Of all the mysteries in this confounding presidential election, this may be the most significant question of all: How many voters will ultimately decide they can't vote for an African-American?
"It's definitely a factor," said Republican pollster Frank Luntz. "How do you count for it? It's almost impossible. But you have to assume you're going to have about a two-point margin of error in almost any poll because of race."
With close races in critical battlegrounds including Florida, Ohio, Colorado and Virginia, two points could make the difference in who wins and loses the election. And if Obama loses in November — with the Democrats having a strong political wind at their back — count on hand-wringing and speculation that America simply wasn't ready to elect an African-American president.
"I don't see what else they could cite, quite frankly," said Stanford University political scientist Shanto Iyengar, who noted that with the anti-Republican climate in the country, any prediction based on historical trends would have had Obama well ahead.
"The fact that the race has been much tighter than that, even before the conventions and vice presidential picks, suggests that there's some kind of drag on the Obama candidacy," he said. "My suspicion is that it has to do with his race."
Or it could be Obama's lack of experience. Or that he's perceived as aloof or too liberal. Or that Republican John McCain, a national figure and respected war hero, has been largely immune from the taint on his party. No one can know with certainty how much race matters and probably no one ever will.
Roughly 20 percent of Democratic primary voters in the critical battlegrounds of Ohio and Pennsylvania, which Hillary Rodham Clinton won, told exit pollsters that race was a factor in their decision. On the other hand, the fact that a black man named Barack Obama defeated a Democratic icon for the party's nomination and now is in a winnable contest for the presidency suggests skin color might not be such a hurdle after all.
"To the people who we believe are truly undecided voters — and we spend a lot of time trying to figure out who's in that pie — race is not going to be a hindrance because they're open to supporting Sen. Obama," said David Plouffe, Obama's national campaign manager.
A pollster calling Sandra Cichon, a 60-year-old Democrat from Spring Hill, would hear her identify herself as an undecided voter. But is she really?
"I can't imagine having a black president, and I think he's inexperienced," she told a reporter recently, eventually acknowledging she was leaning unenthusiastically toward McCain. "I don't think we (Democrats) have a chance to be in the White House with Obama."
Many analysts wonder how many voters answering polls hide their racial biases or mislead survey-takers about their real preferences.
It's known as the "Bradley effect," after former Los Angeles Mayor Tom Bradley, an African-American who in 1982 was leading nearly every poll in the California governor's race but lost to his white opponent. The theory — not universally accepted as valid — is that some white voters tell pollsters they support a black candidate — out of political correctness — but won't vote for one.
In 1989, African-American Democrat Douglas Wilder barely won the Virginia governor's race though polls pointed to a Wilder landslide. That same year, David Dinkins narrowly won the New York city mayor's race despite polls showing a double-digit lead.
In North Carolina in 1990, African-American candidate Harvey Gantt led Republican Jesse Helms in the polls, but Helms won soundly. More recently, a 2006 proposal before voters in Michigan to ban affirmative action looked too close to call, according to polls, but it passed with 58 percent support.
"As pollsters, we have to believe what people tell us. We cannot look into people's soul and know whether or not they're lying," said Democratic pollster Tom Eldon, whose clients include the St. Petersburg Times. "But what we have been told in recent polls is that people are less apprehensive about electing an African-American than electing a woman or electing a man in his 70s."
The Bradley effect may be a bygone relic. In 2006, many observers questioned the polling in Tennessee's Senate race, featuring African-American Democrat Harold Ford and white Republican Brad Corker, but Ford's narrow loss closely matched the polls.
In this year's hard-fought Democratic primaries, exit polls — an especially tricky process that involves talking to voters as they leave their voting precincts — frequently overstated Obama's support. Yet the pre-election polls were mostly close to the mark, with the exception of the Northeast, where Clinton outperformed the polls in states such as New Hampshire, Massachusetts and Rhode Island.
"Everyone has always assumed the South has had more racial issues, and yet the Northeast is where this thing hit much more often," said pollster Luntz, predicting racial issues could also be a factor in the more competitive industrial Midwest where the race may be decided.
Mark Mellman, John Kerry's pollster in 2004, dismissed the Bradley effect but not the potential importance of race.
"Are there people who are not going to vote for Barack Obama because he's black? I'm afraid there are," Mellman said. "Are they going to hide their racism? Probably yes. Are they going to hide their vote? Probably not. There's no evidence that's happening in any systemic way."
Obama is banking on high African-American turnout to give him an edge that Kerry and Al Gore lacked in 2004 and 2000. That could counter any potential racial backlash. In Florida, for instance, Obama's huge get-out-the-vote organization sees great opportunity in the roughly 500,000 registered black voters who skipped the presidential election in 2004.
Still, high turnout among blacks is not enough. Consider Florida:
In 2004, President Bush beat Kerry in Florida by 381,000 votes in an election when Kerry did well among black voters and roughly 917,000 black voters turned out. Even if Obama increases black turnout by 25 percent — a huge challenge — and wins even more black support than Kerry, he could still fall 100,000 votes short of matching Bush's total.
Anything that diminishes support for Obama among white voters could put McCain into the White House. Bush won white Florida voters by 15 percentage points in 2004. A Quinnipiac University poll released last week found McCain winning the white vote in Florida by 24 points, while a Rasmussen poll showed McCain leading by 17 points among white voters.
"It would be naive to think racism isn't a factor, but the question is how much of a factor it is. We don't know," said Peter Brown, assistant director of the Quinnipiac University Polling Institute. "As to the Bradley effect, we won't know that until Election Day."
Times staff writer John Frank contributed to this report.