As we sift through the rubble of Dataworld today, can we find any survivors? I think there are three, although the success of each comes with qualifiers.
First up, the Trumpiest poll in all the land: It’s the LA Times survey, which took a victory lap this morning for predicting a decisive Trump win. They deserve credit for sticking to their guns when the entire polling world was screaming at them that they’re wrong.
For most of the last four months, the USC/L.A. Times Daybreak tracking poll has been the great outlier of the 2016 campaign — consistently showing a better result for Donald Trump than other surveys did.
In light of Tuesday’s election returns, the poll now looks like the only major survey to see the wave coming.
The poll’s findings caused dismay — even outrage — among some readers, especially Democrats, who have denounced it and often criticized The Times for running it.
But just as four years ago the poll was one of the few that did not underestimate President Obama’s support, it seems as though it may have been on its own in not underestimating Trump’s.
That’s correct, by the way. Most pollsters predicted a very tight election between Obama and Romney in 2012. The USC poll at the time, which became the LA Times poll, predicted a more comfortable Obama win. Now they’re two for two, again after flouting a strong consensus in the industry. That’s impressive. The unusual methodology of the poll, in which subsamples from a set group of 3,000 people are continually polled and re-polled over the course of months, will obviously be closely studied and emulated by other pollsters in coming election cycles. Two important notes, though. One: David Lauter, who wrote the “victory lap” piece excerpted above, was himself sufficiently skeptical of the numbers that he went all in on predicting a Hillary landslide a few days ago. Even those who knew the poll best, in other words, didn’t trust it in the end. Two: Although the Times nailed Trump’s share of the popular vote, predicting 46.8 percent versus the 47.5 percent Trump got in reality, it missed badly on Clinton’s share. The LAT survey expected her to pull 43.6 percent of the vote; she ended up with 47.7 percent at last check, ahead of Trump. That is to say, although the Times poll pointed to a decisive Trump win in the electoral college, it actually missed the winner of the popular vote entirely, by more than four points.
Which brings us to contender number two: The IBD/TIPP poll. IBD likes to boast that they’ve been the most accurate presidential pollster over the course of the last four cycles. Last night was another good night for them. They were one of the very few to see Trump winning on Election Day, putting him up 45/43 in the four-way race in their final tracker. That’s a slightly tighter margin than the LA Times poll, anticipating the very tight race that we saw. Interestingly, IBD/TIPP had Clinton ahead in the two-way race, 43/42, a hint of her eventual victory in the popular vote. Putting those numbers together, the takeaway from the poll on Monday was that the election would be very close and actually could go either way depending on how undecideds and the third-party vote shook out — which, broadly speaking, is what happened. They missed, however, on Clinton’s actual share of the vote about as badly as the LA Times did. In fact, the only national pollster to have Clinton and Trump at 46 percent of the vote or better in the two-way race was the vaunted ABC/WaPo tracker, which had called every election since 1992 very accurately. ABC/WaPo doesn’t get a gold star either, though, as they saw a 49/46 race developing, a margin that almost certainly would have made Clinton president if it had happened in reality. They were the closest to getting both candidates right, but they got the outcome wrong. And they missed Trump’s support in the four-way race, predicting a 47/43 Clinton win when what we got was a 47/47 tie.
Contender number three, then: Trafalgar Group, a Republican polling firm. I’m not familiar with them, but respect must be paid to the polls of Pennsylvania and Michigan they dropped on the eve of the election. Until this week, Trump had led in exactly one poll of Pennsylvania all year, a Quinnipiac survey conducted in July before the conventions. He had never led a poll in Michigan — and in fact, had never been so much as within three points in the state. Trafalgar’s poll of PA, released Monday had it Trump 48.4, Clinton 46.5. The actual result last night: Trump 48.8, Clinton 47.7. Pretty good. Their final poll of MI had it Trump 48.5, Clinton 46.8. Actual result: Trump 47.6, Clinton 47.3. Pretty good. And if that’s not enough, here’s the map they dropped yesterday, before results started streaming in:
— Trafalgar Group (@trfgrp) November 8, 2016
That wasn’t a direct hit — Clinton won Nevada, lost Wisconsin, and is barely ahead in New Hampshire today — but predicting that Florida and North Carolina would go red and that Trump would flip nearly the entire Rust Belt when virtually no other data-head would gamble on that? Prettttttty good. You can knock them if you like for not having done a national poll, but those hits in PA and MI are probably the most outstanding calls of the year. Hats off to them.
Two last tidbits for you on polling this year. Strange as it may be, this is actually true:
National polls will wind up being **more accurate** than they were in 2012:
2012: Obama up 1, won by 4
2014: Clinton up 3-4, will win by 1-2
— Nate Silver (@NateSilver538) November 9, 2016
It won’t be remembered that way since most pollsters predicted the winner correctly in 2012 but got the margin wrong whereas they got closer to the margin this year but got the winner wrong, but it is what it is. And finally, per Newsweek, it turns out Trump led the RCP average of national polls this year for a grand total of … 192 hours. He was very slightly ahead for three days in May then slightly ahead for three days in July after the GOP convention. And that’s it. As the saying goes, the only poll that counts is on Election Day.