With only two weeks left till the 2016 Presidential Election, the majority of presidential polls show a consistent lead for Hillary Clinton over Donald Trump at both the national level and in key battleground states. Given historically that polls get more accurate as we get closer to Election Day and the size of the leads in the polls (as high as 12 points in a recent Monmouth poll), is the election over?
Before we start scheduling a Hillary Clinton inauguration, there is a critical factor that could betray the polls, and it is important to keep in mind: turnout.
Every poll has an underlying "likely voter" model. They get far less attention than the top line horse race results, or even the underlying demographic breakdowns. Most pollsters ask a series of questions that they use to elicit how likely it is a given voter will actually go to the polls on Election Day. These questions can range from demographic based (such as age) to past voting behavior to how fervently a respondent says they will support a candidate.
Almost all major polls are now reporting results of "likely voters." But, what happens if the likely voter models are wrong?
In 2012, the final Gallup poll has Mitt Romney leading Barack Obama 50-49. The final result was a 4 point win and electoral college majority for Barack Obama. One of the culprits? The methodology Gallup used to determine likely voters. Although Gallup used similar overall methods as other polling organizations in determining who would turn out on Election Day, one particular question skewed their results. As explained by the Gallup Organization at the time in their post-mortem of what went wrong with their polling:
The more significant variable in Gallup's model that pushed the vote share more toward Romney was the "thought given to the election" variable -- whose removal, along with other changes in the way the likely voter figures were calculated, would have made Gallup's final estimate more similar to those of other firms.
Most pollsters, polling aggregators, and betting markets read the polls as indicating a sizeable Electoral College victory for Hillary Clinton. But, it is important to keep in mind that the only poll that counts is the one that happens on November 8th. The result will be critically shaped by which "likely voters" become "actual voters."
Gallup 2012 Presidential Election Polling Review, p. 4.