More than half the states require elections officials to conduct postelection audits, which are intended to ensure that votes were accurately tallied and that equipment functioned properly. The process involves recounting a set of ballots and comparing the results to the original tabulation. Although these audits can help catch errors, they do not always use the most efficient statistical analytic methods. So several jurisdictions have recently begun to use more efficient risk-limiting audits, which are designed to confirm the outcome of an election while minimizing the number of ballots audited and thus the time required to conduct the audit.
The number of ballots recounted in a risk-limiting audit is determined by the margin of victory in the contest being analyzed, and the process stops once auditors collect adequate statistical evidence to confirm the winner. In a race with a large margin of victory, a risk-limiting audit may require that only a relatively small number of ballots be recounted. But in close contests, the process uses a larger sample of ballots and could, in rare instances, lead to a full recount. In more traditional postelection audits, the lack of methodological flexibility can cause more ballots to be recounted than is necessary.
Cuyahoga County, Ohio, conducted a risk-limiting audit of last year’s gubernatorial race. Incumbent John Kasich received 51 percent of the votes cast in the county, and challenger Edward FitzGerald received 45 percent. The county Board of Elections needed to recount slightly more than 8,000 ballots before it could confidently determine that Governor Kasich had correctly been declared the winner. The board also audited the race for state treasurer, in which incumbent Joshua Mandel received 39 percent of the vote versus 61 percent for challenger Connie Pillich. In this less competitive contest, fewer than 2,500 ballots were needed to certify Pillich’s victory among county voters.
Other jurisdictions have been experimenting with the risk-limiting audit model. The California secretary of state recently completed a three-year pilot program that audited contests of varying size in counties throughout the state. Last year, Arapahoe County, Colorado, conducted a risk-limiting audit of two primary contests and is doing further testing ahead of implementation of a state law requiring all counties to use risk-limiting audits by 2017.