An article in the October 10, 2008 issue of SCIENCE (Vol. 322, No. 5899) by Jennifer Couzin asks "Do Voter Surveys Underestimate the Impact of Racial Bias? " Her article does not give a clear answer. The reason she cannot arrive at a clear answer is that the polls themselves are so full biases that they cannot isolate racial bias separately.
She does pose a good question: have the polls "been skewed by an inability to detect racial bias?" Some pollsters think Obama's poll numbers may be off by 6% due to bias but they cannot tell if the polls are or are not reflective of the bias. In other words, its just a guess.
Couzin tells us this bias is called the "Bradley effect" after L.A. mayor Tom Bradley who lost his 1982 bid to be governor of California even though he was ahead in the polls. The assumption is that because he was Black an unexpressed bias cropped up in the election that had not revealed itself in polls.
The article reports that Harvard political scientist Daniel Hopkins, after analyzing every race for the Senate or for a governor from 1989 to 2006, concluded that there was no evidence for the "Bradley effect" after 1996.
He doesn't expect race to be a big factor in the 2008 election.
However, Couzin reports that the president of the Pew Research Center, a major polling outfit, (Andrew Kohut) fears that because of "reluctant respondents" [people who don't like to take part in polls] the polls don't really have a representative sample of the population. Studies have shown that reluctant responders are more likely to be racially biased. The Obama lead may not really be there in close polls. Suppose we gave McCain the benefit of the 4% margin of error and the 6% bias drop-- he would have 10 points, maybe 14 points, to add to his poll numbers whenever there is a comparison with Obama. This is a worse case scenario but it is a possibility.
Next the article mentions Michael Lewis Beck who teaches political science at the University of Iowa. He has factored racial basis into his voting model and thinks Obama will win the popular vote, Couzin says, but lose in the electoral college.