Gallup Poll Update - What went wrong...?

[The Vorlon]

As you may recall, I have been one of Gallup's stronger supporters, and Gallup's state polls in 2004 were, well... pretty darn bad...

I have been doing a fair bit of emailing back and forth with David Moore, the Senior Editor of The Gallup Poll, and Gallup is going over everything they did, all their assumptions, and all the data with a fine tooth comb to see what blew up and why.

Gallup is doing a top to bottom review of basically everything.  They expect to make a public statement and release a public review likely in February or March of next year. - There will be some BIG changes it sounds like, they are looking at tossing pretty much their entire sampling model.

Gallup blew it, but at least they are behaving with honor and will be open about how they plan to fix it in the future..

This will be very interesting to say the least....

[Alcon]

I respect Gallup for this. They did not cheat - and they did find trends, but screwed up way too much in general. I will give them the benefit of the doubt in 2008, unlike Zogby.

On a side note, I just noticed that Alberta looks a lot like Nevada.

[The Vorlon]

On a side note, I just noticed that Alberta looks a lot like Nevada.

The high today in Las Vegas was 60

The high today in Edmonton was 4

Other than the 56 degree temperature difference the climates are quite similar also...

[Alcon]

On a side note, I just noticed that Alberta looks a lot like Nevada.

The high today in Las Vegas was 60

The high today in Edmonton was 4

Other than the 56 degree temperature difference the climates are quite similar also...

Hehe, I meant in shape...but I suppose both of them have a lot of nothing and then a city or two.

[The Vorlon]

On a side note, I just noticed that Alberta looks a lot like Nevada.

The high today in Las Vegas was 60

The high today in Edmonton was 4

Other than the 56 degree temperature difference the climates are quite similar also...

Hehe, I meant in shape...but I suppose both of them have a lot of nothing and then a city or two.

Alberta has two major population centers.

Edmonton is just under 1 million and Calgary is a tad over a million.

These two cities make up about 2/3rd of the total population.

Nevada is mainly Vegas, Reno is a very distand second I think..

Also the top half of Alberta is all forest, where Nevada is pretty much all desert

[JNB]

   Conventional wisdom went out of the window with this election. For decades, going back to the 30s, it was the higher turnout an election, the better it was for Democrats, and simpily put, that "wisdom" failed. Turnout was higher than expected, Kerry got more votes than what has was projected to have to win. No one expected the winner of this election would have close to 62 million votes.

[ilikeverin]

Could a certain member of a certain political site I know of have a hand in the changes?

[CARLHAYDEN]

It will be interesting to see the results.

I think most people (Gallup included) underestimated the GOTV effort the Bush campaign made this year (particularly in Ohio) or the impact of the Marriage measures on the state ballots.

BTW, I must admidt that I find Edmonton a tad less entertaining than Vegas.

Aside from an occasional bet on football games (go to hell McCain), I stay away from gambling (I know the odds).

[The Vorlon]

Update:

The current debate raging at Gallup is now if indeed it is even possible, given caller ID, call screen, cell phones, etc to actually create/reach a truly "random" or equal probability sample, or if practical reality demands some form of heavy stratification.

It polling terms, given Gallup's methodology, this is +/- the equivalent of the Pope publically questioning his belief in Jesus.

Gallup looks like they are reviewing EVERYTHING, good for them.

[CARLHAYDEN]

While I thing 'weighting' of the sample is necessary given the problems you cited (as well as other reasons), the problem is to project the correct weigting (i.e. the composition of the actual turnout).

In 1974, the turnout was heavily skewed in favor of the Democrats (Republicans were demoralized by Nixon and Watergate), while in 1994 the turnout was skewed in favor of the Republicans (Newt did a good job in directing that campaign).

What will the turnout be in 2006?  It depends in large amount on the efforts of the the major players in the political process and events (will Soros et al continue to pump big money into left wing vote drives, will the RNC continue its GOTV effort?).

[muon2]

   Conventional wisdom went out of the window with this election. For decades, going back to the 30s, it was the higher turnout an election, the better it was for Democrats, and simpily put, that "wisdom" failed. Turnout was higher than expected, Kerry got more votes than what has was projected to have to win. No one expected the winner of this election would have close to 62 million votes.

Actually I think this forum collectively did predict that.

In the middle of October a poll on projected turnout gave an average result of about 56% of the voting age population, slightly over the actual pencentage. Since the voting age population was estimated at 221 million, then the turnout was predicted here to be just under 124 million. Half of that is 62 million, so indirectly the forum picked 62 million votes for the winner.

[12th Doctor]

What were some of Gallup final predictions?

I have a theory on why all the polls were so wrong.  It is called, "realignment".

[Nym90]

Here were some final Gallup state polls that were a bit off.

Pennsylvania--Bush 50, Kerry 46.

Ohio--Kerry 50, Bush 46.

Wisconsin--Bush 52, Kerry 44.

Florida--Kerry 48, Bush 47

[dca5347]

Actually given all the problems with polling right now,I would increase their sampling size & the margin for error(both)

[Sam Spade]

Actually, to me its really the Gallup state polls which need the big-time help.

Their last national poll would have been dead-on, if they would have divided undecideds evenly between the candidates like Pew did, instead of giving 2/3 to Kerry and 1/3 to Nader (wtf?).

[CARLHAYDEN]

As I said BEFORE the election, Gallup generally doesn't do state polls very well.

They have done national polls better (the law of averages helps them here).

[ATFFL]

Let's look at two companies who had a very, very bad year in terms of results and how they responded to the election:

Zogby:  I nailed it!  Nearly perfect.  I remain the best pollster on teh planet.

Gallup:  We really ed up.  Sorry.  We'll be rethinking everything and do our best to figure out how to return to accurate polling.

See why one is higly respected and the other is laughed at?

[CARLHAYDEN]

The question becomes what planet Zogby is from?

[Alcon]

The question becomes what planet Zogby is from?

Upstate New York.

Looking at Gallup's measures (25% of people make $75,000+?!) during the election, I personally think they should have fixed them immediately. But it is good that they at least admit they have a problem - better late than never.