On November 3, Tina Barton ran into a problem. It was Election Day in the US and Barton, a Republican, was city clerk for Rochester Hills, Michigan, a conservative-leaning community near Detroit. As she was uploading some of the voting results, there was a technical issue, which she reported to Oakland County officials. But the voting data wasn’t fixed for two days—by which time the entire country was looking at the state’s election results.
The change was very, very public, and it generated a huge swell of misinformation. This was supercharged on November 6, when Ronna McDaniel, the chair of the Republican National Committee, flew to Oakland County and held a press conference. She claimed that 2,000 ballots had been counted as Republican before being “given” to Democrats.
“If we are going to come out of this and say this was a fair and free election, what we are hearing from the city of Detroit is deeply troubling,” McDaniel said.
Upset at how the situation was being misrepresented, Barton posted a video on Twitter refuting the claims. She’s been the Rochester Hills clerk for eight years, and when she spoke out against McDaniel, she knew she was putting her career on the line. In the video, which has since been deleted, Barton said, “I am disturbed that this is intentionally being mischaracterized to undermine the election process.”
Her remarks went viral, and they were met with threats and anger. In an email to MIT Technology Review, Barton said that “since Ms. McDaniel’s press conference, I have received threatening voice mails and messages.” One caller claimed to be on the way to Michigan. Barton upgraded the security system of her home.
Data shows that during the election, disinformation was highly targeted locally, with voters in swing states exposed to significantly more online messages about voter intimidation, fraud, ballot glitches, and unrest than voters in other states.
In a data set provided by Zignal Labs, we looked at mentions of over 30 terms related to voter suppression or intimidation, fraud, technical errors, and unrest that focused on a particular polling location. Our sample of 16 states found that between October 1 and November 13, swing states had more than four times the amount of localized voting misinformation: a median of 115,200 such mentions, while non-swing states saw a median of 28,000 related messages.
Here’s a chart showing how the volume of messages changed over the days before the election itself.
Bhaskar Chakravorti, dean of global business at Tuft University’s Fletcher School, conducts research on the conditions that leave a community particularly vulnerable to disinformation. He says that this local focus is typical of effective disinformation campaigns, which are usually pinned to a specific place and slice the target audience into its smallest, stereotyped parts. “Clever misinformation” is organized, he says, in the same way that political campaigning is.