MANCHESTER, N.H. — Before the Iowa Republican caucuses last week, we warned you that the uncertainty in the race was high and that the polls might be way off. That’s even more true here in New Hampshire.
At least we know that Donald Trump will finish first on the Republican side? Well, probably. Trump has led all but one poll here since July; his numbers have slumped by a couple of percentage points since his second-place finish in Iowa, but nobody has come especially close to him. And yet, because the uncertainty is so great in New Hampshire, Trump’s victory is not quite assured: Our polls-plus forecast still gives Trump a 31 percent chance of somehow losing here.
The reason for this is that, historically, the more viable candidates there are in a state, the more error-prone the polling has tended to be. Multi-candidate races open up the possibility of last-minute tactical voting and otherwise give voters plenty of options, making them more likely to change their minds at the last minute. Our forecast models account for this dynamic, which is why Trump’s lead is much less safe than Bernie Sanders’s in the two-way race on the Democratic side, even though both candidates lead their nearest competitor by about the same 15-percentage-point margin in the polling average. (I think the model might be a little bit too confident about Sanders — I’d personally put his odds at more like 95 percent rather than 99 percent — but we’ll save that discussion for later.)
So let’s look at New Hampshire from the standpoint of the six leading Republican candidates. For each one, I’ve listed their 90th percentile and 10th percentile forecasts from our polls-plus model to show the most likely range of possible outcomes.
90th percentile forecast: 39 percent
10th percentile forecast: 17 percent
We’ve somehow reverted back to the pattern before Iowa, where the other Republicans weren’t spending much time attacking Trump despite his lead in the polls. In Iowa, voters took matters into their own hands and turned away from Trump at the last minute. Could the same thing happen here?
It’s entirely possible; the polls-plus model projects Trump to finish with 27 percent of the vote — a little less than the 30 percent he has in the polling average. But there’s a huge amount of uncertainty around that estimate. Suppose, for instance, that Trump finishes with a vote share in the mid-to-high 30s. Such a performance would erase many of our doubts about Trump’s ceiling and make him look formidable in South Carolina and beyond.
Conversely, if Trump won but with more like the 26 percent of the vote that Pat Buchanan got in New Hampshire in 1996, he’d look more like a factional candidate who was benefiting from the divided field. And if Trump’s vote share falls into the low 20s or even the high teens, meanwhile, he would be vulnerable to losing New Hampshire outright. It would also raise a lot of questions about whether the polls were oversampling Trump voters.
90th percentile forecast: 25 percent
10th percentile forecast: 9 percent
The polls-plus forecast still has Rubio in second place, but that’s deceptive. He’s only a fraction of a percentage point ahead of John Kasich, with Jeb Bush and Ted Cruz lurking almost as close. In fact, the polls-plus forecast has Rubio with a 64 percent chance of finishing third or worse and he could easily enough slip to fifth or sixth. Here’s our final matrix of probabilities for Rubio and the other candidates: