I have a noon-time speaking engagement today, so this update will need to be brief and a little earlier than usual.
As of this We have logged 22 new statewide polls since this noon on Friday, the bulk of which came from an 11-state release of Zogby's surveys drawn from their non-random online panel. The net result, as of this hour, shifts our classification for just a single state: New Mexico shifts from lean Obama to toss-up, lowering his electoral vote total on our map to 238.
We also had two new surveys in Minnesota, one from SurveyUSA and another from the Minneapolis St. Paul Star Tribune and PSRA. Both show a close race in both the Presidential and U.S. Senate campaigns, narrowing the presidential trend estimate slightly (though leaving our classification unchanged as lean Obama) but shifting the Senate race to a toss-up.
We get asked often about including the Zogby Internet surveys given their problematic performance. The short answer is that the regression trend lines we plot tend to minimize the impact of a single "outlier" result. We get into difficulty when one pollster contributes a disproportionate number of points to the chart and shows systematic differences from other surveys.
Outliers can sometimes exert more influence on the end point of the chart. In this case, the Zogby result in New Mexico (showing Obama leading by just 2 points, 46% to 44%) nudged the Obama lead down slightly on the trend estimate (from +5.4 to +3.9), just enough given to move New Mexico into the toss-up category. The sample sizes are often smaller in New Mexico than in other states, so the margin necessary to color classify a state as leaning to a candidate tends to be slightly bigger. Keep in mind that another survey last week from Rasmussen Reports showed McCain with a two-point edge (49% to 47%) in New Mexico, which also helped narrow the current estimate.