Vote counters made numerous mistakes during an audit of Georgia’s 2020 presidential election in Fulton County, including double-counted and misallocated votes, according to a consent order recently approved by the State Election Board and the county.

The finding doesn’t change the outcome of the audit, which showed that Democrat Joe Biden defeated Republican Donald Trump in Georgia. The audit supported two machine counts that found Biden won by about 12,000 votes.

State election investigators blamed “human error” for mistakes in a system that relied on sorting paper ballots by candidate, counting them by hand, writing totals on sheets of paper and then transcribing numbers into computers, according to the consent order and investigative files obtained by The Atlanta Journal-Constitution through the Georgia Open Records Act.

Double-counting likely occurred when election workers couldn’t tell whether auditing software had recorded their initial tally, leading them to enter numbers a second time. In several cases, they mistyped vote totals or allocated votes to the wrong candidate.

The State Election Board, following an investigation into Fulton County's audit of the 2020 presidential election, ordered the county to “cease and desist” from violating procedures in future audits, implement new audit procedures and adequately train elections staff. Miguel Martinez / miguel.martinezjimenez@ajc.com

Credit: Miguel Martinez/AJC

icon to expand image

Credit: Miguel Martinez/AJC

A rough estimate by the AJC indicates the errors identified by investigators amounted to about 3,000 too many absentee votes counted for Biden during the audit, which was not used as Georgia’s certified vote count. Despite inaccuracies in the ballot batches that were investigated, the overall count in the audit was close to the official machine results.

The State Election Board ordered Fulton County to “cease and desist” from violations in future audits, implement new audit procedures and adequately train elections staff. The board approved the consent order June 21, during the same meetings in which it scuttled a potential takeover of the county’s elections and dismissed allegations of fraud on election night at State Farm Arena.

“The reported inconsistencies were the result of human error in entering the data, which were not discovered in time to make corrections due to time limitations in completing the risk-limiting audit and the sheer amount of ballots, and not due to intentional misconduct,” the consent order states. “The discovered errors were a fractional number of the total votes counted and did not affect the result of the 2020 general election.”

Joe Rossi, a Houston County resident who identified errors in the audit, said the investigation exposed flaws in claims about the election’s accuracy by Secretary of State Brad Raffensperger and Fulton County.

“They never acknowledged the errors in those audits. They continued to lie to the public and say that they verified everything,” Rossi said. “They had no desire, intention or effort to get to the truth. They wanted to get to a number, not the truth.”

Raffensperger’s office stood by the veracity of the audit, which was designed to statistically confirm with a high degree of confidence that the machine ballot count arrived at the correct winner of the election.

“The investigation specifically found that any data entry errors committed by Fulton did not affect the results of the 2020 election. The case is now closed,” Deputy Secretary of State Jordan Fuchs said.

Election officials in Fulton County provided several explanations for mistakes in the audit.

Auditing software by VotingWorks was slow to update data, and elections staff reentered numbers when they couldn’t determine whether the system had accepted the results, sometimes causing duplicate entries, according to a letter from the county. In addition, the rushed nature of the six-day audit of 524,000 ballots cast in Fulton, including 147,000 absentee ballots filled out by hand, left little time to check and correct errors.

“The RLA (risk-limiting audit) was the first of its kind in Georgia. The counties were not properly assisted in preparing for the RLA and were not given tools and resources to allow them to reconcile the data that was entered,” a county attorney wrote to the secretary of state’s office. “A more robust system capable of handling the data and additional time would eliminate the issues.”

Allegations that Fulton mishandled the audit drew the attention of Gov. Brian Kemp, who sent a November 2021 letter telling the board he had vetted the inconsistencies and asking for a quick investigation. Kemp declined to comment Friday.

Ultimately, the audit found 345 additional net votes for Trump in Fulton County, a Democratic stronghold where Biden received 73% of the total vote. Biden received 243,000 more votes than Trump in Fulton.

The point of the audit was to find out whether voting machines determined the right winner, but election audits aren’t intended to confirm the exact vote count, said Ben Adida, executive director for VotingWorks.

“It’s not surprising that some mistakes were made in the process of counting more than 5 million ballots. And it’s important to analyze those mistakes, just like the secretary of state’s office did, to ensure that audit practices always improve over time,” Adida said. “At the same time, the uncovered errors don’t change the outcome or effectiveness of the audit: The winner of the election in Georgia in 2020 was Joe Biden.”

Besides checking for inaccuracies caused by data-entry mistakes, the investigation also examined several ballot batches in which Biden received all votes cast, resulting in counts that awarded him 100 votes and gave 0 to Trump. Election investigators have said the lopsided counts could have been caused by election workers who sorted ballot batches by candidate or entered vote totals incorrectly.

In some cases, handwritten vote counts were correct but they were typed into computers inaccurately. For example, election workers mistook the number “47″ for “97″ because handwriting made the four look like a nine.

Future audits will include software improvements to prevent the entry of duplicate ballot batches and flag missing batches, VotingWorks said.

The audit of the 2020 presidential election was the first in Georgia history. Another audit was conducted on a smaller scale after last year’s midterms, verifying Raffensperger’s reelection over Democrat Bee Nguyen.

Under a state law passed this year, audits will be required of at least one statewide contest after every primary, runoff and special election.


2020 Georgia vote comparison

Initial machine count

Joe Biden led Donald Trump by 12,670 votes

Hand audit count

Biden led by 12,284 votes

Machine recount (final certified result)

Biden led by 11,779 votes

Source: Georgia secretary of state’s office