Who’s Gonna Win?

We all know about horse race polls. If the election were held today, they ask, who would you vote for? But sometimes even more accurate than asking “who ya for” is another survey question — “Who’s gonna win?”

Yup, there’s a difference.

One question measures personal preference, the other asks for a prediction. There is a huge overlap, of course, as people tend to believe their own candidate will win. Three-fourths of Mitt Romney supporters, for example, predicted he’d win in 2012, but one-quarter predicted, accurately, he’d lose. It’s those folks who provide the most useful information. They sense the opinion climate around them and make a prediction at odds, despite the deep psychological drive not to do so, with their personal preference.

Sticking with 2012 for the moment to illustrate, every survey that asked some form of the “who’s gonna win” question, from April through November, had Barack Obama ahead. That’s 14 data points, best I can find, by three survey or news orgs (CNN/ORC, Pew, and ABC/Washington Post).

Over that time Romney never got closer than 14 percentage points on the prediction question and often the gap was much wider. By November, according to CNN/ORC, 59 percent predicted an Obama win and 34 percent predicted a Romney victory. OK, you think, that’s not so great given the actual popular vote was much closer, 51.1 percent for Obama, 47.2 percent for Romney. But CNN/ORC’s preference question in the same poll had it 47/47, so while the margin was off, the direction was better in the “who’s gonna win” question.

Even in the celebrated (or infamous) 2000 election, a Pew prediction question called it right, at least in the ultimate outcome. In that November 2000 survey, 43 percent predicted George W. Bush would win, while 32 percent predicted an Al Gore victory. That same poll’s preference question was flipped, 45–41 for Gore. In other words, the “who’s gonna win” outperformed the “who are ya for” question, even in a close election. So somehow the prediction question better taps something unique, perhaps people’s ability to channel hanging chads before they ever hung, or a U.S. Supreme Court before it ever got involved. Or more likely, the sense of the opinion climate measured by the prediction question somehow, perhaps magically, got to the root of the election and it’s state-by-state results.

So here we are in 2016, with Donald Trump and Hillary Clinton the presumptive nominees. A March poll by CNN/ORC was the first I could find this election cycle to ask the prediction question (scroll down to Q26). That poll has Clinton ahead, 56–42, on the prediction question, and Clinton ahead 51–41 on preference. These questions came long before Trump’s bump and his subsequent problems with certain “Mexican” judges who were actually born in Indiana, so of course the preference numbers have tightened to within the margin of error. No one has yet asked another prediction question, so that will be interesting once it happens. Keep your eyes out, especially from the three survey shops listed above, as they traditionally ask this question a few times during the campaign.