Case 6 — Warren County, 2004
6.1 — Introduction
The 2004 presidential election came down to Ohio. George W. Bush won the state by 118,601 votes out of more than 5.6 million cast — a margin of 2.1 percent — and with Ohio’s 20 electoral votes, he won the presidency. Had John Kerry carried Ohio, he would have won the Electoral College 271–266 despite losing the national popular vote.
What followed was one of the most extensively investigated and bitterly contested election outcomes in modern American history. A congressional investigation documented dozens of irregularities. A federal lawsuit sought to preserve the evidence and determine what happened. A $1.9 million state-commissioned security review found critical vulnerabilities in every voting system used in Ohio. Advocates and journalists compiled detailed accounts of anomalies ranging from statistically improbable vote patterns to a county administration building locked down on election night under what appears to have been a fabricated terrorism pretext.
And then the evidence was destroyed. Fifty-six of Ohio’s 88 counties destroyed their election ballots and records — despite a federal law requiring 22-month retention and a federal court order specifically directing their preservation — as reported by Harvey Wasserman and others following the King Lincoln litigation. No one was prosecuted.
Two decades later, the fundamental question — whether the officially reported results accurately reflected the votes cast in Ohio’s precincts — remains unresolved and, because of the evidence destruction, unresolvable. This is not because the question is unimportant. It is because the independent evidence that could have answered it never existed in a form that survived the destruction.
That is what Actual Vote is designed to change.
This case study does not argue that the 2004 Ohio election was stolen. It does not argue that it was conducted fairly. It argues that the American public deserved — and still deserves — a rational basis for evaluating the outcome, and that the current evidentiary record does not provide one. In counties where Actual Vote volunteers could have recorded precinct-level poll tapes on election night, those records would have constituted independent evidence that existed outside official custody, could not have been destroyed by county officials, and would have either confirmed the reported results or revealed discrepancies. Either outcome would have served democracy better than the current situation, in which no one can know.
6.2 — Background: Ohio as the Deciding State
By November 2004, Ohio had become the central battleground of the presidential election. Both campaigns invested enormous resources: Bush visited the state 18 times during his presidency, Kerry 23 times during the campaign. Almost every week, more than $10 million was spent on television advertising in Ohio alone.
The state’s political geography made it a microcosm of the nation. Liberal urban centers in the north (Cleveland, Akron, Youngstown) contrasted with conservative suburbs and rural areas in the south and west. Southeastern Ohio’s Appalachian communities, historically Democratic but culturally conservative, constituted the kind of swing territory where elections are decided. Ohio’s economy had been hit harder by the early-2000s downturn than most states, losing 200,000 jobs since 2001, giving Democrats hope that economic frustration would translate into votes.
The Secretary of State’s dual role
Ohio’s elections were administered by Secretary of State J. Kenneth Blackwell, who simultaneously served as co-chair of the Bush-Cheney 2004 reelection campaign in Ohio. This dual role — the state’s chief election official actively working to elect one of the candidates whose election he was responsible for overseeing — is undisputed and documented by every source we reviewed. It does not, by itself, prove that Blackwell acted improperly in his official capacity. But it created a structural conflict of interest that would not be tolerated in most institutional contexts where impartial administration is expected, and it is the backdrop against which every subsequent decision he made must be evaluated.
Voting equipment
Ohio’s 88 counties used a patchwork of voting technologies in 2004. Sixty-eight counties still relied on punch-card systems — the same technology whose limitations had been exposed in Florida’s 2000 recount. The remaining counties used either optical scan or direct-recording electronic (DRE) touch-screen systems, supplied primarily by Diebold Election Systems (whose corporate headquarters were in Ohio) and ES&S.
This equipment landscape is relevant to Actual Vote’s applicability. Punch-card counties that used central-count systems — where ballots are transported from precincts to a central location for counting — do not produce precinct-level poll tapes at the polling place. Only precinct-count optical scan or DRE systems with printers generate the kind of precinct-level tape that AV volunteers could record. In 2004, only about 15–20 of Ohio’s 88 counties used precinct-count systems. This is an important limitation that we address directly in the AV counterfactual section below.
Poll tape posting in Ohio
Ohio Revised Code §3505.30 provides that election results are supposed to be posted at each precinct on election night before the election judges leave the polling place. However, inquiries at various Ohio Boards of Elections have found that implementation varied: some counties posted results at each voting location and others did not. Where poll tapes were produced and posted, they would have been available for AV volunteers to record.
6.3 — What Happened: A County-by-County Account
The documented irregularities in Ohio’s 2004 election span multiple counties, multiple types of equipment, and multiple layers of the voting process. We present them individually, noting for each what is documented, what is contested, and what remains unknown.
Warren County: The lockdown
On election night, Warren County — a traditionally Republican stronghold south of Dayton — locked down its administration building and barred all reporters and independent observers from monitoring the vote count. When questioned, county officials claimed they were responding to a terrorist threat that ranked “10” on a scale of 1 to 10, and that this information came from an FBI agent.
The FBI stated that it had no information about a terrorist threat in Warren County. County officials have repeatedly declined to name the agent they claim provided the warning.
Subsequent reporting revealed that the lockdown was not a spontaneous response to a threat received on election day. E-mails dated October 25 and 26 — more than a week before the election — indicate that preparations for the lockdown were already underway. The County Commissioner stated that the decision was made during a closed-door meeting on October 28.
Warren County’s voting results present their own peculiarities. In 2000, Al Gore received approximately 27.7 percent of the Warren County vote in an election where he had stopped running television commercials in Ohio weeks before election day and faced a Nader challenge that siphoned liberal votes. In 2004, John Kerry — who invested heavily in Ohio and faced no significant third-party competition — received approximately 27.5 percent: a gap of less than two-tenths of a percentage point despite dramatically different campaign dynamics. (The Conyers Report rounds both figures to “28 percent”; its own footnotes give Gore 19,142 votes of 69,078 cast and Kerry 25,399 of 92,251.) An analyst who examined precinct-level data found that Warren County showed the greatest increase in votes for Bush by precinct and the greatest increase in the Bush-Kerry margin compared to 2000. The analyst also found that a second lockdown apparently occurred during the recount, after which “very unusual new patterns” emerged.
The Conyers Report concluded (pp. 49–52, citing the Cincinnati Enquirer, Nov. 3–16, 2004): “Given the total lack of explanation by Mr. Blackwell or Warren County officials, it is not implausible to assume that someone is hiding something.” It went on: “Given that no such explanation has been forthcoming and given the statistical anomalies in the Warren County results, it is impossible to rule out the possibility that some sort of manipulation of the tallies occurred on election night in the locked down facility.”
What we know: The lockdown happened. It was pre-planned. The terrorism pretext was not supported by the FBI. Observers were excluded. The vote percentages are unusual.
What we don’t know: Whether any manipulation of vote tallies occurred during the lockdown. The lockdown prevented anyone from observing what happened, and the subsequent destruction of records prevented anyone from verifying after the fact.
Miami County: Phantom votes and a missing audit log
Miami County’s results raised immediate questions. Voter turnout was reported at 98.55 percent — a figure that multiple sources describe as “highly suspect and improbable.” With 100 percent of precincts reporting on election night, Bush had 20,807 votes (65.80%) and Kerry had 10,724 votes (33.92%), for a total of 31,620 voters. Then, inexplicably, nearly 19,000 additional votes were added to the totals, boosting Bush’s vote count from 20,807 to 33,039. Kerry’s vote percentage after the addition remained exactly the same — 33.92 percent — to three one-hundredths of a percentage point.
Roger Kearney of Rhombus Technologies, the company responsible for reporting Miami County’s vote results, stated that the problem was not with his reporting and that the additional votes were added before 100 percent of the precincts had been counted. But he did not explain why the vote count could change for Bush, but not for Kerry, after the late additions, and why the ratio remained identical to the hundredths place.
The Peckarsky Report — authored by Peter Peckarsky (lead trial counsel for plaintiffs in the Ohio election contest Moss v. Bush), Ron Baiman (expert witness for the same plaintiffs), and Robert Fitrakis (co-counsel and executive director of The Free Press) — provides the most detailed account, based on original investigation including interviews with Miami County election officials. We cite it for specific documented claims: Quillen’s on-the-record admissions and the missing audit log. Republican Director Steve Quillen confirmed that 617 extra votes were added to the certified totals by the ES&S electronic vote counting system — 357 for Bush and 225 for Kerry (net +132 for Bush) — votes that were, in Quillen’s own assessment, never lawfully cast. The audit log for the November 2004 election was completely missing: entries stopped on July 23, 2004, and did not resume until April 14, 2005. Deputy Director Miley described the missing audit log as “mysterious.”
When investigators conducted a hand recount of three Miami County precincts in 2006, the hand count totals were lower than the machine count totals in all three precincts, with a total discrepancy of 25 ballots (an average of about 8 per precinct). While this difference is small in absolute terms, the Peckarsky investigators noted that a shift of fewer than 6 votes per precinct statewide from the Bush-Cheney ticket to the Kerry-Edwards ticket would have been enough to change the outcome of the election in Ohio and the nation.
Additionally, the absentee ballot numbers did not add up. The Board of Elections certified that 5,191 absentee ballots were cast by the time polls closed, but the electronic voting system reported only 4,667 absentee ballots counted on election night. If the certified figure was correct, there should have been 5,057 absentee ballots in storage — but only 4,606 were found. Quillen had no explanation for the discrepancy.
What we know: A Republican election director confirmed that the electronic system added 617 votes that were never lawfully cast. The audit log for the election was missing. The hand recount showed machine counts exceeded hand counts. The absentee ballot numbers don’t reconcile.
What we don’t know: Whether the 19,000 late-added votes and the 617 confirmed phantom votes represent isolated equipment glitches, record-keeping errors, or deliberate manipulation. The missing audit log prevents anyone from determining the answer.
Franklin County (Gahanna): The 3,893-vote glitch
This incident was acknowledged by election officials and corrected. But the correction raises a question that the Conyers Report posed and that Blackwell refused to answer: How was it discovered? What procedures were employed to alert other counties? And can anyone be absolutely certain that this particular malfunction did not occur in other counties where it went undetected?
Mahoning County: Vote-flipping
In Youngstown, the Washington Post and other outlets reported that at least 25 electronic voting machines transferred votes for Kerry to votes for Bush. Witnesses described the experience in vivid terms. Jeanne White, a voter and manager at the Buckeye Review, an African-American newspaper, pushed the button for Kerry and watched her vote jump to the Bush column. Another voter reported trying five times before a Kerry vote registered correctly. The Election Protection Coalition confirmed these reports.
In several Mahoning County precincts, more votes were counted by machine than there were signatures in the poll books — meaning the machines recorded more votes than voters who showed up. Thomas McCabe, Deputy Director of the Mahoning County Board of Elections, stated that the machine problem “happens every election” and was “something we have to live with.”
Blackwell refused to investigate or answer questions about the Mahoning County problems.
Important caveat for AV: Vote-flipping on DRE touch-screen machines is a counting-layer problem. If the machine itself records the wrong vote, the poll tape reflects the machine’s (incorrect) count. AV, which verifies poll tapes against official results, would not detect this type of manipulation — the poll tape and the official result would match, both reflecting the flipped vote. This is an honest limitation of AV’s methodology.
Cuyahoga County: Ballots in the wrong machines
In Cuyahoga County (Cleveland), each precinct’s punch-card ballot rotates candidate positions, and each ballot must be fed into a machine calibrated for that specific precinct. At combined polling sites serving multiple precincts, poll workers advised voters to insert their ballots into any open machine — not realizing that the machines were precinct-specific. The result: ballots fed into the wrong machine had their Kerry votes recorded as votes for obscure third-party candidates.
The evidence is stark. In Precinct 4F, the results were: Kerry 290, Bush 21, Constitution Party candidate Peroutka 215. In 2000, all third-party candidates combined had received 8 votes in that precinct. In Precinct 4N: Kerry 318, Bush 11, Libertarian candidate Badnarik 163. The pattern appeared across at least 10 Cleveland precincts, all sharing combined polling sites and strong Democratic histories.
This was not necessarily fraud — it appears to have been a combination of poll-worker error and poor institutional design. But the effect was systematic: it transferred Kerry votes to third-party candidates in Democratic precincts, and no corrective action was taken.
The exit poll discrepancy
Early exit polls on election day showed Kerry leading Bush in Ohio by approximately 3 percentage points. As late as 7:30 PM, exit poll data indicated 52.1 percent for Kerry and 47.9 percent for Bush. The official results showed Bush winning 50.81 percent to 48.71 percent — a swing of approximately 6.7 percentage points from the final exit poll data to the official count.
University of Pennsylvania Professor Steven Freeman analyzed the discrepancy and concluded that it could not have been due to chance or random error, calculating a probability of approximately 1 in 1,000.
However, the National Election Pool’s own evaluation, conducted by Mitofsky International and Edison Media Research, concluded that the exit poll discrepancy was “most likely due to Kerry voters participating in the exit polls at a higher rate than Bush voters.” The report stated: “Exit polls do not support the allegations of fraud due to rigging of voting equipment. Our analysis of the difference between the vote count and the exit poll at each polling location in our sample has found no systematic differences for precincts using touch screen and optical scan voting equipment.” The Caltech/MIT Voting Technology Project independently concluded that “there is no evidence, based on exit polls, that electronic voting machines were used to steal the 2004 election for President Bush.”
Exit polls are not official counts. They have known methodological limitations, and the 2004 discrepancy, while larger than in previous elections, has a plausible non-fraud explanation in differential response rates. We present both the discrepancy and the counter-evidence and do not treat exit polls as proof of fraud.
6.4 — The Voter Suppression Context
The documented irregularities in Ohio 2004 extend well beyond vote-counting anomalies to include systematic problems with voter access — problems that disproportionately affected Democratic and minority voters.
In Franklin County, the Washington Post reported that 27 of the 30 wards with the most voting machines per registered voter showed majorities for Bush, while 6 of the 7 wards with the fewest machines showed large margins for Kerry. At Kenyon College in Knox County, 1,175 voters shared two voting machines (one of which broke down before lunch), producing wait times of up to 11 hours. A DNC study estimated that long lines caused 3 percent of Ohio’s registered voters to abstain — a figure that, if accurate, would represent more voters than Bush’s margin of victory.
Statewide, approximately 93,000 ballots were “spoiled” — meaning they registered either no vote for president or multiple votes. In predominantly black Cleveland ZIP codes, the spoiled-ballot rate was more than double the rate in predominantly white ZIP codes. Democratic precincts had 75 percent more undervotes than Republican precincts.
Secretary Blackwell’s pre-election decisions compounded these access problems. He ruled that provisional ballots cast at the wrong precinct would not be counted, a departure from Ohio’s past practice. He enforced a paper-weight requirement for voter registration forms that resulted in the rejection of thousands of new registrations. The Ohio Republican Party’s “caging” tactics — challenged and found illegal by the Third Circuit — selectively targeted 35,000 predominantly minority voters for eligibility challenges.
An honest note about AV: Voter suppression and ballot-access problems are real, documented, and consequential. But they are counting-layer and access-layer issues that Actual Vote does not address. AV verifies vote reporting — whether the numbers transmitted from precincts to central authorities accurately reflect the votes that were cast. It does not verify whether every eligible voter was able to cast a ballot, whether machines counted ballots correctly, or whether the voting process was fair and accessible. We include this context because an honest account of Ohio 2004 requires it, but we do not claim that AV would have solved these problems.
6.5 — The Investigations
The Conyers Report (January 2005)
Representative John Conyers, Jr., the ranking Democrat on the House Judiciary Committee, directed his staff to conduct an investigation into irregularities reported in Ohio. The resulting 101-page report, released January 5, 2005, documented dozens of irregularities across multiple categories: machine allocation disparities, voter suppression, registration purges, machine malfunctions, and post-election recount problems.
The report’s findings were stark: “We have found numerous, serious election irregularities in the Ohio presidential election, which resulted in a significant disenfranchisement of voters. Cumulatively, these irregularities, which affected hundreds of thousand of votes and voters in Ohio, raise grave doubts regarding whether it can be said that the Ohio electors selected on December 13, 2004, were chosen in a manner that conforms to Ohio law, let alone federal requirements and constitutional standards.”
The report led to a formal objection to Ohio’s electoral votes on January 6, 2005, supported by Senator Barbara Boxer and 31 House Democrats. The objection was defeated: 74–1 in the Senate, 267–31 in the House. No senator other than Boxer supported it. During the debate, Boxer stated she had made the motion not to challenge the outcome but “to cast the light of truth on a flawed system which must be fixed now.”
We note that the Conyers Report is a Democratic staff document and has been criticized for “exaggerated charges.” It was also praised by some as a thorough compilation of documented problems. We cite it for specific factual claims that it sources to external documents, not for its overall conclusions.
Project EVEREST (December 2007)
Ohio Secretary of State Jennifer Brunner — who succeeded Blackwell after his 2006 defeat — commissioned a $1.9 million, 10-week security review of all voting systems used in Ohio. Known as Project EVEREST (Evaluation and Validation of Election Related Equipment, Standards and Testing), the review was conducted by teams from the University of Pennsylvania, Penn State, UC Santa Barbara, Systest Labs, and MicroSolved, managed by Battelle Memorial Institute.
EVEREST found “critical security failures” in systems from all three major vendors: Diebold/Premier Election Solutions, Hart InterCivic, and ES&S. The review concluded that the voting systems used in Ohio’s 2004 election could have been manipulated with relatively simple electronic devices.
King Lincoln Bronzeville v. Blackwell
In July 2011, plaintiffs filed a document in the case purporting to be a contract between Secretary Blackwell and Republican IT consultant Michael Connell, describing a computer architecture that would have allowed real-time access to Ohio vote counts as they were reported on election night — including, the plaintiffs alleged, the ability to modify those numbers remotely.
The case was dismissed without prejudice on February 7, 2012, on Eleventh Amendment grounds (state sovereign immunity). The court found that plaintiffs had not demonstrated an evidentiary basis for the Secretary to continue storing the 2004 election ballots.
6.6 — The SmartTech Question
Michael Connell was a high-level Republican IT consultant whose company, GovTech Solutions, created a website for Secretary Blackwell that presented Ohio’s 2004 election results in real time as they were tabulated. The servers hosting this website were reportedly located in Chattanooga, Tennessee, in a facility that also housed servers for the Republican National Committee.
In July 2008, lead attorney Cliff Arnebeck had written to U.S. Attorney General Michael Mukasey seeking protection for Connell as a witness, claiming that Karl Rove had threatened Connell that if he did not agree to “take the fall” for election fraud in Ohio, his wife would be prosecuted for lobbying law violations. Arnebeck later claimed his source warned that Connell’s life was in danger.
On December 19, 2008, Connell was killed when the single-engine Piper Saratoga he was flying crashed on approach to the Akron-Canton Airport. The National Transportation Safety Board published its final report in January 2010, concluding that Connell had lost control of the aircraft due to spatial disorientation while turning in cloud. Several other pilots in the area had reported severe icing conditions; Connell’s aircraft was not equipped or approved for flight in icing conditions.
Our posture on this: We present the documented facts — the contract, the server architecture, the subpoena, the refusal to testify, the Arnebeck letter, the death, and the NTSB finding. We do not adopt the conclusion that votes were altered through the SmartTech architecture. No court has found evidence that votes were actually manipulated through this system. The case in which the allegations were made was dismissed. The NTSB attributed Connell’s death to pilot error in adverse weather conditions.
What we observe is this: the architecture described in the filed contract — a system that routed election results through servers outside official custody, hosted alongside partisan political infrastructure — represents exactly the kind of reporting-layer vulnerability that Actual Vote is designed to check. Whether that vulnerability was exploited is a question the evidence cannot answer. AV’s independent precinct-level records would have answered it.
6.7 — The Destruction of Evidence
Federal law requires the retention of election records for 22 months following a federal election (52 U.S.C. § 20701, formerly 42 U.S.C. § 1974). In September 2006, Judge Marbley ordered all 88 Ohio county Boards of Elections to preserve all ballots from the 2004 presidential election, reinforcing the federal statutory requirement.
Despite both the federal law and the federal court order, 56 of Ohio’s 88 counties destroyed their 2004 election ballots, records, or both. The reasons cited included fires, floods, storage problems, and requests to dispose of materials. As of the sources we reviewed, no one has been prosecuted for the destruction.
This destruction is the central fact of the Ohio 2004 case from the perspective of independent vote verification. Whatever happened on election night — whether the results were accurate or whether they were manipulated — the evidence that could have resolved the question was destroyed. The destruction was not accidental in any systemic sense: it occurred in the face of both a standing federal law and a specific federal court order directing preservation. The absence of prosecution for this destruction compounds the problem: it establishes that election records can be destroyed with impunity, even under court order.
For Actual Vote, this fact carries a specific and powerful implication. AV records exist outside official custody. They are created by independent volunteers, stored independently, and cannot be destroyed by the county officials, vendors, or political actors whose conduct might be in question. Had AV volunteers recorded precinct-level poll tapes on election night 2004, those records would have survived the destruction of official records by 56 counties. They would constitute an independent evidentiary baseline against which the official results could be verified — or would have been verified years ago, resolving the questions that remain open to this day.
6.8 — The Counter-Evidence and Mainstream Assessment
Intellectual honesty requires that we engage with the evidence and analysis that cuts against the “stolen election” narrative.
The National Election Pool’s evaluation of the 2004 exit polls, conducted by Mitofsky International and Edison Media Research, concluded that the discrepancy between exit polls and official results was “most likely due to Kerry voters participating in the exit polls at a higher rate than Bush voters.” The report found “no systematic differences for precincts using touch screen and optical scan voting equipment.”
The Caltech/MIT Voting Technology Project independently concluded that “there is no evidence, based on exit polls, that electronic voting machines were used to steal the 2004 election for President Bush.”
The statewide recount, completed in late December 2004, changed the result by only 176 votes — though this recount was significantly compromised by the Triad GSI “cheat sheet” episode (in which a voting computer company told counties what their hand counts should show) and by the subsequent conviction of two Cuyahoga County poll workers for pre-selecting recount precincts where they knew hand totals would match machine totals. A compromised recount is not the same as a clean recount that confirms the original result, but it is also not evidence of fraud in the original count.
The Kerry campaign did not formally challenge the election result. The congressional objection to Ohio’s electoral votes was supported by only 1 of 100 senators and 31 of 430 voting House members. Kerry later stated that “the widespread irregularities make it impossible to know for certain that the outcome reflected the will of the voters,” but he did not pursue a legal challenge.
Our position: We do not take sides on whether the 2004 Ohio election was stolen. We take the position that the evidence is insufficient to resolve the question — and that this insufficiency is itself the problem that Actual Vote exists to solve. The anomalies documented in this case study are real, confirmed by named officials and documented sources. The counter-evidence is also real and deserves engagement. What both sides of the debate share is this: neither can conclusively demonstrate its position, because the independent evidence that could have settled the matter was never created in a form that survived — and in 56 counties, was actively destroyed.
6.9 — What Actual Vote Would Have Contributed
What AV could have done
In counties where precinct-level poll tapes were produced — those using precinct-count optical scan systems or DRE machines with printed tapes — Actual Vote volunteers recording tapes on election night would have created an independent record of precinct-level vote totals.
These records would have had three critical properties.
First, they would have survived the evidence destruction. When 56 counties destroyed their election records, AV records — held independently by volunteer citizens — would have remained. The fundamental problem of Ohio 2004 is that the evidence needed to resolve the dispute was destroyed. AV makes this class of destruction irrelevant by creating evidence that exists outside official custody.
Second, they would have detected reporting-layer discrepancies on election night. If vote totals were altered between the precinct level and the officially reported results — whether through the SmartTech architecture, during data transmission, or through any other mechanism — AV’s comparison of precinct-level tape records to official results would have flagged the discrepancy immediately. In the case of Miami County’s 19,000 late-added votes, AV records from before the additions would have shown whether the precinct-level totals supported the revised numbers. In Warren County, AV records from precincts outside the locked-down administration building would have provided independent verification of the count.
Third, they would have enabled detection in hours, not years. The Miami County anomalies were not investigated until 2006 — two years after the election. The King Lincoln lawsuit was not filed until August 2006. The EVEREST security review was not completed until December 2007. Connell was not subpoenaed until September 2008. Throughout this period, evidence was deteriorating and being destroyed. AV records would have been available on election night, enabling comparison before anyone had the opportunity to alter or destroy official records.
What AV could not have done
We are committed to an honest assessment of AV’s limitations, and in this case, the limitations are significant.
AV could not have detected voter suppression. The long lines, machine shortages, ID requirements, and registration purges that disenfranchised an unknown number of Ohio voters are access-layer problems. AV verifies vote reporting, not voter access.
AV could not have detected vote-flipping on DRE machines. In Mahoning County, where electronic machines transferred Kerry votes to Bush, the problem was at the counting layer: the machine itself recorded the wrong vote. The poll tape would have reflected the machine’s (incorrect) count, and AV’s comparison of the poll tape to the official results would have shown no discrepancy — because both would reflect the flipped vote. This is the “matched poll tape caveat” discussed elsewhere in the Actual Vote manual.
AV could not have detected spoiled ballots or improperly counted punch cards. The 93,000 spoiled ballots statewide were a counting problem, not a reporting problem.
AV’s applicability was limited by Ohio’s 2004 equipment landscape. In the 68 counties that used central-count punch-card systems, ballots were transported to a central location for counting. These systems do not produce precinct-level poll tapes at the polling place. AV volunteers could not have recorded tapes that did not exist. This means AV would have been fully applicable in perhaps 15–20 of Ohio’s 88 counties — an important but incomplete coverage.
AV could not have determined whether the SmartTech architecture was used to alter votes. AV detects discrepancies between precinct-level records and official results. If precinct-level records in AV-covered counties showed no discrepancies, that would be evidence of accurate reporting in those counties — but it would not prove that other counties’ results were accurate. If discrepancies were found, that would indicate a reporting-layer problem — but it would not, by itself, identify the mechanism or the responsible parties.
The “either direction” argument
This is the case where the “either direction” framing is most important — and most non-partisan.
If AV records had matched official results in the counties where they were available, that would have constituted strong evidence that reporting was accurate in those counties. It would have given Democrats rational grounds to accept the Ohio result — or at minimum, narrowed their concerns to specific counties or specific types of problems rather than a wholesale claim of stolen election. It would have given Republicans proof that the accusations were unfounded in the verified precincts. It would have reduced the space for conspiracy theories by establishing verified facts.
If AV records had shown discrepancies, that would have constituted immediate, publicly verifiable evidence of reporting-layer problems. It would have triggered investigation on election night rather than years later. It would have preserved the evidence of the discrepancy regardless of what happened to official records. It would have given every American — regardless of party — the basis for demanding accountability.
Either outcome serves democracy. The current situation — where neither side can conclusively demonstrate its position, where the evidence was destroyed, and where the question will remain open forever — serves no one.
The evidence preservation argument
Even if AV had been deployed in only a fraction of Ohio’s precincts, and even if no discrepancies were found in those precincts, the existence of independent records would have established something valuable: a verified baseline of reporting accuracy in the covered precincts. This baseline would have served as a check against the system even where direct coverage did not exist, because it would have established that the reporting infrastructure was functioning correctly in the places where it was independently verified.
The deeper point is this: fifty-six Ohio counties were able to destroy election evidence — in violation of federal law and a federal court order — and face no consequences. This demonstrates that the current system’s evidentiary integrity depends entirely on the good faith of the officials and institutions whose conduct may be in question. AV creates evidence that exists outside this chain of custody entirely. It cannot be destroyed by county officials, by vendors, or by any actor within the system whose behavior an audit might scrutinize.
The transparency value in the most likely scenario
It is worth stating the defensive half of the transparency argument plainly in the context of a case this contested. The most likely single-scenario reconstruction of Ohio 2004 — granting every documented anomaly its full weight — is still a scenario in which the official result was correct within the 118,601-vote margin and the outcome of the presidential race turned on the votes as cast, not on reporting-layer manipulation. In a counterfactual world with broad AV coverage across Ohio’s precincts, this is the scenario that AV evidence would most likely have confirmed. The transparency benefit in that world is not that a stolen election is exposed; it is that the election is vindicated by evidence independent of Kenneth Blackwell, independent of the county boards, independent of the vendors, and independent of the federal litigation. The two decades of corrosive public doubt about Ohio 2004 — the damage done to the legitimacy of a consequential presidential outcome — are damage that was sustained because no such independent evidence existed. A sitting administration vindicated by independent citizen-held evidence is, from a transparency standpoint, just as much a democratic success as a disputed administration exposed by the same evidence. The absence of independent evidence is what permits the controversy to persist; its presence would have resolved it in whichever direction the facts actually ran.
The defensive frame across the collection
Ohio 2004 sits inside a recurring structural pattern the defensive case file takes up across the broader collection. The post-2004 left-aligned argument that Bush had stolen Ohio — advanced in book form by Mark Crispin Miller’s Fooled Again, in House report form by Representative John Conyers, in Rolling Stone by Robert F. Kennedy Jr., and across a substantial body of academic and online analysis — drew on real anomalies (the lockdown in Warren County, the records destruction in 56 counties, the exit-poll discrepancies, optical-scan miscounts in specific precincts) and on contested inferences from those anomalies to fraud. The argument was not unitary. Some of its components were rigorous; some of its components shaded into bad-faith amplification; the components occupied a spectrum that the available evidence has never been able to definitively sort. This is the pattern the Florida 2000 analysis and Baldwin County 2002 analysis describe in their own contexts: contested elections that hardened, in the absence of independent precinct-level evidence, into permanent left-aligned settled belief on one side and permanent right-aligned dismissal on the other, with the underlying evidentiary record insufficient to definitively resolve the dispute in either direction. AV’s defensive value in each of these cases is the same: the primary-source record that lets the analytical critique be tested against arithmetic rather than against the absence of arithmetic, which is what permits the entire spectrum to coexist indefinitely. Ohio 2004’s directional valence (left-aligned accusations against a Republican-favoring outcome) mirrors Antrim 2020 and Prince William 2020 running in the opposite direction. The structural problem and the structural solution are bidirectional.
6.10 — Why It Matters
Ohio 2004 matters for the case for Actual Vote at three levels.
First, it is the clearest demonstration of the evidentiary standard problem. Elections are among the highest-stakes collective decisions a society makes. The 2004 presidential election — which determined the leadership of the world’s most powerful nation during two wars — was decided by 118,601 votes in a single state where dozens of documented anomalies affected an unknown number of votes, where the chief election official was simultaneously campaigning for one of the candidates, where the evidence that could have resolved the resulting questions was destroyed by a majority of counties, and where two decades later, the question of what actually happened remains unanswered.
This is not a standard befitting the stakes. A healthy democracy requires that any reasonable person — regardless of party or which candidate they supported — should be able to look at the available evidence and rationally conclude that the reported results accurately reflect the votes that were cast. Ohio 2004 fails this test. AV would have made it possible to pass it.
Second, it demonstrates that the current system’s only safeguard is the good faith of the participants — and that this safeguard can fail. Blackwell’s dual role placed the integrity of Ohio’s election in the hands of someone with a direct stake in the outcome. The vendor whose CEO had publicly committed to “helping Ohio deliver its electoral votes to the president” supplied the voting machines. (Walden O’Dell, Diebold CEO, August 13, 2003 fundraising letter, as reported by the Cleveland Plain Dealer, the New York Times, and others.) The counties that were ordered to preserve evidence destroyed it instead. The recount was compromised by a vendor providing “cheat sheets” and by poll workers convicted of rigging the precinct selection.
None of this proves that the election result was wrong. But it demonstrates that the system offers no structural guarantee against the result being wrong — no independent check that operates outside the relationships between officials, vendors, and partisan actors. AV provides that structural guarantee.
Third, the “either direction” argument is most powerful in this case. Twenty years of controversy over Ohio 2004 have damaged public confidence in elections, fueled conspiracy theories, and provided ammunition for partisan actors on both sides. If AV records existed and showed that the official results were accurate, that controversy would have been resolved — at enormous benefit to public trust. If AV records showed that the results were inaccurate, that too would have been resolved — at enormous benefit to accountability. The only outcome that benefits no one is the one we have: permanent, irresolvable uncertainty.
6.11 — Further Reading
Wikipedia overviews
- “2004 United States presidential election in Ohio,” Wikipedia. Official vote totals, county-level results, electoral context.
- “2004 United States election voting controversies,” Wikipedia. Comprehensive overview including the NEP/Mitofsky report findings, Caltech/MIT study, congressional objection vote totals, and mainstream characterization of fraud claims.
- “King Lincoln Bronzeville Neighborhood Ass’n v. Blackwell,” Wikipedia. Federal civil rights case; Marbley preservation order; Blackwell-Connell contract filing; 2012 dismissal.
- “Michael Connell,” Wikipedia. SmartTech and GovTech Solutions; subpoena and refusal to testify; Arnebeck letter; death and NTSB finding.
- “Premier Election Solutions,” Wikipedia. Ohio-based Diebold successor; corporate chain through ES&S to Dominion.
Investigative documents (referenced in the case study)
- “Preserving Democracy: What Went Wrong in Ohio” — The Conyers Report. House Judiciary Committee Democratic Staff status report, January 5, 2005. 102 pp. We note this is a Democratic staff document; we cite it for specific factual claims sourced to external documents.
- RFK Jr., “Was the 2004 Election Stolen?” Rolling Stone, June 2006.
- Christopher Hitchens, “Ohio’s Odd Numbers,” Vanity Fair, March 2005.
- Peter Peckarsky, Ron Baiman, and Robert Fitrakis, “Official State’s Electronic Voting System Added Votes Never Cast in 2004 Presidential Election; Audit Log Missing.” Original Miami County investigation. We note that Peckarsky was lead trial counsel for plaintiffs in Moss v. Bush (the Ohio election contest), Baiman was an expert witness for the same plaintiffs, and Fitrakis was co-counsel. We cite this report for specific documented claims — Quillen’s admissions and the missing audit log — not for its overall conclusions.
EVEREST and aftermath
- “Harvey Wasserman on New Ohio Voting Report: ‘The 2004 Election Was Stolen… Finally We Have Irrefutable Confirmation’,” Democracy Now!, December 17, 2007. Wasserman’s advocacy response to the EVEREST report, including the claim that 56 of 88 counties destroyed records. Wasserman states the election was definitively stolen; readers should weigh his characterizations accordingly.
- “Project EVEREST: Evaluation and Validation of Election Related Equipment, Standards and Testing” — Academic team report delivered to Ohio Secretary of State Jennifer Brunner, December 2007. Hosted by Verified Voting. Primary source for the “critical security failures” finding.
Cross-references in this collection
- Case 7 — Butler County, Ohio, 2008 — The GEMS upload bug discovered in Ohio’s 2008 primary, which had been in the code for ~10 years (i.e., during 2004).
- Case 11 — Fraction Magic Attack — The fractional vote architecture in the same family of GEMS-derived software used in Ohio counties.
- Case 4 — Florida 2000 — The companion close-presidential-race case, also turning on a single state’s reporting infrastructure.