Did Cambridge Analytica Sway the Election?

Tufts political scientist Eitan Hersh offers a different take on the controversial firm at a Senate Judiciary Committee hearing
Eitan Hersh being sworn in along with two others at a U.S. Sentate hearing
“In an environment where there is a lot of stimuli—a lot going on in the campaign—a lot of campaign ads don’t really work,” said Eitan Hersh, at left, being sworn in at the Senate Judiciary Committee hearing. Photo: Mandel Ngan/Getty
May 17, 2018

Share

In the controversy surrounding Cambridge Analytica—the political consulting firm that worked for Donald Trump’s campaign—and Facebook, concerns about foreign governments interfering in U.S. elections and privacy violations of Facebook users have been paramount.

But there’s little evidence that Cambridge Analytica was in fact able to sway the electorate in the 2016 presidential election through its use of Facebook data, Eitan Hersh, an associate professor of political science at Tufts, told the U.S. Senate Judiciary Committee on May 16.

“The idea that Cambridge Analytica could use Facebook likes to predict personalities and use those predictions to effectively target ads strikes me as implausible, given what we know about the significant challenges in persuasion in campaigns,” he said. “No evidence has been produced publicly about the firm’s profiling or targeting to suggest that its efforts were effective.”

Hersh testified alongside Christopher Wylie, a former Cambridge Analytica contractor, and Mark Jamison, a visiting scholar at the American Enterprise Institute, at the Judiciary Committee hearing “Cambridge Analytica and the Future of Data Privacy.”

Sen. John Kennedy (R-Louisiana) countered by saying he thought it is “rubbish” to think that people in America are not persuaded by advertising. Hersh responded that “just because campaigns spend a lot of money on advertising doesn’t mean it works. For a long time, campaigns were spending a lot of money on robocalls,” he said. “Countless experiments have shown that robocalls do nothing. In an environment where there is a lot of stimuli—a lot going on in the campaign—a lot of campaign ads don’t really work.”

Political campaigns use data to mobilize people to get out and vote, and, separately, to persuade voters to adopt a point of view, Hersh noted. Campaigns already have access to public records that detail voters’ party affiliation, race, gender, age, and geography, which “is very informative in identifying which voters will support Democrats or Republicans,” Hersh said in written testimony, citing research from his 2015 book Hacking the Electorate: How Campaigns Perceive Voters (Cambridge University Press). That information is used to try and drive voter turnout.

Still, predictions about the race or ethnicity of a voter, for example, are wrong about 25 percent of the time. “So when a campaign sends a message, a quarter of the people receiving the message will be mistargeted. Research suggests that voters penalize candidates who mistarget them,” he said.

Several senators at the hearing brought up the issue of how Cambridge Analytica might have sought to suppress voter turnout, especially among minorities. Would those efforts have been effective? “I don’t think we know the answer,” Hersh said. “As far as I know, all the experiments in this domain have been about increasing participation. But it might be easier to demobilize in the sense that encouraging a person not to take a time-consuming action—voting—might be easier than encouraging them to take the action.”

And if mobilization—or demobilization—is difficult, then persuasion is even harder, Hersh said. “A person who was persuadable yesterday may not be persuadable today,” he wrote. “Moreover, a persuasion effect is quick to decay.”

“Probably nobody in this room,” he told the committee, “or nobody that anyone in this room knows, changed their mind as the result of any campaign ad in the election of 2016. . . . In a presidential election in particular, when there is so much going on, the effect of one ad, one kind of ad, one robocall, is usually zero.”

That said, he added, there is much that is not known about how campaigns are using social media to target voters. “This controversy gives us some anxiety, in part because we don’t really know where the line is between ads attempting to persuade voters and ads attempting to manipulate or deceive voters,” he said.

That’s all the more true because Facebook, he said, “hasn’t really taken seriously its solemn civic role as a facilitator of news and political communications.” Facebook has a lot of data, “and I think it’s acted really inappropriately in terms of how it’s conveyed news, sold ads to just about anyone who wants to, [such as] targeted hate groups,” he said. Hersh said Facebook uses algorithms and then claims “‘It’s not our fault if it goes wrong.’ I think it’s a really terrible way to conduct a business.”

Hersh urged Americans to get their news not from Facebook and other social media, but from “news organizations led by editors who tale seriously their duty to inform the public about the range of news and commentary necessary for informed citizenship.”

Taylor McNeil can be reached at taylor.mcneil@tufts.edu.