Asking the Right Questions About the Polls

Everyone knows the polls are wrong. We just don’t all agree on why.

I don’t know anyone who thinks Donald Trump is going to be the Republican nominee for president, any more than Kang or Kodos were going to be. National parties don’t nominate cartoon characters for president. So what are we to make of his leading in early polls?

Neuroscientist Sam Wang, writing last week at TNR, has part of the answer: 20 percent isn’t really winning. At some point, some candidates will drop out, and their support will swing to another candidate. So why not anticipate that swing?

Wang suggests that pollsters ask respondents to rank all of the presidential candidates, and then we use Instant Runoff Voting to figure out who the winner is.

I generally like the idea. What is missing in most polls is a measure of this lurking support. We argued in The Party Decides that the ideal candidate was not necessarily anyone’s first choice, but someone acceptable to all factions in the party. So we need to know who else voters like.

But there are three big problems in Wang’s suggestion.

1. I think it’s impractical to ask survey respondents to rank 17 candidates, or even many of them. It’s a cognitively difficult task, and respondents, especially on the phone, will be neither willing nor able to do it.

2. More importantly, IRV isn’t how we select delegates for the conventions. Every state is different, and the rules are complicated, but they tend to use variants on winner-take-all or proportional representation. So winning 20% might actually be winning in some states. What is going to happen is that some candidates will drop out, but not necessarily the ones at the bottom, and not necessarily in an IRV-appropriate sequence. I predict Trump drops out if his numbers fade in the fall, but narcissists don’t tend to recognize when they are being rejected, so maybe not. He might linger a long time. So might others. The process of winnowing is complicated, and it involves highly attentive and involved party leaders. Trump has already noticed that the Republican establishment doesn’t like him. The other candidates, too, need to demonstrate their appeal to party leaders and to the rank-and-file. As that happens, some will quit the race, and others will surge to importance. Conducting an IRV vote now won’t foretell how that plays out.

3.Relatedly, the real issue with polls at this point is just that voters don’t know the candidates yet. Even many highly attentive activists don’t. That’s the real reason Trump’s numbers are unreliable. He has a lot of name recognition, while other candidates don’t yet. The debates may change that.

But these are quibbles. Wang is a neuroscientist not a political scientist, so it’s not fair to criticize him for not knowing about the nomination process. It’s a good idea. And a number of pollsters have already been doing it. Mark Blumenthal and the Huffington Post surveyed political activists of both parties, and they asked their second choice, as well as who is simply acceptable to them. Analyzing that data gets at the same dynamic that Wang is rightly saying we should look at.

For example, one could look for nations within the Huffington Post data, as I do in this post here.

[Cross-posted at Mischiefs of Faction]

Hans Noel

Hans Noel is an assistant professor of government at Georgetown University.