Yesterday we reported on the undervote problem in North Carolina -- votes that are cast but don't register a vote for president. In 2004, N.C. had more than 92,000 votes that didn't count a presidential choice -- one of the highest undervote rates in the country, and if repeated in 2008, enough to impact a close presidential race.

We pointed out that several reports -- and N.C.'s own Director of the State Board of Elections -- attribute much of the undervote problem to the fact that straight-ticket voting in North Carolina doesn't include the presidential race. The confusing law is a by-product of the 60's, when state Democratic leaders feared getting lumped in with national liberal presidential candidates and civil rights advocacy.

Of course, there are other reasons why North Carolina voters might not vote for president. At least some is intentional -- voters who want to cast a ballot for state races, but don't have a preference at the presidential level.

But there is some good evidence that a large chunk of undervoting isn't intentional. For example, in 2000 there were nearly 2 million undervotes nation-wide, which prompted several studies into the possible causes.

One study was presented by a team of researchers to the Southern Political Science Association meeting in 2001. While finding that the cause of undervotes is "multi-faceted," they found that the lack of straight-ticket ballots including the presidential race was a big factor:

If states want to reduce the number of unrecorded votes in presidential elections, they might consider adding a straight-party option to the ballot if they do not already have it. [...] The mean residual vote rate is one percentage point higher in states without the straight-party mechanism. [...]

[I]t appears that the straight-party mechanism allows some voters a way to bypass confusing features of the ballot and avoid mistakes in the presidential contest.

What kind of difference would that make in North Carolina? If the 2004 undervote had been 1.57% instead of 2.57%, there would have been 35,800 more votes for president in 2004.

What kinds of voters are most likely not to have a presidential vote counted as a result? The study finds a significant correlation with race and income:

[M]oving from a county with no African-American residents to a county that is 50% African-American will, on average, increase the residual vote rate by one percentage point. Similarly, the mean residual vote rate in a county with a median household income of $44,500 (90th percentile in our sample) is expected to be .7 percentage points lower than a county with a median income of $24,500 (20th percentile).