I posted a couple months ago about how every single time anyone releases a North Carolina poll they get grief for under representing the black vote. That's because the 2004 exit poll said the voting population in North Carolina was 26% black. According to the State Board of Elections it was actually only 18.5%.
Yesterday when we released our South Carolina poll I heard from several people again that they thought we were under representing the black vote at 29%. The 2004 exit poll said 30%, and many think Obama at the top of the ticket will equate to a 10% increase in black turnout which would put it at 33% for this year.
Unfortunately the 2004 South Carolina exit poll was wrong too. According to the official data from the South Carolina Election Commission, 73% of voters were white in the 2004 general election and 27% were non white. That's a sharp strong contrast to the exit poll that said just 67% of voters were white with 30% black and 3% other races.
Race is not the only thing the 2004 exit polls got wrong in South Carolina. The exit poll also stated that just 11% of those voting in the election were over the age of 65. Based on the actual data the electorate was actually 19% over the age of 65.
Given the over representation of African Americans and the under representation of senior citizens in South Carolina, using the 2004 exit polls as a guide is a pretty risky proposition for pollsters since each of those errors would inflate Barack Obama's numbers.
We've only compared the real demographic data to the exit polls in two states- North Carolina and South Carolina. Those were both highly erroneous. Can the 2004 exit polls be trusted anywhere? I hope someone with more time than me will see what sorts of problems there were in other states as well.