Rick Perlstein channels Tom Schaller channeling some academics:
Schaller builds this conclusion on one of the most impressive papers in recent political science, "Old Times There Are Not Forgotten: Race and Partisan Realignment in the Contemporary South," by Nicholas Valentino and David Sears. Running regressions on a massive data set of ideological opinions, Sears and Valentino demonstrate with precision that, for example, a white Southern man who calls himself a "conservative," controlling for racial attitudes, is no less likely to chance a vote for a Democratic presidential candidate than a Northerner who calls himself a conservative. Likewise, a pro-life or hawkish Southern white man is no less likely--again controlling for racial attitudes--than a pro-life or hawkish Northerner to vote for the Democrat. But, on the other hand, when the relevant identifier is anti-black answers to survey questions (such as whether one agrees "If blacks would only try harder, they could be just as well off as whites," or choosing whether blacks are "lazy" or "hardworking"), an untoward result jumps out: white Southerners are twice as likely than white Northerners to refuse to vote for the Democratic presidential candidate. Schaller's writes: "Despite the best efforts of Republican spinmeisters ... the partisan impact of racial attitudes in the South is stronger today than in the past."
Interesting stuff. Meanwhile, Jonah Goldberg has apparently gotten a lot of email from readers who think the "racist" answers to these questions don't really demonstrate racism. I think that's a bit daft, but however you want to characterize the question, the point is that individuals' attitudes toward race (again, however you want to classify those attitudes) have a large impact on voting behavior.