The poll results were averaged except for the few cases where no polling had been conducted since October 1--in those states, the most recent poll was used as the RCP 'average'. There was no polling data available at all in four states. They are colored in black on the map. The rest of the states are colored according to how greatly Trump exceeded/fell short of polling expectations, the darker the greater his over-/under-performance. The five states where the RCP average was within half a percentage point of the actual results are shaded in an off-white color.
|No data for|
One thing that immediately jumps out is the tendency for the winner of non-competitive states to significantly exceed polling expectations (ie Trump doing even better than the lopsided polls suggested in dark red states like West Virginia and Oklahoma and Clinton doing better than the polls predicted in dark blue states like California and Illinois). The correlation between the margin of Trump's victory/defeat and his over-/under-performance relative to the polls is an impressive .63 (p-value = .0000002).
Another thing that astute commentators noted from the beginning but that the polling agencies never fully picked up on was the nationwide east-west divide. The dynamics in play here are the rootedness, manufacturing, and populism of the east versus the transplanted, service-oriented, libertarianism of the west. Trump did better than expected in the east and worse than expected in the west.
A few comments on where the polls went especially wrong:
- They drastically underestimated how the subjects of Hillbilly Elegy would come out in droves for Trump. These are the forgotten Americans he spoke about again and again during the campaign. They were as forgotten by the pollsters as they are by the political class.
- Trump was, save for
- Hispanics--illegal or otherwise--responded to the threat that the reconquista could be coming to an end by shaking off a bit of their apathy to vote at higher-than-expected rates.
Parenthetically, the exit polls should be viewed with the same skepticism that the pre-election polls were correctly viewed with. It's highly unlikely that if blacks dropped from 13% of the electorate in 2012 to 12% in 2016 that the percentage of voters with no more than a high school education dropped from 24% in 2012 to just 18% in 2016, even though that's what the exit polls--which initially indicated a comfortable Clinton win before being sloppily adjusted in light of the actual results--show. We'll have to wait for the Census and eventually the GSS for more precise and reliable data.