Barack Obama’s big win in North Carolina came as no great surprise to me, given the significant African American and up-scale white Democratic demographic representation in the Tarheel State. I was mildly surprised, however, by his near-win in Indiana because Real Clear Politics.com’s last cumulative pre-election poll average showed him trailing Hillary Clinton by four points, and he eventually lost by only two. This difference hardly seems earth-shaking until you consider that in five of six previous primary losses in states outside the South where polling data is available, Obama’s percentages of the actual vote have fallen short, frequently well short, of the percentages indicated in the final poll averages. Obama ran eleven points behind his final poll average in New Hampshire, ten points behind in California, and eight points behind in Massachusetts. In North Carolina, on the other hand, he ran six points ahead of the polls, and in Georgia, South Carolina, and Alabama, his positive margins were eighteen, seventeen, and fifteen points, respectively.

Obviously, polls can’t ever be perfect. For example, despite duly diligent efforts of pollsters to contact voters from every segment of the electorate, it’s always a distinct possibility that the opinions of lower-income groups will be underrepresented. This could be critical in southern states, where African Americans represent a much larger percentage of the Democratic electorate than elsewhere, and it could help to explain the wider than expected Obama victory margin in several southern states, including North Carolina, where some 65,000 new African American voters have reportedly been registered since January alone.

Beyond that, polls are like computer programs. What they tell us can be no more reliable than what we tell them. Even opinion polls completed only hours before the voting polls open cannot account for what happens in the hearts and minds of the electorate in the all-important meantime. Surely, part of Obama’s shortfall in New Hampshire can be chalked up to a big but late shift of women to Senator Clinton in the aftermath of her well-publicized, ostensibly “unguarded” emotional moment a few days before the vote.

If last-second shifts help to explain why Obama has performed better in the polls than at the polls in several non-southern states, what explains the shifts themselves? Trying to make sense of Obama’s surprising loss in New Hampshire, some observers invoked the “Bradley Effect,” referring to the 1982 California gubernatorial race in which former LA mayor Tom Bradley, an African American, enjoyed a clear lead in the polls only to come up short on Election Day. The following year, polls had Harold Washington headed for a landslide victory in the Chicago mayoral election before he barely managed to squeak out a win. In response to these and several similar cases, the pointy-headed obfuscators of the blatantly obvious quickly trotted out “social desirability bias” to describe the phenomenon of whites misrepresenting their electoral intentions because they are too embarrassed to tell pollsters, especially black pollsters that they don’t want to vote for a black candidate. (It doesn’t sound scientific enough for some folks, I guess, but where I come from, we just call this kind of behavior “lyin.’”)

It’s hard to know how much of a role plain old prevarication by whites has played in generating the disparity between Obama’s polling percentages and his vote percentages outside the South, but I’m guessing it may be a significant one. If so, this raises some questions: Why would white Democrats outside the South be more likely to say they support a black candidate and then fail to actually follow through in the voting both? Are, as some suggest, northern whites simply more sensitive to the social stigma attached to any expression of opinion that might somehow be construed as racist? I love this interpretation because it makes northern whites who try to conceal their racial hang-ups seem more enlightened than southern whites who presumably have no such qualms. Unfortunately, this thesis doesn’t quite square with the fact that, to cite but one example, the percentage of white Democratic voters who actually admitted in exit polls that race had “mattered” in their voting choice was slightly higher in Indiana than in North Carolina, and the percentage (78) of those same whites who also voted for Clinton was a whopping sixteen points higher among Hoosiers than Tar Heels. However directly it may fly in the face of enduring regional stereotypes, from where I sit, Obama’s much stronger performance among whites in states like Georgia and Texas than in states like Ohio and Pennsylvania would seem to suggest that maybe whites in the South who say they support Obama are simply more likely than whites elsewhere to really mean it. This possibility likely says less about the greater racial tolerance of white southerners as a group, of course, than the relatively stronger economic and educational bonafides of southern white Democrats, among whom blue-collar types are as rare as a full set of teeth at a Porter Wagoner concert.

In any case, with the nomination of a black candidate now a virtual certainty, Democratic campaign strategists would at least seem to have a better idea of where they stand with white Democratic voters in the South than with Democratic whites elsewhere in the country. Regrettably for the Dems, white Democrats hardly dominate the electorate in these parts. So far as I can tell, even Bill Clinton, who won four southern states in both 1992 and 1996, only managed about 35 percent of the total southern white vote. I’m eager to see how close Obama, who wants to become our next “first black president,” can come to that figure.

In reality, it will probably be white voters outside the South who ultimately tell the tale for the Democrats this November. Constructing a reliable appraisal of the party’s prospects this fall depends on whether pollsters can find better ways to assure that the tale they hear from these voters between now and then is by any means a truthful one.