The Republican presidential candidate is trailing in just about every national or battleground state poll. To certain minds, this can mean only one thing, and it’s not that the Republican presidential candidate is losing the election. It’s that the polls are skewed.
Poll-skewing accusations come in different flavors. In 2012, the unusually widespread idea among conservatives was that the liberal media organizations sponsoring these polls were fiddling inappropriately with their samples and making them more Democratic. These accusations were poorly evidenced, the polls were mostly accurate, and President Obama won re-election.
The sampling wars will be inflicted upon us anew if Donald Trump is able to significantly close polling gaps to within a few percentage points. But a larger deficit of, say, 5 to 10 percentage points in national polls, with corresponding marks in the state polls, requires a blunter argument from surrogates tasked with explaining how Donald Trump Is Actually Winning. It requires an argument like, say, this one from Trump’s new campaign manager, Kellyanne Conway, herself a pollster:
“Donald Trump performs consistently better in online polling where a human being is not talking to another human being about what he or she may do in the election,” she argued. “It’s because it’s become socially desirable, if you’re a college-educated person in the United States of America, to say that you’re against Donald Trump.”
What Conway is introducing is a sort of reverse Bradley effect. The Bradley effect, named after former Los Angeles Mayor and California gubernatorial candidate Tom Bradley, is used to describe a phenomenon whereby a black candidate underperforms his or her polling on Election Day, thanks to white voters who say one thing when on the phone with a pollster and do another when in the privacy of the voting booth—i.e., they vote for the white guy. Conway is positing a different effect among college-educated voters, the segment that’s killing Trump right now: They feel ashamed to support Trump publicly because it’s not perceived as “socially desirable,” and so they won’t admit it to anyone: friends, family, strangers conducting surveys at the other end of the phone line. The closest recent analogue in electoral politics would probably be the “shy Tory.” Let us call this, then, the bashful fascist effect.
Is there any truth to Conway’s theory about polling discrepancies? Some! But it’s nowhere near large enough, at this point, to make up Trump’s deficit. If college-educated voters say they find Trump repugnant, it may be that’s really how they feel about him.
Conway claims that through her internal polling, she’s put a number on this “undercover Trump voter” phenomenon, but that she “can’t discuss it.” We, however, can put a number on it through the available public polling. The Huffington Post’s Pollster.com chart customization tool allows you to sort polling results by methodology: automated phone, interactive voice response, online, live-caller, and mixed (which is just SurveyUSA).
As of this writing, and with all polls factored into the model and under moderate smoothing, Clinton’s lead in the national polling average is 7.4 percentage points, 47.5 percent to Trump’s 40.1 percent. If you eliminate live-caller—i.e., talking to a human being who would totally judge you if you said you were supporting that lout Donald Trump—the margin narrows to 5.1 percentage points, 44.2 percent to 39.1 percent. If you count only live-caller polls, Clinton’s lead is 9.6 percentage points, 48.7 percent to 39.1 percent. So Trump slips 4.5 percentage points in live-caller polls versus all others, and 2.3 percentage points to the overall average. (Most of the big-name polls use live-caller.)
Trump “consistently performs better,” as Conway says, in the sense that the race is tighter outside of live-caller polls. But it’s not his support that changes. You’ll notice that movement is almost entirely within Clinton’s support, while Trump’s is basically the same throughout. If there’s any inter-methodology disparity here, it’s that respondents are more likely to tell a fellow human that they support Clinton than they are to tell some tricksy robot.
“Trump does do a little better in automated than live caller polls,” Tom Jensen, director of the Democratic-leaning Public Policy Polling, tells Slate in an email, “but it’s the difference between him losing nationally by 7 or 8 points or losing by 4 or 5 points—he’s losing either way and it just comes down to a degree of magnitude.” Jensen does “think it’s possible that Trump will do better on Election Day than the polls show—Republicans outperformed their poll numbers in 2014.” This doesn’t help Trump that much, though, given the way his polling numbers in critical states lag his national numbers. If “Clinton wins the Virginias and Colorados of the world by 6 or 7 points instead of 10 or 11 points,” Jensen explains, “it doesn’t make much of a big picture difference in who wins the key battleground states and who wins the White House.”
And it’s also possible that the automated and online polls are overstating Trump’s support. “They also showed Trump often winning by 30 points nationally during primary season when the popular vote in the primaries was coming out in a way that was closer to Trump winning by 10 or 15 points,” Jensen writes. It’s fun to click a Trump button online, but perhaps less so in the voting booth when the stakes are real. And they’ll be far more real when the choice is to elect a president rather than to nominate a candidate.
Conway is not entirely off in her reading of the public polls, but unless her internal numbers are far more substantial than anything the public surveys show, it doesn’t come close to patching up the massive political problem she breezily identifies while discussing polling methodology. It was astounding to hear Donald Trump’s campaign manager, in public, utter this sentence: “It’s because it’s become socially desirable, if you’re a college-educated person in the United States of America, to say that you’re against Donald Trump.”
She portrays it all as affect: secret college-educated Trump voters at their wine-and-cheese English literature parties—or, apparently, when on the phone with pollsters—tut-tutting the boorish salesman just because that’s the proper social etiquette. It may be, though, that it’s become socially desirable for college-educated people in the United States to say they’re against Donald Trump because they really are against Donald Trump, whom they don’t perceive as a desirable president. If it becomes socially unacceptable among college-educated voters to support Trump, then they’re not going to support Trump in the voting booth, not just when surveyed by a pollster.
Conway understands this. She knows that if Trump can’t seriously turn around his numbers among college-educated whites, then it’s a rout. The way to turn around these college-educated whites is in part to make it socially acceptable for them to support Trump. This explains Trump’s recent efforts to “court black and Hispanic voters,” which is mostly just an effort to signal to college-educated whites that he’s not a remarkably racist oaf. Unless the Trump campaign can change those impressions at this late, baked-in stage, it’s more the election that’s skewed against Trump than the polls.