Mark Blumenthal | October 21, 2008
With two weeks to go until the election, the pace of new poll releases is accelerating, and while yesterday's batch of new polls produces some notable shifts and at least one odd outlier, the overall pattern is stability.
We logged 19 new statewide polls yesterday. Of the new polls that update results collected by the same pollster earlier in October, most as usual showed very small, non-significant changes although the overall pattern hints of a slight improvement for McCain: 7 of 11 show slight shifts to McCain, 4 show slight shifts to Obama
The impact of all of the new polls on our trend estimates in the more competitive states was mixed, showing a pattern that suggests random statistical noise. The margins moved in Obama's direction in 6 states and in McCain's direction in 5. Over the last week, the shifts are still slightly in Obama's favor: 14 move toward Obama, 5 to McCain and one unchanged.
The biggest shift yesterday came as the result of two new automated surveys in Virginia from SurveyUSA and Rasmussen Reports showing Obama with large leads -- 6 and 10 percentage points respectively. The new polls increase Obama's lead on our trend estimate to nearly nine points (52.4% to 43.5%), enough to qualify for dark blue "strong" Obama status on our map.
[Update: The orginal version of this post reflected a data entry error for the latest Fox/Rasmussen poll in Florida. The correct margin -- McCain 49%, Obama 48% -- dropped Obama's lead on our margin to 3.3%, just enough to shift Florida and its crucial 27 electoral votes back to "toss-up" status=.
Yesterday also produced at least one example of the sort of outlier that leaves poll watchers scratching their heads. Rasmussen/Fox News reported a new Ohio survey showing McCain leading by 2 point (49% to 47%) representing a slight but non-significant improvement for McCain since their survey last week. At the same time, Suffolk University released a new survey showing Obama leading by nine points (51% to 42%). Our trend estimate performed as intended, muting the impact of the Suffolk result and leaving Obama with a small advantage (48.4% to 46.7%), well within the "toss-up" range.
Finally, I want to look more closely at the national trend, which has been receiving quite a bit of attention over the last few days. Our overall trend estimate, based on all available national polls, shows the same pattern seen by other aggregators: A one and a half point upturn in McCain's total from a smoothed low low of 42.2% a little over a week ago to 43.7% as of this writing. Meanwhile, Obama's total has dropped about a half point, from a high of 50.2% a week ago to 49.7% now.
Several observers -- from Pollster reader DTM to the Washington Post's Ben Pershing -- have speculated that the drop appears to result mostly from a change in polls influencing the trend estimate over the last week (an issue also summarized yesterday by the Wall Street Journal's Carl Bialik). We saw a pause last week in new releases of by non "daily-tracker" national pollsters, presumably an issue of timing around the final debate. Since, as I noted yesterday, three of the trackers typically show a slightly closer margin, perhaps their greater influence has caused the narrowing margin.
One reason why I'm hesitant to dismiss the slight narrowing altogether: If I use our chart's filter tool to look at only the six daily trackers that have been tracking since the beginning of October (which excludes the new IBD/TIPP tracker), which produces a much more "apples-to-apples" comparison, we still see a slight narrowing of the trend.
Still, these are relatively small changes and, as of yesterday at least, do not appear to represent a continuing trend. Yesterday, four of six trackers showed slight shifts to Obama, two to McCain since the previous release on Sunday. If we add the GWU/Battleground tracker (which doesn't interview or release on weekends), and compare to Thursday (the last with interviews conducted before Wednesday's debate), we see three surveys shifting to Obama, three to McCain and one showing no change.