Facebook users who deactivated their accounts for six weeks before the 2020 U.S. presidential election weeks may have been less likely to cast a vote for Trump as a result, according to a new study that is the largest of its type to date.

The study, “The effects of Facebook and Instagram on the 2020 election: A deactivation experiment,” was published in Proceedings of the National Academy of Sciences on May 13 by Hunt Allcott (Stanford), Matthew Gentzkow (Stanford), Winter Mason (a computational social scientist at Meta), and Joshua A. Tucker (NYU). Gentzkow presented the paper at a Harvard Kennedy School seminar I attended in January. The research was conducted as part of the U.S. 2020 Facebook and Instagram Election Study, a collaboration between academics and Meta researchers.

In the study — the largest of its type to date — 19,857 Facebook users and 15,585 Instagram users were paid to deactivate their accounts for six weeks.¹ A control group deactivated their accounts for just one week.²

The researchers matched participants to their state voting records, to their publicly recorded campaign donations, and to Meta platform data. A subset of participants also “opted into passive tracking that directly recorded their use of news and social media apps and websites in exchange for an additional payment.”

The researchers write:

Our estimates are only directly informative about the set of people who agreed to participate in the study and were willing to deactivate their accounts for the payments we offered, and we do not know what the effects might have been in a different year or outside of an election period. However, we do think our results can inform readers’ priors about the potential effects of social media in the final weeks of high-profile national elections.

Here’s what happened when people deactivated their Facebook accounts:

Less knowledge of current events

The people who deactivated their Facebook accounts for six weeks had less “news knowledge”; in other words, when the accuracy of their beliefs about international and domestic news events was measured, the people off Facebook did worse. (They were asked about statements like “A militia’s plot to kidnap Michigan Governor Gretchen Whitmer was foiled by undercover agents”; “Pope Francis voiced support for same-sex civil unions”; “Amy Coney Barrett, Donald Trump’s nominee, became the newest Supreme Court Justice.”)

Slightly less belief in misinformation

People who deactivated their Facebook accounts were also slightly less likely to believe in common misinformation, the researchers found; deactivation had a “marginally significant positive effect on fact knowledge.” (Some examples of inaccurate claims tested: “Amy Coney Barrett said a woman needs a man’s permission to own property”; “The U.S. government has a plan to force a COVID-19 vaccine on everyone”; “Donald Trump held a Bible upside down in front of a church.”)

Slightly fewer votes for Trump

People who deactivated their Facebook accounts may have been less likely to vote for Trump, the researchers found. While this effect “falls just short of our preregistered significance threshold of Q < 0.05,” as they put it in the paper — i.e., the threshold they defined before running the study — the effect was also “not tiny,” Gentzkow noted in the seminar, adding, “If anything, a very small number of people actually shifted their votes from Trump to Biden when they were deactivated.”

Here are two ways to think about what “not tiny” means. Here’s how the paper’s authors describe it:

The point estimate for the effect of Facebook deactivation on Trump vote is a reduction of 0.026 units (P= 0.015, Q = 0.076, 95% CI bounds = -0.046, -0.005). This effect falls just short of our preregistered significance threshold of Q < 0.05…

To put the magnitude of a 0.026 unit effect on our Trump vote variable in context, note that this would result if Facebook deactivation caused 1.3% of Trump voters to switch to Biden, or if it caused 2.6% of Trump voters to not vote. SI Appendix, section D.4 shows that the associated point estimate for the effect on the Trump–Biden two-party vote share within our sample is 1.16 percentage points.

Or, here’s how Gentzkow put it to me when I asked him in a follow-up email:

“The estimates imply that being off of Facebook reduced the share of voters choosing Trump by about 1 percentage point. That’s what would happen if 1 percent of voters switched from Trump to Biden, or 2-3 percent of Trump voters decided not to vote.”

Small or zero effect on polarization, perceived legitimacy of the election, and turnout

Deactivation did not have a detectable effect on beliefs in the election’s legitimacy (whether voting was suppressed, whether there was electoral fraud, etc.) or on voter turnout (as measured by matching participants to state voting records). It also did not have a detectable effect on affective polarization or issue polarization.

“This suggests that if Facebook access contributes to political polarization, the effect is either small or accumulates over a longer period than 5 weeks,” the authors write.³

Less online participation

Participants didn’t just take all the time they’d spent on Facebook and shift it to other sites.

The Facebook users had been spending an average of 43 minutes a day on Facebook before they were deactivated. When they were deactivated, they shifted some of that activity to other social apps and news apps, but “if you add all those things together, it’s like a quarter of the total time they were spending on Facebook,” Gentzkow noted. Deactivated users signed fewer online petitions and posted less about politics.

“Facebook does have the impact of getting people more engaged with politics online,” Gentzkow said.

1. The members of the test group were paid $150 to deactivate their Facebook account for six weeks, from September 23 to November 4, 2020. The members of the control group were paid $25 to deactivate their account for one week, from September 23 to September 30, 2020. (To be eligible for the experiment, participants had to be willing to have their accounts deactivated for either amount of time.) Meta implemented the deactivations directly, and people who tried to sign back in were warned they’d lose their payment if they did. [↩]
2. You can’t do a double-blind Facebook deactivation where people don’t actually know whether they’ve deactivated their accounts, Gentzkow pointed out — “we don’t have a sugar pill for Facebook.” Changing the amount of time people deactivated their accounts, rather than having one group deactivate their accounts completely, made it easier to compare the results. [↩]
3. This finding differed from what Hunt and Gentzkow saw in a smaller 2018 study, where “we found a more significant effect on issue polarization” when people deactivated their Facebook accounts. [↩]