Using Gallup's Own Numbers Obama Was Up 59% -41% In Sunday's Subsample

[DemocratSinceBirth]

It was 47% -47% with samples from 10/4, 10/5 and 10/6:


http://www.gallup.com/poll/157907/romney-narrows-vote-gap-historic-debate-win.aspx


It was 50%-45% with samples from 10/4, 10/5. 10/6, and 10/7:



http://www.gallup.com/poll/election.aspx


For Obama to go from a 47% -47% tie on the third day of a sample to a 50%-45% lead on the fourth days of a sample the sample days have to look essentially like this:



It would have been

47-47 10/4
47-47 10/5
47-47 10/6
59-41 10/7

[mzmolly]

1. Is that even reasonable?

Seems like quite a swing, in a single day? Especially without a major scandal or???

[DemocratSinceBirth]

3. I'm Just Playing The Game Gallup Did

They release seven day samples. All of a sudden they decided to release a three day sample that showed the race tied. However their seven day sample shows a 50% -45% Obama lead. I decided to do a four day sample, using Gallup's own numbers. To get to 50% -45% in a four day sample where they were tied 47%-47% after three days Obama would have to won the fourth day 59%-41%.

[mzmolly]

4. Gotcha.

Thanks.

[elleng]

2. So you're suggesting that media

is misunderstanding/misinterpreting/misreading the Gallup numbers, not actually noticing the days corresponding to the numbers? And that MSNBC among others should be so informed?
(Trying to understand.)

[BumRushDaShow]

6. YES - MSNBC JUST corrected the poll 5 minutes ago. nt

[elleng]

10. Thanks. Good.

NOW for the Headlines!

[DemocratSinceBirth]

7. Romney Closed To A Tie On 10/4, 10/5 , And 10/6 And Got Crushed On 10/7

.

[nsd]

5. The 50-45 split is from a 7-day average.

It includes not just 10/4 - 10/7, but also 10/1 - 10/3.

ETA: We don't know what the 10/4 - 10/7 average was.

[DemocratSinceBirth]

9. More

Got This From Intrade- Could Obama Have Been Up 56% -42% In Sunday's One Day Gallup Sample?

Last edited Mon Oct 8, 2012, 05:58 PM USA/ET - Edit history (1)

I'm surprised nobody posted this:

From Gallup's Data, the split was 47/47 post debate and 50/45 pre-debate

So the split looks like:

9/30: 50/45
10/1: 50/45
10/2: 50/45
10/3: 50/45
10/4: 47/47
10/5: 47/47
10/6: 47/47

This would give you an average of 49/46, as it was before today's numbers.

Now we have:

10/1: 50/45
10/2: 50/45
10/3: 50/45
10/4: 47/47
10/5: 47/47
10/6: 47/47
10/7: ? / ?

For the rolling average of 50/45 that was posted today, today's data must be much more heavily favorable towards Obama to make the numbers work. 56/42 is most Romney-favorable number that makes this work

[nsd]

17. But this assumes that the split was 50-45 on each of the three days of the pre-debate average.

We don't know that to be true. All we know is that yesterday (10/7) was better than last Sunday (9/30).

[DemocratSinceBirth]

20. On The Eve Of The Debate The Pres Was Up 49%-46%

.

[nsd]

23. But isn't that a 7-day average including several days from the week before the debate?

It doesn't tell us what happened on Sunday, 9/30 which is the crucial thing.

[central scrutinizer]

21. assumptions?

Are your numbers what you assume each day's individual polls were?

If so, then I doubt that each of those seven days had the same results. I would expect variation due to sampling variability and it might be that 9/30 was a relatively good day for Rmoney and when that fell off the seven day calculation it would bump up the rolling average. Which would mean that Sunday was a good day for Obama, but not 56/42 good. We should know a lot more in a couple more days when the Thursday (day after debate) poll falls off the seven day average.

[cojoel]

26. this version seems correct for a 7-day simple moving average

But I can't find if that is what Gallup uses, of if they use a moving average that is weighted, or modified, or some other approach.

[B2G]

11. Isn't the 50-45 number for the 3 days BEFORE the debate?

And it went to 47-47 AFTER the debate?

Am I not reading this correctly?

[DemocratSinceBirth]

13. Gallup Released A Three Day Tracker That Excluded Sunday That Showed A Tie

Gallup then released a seven day tracker that included Sunday that showed the pres with a huge lead.

Conclusion-The president had a huge Sunday.

[B2G]

15. What happened on Sunday?

I really don't get some of these wild swings. And I wish more of these polls would release the demographics about who responded.

[BumRushDaShow]

16. It was the full result of the lower jobs numbers

the lower deficit (down by $200,000,000,000) and other positive economic news that was conveniently left out of the polling and "tie" narrative.

[DemocratSinceBirth]

18. Each Day's Sample Only Consists Of Four Hundred Fifteen Or So Respondents So There's A Huge MOE

That being said, just willy nilly taking three days out of that sample which is what Gallup did with their shock poll strikes me as irresponsible.

[BumRushDaShow]

25. And for the media to run with that to show a "tie"

against all other indicators, was an even WORSE decision. And the sad thing is that so many on DU just piled on.

As they say - "There are lies, damn lies, and statistics".

[nsd]

14. It's both.

Early this morning Gallup released 3-day averages for before and after the debate. These showed 50-45 and 47-47 respectively.

At 1 pm EST, Gallup released a 7-day average showing 50-45.

[OldDem2012]

8. Gee...I wonder what the Romney camp is thinking now? nt.

[elleng]

12. 'Thinking' that We the People are THINKING???

[woolldog]

19. That's not accurate. nt

[DemocratSinceBirth]

24. On The Eve Of The Debate The Pres Was Up 49%-46% In The Seven Day Gallup Tracking Poll

Gallup has stated they were tied at 47% on 47% 10/4, 10/5, and 10/6. To be at 50%-45% at the end of a tracking poll that began 10/1 , the data suggests Obama beat Romney in yesterday's sample by fourteen points or more.

[woolldog]

27. Okay

but how do you know what the 3 day average was for the 3 days before the debate?

We know the 3 day average after the debate = 47

We know the 4 day average for 3 days before debate + Sunday = 52.3

You're assuming what the numbers are for the first 3 days. Just because the average on the eve of the debate was 49 or 50, doesn't mean the average for the 3 days leading up to the debate was 49 or 50.

Right? Am I missing something?

[reformist2]

22. The Gallup tracking poll is a 7-day not a 4-day.

Having said that, Obama's Sunday number could be anywhere from 54-61.