Across the world, election officials continue searching for tools that strengthen trust in democratic outcomes while keeping election administration efficient and transparent. One way to check vote integrity now gaining adoption is the Risk-Limiting Audit (RLA) which brings mathematical rigor and adaptability to post-election auditing.
An RLA is a scientifically designed audit that provides strong evidence that the reported winner truly won — and expands to a full hand count if more confirmation is needed. What makes RLAs distinctive is their adaptability. Instead of reviewing a fixed number of ballots, they adjust based on the competitiveness of a race. Contests with wide margins typically require only a small review, while closer races naturally invite more examination. This flexibility helps jurisdictions use resources wisely while still providing robust verification.
RLAs operate with an explicit risk limit, a mathematically defined maximum probability of confirming an incorrect outcome. The audit expands only when needed. If the preliminary result is well supported by the sampled ballots, the review can conclude quickly. If not, the process continues until the outcome is confirmed with high statistical confidence or until a full hand count determines the true winner. RLAs can be conducted in different ways depending on local systems: some are ballot-polling audits that sample and hand-interpret ballots directly, while others are ballot-comparison audits that compare sampled paper ballots to their corresponding digital records, often requiring fewer ballots when technology supports it. In either case, the goal is the same: validate the outcome efficiently and reliably.
A key feature of every RLA is transparent randomness. Ballots are selected through a publicly verifiable random process (often using dice rolls or other observable methods) so that no one can predict or influence which ballots will be checked. This open, random sampling is central to why the results are statistically trustworthy and understandable to the public.
The first practical pilots took place in the early 2010s, when Merced County, California (2011) and Arapahoe County, Colorado (2013) tested early versions of RLA procedures. After years of refining the method, Colorado became the first state to conduct statewide RLAs in 2017, carrying out ballot-comparison audits across its counties. That milestone showed that RLAs can be successfully implemented at scale.
Outside the United States, Denmark has also piloted RLAs in real election contexts—including a pilot on Denmark’s May 2014 European Parliament election results and a national referendum held on December 3, 2015. Mongolia has likewise considered implementing RLAs, though there are no documented completed, binding RLAs of real elections there so far.
At a time when public trust is paramount, RLAs provide an evidence-based, voter-centered approach that enhances confidence in election outcomes while complementing the broader ecosystem of post-election verification tools. Importantly, RLAs are designed to confirm whether the reported outcome is correct, not to investigate the cause of any discrepancies. They can uncover errors or, in rare cases, signs of misconduct—but their unique promise is outcome verification: if the outcome is wrong, the audit is very likely to catch and correct it.