A study led by a University of Wisconsin researcher shows the first empirical documentation of digital voter suppression on social media and foreign election interference in the 2016 U.S. presidential election. The study was published Jan. 26 in the official journal of the U.S. National Academy of Sciences.
The study focuses on voter suppression on social media, specifically looking at targeted ads on Facebook, a platform that has seen the most voter suppression, according to lead researcher UW School of Journalism and Mass Communication professor Young Mie Kim.
Unlike advertisements in traditional print and broadcast news, targeted advertising is essentially invisible to social media users, unless they are the ones being targeted by the ad’s sponsors, according to Kim.
“This means these ads are not publicly available,” Kim said. “So it was almost impossible for researchers like me to study what kind of political ads people are exposed to on social media.”
For the study, Kim developed an app named EScope, which worked similarly to an ad blocker, capturing ads and showing them to study participants. The study pulled from a representative sample of more than 10,000 U.S. voting-age population participants, who used EScope for six weeks until election day, according to Kim.
The app captured participant data on all the ads, who sent them and when the user was exposed. Every week Kim and her team asked users to complete a survey which recorded data on demographics, ideology, candidate preferences and policy position for each of the participants.
“By merging these — the ads and the survey questions — I was able to reverse engineer a model to identify targeting patterns,” Kim said. “I found that … [in] the counties where the racial minority population is more than 50% of the total population, these people received 10 times more voter suppression ads.”
Kim and her team, Project Digital Ad Tracking and Analysis, studied the targeting of U.S. voters on Facebook during the 2016 election and identified thousands of campaign-style ads that fostered fear, included threats and aimed to sow division among the public, according to the study. The majority of the sponsors for those ads were unidentifiable and untraceable, as they were not found in Federal Elections Committee or Internal Revenue Service data, according to the study.
This study by Kim and her team takes this research a step further.
The sponsors of these ad campaigns are not all known, but 17% of the campaigns studied are connected to Russian disinformation operations in the 2016 election, according to Kim.
“A lot of journalists said that Russian election interference is all about sowing division, so you target a conservative and then send a conservative message and then target a liberal and then send liberal messages,” Kim said. “But I found here in this study is that, no, it was not about mobilizing people, it was actually about demobilizing certain types of people.”
Participants were matched in pairs based on similarity, then one was exposed to voter suppression, and the other was not. The study shows that those who were exposed to targeted ads of voter suppression did not turn out to vote, with a difference of 1.9% of the population.
This trend is seen at a higher rate in non-white voters in minority counties in battleground states, such as non-white voters in Milwaukee County, Wisconsin, according to Kim. The difference in voter turnout between voters of this group compared to white voters in the same states is 14.2% according to Kim. In 2016, actual voter turnout differences between white and non white voters were around 12%, according to Kim.
“I think this study provides a smoking gun that closes and ends the discussion,” Kim said.“This study provides some empirical evidence for a — georacially targeted voter suppression ads [were] actually effective, and b — targeted digital voter suppression might explain the huge turnout gap between non-white voters and white voters, especially non-white voters, in battleground states.”
