First, Donald Trump has turned the tables on American politics. He broke all the rules, made every possible mistake and still won. He transcended campaign tactics because he represented something bigger to many voters, and that was the opportunity for a revolt, a flushing out of the system.
An old friend and colleague sent me a note this morning which best captured the moment: “The people have taken the Establishment by the lapels, given it a good throttling and shouted: ‘WE’RE the boss, remember?!’”
Second, Trump’s victory was stunning, but his ability to bring out voters who helped Republicans down ballot was equally impressive. For months, the media narrative was: Will Trump sink Congressional Republicans? In the end, he saved a lot of them.
The GOP had great exposure in the Senate elections going into 2016, with 24 seats on the block while Democrats only had 10 seats at stake. At one time or another, a slew of Republican incumbents looked like sitting ducks––Portman in Ohio, Burr in North Carolina, Toomey in Pennsylvania, Kirk in Illinois, Rubio in Florida, Blunt in Missouri, Ayotte in New Hampshire and Johnson in Wisconsin. But, in the end, only two lost. The same goes for the House, GOP losses were much smaller than either side expected, despite a large number of vulnerable freshmen members.
Third, who would have thought Clinton would join the ranks of Al Gore, Samuel J. Tilden and Grover Cleveland––losers who won the popular vote but lost the Electoral College? Based on the latest returns, Trump has at least 289 electoral votes (more than the needed 270), but Clinton is leading him by 219,347 votes in the popular vote. If this holds up, Democrats will have won the popular vote in six of the last seven presidential elections. While that fact is historically interesting, it’s not much of a consolation prize for today’s shocked Democrats.
Fourth, public polling was often a mess––but not always wrong. You could see that in this newsletter for weeks, where numbers fluctuated and diverged all over the place without, seemingly, much reason.
But, before we rush to conclusions, let’s look at what actually happened:
NATIONAL POLLING: Trump ended up winning a race that looks quite different in actual vote totals than the national polls were describing at the end. The final national polling average we calculated in this newsletter on the basis of 13 polls showed Hillary Clinton with a 3.5-point popular vote advantage. But, her actual popular vote lead, based on current numbers, is about two-tenths of a point––more than 3 points off the average.
Interestingly, the national polling average that we published Thursday before the election showed Clinton leading the popular vote by only half a point, which turned out to be much closer to the actual results than the final averages on Monday and Tuesday. That raises a lot of questions about why the polls missed what appears to have been a late-breaking Trump surge. We will discuss those issues in future editions of this newsletter.
Before we look at individual national polls, keep in mind that they only measure the nationwide popular vote. Twelve of the final 13 reported polls showed Clinton winning that vote––and she did, in fact, win it. The problem was that the polls missed the size of Clinton’s actual margin and, in some cases, they missed it by an embarrassingly wide mark. The fact that Trump ultimately won the presidency, even though he did it based on state electoral votes and not on national popular vote totals, makes the polls seem even worse.
Of the 13 final national polls, only the IBD/TIPP poll had Trump ahead, and that was by 2 points in the 4-way trial heat. In terms of comparing final poll results to the actual vote count, three polls came closest: IBD/TIPP, Rasmussen and McClatchy/Marist. They were all roughly 2 points off the outcome. But, because IBD/TIPP was the only reported poll that put Trump ahead in the popular vote, they will get the accolades for the best performance––even though Trump actually lost the popular vote. Rasmussen and McClatchy/Marist may turn out to have been slightly closer to the actual popular vote margin than was IBD/TIPP.
The two polls that gave Clinton her biggest vote leads were the Monmouth Univ. poll and the NBC/Survey Monkey poll, which both had Clinton ahead by 6 points. Five polls had Clinton ahead by 4 points: NBC/WSJ, ABC/Washington Post, CBS, Gravis and Economist/YouGov. Three polls had Clinton ahead by 3 points: Bloomberg, Morning Consult and Reuters/Ipsos.
Finally, there is the matter of the LA Times/USC election forecast, which is based on methodology different from that used by other pollsters. On Monday, they showed Trump ahead by 5 points and on Tuesday they showed him up by 3 points. While one can say that their forecast correctly caught the pro-change mood of the electorate that led to Trump’s winning the presidency, their final number was still about 3 points off the actual popular vote count.
On polling methodology: Clinton led by an average of 3 points in the eight final national polls conducted by live phone interviews via landline and cell (NBC/WSJ, Fox News, CBS, ABC/WP, IBD/TIPP, McClatchy/Marist, Monmouth Univ., Bloomberg). She led by an average of 4 points in the four final polls conducted by online interviewing (Reuters/Ipsos, Morning Consult, Economist/YouGov, NBC/Survey Monkey). And, she led by an average of 3 points the two final polls conducted by automated calling (IVR) supplemented by online interviews (Gravis, Rasmussen).
STATE POLLING: The day before the election, we said in this newsletter that, “A third of today’s state polls are wrong. The problem is that we don’t know which third.” Now we know.
Here are examples in key states:
Wisconsin: Every poll was off. Remington (R) had Clinton ahead by 8, Loras and Marquette Univ. had Clinton ahead by 6. Trump won the state by a point.
Florida: Most polls showed a close race, and the actual result was close, but only a GOP poll had Trump ahead. Opinion Savvy had Clinton ahead by 2, Quinnipiac and Gravis had her up by 1, CBS/YouGov had it a tie and Trafalgar (R) had Trump ahead by 4. Trump won the state by about 1.4 points.
Michigan: Every poll had Clinton ahead. Mitchell and PPP(D) had Clinton up by 5 and the Detroit Free Press had her up by 4. Trump is leading the state by three-tenths of a point.
Pennsylvania: Morning Call/Muhlenberg College had Clinton ahead by 4, Harper (R) had it a tie and Trafalgar (R) had Trump ahead by 1. Trump won the state by about a point.
North Carolina: The polls showed it tight, but missed the late break for Trump. NYT/Siena had it a tie, Quinnipiac had Clinton ahead by 2 and Gravis had Clinton ahead by 1. Trump won the state by about 3.8 points.
Iowa: Only the Des Moines Register poll caught the extent of Trump’s strength. Loras had Clinton ahead by 1, Quinnipiac had it a tie, Emerson had Trump ahead by 3 and the Des Moines Register poll put Trump in the lead by 7. Trump won the state by more than 9 points.
Ohio: Emerson had Trump ahead by 7, CBS/YouGov had him ahead by 1 and the Columbus Dispatch had Clinton up by a point. Trump won the state by more than 8 points.
Arizona: Three polls (NBC/WSJ/Marist, CNN, Emerson) taken at least a week before the election had Trump ahead 4-5 points. Trump is now leading the state by about 4 points.
New Hampshire: The final polls bounced all over. The WMUR/UNH poll had Clinton ahead by 11, Emerson had her up by 1, the Boston Globe/Suffolk and UMass Lowell had it a tie, ARG had Trump ahead by 5 and Gravis showed Trump in the lead by 2. Clinton is leading in actual vote count by two-tenths of a point.
Virginia: All polls pointed to a Clinton victory. Christopher Newport Univ. had Clinton ahead by 6, PPP (D) and Gravis had her ahead by 5 and Remington (R) had her up by 2 and Roanoke College gave her a 7-point lead. Clinton won the state by 4.9 points
There you have it. About a third of the state polls came very close, about a third were way off and the rest were somewhere in the middle.
BOTTOM LINE: In terms of public polls that are sponsored and covered by the media––which is the focus of this entire analysis––there needs to be a major reassessment that looks at polling costs, methodology and quality in an era of technology change. There is also a need to look at the amount of attention paid to polls as a news source and the often erroneous presentation of polls as a predictive tool.
OVERVIEW OF EXIT POLL RESULTS
Among voters nationwide
Male: Trump +12 (53–41)
Female: Clinton +12 (54–42)
18-29: Clinton +18 (55–37)
30-44: Clinton +8 (50–42)
45-64: Trump +9 (53–44)
65+: Trump +8 (53–45)
White: Trump +21 (58–37)
Black: Clinton +80 (88–8)
Latino: Clinton +36 (65–29)
Asian: Clinton +36 (65–29)
White men: Trump +32 (63–31)
White women: Trump +10 (53–43)
College graduate: Clinton +9 (52–43)
Non college educated: Trump +8 (52–44)
White college graduates: Trump +4 (49–45)
Whites no degree: Trump +39 (67–28)
Under $50K: Clinton +11 (52–41)
$50K to $100K: Trump +4 (50–46)
$100K or more: Trump +1 (48–47)
Democrats: Clinton +80 (89–9)
Republicans: Trump +83 (90–7)
Independents: Trump +6 (48–42)
Married: Trump +10 (53–43)
Unmarried: Clinton +17 (55–38)
Married men: Trump +21 (58–37)
Married women: Clinton +2 (49–47)
Unmarried men: Clinton +1 (46–45)
Unmarried women: Clinton +29 (62–33)
Protestant: Trump +19 (58–39)
Catholic: Trump +7 (52–45)
Jewish: Clinton +47 (71–24)
No religion: Clinton +42 (68–26)
Union household: Clinton +8 (51–43)
Nonunion household: Trump +3 (49–46)
Veteran: Trump +27 (61–34)
Non-veteran: Clinton +5 (50–45)
by Ron Faucheux, President of Clarus Research Group, MSLGROUP’s full service polling and research company
Cover image: Mark Newman, Creative Commons license