Why States Should Be Watching Rhode Island’s Election Audit Pilots

Rhode Island is on pace to become the second state to use risk-limiting election audits, beginning with the presidential primary in 2020, having conducted three trial runs earlier this month.

These audits—essentially a check to ensure votes are being counted correctly—limit the risk a wrong election result is certified, and Colorado performed the first in November 2017.

The Rhode Island General Assembly passed legislation earlier that same year requiring the audits by 2020. But unlike Colorado, an all-mail ballot state, the Ocean State’s audits will have more far-reaching implications.

“We’re a far more typical state where people go to polling places to vote, so we had to build our own custom tool.” John Marion, executive director of Common Cause Rhode Island, told Route Fifty. “If we can do this successfully we’ll have proven that the average state, where voters go to a polling place to vote, not an exceptional state like Colorado, can successfully audit election.”

Common Cause, a watchdog group, had been pushing risk-limiting audit legislation in Rhode Island since 2013—the state lacking any type of audit—but it wasn’t until Russian interference in the 2016 presidential election and a simultaneous miscount in the town of North Kingstown that state legislators came around.

North Kingstown used one type of ballot to program its voting machines and another on Election Day, resulting in a referendum to float a bond to replace local septic systems with sewers failing by a noticeably lopsided 8,471 votes to 5. Tabulators were reprogrammed and the outcome reversed, voters approving the ballot measure 9,481 votes to 4,526.

“Local officials quickly realized that, had it not been a blowout, an incorrect result would have been certified,” Marion said

Rhode Island can officially implement the audits in any election between now and its 2020 primary, but first the state must settle on a method.

Three different audit processes were piloted on Jan. 16 and 17 using ballots from November’s contests in Bristol, Cranston and Portsmouth.

“A pilot is important because it’s one of those things that gives election officials, themselves, the confidence they can actually do a risk-limiting audit,” said Wilfred Codrington III, counsel with the Brennan Center for Justice Democracy Program.

For Bristol precincts, a ballot-comparison audit was tested similar to Colorado’s. In that state, a sampling of ballots is reviewed, comparing voter selections to how those selections were interpreted by the tabulator.

“That’s really the most efficient type of audit,” Marion said. “It requires the least amount of ballots to be reviewed in the first round.”

The problem with such an audit is a unique ID must be placed on the ballot so it can be matched with the cast vote record, and that record can’t be randomized by the software. In Rhode Island, and other states, most polling place scanners don’t give ballots serial numbers and subsequently randomize the cast vote record to protect the voter’s identity.

Colorado’s vote-by-mail process randomizes ballots automatically, so they serialize them as they’re counted.

In Rhode Island’s case, ballots had to be rescanned a second time through a different tabulator that imprinted a serial number and didn’t use software randomizing the cast vote record. The state’s mail ballot tabulators already do both things, Marion said, but only about 10 percent of Rhode Island voters vote by mail.

A batch-comparison audit was piloted for Cranston precincts. In it, a randomly selected group of 250 to 300 ballots, possibly belonging to an entire polling place, is audited comparing total vote count to the voting machine’s count.

“It’s not as efficient as a ballot-comparison audit,” Marion said. “But it doesn’t require rescanning them.”

Lastly, a ballot-polling audit was tested for Portsmouth precincts. Those audits randomly select ballots and compare their results to the reported outcome to see if the former is representative enough of the latter.

The Massachusetts Institute of Technology also piloted four methods for randomly ballots at the same time.

No decisions have been made by the state Board of Elections as to an audit type or random selection method for ballots, and it plans to publish all the pilot data on its website first.

“That is a big decision Rhode Island will have to make,” Codrington said. “I’m not sure I have a preference, but the ballot-comparison model is great just because conceptually it’s a little easier to explain to the layperson.”

That goes a long way in gaining public trust, Codrington added.

Mark Lindeman, senior science and technology policy officer with Verified Voting; Mayuri Sridhar, an MIT graduate student; and Tom Murphy, a New York software developer working on a volunteer basis, are operating as Rhode Island’s team leads during the pilot phase. The open-source software that is being used is available on GitHub, and further development must happen once the audit process is finalized.

Most states audit a small percentage of all ballots, but that doesn’t provide confidence the entire election was sound—just that portion. If a risk-limiting audit detects a problem with the vote tally, the auditing process escalates to a full hand count.

“This is about building trust in the system, which is a foundation for a democratic form of government,” Marion said.

Rhode Island’s biggest voting machine vendor, Election Systems & Software, was on hand to observe the pilots, as was a voting equipment distributor for Clear Ballot.

The state Board of Elections anticipates conducting more pilots, Marion said, and an official risk-limiting audit could eventually be done ahead of the 2020 primary should there be a special election.

State law makes these audits for state primaries optional because Rhode Island’s are the latest in the U.S., and there are concerns they’ll occur too slowly if the state can’t get overseas and military ballots out in time, Marion said.

Rhode Island is doing its best to put in place an audit that verifies election results and restores public confidence in the process, Codrington said.

“It’s just a positive step forward for the state and the country in this environment, especially with the president talking about voter fraud, and you have clearly the Russian interference in the election,” Codrington said. “I think there’s less voter confidence in the results generally across the country.”