Count U.S. Out: How polls got election day so wrong


On Tuesday November 8, Donald J. Trump was victorious in the United States presidential election. This win has been largely attributed to the electoral college, a system in which certain states are granted more electoral representatives, depending on population and size. This win came as a surprise to many people, both in the United States and around the world, as nearly every poll predicted democratic candidate Hillary Clinton to win.

As the election news begins to settle down, people are taking a look at how the polls were able to get the election results so detrimentally wrong, especially for an election that appeared to be a clear landslide.

Election polls are conducted in many different ways by polling companies, who are groups independent from the government. Polls happen on the internet, in person, and through phone calls. Theoretically, participants are randomly selected by interviewers to answer questions – in election polls specifically, the main question being who they will vote for. Pollsters then weight the data, and verify that there is an equal representation of genders, races, ages, and religions.

Polls as a whole are often seen as flawed, mainly because the data collected often inadvertently excludes large groups of people. Since pollsters often use data collected from landline phone calls, they often miss 18-30 year olds, many of which only use cell phones. Additionally, there are about 7 million Americans who temporarily reside in other countries but are still eligible to vote (such as teachers, soldiers, etc.) that are not included in polls. However, as younger voters were projected to (and did) vote for Hillary Clinton, the usual discrepancies don’t account for the error in the projected win.

“Late swing” votes are partially attributed to the inaccuracy of polls, referring to voters who decided at the last minute to vote for Trump over Clinton. Two out of three polls on election day put Trump in the lead, whereas the 21 polls published the day before the election put Clinton ahead. These variations can be attributed to voters who changed their mind before heading to their voting precinct, or who had yet to make up their minds when participating in polling.

Sources have claimed that the rhetoric surrounding the election could have potentially led to discrepancies amongst polls and results. Because of Donald Trump’s constant claims of government corruption, it’s possible that Trump supporters simply did not answer calls made by pollsters, or lied during exit polls, thinking that these independent poll groups were somehow connected to the government. Additionally, Trump’s speeches carried a haunting Nixon-era message of the ‘silent majority,’ meaning that most people felt the same way about issues that Trump did, but remained quiet on their views to avoid controversy. Some have predicted that a significant group of voters did not state in polls that they would vote for Trump due to a social desirability bias, but turned out on election day when they felt their vote was private. Some voters may have feared judgment from their fellow citizens, as a largely anti-trump movement was seen before and after the election, and may not have disclosed their support of a controversial candidate to pollsters.

Now that the election is over, pollsters are closely examining their methods to determine what must be fixed before the next election cycle. Jon Cohen, former head of polling at the Washington Post and present Senior Vice President of Survey Monkey, told the Washington Post that the state of polls on November 9, the day after the election, was “unsettling… and facing a moment of reckoning.”