After Donald Trump’s political-world-shattering upset of Hillary Clinton, the polling industry finds itself facing an existential crisis. A vast majority of the key polls were not just wrong, they were humiliatingly wrong. Though a very select few — most notably the LA Times, IBD/TIPP, and Trafalgar Group — actually got it about right, most pollsters ended up grossly over-sampling Democrats and failing to account for Trump’s “hidden” supporters. Below is a comparison of the final polling data going into the election Tuesday morning and the final results, which often ended up looking quite different.
On the morning of the election, Real Clear Politics’ average of the national polls showed Clinton with a 3.3% national lead over Trump and a projected Electoral College victory of 272 to 266. Instead, Clinton ended up with only a 2.2% popular vote advantage and suffering a devastating 227 to 304 loss. At no point before the election did the state poll averages show Trump winning the necessary 270 electoral votes.
Below is Real Clear Politic’s No Toss Ups Map, based on the average of the state polls as of election day morning, projecting that Clinton would beat Trump 272 to 266:
Though the poll averages showed Clinton winning Pennsylvania, Michigan, and Wisconsin, Trump ended up winning all three, outperforming projections by 3 points, 4.4 points, and a stunning 7.5 points, respectively. Here is how the electoral map actually shaped up: Trump over Clinton, 306 to 232 (Michigan going to Trump and NH to Clinton):
Here is the New York Times‘ graph of their projected electoral vote count, which dramatically shifted over the course of the evening, eventually reaching a 306 to 232 win for Trump:
BATTLEGROUND STATES
Below are RCP’s poll averages for sixteen battleground states going into election day, followed by the actual election results. Some of the averages ended up projecting the winner relatively closely, but others were drastically off (including Wisconsin, Iowa, Ohio and Michigan). Among the polls used to produce the averages, many were even more wildly divergent from the actual vote, far beyond the margin of error. A majority of the surveys over-projected for Clinton and overestimated third party candidate votes. The states are listed from the most inaccurate to the most accurate projections.
WISCONSIN – off by 7+
- Polls: Clinton +6.5 (46.8 – 40.3)
- Results: TRUMP +1 (48 – 47)
IOWA – off by 7
- Polls: Trump +3 (44.3 – 41.3)
- Results: Trump +10 (52 – 42)
OHIO – off by 4+
- Polls: Trump +3.5 (45.8 – 42.3)
- Result: Trump +8 (52 – 44)
MICHIGAN – off by 4+
- Polls: Clinton +3.4 (45.4 – 42)
- Results: TRUMP +1 (48 – 47)
MINNESOTA – off by 3
- Polls: Clinton +5 (45.3 – 40.3)
- Results: Clinton +2 (47 – 45)
NEVADA – off by 3
- Polls: Trump +0.8 (45.8 – 45)
- Results: CLINTON +2 (48 – 46)
NEW MEXICO – off by 3
- Polls: Clinton +5 (45.3 – 40.3)
- Results: Clinton +8 (48 – 45)
NORTH CAROLINA – off by 3
- Polls: Trump +1 (46.5 – 45.5)
- Results: Trump +4 (51 – 47)
PENNSYLVANIA – off by 3
- Polls: Clinton +1.9 (46.2 – 44.3)
- Result: TRUMP +1 (49 – 48)
MAINE – off by 1+
- Polls: Clinton +4.5 (44 – 39.5)
- Resutls: Clinton +3 (48 – 45)
ARIZONA – off by 1
- Polls: Trump +4 (46.3 – 42.3)
- Results: Trump +5 (50 – 45)
COLORADO – off by 1
- Polls: Clinton +2.9 (43.3 – 40.4)
- Results: Clinton +2 (47 – 45)
FLORIDA – close
- Polls: Trump +0.2 (46.6 – 46.4)
- Result: Trump +1 (49 – 48)
GEORGIA – close
- Polls: Trump +4.6 (48.4 – 43.8)
- Results: Trump +5 (51 – 46)
NEW HAMPSHIRE – close
- Polls: Clinton +0.6 (43.3 – 42.7)
- Results: Clinton +1 (48 – 47)
VIRGINIA – correct
- Polls: Clinton +5 (47.3 – 42.3)
- Results: Clinton +5 (50 – 45)
NATIONAL POLLS
The current total popular vote count shows Clinton with just a 0.2% advantage. Heading into election day morning, RCP’s national average for the four-candidate race gave Clinton a 3.3% advantage (45.5 – 42.2). Ten of RCP’s eleven featured polls showed Clinton ahead, her largest lead 6 points (Monmouth). Only one of the eleven polls, IBD/TIPP, gave Trump an advantage (+2 points):
RCP’s head-to-head survey average showed Clinton holding an almost identical 3.2-point lead (46.8 to 43.6). Clinton led in nine of RCP’s ten featured polls, her largest lead 6 points (Monmouth). Trump led in only one of the ten polls: the LA Times/USC tracking poll (+3 points), which consistently showed more favorable numbers for the Republican than other polls, a result the pollsters attribute to the survey including a “bloc of disaffected [Trump] voters” ignored by others. IBD/TIPP provided the closest prediction to the actual results: Clinton +1.
BETTING ODDS
On the morning of the election, the average of the betting odds heavily favored Clinton, 81.6% to 18%. Like many other forecasts, the New York Times began its election day projecting Clinton with an over 80% chance of winning. But here’s how the percentages changed when the results started coming in:
After months of skewed data from pollsters and all the failed forecasts, it’s hard to imagine the American people taking the polls quite as seriously in the near future. Trump ended up being right about the “biased” polls. When in doubt, trust what you see with your own eyes.
This article has been updated to include the final Electoral College and national vote data.