What the Polls Are Missing About Trump
Donald Trump “needs a miracle” to win, says Stuart Rothenberg, an old-style political analyst. On the other hand, political scientist Drew Linzer estimates Trump’s odds at about 1 in 4, or as he puts it: “Flip a coin twice: if you get two heads, that’s President Trump.” Nate Silver’s “polls-plus” model has Trump with a 25 percent chance, too.
Linzer’s and Silver's estimates are somewhat generous to Trump compared with those of some other forecasters, but none of them is in the "needs a miracle" range.
Here's how to read what polling forecasters say -- and why their estimates can't fully explain how difficult a Trump comeback might be.
Predictions such as those at FiveThirtyEight basically do two things. First, they construct averages from all the polls to estimate where the contest is right now. Then they adjust for how things normally change given the time remaining before the election.
These predictions could go wrong if the polls themselves are systematically wrong for some reason. This can happen, although they are unlikely to be off by more than a percentage point or two, and that can be in either direction.
The predictions can be wrong in a different way, even if the polls give an accurate snapshot of current voting intentions. They might not be able to account for factors that aren't usually present between now until Election Day.
Here are three ways in which Trump is different, all of which would appear to work against him:
- His inability to control his mouth. Normally, a candidate's gaffes turn out to be unimportant; they may dominate the news for a day or two, but then go away. Trump has proved that he can't stop himself from making wild and controversial statements, and then he litigates them on cable news and at his rallies for days after. Some effects from this are already reflected in current polls and partially account for his abysmal favorable-unfavorable numbers. Normally we could expect some voters to return to him after the short-term fallout fades. Not so with Trump so far in the general-election campaign.
- His lack of a proper campaign. This hurdle is already incorporated in the models since part of Hillary Clinton’s lead is presumably due to Trump's lack of response to her ads against him. Other consequences could include the impact of Clinton’s professional get-out-the-vote operation compared with Trump’s failure to launch one (he’s relying on formal party organizations to do it for him). This could cost Trump a percentage point or two that may never show up in the polls before Election Day.
- His failure to unify the party. It guarantees that leading Republicans will condemn whatever outrageous things he says, thus sending mixed messages to Republican voters.
It’s impossible to estimate the magnitude of these three effects. But if current poll-based predictions give Clinton a 75 percent to 85 percent chance of winning, then adding in the rest of what we know probably pushes Trump up to the “needs a miracle” mark.
Well-constructed polling averages reduce the uncertainty that any individual poll has, even very good ones. But new uncertainty is introduced in the time between now and the election. A good projection model not only gives a reliable estimate for where the election will wind up, but also provides an idea of how much uncertainty goes along with that prediction.
That is, some voters may not like Trump because he is always saying outrageous things. They already dislike him enough that they are unlikely to dislike him more for what he says next week or next month. But some voters who still like him at this point, despite what they've heard him say so far, could turn against him if he continues to behave in the same way.
This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.
To contact the author of this story:
Jonathan Bernstein at email@example.com
To contact the editor responsible for this story:
Katy Roberts at firstname.lastname@example.org