WASHINGTON — Barack Obama leads Hillary Clinton in South Carolina, where their increasingly bitter rivalry has opened a deep racial divide among Democrats days before the party's first primary in the South on Saturday, according to a new McClatchy-MSNBC poll.
African-Americans in South Carolina break solidly for Obama, with 59 percent supporting the Illinois senator, 25 percent behind New York Sen. Clinton, 4 percent for former North Carolina Sen. John Edwards and 12 percent undecided.
White voters see the primary from the opposite direction: 40 percent support Edwards, 36 percent back Clinton, 10 percent are behind Obama and 14 percent are undecided.
"It's still a racially divided state," said J. Brad Coker, the managing partner of Mason Dixon Polling & Research, which conducted the survey.
Never miss a local story.
With African-Americans expected to dominate the voting, their strong support gave Obama the overall edge.
The statewide landscape, as of Wednesday night:
-- Obama, 38 percent.
-- Clinton, 30 percent.
-- Edwards, 19 percent.
-- Undecided, 13 percent.
The poll's error margin was plus or minus 5 percentage points.
As if has everywhere so far this year, the contest in South Carolina remained somewhat volatile.
Edwards gained ground in this poll, up 6 points from the week before, primarily because of white men. He's looking for a repeat of his 2004 primary win in his birth state.
Clinton was waging a spirited challenge, returning to the state Thursday after a two-day absence while her husband campaigned there full time and her ads continued to air on television.
At the same time, a nasty debate Monday in Myrtle Beach eroded "likability" ratings for all three candidates, more than 1 in 10 likely voters remained undecided and 1 in 5 who did express support for certain candidates said they could still change their minds.
Obama owed his lead largely to African-Americans, particularly men, 66 percent of whom supported him; black women gave him 55 percent support.
Obama draws the least backing from white women, only 8 percent, and only 11 percent from white men.
He led among younger voters, drawing 47 percent of those 49 and younger, more than supported Clinton (22 percent) and Edwards (20 percent) combined.
On issues, Obama had a slight edge over Clinton among voters who said the economy or health care was their top concern. The economy dominates kitchen tables in South Carolina as it does elsewhere this winter, up sharply on the priority list from December. Health care trails in second place. No other issue comes close.
Obama also has forged a solid bond with his supporters; they were the least likely to switch to other candidates.
Clinton's strength is white women. She gets 43 percent of their support, compared with Edwards' 34 percent and Obama's 8 percent.
She also has an edge among voters older than 50.
Edwards was the only candidate whose support had increased from the week before. One possible reason was Monday's debate, in which Clinton and Obama clashed bitterly while Edwards tried to be the voice of reason. The next day he said he was the "grownup" in the debate.
But his ability to gain more could be limited by his small base of support. The one group where he leads is white men. He has 47 percent of them, Clinton 29 percent and Obama 11 percent. Most white male South Carolinians are Republicans, however, who weren't included in the survey.
"Edwards has no appeal at all among minority voters and they're half the vote. That limits his ability to move up. He's not going to take black votes away from Obama, and not that many white women from Hillary," Coker said.
"Edwards is picking up some support from whites. But he's still not likely to be a threat to win or even finish second. He may be helping Obama by pulling white voters away from Hillary."
HOW WE POLL
The McClatchy-MSNBC Poll is a snapshot of voter opinion at the time it was conducted. It isn't a prediction of how people will vote on Election Day.
The Mason-Dixon poll of 400 likely Democratic primary voters in South Carolina was conducted by telephone Tuesday and Wednesday.
Those interviewed were selected by a random variation of telephone numbers from a cross-section of telephone exchanges. That means that anyone in the state with a phone line had the same odds of being called as anyone else, except for people who use only cell phones. Cell phone numbers aren't in the exchanges.
The margin of error was plus or minus 5 percentage points. That means that 95 percent of the time, the correct numbers could be up to 5 percentage points above our poll's percentage-point findings or up to 5 percentage points below them. The other 5 percent of the time, the correct numbers could vary even more.
The sampling margin of error doesn't include other variables that could affect results, including the way questions are worded or the order in which they're asked.