ballotpolling

Ballot polling is a method used in post-election auditing to verify the reported outcome by examining a random sample of ballots rather than the entire ballot set. It is a form of risk-limiting audit (RLA) in which a statistically valid sample is drawn from the paper ballots and tallies for the leading candidates within the sample are compared to the officially reported outcome. If the sample results are consistent with the reported result, the audit reduces the probability that the outcome is wrong; if they are not, additional ballots are inspected and, if necessary, a full hand count is conducted.

In practice, ballot polling relies on a pre-defined risk limit and a sampling plan. Ballots are selected at random, manually reviewed, and classified according to how they would have voted in the contest. The observed sample outcome is then used to infer whether the overall result would likely be the same if all ballots were counted. The process typically requires a trustworthy paper ballot trail and a secure chain of custody for the ballots.

Ballot polling differs from ballot-level comparison audits, which compare each ballot’s markings to its electronic record. Ballot polling does not require direct one-to-one comparison with the machine tally in the initial sample, but it does require randomization and statistical reasoning to justify conclusions. It is one of several RLAs used to provide post-election confidence while limiting the scope of manual counting.