The widow who has voted by post since her husband died. The nurse coming off nights. The soldier stationed in Germany. The farmer forty miles from the nearest county office, and the woman on a zero-hours contract whose shift does not bend to an earlier deadline.

Between them they are 43% of the American electorate. This autumn, some of their envelopes will not be opened — because they arrived on the Wednesday instead of the Tuesday, or because a document they have never needed is now required. None of them has been told which. Two courts disagree, and both are still sitting.

Nobody has counted them. And in an argument where almost every important number is missing, the side that has to produce one loses. That is what this fight is really about.

The fraud case is closed. That is when the argument started

Identified fraud in mail voting runs at roughly 2.5 cases per 10 million ballots — between 6 and 46 cases in each general election year from 2016 to 2022, and that is the figure you get after assuming the most generous possible level of underreporting. A peer-reviewed study found no evidence that voting by post raises the overall risk at all.

That looks like the end of it. It is not, and the reason came from the side the figure favours:

"That is a count of identified cases. Detection rates for ballot fraud are unknown. Defenders of the order could — and likely will — argue the figure reflects investigative capacity, not true incidence. We cannot rebut that without acknowledging the methodological gap."MiniMax

Nobody knows how much fraud goes undetected. Nobody knows how many voters lack the documents. Both sides are arguing from a hole. So the question is no longer how much fraud is there — it is who is required to fill the hole before they are allowed to win.

Filling it, and the trap in filling it

One answer is: fill it. Count how many ballots arrive late, how many voters lack the papers, what it costs to run an election in twenty-three states under one court order while the Postal Service operates under another that contradicts it.

"Quantification is not endorsement; it is ammunition."GLM

The other answer is that picking up the shovel is the concession:

"To pivot to administrative cost analysis is to concede the legitimacy of the restriction itself. It is like debating the optimal speed limit on a road where no traffic violations have occurred."MiMo Flash

This is not a philosophical nicety. It is the choice in front of every state attorney general, every campaign group and every person arguing about this at a kitchen table. Start counting the harm and you have quietly agreed that harm is the currency — that if the number came out small enough, the rule would be fine. Refuse to count, and you walk into a courtroom holding a principle and nothing else.

"Principle without evidence loses cases. Evidence without principle loses moral clarity."MiMo

Which would settle it, except for one thing: some of the people this rule lands on cannot be counted at all, and that is not an accident of statistics.

The ones who never become a number

"Not everything that matters can be measured, and not everything measurable is what matters."DeepSeek

The person who never attempts to register because she has no birth certificate does not appear in a registration statistic. The voter who stays home because two courts disagree and nobody can tell him whether his ballot counts does not appear in a rejection statistic. He appears in a turnout figure, indistinguishable from apathy.

"Fear, not fraud, can keep people from participating."

So the demand for a number is not neutral. Ask the widow's side to prove the harm and you have asked them to produce evidence that the rule itself is designed to erase. The harm the rule claims to prevent has been measured and is tiny; the harm it may cause is unmeasured, and its being unmeasured is being offered as proof that it is small.

That is why the closing answer was not about evidence at all:

"When a restriction lacks any demonstrated public benefit and risks disenfranchising people, the tie goes to the voter, not the state."DeepSeek

And why the sharper reply was that calling it a tie already gives away too much:

"The default is access. Any deviation is a breach of contract with democracy itself."MiMo Flash

The size of the hole

Here is what nobody could put a number on, after an hour of trying. How many eligible voters lack documentary proof of citizenship. What share of postal ballots currently arrive after election day. What implementing the order would cost to administer. Whether coercion inside a household — far harder to detect than a forged ballot, and barely studied — matters more than the individual fraud everyone does measure.

That list is the whole argument in one place. Each missing figure is a job, and the fight is over whose desk it lands on.

There was also a long detour into replacing the post with cryptographically verifiable digital voting, and it collapsed on the same test it was being used to escape: it arrived with no costs, no pilot data and no security audit. The 2010 Washington D.C. internet-voting pilot was broken into by university researchers and abandoned. Estonia, twenty years in and still the only country running nationwide internet voting, draws peer-reviewed criticism on precisely the point no cryptography has ever solved — the voter's own laptop. Asked to prove its case, it could not either.

None of which helps the widow. She is still holding an envelope, waiting on two courts that disagree, with nobody able to tell her whether it counts — and, if it does not, no number anywhere that will record that she was there.