Several polls in battleground states over the last few weeks have shown gains for President Obama.

Several of these polls had many more respondents who self-identified as Democrats than Republicans. The imbalance in the respondents’ political identification caused many Republicans to complain that the polls were skewed.

Many reporters and pollsters dismissed these complaints out of hand. A few offered articles and extended explanations on why most pollsters do not try to balance the polls by party ID. Most voters probably don’t care about the specifics of polling but would like to know what basic things they should consider when evaluating polls. Let me offer four points to keep in mind.

First, and most important, polls do not represent truth. Reporters, particularly those on TV, will usually something along the lines of “poll results show candidate Jones is ahead.” This phrasing is understandable but incorrect. A poll obtains responses from a sample of the population to get an idea of what the population as a whole thinks about some issue or candidate. The results are not exact, which is why they are always reported with a margin of error. 

Second, the margin of error applies to each statistic in the results. Thus, if a poll has Jones at 45 percent and Smith at 41 percent with a margin of error of plus or minus 3 percent, it means 95 times out of 100 we would expect responses for Jones to be between 48 percent and 42 percent and for Smith to be between 44 percent and 38 percent. This is the 95 percent confidence interval. Of course, we also expect the results to be outside the margin 5 percent of the time. Sometimes those outlier polls are obvious, sometimes they are not.

Third, because polls do not represent true values, one cannot directly compare one poll with another even when they are by the same pollster. It’s even less appropriate to directly compare polls by different pollsters given the different methods and criteria they use to conduct their polls. That said, one can sometimes get a feel for a trend in a race if all the polls seem to be moving in the same direction at the same time.

Fourth, regardless of polls, turnout is the key. Pollsters are only sampling “likely voters” this close to Election Day, but who actually votes is always a big question. Pollsters use various techniques to balance the demographics of the respondents, but they have to take the person’s word that he or she plans to vote. In 2008, about 79 percent of Iowa’s registered Democrats and Republicans voted. In contrast, only about 61 percent of those registered as No Party voted. By age group, more than 80 percent of those 50 years old and above voted. At the other end of the scale, fewer than 60 percent of those 18 to 24 voted. Campaigns can outperform poll results by working hard to turn out groups that may include less reliable voters.

On the whole, polls can be useful to get a general sense of a race. In close races, however, their main value is to emphasize the need to focus on turnout.

Associate Professor Timothy Hagle