Virginia Code calls for a statewide audit every five years.
RICHMOND – Virginia’s election is going through an audit. That part we know, as the state’s Department of Elections made the announcement last week. But what does that mean? .
The audit happens once every five years under Virginia Code. The goal here is to basically walk through the process for every city, county and region, outlining any areas that need to be improved.
Risk Limiting Audits
Auditors are performing a risk-limiting audit (RLA) on the Commonwealth’s November election. These audits are a type of post-election tabulation that examines a random sample of paper ballots for evidence that the reported winner actually won.
This type of audit is possible because Virginia only uses paper ballots. On Election Day, electronic voting machines tabulate votes. An RLA audit bypasses voting machines to avoid any discrepancies in vote counts caused by machine error.
“That’s why an RLA is so important in this moment. You can trust an RLA whether you trust the tabulators or not. Because an RLA is completely independent of the tabulators. An RLA answers the question ‘would this person have won if we hadn’t used machines to count the votes at all?’ It’s a quality assurance check on the overall result of the election, not any one ballot,” said Monica Childers, product manager at VotingWorks.
VotingWorks is a non-partisan nonprofit that produces voting machines and performs RLAs. They are facilitating Virginia’s ongoing RLA, and have overseen statewide audits in Georgia, Pennsylvania, Rhode Island, and Michigan.
Audit Begins With A Roll of the Dice
It all starts with the roll of dice.
“When the audit begins, twenty dice are rolled in a public ceremony of sorts to create a long random number. This number, or ‘random seed,’ is used to seed the algorithm in the audit tool that randomly selects ballots for audit from across the state,” said Childers. “Since which particular ballots are audited depends on that number, generating it randomly, in public, at the beginning of the audit ensures that no one can ‘game’ the audit by knowing which ballots we will look at ahead of time.”
Based on that number, the audit tool provides each jurisdiction with a list of the ballots it selects for auditing. How many ballots they request for the audit depends on how close the election was in each locality. It also depends on what proportion of the Commonwealth’s overall vote count that jurisdiction makes up.
Ballots, once chosen for the RLA, are collected in-person.
“Audit boards of at least two people then retrieve those particular paper ballots from storage, examine each ballot by hand, and record the votes they see on those ballots, then report those votes back into the audit tool,” said Childers.
Next, the audit tool examines the audited votes from across the Commonwealth to confirm the winner of the election. Several rounds may be necessary to certify the results, but that’s not out of the ordinary, according to Childers.
“A few rounds of auditing isn’t necessarily indicative of a problem with the original election results. Because the audited ballots are a random sample, sometimes the random sample just isn’t big enough to give an accurate representation on the first round. An audit expanding to a full hand count, on the other hand, is very rare,” Childers said.
Why Bypass Voting Machines?
Though they’re choosing to perform an RLA, that doesn’t mean the public can’t trust overallresults generated by machines. According to representatives of the Virginia Department of Elections, most discrepancies are due to voting machines. However, they don’t happen often.
“The main cause for a discrepancy in elections would be an error in a voting machine. However, all voting machines are certified to federal and state standards and are required to undergo rigorous testing prior to deployment in an election.,” said Andrea Gaines, Director of Community Relations & Compliance Support at the Virginia Department of Elections. “Virginia uses only paper ballots which allows us to conduct these audits to confirm that the machines counted the ballots properly.”
Meg Schiffres is Dogwood’s associate editor. You can reach her at firstname.lastname@example.org.