By 7 a.m. on Saturday, November 14, the windowless, cavernous hall six flights below street level at the Georgia World Congress Center in downtown Atlanta was bustling. Hundreds of county government employees wearing face masks were seated in pairs at 170 tables with pens, forms and bottles of hand sanitizer. They had come to manually count every vote for president in Fulton County, the epicenter of Joe Biden’s still unofficial 14,000-vote victory over President Trump.

Their task was sorting and counting more than half-a-million paper ballots cast in Georgia’s most populous county. As observers from political parties, campaigns and voting rights groups roved, watched and sometimes leered, the assembled librarians, social workers and other civil servants proceeded.

They started by first counting absentee ballots, where voters filled in ovals next to a candidate’s name. By midday, they turned to votes cast in person on Georgia’s new voting system, where voters used a touch screen to make their choices and a ballot was printed out with their votes in readable text and a scannable code.

“It couldn’t be more transparent what we are doing,” Matthew Tyler, a broad-shouldered man and victim-witness advocate for the Fulton County District Attorney’s office, said afterward. “I count them three times. I count them five times. I am happy to spend my weekend like this.”

Throughout the day, supervisors wheeled suitcase-like storage cases to his table and the other counting stations. Reams of paper ballots were taken out. The ballots first were sorted into piles by the presidential choice. The piles were then counted and stacked 10 ballots at a time. When finished, an orange reporting sheet for that batch was filled out. The supervisor brought it to an official at a computer at one side of the hall, who sent each batch’s results to Georgia’s secretary of state.

“I just think that democracy is so serious and so reverent,” Tyler said, reflecting on a process that was slow, steady and careful. “As a young Black man, I know that so many people died before me literally for the right to vote. It’s my responsibility. I take it very personally.”

Tyler wasn’t the only one taking the next phase in Georgia’s election personally. Two days earlier, Secretary of State Brad Raffensperger, a Republican with an engineering background, abruptly decided that the state’s 159 counties could not do a statistical audit required by a 2019 law that he had helped write. The audit’s goal was ensuring that the electronics in a new statewide voting system were accurately counting the hand- and machine-marked paper ballots.

That statistical process, called a risk-limiting audit (RLA), is not the same as a recount—which comes later and only if the official results are close. A recount is a different legal process, one where candidates’ lawyers can challenge the ballots. An audit is not adversarial, but more of a quality-control exercise.

Georgia’s RLA was promoted as an efficient way to double-check vote counts. If the results are not close, several hundred randomly chosen ballots can assess the vote-counting accuracy in a big urban county. But because 14,000 votes divided Biden and Trump in an election of nearly 5 million votes in the state, the mathematical efficiencies evaporated. Raffensperger’s decision to manually count every ballot to satisfy the state’s RLA law left almost everyone involved in the process feeling some heat.

After the hand count was announced, Raffensperger was attacked by President Trump. Georgia’s two Republican senators called for his resignation. Trump supporters followed the cues from Trump’s tweets and Fox News’ hosts who stridently supported the president, including some supporters who converged on Fulton County’s audit with credentials as Republican and Libertarian Party observers. Meanwhile, the county’s election officials had work to do.

“We were preparing for an RLA on Thursday, which probably would have pulled 200 to 250 ballots,” Fulton County Registration and Elections Director Richard Barron recounted at a press conference on Sunday, November 15. “We found out on Wednesday that we were going to do a full hand tally [of] 528,777 ballots. That’s quite a difference from 200 to 250. To put this together that quickly was onerous.”

But Barron’s staff were not the only ones scrambling. Local supporters of President Trump—assisted by a longtime election integrity activist and critic of Georgia’s voting systems, including its newest system, Garland Favorito of Voter GA—were determined to find evidence that could be used in court to challenge Biden’s win.

On Friday, November 13, as the Congress Center was being set up, Favorito, a retired information technology professional, issued a statement saying that he saw Biden’s total jump ahead by 20,000 votes—while Trump’s decreased by 1,000 votes—as Fulton County’s results were updated. He wanted to see data transmission logs, saying a programming error of the same voting system in a rural Michigan county had “erroneously swapped” 6,000 votes from Trump to Biden.

The Michigan incident, which was caught and corrected, alarmed Trump supporters and became part of a growing partisan narrative that unseen dark forces had stolen the election from the president.

Skeptical Observers

By midday Saturday, Favorito and a dozen pro-Trump observers had seen enough to confirm what they had feared and believed, that Georgia’s most vote-rich county had nefariously tipped the scales for Biden. As the hand counting proceeded, they saw stacks of ballots on counting stations that were almost entirely for Biden. Few Trump votes were piled up. (Fulton County’s latest results, posted a day before the audit, showed Biden had 72.65 percent and Trump had 26.16 percent of the vote.) Favorito said he and his monitors had been tracking this trend all morning.

“I live in Fulton County. That’s ballot box stuffing,” he said. “There’s no place in Fulton County that’s going to be 100 percent for Joe Biden or even 99 percent. Seventy-thirty, maybe. Eighty-twenty? Something’s wrong.”

After walking through the rows of tables, Garland’s team of a dozen observers had been noting which counting stations had the disproportionate piles. They were not allowed to talk to the audit board workers. They copied the ballot batch numbers on the storage cases. After taking notes, a small group huddled on the side.

“I don’t know how to report this when there are huge discrepancies,” said the wife of a GOP legislator who didn’t want her name used.

“I’ve written the precincts down,” she said.

“There is no way in hell that’s correct,” replied Favorito, who was usually soft-spoken and courtly.

“I’ve taken pictures,” she continued, pulling out a legal pad. “These are all of my batch numbers where I saw huge discrepancies.”

How did they know that these were discrepancies?

“These are 95-5,” Favorito said.

“Is there a Republican lawyer here, Garland?” the legislator’s wife asked.

Another woman (who didn’t want to be named) broke into the conversation. At one table, she said that the absentee ballots looked “too white and they didn’t look like they were crinkled at all—like they didn’t go through a machine or anything.”

“Did you write that down?” Favorito asked. “That has to be a legal challenge.”

A few minutes later, the two women said that they did not fully understand what they were seeing. They were flustered that they could not ask the counters any questions. They did not trust the county election officials, who they suspected somehow had fabricated ballots. They did not seem to realize that the audit was a small part of a larger post-Election Day sequence. They believed that the audit boards were accurately counting ballots, but they didn’t trust the ballots.

The audit’s narrowly focused tasks flustered other Trump supporters who joined the conversation. The night before, Trump attacked Raffensperger on Twitter, calling him a “RINO”—Republican in name only—and said that he “won’t let the people checking the ballots see the signatures for fraud,” which is not part of this audit’s process. Ballots do not have signatures on them. Absentee ballot return envelopes have signatures and other identifying information that is verified before the ballots are taken out and counted. That two-step process ensures that one’s ballot remains secret, and it occurred before the November 14 audit.

“It’s ridiculous,” groused another woman. “They don’t do the audit of the signature to match the mail-in ballot? So why are they counting?”

When it was pointed out that the signature verification was done earlier, she replied, “How do they explain dead people voting then?”

When asked to explain how she knew if votes had been cast in the names of the deceased, she replied, “How do you know they weren’t?” After a pause, she continued, “Because it happens in other countries. Come on, is America the first country where dead people don’t vote? This happens in every country.”

“There were people who voted in this election who were over 100 years old,” said David Turner, a computer programmer and colleague of Favorito’s. “They died years ago. But their names [are on voter rolls]. They [state officials] haven’t cleaned up the database, the voter rolls. They [dead people] keep voting.”

The night before the audit, Fox News’ Tucker Carlson apologized on air, saying, “One of the people who voted in last week’s [Georgia] election isn’t dead… A whole bunch of dead people did vote… but James Blalock was not among them. … So apologies for that and of course we’re always going to correct when we’re wrong.”

Fulton County’s Audit

Claims of fraudulent voting are neither new nor confined to Fulton County, where election officials rejected the allegations of fabricated ballots sullying the process. Voter fraud claims are cliches that have been promoted by Republicans for many years. They have been raised in Trump’s post-election litigation and have been widely rejected by numerous state and federal judges. The angst behind these accusations is clearly born of frustration that their candidate did not emerge victorious in Georgia’s first unofficial presidential election results.

However, the perception that Fulton County was counting illegally manufactured ballots—because there were much bigger piles of Biden votes and seemingly too few Trump votes—deserves scrutiny. It turns out that these partisan observers misunderstood what they thought they were seeing.

The batches of ballots brought to the counting tables were not all of the votes that had been cast in a neighborhood, explained several county election officials on November 15. However, the Trump supporters assumed that every labeled batch of ballots represented a local neighborhood’s precinct.

“They don’t necessarily correspond to the precincts,” Barron said, at a press conference.

“The reason that they are not sorted by precinct is that Fulton County had much higher percentages of absentee [ballots] and [in-person] early voting,” explained Jessica Corbitt, Fulton County’s director of external affairs. “Over 80 percent of our voters cast their ballot before Election Day. Those processes, especially in early voting, [are] where the bulk of the votes were cast. At State Farm Arena, more than 30,000 people voted. So every 3,000 or so votes, you batch the ballots.”

Across Fulton County, voters from the same neighborhood chose different ways to cast their ballots. Some people voted early at the in-person voting centers, creating batches of ballots that were bundled as they came in. Some people returned their mailed-out absentee ballot in the mail, which was vetted and counted separately. Some people did not trust mail delivery, so they took their absentee ballot to a voting center, where it was canceled, and they used the new touch screen system. Some people waited until Election Day and voted in person at a local precinct.

In other words, the Trump supporters mistakenly assumed that they were seeing all of the ballots from a neighborhood—or precinct—being counted at the same time. What they were seeing, instead, were piles from different slices of the process.

“We had 30 early voting locations all over the county,” Corbitt said. “The Alpharetta Library was the second-highest [turnout], which was in a more Republican-voting part of the county. There are a lot of stereotypes of where people in different parties live. One of these stacks somewhere [in this hall] is the Alpharetta Library… Across the county, we had high turnout, high early voting participation, high absentee ballot participation.”

Corbitt’s explanation was unlikely to sway Trump supporters—who had other complaints. A young reporter for America’s Voice, a conservative outlet, said that she did not trust county employees to accurately count votes because they most likely were Democrats. She asked Barron at the November 15 press conference why he hadn’t assembled more bipartisan teams. The county election director said that he turned to county employees and precinct managers to quickly assemble hundreds of needed workers and to minimize the county’s cost.

Corbitt and Barron’s explanations underscored that election observers are not always familiar with the intricacies of the process that they came to observe.

“Some of it is genuine misunderstanding,” Corbitt said. “Not everyone is an election professional. Not everyone has the same experience observing elections. If someone is relatively new to this, it is a relatively multifaceted process. I think there’s a lot of innocent misunderstanding. I do think there is intentional misinformation, which is what I’m more concerned about. I think that’s dangerous and irresponsible. But if people just don’t know, we are more than willing to be transparent and explain the process.”

Favorito said that the hand count, as far as it went, was a step in the right direction. Fulton County was on track to finish counting on November 15—sooner than expected. Across the state, one-third of the counties had finished. The rest faced a late November 18 deadline.

“I’m really pleased that the secretary ordered the full hand count,” Favorito said afterward. “There are people [some Trump supporters and observers] who don’t understand that this is the first step in the process. The first step is to verify the electronic totals are correct. Then we have to decide if there are illegal ballots, and how to challenge them. That’s the next step.”

Other election integrity activists from out of state had sharper criticisms of Georgia’s hand count audit—such as asking why it couldn’t produce easily understood precinct-based vote results, and why hand count totals were being input into a computer at every county and sent to the secretary of state to compile, which seemed at cross-purposes to counting every paper ballot by hand.

But answering those questions would wait until after 2020’s elections were finished. In the meantime, Favorito had another worry. A colleague, Scott Hall, had stayed up until 2 a.m. on November 15 to crunch Fulton County’s voter registration data. After Barron’s press conference, he held forth with reporters for PBS NewsHour, the New York Times and right-wing outlets and insisted that his analysis showed that 52,000 Fulton County residents voted more than once for president.

“I’ve heard that gentleman who was one of the observers,” Corbitt said. “We obviously don’t believe that’s true. I’m sure you knew that.”

Several hours later, Hall reran the numbers and retracted his assertion, Favorito said. “We were collecting evidence, and he goes and blurts it out before we confirmed it.”

Georgia’s deadline to certify 2020’s election results is Friday, November 20. After certification, losing candidates in very close races can request a recount. That’s a different procedure, legally, from the audit. If there is a recount, each county’s paper ballots would be counted by high-speed electronic scanners.

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.