Earlier this month, I wrote a story outlining the extremely narrow but not impossible path that Evan McMullin could take to the White House, and since then, McMullin has become a genuinely hot topic. Nate Silver followed up with a post about how our forecast model is handling Utah. And we’ve talked about how McMullin’s chances of being the first third-party or independent candidate to win a state since 1968 may turn out to be one of the last cliffhanger results in this race.
Despite all the drama, there hasn’t been much polling in Utah, and polls that include McMullin remain fairly scarce. Polls that ask voters about him the same way they ask about the better-known candidates are scarcer still. In particular, three of the polls currently used by the FiveThirtyEight model have issues that could be affecting the results:
- The YouGov poll that gives Donald Trump a huge lead (and is the heaviest-weighted poll in our model) included McMullin only as an option after a respondent selected “someone else” when presented with a choice between Donald Trump and Hillary Clinton.
- The Monmouth University poll that also gave Trump a significant lead had essentially the opposite issue. It asked about McMullin on its initial list of candidates, but then when following up with undecideds, it asked only about Trump and Clinton.
This follow-up question is not itself unusual — pollsters will sometimes “push” undecided respondents by asking “if you had to vote for … who do you lean toward.” Respondents can still say they’re undecided, but many choose a candidate at this stage. (I asked Monmouth’s polling director, Patrick Murray, why the poll didn’t ask about McMullin when following up with undecided voters, and he explained that Monmouth pollsters believe that, most of the time, undecided voters break for the major-party candidates. But he also said Utah this year may be a special case.)
- The Rasmussen poll that had McMullin within 1 percentage point of Trump listed McMullin as an “Independent Conservative.” This phrasing will not be on the ballot, where he will be listed as “Unaffiliated,” but in a state as conservative as Utah, the way Rasmussen labeled McMullin could give him an advantage in that poll.
Generally, we don’t get too far into the weeds about methodological questions. Different pollsters have different methods, which may be reflected in their “house effect” (which we adjust for) and ultimately in their pollster ratings (which affects how heavily we weight them). The best protection against potential method bias is simply averaging a lot of different polls and methods, which has a strong historical track record of accuracy.
But Utah is an unusual case. Polls are scarce there, the dynamic has changed quite quickly, and these factors can significantly affect polling results. Our model already excludes Utah polling that doesn’t poll McMullin at all, and that will remain our official approach. But unofficially, we thought we’d take a look at how things might be different if we excluded all polls that didn’t treat McMullin the same way they do the major-party candidates.
Unsurprisingly, this paints a better picture of his chances. As of 3:30 p.m. Eastern time on Oct. 24, our polls-only model gave McMullin a 13.9 percent chance of winning Utah, and the now-cast gave him a 22.5 percent chance. Remove the three polls discussed above, and those numbers rise to 23.5 percent and 38.4 percent, respectively. In other words, excluding a small number of suspect polls improves McMullin’s chances in Utah by about 70 percent, which inches him ever closer to even money. Note that though the now-cast should normally be used cautiously, it might be better equipped than other versions of our model to respond to McMullin’s late surge. First, because it reacts to change much more quickly than other versions, and second because the lack of historical precedents for a candidacy like McMullin’s makes it harder to forecast what might happen between now and election day, and the now-cast bypasses that uncertainty by predicting what would happen if the election were held today.
As McMullin’s chances in Utah improve, so do the the chances of an Electoral College deadlock, though that rise is not as dramatic. In our polls-only model, the chances that no one will win a majority of the Electoral College improve from 0.5 percent1 to 0.7 percent. In the now-cast, they improve from 0.9 percent to an even 1 percent.
That brings us to the ultimate question: How much does this affect McMullin’s chances of winning the presidency? If the Electoral College deadlocks, the next president would be selected from the top three electoral-vote-getters (presumably Clinton, Trump and McMullin, barring any shenanigans) by state delegations in the incoming House of Representatives — in January. What would happen in that scenario is still entirely speculative.
But regardless: Pollsters, please include McMullin in your polls of Utah (at least) and treat him like a candidate who has a chance to win. Because he does.