The Pew Research Center declares, “polls conducted in campaigns pitting white and black candidates against each other are doing a better job of accurately predicting the outcome of the election now than in the past, suggesting that hidden biases that confounded polling in biracial elections in the 1980s and early 1990s are no longer a serious problem.”

I have two minor quibbles with their report. They spend quite a bit of time looking at the races in 2006, including the campaigns of African-American Republicans Ken Blackwell in Ohio, Lynn Swann in Pennsylvania, and Michael Steele in Maryland.

If the “Bradley Effect” measures fear of being accused of racism (instead of the popular belief that it measures racism) we’re left wondering whether those who oppose African-American Republican candidates would be accused of racism. Were white Democrats Ted Strickland, Ed Rendell, and Ben Cardin accused of racism? (I wouldn’t be surprised if someone at some time during those campaigns made the accusation, but I think we can safely conclude that the dominant theme in the coverage of those campaigns was not how those Democratic candidates were standing in the way of racial progress.)

Second, they write:

Black Republican candidates for governor lost by wide margins in Ohio (by 23 points) and Pennsylvania (20 points), but the average of the final independent polls in each state showed similar margins (21 and 23 points, respectively).

Pennsylvania, yes, but looking back at RealClearPolitics’ archives, I calculate the average a bit differently.

Survey USA’s final poll put Strickland’s margin at 17 percent. The University of Cincinnati put it at 22 percent. Mason-Dixon put it at 19. (A bit earlier than those polls, CNN got the margin precisely right, and Zogby put it at 8 percent.)

Strickland finished with 60.4 percent, Blackwell with 36.8 percent, a margin of 23.6 percent.

(One poll wildly overestimated the margin; the Columbus Dispatch put it at 36 percent. For some reason, the guys at RCP did not include that poll in their final average.)

In that race, the African-American’s margin of defeat was 4.2 percent worse than the RCP average.

Now, taking a look at this year’s Ohio Democratic primary… Quinnipiac put Hillary Clinton up by 4, University of Cincinnati put her up by 9, Suffolk put her up 12, PPP put her up 9, Survey USA put her up 10, Rasmussen put her up 6, and Zogby said it was a tie. (!) The final average was 7.1 percent; Clinton won by 10.1 percent.

In that race, the African-American’s margin of defeat was 3 percent worse than the RCP average.

If you had to name a state where Obama is going to underperform his margin in the final polls, Ohio would seem like a pretty good guess.