CNN.com (that glorified tabloid) ran an article after the Mississippi primary which noted:
For Mississippi, [the primary is] a moment to bask in the national spotlight. And for a state with images of a strictly segregated past, the Democratic primary is a chance to alter some long held stereotypes.
Let's go to the numbers, shall we? Keep in mind Obama won 61% of the vote (or if you prefer, Hillary won 37%), so we're looking for deviations from those numbers.
To begin, 70% of whites voted for Clinton; 92% of blacks voted for Obama. Wow, way to alter some stereotypes. Too harsh? Consider that Obama won just over 60% of voters who thought race was "important," and he also won just under 60% of voters who felt race was "not important". Seems a bit contradictory at first glance -- does this mean Mississippians were lying about their opinion of racial significance?
Four-way exit poll data interacting racial significance and race isn't available (send it to me if you find it!) but we can try to get an answer anyway. What we know:
- 50% of voters were black; 48% were white
- 31% felt race was important; 69% disagreed
- White voters went 72%/28% for Clinton; black voters went 90%/10% for Obama
Based on this, we can set up two scenarios. In both scenarios, we assume voters who think race is important vote along the racial breakdowns (i.e. the 90%/10% split for Obama). As for voters who think race is unimportant, in the first scenario we assume they are telling the truth and have them vote along the general breakdown (61% Obama vs 37% Clinton). In the second scenario, we assume they are lying and have them vote with the racial breakdown. I feel comfortable with these assumptions because a) nearly every other measured category agrees with the general breakdown, so it seems a reasonable proxy in scenario 1 and b) given the clear racial split, it appears that race being important would cause people to vote with their race, not against it.
Here are the results:
Many of the numbers from both scenarios match reality closely -- for example the totals for Obama (59% and 58%) and Clinton (37% and 40%), and the percent of people who find race important AND vote for Obama (60% in both) -- but the racial breakdown (yellow highlight) is only correct in scenario 2. In scenario 1, where we assume people who find race unimportant vote with the general trend, only 71% of blacks end up voting for Obama; if we assume everyone votes with the racial breakdown then 90% of blacks end up voting for Obama and this is the result we see in reality.
To be fair, the assumptions here are a bit restrictive and it is possible that the real data would disagree, but these findings are highly suggestive that even people who claimed race was unimportant voted along racial lines; this is the only way to arrive at the true results. Is it possible that people voted with their race AND race was a non-issue? Yes, but highly unlikely.
For the curious, gender showed no significant bias -- Obama got 61% of men and 58% of women, and fared a few points better among people who felt gender was important.
Most other categories fell into the general 60/40 split, with a notable exception -- 75% of Republicans voted for Clinton. --wait, what? Yeah, Republicans made up 12% of voters, meaning 9% of voters hailed from the GOP and voted for Clinton. Interestingly, among the 13% of voters who found John McCain "strongly favorable," 70% went for Clinton -- another (or the same?) 9% of outlier votes for Clinton. Might the rumors be true?
And finally, of the voters who find Clinton "honest and trustworthy" (49%), 46% voted for Obama. Conversely, of the people finding Obama "honest and trustworthy" (70%), just 17% voted for Clinton. Is this the polarizing impact of negative campaigning?
But now I'm probably just playing with numbers.