.jpg)
Nate Silver, creator of the polling aggregator website fivethirtyeight.com, spoke to an overflow crowd of students, guests and faculty on Jan. 22 at Fordham.
From a podium in the cafeteria atrium at the Lowenstein Center, Silver used his address, “Polls, Predictions and the Role of Internet in the 2008 Elections,” to discuss which pollsters most accurately called the presidential race.
He also answered audience questions about topics as varied as the effects polls can have on voting patterns and the future of Internet polls.
He did not, however, reveal exactly how fivethirtyeight.com works, except to say that his site assigns polls a weighting based on their track record, sample size and timeliness. The weighting, he said, separates his site from another site, realclearpolitics.com, which also predicted that Barack Obama would defeat John McCain last November.
“Online poll aggregators were a key innovation, and one of the main highlights of the 2008 election cycle,” said Costas Panagopoulos, Ph.D., assistant professor of political science and director of the Center for Electoral Politics and Democracy at Fordham. “Silver and his website provided timely and reliable data and analysis about trends in candidate support over the course of the election.”
Silver’s analysis showed that Zogby, AP-GFK and Insider Advantage were the most accurate of all polling firms, although the percentages separating them were small. Comparing polling firms can be misleading, he said, because pollsters such as Rasmussen Reports include states like Alaska, whereas Zogby sticks to so-called battleground states.
Quinnipiac University, which is just north of New York City, did not conduct enough national polls to warrant inclusion in his rankings, but he noted that the university does a good job because it knows its area well.
“If you know your state, you know your region, then you can get a little bit of extra mileage out of it. The regional pollsters seem to do best, better than the kind of whole-hog national pollsters,” he said.
One of the most revealing visual aspects of Silver’s presentation came when he showed a color-coded map of the United States much like the red and blue ones familiar to election-watchers.
Dark blue areas showed where Obama performed better than most polls. White areas showed where polls accurately predicted how he would perform, and dark red areas showed where he did worse than had been predicted. There was a lot of red in the Appalachian region, he noted.
“Was there some sort of Bradley Effect, where people said, ‘I’m going to vote for Obama,’ and then decide, ‘I’m not going to vote for a black guy,’ once they got into the voting booth?” he asked.
“Maybe there’s a regional Bradley Effect, but I don’t think there was a national one,” he said. “Washington, Maine, Rhode Island and Massachusetts are also very white states, so I don’t think it had any effect on them.”
Silver cautioned that there was no perfect solution to modern polling problems, such as the increase in voters who only use cell phones. Completely random calls to 100 voters at home during the day will statistically reach more women and senior citizens, for example, so as more people go completely wireless, pollsters may need to spend money to pay for calls to cell-phone owners.
Silver’s appearance was sponsored by the Center for Electoral Politics and Democracy and was co-sponsored by Edison Media Research and the New York Association of Public Opinion Researchers.
