Increased participation and competitiveness expected in today's presidential elections is intensifying the focus on the only predictor of Election Day outcomes: polls.
Campaigns and journalists frequently point to numbers such as daily tracking polls, battleground polls and even Intrade.com's political "stock markets" to show trends in candidate support.
According to The Chronicle's election poll, conducted Oct. 27-30 among, 74.8 percent of Duke undergraduates eligible to vote favor Democratic Sen. Barack Obama, compared to 23.3 percent who favor Republican candidate Sen. John McCain.
And of those who said they have already voted-61.9 percent-nearly 80 percent said they voted for Obama.
Because the upcoming presidential election is between a black and a white candidate, many observers have said the so-called "Bradley effect" may explain some of Obama's lead in the polls. The effect takes it name from the 1982 California gubernatorial race, in which Republican George Deukmejian defeated Democratic candidate Tom Bradley despite polls showing Bradley with a substantial lead before the election.
The theory is that poll respondents declared support for Bradley, even though they supported Deukmejian, so as to not seem racist to pollsters. Yet even after years of analysis it remains unclear whether the effect was the actual cause of the discrepancies in polling.
With polls showing Obama leading by less than five points in many battleground states, a Bradley effect could mean that McCain is victorious today.
Experts in polling and politics, however, downplayed the importance of the effect.
"I am doubtful that the Bradley effect will play a role," said David Brady, associate professor of sociology. "Any respondent can think of 100 reasons to vote for McCain without seeming racist."
Others argue that even if the effect does exist, racial dynamics have changed significantly since 1982.
"The Bradley effect dates back 20 years or more," said Ferrell Guillory, founder and director of the Program on Public Life at the University of North Carolina at Chapel Hill. "I wonder, though, if [the effect] is offset by the increased excitement with newly registered black and young voters."
Although a high turnout among minority voters could outweigh any possible Bradley effect, it is difficult for polling organizations to try to determine which voters are going to show up at the polls.
"I don't think the pollsters have the faintest idea who is going to show up on Election Day," said David Rohde, Ernestine Friedl professor of political science.
Still, interest in the election may be an indicator. According to The Chronicle's poll, 83.7 percent of students said they "care a great deal" about who will win the election.
Beyond the Bradley effect, there are also issues with the polling techniques themselves.
Demand for daily updated polls has put immense pressure on pollsters to conduct statistically valid polls on short notice, said David Banks, professor of the practice of statistics.
The Chronicle's poll had a 2.4 percent margin of error, which is considered to be relatively low. In other words, for questions on which respondents are at a roughly 50-50 split, there is a 95 percent chance that the number of lesser value is actually within 2.4 percentage points.
Errors based on sample size are actually "the least troublesome of the causes of uncertainty" Banks said, adding that biases play a much bigger role in the uncertainty of polling results.
In fact, the Bradley effect itself is considered to be a bias, a social desirability bias in which respondents give answers they perceive to be socially acceptable rather than their true intentions.
"Pollsters have gotten very smart and have managed to solve almost all of the problems seen in previous elections," Banks said. "[The pollsters] are bad at problems they haven't seen yet."
Get The Chronicle straight to your inbox
Signup for our weekly newsletter. Cancel at any time.