![[Adobe Stock]](https://sp-ao.shortpixel.ai/client/to_webp,q_glossy,ret_img,w_750,h_375/https://www.psypost.org/wp-content/uploads/2024/08/voter-750x375.jpg)
Digital advertisements designed to discourage voting were heavily aimed at specific demographic groups during the 2016 United States presidential election. People who saw these undisclosed political advertisements were less likely to cast a ballot compared to those who did not. The research, published in PNAS, presents real-world data connecting personalized social media messaging to offline voting behavior.
Political campaigns have a history of trying to demobilize selected segments of the population. This practice is known as voter suppression. It involves targeted strategies intended to discourage or prevent opposing demographic groups from casting ballots.
Historically, voter suppression manifested through physical intimidation or strict localized regulations. In previous eras, tactics included regulatory devices such as poll taxes, stringent identification laws, and deliberately confusing information about polling locations. Today, these targeted efforts have increasingly shifted to the digital sphere. Modern platforms operate on customized feed algorithms that allow messages to reach specific individuals.
Advertisers use microtargeting to reach these specific audiences online. They rely on vast amounts of data regarding user interests, geographic locations, and demographic backgrounds. Social media companies package this data into consumer categories, which allows political groups to deliver customized messages to very narrow slices of the public.
Government reports later showed that Russian operatives purchased platform advertisements using historical search terms associated with the African American Civil Rights Movement to find targeted users in 2016. Many of these digital strategies operate in regulatory blind spots. The messages frequently come from undisclosed campaigns that do not file financial reports with traditional tax agencies or federal election regulators. Because these sponsors remain anonymous, misleading election content can spread unchecked across social networks.
Measuring exactly who saw specific advertisements and tracking whether those people voted is extremely difficult. Most prior studies relied on computer simulations or asked people to self report their voting histories, which can be inaccurate. Young Mie Kim, a media researcher at the University of Wisconsin Madison, recognized this gap in the research.
She worked with Ross Dahlke, Hyebin Song, and Richard Heinrich to design an observational study measuring direct exposure to anonymous negative election advertisements. The team wanted to know exactly who received these messages. They also sought to evaluate whether the visual exposure was tied to actual turnout at the ballot box.
To monitor advertising exposure, the researchers asked thousands of volunteers to install a custom digital tracking application. The tracking program functioned similarly to a conventional ad blocker. Instead of blocking the promotional content, the program cataloged each advertisement and its associated data on a secure research server. During the six weeks leading up to the 2016 election, the application recorded every political advertisement displayed on the participants’ social media feeds.
A major challenge in studying social media influence is accounting for user choices, often called self selection bias. When individuals browse unpaid posts on social networks, they actively select which accounts to follow and interact with. This mechanism makes it difficult to separate preexisting political beliefs from the influence of new information.
Digital advertisements operate differently because they are delivered solely based on algorithmic targeting rather than user subscriptions. A person encounters a promotional message simply because the sponsor paid to put it in their feed. By analyzing these forced exposures, the researchers could remove self selection from the equation, adding validity to their measurements of electoral influence.
The researchers also asked the participants to complete a survey about their political leanings and demographic backgrounds. Following the election, the team partnered with external data firms to link these profile surveys and advertisement logs with official localized voting records. This allowed the researchers to confirm whether a person actually voted without having to rely on the individual’s memory.
Kim and her colleagues reviewed the collected advertisements to identify specific forms of voter suppression messages. They looked for content encouraging election boycotts or promoting third party candidates primarily to split votes. For the central statistical analysis, the team isolated tens of thousands of messages sponsored by anonymous entities.
The researchers identified common themes utilized by the anonymous sponsors. Campaigns often spread deceptive information about voting mechanics, such as telling users they could vote from home using a text message or social media post. These tactics were built directly upon historic efforts to depress voter turnout, tailored to modern digital consumption habits.
The research team documented a highly specific pattern of distribution for these advertisements. Non-White voters residing in counties with high populations of racial minorities within battleground states received a disproportionate volume of negative voting messages. The data showed that these specific demographic and geographic groups were intensely targeted compared to white voters living in less competitive electoral regions.
To estimate the effect on voting behavior, the researchers used a statistical adjustment technique known as entropy balancing. This method creates groups of exposed and unexposed people with closely matching traits. By pairing individuals who shared the exact same age, income, education, and political ideology, the researchers could compare variations in their final voting habits. Since the exposure happened before the election, the timeline ensures the advertisements preceded the voting behavior.
Across the entire sample population, exposure to voter suppression advertisements was connected to lower voter turnout. On average, the voting rate of people exposed to the advertisements was about two percent lower than those who never saw the messages. Several battleground states in 2016 were decided by margins of less than one percent, meaning even subtle shifts in voter participation could alter final electoral outcomes.
The researchers noted an even larger drop in turnout among the specific groups tracked most heavily by the targeted algorithms. Non-White voters living in minority population centers within battleground states experienced the largest declines in voting rates after exposure. The targeted subpopulation saw a voting drop of roughly 14 percent compared to counterparts who did not encounter the negative election messages. This indicates that the advertisements had distinct and varied effects depending on the demographic profile of the matched audience.
To verify their work, the researchers tested the data against multiple control groups. They compared the targeted subjects with voters who interacted with generic political messaging and voters who saw no political advertisements at all. The patterns of suppressed turnout remained consistent across the different groups. The researchers also noted that people exposed to positive political advertising saw slight increases in total turnout, highlighting the unique depressive effect of the suppression messages.
The study relies entirely on observational data rather than an actively manipulated, randomized experiment. Although the researchers used matching techniques to account for confounding variables like income and political ideology, unknown factors could still theoretically influence the results. A person’s local community environment, for instance, might impact their decision to visit a polling location on election day. Consequently, the team advises caution when making direct causal assumptions about the digital advertisements and individual voting decisions.
The results are also specific to the political context of the 2016 presidential contest, as the digital advertising landscape and social media moderation policies shift continuously with each election cycle. Future observational research could focus on other election periods to build a more comprehensive understanding of how customized online messaging affects localized voting habits. The study, “Targeted digital voter suppression efforts likely decrease voter turnout,” was authored by Young Mie Kim, Ross Dahlke, Hyebin Song, and Richard Heinrich.
