What a difference a poll makes
A spate of widely publicized newspaper and network polls over the past week have shown Barack Obama opening a big lead over John McCain. But other surveys tell a somewhat different story, suggesting the presidential race is still close, and the Republican has even gained ground in recent days.
The reason for the divergence: Pollsters are facing new challenges this year, trying to gauge whether the electorate is changing, and how much.
Surveys giving Sen. Obama a large and growing lead tend to assume that a growing proportion of voters are Democrats, and a shrinking percentage Republicans. They also point to a big increase in turnout, particularly among voters under the age of 30. Surveys showing a closer race assume less change in party affiliation in particular.
The polls owe their wide variations, in part, to differences in how they determine likely voters. Gallup actually conducts two separate daily polls, one that includes all surveyed adults who say they will vote, and a second that is more restricted, using a decades-old methodology that determines “likely voters” in part by examining historical models on the types of voters who have showed up at the polls.
In the first Gallup sample, Sen. Obama leads Sen. McCain by six points. The second group yields the two-point gap. Both polls were conducted from Oct. 13-15.
Differences over how to accurately gauge party affiliation also help account for the discrepancies. Some pollsters argue polls should be statistically “weighted” so that their results achieve a partisan composition that reflects long-term national averages — particularly if a poll shows that one party gets an unusually large share among the respondents, compared with past elections.
Pollster Scott Rasmussen, for example, weights current polls so that Democrats outnumber Republicans by a 39.3% to 33% margin, while pollster John Zogby adjusts polls so that Democrats account for around 38% of the electorate and Republicans, 36%. So even if a particular sample of calls shows different ratios, the pollsters adjust to fit that formula.
“What troubles me is when I see some of my colleagues have 27% of the respondents that are Republicans. That’s just not America, period,” says Mr. Zogby, whose polls have shown Sen. Obama with a lead ranging from two to six points this month. He argues that while party affiliation fluctuates over time, it doesn’t change “day-to-day, and it never fluctuates by eight points in a short time period.”
Other pollsters argue that polls should use whatever partisan mix results from a particular survey rather than arbitrarily establishing party affiliation weights. “How do you know that’s right? I mean, they’re making up numbers,” says Susan Pinkus, who conducts the Los Angeles Times-Bloomberg poll, which isn’t weighted. In this week’s poll, the respondents were 34% Democratic and 26% Republican.
Both campaigns are running large vote turnout operations, and the Obama campaign is counting on unprecedented turnout from young voters, which further complicates efforts to determine likely voters. “It’s more art than science in many cases. They’re very difficult decisions to make,” says Neil Newhouse, a Republican pollster who conducts the NBC News-Wall Street Journal poll.
Predicting turnout among young voters remains particularly challenging because many of those voters don’t use landline phones that pollsters traditionally rely on to achieve a balanced sample. Pollsters have also struggled with accurately predicting minority turnout and how race could influence the current election.