Op-Ed: When “Random” Sampling Overlooks Key Indicators in Election Audits
By Patrice Johnson
Chairperson, Michigan Fair Elections Institute
Picture a health inspector tasked with randomly auditing local restaurants but somehow manages to overlook the city’s McDonald’s and Burger King—both notorious for their myriad health code violations. Now, envision that this inspector is appraising his own work before declaring every restaurant a model of health and safety. Would you trust the validity of that report?
Yet, this is eerily reminiscent of the recent post-election audit conducted in Michigan. In October 2025—ten months post-election—Secretary of State Jocelyn Benson’s office proudly announced the state achieved a remarkable 99.97% accuracy in its ballot counting. A commendable figure, no doubt. However, the so-called “random” sample deliberately avoided Detroit and Troy, Michigan’s largest cities that have historically faced election administration issues.
When the so-called randomness conveniently skips critical areas, trust in the process diminishes.
This audit was not only delayed—taking an astonishing ten months to complete and publish, as pointed out by election security experts Susan Greenhalgh and Dr. David Jefferson, who termed it “inadequate” and “lacking transparency” in their detailed analysis. Comparatively, the 2020 audit wrapped up in merely five months. What accounted for the delay in the 2024 audit?
Timeliness is crucial.
Risk-limiting audits are designed to provide prompt statistical assurance immediately following elections, during peak public interest. Therefore, publishing results ten months after the fact, when media attention has waned and public memory has shifted, severely undermines this principle. In a swing state already grappling with contentious election administration, such delays are likely to exacerbate, rather than ease, skepticism.
However, timing isn’t the sole issue. While the audit confirmed the accuracy of ballot counting, it failed to address whether those ballots originated from eligible voters at legitimate addresses, whether signatures on the over two million mail-in ballots aligned with registration records, or whether Michigan’s voter databases adhered to federal maintenance requirements.
Consider this analogy: A referee who correctly counts touchdowns but neglects to check whether players are within bounds or even qualified to play has not validated the outcome of the game in any meaningful sense. Michigan’s audit may have confirmed counting accuracy, but it failed to inspect the broader context.
The Michigan Fair Elections Institute identified the significance of examining that broader context. Through a rigorous statistical analysis of 384 randomly selected registrations from the state’s 542,121 voters inactive since 2019, we uncovered worrisome patterns in our Ensuring Electoral Integrity report.
Our findings demonstrated that 27.6% of the sampled registrations contained discrepancies, projecting issues that included deceased voters (roughly 21,141 statewide), out-of-state dual registrations (around 53,702), and invalid addresses (about 4,225).
Almost three-quarters of these registrants either had never voted or last participated in the electoral process prior to 2010, amounting to a staggering total of 398,121 inactive registrations throughout the state. With Michigan housing 558,627 voters inactive since 2019 or earlier, it far exceeds the four-year maximum set forth by the National Voter Registration Act by 375%.
To further investigate, we mailed confirmation letters for residency verification; 9.11% of those letters were returned undeliverable, with Detroit presenting an alarming 57.14% indeterminate residency rate among these long-inactive registrations. Yet the state’s so-called “random” audit managed to completely overlook these significant indicators.
Our extensive MFEI Investigation into Michigan Elections also documented substantial alterations made to voter history records months after certification, flagged 82,467 voter IDs that were temporarily marked as duplicates in October 2024 due to a “glitch,” and highlighted the premature destruction of electronic pollbook data, contradicting federal mandates requiring 22-month retention. Michigan appears to harbor approximately 800,000 potentially ineligible registrations, significantly surpassing federal guidelines.
Moreover, the audit failed to consider equipment security vulnerabilities. Under Michigan law, voting systems mandates certification by an independent authority accredited by the National Association of State Election Directors and the State Canvassers Board, per MCL 168.795a. However, Michigan’s own certification process displays a lack of independent validation. Research conducted by University of Michigan computer scientist Professor Alex Halderman pinpointed four alarming vulnerabilities in the state’s voting equipment, including malware risks capable of covertly altering vote totals. Despite these documented threats, the proprietary source code remains concealed from public review; nevertheless, the audit asserted machine reliability without addressing the integrity of the certification process.
Michigan deserves better standards.
The Supreme Court’s ruling in 2018 Husted v. A. Philip Randolph Institute reaffirms that states must proactively remove long-inactive registrations. It is imperative for Michigan to rectify its records concerning the over 558,000 voters inactive since 2019 or earlier.
Furthermore, we must uphold the 22-month federal record retention requirement rather than hastily destroying pertinent evidence. We must also permit independent computer scientists to thoroughly review voting equipment source codes as federal cybersecurity experts advise.
Looking ahead, future audits should employ genuinely independent methodologies, directing bipartisan observers—not the Secretary of State’s office—to select precincts via transparent randomization that explicitly includes higher-risk areas such as Detroit. No more “random” samples that conveniently bypass the crucial jurisdictions.
Michigan requires thorough verification of mail-in ballot signatures along with the application of the same statistical rigor, which asserted 99.97% accuracy in ballot counting, to assessing voter registration databases, potentially through independent federal review under the Help America Vote Act and National Voter Registration Act.
While Michigan’s achievements in accurately counting ballots are commendable, when the health inspector’s “random” restaurant inspection conveniently bypasses the establishments with the most severe violations, when the referee tallies touchdowns yet neglects to verify boundaries or player eligibility, and when election audits take ten months yet fail to account for Detroit, the public is right to question whether these processes are genuine transparency exercises or mere theatrics.
True election integrity hinges on assessing all crucial elements: accurate counts, legitimate voters, validated signatures, well-maintained databases, secure equipment, independent oversight, and timely results. Michigan has excelled in one aspect, yet the others await our robust commitment to true transparency that honors every voter.
Patrice Johnson chairs the Michigan Fair Elections Institute, a 501(c)(3) nonprofit dedicated to election integrity. Reports available at mifairelections.org/library.
