Jill Lepore: The problems inherent in political polling

«Stay out of American elections,” Bernie Sanders warned Vladimir Putin, after U.S. intelligence officials informed him that the Russians are meddling in the Democratic Presidential primaries on his behalf. They’re still shilling for Donald Trump, too, setting up fake social-media accounts and spreading fake news, which apparently bothers the President no more now than it did the last time around. But it is difficult to defend the integrity of American elections against foreign interference when Americans have come to accept so much domestic interference. Michael Bloomberg is attempting to buy his way to the Presidency and, while plenty of people have complained about it—“That is called oligarchy, not democracy,” Sanders said—no one has done anything about it, and it was the Democratic National Committee, not the Internet Research Agency, that made it possible for Bloomberg to purchase a place in the debates.

Before 2015, when Fox News put Trump at the center of its debate stage, and asked him the lion’s share of the questions, polls had never been used to determine which major-party candidates would be allowed to participate in a televised debate, or where they would stand, or how many questions they would get. Reputable polling organizations, including Pew and Gallup, did not participate in this charade; pollsters at Bloomberg Politics were among those who complied. Four years later, notwithstanding how badly this worked out for Republicans, the D.N.C. decided to use the same method, a decision that doomed the slow-starting campaigns of the likes of Michael Bennet and Julián Castro. If the method narrowed the field, it did not improve the calibre of the candidates. And almost no one blinked an eye.

Nearly every major polling outfit miscalled the 2016 Presidential race. Most did a lot better at predicting the 2018 midterms. Still, a majority of Americans don’t trust polls. Polls measure something, but it’s often the wrong thing (fame, money). They’re like S.A.T. scores. The problem isn’t really their accuracy; it’s the damage they do. When modern polling began, in the nineteen-thirties, George Gallup claimed that it rekindled the tradition of the town meeting, but most members of Congress considered it to be, as one wrote, “in contradiction to representative government.” In 1949, the political scientist Lindsay Rogers complained that “pollsters have dismissed as irrelevant the kind of political society in which we live and which we, as citizens, should endeavor to strengthen.” Democracy requires participation, deliberation, representation, and leadership—the actual things, not their simulation.

Tweeting is to talking what polling is to voting. Twitter was launched in 2006. Straightaway, people began using it to wage political campaigns. Barack Obama’s 2008 campaign used it to raise money. Conservatives used it to undermine the press. “Using Twitter to bypass traditional media and directly reach voters is definitely a good thing,” Newt Gingrich said, in 2009. A lot of people thought, and still do, earnestly, that Twitter is good for democracy. “Twitter is one of the places where you actually have your own soapbox,” a user said, in 2011. Twitter dubbed the 2012 race “The Twitter Election,” and, two years later, published “The Twitter Government and Elections Handbook,” which described Twitter as “a real-time measure of public opinion.” Just as Gallup had done decades earlier, Twitter advertised its platform to politicians as “The Town Hall Meeting . . . In Your Pocket.”

All this happened even as a growing body of empirical research demonstrated that the more politically charged the tweet, the more likely it is to reach a large audience, that people who get political information from Twitter are radicalized by the experience, and that Twitter, like Facebook, serves as an excellent medium for propaganda. So wholly did the tiny world of Twitter seem to be the world that, in 2017, the Supreme Court ruled, in Packingham v. North Carolina, that social media is a public sphere, a decision that rested on the Court’s belief in its ubiquity. “Everybody uses Twitter,” Justice Elena Kagan said. In fact, only about one in five Americans has used it. Most people who have a Twitter account rarely use it, and very few of those who do post about politics. The ones who do, post a lot—sometimes even as much as the President—and they’re atypical in other ways, too. A study from a decade ago found that the average political tweeter is “a white male in his 30s or 40s who has moderate-to-high household income and considers himself to be a political junkie.” The Twitterati have become more diverse in the years since; Black Lives Matter and MeToo arose on the platform. Still, it remains a very poor proxy for the electorate. In 2018, according to the Pew Research Center, ninety-seven per cent of all tweets posted by American adults about national politics were posted by ten per cent of tweeters. A disproportionate number of the people in Twitter’s town hall are the sorts of people who were eligible to vote in 1820, before the first, Jackson-era expansion of the electorate: the wealthy, the educated, and the hyperpartisan. Twitter isn’t the future of American democracy; it’s the past.

A simulation of democracy taking the place of the real thing has been a long time coming. It began, arguably, during the 1960 Presidential election, when John F. Kennedy’s campaign hired a pioneering predictive-analytics company, called the Simulmatics Corporation, to provide advice on how a Democrat could win back the White House, using an invention that it called a People Machine. Simulmatics aggregated polls (not unlike the way that FiveThirtyEight aggregates polls), divided the electorate into four hundred and eighty voter types, came up with an algorithm to model their voting behavior, and then conducted a simulation of the election (quite similar to that conducted by the Washington Post’s new Simulator). The firm advised Kennedy to speak forthrightly about his Catholicism, and, after he won, Simulmatics took credit, which led to reports that the President-elect had relied on “a secretly designed robot campaign strategist nicknamed a ‘people-machine.’ ” Had he cheated? Should that kind of thing be illegal? People asked those questions, but then, after a while, no one blinked an eye.

“Some critics say this is dehumanizing,” a Simulmatics executive admitted. But, he asked, “Why should politicians operate in the dark? If there are two People Machines working against each other in a political campaign—that would be progress.” Everyone’s got a people machine, lately. You’ve probably got one in your pocket. It is not progress. ♦