Most people across the world woke up last Wednesday to an unexpected result of the presidential election, mainly because 98% of all polls showed a substantial, steady lead by Hillary Clinton. And this was not a singular event.
During the past year or so, most polls failed to correctly predict the results of important votes. Starting with the 2015 general election in the UK, where every polling company had Labour and Conservative pretty much neck-and-neck when in reality the Conservatives won by seven points.
The Brexit vote followed, where most failed to acknowledge the possibility of a leave result, and finally the US presidential election, where every poll up until the election showed a clear win for Hillary Clinton.
What Led To These Shortcomings?
First, there is the difference between online-based and phone-based polls. As with Brexit, where throughout the referendum campaign, polls conducted by telephone generally showed Remain ahead, but polls conducted online generally showed a very tight race.
In the US election it was somewhat similar, online polls showed a Clinton win by a much smaller margin than those conducted by phone and although they still got it wrong, online polls definitely benefited from a degree of anonymity they provide.
Phone polls, however, suffered due to all the ridicule in the public space about the typical Trump voter. Most of the media reporting on the election showed a clear bias against Trump and therefore it should not come as a surprise that many people (shy “Trumpers”) never showed their real support publicly (in this case, talking to a pollster over the phone) or even lied about their candidate preference up until the election.
However, even though there was a difference between online and phone polls, they all showed a Clinton victory. Therefore the effect of shy Trumpers could not have been the only mistake in the polls, which leads to other possible explanations: sampling error and turnout modelling.
US pollsters have access to loads of information on which people actually do vote, allowing them to weight their samples to the profile of actual voters in a state. This method proved functional in the past, but it also brings a potential risk if there is an unexpected increase in turnout among demographics that don’t usually vote, which was probably the case in this election.
Less Interest In Participating
Another thing is the sampling error. There is a palpable decrease in the willingness to participate in polls, which makes it harder for pollsters to find a representative sample of the population out of those who are still enthusiastic about polling.
Moreover, since the rise in online and phone polling and the demise of newspaper polling, the results of polls are being skewed towards people who are well educated, pay attention to politics, and follow the political media, simply because those people living in rural areas are harder to reach.
And since most of the media in this election supported Hillary Clinton, it does not come as a big surprise that the polling samples and thus the poll themselves were skewed towards the likelihood of her victory.
To draw a complete picture it is also necessary to add that not all polls had it wrong and that a surprise is not always the result of wrong polls. Brexit was not a surprise because the polls were wrong, but because people and mainly media had not believed or paid attention to that polling evidence and thus took polls showing Remain ahead more seriously.
Also, some polls, mostly from the LA Times, correctly predicted Trump as the future president. As to why they did not fail to correctly predict the future president they said :
“Trump voters were notably less comfortable about telling a telephone pollster about their vote. Voters who backed a third-party candidate were even less comfortable responding to a poll. Women who said they backed Trump were particularly less likely to say they would be comfortable talking to a pollster about their vote.”