Election Audits Explained: What They Test and What They Do Not
A plain-language guide to risk-limiting audits, ballot-image reviews, recounts and common misconceptions.
Audit and recount are not synonyms
A recount repeats a counting process; an audit tests evidence against the reported outcome under a defined method.
Read the contest and sample
Risk-limiting audits use statistical sampling designed to provide evidence about a reported winner.
Check the paper trail
For day-to-day verification, the practical consequence is straightforward: Audits are strongest when they compare human-readable paper records with electronic totals.
Inspect discrepancy reporting
Small discrepancies can occur without changing outcomes; the relevant question is scale and cause.
Know the escalation rule
Audit methods specify when larger samples or a full count are needed.
Use current jurisdiction documents
States publish different audit procedures and reports; the local method governs the claim you are checking.
When competing summaries describe the same audit differently, open the report itself and locate the table or conclusion each side is citing. Many disputes become clearer once the exact tested population and error definition are visible.
Audit reports often recommend better documentation, procedures or controls even when the tested totals are accurate. Those recommendations are evidence about process quality, not automatically evidence of fraudulent votes. Record the finding, the sample size, the tolerance or method, and the action recommended by the auditors.
Separate findings from recommendations
Read the scope before quoting the conclusion. A county-level test cannot establish facts about every county, and a machine audit cannot answer an allegation about voter registration. Conversely, a narrow discrepancy does not demonstrate that the certified outcome is wrong unless the size and direction of the discrepancy can plausibly affect the result.
An audit is not a single procedure. A hand-count audit compares paper ballots with reported totals in selected batches or jurisdictions. A risk-limiting audit uses a statistical method to test whether the reported winner is consistent with the paper record. Equipment testing can examine tabulators without re-evaluating voter eligibility or every chain-of-custody question.
Identify what the audit was designed to test
Also note who selected the sample and whether observers could inspect the process. Transparency does not make an audit infallible, but it gives outside reviewers a way to reproduce the reasoning and challenge specific steps rather than debating the label “audit” in the abstract.
Find the reported total, audited total, discrepancy and allowable threshold. A report may describe dozens of administrative corrections while showing a tiny net vote difference. Conversely, a small sample can be statistically powerful when the audit method was designed around the contest margin. The method determines how the numbers should be interpreted.