Risk-limiting vote-tabulation audits: The importance of cluster size
Philip B. Stark · CHANCE · 2010
Post-election vote-tabulation audits compare hand counts of votes in a collection of groups (clusters) of paper records (“voter-verified paper audit trail,” VVPAT) to reported machine counts of the votes in the same clusters. Votetabulation audits can serve a variety of roles, including process monitoring, quality improvement, fraud deterrence, and bolstering public confidence. All of these raise statistical issues. This note focuses on audits that check whether the machine-count outcome 1 is correct. The outcome is the set of winners, not the numerical vote totals. The machine-count outcome is correct if it agrees with the outcome that a full hand count of the paper audit trail would show. Hand counts can have errors, but many jurisdictions define the correct outcome to be the outcome a hand count shows. Moreover, when the hand count of a cluster of ballots disagrees with the machine count, jurisdictions typically repeat the hand count until they are satisfied that the problem is with the machine count, not the hand count. Generally the only legally acceptable way to prove that a machine count