Donald Trump’s surprise win over Hillary Clinton in the recent U.S. presidential election spurred a lot of questions about the opinion polls leading up to Election Day and how they were conducted. Harvard University political scientist Gary King said the attention is focused on the wrong group.
“The story is not really about the pollsters; the story is really about the people,” said King, who also leads Harvard’s Institute for Quantitative Social Science.
Certainly, the campaigns turned up twist after twist — Clinton’s swoon at a 9/11 memorial service, seemingly giving credence to reports about her ill health; infighting among top brass in the Trump campaign; leaked Democratic National Committee emails indicating that officials conspired against Clinton’s primary challenger, Bernie Sanders; a surfaced 2005 video of Trump boasting about groping women.
Changing hearts and minds?
Opinion polls zigzagged — Clinton was up by a few percentage points, then she was up by a lot more, then the race tightened and it was tougher to call. Through it all, the media, King said, characterized voters as indecisive, favoring one candidate and then the other as pundits pontificated and “fake news” circulated on social media.
News website Politico asked pollsters about fluctuating polls before the Nov. 8 election and found that when there was bad news about, say, Trump, the Republican candidate, registered Republican voters were less likely to answer questions about the election. Ditto for news on Clinton and Democratic voters.
Voters themselves are remarkably stable, King said. They tend to choose early on whom to support and then move toward that choice throughout the campaign. (King co-wrote a paper on the tension between variable polls and predictable elections in 1993 and stands by it today.)
“If you watch CNN, they’re all obsessed about the people being swayed by this, and the people being swayed by this and they only talk about the horse race,” King said. “Very few people know people who flipped and flopped across the campaign.”
Once the voters get a general understanding of who candidates are and what their ideologies are, King said, forecasters can pretty reliably predict the outcome of the election, at around the times of the Republican and Democratic conventions, plus or minus three or four percentage points.
What happened in this election, he said, wasn’t that voters switched from planning to vote for Trump to Clinton and then back to Trump or vice versa. It was just that the race was incredibly close.
“The facts about people are that people are very stable and very predictable and in this sense trustworthy,” King said. “They’re not swayed by crazy little events in the campaign.”
Rising use of mobile phones means pollsters have to rethink how to run opinion polls during elections. Read about it in this SearchCIO news story.