Polling, and how people pick their party

This article was originally on a blog post platform and may be missing photos, graphics or links. See About archive blog posts.

Today’s poll story showing Barack Obama with a 12-point lead over John McCain drew almost as much reaction from readers for what it had to say about voters’ party affiliation as for what it said about the candidates’ popularity.

The questions centered on this passage in the story: ‘In this national poll’s random sample of voters, 39% identified themselves as Democrats, 22% as Republicans and 27% as independents. In a similar poll a year ago, 33% identified themselves as Democrats, 28% as Republicans and 30% as independents.’

Susan Pinkus, director of the Times Poll, and Doyle McManus, Washington bureau chief, respond to the readers’ concerns:

The result for self-described Democrats seems high, acknowledged Pinkus, but, she wrote in an e-mail to readers, “This is what the poll got from a random sample of 1,233 adults nationwide, including 1,115 registered voters (which includes listed, unlisted and cell phone users).’


McManus, who reported the poll results in the front-page article, explained it in an e-mail he sent to readers who asked. He wrote, ‘Please note that this question is not about how people are registered. It’s about how voters identify themselves when asked, ‘Regardless of your party registration or how you have voted in the past, do you think of yourself today as a Democrat, or a Republican, or an independent, or as something else?’' The answers to that question often ebb and flow depending on the popularity of each political party.”

The article itself put it this way: “After the terrorist attacks of Sept. 11, 2001, for example, when President Bush’s popularity soared, the number of voters who described themselves as Republicans rose as well. During the last three years, as Bush’s popularity has slumped, the number of voters who describe themselves as Republicans has also dropped.’

As the fine print explained to readers, this poll was weighted slightly, where necessary, to conform to the Census Bureau’s proportions of sex, race, ethnicity, age and national region. However, Pinkus emphasizes, ‘The poll is NOT weighted for party identification, since party ID is a moving variable that changes from one election to another, or when one party may be favored more than the other.’

On that point, McManus’ went on: ‘There’s a long-running debate among pollsters on how to handle the ebbs and flows. In one camp, you go out and poll a random sample, and report whatever number you get. (We’re in that purist camp.) In the other camp, you ‘adjust’ the sample to bring it closer to whatever number you think is the right one -- but that means the pollster is inserting his/her judgment into the process. And there’s a margin of error -- in this case, plus or minus three. So our 2007 poll with 28% Republicans and our 2008 poll with 22% Republicans could reflect ‘real’ numbers that were constant at 25%. Again, those aren’t registration numbers; they are ‘how do you think of yourself’ numbers.’

Pinkus concurs, ‘This number fluctuates over time. Comparing other polling organizations’ party ID numbers, CBS News/N.Y. Times had Democrats 14 points more than Republicans, Democracy Corps had a 12-point gap in favor of the Democrats and ABC News/Washington Post had Democrats with 14 points more than Republicans.’

(Some readers also thought that the total -- 39% identifying themselves as Democrats, 22% as Republicans and 27% as independents -- didn’t seem to add up. What the article didn’t note was that 8% said they consider themselves as something else, and 4% refused to say.)

Of the results published in today’s Times, Pinkus concludes: “It is early in the election and many things can happen that could shift party loyalty, but as of today, this is where the country stands.’