“Stunning” is the word CNN used to describe Donald Trump’s election night victory. With nearly all pre-election polls showing showing Clinton up by several points, FiveThirtyEight giving her a 72% chance of winning the election, and The New York Times predicting an 85% chance of a Clinton victory, ‘stunning’ was a well-chosen adjective for Tuesday’s result.
With the Clinton campaign staff diving into the post-mortem, there is one question on everyone’s minds: how did the polls get is so wrong?
The most likely explanation is that polls were tainted by social desirability bias; namely, those sampled misled pollsters due to the undesirability of being labelled a Trump supporter in many public circles. Social desirability bias is most frequently noted in what is known as the Bradley Effect. Named after an African American mayoral candidate in Los Angeles in the 80s, who lost his race despite having been up in the polls, the Bradley Effect refers to the tendency of a significant portion of voters to report being undecided or intending to vote for the non-white candidate even when that is not the case. The type of social desirability bias exhibited here is effectively the reverse; instead of pretending they might vote for the non-white candidate to avoid projecting an impression of racism, some Trump supporter may have pretended they might not vote for him to avoid projecting an impression of racism. Although those being sampled do not presume that anyone else will find out their stated preference, they may lie for fear of judgement from the pollster or because they themselves have not yet entirely come to terms with their decision or its motivations.
Another likely factor is that those who were undecided when called by pollsters, may have split overwhelmingly in favor of Trump once they had entered the voting booth. This could be explained through the public perceptions of the candidates: one a career politician, the other a political outsider. A person totally dissatisfied with both may have chosen the latter as a means of registering their disdain for the system or due to a ‘why not?’ philosophy.
It also may have come down to what all elections come down to: voter turnout. Voter turnout for Clinton was lower than expected, and voter turnout among the white, non-college-educated segment of the population, which leaned overwhelmingly toward Trump, was higher than expected. This may be attributed to lack of enthusiasm for Clinton in comparison to other Democratic nominees, the media’s portrayal of the election as a lock for Clinton, Trump’s references to election rigging, or a number of other factors.
Only one thing is clear: the polls got it wrong.