Conclusion: From Evidence to Participation

A reader who has reached this point in the case studies section has sat with something that is not easy to sit with. Not any particular incident — any one of them, considered on its own, is a manageable piece of information. What is not easy is the synthesis. Across more than a century and a half, across thirteen states and eighteen stories, across every major voting-system vendor that has operated in American elections, with both major parties appearing as perpetrators and as victims, the same structural weakness surfaces again and again. A scanner counts correctly. A tape prints correctly. And then, somewhere between the tape and the published official result, something happens that no independent check was positioned to catch, and that no surviving evidence is available to resolve. The pattern is not about any single event. It is about the conditions under which those events were possible, and under which they will continue to be possible until the conditions change.

The conditions are structural. The reporting chain that runs from the scanner at a precinct to the number a reader sees on a Secretary of State’s website is not, in most American jurisdictions, independently verified. Counting is audited, in increasing numbers of states, under increasingly rigorous regimes. Ballot eligibility is adjudicated, often contentiously. Registration is checked. The step between the precinct’s count and the published total — the step that is arithmetically simplest, most tractable, and easiest to verify — is the step the existing verification regime largely skips. And for the one mechanism that would allow independent checking after the fact, access to primary-source evidence, the access regime in most states is structurally hostile. Records are denied. Denials are upheld. Tapes are destroyed while appeals are pending. Trade-secret law protects the software. The evidence needed to verify the reporting layer is, by several independent mechanisms operating at once, kept out of public hands.

This is the gap that Actual Vote is designed to close. Not by reforming the access regime, which is slow and contested work properly pursued by others. Not by auditing the vendors, which is the business of federal certification authorities and state procurement offices. Not by replacing the election officials who run American elections, who — as the methodology document takes pains to say — are, in the overwhelming majority, doing their work competently, in good faith, and often under conditions of inadequate resources and public distrust that no one should envy. The contribution Actual Vote is positioned to make is narrower, and it is immediate. It is the production, by ordinary citizens at the precinct, of an independent record of what the tabulator printed at the close of polls. That record exists outside the chain of custody that every access-to-evidence vulnerability described in the taxonomy is designed around. It is captured before any of the failures the case studies catalog can reach it. It is held redundantly, by the volunteers who produced it, and it is available for comparison against the official result by anyone who cares to make the comparison.

A word on the reader’s standing in the argument. This collection has been addressed to readers approaching it from any political direction. The argument does not depend on the reader’s party affiliation, on prior beliefs about any particular recent election, or on prior engagement with election-integrity questions. The case studies deliberately document failures that have affected Republican and Democratic candidates, rural and urban jurisdictions, red and blue states, and every decade in which American elections have been mechanically tabulated. The taxonomy’s categories apply equally well to the next election regardless of who wins it. If anything in the collection has felt like a prelude to a partisan conclusion, that is a misreading. The case is for a verification mechanism that cannot be made to care about which side it vindicates on any particular night, that produces evidence readable by any citizen of any persuasion, and that is structurally unavailable for partisan capture. That is the only kind of verification that can actually land with a public as divided as the American public currently is. It is also the kind of verification that would remain valuable in a future in which the present divisions have shifted, softened, or disappeared.

The collection’s defensive companion piece extends this point. The case for AV is not only that it would have caught the documented errors in the cases collected here. It is also that it would have provided the evidentiary foundation on which post-election disputes — including disputes in which the official record turns out to be correct — could be resolved against citizen-held primary-source evidence rather than against the absence of it. Bad-faith accusations of vote-reporting fraud have come from both directions of the political spectrum across more than a century of American elections, and they have grown most readily in the absence of independent precinct-level evidence. Antrim 2020 shows the dynamic operating against a Democratic-favoring outcome; Florida 2000, Baldwin 2002, Warren 2004, and Cook 2016 show variants operating against Republican-favoring outcomes. The defensive value of AV is structurally identical in all of these cases, just as the offensive value is. A verification method that protects only one side from one kind of harm is not a verification method. AV is the methodology this collection’s evidence demands precisely because it is the same methodology in every case.

The evidence this collection has asked the reader to consider is now in front of you. What comes next is participation. If the argument has been persuasive — not that any particular past election was wrong, but that the reporting layer in American elections is under-verified in ways that matter for the confidence a rational citizen can have in its outputs — the most direct thing you can do is help close the gap in the next election you can reach. Use the Actual Vote app at your precinct when the polls close. Photograph the poll tape where your jurisdiction permits it, and do so whether or not you think anything is likely to go wrong in your precinct, since the evidence matters most in the cases where its absence would otherwise be total. Share the app with friends, family members, and neighbors who vote in precincts where no one else is likely to be capturing evidence. Support, institutionally, the work of extending the evidence base across as many jurisdictions as possible. The value of the corpus scales with its coverage: evidence from one precinct tells us about that precinct; evidence from a thousand precincts tells us about a jurisdiction; evidence from every precinct in a state tells us, with high confidence, about the integrity of that state’s reporting layer.

None of this is heroic. None of it requires technical expertise, political commitment, or any specific belief about what may have gone wrong in any past election. It requires only the willingness to take a photograph, at a particular moment, in a place you would have been anyway, and to contribute that photograph to a public record that anyone is free to examine. The case for independent evidence, in the end, is that it exists. The way it comes to exist is through the ordinary actions of ordinary people doing an ordinary thing at an ordinary moment on an ordinary night. That, after everything else this section has shown, is the single most consequential thing the case studies are asking of the reader who has made it to their end.