Gallup did something unprecedented today, and while they may have had good intentions, it was egregious nonetheless. As most of you know, they took the extraordinary step of breaking their 7-day tracking poll into two 3-day periods, and lopped off the seventh day.
I think they did this for a couple of reasons. First, their 7-day window is simply too long to reflect rapidly changing events - and the debate was surely that. While their 7-day tracker showed Obama with a +2 lead as of Sunday (Saturday to the prior Saturday), their data from the 3-days immediately following the debates was even. So they announced at 10 am this morning that the three days since the debates were even, but the three days prior to the debates were Obama +5. From what we saw from Pew, PPP, Ipsos, and Rasmussen, this is probably right on the mark. It is fairly obvious that Romney had a good day Thursday, a monster day Friday, and then Obama recovered somewhat on Saturday and much more on Sunday. I would bet that any poll spread evenly from Thursday-Sunday would show a tie in the likely voter model (The bulk of the Pew sample was Thursday, Friday, and Saturday).
So our friends at Gallup already had an issue where their poll did not reflect a rapidly changing landscape, so they made changes in their methodology to reflect this. That I have no problem with so long as they disclose it - which they did. They are not responsible for a lazy media that does not report it accurately.
So where is the malpractice. Yesterday was a very good day for the President. We saw it in Rasmussen where it is easier to extrapolate daily results due to their tighter 3-day band. But since Gallup now gave us their two most recent 3-day band, we could do the same thing. Obama had a monster day in Gallup on Sunday. If the average of the 3-day pre-debate sample for Obama were +5, and the average of the three days immediately following the debate were even, Obama would have to have something like a +19 day yesterday to get a +5 seven day sample. Now Gallup they may feel that the pro-obama monster Sunday number to be an outlier, especially if they are also in the field with the USA Today poll, but to not disclose the big Obama day and essentially bury it in the 7-day number that still includes three pre-debate days is unconscionable.
Please note that I do not have any issue with Gallup's numbers - I think they are right on the mark. When you consider that they are still in a registered voter model and the likely voter screen probably nets 3 - 4 additional points for Romney, they are consistent with pew over the same 3-day period. My point is that they took great pains to highlight the Romney debate bounce, and then did nothing to show that the bounce was likely receding. In fact, they obfuscated it!
My advice to Gallup is that if events on the ground warrant breaking the 7-day poll into two parts, they continue to run to multiple tally's until the period following the event catches up to the 7-day number. That is, when they released the 7-day number at 1:00 today, they should have also released the 4-day subset of post debate days. Tomorrow, they should release the 7-day number and the 5-day post debate subset, and continue this until the mini-tracker aligns with the full tracker.