Testing the Mettle of the Pollsters

With just over a week to go, Sen.

With just over a week to go, Sen. Barack Obama is solidly ahead in practically every nationwide poll. And in some key battleground states, Obama’s margins are at or near double digits. But each new poll brings with it the reminder that this year’s pre-election numbers present an enormous challenge to political polling, and the ultimate question of whether the numbers are right.


How much higher will turnout be among black voters this year? What about the youth vote and newly-registered voters? These are all questions pollsters have to answer, accurately, before Election Day for their surveys to be on the money. For the pollsters themselves, it means a whole lot of pressure Nov. 4.


“Election night could be fun,” says Jeff Jones, managing editor of the Gallup Poll. “Or it could be a complete nightmare if your polls are off.” 


Jones’ poll is one of many that have been hit with some criticism this year. Some say Gallup’s likely voter model, the variable that stands to make or break pollsters this Election Day, doesn’t catch enough new voters.


It’s one of the reasons that Gallup is actually using two likely voter models—one that estimates turnout based on the model Gallup used four years ago, and another that assumes higher turnout than normal among certain segments of voters. The latest Gallup tracking poll using the traditional model gives Obama just a 5 point lead, 50 percent to 45 percent. Using Gallup’s expanded likely voter model, Obama has a slightly higher lead—52 percent to 43 percent.     


Adding to the questions are polls that show drastic differences so close to Election Day. Take the recent Pew Research poll which showed Obama with a staggering 14 point advantage. Then there's the latest AP poll that has Obama and McCain in a statistical dead heat.


On Friday the AP tried to explain some of the differences offering this: “In identifying likely voters, the AP does not build in an assumption of higher turnout by blacks or young voters.”


One of the most accurate pollsters over the course of the last three presidential cycles—Zogby International—is also accounting for what it thinks are big changes in the electorate this year compared to 4 years ago.


Zogby International's Fritz Wenzel says, “We feel really confident that we’ve made the proper adjustments in our turnout models."


So who has it right and who has it wrong? Obviously we’ll know next week, but the answer will undoubtedly be a consequential one for the polling industry.   


“For the industry as a whole, yes it’s critical,” says Nancy Mathiowetz, professor at the University of Wisconsin and past-president of the American Association for Public Opinion Research, or AAPOR. “From that perspective, I’m concerned. There are all sorts of things that might not be reflected in the current polls.”


And if the polls are off come Election Day, Mathiowetz says she hopes pollsters are willing to open up their methodology to review.


After most of the polls in New Hampshire’s Democratic primary this past January were off by double-digits, APPOR launched a probe of sorts into the discrepancies. A report with AAPOR’s findings was expected to be out this past spring, but the industry group hit a number of speed bumps.


The main problem, according to Mathiowetz, was reluctance on the part of some pollsters to provide the detailed methodological information AAPOR was asking for.


Mathiowetz told Politics magazine on Friday that the group’s report is being finalized, but likely wouldn’t be released until after the November election. No hint on the group’s findings though.  


Shane D’Aprile is web editor at Politics magazine. sdaprile@politicsmagazine.com

Share this article


Submit a comment

Required field are marked with “*”.