Half a Point

The first: business groups had bought television advertising against Andrés Manuel López Obrador in violation of Mexican electoral law. The second: Felipe Calderón was the president-elect of Mexico, by 233,831 votes out of roughly 41.5 million cast — a margin of 0.58 percent.

Both findings are in the same document. Both were reached by the same magistrates on the same day. And the second one was final in the strongest sense the Mexican constitution allows: under the reforms of the 1990s that built the Tribunal, there is no appeal from it on electoral matters, no supreme court sitting above it, no further body to petition. September 5 closed the legal question permanently.

It closed nothing else. Eleven days later, on September 16 — Independence Day — a convention in the Zócalo named López Obrador the legitimate president of Mexico. On November 20 he took a symbolic oath and a sash. He kept a shadow cabinet and toured the country as a head of state without a state for years afterward. Calderón was inaugurated on December 1 and governed for a full six-year term, and for a full six-year term a substantial fraction of the country held that he had not been elected.

The easy reading is that this is a story about a man who would not lose gracefully. That reading is available and it is not wrong about López Obrador's temperament. It is wrong about the mechanism. The interesting thing that happened on September 5, 2006 was not a candidate's refusal. It was an institution discovering that its instruments and its problem were different shapes.

No Casilla to Void

Mexican electoral law, then as now, sets a threshold for annulling a national election. An irregularity does not void a result because it happened; it voids a result because it was determinante — determinative of the outcome. That is a sensible standard. Every serious electoral system has some version of it, because the alternative is that any provable infraction anywhere cancels a national vote, which is an invitation rather than a safeguard.

It would be convenient to say that annulment was the Tribunal's only tool, and it would be false. The magistrates could work at much finer grain, and did. They ordered a recount at roughly nine percent of polling places. They annulled the returns from individual casillas where the tally sheets did not survive inspection. They could refer the advertising to the Federal Electoral Institute for administrative sanction against the associations that bought it. Station-level annulment, referral, fines — the repertoire was graduated, and the Tribunal used the graduations.

Now look at what every one of those instruments has in common. They are all local. Each one attaches to a physical object or a legal person: a ballot, a tally sheet, a polling place, an association with a bank account. That is exactly why they are precise.

The harm at issue was neither an object nor a person. Illegal television advertising, bought by private business groups, running nationally against one candidate, at some unknown level of effectiveness, is a continuous quantity smeared across forty-one million people. It moved some number of votes. That number is unknown and probably unknowable, but it is a real number and it is somewhere on a line. There is no casilla you can void to undo a television campaign. No tally sheet has the ad buy on it. Fining the association that placed the spots does not give anyone back the vote they would otherwise have cast.

So the graduated remedies were real, and not one of them could reach this class of harm. What remained at the level of the national result — the only level at which a diffuse national harm exists at all — was two values. Annul, or certify.

And so the Tribunal rounded. It found the advertising illegal, found the effect unproven at the scale the law requires, and rounded the harm to zero, because zero is the only value the certification output can express. Everything below the annulment threshold is, in the language of the ruling, legally identical to nothing at all. Not "a small crime with no remedy." Nothing.

There is no evidence — none produced in 2006, none produced in the twenty years since — that the count itself was rigged in a way that reversed the outcome. The partial recount voided nearly 238,000 ballots for irregularities, a number larger than the margin, and that figure has been waved around ever since as though it were a stolen bloc. It is not. Voided ballots are voided in both directions; irregular tally sheets are noise, not theft, and the recount did not move the lead. Anyone who says the illegal ads moved 234,000 votes is saying something they cannot show, and the burden is theirs. Calderón's presidency was legal, and the Tribunal that said so was doing its job.

None of that touches the problem. The problem is not who won. The problem is what an institution does with a wrong it has formally acknowledged and structurally cannot price.

Two Branches, Same Facts, Opposite Magnitudes

Watch what the rest of the Mexican state did next.

Within roughly a year, Congress amended the constitution and rewrote the electoral code to prohibit precisely the conduct the Tribunal had found insufficient. The 2007–2008 reform banned the private purchase of radio and television airtime for electoral purposes outright. All broadcast campaign advertising was routed into state-allocated time administered by the Federal Electoral Institute. Restrictions were placed on attack advertising and on the appearance of sitting officials in government publicity. It was one of the most sweeping media-and-money reforms any democracy has passed in the modern era, and it was aimed squarely at the thing the Tribunal had waved through.

Read those two acts side by side. The Tribunal looked at the illegal advertising and concluded it was not large enough to matter. The Congress looked at the same illegal advertising and concluded it was dangerous enough to require a constitutional amendment. Two branches, identical facts, opposite magnitudes.

They are not contradicting each other. They were asked different questions. The Tribunal was asked a threshold question about one race: did this exceed the line? Congress was asked a proportionate question about the system: is this tolerable in general? Only the second question has an answer that can be sized to the harm. The first question can only say yes or no, and it said no, and the no was correct on the law and false to the finding recorded in the same opinion.

What that combination produces, from the outside, is the single most destabilizing message an institution can send: you were right about the wrong, we are not going to do anything about it, and the matter is now permanently closed. If the Tribunal had found no illegality, López Obrador would have had a grievance and no evidence. If it had annulled, he would have had a remedy. What it actually handed him was evidence without remedy, signed by the state, notarized as final.

The Interval

The parallel presidency, read as theater, is absurd — a sash, a cabinet with no ministries, a tour of municipalities that had already sworn allegiance to somebody else. Read as infrastructure, it is something else: a national organization, maintained in continuous operation, with a permanent standing claim about the illegitimacy of the government it shadowed.

The arc from there is a matter of public record. López Obrador ran again in 2012 and lost. He broke with the PRD and founded Morena in 2014. He won in 2018 with about 53 percent — the largest presidential share of Mexico's democratic period — and Morena took Congress with him. In 2024 Claudia Sheinbaum won with roughly 60 percent and the coalition took the supermajorities that a constitutional amendment requires.

And with those supermajorities came the judicial overhaul. On June 1, 2025, Mexico became the first country on earth to elect its entire national judiciary by popular vote: nine Supreme Court justices, five members of a new disciplinary tribunal, 464 circuit magistrates, 386 district judges. Also on that ballot, further down: seventeen magistrates of the Electoral Tribunal of the Federal Judiciary — two for the Superior Chamber, fifteen for the regional chambers. The body that certified Felipe Calderón now seats its members by election. Turnout was about 13 percent. Morena-aligned candidates won across the board.

It would be too tidy to call this revenge for 2006, and the tidiness would be a lie. The judicial overhaul had its own immediate causes — six years of collision between López Obrador's government and a Supreme Court that kept striking down his energy and electoral legislation — and the legislators who voted for it were not litigating a nineteen-year-old margin. Causation here is not a plot. It is an accumulation.

But it is worth being exact about how an accumulation of this kind actually works, because the usual metaphor is wrong. Grievances do not go on the books and sit there accruing interest, waiting for someone with standing to collect. Nothing accrues on its own. What happened between 2006 and 2018 is that a specific man kept a specific grievance in continuous operation for twelve years, at his own expense, and then converted the organization built to hold it into a political party, and then won the country with it. The sash and the shadow cabinet were not theater in retrospect. They were the early inventory of Morena.

That is the fragile version of the claim and it is the true one. An institution that acknowledges a wrong and declines to remedy it is not manufacturing an inevitability. It is leaving an opening, and whether anyone walks through it depends on whether a person with the stamina to carry a grievance for a decade happens to be standing in the room when the ruling comes down. In September 2006, one was.

So the ruling was never overturned. It did not need to be. The court that issued it was put on a ballot nineteen years later, and thirteen percent of Mexico decided who would sit on it next.

There is a difference between a decision that is final and a decision that holds. Finality is a property of the legal record, and it is available on demand — any competent tribunal can produce it in an afternoon. Holding is a property of everyone who was affected, and no tribunal can produce it at all. On September 5, 2006, the Electoral Tribunal of the Federal Judiciary had the authority to deliver the first and no instrument whatsoever for the second, and it did what institutions do when the only tool in the drawer is the wrong size. It used it, called the matter closed, and started a clock.

The Remedy Rounds Too

It would be satisfying to stop there, with the clock running out on schedule and the account finally settled. The essay's own standard will not permit it.

Apply to June 1, 2025 exactly the test applied to September 5, 2006. What is the instrument, and what shape is the thing it is being asked to measure?

Judicial legitimacy is continuous. It is assembled out of a thousand particulars — how a magistrate reasons, whether she can be reached, whether the party who lost in her courtroom believes she read the file. A ballot cannot express any of that. A voter facing lists of hundreds of names, for offices whose functions are not widely known, casts one bit per candidate: yes or no, in practice by way of a slate recommendation handed out at the door. That is not a finer instrument than appointment. It is a coarser one, deployed at enormous scale, and it rounds a continuous quantity in precisely the way the Tribunal did — across more than eight hundred offices at once.

It also produces less of the thing this whole argument has been about. Whatever else can be said of the magistrates who ruled in 2006, they were seated by a process the political system had broadly signed off on, which is why López Obrador had to build a movement rather than simply point at a defect. The magistrates seated in 2025 were chosen in an election roughly thirteen percent of the electorate attended. If the complaint against September 5, 2006 is that a decision can be final without holding, then a judiciary seated by 13 percent has not answered the complaint. It has restated it, with a much larger denominator.

So the arc does not close. 2006 did not produce 2025 the way a diagnosis produces a cure. It produced it the way an unpriced harm produces a second unpriced harm. The nineteen years between them are not the story of a wrong being righted; they are the story of the same operation being run twice, in opposite directions, by different hands.

Which means the diagnosis was one level too shallow the whole time. The problem in 2006 was not that the Tribunal was holding the wrong tool. There is no right tool. Legitimacy is not a substance that some sufficiently well-designed mechanism manufactures and dispenses — not a tribunal, not a ballot, not a supermajority. It is a property of who was actually included in producing an outcome and whether they can find themselves in it afterward. Mexico has now put that question to the same institution twice. In 2006 the answer was: the magistrates of the Superior Chamber, with no appeal from them to anyone. In 2025 the answer was: thirteen percent, once.

Both are final. Neither is an answer to the question of whether it holds, because that is not a question any of these instruments is built to ask, and the clock does not stop when nobody is measuring.