After reviewing the state polling results for eight-one polls in the fifty states and the District of Columbia, I have found the following results for apparent bias.
Eighty-one polling agencies or groups have performed polls in the states during the 2004 U.S. Presidential Election campaign since the primaries. As noted in an earlier post, I drew up the average poll results by state, and was thus able to determine a bias point for each poll in a state. I decided to draw up trends for only those polls which take surveys in at least three states. Thirteen polls met that requirement, and this post shows their results.
I have ranked the polls in order of Average Variance from the state average in those states they polled, also noting the widest variance (average and widest variance gives you a decent idea of a realistic Margin of Error), as well as average bias for/against President Bush and Senator Kerry.
1. Strategic Vision - Average Variance Poll (AVP) 2.0%, Widest Variance Poll (WVP) 4.8%, Bush -0.4% Average Variance Bush (AVB), Kerry -0.5% Average Variance Kerry (AVK).
2. Rasmussen - AVP 2.4%, WVP 5.7%, AVB -0.4%, AVK -0.2%.
3. U/Minnesota, HHI - AVP 2.6%, WVP 3.7%, AVB +0.5%, AVK -1.5%.
4. Research 2000 - AVP 2.8%, WVP 5.1%, AVB -1.2%, AVK -0.2%.
5. Survey USA - AVP 2.8%, WVP 9.4%, AVB +0.9%, AVK +1.2%.
6. Zogby Interactive - AVP 2.9%, WVP 5.4%, AVB +0.7%, AVK +1.7%.
7. Quinnipiac University - AVP 3.3%, WVP 5.7%, AVB -0.8%, AVK -2.0%.
8. American Research Group - AVP 3.3%, WVP 6.4%, AVB -0.8%, AVK +0.1%.
9. Mason-Dixon - AVP 3.4%, WVP 5.8%, AVB -0.1%, AVK -2.4%.
10. CNN/USA Today/Gallup - AVP 5.2%, WVP 8.6%, AVB +3.0%, AVK -2.7%.
11. LA Times - AVP 5.2%, WVP 9.8%, AVB -1.9%, AVK -2.7%.
12. MRI - AVP 8.0%, WVP 13.6%, AVB -1.5%, AVK -2.0%.
13. Fox News - AVP 8.3%, WVP 10.5%, AVB +0.5%, AVK -7.0%.
Details to follow in my next update. For now, note that while not impressive in its accuracy, the Zogby internet polls actually hold up fairly well. Fox News and MRI, however, have abysmal overall accuracy, they trend to be outliers.