What Barack Obama has to do to win:

Although Obama and his people are laboring mightily to expand the electorate—by driving up turnout among black voters and the young—even if they succeed wildly in that endeavor, the central hurdle to Obama’s election will be white voters. The pollster Thom Riehle, who founded the AP/Ipsos survey and is now a partner at the firm RT Strategies, calculates that even if black turnout rises by 25 percent from 2004 (and Obama wins 92 percent), if Hispanic turnout holds steady (and Obama wins 60 percent of it, seven points better than John Kerry did), and the under-50 vote rises by 5 percent (and Obama wins half of young white voters), the Democrat would still need to win 40 percent of the overall paleface vote to prevail in November, one point less than Kerry garnered and two points less than Al Gore did in 2000.

“This is a daunting task as the first black candidate for president,” Riehle tells me. “To get there, he’s got to win roughly the same proportion of Democrats and Democratic-leaning independents that all other Democrats get. If he doesn’t, he’s in a world of trouble. He can’t win it just by changing the electorate.”

[...] The question is just how accurate any of these numbers are. Polling on African-American candidates has often been unreliable in the past, overstating support for them, coughing up large blocs of alleged undecideds who actually have no intention of voting for a black contender but are too embarrassed to say so. “I know a lot of Republicans who are aware of surveys in this race that ask the ballot question ‘Who are you voting for?’ and then ask the ‘Who are your neighbors voting for?’ question,” says a GOP operative, referring to a common pollsters’ tactic of seeing through obfuscation revolving around race. “And between the first and second question, you see a five-to-ten-point shift in the answers. There’s a great big lump under the rug.”

Here’s hoping last night’s speech changed some minds.