Much of Obama's advantage lies in greater-than-usual participation by Democrats; they outnumber Republicans by 7 percentage points among likely voters in this poll, as they have consistently this season. In 2004, by contrast, Democrats and Republicans turned out in equal numbers.
Obama's lead depends in large part on a larger-than-usual differential between Democrats and Republicans.
Not only are there more Democrats, Obama's doing a bit better in his own party; 91 percent of Democrats support him, compared with McCain's support among Republicans, 84 percent. Republicans are more often reliable party voters, but 12 percent of them now favor Obama, better than John Kerry or Al Gore's share of Republicans (6 and 8 percent, respectively), and about matching Bill Clinton's in 1996 (13 percent).
Swing-voting independents, meanwhile, divide very closely, 48-47 percent, as do married women, another potential swing group.
Among other groups, McCain leads Obama by 6 points among whites, 51-45 percent; Republicans have won whites by a wider margin, averaging 13 points, in the last eight presidential elections.
Obama's support from blacks is nearly unanimous, and he holds a substantial advantage among Hispanic voters as well.
Young voters remain Obama's best group, with a 2-1 advantage among those under 30; a question as ever with this group is the extent to which they turn out. Still, among likely voters 30 and over Obama retains an edge, albeit a much smaller one, 51-45 percent.
Among religious groups, evangelical white Protestants remain one of McCain's best constituencies, with a 72-23 percent lead over Obama. White Catholics, customarily a swing group, continue to favor McCain, now by 55-42 percent -- a surprise given Obama's lead overall. Part of the reason is non-evangelical white Protestants; in the past a more pro-Republican group, they now divide closely, 50-46 percent, Obama-McCain.
METHODOLOGY:This is the first in a series of ABC News/Washington Post tracking polls that will continue daily until Election Day. Interviews were conducted by telephone Oct. 16-19, 2008, among a random national sample of 1,336 likely voters, including landline and cell-phone-only respondents. Results have a 2.5-point error margin for the full sample. Sampling, data collection and tabulation by TNS of Horsham, PA.