Tracking Poll Update: McCain Falling on the Economy?

by Jonathan Singer, Tue Sep 16, 2008 at 10:27:12 AM EDT

Here are today's numbers:

	Obama	McCain
Diageo/Hotline	46	42
Gallup	46	47
Rasmussen Reports	47	48
Research 2000/dKos	48	44


Average:	46.75	45.25

The most worrisome numbers out of today's polling for John McCain are not that he has lost his average lead across the surveys (after the four polls showed an overall tie yesterday), but rather an indication that he may be slipping -- and badly -- on the economy. Take a look at this trend found in the Diageo/Hotline polling:

Who would do best job handling the economy?

Of course it is worth waiting to see if these numbers are borne out in other surveys, as it is never good policy to base too hard of conclusions on the results of a single poll. What's more, it should be noted that Gallup polling a week ago showed McCain gaining on the economy (though, again, that polling was in the field quite some time before this Diageo/Hotline poll, at a time before the markets really began to hurt). (Update [2008-9-16 15:1:31 by Jonathan Singer]: I missed it, but Rasmussen gives McCain a 49/45 edge on the economy. Hmm... Let's watch this metric.) That said, this is a number worth keeping an eye on, because if voters do not believe that McCain is credible on the economy at a time when the economy appears headed towards one of its worst spells in recent memory (and perhaps in several decades), it's going to be awfully difficult for the Repulicans to pull this one out.

Update [2008-9-16 14:51:41 by Todd Beeton]:Looking at the two candidates' favorable/unfavorable ratings over the last few DailyKos/R2000 polls (9/16, 9/15, 9/14, 9/13), I think it's pretty clear that what we're seeing here is more than just the end of a bounce. The media coverage of John McCain's lies and the hammering home of just how little honor McCain has is taking a serious toll.

Obama 54/37 (54/38, 54/39, 55/41...)
McCain 49/45 (51/45, 53/44, 55/43...)

Tags: Economy, Tracking Poll Update, White House 2008 (all tags)

Comments

83 Comments

Sort by

permalink

so do Ras and Gallup have likely voter screens?

I just don't remember the answer. It sure seems like the screens heavily favor McCain at the moment.

by John DE 2008-09-16 10:29AM | 0 recs

permalink

Democratic party is party of prosperity

This seems to be very consistent over the last 80 years.

Democratic administrations have been shown to be periods of job expansion and sustainable economic growth, as well as growth of stocks..

GOP administrations have been the opposite.

Why don't more people realize this?

by architek 2008-09-16 10:53AM | 0 recs

permalink

do Ras and Gallup have likely voter screens?

Ras polls likely voters, Gallup polls registered voters. However, neither of them releases their internals. That lack of transparency was what pushed Kos to commission his own tracking poll.

by Angry White Democrat 2008-09-16 02:52PM | 0 recs

permalink

Wow

I wonder why people would be losing faith in McCain's economic credibility!

Once again, today, Obama puts out a plan with real details and actual policy breakdown. Once again, today, McCain puts forth a bunch of empty promises that not only don't really explain how he would reform Wall Street and corrupt executives, but also don't explain why he would start doing that.

by vcalzone 2008-09-16 10:36AM | 0 recs

permalink

Re: Wow

I was stunned to hear, "We must strengthen capital requirements, particularly for complex financial instruments like some of the mortgage securities and other derivatives at the center of our current crisis" as an applause line. Obama is growing into talking policy while not losing charisma quite nicely.

by thezzyzx 2008-09-16 10:45AM | 0 recs

permalink

GOP is too involved in the downturn..

they are also, too emotionally invested in its causes, (such as deregulation) to realistically, honestly turn things around.

They would be too obsessed by trying to cover things up or gloss them over, we would not see a real recovery.

by architek 2008-09-16 10:56AM | 0 recs

permalink

Re: Wow

I understand the crowd tried to get a chant started "strengthen capital requirements, baby, strengthen capital requirements!" but it didn't catch on for some reason. Still, it's good policy.

by Steve M 2008-09-16 11:06AM | 0 recs

permalink

Re: Wow

Just in case I wasn't clear, he actually did get a rousing ovation after many lines like that. It was kind of bizarre.

by thezzyzx 2008-09-16 12:42PM | 0 recs

permalink

Where's Jerome?

I need somebody to crap all over this good news!

(Just kidding, dude).

by Spiffarino 2008-09-16 10:40AM | 0 recs

4a2fab88-ed3b-47a4-8f14-3ec8a24504ac_normal

permalink

Re: Where's Jerome?

yea, Markos is letting Research2000 convince him that Democrats are going to out ID Republicans by 9% on election day. He know better, but likes to feed the needers like you the candy anyway.

by Jerome Armstrong 2008-09-16 11:07AM | 0 recs

permalink

Re: Where's Jerome?

The non-partisan (or at least bi-partisan) folk at The Hotline are also being duped, then, as are the 11 million more voters registered as Democrat than Republican.

by Jonathan Singer 2008-09-16 11:16AM | 0 recs

permalink

Re: Where's Jerome?

And so it Charlie Cook:

This Gallup/USA Today poll also showed near-parity in party identification after months of showing a Democratic edge, suggesting either a massive shift in party identification, which is pretty unlikely, or a poll that is too heavily weighted toward Republicans.

by Jonathan Singer 2008-09-16 11:18AM | 0 recs

permalink

Jonathans wishful world

It wouldn't be the first time. Look, it's pretty clear, to anyone that looked at past results, what to expect. And its pretty clear that if you look at the likely voter trend on Rasmussen that this election is going there as well. Same with what Gallup is seeing with LV's.

by Jerome Armstrong 2008-09-16 11:28AM | 0 recs

permalink

Re: Jonathans wishful world

Yep, Gallup's 55 percent to 42 percent likely voter spread in favor of George W. Bush at this very moment in the 2004 campaign was dead on, and all the other polling showing the race between Bush and John Kerry to be close was dead wrong...

by Jonathan Singer 2008-09-16 11:42AM | 0 recs

permalink

In defense of Jerome

I don't think that being pragmatic about the state of the race is "crapping all over" it.

If we don't go into the home stretch with a belief that this thing is going down the wire, then we will be in big trouble.

Barring a radical news event, it is going to come down to our energy and ground game vs. their lies and racism, all within the margin of error.

by activatedbybush 2008-09-16 11:54AM | 0 recs

permalink

Re: Where's Jerome?

Ed Reiley is using the same model, yes, 8-11% in Dems favor. There's nothing wrong with polling RV's Jonathan, but positing that as the same as likely voters is a bit of a sham, right?

by Jerome Armstrong 2008-09-16 11:24AM | 0 recs

permalink

Re: Where's Jerome?

Yes, it's a sham -- but it's a sham in the same way that relying on likely voter numbers and a likely voter model that will produce wildly greater swings than are actually apparent in the electorate two months out from election day is a sham. Take Gallup in 2004, for instance. At this point in the election, they had George W. Bush up 13 points among likely voters (9/13-15/04, B-55/K-42), a spread that was not borne out in other polling in the field at the same time and was not borne out on election day.

I don't know what the answer is, except to say that there are problems with vesting too much faith in either model -- likely voter or registered -- at this moment in the campaign, and that it is better to base conclusions on as broad a range of data as possible (i.e. not just Gallup).

by Jonathan Singer 2008-09-16 11:39AM | 0 recs

permalink

Ugh

That AP statistic

Nationwide, there are about 42 million registered Democrats and about 31 million Republicans, according to statistics compiled by The Associated Press.

is meaningless. I checked my stats from 2004 and saw 33 million Rs vs 42 million Ds, that left about 80 million unknown and a Democratic loss despite a "9 million registration advantage". You can't draw conclusions based on which states are willing to report party reg.

by souvarine 2008-09-16 12:34PM | 0 recs

permalink

Re: Where's Jerome?

I guess you missed that link was going to a poll thats two days old. Their nowest:

Today's survey, conducted 9/13-15 by FD, surveyed 909 RVs and has a margin of error of +/- 3.3%. Party ID breakdown for the sample is 41%D, 36%R, 19%I.

See where its going?

by Jerome Armstrong 2008-09-16 07:55PM | 0 recs

permalink

Re: Where's Jerome?

Once again, clearly missing the unambiguously good news in the trends, which still hold whatever party ID reweighting you do.

by dtox 2008-09-16 11:22AM | 0 recs

permalink

Re: Where's Jerome?

I still love you, Jerome.

When Obama gets elected, we will need people like Jerome to stir us into not pressing action. And I think he's keeping sight of that.

by vcalzone 2008-09-16 11:27AM | 0 recs

permalink

Re: Where's Jerome?

I deal in reality. Not wishful thinking that has no basis on past results. I'd rather be surprised of a landscape that's changed to favorable after the outcome than depend on projecting one that gives false indications.

by Jerome Armstrong 2008-09-16 11:31AM | 0 recs

permalink

Re: Where's Jerome?

It helps your case that you were one of the voices saying that Palin would be a big problem.

by vcalzone 2008-09-16 11:57AM | 0 recs

permalink

Re: Where's Jerome?

Although, that being said, I think Sullivan is right, too, when he says that ultimately, Palin will be McCain's Waterloo.

by vcalzone 2008-09-16 12:11PM | 0 recs

permalink

Re: Where's Jerome?

This is what I think as well. Whatever short term benefit she has given McCain (and it is undeniable in the polling) I think will be reversed into that much drag on the ticket once the media have finished airing all of her bizarre dirty laundry (as well as thoroughly demonstrating that she should not be a heartbeat away under any circumstances).

In re Jerome's "realism"--I think that its useful to have the Eeyore types around because it does keep us grounded and eyes on the big picture. I am always looking for evidence however that this is a pragmatic attitude by Jerome and not an emotional one.

by wasder 2008-09-16 12:24PM | 0 recs

permalink

Re: Where's Jerome?

I have no vested interest at all, there's nothing other than wanting to be proved wrong with where I see this all going.

by Jerome Armstrong 2008-09-16 07:53PM | 0 recs

permalink

Re: Where's Jerome?

Amazing how a joke started this ball rolling. Did you miss the "just kidding" part?

I actually look forward to your analyses. You motivate me to contribute to Barack 'til it hurts.

by Spiffarino 2008-09-16 09:23PM | 0 recs

permalink

I think Jerome is feeling kind of

Blue. Point to the good news and he manages to predict doom and gloom til he turns blue. Point out the underlying fundamentals that trend so blue. Point out the fallacies and flaws in the logic.

What does your crystal ball say about the Senate Jerome?

by Sanguine Giant 2008-09-16 04:16PM | 0 recs

permalink

Re: I think Jerome is feeling kind of

Point taken.

Cheers!

by OmniStipes 2008-09-16 04:26PM | 0 recs

permalink

Re: I think Jerome is feeling kind of

But let's not be distracted.
Lots of democrats still have worries and doubts.
Understandably, after years of a Rove Morris
everyone is scared of an October surprise

personally I think Jerome serves a useful function
only when we've got all former Hillary supporters
invested in Obama rather than Mccain
no way does he want republicans in power
though he might want our enthusiasm checked

by duende 2008-09-16 05:29PM | 0 recs

permalink

Nah I think that

asinine comments
sort of drain
the motivation that
revolutionizes this race
on the order of
legions of volunteers and
other staff
going all out.
Yes we can.

by Sanguine Giant 2008-09-16 06:58PM | 0 recs

permalink

Re: Tracking Poll Update:

McHoover says he'll appoint a COMMISSION to study the economy! His folks are obviously in way over their heads at anything but political gimmickry like smear ads and photogenic VP picks.

by ObamaBiden 2008-09-16 10:46AM | 0 recs

permalink

"Appointing a commission" is a code word

for not doing anything...

Just to let you know.. in case you hadn't realized that.

They smile when they hear that..

by architek 2008-09-16 10:57AM | 0 recs

permalink

Re: "Appointing a commission"

Obama had a great line speaking today when he said (paraphrasing here) "We don't need a Commission, we KNOW what's wrong with the economy." Demonizing commissions themselves is hard, since the trad media idiots love them so. But playing the angle that a commission is not needed and would forestall action could be a winner.You'll never get the Tweeties of the world to come out against a commission unless you get them saying "what the hell do we need a commission for?" first.

by Beomoose 2008-09-16 11:04AM | 0 recs

permalink

Re: "Appointing a commission"

"We will appoint a commission" is political hack-speak for, "I don't know what I'm doing, so I'll purchase somebody to blame it on."

by Spiffarino 2008-09-16 11:06AM | 0 recs

permalink

Re: "Appointing a commission"

I completely agree. But winning elections requires driving the media narrative. Witness Palin: objectively she's in no way appealing to Clinton supporters, but the Narrative up until recently was she WAS. That hurt Obama for about a week.

The Trad Media LOVES commissions, crapping on them can easily backfire. Last thing you need is David Gregory going "don't you support the 9/11 commissions, Senator?" So its better to drive the point that a Commission would get in the way, we already know how to fix things we just need to DO IT.

by Beomoose 2008-09-16 11:13AM | 0 recs

permalink

Re: "Appointing a commission"

Commissions are good when you have no power. If you're running for President, it's an insane idea. If you don't know what you want to do, get out of the way.

by vcalzone 2008-09-16 11:23AM | 0 recs

permalink

Re: "Appointing a commission"

"Commission" is really just another way to say, "kick the can down the road."

I don't think that's what concerned voters want from the next prez.

by ObamaBiden 2008-09-16 12:42PM | 0 recs

permalink

Re: "Appointing a commission"

That is precisely right. McSame is telling the electorate, "Gee, I don't know what to do, but I can get a bunch of people together who do!"

That's not leadership. A leader says, "I know what's wrong and I can fix that." That's Obama's message today and it's going to turn the polls around, if only slightly.

by Spiffarino 2008-09-16 09:21PM | 0 recs

permalink

Props for snagging that user ID

by activatedbybush 2008-09-16 11:55AM | 0 recs

permalink

Rasmussen has a result I find hard to believe

"Forty-nine percent (49%) of voters say they trust McCain more than Obama on the economy while 45% trust Obama."

Does anyone think this is possible? If so, are 49% of voters completely insane?

by dtox 2008-09-16 10:52AM | 0 recs

permalink

Re: Rasmussen has a result I find hard to believe

Ras is pretty much becoming the "McCain best-case scenario" poll. A good indicator of the theoretical top end of his support, which is good to know. But I wouldn't live and die by their numbers.

by Beomoose 2008-09-16 11:01AM | 0 recs

permalink

McSame is the devil they know..

But people know they are in serious trouble. Obama needs to show them how he will protect them, personally from economic disaster.

The most common economic disaster is healthcare related for most people.. If Obama puts forward a real universal health plan, he will win. Otherwise, expect inertia to doom us to four more years of McSame because of his failure to seize the initiative.

by architek 2008-09-16 11:03AM | 0 recs

permalink

Re: McSame is the devil they know..

I know how invested you are in the issue, and I absolutely agree with you on the need for universal healthcare.

However, with respect, I think you're completely wrong on what that means electorally. Anything that's easy to brand as "socialized" will end up doing him more harm than good. I think the campaign's found a good line with showing McCain as being out of touch with people's economic woes, and it seems to be working.

by dtox 2008-09-16 11:14AM | 0 recs

permalink

Actually...

Obama doesn't have to change his health care policy at all--and if he did, he'd just get pounded for "capitulating to Clinton". If he switches his position now, it will kill his candidacy.

Fortunately, McCain has handed Obama the health care argument on a silver platter. Obama just needs to seize on this and make it a cornerstone of his ads and speeches:

http://www.nytimes.com/2008/09/16/opinio n/16herbert.html?_r=1&hp&oref=sl ogin

KISS: John McCain will tax your employee health insurance. Boom. There you go.

by Elsinora 2008-09-16 12:42PM | 0 recs

permalink

Re: Actually...

Hey, I wanted to let you know that the Obama tax calculator link you posted in your last diary has been getting a lot of use in my e-mail!
Every time I get a viral e-mail from someone, I send them the tax calculator. Don't know if it's doing any good, but it feels good. :)
So I wanted to say thank you!

by skohayes 2008-09-16 02:48PM | 0 recs

permalink

faves/unfaves

But we can't trust these fave/unfave numbers because the poll is sponsored by DailyKos!

That's what they'll say, anyway. And it's BS. He was polling pretty much the same as Obama four days ago. Now he's significantly down from those numbers. This is all in the same poll. So even if you want to claim that the poll has a Dem lean, there's a definite drop.

PS - Palin's fave/unfaves in the Kos poll are even worse.

by OGLiberal 2008-09-16 11:04AM | 0 recs

permalink

here candy....

Right, and the Battleground poll, showing Palin with 53/29 favorabless/ unfavorables, and Biden 49/29 is the outlier...

by Jerome Armstrong 2008-09-16 11:12AM | 0 recs

permalink

Re: here candy....

Is that the one that was in the field from the 7th to the 11th of this month?

If so, you've completely missed the point OGLiberal was making about the trend.

by dtox 2008-09-16 11:20AM | 0 recs

permalink

Re: here candy....

taken for a ride.

by Jerome Armstrong 2008-09-16 11:33AM | 0 recs

permalink

Re: here candy....

What does that even mean?

by dtox 2008-09-16 11:39AM | 0 recs

permalink

Re: here candy....

I think it's meant to be cryptic and vague...enhances the mystery.

by OmniStipes 2008-09-16 02:55PM | 0 recs

permalink

Re: here candy....

Exactly the point. Look at pretty much every poll for which there are internals saying the same thing, and hmm, there might be a trend here.

Reality based, remember?

Pretty much every poll shows Palin's (and McCain's) favorables dropping and unfavorables rising. Ras shows it right along with the others, and Ras is hardly biased in favor of Democrats.

Or you can cherrypick one old poll that's before the trend begins and try to use to to refute a trend that apparently most everyone else has noticed.

by Texas Gray Wolf 2008-09-16 12:00PM | 0 recs

permalink

Re: here candy....

As of today, Newsweek is confirming a 10 point drop in their polls, too... Are they super-biased as well?

by LordMike 2008-09-16 11:33AM | 0 recs

permalink

Re: here candy....

You could look into the difference between RV's and LV's, look at past results, and not have to resort to seeing bias instead of just different methodologies, some more based on the likely outcome than others.

by Jerome Armstrong 2008-09-16 11:35AM | 0 recs

permalink

Re: Tracking Poll Update

I hope Obama has finally figured out how to deal with Palin: ignore her and concentrate on the economy.

Obama's speech provides the substance to drive a positive message home, now if he and his surrogates can package his points into soundbites he might be able to win the message war in the remaining weeks. His negative messages have not shown much damage to McCain in Ras or Gallup.

I won't put any faith in R2000 until other polls start confirming the trends it shows.

by souvarine 2008-09-16 11:11AM | 0 recs

permalink

Re: Tracking Poll Update

Um -- all the polls are showing a decline in McCain's and Palin's favorables and a rise in their unfavorables. All of the polls are showing the same trendlines at the top, and all of the ones which release internals are showing the same trends internally. There's a difference in the raw numbers, but not in the trends.

by Texas Gray Wolf 2008-09-16 12:02PM | 0 recs

permalink

Rasmussen isn't

Rasmussen favorable ratings:

Obama: 55/44 (54/45, 53/45, 55/44...)
McCain: 56/42 (57/42, 57/42, 56/42...)

by souvarine 2008-09-16 12:21PM | 0 recs

permalink

Re: Rasmussen isn't

Yes but Rasmussen changed their targets in the middle of that sampling period to be more Republican focused which could potentially mask a decline. If the target had been constant, we'd probably be seeing one.

by thezzyzx 2008-09-16 12:38PM | 0 recs

permalink

Re: Tracking Poll Update

The reason his negative response has actually resulted in an increase is because nobody likes somebody who gets punked and doesn't step up. Obama showed that he wasn't going to back down and that raised his cred a little bit as a leader. Even if only subconsciously, that's what happened.

by vcalzone 2008-09-16 12:13PM | 0 recs

permalink

This econ news

is the boon that Obama needed. I just hope if he gets elected, it doesn't get worse and lead to a Carter-like thing where he gets left holding the bag.

by Lakrosse 2008-09-16 11:15AM | 0 recs

permalink

Re: This econ news

That's going to be the case for either, without a doubt in my mind.

by Jerome Armstrong 2008-09-16 11:35AM | 0 recs

permalink

It almost kinda makes me

rather see McCain/Palin hold the bag because if we're in the Great Depression II during McCain/Palin's term, Hillary Clinton will be elected in a landslide in 2012, whereas if Obama holds the bag, his weaknesses on "American values" already hinder him and the economy would be two heavily mortal blows. But the history books, the war, and environment are too important to let McCain in. I will not tell my kids about how Republicans reigned for 20/28 years even after GW Bush.

by Lakrosse 2008-09-16 11:44AM | 0 recs

permalink

Re: It almost kinda makes me

Two words: Supreme Court. Even if the Democrats take moderate control of the Senate, a John Roberts-style conservative will be confirmed. Quite possibly we get another Alito. That's especially true if the economy tanks worse and they manage to convince people it's the fault of those obstructionist Democrats in the Senate.

Potentially three Supreme Court vacancies in McCain's term, two who are on the centrist/moderate end of things. Flip those two to two Roberts (or worse, two Alitos) and the problem for the next President is that anything they try to do that would actually help can and will be declared unconstitutional.

Who cares if Clinton or anyone else wins in 2012, if there's little possibility they'll be able to do anything that matters?

Also, leave McCain in charge for 4 years and what money we have now will be gone. Hard to implement things like universal health care when, say, 50% of the federal budget is interest.

Sure, we survived 8 years of GWB. Maybe it won't be that bad. Maybe we'll hold the House and Senate and be able to block all the stupidity for 4 years without looking like we are the problem. But it's way too unlikely to gamble on.

by Texas Gray Wolf 2008-09-16 12:07PM | 0 recs

permalink

Re: This econ news

is that a "Goldilocks Recession"? not too bad, not too good?

I kid, I kid.

by Fluffy Puff Marshmallow 2008-09-16 11:39AM | 0 recs

permalink

I like the recent Obama ad hitting McCain on his

stupid strong economic fundamentals statement. The advert is highly effective and drives home the point that McCain is clearly out of depth in how to tackle the underlying weakness of American economy for next 8 years. The tide is definitely turning and I expect the swing back to Obama/Biden ticket..especially as the Sarah Palin factor becomes yesterday's news..

by louisprandtl 2008-09-16 11:43AM | 0 recs

permalink

McShame

Wow, I can't believe that I voted for McShame in the 2000 CA primary (though voted Gore in the general). First he cozies up to Chimp after being swift-boated in the 2000 primaries, then he becomes hard-right theocon in the 2008 primaries. And how he's just a liar and a fraud.

I would never have voted for the man this year given the R hanging after his name and the damage that the GOP has done under Chimp. But he could have salvaged my respect.

John McCain, you are a disgrace to the office you hold, and to the uniform that you once wore proudly and with honor.

by activatedbybush 2008-09-16 11:45AM | 0 recs

permalink

McShame's stabilizing position on the economy

should be 0.05%. That is the percentage of people in the US who would directly benefit from his economic policies (of course, they will increasingly live in a third world banana republic country behind fortified walls and hired bodyguards - but hey, they will get their generous tax cuts won't they?)

by activatedbybush 2008-09-16 11:47AM | 0 recs

permalink

I can't believe it's even close at this point

We really should require a civics and current events test to be eligible to vote. How in the world can people buy into 4 more years of the same old crap?

by activatedbybush 2008-09-16 11:49AM | 0 recs

permalink

Of course

We all knew the Pallin bounce would get the evangelicals to pee themselves. What was GREAT is it got everyone to pay attention. They know now what we know . . . the Republicans are pathetic.

by scytherius 2008-09-16 12:14PM | 0 recs

permalink

Re: Tracking Poll Update: McCain Falling on the Ec

This is good news! However wouldn't it be better to average the polls in proportion with the number of interviews made? As the average is done now each Diageo-Hotline or Research 2000/dKos respondent is worth three respondents of the other two polls.

by blueosprey 2008-09-16 12:19PM | 0 recs

permalink

Re: Tracking Poll Update: McCain Falling on the Ec

It's far more complicated than that if you really wanted to compare polls accurately. You'd have to also take into account how they determine sampling pools, the mechanism they use to collect results (landlines, cellphones, web based, paper), and a host of other variables. The bottom line is that the "poll of polls" composite a lot of places are starting to use is really very bogus and is probably making most statisticians and pollsters violently ill. But, it is a VERY crude way of seeing overall trends if you look at the composite over time. The problem is, most people are looking at it as a day-to-day measurement.

by tlhwraith 2008-09-16 12:29PM | 0 recs

permalink

Re: Tracking Poll Update: McCain Falling on the Ec

I totally agree with what you say. It is true that it is difficult to make sense of averaging the polls in any simple way. In a sense it is even difficult to compare the same poll in different days as it seems sometimes they change their LV filters and so on. However there are simple things that can be taken into account like the # of respondents.

by blueosprey 2008-09-16 12:41PM | 0 recs

permalink

Ras vs R2K

If Ras represents "reality", then the Hotline result is statistically implausible (0.5% chance using chi-square test).

However, if R2K represents reality, then all the polls fit nicely within bounds of random variation. Even the most improbable poll (Hotline) still has a 33% chance to occur.

Taking each poll as "given" and multiplying the chi-square probs of the remaining polls we have:

Hotline as given: 1.03% probability of the rest
Gallup as given: 7.66% probability of the rest
Ras as given: 0.11% probability of the rest
DKos/RAS as given: 12.90% probability of the rest

From the perspective of overall statistical likelihood, my money in this case is on R2K. I don't think it deserves the bad rap it gets. Ras is hosed - either they are screwing something up, or they are the only one getting something right.

by Neef 2008-09-16 12:35PM | 0 recs

permalink

Re: Ras vs R2K

If Rasmussen represents reality then Diageo is off by less than two standard deviations (sigma=3.3%) then by a ballpark computation I get at least 7% chance of that happening. How do you get your numbers?

by blueosprey 2008-09-16 12:57PM | 0 recs

permalink

Re: Ras vs R2K

low tech: excel chitest function. Ras expected distribution:

obama 47
mccain 48
other 5

tested against Hotline observed:

obama 46
mccain 42
other 12

leads to a pval of 0.5%

What did your calcs look like?

by Neef 2008-09-16 01:11PM | 0 recs

permalink

Re: Ras vs R2K

I'll try to do what you say. I wasn't doing anything as complicated it is just that your result seemed weird for the following reason. Assuming Ras to be correct then for example the result for Mc-Cain is 6% off in the Hotline result which is around 2sigma. Now assuming a normal distribution that should have around 7% probability of happening. If one believes such polls can be compared one could first average the results of all the polls with with weights proportional to the number of respondents in each poll and then look how far from the average each poll is and compare with its sigma. I would guess in such a case the average will fall in the one sigma average of each respective poll.

by blueosprey 2008-09-16 01:36PM | 0 recs

permalink

Re: Ras vs R2K

sorry the last sentence should read:
I would guess in such a case the average will fall in the one sigma interval of each respective poll.

In any case I found your post quite interesting

by blueosprey 2008-09-16 01:39PM | 0 recs

permalink

Re: Ras vs R2K

I see what you're doing. In that analysis, you'd also have to account for "percent no answer", which has it's own error distinct from "percent obama". Even if the "percent obama" were identical between the two samples, and the "percent no answer" were different, there would be distance. It's actually a multidimensional probability distribution, though the MOE is never reported that way (i.e. it's wrong). On top of that, the percents they report are almost never the percents they actually saw, due to demographic weights. And then all the stuff tlhwraith mentioned.

chi-square is nice and clean, because you're only testing the final numbers. Rasmussen has essentially made a hypothesis about the proportion of voters (however they arrive at it), and you can test Hotline's proportions against that hypothesis. In that analysis, you don't really need sample size - you no longer care about the error inherent in the results, you care about the likelihood of the results. fun stuff =)

by Neef 2008-09-16 02:08PM | 0 recs

permalink

Re: Ras vs R2K

Sure in principle given the no answer/other possibility one would want to have the variance matrix instead of a single number to make better comparisons. However I think that given the differences in methodology/small population involved
it is better to discard the third answer completely and just normalize the Obama+McCain result in each poll to 100. You can readily see in the chi square test it is the no answer/other case that makes chi big.

by blueosprey 2008-09-16 03:18PM | 0 recs

permalink

Re: Ras vs R2K

If you could put that into English for the rest of us statistically challenged people, I would really appreciate it. :)

by skohayes 2008-09-16 02:55PM | 0 recs

permalink

Re: Ras vs R2K

I can try to explain what I think is going on:
Suppose for a moment that the world is an optimal place where one can imagine performing a poll interviewing say 1000 voters in an exactly random way. Then different polls of different random samples will still differ in their results. This is what is encapsulated (with caveats see above) in the error estimate provided with the polls. The percent error scales inversely to the square root of the sample size. Suppose now you know the real percentages say (50-50) then there are ways to compute the probability of getting say (48-52) interviewing 1000 people.
What nees is saying is that taken each poll as the true distribution the probability of obtaining the others by random sampling is maximum for the R2K poll. This can be an argument for the R2K poll being the closer to the actual "true" distribution.

by blueosprey 2008-09-16 04:10PM | 0 recs

permalink

Heh, yes

If we pick the Ras poll as the "best" poll, the Hotline poll should never have happened. It's unlikely given Ras' version of the voters.

If we pick the R2K poll as he "best" poll, then all the other polls are very plausible. Just natural random variation.

Since all the polls DID happen, it makes sense to choose the poll that makes them ALL most likely. Which is the R2K.

by Neef 2008-09-16 06:26PM | 0 recs