A peer‑reviewed paper published Monday says a wave of racially coded, digitally targeted political ads likely nudged thousands of nonwhite voters away from the polls in the 2016 presidential election, an effect the authors say was big enough to matter in razor‑thin swing states like Wisconsin.
The study, appearing in the Proceedings of the National Academy of Sciences, links ad‑exposure records to individual voter files to see whether people who actually saw certain ads later sat out Election Day. The authors argue that this person‑level approach reveals a modern, algorithm driven form of voter demobilization that operates quietly through social feeds instead of at the polling place.
According to University of Wisconsin–Madison, three researchers from the university and a coauthor who previously studied there are listed on the paper. Coauthor Young Mie Kim told Wisconsin Public Radio that platform algorithms and microtargeting tools made it possible to zero in on people who were more easily discouraged from voting.
Using matched logs of who saw which ads and public turnout records, the team estimates that people exposed to these online suppression messages were, on average, 1.86 percentage points less likely to cast a ballot. Scaled nationally, that comes out to roughly 4.7 million votes, according to the paper.
The effect was far from uniform. The authors say nonwhite voters living in minority‑majority counties in battleground states showed turnout rates about 14.2 percent lower than a comparable group that did not see the suppression ads, a pattern they describe as both racialized and tightly focused on key geographies.
Many of the ads were purchased by groups that left almost no public trace and did not register with the Federal Election Commission or the Internal Revenue Service. The study also links some of the content to the Russian Internet Research Agency. The campaigns featured messages urging voters to back third‑party options, attacking candidates from the same party, and encouraging some users to skip the election entirely, according to contemporary coverage of the report.
How the researchers did it
The authors leaned on ad‑tracking data from users who consented to have their online political advertising monitored, then matched that information to public voter files. That gave them a rare, direct line from the ads that appeared in a person’s feed to whether that person later showed up on the voter rolls.
The method builds on earlier work from Project DATA at UW–Madison and other ad‑tracking efforts that try to reconstruct opaque targeting strategies from outside the social media platforms themselves. In an environment where campaign ads can vanish from public view as quickly as a refreshed timeline, this kind of tracking is one of the few ways to see which messages actually landed where.
Legal and policy questions
The study’s authors and voting‑rights advocates say the findings highlight gaps in disclosure rules and enforcement that let hidden players buy targeted political influence without the transparency that usually comes with campaign spending. Groups have been pushing for clearer regulations and more robust platform reporting so investigators and regulators can tell who is paying for specific political messages online and who is being targeted.
Policy analysts in the report say federal disclosure standards have not kept pace with algorithmic targeting tools that can slice audiences into extremely narrow segments. Without stronger rules, they argue, digital voter suppression can continue to operate largely out of public view.
Why Wisconsin mattered
The paper pays particular attention to battleground states such as Wisconsin, where margins were razor thin. Hillary Clinton lost Wisconsin by about 22,748 votes in 2016. When the authors’ national turnout estimate translates into millions of votes, they argue that tightly focused demobilization efforts in small geographic areas could plausibly have altered Electoral College outcomes in some close states.
The study does not claim to pinpoint a single cause for any state’s result, but it frames Wisconsin as a case study in how relatively modest shifts in turnout among specific groups can loom large when the statewide gap is only a few tens of thousands of ballots.
What to watch ahead
Researchers say the timing of the findings matters as campaigns and platforms continue to tweak their targeting tools and as new AI systems change how content is selected and boosted. The paper’s authors and election‑integrity advocates argue that better disclosure, stronger transparency from platforms, and clearer rules for online political advertising are needed to spot and stop digital voter suppression before the next high‑stakes election cycle.
For Wisconsin, the lesson is close to home. A study that ties specific online ad exposures to real‑world turnout suggests that voter suppression no longer needs blocked doors or bogus flyers. It can run quietly through recommendation engines and ad auctions. The researchers say the rule book for political advertising on the internet is overdue for an update if regulators want to keep up with the tactics already in play.
