BRAVO: Ballot-polling Risk-limiting Audits to Verify Outcomes

Risk-limiting post-election audits guarantee a high probability of correcting incorrect electoral outcomes, regardless of why the outcomes are incorrect. Two types of risk-limiting post-election vote tabulation audits are comparison audits and ballot-polling audits. Comparison audits check some of the subtotals reported by the vote tabulation system, by hand-counting votes on the corresponding ballots. Ballot-polling audits select ballots at random and interpret those ballots by hand until there is strong evidence that the outcome is right, or until all the votes have been counted by hand: They directly assess whether the outcome is correct, rather than assessing whether the tabulation was accurate. Comparison audits have advantages, but make large demands on the vote tabulation system. Ballot-polling audits make no such demands. For small margins, they can require large samples, but the total burden may still be modest for large contests, such as county-wide or state-wide races. This paper describes BRAVO, a flexible protocol for risk-limiting ballot-polling audits. Among 255 state presidential contests between 1992 and 2008, the median expected sample size to confirm the plurality winner in each state using BRAVO was 307 ballots (per state). Ballot-polling audits can improve election integrity immediately at nominal incremental cost to election administration.