Q&A: Poli-Sci Blogger John Sides

The GWU professor on what we can—and can’t—learn from early polling

The 2012 election is almost eighteen months away, but politics junkies are already being treated to polls asking if people plan to vote for Barack Obama in his re-election campaign, and testing how he fares against specific Republican opponents. What can these polls tell us so far ahead of the election, and what opportunities—or perils—do they create for reporters? To find out, CJR contributor Greg Marx recently corresponded over e-mail with John Sides, a professor of political science at George Washington University who, along with his colleagues at The Monkey Cage, was recently named “Blogger of the Year” by The Week. Their exchange is below.

Let’s start with the big picture: How much value do these polls have this far in advance? How much can they tell us about how the public views Obama or the Republican candidates—and what’s the relationship between those views now and what they might be next year?

The simplest fact about polling a presidential general election is this: early polls are much worse at forecasting the outcome than later polls. In some cases, early polls can at least be pretty good: when a popular incumbent is running for reelection, it’s not hard to forecast that he’ll win. But in other cases, it’s much more difficult to use polls to forecast outcomes. I pay relatively little attention to early polls. I am watching presidential approval and changes in macro-economic indicators.

What about the different ways of asking trial heat questions? Of course, when you pit Obama against any specific Republican opponent, how well that Republican does depends a lot on name recognition. This is why, for example, Romney tends to poll better against Obama than Pawlenty or Huntsman. And even Romney himself isn’t as well-known as he will be if he becomes the nominee. So his poll numbers will increase as well. In sum, barring some catastrophe, there’s every reason to anticipate a race that’s closer than 2008, because of the economy among other things. I don’t think we need polls to tell us this.

What are some factors that might influence early polls? Are there points in the campaign at which those factors tend to balance out, and the polls become more predictive?

Early poll numbers for the incumbent president or his party’s nominee are driven by the state of the country and approval of the incumbent. And, as I suggested before, they depend on how well-known various candidates are. As the campaign goes on, the polls do become more predictive of the outcome. Here is some forecasting research by Christopher Wlezien and Robert Erikson (pdf) that makes this point. Looking at the 1952-2000 presidential elections, they find that polls begin to be predictive of the outcome—over and above the predictive effects of the economy and presidential approval—in July of the election year, and then their impact increases with time. Obviously, on the eve of the election, the polls will typically predict the winner with great accuracy, although at that point you don’t need a poll to tell you who’s likely to win.

One of the insights of political science is that most people—and many voters—don’t follow politics very closely, especially when there’s not a major election on the horizon. At the same time, we’ve entered a period that’s often described as a “permanent campaign,” where the process of legislating takes on the characteristics of campaigns and the actual campaigns themselves seem to start earlier and earlier. Is there any reason to think this shift affects the utility of early polls—for example, by increasing the amount of information voters receive?

I haven’t seen anybody test this proposition—namely, that early polls are better forecasters now than in the past, before the “permanent campaign.” My guess is that they aren’t. I’m not sure that the public really has more information about the candidates now, despite all the early campaigning. In part, this is because the onset of a permanent campaign coincides with lots of other big political, societal, and technological changes that mitigate against the public’s becoming more informed. For example, as Markus Prior argues in Post-Broadcast Democracy, the rise of cable television and the Internet has allowed voters without much political interest to avoid news about politics.

Do you see ways in which early polls could help reporters make decisions about coverage? For example, can they tell us whether someone is a credible candidate? At the same time, can they create analytical traps for the press?

It depends on what “credible” means. One definition is “qualified to serve.” By that definition, polls are not good at all. Lots of worthy candidates never poll that well. In the 2008 Democratic primary, for example, you might think of Biden, Dodd, or Richardson.

Another definition is “likely to win” or “viable.” Polls are better at this, but not always very good in the early going. When looking at presidential primaries in particular, polls have some value—as Nate Silver has argued—but probably the best predictor of how many convention delegates a candidate will win is the number of endorsements he or she gets from party leaders. This is the conclusion of a systematic study of recent presidential primaries by Martin Cohen and colleagues in The Party Decides, where they looked at the simultaneous effects of endorsements, fundraising, news coverage, and poll numbers. Why elite endorsements matter is still a bit opaque, but if I were a reporter chasing the candidates, I’d be paying attention to the least visible parts of the “invisible primary”—the interactions and negotiations among party leaders—and less attention to the poll du jour.

Politico reported on a poll this week that had an interesting finding. By a 42-57 margin, people disapprove of Obama’s performance on the economy; respondents were also very worried about the economic outlook. That seems like it should be big trouble for the president, especially since economic conditions have a powerful influence on election outcomes. But in the same poll, a majority of voters said they approve of Obama’s overall performance, and a bigger majority said they were open to voting for him.

Does that strike you as surprising? The writeup of the poll concluded that Obama is, “at least for now—surprisingly immune to economic fears.” Would you agree?

I don’t think it’s too surprising. According to one prominent theory, articulated by John Zaller in The Nature and Origins of Mass Opinion, people answer survey questions by (unconsciously) summarizing a handful of considerations most accessible to them—that is, the consideration on the tops of their heads. As that mix of considerations changes, so might people’s answers.

At the moment, Obama’s approval is a bit higher than it had been because of the killing of Osama bin Laden. So that is likely a more salient consideration for some Americans. Thus, when the pollster calls, their overall approval of Obama reflects that consideration more than considerations regarding the economy. But when questioned directly about the economy, unemployment and gas prices become more salient considerations, and people evaluate Obama less favorably. I would expect that as bin Laden’s killing fades from the headlines, the economy will become a more salient consideration and—barring notable improvement in the economy—approval of Obama will decline again.

What else should reporters and readers keep in mind about this subject?

Every time I teach public opinion to undergraduates, I tell them never to react to a single poll. Wait two weeks, or at least until a handful of other polls come out. Then maybe you’ll know whether something has changed.

Has America ever needed a media watchdog more than now? Help us by joining CJR today.

Greg Marx is an associate editor at CJR. Follow him on Twitter @gregamarx. Tags: , , , ,