permanent the following year. The idea was older than the law — an ecologist group had launched a voluntary ‘Un Día sin Auto’, a day without the car, back in 1984 — and the emergency made it compulsory: it began on 20 November 1989 as a winter-season measure, and in 1990 it was made permanent, with the vehicle-verification programme established alongside it. Like many emergency measures, it never ended.
1989one of the oldest plate-based driving restrictions still running anywhere
4peer-reviewed studies of the ban; not one reports cleaner air
11 monthsone study's measure of how fast second cars filled the gap
8pollutants tested for the 2008 Saturday extension, with virtually no effect
The hologram is the interesting part
Strip away the plate digit and what remains is a measurement with teeth. A compulsory emissions test every six months assigns each car a hologram, and that reading decides how hard the ban bites: holograms 00 and 0 drive every day, hologram 1 also rests two Saturdays a month, hologram 2 every Saturday. A clean car buys its way out of the restriction by being clean; a dirty one is restricted more.
That is a genuinely unusual arrangement, and it is why this entry sits where it does in the atlas. Most air-quality programmes measure the city. This one measures the individual vehicle and attaches a consequence to the result, twice a year, for millions of cars. It is the same logical structure as a flood barrier that closes on a forecast, or a congestion charge that reads a plate — a reading, a threshold, an automatic effect on somebody's day.
There is a second measurement layer above it. When the metropolitan monitoring network pushes ozone past the contingency threshold the regional environment commission declares a ‘Fase I’ that pulls even exempt cars off the road. CAMe activated five ozone Fase I contingencies in the first four months of 2025 alone, and the mechanism was still firing the following spring, holding a Fase I in place through a Sunday in April 2026. So the system has both a slow trigger, the six-monthly test, and a fast one, the live ozone reading.
Four decades in nine dates
| When | What happened | Who says so |
|---|
| 1984 | “Un Día sin Auto” begins as a voluntary citizens’ campaign by an ecologist group | Press chronology (El Universal) |
|---|
| Nov 1989 | Launched during an ozone emergency, at first for the winter season | Press (El Financiero / El Universal) |
|---|
| 1990 | Made permanent; the vehicle-verification programme is established | Press chronology (El Universal) |
|---|
| 2008 | Davis publishes the null result in the Journal of Political Economy | Peer-reviewed |
|---|
| Jul 2008 | Saturday restrictions added, tied to the hologram | Peer-reviewed / operator |
|---|
| 2013 | Gallego, Montero and Salas measure the second-car response directly | Peer-reviewed |
|---|
| 2017 | Two further studies: the Saturday extension fails its own test, and a survey-based study rejects the second-car explanation | Peer-reviewed |
|---|
| 2025 | Five ozone contingencies declared in the first four months alone; camera enforcement reported in the press | Press |
|---|
| 2026 | A Fase I contingency holds through a Sunday in April; the programme applies unchanged | Authority (CAMe) |
|---|
What four studies found
The evidence base here is unusually good. It is also almost entirely negative — and where it disagrees with itself, it disagrees only about the mechanism, never about the result.
The load-bearing study is Lucas Davis's 2008 paper in the Journal of Political Economy, which examined the programme's first two decades and reported that ‘across pollutants and specifications there is no evidence that the restrictions have improved air quality’. That sentence has now stood for eighteen years. Worse for the policy, the same paper found evidence that the ban raised the total number of vehicles in circulation and shifted the fleet toward high-emission cars.
The mechanism seemed to be pinned down five years later. Gallego, Montero and Salas, in the Journal of Public Economics, measured an initial drop in car use followed by a rise in the household vehicle stock within eleven months. That, for a decade, read as the whole story in one clause: the restriction works exactly as designed for about a year, and then the households that can afford to buy their way out of it do, with the cheapest vehicle that will fill the missing weekday. The cheapest available vehicle in Mexico City is an old one, and an old one pollutes more.
When the city extended the ban to Saturdays in 2008, that extension became its own natural experiment, and Davis tested it in Scientific Reports in 2017. The result: the July 2008 Saturday extension had virtually no discernible effect across eight pollutants, and no sign that drivers switched to metro, bus or light rail. The second half of that finding matters as much as the first. The theory of the policy is that a driver denied a car takes public transport. The measurement says they did not.
A fourth study complicates the story rather than completing it. Guerra and Millard-Ball, working from household travel surveys, confirmed that the ban does little to reduce driving — but they reject the prevailing theory that its lack of success is due to perverse incentives for households to buy second cars. Most households, they found, avoid the ban by shuffling weekly travel routines: the family car simply runs its errands on the days it is allowed out. The four studies disagree about why the restriction failed to clean the air. None of them disagrees that it failed to.
Why it is still running
A policy with four negative evaluations that has survived thirty-seven years is worth taking seriously rather than mocking. Several things keep it alive at once.
It is legible. Every driver understands it, which is rare in air-quality policy and matters for compliance. It is cheap: no gantries, no accounts, no billing system — for most of its life, enforcement was a police officer reading a plate. And it is visible, which for an emergency measure is close to the point; a government facing an ozone crisis can announce it on Monday and have it running by Thursday.
There is also a harder argument, and the record does not settle it. The counterfactual is not “Mexico City with clean air”; it is Mexico City with the same fleet growth, the same geography — a basin at 2,240 metres that traps its own exhaust — and no restriction at all. None of the four studies claims the city would have been better off without the programme. They claim it did not deliver the improvement it promised, which is a narrower and more defensible finding.
What this atlas cannot tell you
Two newer parts of the system are unevaluated, and both are the parts that would most plausibly change the verdict.
The first is automated enforcement. Press reporting describes photo-enforcement cameras in Mexico City that sanction Hoy No Circula violations as well as speeding. If that is now the primary enforcement mode, the compliance economics have changed materially since the studies were done — but this rests on a single media report, and no operator document or evaluation confirms the scale.
The second is the contingency trigger. The Fase I mechanism is the part of the system that behaves like a modern sensing programme: a live reading crossing a threshold and producing an immediate, automatic legal effect. Nobody has isolated it. The four studies measure the standing weekday restriction; none of them separates out what happens on the handful of days a year when the ozone network forces everyone off the road. It is entirely possible that the fast trigger works and the slow one does not, and the published record cannot say.
