This article was produced by Voting Booth, a project of the Independent Media Institute.

As early voting began in Nevada’s 2020 Democratic presidential caucus, thousands of people had to wait for two hours or more before voting. The bottleneck was due to a shortage of pre-programmed iPads that the Nevada State Democratic Party gave volunteers to check in voters.

“I have been in line for one hour and 58 minutes. This is normal from what I can tell from this line today,” said a researcher at the University of Nevada, Reno, who didn’t want to give her name after voting on campus on February 15, the first day of the four-day early voting period. “This is my second voting location today. The line at the first one was even longer.”

“We’re excited. They’re all standing in line. They’re all very nice,” said Sarah Mahler, the UNR voting site lead and Washoe County Democratic Party chair.

“It’s never been done before in Nevada. It’s never been done before in our nation. It’s the first time we’ve utilized early caucus voting.”

Despite Mahler’s enthusiasm, there were signs that delays or problems counting early votes could shadow 2020’s third Democratic presidential nominating contest. These complications would unfold behind closed doors at vote-counting hubs run by the Nevada State Democratic Party (NSDP). The possible problems concern the system used to scan and count tens of thousands of paper early ballots, as well as the database tracking all of the voters and their votes.

Nearly 75,000 people voted early, the NSDP said. That process continues on Saturday, February 22, with precinct caucuses across the state.

Top party officials have not responded to numerous requests to comment about the last-minute voter-tracking and vote-counting system that it will be using after it jettisoned the same reporting and counting technology that failed in Iowa’s Democratic Party presidential caucuses on February 3.

“Nevada Dems remain committed to executing the most accessible, expansive and transparent caucus yet,” Alana Mounce, NSDP executive director, said in a “not for distribution” memo on February 14 that was a “ballot processing update.” It said party officials would scan the early voting ballots in two locations where few observers would be allowed to watch. Those observers would not include “the public or the press.

Leaders of the state party, not government election officials, are running the caucuses. That means that they had to create the entire voting system, from voter registration to the counting of votes. They have done that by hiring many private contractors and consulting with the Democratic National Committee’s technology staff to vet cybersecurity precautions. But other potential issues besides cybersecurity received less scrutiny.

“Running elections is hard. It is not for the faint of heart or rank amateurs. And it takes a lot of logistics to do it,” said Larry Moore, senior vice president of Voatz (a firm developing a smartphone voting system) and former CEO of Clear Ballot, which is the nation’s most comprehensive vote auditing system.

At an early voting site on Reno’s north side visited by Voting Booth (and elsewhere, according to media reports from Las Vegas), weary volunteers were seen skipping a key check-in step to alleviate the long wait to vote. (Voting site coordinators said the party underestimated the turnout and did not have more iPads to use to help check in voters.) Thus, the response at some early voting sites was to skip a check-in step that electronically tied voters to their filled-out ballots. That data is at the heart of the party’s digital vote-counting system.

After waiting in line, voters would first show an ID to check in with a volunteer. That party worker would check if that person was a registered voter by looking them up on a PDF document on a party-owned iPad. (If voters weren’t registered, they would be asked to fill out a state registration form and join the Democratic Party.) But there was supposed to be a second step before the voter was given a paper ballot. That step involved entering the voter’s identifying information, including a PIN number matching a sticker that the volunteers put on their paper ballot, into a Google form on another party-owned iPad.

Harried volunteers were seen skipping the Google form step to speed up the lines. That omission meant that the data that the NSDP planned to use to pair those voters to their ballots and choices ranking the candidates (a caucus is not a secret ballot and involves two rounds of ranked-choice voting) would be incomplete, unless other party workers later retraced and filled in those gaps.

The NSDP spokeswoman and its caucus director did not respond when asked to comment on this issue, which could complicate tallying the early voting results. But Mahler raised her eyebrows when told of this skipped step. At the early voting site that she was running at the university in Reno, volunteers patiently made sure that all voters verified their information and PIN numbers that were entered into the Google form.

The NSDP is creating two records of the early voting: a digital and a paper record. It plans to use only its digital system to count votes, at least in the first instance, according to the “ballot processing” memo to staff on February 14.

While national media have warned about “chaos” that could affect the voting and results after statewide caucuses on February 22—echoing Iowa’s meltdown when its digital counting systems failed—issues that might arise earlier in the process are being overlooked. The skipping of the Google form step (and thus omissions in the underlying data used to generate results) was only one of the signs that the system being used during early voting could face delays or produce inaccurate results.

Caucus voting is not the same as voting in a primary where there is only one round of voting and the candidate with the most votes wins. The caucus is not as simple. Early voters are asked to rank three to five presidential candidates on a paper ballot by filling in ovals in rows ranking their choices next to a candidate’s name. Their first choice who gets 15 percent or more of the vote in their precinct is counted.

The “not for distribution memo” from NSDP executive director Mounce makes it clear that the party will scan the early paper ballots to analyze the ranked choices and tabulate the results. (Mahler, the Washoe County Democratic chair, reiterated the party plans to import those local early voting results onto the iPads that all of the NSDP precinct chairs will use in the February 22 statewide caucuses.)

But the party has released no details on that scanning operation, and the little that’s known about it worries seasoned election experts. Some career election officials who since their retirement from government have developed software for counting ranked-choice ballots (one version of which will be used to tally the early voting ballots, while another ranking system will be used in precinct caucuses) said that the NSDP is heading into problematic waters as it tallies early votes.

These officials and computer scientists who study voting all noted that the paper ballot used for the NSDP’s early voting did not have alignment or “timing marks” on its perimeter, which is how scanners in government elections are calibrated and tested to correctly read the ballots’ ink-filled ovals. (These marks help computers create a grid that then links ink-marked ovals to choices on the ballot—no matter what end of the paper is put into the scanner.) Also, the NSDP ballot is the size of half of a sheet of typing paper with fairly small ovals to be marked by pen, which additionally means there’s less leeway for the scanning software to read ovals.

“I would think it could be hard to scan ballots with no timing marks,” said Duncan Buell, chair of the computer science and engineering department at the University of South Carolina, Columbia, and a county election official. “This ballot looks to have rather small places for marking. I can think that timing marks might be unnecessary if the bubbles were an inch big, but as we all know from scanning things on a photocopier, a skew of [a] quarter-inch or so can be routine.”

Reflecting on the absence of timing marks and small ballot design, Buell said, “I am surprised that Nevada would not have had that done by people who have a lot of experience in scanning.”

Other longtime election officials who have retired from government posts and are now working on ranked-choice ballot technology were worried that the NSDP was not using any system produced and certified by election professionals. They were worried that if Nevada experienced scanning problems—meaning it would have to revert to counting tens of thousands of early ballots by hand—that their work to nationally advance ranked-choice voting would be undermined.

“It would be very scary if they have come up with a home-made product [scanner and software] that they think will work,” said Gary Bartlett, who was North Carolina’s election director for two decades and now directs the Ranked Choice Voting (RCV) Resource Center.

“It looks like a paper ballot that you hand-count,” said George Gilbert, who works with Bartlett, after seeing a photograph of the NSDP’s early ballot. “It would take a while” to do that should hand-counting thousands of ballots be necessary, he said.

Gilbert has developed an open-source software tool to use for counting in ranked-choice elections. But neither he nor Bartlett was aware if the NSDP was using that open-source software, he said, adding that Nevada party officials were not interested in working with the RCV Resource Center.

Nor is the state party working with other groups like FairVote, which also advises election officials on ranked-choice voting, according to its president, Rob Richie. FairVote provided memos to the NSDP, he said, but it also was not asked to help. “They’ve been pretty closed-lipped. They’re not consulting folks like us.”

Richie said that Bartlett and Gilbert were among the nation’s top technical experts on ranked-choice systems. Gilbert said that his counting software would work if all of the ranked votes were accurately entered into an underlying spreadsheet, which, presumably, is the role that the Google form would play for the Nevada party.

But that scenario presumes that the party’s data set was complete (an unknown number of volunteers skipped entering voter information and ballot PIN numbers into the Google form), and that all of the paper ballots were accurately scanned and successfully imported to the underlying Google form spreadsheet.

“I have no idea how accurately they can do that or how rapidly they can do it,” said Gilbert. “That’s the question that has to be asked: How will this ballot be scanned?”

Bartlett added that there were many things that could go “wrong” with scanners.

“There are so many small things that can go wrong with that type of voting system when you fill in an oval,” he said. “Sometimes you have somebody with dexterity problems who cannot mark the target area dark enough for the scanner to read. Other times you see weird things where you [as a human observer] can read a voter’s intent, but a scanner cannot read the voter intent.”

Other Possible Delays

Another issue that could cause delays in counting Nevada’s early votes would echo a problem faced in Iowa, where that state’s party found that it had understaffed its operations center—when it had to shift from relying on an electronic system that failed to manually recording results from precinct caucuses across the state.

If there was an issue with scanning ballots, the February 14 memo from Mounce said that a “Ballot Review Team comprised of the General Counsel for the Nevada State Democratic Party and two other individuals appointed by the [party] Chair… will review, by hand, each ballot.”

In other words, in the party’s two early vote-counting centers (that are closed to the press and public—one in Reno and one in Las Vegas), only three people, so far, are empowered to hand-count what could be many ballots out of the nearly 75,000 early votes cast.

FairVote’s Richie suggested that the NSDP was making more work for itself by having conflicting counting rules for early and regular caucuses. Mounce’s memo said that the party will not count any early ballot if a voter only made one presidential choice (although it will count their ballot if that voter picked the same candidate three times). In contrast, caucus-goers can leave after the first round, and their vote will count.

Curiously, Mounce’s memo also said that the “Ballot Review Team” would judge voter intent issues on early ballots, but the party’s “Nevada 2020 Caucus Recount Manual” said that post-caucus recounts filed by candidates “will not allow for challenges to the intent of a voter’s preference”—an apparent inconsistency.

But the big picture is that seasoned election officials and technology experts have doubts that the NSDP’s automated electronic vote-counting system for early votes will be accurate—because the paper ballots don’t have alignment markings, and because the underlying cast vote record (via Google forms) may be incomplete.

These concerns all come before Nevada’s statewide precinct caucuses on Saturday, February 22. If the party encounters any of these problems behind the closed doors of its counting centers, it is an open question as to whether or not it would publicly acknowledge them. The presidential campaigns can have one observer at the vote-counting centers, Mounce’s memo said. (Election protection lawyers working with the campaigns did not reply to requests to comment for this report.)

Meanwhile, as the party looks toward Saturday’s statewide caucuses, it sent out a release on Wednesday announcing the “State Party will host more than 55 trainings in person and online before Caucus Day” and included a link to its training hub. “We need Nevada Democrats across the state to volunteer to help us make our 2020 Caucus the most expansive, transparent, and accessible caucus yet,” the website said.

Steven Rosenfeld is the editor and chief correspondent of Voting Booth, a project of the Independent Media Institute. He has reported for National Public Radio, Marketplace, and Christian Science Monitor Radio, as well as a wide range of progressive publications including Salon, AlterNet, the American Prospect, and many others.