Race of Interviewer

[cinyc]

I don't know whether Lake or Gallup/NBC are correct, but  I think the IVR polls are sometimes more accurate in a race involving an African American because there is no human being on the phone to judge your answer.   A lot of older respondents ashamed to say they won't vote for the black candidate tell human pollsters that are undecided when they are not.   SUSA came closest to predicting the exact Massachusetts gubernatorial outcome - perhaps for good reason.

[cinyc]

Well, first of all, I have not discounted the possibility that some of the "Bradley Effect" is black voters feeling pressured to say they are voting for a black candidate.  That is is one of the things we should check after the election. 

However, there does seem to be one out there.  In looking at the late polls in Blackwell (OH), Steele (MD), and Patrick (MA), it does seem to be from 2-4 points to 1-7 points, which jibes with the 3-5 point range mentioned.  (I actually think it will be about 1-2 points, but closer to 1.)

Now, why not Ford (TN)?  That might be the key question.

(One theory that I'll put forward is that the "call me" commercial might have forced the issue out in the open.  Because the campaign was seen as racially charged, voters were more open about their choices.)

...or Tennessee is in the south, where people are more open about their racism than the (supposedly) more liberal and tolerant northeast and rust belt.