The 2016 presidential election was highly unusual, the outcome stunned many Americans (including experts), and the results will reverberate for quite some time. Two observations here: A. the outcome boils down to 5 key factors, and B. the swing-state polls were technically correct.
A. Five keys to Trump’s “surprising” victory (there are other factors, for sure).
- Soft Polling on Trump. Due to the social stigma and embarrassment that came with admitting support for Trump, it is highly likely that pollsters did not capture the full extent of his support. Additionally, polls can’t easily measure “intensity” of support either (and usually don’t try). Trump supporters were more intensely supportive of him than Hillary’s supporters were of her.
- Enthusiasm Gap. The Obama Coalition just did not come out for Hillary. Trump brought millions of previously-inactivated voters into the process who were highly enthusiastic. Evidence for this can be seen in the throngs of people who attended his rallies.
- Exit Poll Data Concerning Voters Who Highly Disliked Both Candidates: For the most part, I will leave the digging through exit poll data to others. However, one fact that sticks out here is that 1 in 5 voters highly disliked both candidates, and they broke 2 to 1 for Trump which undoubtedly helped him at the margins in close counties/states.
- Trump’s Over performance Among Rural, White Voters v. Hillary’s Under Performance in Urban Areas. Trump had huge margins (often greater than 30%) across a massive swath of rural America while Hillary could not match those margins in most urban areas.
- James Comey’s Letter to Congress 10 Days Prior to the Election: This unprecedented action clearly slowed Hillary’s momentum while Trump simultaneously became a much more disciplined candidate.
B. In spite of the oft-repeated question of “How could the polls have gotten this so wrong?” the fact is that the all-important swing-state polls were actually (technically) correct. Let’s look at some data. The table below shows the 16 swing states, the final pre-election projection average for the “expected” winner, and the outcome. As can be seen, in every case except Wisconsin, the outcome was either within the margin of error (3-4%), or was not different from the projection even though the final margin was larger than predicted. The outlier—Wisconsin—is probably due to a state-wide campaign blitz by Paul Ryan, Scott Walker, and Mike Pence in the final weekend, and the massive margins in rural areas produced by Trump supporters. In the three states that were outside the margin of error, but in which the outcome was not flipped–Iowa, Missouri, and Ohio–the electorate’s demographic is much more closely aligned with the primary Trump coalition–older, white, rural and blue-collar. Given Trump’s over-performance among this cohort, it is not surprising that these states beat the pre-election projections on totals, but not outcome.
Table 1. 2016 Swing-State Polling Data
Final Pre-Election Poll
|Trump +4.0||Trump +4.3|
|Colorado||Clinton +2.9||Clinton +2.1|
|Florida||Trump +0.02||Trump +1.3|
|Georgia||Trump +4.8||Trump +5.7|
|Iowa@@||Trump +3.0||Trump +9.6|
|Maine||Clinton +4.5||Clinton +2.7|
|Michigan**||Clinton +3.4||Trump +.03|
|Minnesota||Clinton +1.0||Clinton +1.4|
|Missouri@@||Trump 11.0||Trump +19.1|
|Nevada**||Trump +0.08||Clinton +2.4|
|New Hampshire||Clinton +0.06||Clinton +0.03|
|North Carolina||Trump +1.0||Trump +3.8|
|Ohio@@||Trump +3.5||Trump +8.6|
|Pennsylvania**||Clinton +1.9||Trump +1.2|
|Virginia||Clinton +5.0||Clinton +4.9|
|Wisconsin**||Clinton +6.5||Trump +1.0|
**Outcome flipped. @@indicates the outcome was outside the margin of error, but not flipped. Data Source: http://www.realclearpolitics.com