Updated graphs on the horse race:
Note: regarding the results drawn in for the past week (week 40), those of GWU, LA Times, Democracy Corps (all up for Bush) and Rasmussen weekly (no change) date from before the debate; those of Zogby, Gallup and Newsweek (all down for Bush) date from after the debate.
This is what the resulting average looks like:
Methodological note: I hadn't thus far ever included Rasmussen in these graphs, since they do a daily tracking, and how do you process a daily tracking poll into ten-day updates? Calculate averages yourself? Unh-huh.
But since I have shortened the intervals in the graph to week-long rather than ten-day periods in this closing stage of the race, and since Rasmussen
does provide week-by-week polls as well, I decided to include those, at least, after all.
It would have been rather odd to start including Rasmussen only from the moment that I changed to weekly intervals, though, especially since that happened to coincide with the post-RNC Bush bounce. That could have looked just a little partisan, since the Rasmussen numbers are so stable, and thus served to dampen that bounce in the average numbers. So I went out on a limp and included them, after all, all the way back to January - just taking the average of two of those week-by-week numbers whenever the ten-day interval forced me to. Not really scientific, but what do you do. That explains, in any case, why the past (average) numbers now may look slightly different from what you might remember.