Rick Brady of Stones Cry Out has two new takes on the difference between the 2004 exit polls and the election result were statistically significant within the 11 battleground states:
He’s got that unadjusted exit poll data (at least for 11 states) everyone is looking for…
Based on a quick look his data (Exhibit) it sure seems likely that the exit poll vs. election result discrepancy was a systemic bias of around 3% against Bush across the board. Notice how Kerry’s poll and actual percentages are remarkably consistent across the battleground states. In nearly all cases the Bush percentage is under reported by right around 3%.
Anecdotally we have heard that even back to the 2002 midterm elections the Republican vote was consistently under-polled in the exit polls, and the 3% number was usually mentioned.
It would be interesting if we could determine the causes of this 3% under estimation, but based on the last two elections exit pollers in the future might just want to go ahead and add 3% to the Republican percentage.