Two Good Reasons- The Wilder Effect and the possibility of MASSIVE black voter turnout.
The Wilder Effect (Also known as the Bradley Effect and the Dinkins Effect) states that in order to either conceal prejudice or to seem more "progressive", people talking to pollsters will LIE and say they plan on voting for a non-white candidate over a white one. This patter will continue until they get into the voting booth and pull the lever for the white candidate over the non-white one.
Some relevant numbers for you
Chicago Mayoral Race 1983: Washington (Black) vs. Epton (White)-
- Polling- Washington 52, Epton 38
- Result- Washington 51.5, Epton 48.1
- Polling- Dinkins 53, Guliani 39
- Result- Dinkins 50, Guliani 48
- Polling- Grant 47, Helms 41
- Result- Grant 47, Helms 53
- Polling- Jindal 48, Blanco 43
- Result- Jindal 48, Blanco 52
On the other hand you have the possibility of tens of thousands of new voters coming to the polls for Obama that pollsters have no idea how to count. Massive numbers of newly registered African American voters may be able to do things for Obama thought impossible, winning deep south states like Mississippi, Alabama and Georgia. These voters are also motivated largely by racial preference, though it's in a different light. Obama himself predicted that African American turnout across the nation would be "up 30%, minimum". In some southern states where the number of African Americans voting is normally far below the actually number of African Americans turning out, that would be an amazing figure, giving obama a boost well above his current polling numbers.
In the end we could finish with the Wilder Effect balancing out these new voters in a zero sum game bound to confuse everyone involved. It's the preferable outcome unfortunatly, otherwise you'd get a situation where you'd have a candidate rejected because of race or accepted because of it (both bad choices for america).
So what do you think? Will the Wilder Effect put McCain over the top? or will new black voter turnout do him in?