ASA Releases Statement on Risk-Limiting Audits
The ASA Board of Directors recently released a statement recommending that routine risk-limiting audits be conducted and reported in all federal elections, as well as in statewide and other governmental election contests. The statement also urges election officials to seek statistical advice about how to efficiently attain the desired accuracy.
“Risk-limiting audits are designed to limit the risk of certifying incorrect winners,” said Sastry Pantula, ASA president. “A well-designed audit often can confirm a correct electoral outcome after examining only a small fraction of the ballots cast. If, however, an outcome is unclear or incorrect, a risk-limiting audit protects against prematurely accepting it. We need laws that call for such audits—laws with teeth.”
“Most current laws mandating post-election audits call for looking at a fixed percentage of ballots,” continued Pantula. “They may select many more ballots than are needed to do the job or, sometimes, not enough to provide a satisfactory check on the outcome. Good statistical design can efficiently and economically guarantee a large chance of finding and correcting erroneous electoral outcomes. Statisticians can help design risk-limiting audits that make better use of election officials’ time and taxpayers’ money.”
In a risk-limiting audit, ballots are sampled for review in batches (a batch is a group of ballots for which the machine total has been separately tallied). The total number of ballots examined to confirm a correct outcome generally increases with 1) a smaller margin of victory, 2) larger-sized batches of ballots, and 3) more errors in the machine count. Thus, states should seek to buy electoral machinery that makes small-batch reporting easy and minimizes errors in recording voter intent. The audits, themselves, can help identify the kinds of machines and ballot designs that lead to the fewest errors.