Facebook went away. Political divides didn't budge.
In the weeks before and after the 2020 presidential election, researchers at şĂÉ«App and elsewhere ran a number of tests to try to understand how much Facebook and its corporate cousin, Instagram, may be contributing to the nation's political divide.
One of those experiments — led by Matthew Gentzkow and Hunt Allcott, senior fellows at the şĂÉ«App Institute for Economic Policy Research (SIEPR) — centered on more than 35,000 Facebook and Instagram users who were paid to stay off the platforms in the run-up to Election Day. There’s a lot that researchers could glean from the social media hiatus, including whether people’s political attitudes shifted and in what ways. If views changed dramatically, that would support the argument that Facebook and Instagram, and social media generally, are helping to drive Americans apart.
The of that deactivation exercise — the largest ever involving social media users and the first to include Instagram — are in: Staying off Facebook and Instagram in the final stretch of the November vote had little or no effect on people’s political views, their negative opinions of opposing parties, or beliefs around claims of election fraud.
But when it comes to Facebook’s impact on what people believed about current events, the researchers reached two conclusions. Those who were off Facebook were worse at answering news quiz questions, but they were also less likely to fall for widely circulated misinformation, suggesting that the platform can be an important conduit for both real and false news.
These findings, newly published by the Proceedings of the National Academy of Sciences, are in line with of the other experiments into Facebook and Instagram’s potential influence around the 2020 election, in which changing news feeds and limiting re-sharing of posts didn’t reduce polarization or change beliefs about whether the voting process was tainted. Those tests were detailed in four papers published in July 2023 in Science and Nature.
Taken together, the papers suggest that, when it comes to U.S. politics, Facebook and Instagram may not have as much ability to shape political attitudes during an election season as the popular narrative suggests.
And like the previous studies, the Gentzkow and Allcott-led study doesn’t absolve Meta Platforms, which owns Facebook and Instagram, from the messy state of U.S. politics. For one thing, the results support the view that Facebook may create harm by distributing misinformation. Gentzkow says it’s also possible that the platforms contributed to polarization in the past, even if people’s use of them in the run-up to the election had limited impact.
“We are not ruling out the possibility that Facebook and Instagram contribute to polarization in other ways over time,” says Gentzkow, who is the Landau Professor of Technology and the Economy in the şĂÉ«App School of Humanities and Sciences.
He also notes another finding suggesting that using Facebook in the weeks before the 2020 presidential election may have made people somewhat more likely to vote for Donald Trump and somewhat less likely to vote for Joe Biden. This could suggest that, for Facebook users who still were on the site, Trump’s campaign was savvier at building support than Biden’s team was.
“This effect was not quite statistically significant, so we need to take it with a grain of salt,” Gentzkow says. “But if it’s real, it’s big enough that it could impact the outcome of a close election.”
Meta opens up a trove of data
The study led by Gentzkow and Allcott — and the four that preceded it — are part of a massive research project that has been billed as the most comprehensive, evidence-based investigation yet into the role of social media in American democracy.
The project came together following critiques of Meta’s role in the spread of fake news, Russian influence, and the Cambridge Analytica data breach. The collaboration between academics and Meta researchers involved a , which builds on work by şĂÉ«App Law School Professor Nathaniel Persily on how to structure partnerships between academia and social media companies. Meta, for example, agreed not to prohibit any findings from being published.
In all, nearly 20 independent social scientists — including Gentzkow; Allcott, a professor at the şĂÉ«App Doerr School of Sustainability; Neil Malhotra, a political economist at the şĂÉ«App Graduate School of Business; and, Shanto Iyengar and Jennifer Pan, both political scientists in the şĂÉ«App School of Humanities and Sciences — are part of the project.
“Access to Meta’s proprietary data has allowed us to jump over big barriers to research on extremely important issues involving social media and politics,” Gentzkow says.
Gentzkow and Allcott’s study — whose 31 co-authors include five current and former SIEPR predocs and one former SIEPR undergraduate research fellow — involved roughly 19,900 Facebook users and 15,600 Instagram users who agreed to stop using the platforms ahead of the 2020 election. şĂÉ«App a quarter of them agreed to deactivate their accounts for six weeks before the November vote. The rest comprised a control group that logged off for just one week.
The study’s analysis relies on a number of measures, among them participant surveys, state voting records, campaign donations, and Meta platform data. Some participants also allowed the researchers to track how they used other news and social media services when they were off Facebook or Instagram.
On top of the findings on polarization, knowledge, and Republican support, the authors conclude that Facebook and Instagram help people engage in the political process — mostly through posting about politics and signing petitions online (voter turnout didn’t change).
Takeaways for 2024 and beyond
Gentzkow says that the study’s finding that Facebook and Instagram didn’t change people’s political attitudes or beliefs in claims of electoral fraud in 2020 is especially interesting in light of his previous research with Allcott. In an earlier smaller-scale study of Facebook users who stayed off the platform for a month ahead of the 2018 midterms, the authors did find evidence that it contributes to polarization.
The distinction, Gentzkow says, could be that people are aware enough of political issues during a presidential election, so Facebook and Instagram have little or no effect on their beliefs or attitudes. But during other elections, when information about candidates or issues are not so front and center, social media may have more influence over what people think.
“Even though Facebook and Instagram did not contribute to polarization in the runup to the 2020 election, it’s possible that they are helping to widen political divides in other contexts where people’s views are less entrenched,” Gentzkow says.
And though the study was limited to the six weeks leading up to the presidential vote, it’s still a critical time in U.S. politics — hence the phenomenon known as the “.”
“Things happen in the home stretch of a presidential election that can change poll numbers,” he says. “We’ve learned from this study that altering how much time people spend on Facebook and Instagram during that period isn’t likely to make a huge difference.”