Every once in a while during a political campaign, I get a little cranky:
What I had in mind is the kerfuffle over Hillary Clinton’s poor poll numbers in a recent Quinnipiac poll in several swing states. But I’d say the same thing about the Hillary Clinton’s good poll numbers in national polls pitting her against various Republicans (e.g., Jeb Bush). And I would say the same thing about the GOP primary polls too. I don’t put much stock in any of them.
Here’s the problem. We know from presidential general elections that polling this early is not correlated with outcomes in November. We know from presidential primary elections that poll surges are often ephemeral, in part because they reflect how the media covers primary candidates. For both reasons, we should discount where the candidates stand in current polls. In fact, it could easily be the case that current polls tell us almost nothing about where the presidential primary or general election will end up.
My sense is that the lessons from political science research and the history of previous campaigns just haven’t sunk in. No one wants to start a piece about a campaign by saying, “A new poll is out today, but history shows that it tells us little about who will actually win the nomination/election.”
You’re more likely to get an analysis of the poll and what it could mean, followed by a caveat about how early it is. But often the caveat basically calls into question the motivation for the analysis.
My tweets elicited some direct or apparent responses from political reporters. Here are three:
Alex Burns is absolutely correct about the GOP primary polls. To me, this shows that the debate organizers also haven’t learned from political science and history (to say nothing about what they haven’t learned from statistics). I don’t know what a better system for choosing debate participants might be. But I do know that current standing in primary polls at this moment is, at best, a weak indicator of a candidate’s viability.
(Burn and I discussed it a bit more here. As can happen even on Twitter, an apparent disagreement gave way to at least some general agreement.)
My Post colleague Aaron Blake raises a more fundamental question: how should campaign reporting deal with polls, given that they are not good predictors at this point?
I certainly wouldn’t say that campaign reporting should ignore anything that fails to predict an election 15 months away. I also watch more than the last 2 minutes of a basketball game.
But it does seem to me that the sheer volume of coverage is vastly out of proportion to the value of these polls. To use Philip Bump’s analogy, it’s as if there were thousands of news articles being written seconds into a basketball game speculating on the impact of who wins the tip-off. So, yes, there should be less coverage of early polls.
But I think there can still be worthwhile coverage of early polls. I would tend to ignore trial-heat match-ups for the general election. A Clinton-Bush or Clinton-Walker poll now doesn’t tell us much about November 2016. If those match-ups are discussed, I would pair it with more historical context that demonstrates how early polls have often been wrong and also venture to explain why they are wrong and, possibly, what direction they are likely to move.
For example, when Clinton’s favorability numbers fell in the spring, I noted that at least some of that decline was actually among Democrats but that presidential campaigns usually rally partisans — meaning it’s likely Clinton’s numbers among Democrats will turn around. During the 2012 campaign, I noted that Romney’s deficit in terms of his perceived “empathy” was real but that, historically, it has not been necessary for Republican candidates to close this deficit in order to win. Charlie Cook made a similar good point this year about negative perceptions of Clinton’s trustworthiness: certainly these aren’t good numbers for her, but we shouldn’t focus on only this one dimension.
Primary trial-heat polls are harder to ignore — especially when they inform debate participation. Here again, though, I think context is valuable. One piece of context is how fluid the numbers can be. Another is that the fluidity depends a lot on media coverage. Another is that other indicators are likely better signals of candidate strength.
There are also lots of reasons to report on polls beyond the horse race — such as figuring out the national mood and the public’s stances on key issues. See, for example, Blake on immigration or the Islamic State. Or we can report on polls of more specialized populations, like GOP activists, whose views may be more important in the primary.
So my point isn’t that we shouldn’t cover pre-election polls 15 months away from the election, it’s more that we can cover them better — in ways that truly reflect their weaknesses and strengths. But that would still require a significant shift in news coverage.