United States presidential election, 2004, exit polls

The exit polls were conducted by Edison Media Research and Mitofsky International for the National Election Pool. The NEP has a policy of adjusting the final polls to match the vote as reported by state and local governments. Due to this policy later polls may be tainted so as to hide election fraud. Women do seem to be either more likely to participate in a poll, or more likely to vote early. In the early polls women represented 58% of the sampled voters, by poll closing only represented 52% of the voters as reported by local governments, so women were likely oversampled in the early polling. The 52% figure is consistent with historical turnout numbers. Blacks may also have been undersampled, since black turnout increased 25%, much greater than the increase in turnout by other groups.

NEP has not made the raw numbers and sampling methods available to pollsters to study. Historically, exit polls are accurate at predicting the winner. This is because the largest source of error in conventional polls is in determining a probability a voter will show up in a polling place and vote. Exit polls do not have to cope with this variable. However the difference between the polls and the vote as reported by state and local governments in Florida, Ohio, and Pennsylvania was substantial, with one analysis initially claiming that the probability that it would be due to random error was 1 in 250,000,000. (This figure was later revised to 1 in 662,000.) This has led Republicans and Democrats alike to cry foul. Karl Rove has implicated a vast left wing conspiracy to tilt the exit polls, while Democrats have cast doubt on the accuracy of the un-auditable Diebold Election Systems and Election Systems & Software DRE software written by Bob Urosevich.

Last CNN exit polls
State Bush (%) Kerry (%) Sample Size Predicted
Differential
Florida 49.8% 49.7% 2846 Bush 0.1%
Colorado 49.9% 48.1% 2515 Bush 1.8%
Iowa 48.4% 49.7% 2502 Kerry 1.3%
Michigan 46.5% 51.5% 2452 Kerry 5.0%
Wisconsin 48.8% 49.2% 2223 Kerry 0.4%
Minnesota 44.5% 51.5% 2178 Kerry 9.0%
Nevada 47.9% 49.2% 2116 Kerry 1.3%
Ohio 47.9% 52.1% 1963 Kerry 4.2%
New Mexico 47.5% 50.1% 1951 Kerry 2.6%
Pennsylvania 45.4% 54.1% 1930 Kerry 8.7%
New Hampshire 44.1% 54.9% 1849 Kerry 10.8%