Explainer: Risk Limiting Audits
Risk-limiting audits are a standard part of post-election procedures in some states, and becoming increasingly popular and used across the country.
First used in elections in Colorado in 2017, the process is rooted in statistical principles which is why RLAs are quickly becoming an election-integrity best practice. In fact, they’re required in five U.S. States (Colorado, Georgia, Nevada, Rhode Island, and Virginia), and more states and counties across the United States are exploring using this cost-effective, statistics-driven audit method.
What are Risk Limiting Audits?
A risk-limiting audit isn’t a recount. It’s a method of cross-checking paper ballots with machine totals in order to confirm election results.
To conduct a risk-limit audit, auditors use a published, statistical formula to determine the number of ballots they need to sample. This formula takes into account factors like the margin of victory and the total number of votes cast.
An election with a larger margin of victory will require a smaller sample size to satisfy the risk limit than an election with a smaller margin of victory. An election with more ballots cast will require a larger sample to satisfy the risk limit than an election with fewer ballots cast.
Why Are States Adopting Risk Limiting Audits as Standard Practice?
Full recounts and audits are expensive and labor-intensive to conduct. Auditors have to haul, store, and hand-count thousands of ballots, all while respecting the chain of custody to ensure ballots aren’t tampered with. It’s much more efficient to use smaller sample sizes to analyze election results, while still maintaining a very high level of certainty.
Conventional audits require a certain percentage of ballots to be randomly selected, without consideration for how many ballots were cast or the margin of victory in races. The results, therefore, have varying degrees of certainty in the outcome. These types of conventional audits provide no certainty that the certain percentage of ballots that were randomly selected were actually representative of all ballots cast. In other words, conventional audits provide no way to know if the sample ballots that were examined were a fluke.
Risk-limited audits can be completed faster and more efficiently, while also ensuring a consistent and high level of accuracy across elections of varying sizes. By applying statistical formulas to the practice of RLAs, auditors can review a smaller set of ballots while still confirming the results of an election with confidence.
What is the “Risk” in a Risk-Limiting Audit?
The “risk limit” describes the certainty of the audit results. In most states, the risk limit is set at 10%, meaning there is a 90% chance that the samples examined provide statistical confirmation of the election outcome. There will always be some risk inherent in an audit unless auditors complete a full recount of every ballot, but by using statistical calculations, auditors can ensure certainty of results without examining every ballot.
What Happens During a Risk-Limiting Audit?
The precise methods for conducting a risk-limiting audit will vary by state, and sometimes by county. Sometimes, auditors will examine each ballot individually. In other instances, they will only examine some ballots within a batch. Some states will use physical items to generate random numbers to determine which ballots should be examined.
Regardless of minor variations, the method for conducting the audit and the work of performing the audit should be made public as a best practice, to ensure confidence in the outcome.