Can we deduce the size of the upcoming Gallup bounce based on today's Job Approval and head-to-head polling data? I think some inferences can be drawn. Here are recent Job Approval numbers from Gallup:
Actual Gallup Numbers
3-Days Approve Disapprove Margin
08/28-30 45 46 -1
08/29-31 45 46 -1
08/30-09/1 43 48 -5
08/31-09/2 45 48 -3
09/1-3 45 48 -3
09/2-4 47 47 +0
09/3-5 49 45 +4
09/4-6 52 43 +9
What we know from these 3-day averages is that the polling for Thursday specifically must have been about 15 points better than the polling for Monday. Because when Thursday's number was added and Monday's number dropped out of the average, Obama's lead grew from 4 to 9, a gain of 5 points. To move the average 5 points would require a one day improvement of 15 points. So we can work backwards and make a guess about the specific days numbers. Like this:
Specific Day Estimates
Date Appr Disappr
9/2 46 48
9/3 47 47
9/4 48 46
9/5 52 42
9/6 55 40
If you compute 3-day averages using my estimates for each day, they should work out to match the numbers in the above chart.
My only conclusion is that Thursday was an awesome polling day for us and Wednesday was pretty darn good too. Friday's number will replace the 48/46 numbers for Tuesday. If it's anything like the last 2 days, the Approval Gap could go from +9 to maybe +12.
But what does this tell us about the 7-day-tracker pitting Obama against Romney? Is it reasonable to assume that a rise in Job Approval for Obama will also move his head-to-head number? I think it's reasonable but it's hard to compare given the 3-day window vs. the 7-day window. And sometimes these numbers don't always move in the same direction. One reason they don't is that asking people about their Obama job approval doesn't say anything about Romney. Over the last several months there have been times when Obama job approval lagged but his lead over Romney remained or even edged up, based mostly on negative news about Romney.
Nevertheless, as we go closer and closer to election day these numbers will likely move in concert. Here are the recent Gallup head-to-head numbers:
Official Gallup Numbers
Date Range O R O lead
08/22-28/2012 47 46 1
08/23-29/2012 47 46 1
08/24-30/2012 47 46 1
08/25-31/2012 47 46 1
08/26-09/1/2012 47 46 1
08/27-09/2/2012 47 46 1
08/28-09/3/2012 47 46 1
08/29-09/4/2012 47 46 1
08/30-09/5/2012 47 46 1
08/31-09/6/2012 48 45 3
Using the same logic as above, how big does Thursday's number have to have been relative to the departing Thursday number from a week ago to move the 7-day average here up 2 points? Answer: about 14 points better than last Thursday.
So, a poll that didn't budge for 9 days improved 2 points for Obama in a single day. Day-to-day numbers could look like these:
8/28 47 46
8/29 47 46
8/30 47 46
8/31 47 46
9/1 47 46
9/2 47 46
9/3 47 46
9/4 48 45
9/5 49 45
9/6 55 40
While it's unlikely the daily numbers were quite that smooth during the RNC we know they didn't vary much or the average would have shifted a little at some point. I have Thursday's number 14 points better than the previous Thursday number to account for the sudden 2-point improvement we saw in today's release.
You'll notice the last numbers, yesterday's, are the same in both the Approval tracker and the head-to-head tracker: 55 and 40. While the "real" numbers may not be identical I do believe they are somewhere near this magnitude and likely similar to each other.
So how much will the Gallup tracker move? It depends on how many days like Thursday show up in their results. Nate Silver suggests bounces can last weeks. If so, we may see a sizable bounce number from Gallup. Possibly nearing double digits. Of course, other pollsters (yes, Rasmussen) aren't going to show as big a move. The average of the bounce numbers might be around 5 according to Nate's latest post. All three tracking polls, even Rasmussen, showed improvement for Obama today.
The other question is how much of the bounce will fade away. If voters solidified their decisions over the last 2 weeks, some of the bounce may not be a bounce at all but a new normal.