Risk-limiting audit

A risk-limiting audit (RLA) is a method of verifying election outcomes by manually inspecting a random sample of paper ballots and comparing it with the reported results. It is designed to limit the risk of certifying a wrong outcome: the audit continues until there is strong statistical evidence the outcome is right, or until it escalates to a full hand count.

The size of the sample depends on the margin of victory. A race decided by a wide margin needs only a small sample, while a close one requires checking more ballots or all of them. RLAs depend on a paper record of every vote, such as marked paper ballots, and on organized chain-of-custody procedures. Colorado was the first state to use them statewide in 2017, and several other states have adopted or piloted them.

Election-security experts of many viewpoints favor RLAs because they provide evidence that machines counted accurately without requiring a full recount. Critics of how they are implemented point out that some states' audits are limited or do not verify all contests, and that voters may not understand the statistical approach. Compared with a recount, an audit is a verification step before election certification, and it provides a defense against claims of error.