A group of major pollsters just released an autopsy report to explain why the polls were such a disaster in 2016
The analysis, released Thursday by the American Association for Public Opinion Research, found that the biggest culprit was state-level polling underestimating the level of Trump's support, most importantly in Pennsylvania, Michigan, and Wisconsin.
There were a number of reasons for the discrepancy, the study concluded.
The big three were a substantive change in vote preference during the campaigns final days, a failure to properly adjust for an overrepresentation of college graduates, and many Trump voters failing to reveal their preferences until after the election. The last point could have also been the result of late-deciding Trump voters, the report said.
On the first point, the autopsy found that 13% of voters in Wisconsin, Florida, and Pennsylvania - three states Trump won narrowly over Clinton and helped catapult him to an unlikely victory - decided on their presidential choice in the final week of the election. Overwhelmingly, those voters broke for Trump: In Wisconsin, they chose Trump over Clinton by a 30-point margin. In Florida and Pennsylvania, the margin was 17 percentage points.
The analysis makes note of a couple of other points - namely, that a change in turnout from 2012 to 2016 likely also caused state polls to mis the results. It also said that "ballot order effects" may have effected some state contests. Trump appeared above Clinton on the ballot in several key states, such as Michigan, Wisconsin, and Florida. The study estimated that about one-third of a percentage point bump can be attributed to appearing first on the ballot. The study said that could, at best, be a very minor explanation for "polling problems," however.
"Collectively, well-resourced survey organizations might have enough common interest in financing some high quality state-level polls so as to reduce the likelihood of another black eye for the profession," the autopsy report said.
Reminiscing on what went wrong just days after the November election, Patrick Murray, the head of Monmouth University's polling institute, a firm that conducts a number of state polls, told Business Insider "polls might not be capable of predicting elections."
Murray's final Pennsylvania poll showed Clinton with a 4-point lead with a 4.9-point margin of error, which still was not big enough to capture the margin - 1.2 points - by which Trump would win the state.
His theory for what happened at the time: "Non-response among a major core of Trump voters."
The Thursday analysis featured pollsters from Pew, SurveyMonkey, The Washington Post, Gallup, Marist, YouGov, among others. You can read the full report here.