Obama surge in Gallup guaranteed after debate tonight. Bank it.

[geek tragedy]

Why?

Nothing to do with the debate, but rather the fact that Romney's 7-day average got thrown off by 1-2 outlier days--specifically the results from 10-15 and 10-16. Those two days produced the ridiculous spike in Romney's LV edge, as well as showing Romney pass Obama in RV.

The 15th rolls off tomorrow. The 16th rolls off on Wednesday.

Expect a big swing--3-5 points--for Obama as those days are replaced by normal days.

And get prepared to spin it as a surge for Obama, momentum, etc.

[woolldog]

1. how do you know it's those 2 days?

[geek tragedy]

2. Just look at the daily results. They spike in favor of Romney on those two days, and then hold

constant.

Obama was up 48-46 in RV for the interviews through the 14th. Two days later he was down 48-46.

Similarly he was down only 2 in LV 49-47 for interviews through the 14th. Two days later he was down 6.

[FBaggins]

4. You assume that seven days prior to that were "normal" days.

How do we know that part of the bump wasn't caused by a particularly good day for the president (form 14 days ago) dropping off?

[geek tragedy]

5. Obama had lead in RV for two straight weeks before those two days entered the sample

They were last tied on September 20. Obama lead then until the day the 15th of October results were included in the 7-day total.

[FBaggins]

7. Sorry... you've ignored the point.

You can get fairly substantial 1-day numbers that don't show up in the rolling average because of rounding (and there's LOTS of variability that can be lost if your only standard is "had a lead in RV" rather than specific numbers)... then any one-day movement could also be the result of a change today or a change in the other direction a week ago dropping off (or both).

A move as large as what we saw certainly implies that the most recent day was much worse than the one seven days prior, but we don't know whether that's because it's an objectively bad day... or just that the one from a week ago was better than the neighboring days.

Let's take an example. Say you have six straight days of exactly the same data. The president leads 47.4% to 45.6% (but that gets reported as 47-46. The next day shows a big bounce for Romney (51.2-42.6 one-day sample). That's a really big day for Romney... but the headline number still reads 47-46 for the president. The number could be thrown way off a week from now when this one drops off, but you won't know whether it's because that day was out of whack... or that it replaced one that was out of whack. So you won't know for sure what to expect two weeks from now when that number falls off.

[geek tragedy]

11. When you have 14 straight days of relative steadiness followed by an abrupt

switch to one candidate over two days, followed by more steadiness, that clearly indicates that the two days were especially beneficial to Romney.

The fact that the numbers were so steady in RVs indicate that no days were particularly good or bad--if they were, they were synced perfectly on a weekly basis (unlikely).

Just watch--there WILL be movement in the next two days towards Obama.

[FBaggins]

12. So you didn't read what you responded to?

I just gave you an example of days that remained perfectly steady even though the underlying data shifted dramatically.

We didn't see "14 straight days of relative steadiness" preceeding those two days. We saw significant volatility. You're just glossing over it by calling any day that the president leads among RVs a "steady" day.

[geek tragedy]

14. Look at the weekly numbers

october 8-14: Obama +2
October 15-21: Romney +1

Where did the movement occur: October 15 and 16--that's where the lines cross. And then stabilize.

The race was stable enough for one candidate to hold the lead in 14 straight days.

We'll see who is right tomorrow.

[MjolnirTime]

13. nice attempt, fbag

[SCliberal091294]

3. Correct...

On one of those days Romney surged 4 points. The next day he gained 1-2. It will be a virtual tie in national polling.

[yellowcanine]

6. May be but I still think Gallup is doing something different with their likely voter model

which is skewing their results.

[geek tragedy]

8. The movement was actually in the RV totals, if you examine the poll.

There was a +4 net swing to Romney in the RV totals and a +4 swing in the LV totals. So, it wasn't that extra Obama voters were being screened out, it was that the daily samples were especially pro-Romney from the outset.

Gallup's LV screen is +4-5 in favor of Romney in general. A more reasonable number would be +2-3, but that's a separate issue.

[budkin]

9. Honestly I feel like it's going to increase for Romney because they make no sense

[bushisanidiot]

10. Romney has no where to go but down (from gallups pumped up numbers)

once gallup starts to align with the rest of the (real) polls being conducted, to save face of course,
romney will be lagging about 3 points behind the president in averaged polling.

romney is liable to do something very desperate (and embarrassing) in the next week.
watch for it.

[woolldog]

15. So did the surge materialize today?

[geek tragedy]

16. Numbers not up yet--annoyingly enough. nt

[MSMITH33156]

17. They are unskewing them

as we speak.

[woolldog]

18. lol

[geek tragedy]

24. Obama gained +1 in LV, Romney unchanged

Disappointing, but we'll see what tomorrow brings.

[woolldog]

25. grrrrr

[Jennicut]

19. Romney down one in LV and RV. Obama up one in RV.

No idea what that means though.

For Gallup, all I look at is their RV since the LV screen is ridiculous.

48 47 R O RV
51 45 R O LV

[geek tragedy]

20. Those are yesterday's numbers.

Poll addicts are jonesing for their Gallup fix.

[FBaggins]

21. That's yesterday's data

Still waiting on today's 1pm update

[Jennicut]

22. Sorry.

What is taking them so long? They can't make it look good enough for Mittens.

[FBaggins]

23. RV no change. LV 1-pt gain.

48-47 RV
51-46 LV

[geek tragedy]

27. Apparently the 16th was the really terrible day of the two.

Obama gained +2 in RV and +2 in LV after it dropped off.

[FBaggins]

29. Again... you can't assume that without Gallup reporting specifics

There are three days involved here. Last night's data... the data from the 16th... and the data from the 9th. (Then all of that can be muddled by rounding issues - but we've gone over that already)

You see a move in one direction on a given day (the 16th in this case), but you don't know how much of that move is due to a "good night" on the 16th and how much of it might have been a bad night dropping off from the 9th. Now we see a good move in today's report... but we can't say how much of today's move is due to the 16th falling off, and how much of it might be due to the president actually gaining last night (due perhaps to performance in the final debate?).

If the polling universe were more consistent, you could tease a little more out of the reported data by comparing the figures to the tracking polls that run on three-day averages... but the polling universe is anything but consistent.

[geek tragedy]

30. October 1-7: Obama +5 in RV, October 8-14: Obama +2 in RV

October 17-23: Obama +1 RV

The only days Romney lead in RV were those days including the 16th. The only days showing him tied or ahead were those including the 15th and/or the 16th.

We can engage in hypotheticals about perfectly correlated pairs spaced 7 days apart as producing a stable race, or we can look at Occam's Razor and conclude that those two days were between them VERY good for Romney.

[FBaggins]

35. And now you're conflating week-long periods with discussions around single days.

You also seem to regularly oversimplify moves as "we led" or "we were behind" as if that means things didn't change... as if a two point drop from a 3-pt lead to a 1-pt lead is statistically less significant than a two point drop from a 1-pt lead to a 1-pt deficit.

or we can look at Occam's Razor

Except that that isn't what you're doing. Your explanation isn't simpler... in fact it's more complex. Occam's razor would tell us to look at today's report an conclude that last night's data was better for the president than the week-old data that fell off. It would not then try to decipher whether that's because one day was unusually good... or one day was unusually bad...

... or the far more important possibility - that the race may have actually moved in our direction (IOW, that it wasn't a statistical anomaly at all).

The move wasn't as large as you expected. Holding on to your prior position forces you to then believe that the underlying reality of the race (the "true" move that wasn't statistical "noise" clearing) is that we're slightly losing ground in the two days of polling that are post-debate.

Do you really want to believe that? I sure don't.

[geek tragedy]

36. Here's what I projected:

Expect a big swing--3-5 points--for Obama as those days are replaced by normal days.

Obama is up +4 in RV over the past three days, plus 2 from yesterday to today alone.

Obama is up +3 in LV over the past two days.

What we know is that the 15th and 16th were much worse for our side than either the 8th and 9th or the 22nd and 23rd were. With no reason to suggest that 8-9 and 22-23 were abnormal, the logical inference is that the 15th and 16th were especially bad.

[FBaggins]

37. Right... and we didn't get it.

Obama is up +4 in RV over the past three days

So now a three day move is the result of two bad days? You expected 3-5 points in two days and we have an RV shift of 1-3 points in those two days.

You said on Monday that we would know who was right on Tuesday. You had been focusing on RV and that didn't change at all. So rather than realize that this is far more complicated than you assumed, you just decided that only one of the days really counts (or that now it's three)?

What we know is that the 15th and 16th were much worse for our side than either the 8th and 9th

Very good. Just as I said. What we don't know is to what extent that was something off on the 8th/9th or something off on the 15th/16th. Or a bit of both or a change in the race itself... or an artifact caused by rounding (or some combination of all of those). I point you back to Occam. You shouldn't be trying to read a "why" into something that does not have enough data to judge.

I point out further that smart people have been trying to do this for years now (back-fitting a rolling average to guesstimate what the "real" number is)... and as I said earlier - it's can't be done without the pollster giving you specifics for at least some of the days. You can look back at this forum to see predictions of which way polls were going to move (e.g., because the two days after the first debate were the only real Romney bounce and they were about to drop off). They have well below a 50/50 reliability.

We don't really know that yesterday saw no RV movement. It could have been more than a point... but hidden in the rounding. Similarly... today's 2-pt move could be significantly larger... or just a few tenths of a percent. Because Gallup doesn't report to the tenth of a percent. All those days that you grouped together actually saw several reported shifts (that inexplicably you're ignoring because the lead didn't change hands).

With no reason to suggest that 8-9 and 22-23 were abnormal

Sorry... that too makes no sense. There isn't any external reason to believe that ANY of them were abnormal. You can't say that something was fishy and then declare that two of the three were not fishy... so the third must be.

So let me ask you the question again: Do you believe that the president lost ground since the final debate? That the movement we've seen thus far in a positive direction is merely the result of even worse days falling off?
Yes or no?

[geek tragedy]

26. Update: Obama retakes the lead in RVs, closes to within 3 on LV.

Obama 48, Romney 47 in RV
Romney 50, Obama 47 in LV

Two days ago, Romney lead LV 51-45 (+3 Obama shift)

Three days ago, Romney lead RV 49-46 (+4 Obama shift)

[boingboinh]

28. K&R in respect

[woolldog]

32. Good call.

Can you give me some good lotto numbers for Friday?

[DemocratSinceBirth]

33. Kudos To Geek Tragedy

For his prediction. The doubters in this thread aren't only licking their wounds, they're licking each others.

[FBaggins]

38. And now?

Today's poll is virtually indestinguishable from when the OP was posted.

Did Romney have another pair of huge days, or is it time to recognize that we can't backfit 7-day rolling averages?

[geek tragedy]

39. Yesterday was a crummy day of polling for Obama--his approval rating

dropped a net six points.

One can consider whether the occurrence of so many crummy days bodes ill for his campaign.

[Johnny2X2X]

34. Love it

Really helping me breathe easier, the approval numbers are extremely strong and should they hold up throughout this week, the other numbers will follow. Obama could gain 4 or 5 more points by week's end IMO.