The main problem with this sort of study is the very small number of African American major party nominees for state office. I collected election results from all the U.S. Senate and gubernatorial elections conducted in 2002, 2004, and 2006. In these 181 contests, there were only 10 African American major party candidates: 5 Democrats, 5 Republicans. Even worse, I had to throw out two of those cases -- Barack Obama and Alan Keyes, who ran against each other in the 2004 Senate race in Illinois. Presumably, any African American "penalty" would have been cancelled out in that election.
I found no statistically significant race effect. The effect I measured is, in fact, negative, but it's less than two points and has a t-score of -.47. This means either that blacks aren't systematically punished in statewide elections (yay!) or that I haven't gathered enough data (boo!). But I'm not sure how valuable additional data-gathering would be -- there just aren't that many African American statewide candidates going back into the 1990s, and even if an effect did turn up then, one could argue that it no longer exists.
The story is a bit more interesting if we just focus on the black Democratic candidates. I used my control variables (the Democratic presidential vote in each state, incumbency, the percent of the state's population that's African American, region, and year and contest dummies) to generate an "expected" Democratic vote in each contest. In each of the four contests, the African American Democrat ran behind the expected vote, although not by much:
- Deval Patrick (Gov cand., MA, 2006): Expected vote: 57.2; Actual: 56
- Harold Ford (Sen cand., TN, 2006): Expected vote: 50.4; Actual: 48
- Carl McCall (Gov cand., NY, 2002): Expected vote: 37.5; Actual: 33
- Ron Kirk (Sen can., TX, 2002): Expected vote: 44.3; Actual: 43
Incidentally, the story is much less consistent for the black Republican candidates. In two cases, they ran ahead of the expected vote, and in two they ran behind. None got within 10 points of winning, though.
So what does this mean? It might mean nothing. Then again, if we had more data, we might find that this modest vote penalty is real. Which would mean that McCain has a one-to-two-point advantage going into the November election just based on race. I imagine Obama is planning to win by more than that anyway, but it's something to consider.