Philip Bump dealt ably with the completely faulty premise behind the Podesta "oversampling" claim on Monday. But what about the second claim? Would Trump voters be more likely to stay home if they think he's going to lose?
Here's what's going on here: Trump is basically offering another pre-excuse for losing. The polls are rigged against him, he says, and this will cause his supporters to stay home since they think the race is lost. It's a neat trick, because it allows Trump, if and when loses, to claim that he wouldn't have lost without the rigged polls. He'll say he was going to win, but then the polls discouraged his voters.
So let's imagine, just for the sake of argument, that the polls do show Trump worse off than he is in reality — whether through rigging or just plain old human error. What, then, would be the effect on turnout?
A 2015 study from Binghamton University dealt with this very subject. And it offered a mixed verdict.
Using data from the 2000, 2004, 2008 and 2012 presidential elections, the study's authors set about evaluating a series of hypotheses. Among them:
- That "turnout should be lowest when elections are uncompetitive and highest when they are competitive."
- That "the more lopsided [the] predicted election results, the lower turnout should be."
And finally, the Trump theory:
- That "lopsided polls should in hindsight be skewed against winning candidates" because "those who expect to be on the losing side ... should stay home more than those who expect their party to win."
The first two hypotheses are depicted here — with No. 1 being "Efficacy" and No. 2 being "Partisan Efficacy."
You'll notice that Hypothesis No. 2 — "Partisan efficacy" — dips more than Hypothesis No. 1 — "Efficacy" — when the party is on the losing side of the ledger. If that turned out to be true, it would lend credence to Trump's theory.
The authors decided that the best way to test these hypotheses against one another was to see whether late polls in states across the country actually showed the winner doing worse than the actual election result — the idea being that they won by more than the polls showed because their opponents' voters were discouraged and didn't turn out to vote as much.
The study did find that turnout increases with competitiveness. It also deduced that the losing candidate's voters appear to be less likely to turn out to vote.
The study's authors write that "the combination of low turnout and skewed (toward losers) polls strongly supports the argument that voters use the information from pre-election surveys not so much to determine how they will vote but rather whether to do so.... Erstwhile voters who expect their preferred candidate to lose stay at home in droves; those who expect to win also might be less inclined to go out and vote, but they do so at a greater rate than their counterparts who expect to lose."
But — and there is a but — saying bad polls hurt a candidate's turnout effort isn't the same as saying they cost the candidate the election. The study concludes, given that turnout drops off much more when the polls are much more lopsided, that this decline is unlikely to change the result of a closer race:
Were polls wildly wrong — if they predicted a lopsided victory for a candidate whose “true” support were less than the opposing candidate — our data indicate that they could conceivably manufacture a victory for the “wrong” candidate, or at least a more competitive election than either that predicted or what would have occurred in the absence of polling. More likely, however, polls get the expected competitiveness and probable winner right, and voter response to information from those same polls cements and (in the case of noncompetitive elections) magnifies their predictions.
In other words, there is data to support that idea that bad polls might turn off some Trump voters. And as the media keeps talking about how he has little chance to win, that might discourage them even more.
But since we remain in what is generally a single-digit race, unless the polls are "wildly wrong" — and there is little evidence of that — it's very unlikely that the effect would be large enough to change the result.
All of which doesn't mean Trump won't make that very argument come Nov. 9.