http://www.federalreview.com/compositepoll.htmTHERE IS A BOUNCE
Bush 47.4% – 242 EV | Kerry 50.4% – 296 EV
EV without Toss Up states (under 2% margin): Bush 200| Kerry 253| Toss 85
August 3, 2004
With conflicting results in different polls and most showing changes within the margin of error, it's difficult to tell if Kerry/Edwards got any bounce from their convention. Clearly, they did. Although it is certainly not as large a bounce as we have seen historically, that should be no surprise with as few people reportedly undecided as we have in this year of polarization.
Although many polls show changes within their own margins of error, when several polls move in the same direction, the odds are that there has been real movement. The meta analysis techniques employed in the Composite Poll confirm that movement. From the pre-Convention Composite Poll to Tuesday, August 8, Kerry gained 1.4 points and Bush lost 0.5. That equals the biggest one week swing since we started this analysis on March 8. Kerry now leads by 3 points. A respectable lead in such a close race.
While much has been made about the odd results of the USA Today / Gallup Poll, I note that it has traditionally been the most stable poll this year, and its weekly variance from the weekly mean has been generally to Kerry's favor by 0.4 points (i.e., it is barely more Kerry favorable on the average). Any suggestion that the Gallup poll is generally pro-Bush is simply untrue. However, it is possible that this week's poll is a bad sample. But the nature of polling is such that 5% of the time, you can get ridiculous results. Alternatively, there's a chance that Gallup is the only one with the correct turnout model. Personally, I think it's reasonable to discount this week's results as an outlier - showing Bush with a 4 point lead.