Every day, our journalism dismantles barriers and shines a light on the critical overlooked and under-reported issues important to all North Carolinians.
Before you go …
Thanks for reading. If you like what you are reading and believe in independent, nonpartisan journalism like ours—journalism the way it should be—please contribute to keep us going. Reporting like this isn’t free to produce and we cannot do this alone. Thank you!
This fall, North Carolina will pilot a new kind of postelection audit, the gold-standard method to ensure the candidate declared the winner in a race actually received the most votes. The action is the first step in a likely yearslong process of improving the state’s postelection audit strategies.
Currently, the state uses a “sample audit,” whereby election officials hand-recount two random precincts to make sure the results are right.
For most elections, North Carolina’s sample audits count far more ballots than is necessary to be confident that the election results are accurate, creating a significant and unnecessary burden on election officials. For very close elections, the state’s current sample audit may recount too few ballots to be highly confident in checking the results.
Risk-limiting audits were designed to right-size this problem, what N.C. State Board of Elections Chair Damon Circosta referred to as an “optimization” of the system.
A risk-limiting audit randomly samples ballots from across voting methods. Election officials hand-count the sample and then use an equation to see how likely it is that the paper ballots show a different outcome than the computer-counted results. If the ballots show a potentially different outcome, a bigger sample is pulled. The process is repeated using progressively larger samples.
If it looks as if the paper ballots aren’t backing the electronic outcome, an entire recount occurs.
Bolstering a functioning system
Already, the state runs five kinds of postelection audits before certifying any election and provides extensive data on its voting systems. The goal is to increase voter confidence in election results and improve the efficiency of running elections, Circosta said.
“Let’s just be clear: We are as confident as anybody can be anywhere in the world our elections are secure,” Circosta said.
Even so, the state is “committed to being on the cutting edge” of election audits by creating “another layer of verification that we hope will lead to greater overall confidence in elections,” according to Pat Gannon, spokesperson for the state board.
North Carolina joins a trend sweeping the nation. Since 2016, Colorado, Rhode Island and Virginia passed laws requiring these audits, with another nine states either piloting the audits or having them as an option, according to the National Conference of State Legislatures.
Foreign attacks on U.S. election systems, which came to light in 2016, and domestic disinformation campaigns drive the proliferation of these audits, according to Mark Lindeman, co-director of Verified Voting, a nonprofit focused on improving how election administrators manage the casting and auditing of ballots.
Circosta started pushing for risk-limiting audits when he was appointed to the state board in 2019. The state board first planned to pilot these audits after the March 2020 primary, with a self-admittedly ambitious goal of having statewide risk-limiting audits for the general election; but the pandemic swept away those aspirations.
Now, the state is starting small with the municipal elections, which Lindeman calls a “brilliant move.” This lets election administrators count fewer ballots to complete the audit and to learn a new system without the stress of statewide or nationwide elections.
The state board did not commit to running the audits in the 2022 elections, instead saying it will wait to see the results of the fall pilots and will consider other auditing methods.
But if it goes well, Circosta said, it could eventually replace the state’s current system.
The shift would require a change to state law, a request the state board has not yet made. The National Academies of Sciences, Engineering and Medicine suggested in a 2018 report that all states mandate risk-limiting audits within 10 years.
In the long term, moving to risk-limiting audits could save election officials some work, but at the beginning, it is another item on their already full to-do lists.
For Circosta, the effort will be worthwhile if the audits can increase voter confidence in the election results.
“The whole point of any of these postelection procedures is so that people can see them and understand them and see how we secured that vote,” he said.
‘Eliminate the need to trust other people’
Democracies function on the will of the governed, which means that enough groups distrusting election results could result in revolts, violence and attempts to overthrow elected governments, as the Jan. 6 insurrectionists showed.
Voter confidence is essential to a functioning democracy, and a risk-limiting audit builds confidence by starting with the assumption that the results are not accurate, a stronger way to review the results than assuming they are correct at the outset.
Philip Stark, a statistics professor at the University of California, Berkeley, and a federal Election Assistance Commission board member, designed risk-limiting audits with voters in mind. The math it took to create the audits is a bit complicated, but the math to follow along is quite simple, he said.
“We’re trying to eliminate the need to trust other people in order to justify trust in the election outcome to the extent that that’s possible,” Stark said.
The audit is a process that requires transparency, he said. Just as state and county election officials do for the current sample audit process for the state, the public should have access to observe each step. The process includes rolling 10-sided dice to generate a random number that helps pick random ballots to sample, pulling and counting those ballots, and publicly presenting all the results.
The process is designed to counter what Lindeman called an ongoing attack on verifiable facts.
Following the 2016 election, former President Donald Trump falsely claimed that undocumented immigrants cast millions of ballots. Since 2020, he or his representatives have claimed Italian military satellites changed election results, a voting machine manufacturer manipulated results, mail-in ballots were illegally processed, and other claims dismissed by courts as untrue or frivolous.
There’s no way to stop these kinds of claims, Lindeman said. But there is a way to show how the election was run and a way to publicly test if the results can be trusted.
“The basic principle is that seeing is believing,” Lindeman said. “That gives us a chance of cutting through the fog and disinformation.”
Risk-limiting audits are the final steps in a well-run election, not holy water that will somehow bless the results after a series of administrative mishaps, according to Stark.
If there’s garbage in, there’s going to be garbage out, no matter what the numbers say, he said.
In the 2018 election in North Carolina’s 9th Congressional District, Republican candidate Mark Harris paid McCrae Dowless to manage the absentee ballot portion of the campaign, according to a state board investigation. Dowless illegally collected, altered or destroyed ballots, according to the state board order.
Ultimately, the State Board of Elections noticed the irregularities and ordered a new election. A risk-limiting audit would not have caught that kind of fraud in an election. It is only meant to make sure that the machines that counted the votes did not give the victory to the wrong candidate.
A risk-limiting audit could help defeat false claims that a voting machine vendor changed results by showing that the real, paper ballots backed the winner of the election, or it could reveal an otherwise invisible tabulation error that would hand the election to the wrong candidate.
The very process of conducting and preparing for a risk-limiting audit can help election officials run a smoother and better-documented election.
“The jurisdictions that I’ve worked with … generally report that they understand their own election processes better once they start to gear up to do risk-limiting audits because you do need to keep track of things,” Stark said.
That includes making a ballot manifest, which shows where and how ballots are stored. The information is readily available in the election information that each county manages, but it is not pulled into a single document.
VotingWorks, a nonprofit working to build a “publicly accountable voting system,” created an open-source tool called Arlo to help states manage their audits. The tool will help counties create ballot manifests and with the mathematically complicated part of running a risk-limiting audit, which is knowing how many ballots to sample and of which ballot types — by-mail, early voting, provisional, overseas or Election Day.
North Carolina will be the first state to use Arlo on its own rather than paying VotingWorks to run the audit. Virginia Vander Roest, VotingWorks’ election implementation manager, said the state’s election staff is more than capable of managing the audit.
Election officials will have to randomly select ballots from those batches, mark where they pulled the ballots from, count them and show the results. The process will be tracked in Arlo, which produces an audit report that Gannon said will be made publicly available.
Counties testing the waters
The state board does not know which municipal elections will pilot risk-limiting audits in October, and it has a contingency plan in place should it need more pilots in the November elections.
The election directors of both Buncombe and Union counties volunteered to run the pilots for their November elections.
“If the state board is going in this direction, we want to know what it’s about,” Corinne Duncan, Buncombe County’s election director, said.
The state board is pushing for risk-limiting audits, Duncan said, and if it mandates them, the counties will have to follow suit.
The state board is using existing staff and resources to create the new auditing program. County election officials will have to do the same.
“It is extra work. We can do it. Our counties will fund us,” Duncan said. “I also want to say that we always say that. We always do more, and it’s just we keep being asked to do more and more and more and more and more.”
Duncan is motivated to pull off risk-limiting audits for the same reason Circosta began pushing them. In this divisive political environment, bolstering voter confidence is now a key part of running the election.
“We know that our results are good,” Duncan said. “We want to do what we can to show the rest of the world that’s true.”