What This Requires of Us

I. Two things the record establishes, and one it does not

The eighteen cases in this collection are not a survey. They were selected because they are documented, and the documentation is uneven for reasons that are themselves part of the problem. They cannot tell you how often American elections are misreported.

They establish two narrower things, and they establish them well.

The reporting layer is structurally under-verified. Counting is audited in a growing number of states, under regimes that are getting better. Registration is checked. Ballot eligibility is adjudicated, often bitterly. The arithmetic step between what a precinct’s tabulator printed and what a Secretary of State published — the simplest step, the most mechanical, the easiest of all of them to check — is the step almost no one checks. That is not an inference from the cases. It is the condition the cases keep revealing.

When that layer fails, the evidence needed to resolve it is frequently unavailable. Records were destroyed in Ohio in 2004 in the face of a federal retention statute and a preservation order. Ballots were destroyed in Alabama in 2002 on a lawful schedule while the dispute was live. Tapes went unopened in Kansas across five years and two lawsuits. Trade-secret law shields the software. Public-records denials are upheld. The pattern is consistent enough that the taxonomy treats access failure as a category of vulnerability in its own right, and the missing-evidence rule exists in American law precisely because evidentiary vacuums of this kind are a recognized problem that courts have had to develop second-best tools for.

What the record does not establish is that any American election was stolen. No case study here makes that claim, and this page does not make it either. Several of the cases are unresolvable, and unresolvable is a different finding from wrongdoing — a distinction this collection has been careful about and intends to remain careful about. The argument below does not require the stronger claim. It requires only the observation that a system which cannot answer the question is a system that has failed at something, whatever the answer would have been.

II. The standard already exists in American law

The United States has confronted this problem before, in a domain where the stakes were money rather than self-government, and it solved it.

After 1929, Congress concluded that a market in which companies reported their own results, on their own terms, with no independent verification, was not a market that could sustain public confidence — regardless of how many companies were honest. The Securities Act of 1933 and the Securities Exchange Act of 1934, which created the Securities and Exchange Commission, did not attempt to make corporate officers more trustworthy. They made trustworthiness unnecessary. Public companies must file standardized disclosures on a fixed schedule. Those disclosures must be audited by an independent firm that does not work for management. The filings go into a public database any citizen can search for free. The auditor’s independence is not a courtesy; it is a statutory requirement with penalties attached.

The result is a system in which nobody has to trust anybody. That is the point. An investor evaluating a company is not asked to form a view about the character of its chief financial officer. The verification is structural, routine, boring, and mandatory, and because it is all four of those things it works.

American elections are held to a materially weaker standard than American quarterly earnings. A precinct’s results pass from a tabulator to a county to a state to a public website, and at no point in most jurisdictions is any independent party required to confirm that the number that went in is the number that came out. A public company that reported its revenue the way a county reports its vote totals would be in violation of federal law.

This comparison is not ours originally. It is the observation that started this organization. America Counts’ co-founder came to election technology from the chief executive’s seat at a public company, where he had spent years subject to exactly this regime, and asked the obvious question. (How that became Actual Vote is told in the technical manual.) If the public interest in honest accounting justifies mandatory independent audit, what does the public interest in honest elections justify?

III. What we think should follow

Four things, in ascending order of difficulty.

Poll tapes should be posted publicly, everywhere, as a matter of law. This is the cheapest reform available and the one with the widest reach. By our state-by-state survey, twenty-five of the fifty-one jurisdictions already require posting or reliably practice it. A results tape posted on the door of a polling place at the close of voting is a public record of what the machine in that room produced, created at the moment of lowest possible opportunity for interference, and readable by anyone who walks past.

For many of the remaining jurisdictions this is a procedural change and nothing more: the scanner in the polling place already prints the tape at the close of voting, and the reform is a line in a poll-worker manual telling someone to tape it to the glass. Not all of them, though, and the exception is worth naming rather than eliding. Where ballots are tabulated centrally instead of at the polling place — the all-mail states, the central-count jurisdictions, and the remaining paperless equipment — there is no precinct tape to post, and no manual revision produces one. Those places need a different remedy, whether that is prompt precinct-level publication of central-count results or new equipment, and pretending otherwise would understate what the reform costs. Which jurisdictions fall on which side of that line is an equipment question best answered against the Verified Voting equipment database, not asserted from a policy page.

Official results should be published at precinct level, in machine-readable form, on a fixed schedule. A jurisdiction that posts tapes but publishes only county aggregates has made verification impossible in a different way. The comparison that matters requires both halves.

Independent verification of the reporting layer should be required, not merely permitted. The technical manual’s recommendations set this out in operational detail. The principle is the one the 1934 Act settled: verification that depends on the good will of the party being verified is not verification. It should be a duty owed to the public, performed by someone with no stake in the outcome, on every election, whether or not anyone has complained.

And the country should build the profession to do it. This is the piece that does not exist yet, and the one we think is ultimately necessary. Audited financial statements are not produced by volunteers; they are produced by a credentialed profession with published standards, examinations, continuing education, independence requirements, and consequences for failure. Election auditing has no equivalent. There is no body that sets the standards, certifies the practitioners, or holds them to a code. America Counts believes there should be one — a national federation of election auditors, professional in the way the accounting profession is professional, whose members are qualified to do this work and accountable for how they do it. We are working on what that would look like and will have more to say when we do.

A federal statute would reach federal contests only. In practice that is a smaller limitation than it sounds: almost no jurisdiction is going to run two parallel election systems, one for the federal race at the top of the ballot and another for everything below it. The federal standard would become the standard, the same way securities regulation set norms well beyond the companies it strictly bound.

IV. Why this is not a partisan argument, and how you can tell

Every reform above is indifferent to who wins. A verification requirement catches an error that helps a Republican exactly as readily as one that helps a Democrat, and the case studies in this collection document both. Baldwin 2002 and Antrim 2020 point in opposite directions. So do Cook 2016 and the Ohio 2004 lockdown. Any reader who finds the argument congenial only because of who they think it would embarrass has misread it, and the reform they are imagining is not the one being proposed.

There is a version of this argument that we are deliberately not making. We are not claiming that American elections are rigged, that officials are untrustworthy, or that any recent result was wrong. The overwhelming majority of election administrators in this country do careful work, in public, for inadequate pay, under a level of suspicion that few of them deserve. Requiring independent verification is not an accusation against them. It is the same courtesy the law extends to every public-company accountant in America, which is to say: it is a structure that lets honest people prove they were honest, and stops the question from turning on anyone’s word.

V. What it asks of you in the meantime

None of the above will happen soon. The conclusion to this collection describes the thing that can happen now, at no cost, without waiting for any legislature: a person with a phone, standing outside a polling place at the close of voting, recording a piece of paper that is already public. That is the whole intervention. It works because the evidence it creates is held outside the custody of the institution being checked, which is the one property every case in this collection shows to be missing.

The legal architecture is what we should build. The recording is what we can do on the next election night.