Contrary to popular myths, the national political polls in 2016 were very accurate. According to the Real Clear Politics average, Hillary Clinton was projected to win the popular vote by 3.3 percentage points and she actually won it by 2.1 percentage points. Ten of the last twelve national polls released were within two percentage points of the actual margin. One of the others overestimated Clinton’s margin by four points and one underestimated it by four points.
That’s about as accurate a projection as you could hope for!
However, while the polling was good, the analysis of the polling was not.
Many in the media and political worlds simply could not imagine a Clinton loss. Some looked at the polling, noted the margin of error, and assumed the Democratic nominee would win by far more than three percentage points. Some thought she could win a few traditionally Republican states while few imagined the reverse could happen.
The problem was not the polls, but the analysis.
This failure of polling analysis was not supported by the underlying data. Heading into Election Day, the Real Clear Politics projections showed that Clinton was clearly favored to win just 203 of the needed 270 Electoral College votes. Donald Trump was favored to win 164 and an astounding 171 were in the Toss-Up category. Reviewing the toss-up states at that time, I noted publicly that it was fairly easy to envision how Trump could reach 263 Electoral College votes.
That reality should have dented the overwhelming confidence of a Clinton victory expressed in the media and political worlds. Polling data showed a race close enough that a surprise in a single traditionally Democratic state could elect Donald Trump. But that possibility was largely ignored in media coverage of the race.
On Election Day, of course, then-candidate Trump stunned the political world by capturing three traditionally Democratic states: Pennsylvania, Michigan, and Wisconsin. Once again, the surprise was more a failure of analysis than public polling. In Pennsylvania, the polling averages showed Clinton ahead by just 1.9 percentage points. In Michigan, she was up just 3.4 points. Both results were clearly in toss-up range. Additionally, in both states, the final public poll released showed Trump ahead.
The only true polling misfire was in Wisconsin, where Clinton was projected to win by 6.5 percentage points and Trump won by just under a point.
The bottom line is that the actual public polling in 2016 was far better than the pundit’s analysis of that polling data.