By Joe Sheller / Guest Column
In the final days before the Iowa Caucuses on Feb. 1, expect more headlines like this: “Sanders overtakes Clinton in Iowa, new poll says.”
The Gazette put that headline on a Jan. 12 report by Quad City Times writer Ed Tibbetts. Meanwhile, “Trump and Cruz Still Lead in Iowa,” according to a Jan. 11 story posted by KGAN.
The media are often criticized, with some legitimacy, I think, as treating our politics like a horse race. Many of the election news stories are driven by who is on the rise and whose poll numbers are slipping. The real question for the electorate is what Bernie Sanders or Hillary Clinton — or Donald Trump or Jeb Bush or Rick Santorum, etc. — would do as president.
Here are four points to remember when consuming news media coverage of polls:
First, not all polls are equal. A candidate will often claim that his or her own polling offers different results than what the media are reporting—and, for once, the candidates are probably being candid. Candid, however, is not the same as objective. A poll done by a PAC or candidate usually reflects the bias of the person or thing who sponsored the poll. That bias can be insidious—poll results are shaded, for instance, by how questions are worded or even the order in which they are asked.
For the Iowa Caucuses, the Iowa Poll from the Des Moines Register and the Quinnipiac Poll from Quinnipiac University in Connecticut are two relatively reputable sources of poll data. RealClearPolitics.com is a nice online source to consult if you like poll data, because it aggregates information from many sources.
Second, understand what a poll actually can show. A poll is not a projection, but rather a point-in-time snapshot of volatile, changing opinion. Pollsters aren’t trying to say what will happen Feb. 1, merely what likely caucus-goers thought on whatever days the polls were conducted. Even the best poll results have a short shelf life.
Third, keep in mind that what will happen on Feb. 1 is not a “vote.” It’s a complex network of local precinct party meetings. By its nature, the Iowa Caucuses exclude many people. If you’re not registered to vote as a Republican or Democrat, you’re not welcome. (True, the parties will allow you to declare your allegiance that night, but still, a significant segment of voters chooses to remain outside of those two political parties).
The caucus nature of the Feb. 1 party meetings requires some level of commitment from participants beyond what a vote requires. You can’t just show up, mark your ballot and go home. You have to spend several hours at a meeting if you want to be part of the process.
The parties differ, too, on how preferences are reported and what they mean.
What does this have to do with polling and poll results? Well, most poll results are qualified by stating that “likely caucus-goers” were questioned. Accurately identifying such people puts another layer of complexity on top of an already convoluted polling process, and makes Iowa polling a bit dicier to do, or report, well.
Which brings us to point four: a poll is only as valid as its methodology. The best poll has a margin of error — that is, a statistical range in which the “real” answer is highly likely to lie. A poll represents a sample that is supposed to be projectable onto a full population—and that causes some level of imprecision in the best of results.
So, for example, when the Jan. 12 story the Gazette published has language about a “5-point lead” for Bernie Sanders, it’s going astray. We don’t know from that result whether the actual lead that Sanders had was 1 percent or 9 percent, if there had been a 4-point margin of error, which is a pretty good margin. In fact, the story said the Quinnipiac poll it was reporting “has a margin of error of plus or minus 4.1 percentage points.” It doesn’t take a math genius to recognize that a 5-point lead with a 4.1-point margin of error can mean less than it seems to.
In 1948, the Chicago Tribune was partly relying on strong poll results for the Republican candidate when it erroneously reported, in early editions printed before the final vote tally in California, that “Dewey Beat Truman.” The most famous news photograph of that election shows a grinning President Truman, who has learned that those California votes ensured his continued residency on Pennsylvania Avenue, holding up that embarrassing early edition of the Tribune with the banner headline displayed.
Polls don’t really lie. But they don’t always tell the truth that news consumers might think they do. So, don’t let the poll numbers color your perception of the upcoming caucuses too much.
That would be an error with a big margin.
Joe Sheller is an associate professor of communication and journalism at Mount Mercy University in Cedar Rapids. He can be reached at jsheller@mtmercy.edu.
