Why is RealClearPolitics averaging in the Daily Gallup results more than once?

[Brotherjohn]

I'm confused as hell and don't know WTF they're doing.
http://www.realclearpolitics.com/epolls/2008/president/us/general_election_mccain_vs_obama-225.html

They are including one set of numbers derived from the daily Gallup polling, as a separate poll (labelled "Gallup Tracking"). This is an interpretation of the Gallup Daily Results (averaging the 2 LV scenarios), and RCP, at least, is pretty clear on that.

Then there's another "separate" poll, also averaged in their "RCP Poll Average", which appears to be another interpretation of the same data (labelled "USAToday/Gallup*"), though they're not as clear on how they derived that one. You jump to the link of the first one, and RCP explains themselves. You jump to the link for the 2nd, and you get a USAToday page that explains nothing. I have no idea how they got that number. It's obviously an average, but it doesn't work if you just average the 2 LV and the RV numbers.

What is clear, though, is that RCP is treating one poll as two. And they seem to be planning on doing this every day, tilting the average to whatever Gallup results show. Gallup's info appears to indicate that these LV's are just a subset of the RVs. But even if they were not, the 2nd poll included in the RCP average is NOT simply the RV poll.

So Gallup is now releasing 3 polls a day (or 3 interpretations of the same polling data: RV, LVI and LVII). Fine, if they want to. But if you're a polling site keeping track of a "poll average", I'd think you just use one of those by some consistent criteria. And for RCP, that would be the one they have been using all along for the Gallup Daily: RV.

UPON FURTHER READING, RCP says (in the link from the "Gallup Tracking" poll, then linked to "for more explanation" at the bottom): "In order to not double count the Gallup Daily Tracking poll survey results in the RCP National Poll Average", they are averaging the two LV scenarios. BUT THEN they STILL include this LV average, AND a second version of the Gallup Daily numbers!

[Brotherjohn]

1. Okay, WORSE! Now they're including FOUR versions of the same polling in their current RCP Average!!!

http://www.realclearpolitics.com/epolls/2008/president/us/general_election_mccain_vs_obama-225.html

[Undercurrent]

5. What?

I double down on my last post. They're flat dishonest!

[Undercurrent]

2. Sheesh!

Thanks for explaining what they're doing at RCP. Pretty shady stuff.
I wondered about this last night:
http://www.democraticunderground.com/discuss/duboard.php?az=show_mesg&forum=132&topic_id=7435546&mesg_id=7436739

[Brotherjohn]

6. "Explaining"?! I STILL don't understand!!

No matter how I crunch the numbers, I don't get what's listed in their current table to match their current average.

I thought maybe they're just LISTING all the Gallup variations, but only AVERAGING one variation... but the numbers still don't add up.

???&**^%#*$I???

[Schulzz]

3. I don't know about Gallup, but I have the impression that he (RCP) tends to favor Republicans a bit.

He didn't use PPP's Colorado poll (Obama +10) or Democracy Corps's national poll, but uses every Strategic Vision poll.
He also doesn't include R2000 tracking, but every other tracking poll, even added a new one yesterday.

[VWolf]

4. I agree that they lean right

I use their electoral map as a worst-case scenario.




(One that has Obama STILL winning handily!)

[Brotherjohn]

7. Okay, more info posted on RCP site. Processing... My head hurts. Anyone? (nt)

[Brotherjohn]

8. BY JOVE, I THINK I'VE GOT IT!!! (but my head still hurts)

They now make clear that they're weighting the two LV models 50-50.

It also appears they're not using the RV poll results anymore.

I THINK they're also considering the "Gallup Tracking" (LVI & LVII, or "Expanded" and "Traditional") as one poll (weighting 50% each), and the "USA Tday/Gallup" as another poll (also two versions, weighted).

That's fine IF they are separate polls, separate samples. I'm not entirely clear that that's the case, but they are now using separate names to refer to the two polls. The two just happen to be Gallup-affiliated and both now using two LV scenarios.

Pollingreport.com is now listing the "USA Today/Gallup" in their "Presidential Trial Heat Summary" section, and the "Gallup Tracking" in their "Daily Tracking Polls" section. I assume that means the USA Today/Gallup poll won't be an every day poll.

DAMN this polling obsession, and DAMN RCP and Gallup for making things difficult!!!

[RichardRay]

9. Are you looking at Gallups multi day tracking polls?

Both RCP and 538 infer a single day poll result from the mid-point value of a multiday poll, include that in their daily average and weight each day by the number of days in the tracker to avoiding overcounting. It's all driven by wanting to include some sort of an estimate of the meaning of the tracking poll (which runs over several days) in the daily averages.

[Brotherjohn]

10. Yeah, I think I get that. What I didn't know before and think I see now is that...

... the Gallup Daily Tracking poll is a different poll from the USA Today/Gallup poll (which will not be daily).

As an afterthought, I guess this has always been the case (there's a periodic USA Today / Gallup poll). But it became confusing because that poll started using the new methodiology (LV I & II) and relased the same day that Gallup Daily started using the same methodology.

All periodic national polls, I'd think (ABC/WaPo, CBS, etc, etc,), do the weighting you mention, because they are all usually compiled from data over several days.