‘Hopefully the Errors Were Random:’ FiveThirtyEight Analyst Talks Election Polls

FiveThirtyEight

Harry Enten, a senior political writer and and analyst for FiveThirtyEight, spoke Tuesday about why many newspaper and data journalists were so surprised that Donald Trump won the presidential election last week. The talk was sponsored by Boston College’s communication department and co-sponsored by the American studies and African and African diaspora studies programs.

Matt Sienkiewicz, an assistant professor of communication at BC, started the discussion with Enten by directly addressing the elephant in the room—that FiveThirtyEight and other news media organizations were incorrect in their nearly universal predictions that Democratic candidate Hillary Clinton would win the presidency.

FiveThirtyEight, one of the media sites most bullish on a Trump victory, pegged Trump’s chances of winning the election at 28.6 percent. Many other journalists and pollsters were less so—the Princeton Election Consortium put his chances at less than 1 percent.

Following Trump’s sizable victory, many commentators have lambasted pollsters for doing a poor job. Enten said he believes that such criticism is largely overblown, however, as most polling outlets only provide percentage chances, not guarantees, that a certain candidate will win an election.

While FiveThirtyEight predicted a Clinton victory as the most likely outcome, Enten said that the commonly repeated refrain that data analysts like him led the nation astray is misleading.


“Hopefully, the errors were random, and not systematic, and will work themselves out by the next time around.”

-Harry Enten, a senior political writer and analyst for FiveThirtyEight


“If you flip heads twice on a coin, that happens about 25 percent of the time. I don’t think that people quite recognize how uncertain the result actually was,” Enten said. “Probabilities actually mean something.”

Enten said that the most surprising results were in Midwestern swing states which, while their residents were expected before Nov. 8 to vote for Clinton in large numbers, ended up going to Trump.

“Errors are correlated across states,” Enten said. “That is, similar states tend to vote together. When similar states have polling errors, they all tend to have polling errors, and that’s exactly what happened.”

While Clinton did receive more popular votes than Trump, the Republican candidate won a majority of electoral votes. Enten cautioned against thinking that, had the Electoral College not existed, Clinton would have won the election. It’s impossible to know what strategies would have been employed by either candidate had the conditions of the race been different, Enten said.

Most political prognosticators believed that Trump was doomed to fail due to his antagonism and lack of outreach to minority voting blocs, which now comprise a greater portion of the electorate than ever before, Enten said. Trump, however, by “running up the score” with white voters, managed to sidestep this hurdle. The low turnout among minority voters also hurt Clinton in critical voting districts.

The idea that significant American demographic shifts, such as a shrinking white population, would spell doom for the Republican Party was proven false, at least in the short run, Enten said.

“Demography is not destiny,” Enten said. “Parties adjust in the ways that they need to in order to win.”

While he said that it’s hard to pinpoint the precise reasons for the widespread underestimation of Trump’s support among voters, Enten singled out factors that might have contributed to the widespread inaccuracy of pre-election polling.

Neither Gallup nor the Pew Research Center, two of the most respected polling institutions in the world, conducted presidential polls this year, Enten said. Journalists, therefore, made predictions based on less reliable sources. The use of robo-polls, which are less accurate than traditional techniques, may have also played a role, he said.

Polling is not an exact science, and faulty predictions should not cause serious concern, Enten said.

“Hopefully, the errors were random, and not systematic, and will work themselves out by the next time around,” Enten said.

Featured Image by Liam Weir / Heights Staff