Tarrance projects a toss-up

Oh boy. As you know, Battleground (pdf) is a polling firm done by a Democratic firm and a Republican firm. In 2004, the final of the Republican firm, Tarrance, nailed the actual vote-- off by one decimal point.

They've come out today with their final projection of 50.2 for Obama and 48.3 for McCain, .9 for Barr and .6 for Nader.

Sure, this looks good enough, but when you factor in that 20M of the total of 130M votes comes from CA & NY, where Obama racks up nearly a 2:1 majority going by the polls, it means that McCain would lead in the other 48 states, where the battlegrounds are happening.

So now we have a real polling gap, with margins going from 11% (Zogby/Gallup) to 2% (Tarrance), and the rest in-between. Going off of DemFromCT's compilation of 13 different polls, the average difference is 7.35% in favor of Obama.

I hope they get the extension in Virginia, it was denied yesterday by a judge. I doubt they get it, as even during the primary there were big problems, and we got no extension. You just have to be in line by 7. My polling location was changed (signs pointed it out clearly), though I did drive by and see the lines, I'm going to take my 3 year old daughter with me later tonight to vote.

Tags: 2008 (all tags)



Re: Tarrance projects a toss-up

What's the response rate on the day before election day? I heard both of NBC News' pollsters -- Democrat Peter Hart and Republican Neil Newhouse -- explain last week that they never poll on the Monday because people don't pick up the phone, or if they do they won't assent to be surveyed, leading to kooky results.

by Jonathan Singer 2008-11-04 06:55AM | 0 recs
Re: Tarrance projects a toss-up

I think what they do is take their final poll from yesterday, and weight it, which makes it even stranger as that one showed movement to Obama.

by Jerome Armstrong 2008-11-04 07:15AM | 0 recs
or you could exclude Texas

Why would you go by the polls in CA and NY and not in CO and VA?  That's bizarre.

And everyone got the election to within their margin of error.  Anything else is just happenstance.

by John DE 2008-11-04 07:00AM | 0 recs
Re: Tarrance projects a toss-up

Actually, the Democratic landslides in NY and CA are balanced by the Republican landslides in the deep south, rocky mountains, and Texas. Contrary to myth, the Democratic vote is not more inefficiently distributed than the Republican vote. I did some statistical analysis on this here: http://third-estate.blogspot.com/2008/06 /do-republicans-have-advantage-in.html

by thirdestate 2008-11-04 07:02AM | 0 recs
Re: Tarrance projects a toss-up

I was going off ow what TIPP had said in that audio I posted. '08 isn't like '04 in that regards, Obama is much closer in all of the red states, some of them very close.

by Jerome Armstrong 2008-11-04 07:06AM | 0 recs
Re: Tarrance projects a toss-up

Plus, if they are right that the race has tightened nationally, it will also have tightened in CA and NY. It doesn't seem right to say: Obama's lead has shrunk nationally (evidence: Tarrance), but not in these two large states (evidence: all the other polls), so McCain must be ahead everywhere else. We would need Tarrance's specific numbers in CA and NY showing Obama's margin holding there.

by thesleepthief 2008-11-04 07:07AM | 0 recs
Re: Tarrance projects a toss-up

Republican Pollster...

I also notice they have Obama getting barely 1% of the undecideds which is not in line with a lot of other polls.  

Wierd undecided breaks and old turnout model... we shall see how close they really are.

by yitbos96bb 2008-11-04 07:20AM | 0 recs
Re: Tarrance projects a toss-up

I agree.

I don't intend this to be a dig at anyone in particular, in fact, it's heartening to see people working with the statistics to try to mine some more information out of it.  I know I've learned a lot from all the blogs I've visited this cycle.

Nonetheless, it's my sincere hope that interested people will look into investing in a statistics class at their nearest institution of higher learning.  I see all sorts of weird inferences being drawn from these polls, logical leaps, particularly ecological fallacies, even a lack of understanding of what a mean represents.  

I think we'd all be better off if people took some time to try to understand what constitutes a statistically valid claim.

by sneakers563 2008-11-04 12:08PM | 0 recs
Re: Tarrance projects a toss-up

I'd be curious to see that, but common sense would tell you the same thing, based on how rare it is for someone to win the electoral vote and another to win the popular vote.  The Republicans had more efficient distribution in 2000, but the Democrats had it in 2004, and in both cases the final difference wasn't a full 2 percentage points.

That's just dumb.

by Jess81 2008-11-04 07:07AM | 0 recs
Now this is my kind of post!!!


by Concern Troll 2008-11-04 07:05AM | 0 recs
Past Performance

As you know from October if you have a 401(K), past performance is not guarantee of future performance.

Seriously, I listened to Republican talking heads for the past two weeks touting the IBD poll, saying that it must be the more reliable this time just because it was the most reliable in 2004. I thought that now that IBD is in line with the other polls, we wouldn't hear that faulty analysis anymore...and particularly not here.

Maybe every other poll in the world is wrong, and McCain will eek out a victory -- it's not outside of the realm of possibility. But it is clearly outside the realm of probability, and until such time as there's actual conflicting data to disprove the polls, I'm not going to cherry-pick bad polls to feed my anxiety.

by fsm 2008-11-04 07:06AM | 0 recs
Re: Tarrance projects a toss-up

People remember Zogby for having blown the electoral college count in 2004, but he got the final popular vote numbers pretty well.

by Jess81 2008-11-04 07:09AM | 0 recs
Jerome you know this is dishonest

In 2004, about half of the polls - if not a majority of the polls - projected a similar margin to this polls.

This year, this is clearly the most pro-McCain outlier, actually by a fair margin. The next closest "pro-McCain" poll has the margin at 5% in favor of Obama.

And to top if off, success in '04 doesn't mean success in '08. Different polling companies do better in different years. For example, Zogby and CBS were the most accurate in '08.

by Ben P 2008-11-04 07:11AM | 0 recs
I mean in '00

by Ben P 2008-11-04 07:11AM | 0 recs
Re: Jerome you know this is dishonest

This isn't even the most pro-McCain outlier - it's the Republican half of the most pro-McCain outlier.

by Jess81 2008-11-04 07:16AM | 0 recs
Re: Jerome you know this is dishonest

You know, you can first grow up and stop calling people liars, and then get the years right. Look, its their poll, not mine.

by Jerome Armstrong 2008-11-04 07:16AM | 0 recs
Not calling you a liar

Just saying you are misleadingly representing data.

Well, this is the last time I ever visit this site.

by Ben P 2008-11-04 07:20AM | 0 recs
Re: Not calling you a liar

I'll take your word for it.

by Jerome Armstrong 2008-11-04 07:27AM | 0 recs
Re: Holy crap, this guy needs a Xanax

I thought your post was very balanced between the sky is falling and dismissal.  You've definitely been more negative than most this cycle, but that just counterbalanced the irrational exuberance elsewhere.  Thanks for the post.

by edparrot 2008-11-04 07:36AM | 0 recs
Re: Holy crap, this guy needs a Xanax

I concur

by yitbos96bb 2008-11-04 07:47AM | 0 recs
Re: Not calling you a liar

Dude, I've been pretty critical of Jerome this cycle, but you are over reacting... he never said things would turn out like this.  


by yitbos96bb 2008-11-04 07:37AM | 0 recs
Re: Jerome you know this is dishonest

He didn't call you a liar and he corrected his mistake.  Boy, you are a testy one aren't you?

by realtarheel 2008-11-04 07:21AM | 0 recs
Re: Jerome you know this is dishonest

Liar, Liar pants on fire... hang them from a telephone wire...

Big doody head :-P

(yes, this was of the snarkiest snark)

Sorry couldn't resist.

by yitbos96bb 2008-11-04 07:36AM | 0 recs
Re: Jerome you know this is dishonest

lol, I don't have a comeback to that, let me go consult my 9 year-old first~!

by Jerome Armstrong 2008-11-04 09:58AM | 0 recs
Re: Jerome you know this is dishonest
engage in the discourse or just go to KOS; you'll see plenty there that only requires "Amen".
by ChitownDenny 2008-11-04 08:42AM | 0 recs
I think what Jerome meant was

just one more reason to GOTV!!!

by linc 2008-11-04 07:11AM | 0 recs
Like the old saying goes

Even a broken clock is right twice a day!

by Jon Niola 2008-11-04 07:16AM | 0 recs
Re: Like the old saying goes

And a broken calendar is right once an era!

by french imp 2008-11-04 07:25AM | 0 recs
Re: Like the old saying goes

Not if it's digital

by Skex 2008-11-04 08:22AM | 0 recs
Re: Tarrance projects a toss-up

Sure, this looks good enough, but when you factor in that 20M of the total of 130M votes comes from CA & NY, where Obama racks up nearly a 2:1 majority going by the polls, it means that McCain would lead in the other 48 states, where the battlegrounds are happening.

When I first read this, it seemed like you were saying that Tarrant was predicting McCain would win the other 48 States. I nearly fainted!

by QTG 2008-11-04 07:28AM | 0 recs
And Jerome rescues his reputation!

That didn't take long. I guess he didn't want to lose his base.

by dtox 2008-11-04 07:28AM | 0 recs
Re: Tarrance projects a toss-up

Based on final margin of 49-44 with 7 undecided, the final projection assumes that undecided break for McCain about 4 or even 5 to 1.  There are a number of reasons to be skeptical of such a disparity, best set out by Mark Blumenthal at pollster .com.  PEW and others have also argued that about half of undecided generally do not vote, further watering down such a swing to McCain.  

Finally pollsters of all people should recognize the shakiness of a claim that we are the best based on a three out of four n performance.

We'll soon see.  

by Kensingtonbill 2008-11-04 07:29AM | 0 recs
Re: Tarrance projects a toss-up

There's quite a bit of polling indicating that undecideds have been breaking slightly for Obama this far, as well. This 4:1+ split seems extremely unlikely.

Since there's no detailed data within the PDF, it's hard to otherwise comment on their internals. Color me extremely skeptical, though, when every other pollster (including ones with strong R leans) has this at a 6-point race or wider.

by Texas Gray Wolf 2008-11-04 09:52AM | 0 recs
Re: Tarrance projects a toss-up

Jerome - You're only seeing half the picture.  The Battleground daily tracking poll was indeed an R-D effort this year.  Their final actual poll number from their tracker was 49-44, which was 1pt lower for Obama than what they logged two days ago, which was 50-44.  So in the end, they came up with Obama +5 with 7% undecided.  That's based on actual poll data, not projections.

What each pollster then did was take that 49-44 results and, using their own respective models, project how the undecideds would fall.  Tarrance - the GOP side - came up with 50-48.  Lake - the Dem side - came up with 52-47.

RCP puts both results in their average and weighs them at 50% each, similar to what they were doing with Gallup's 2 LV models.  The overall effect on the overall RCP average was minimal.

Whose model do you trust?  Or maybe we should just look at the 49-44 actual final result and make our own guesses re: how the undecideds will fall.  Lake split them 50-50.  Tarrance gave McCain 4% and Obama 1%.  Do you think undecideds will go that strongly for McCain?  I can see a 3-2 split.  But 4-1?

So, in light of this, Jerome, I'd like to get your thoughts.

by OGLiberal 2008-11-04 07:37AM | 0 recs
Re: Tarrance projects a toss-up

Well, from what I can tell, Tarrance and Lake both did not have different results before 2004. They apparently disagreed then on the final weighting, and derived two different finals based on the last numbers. These are the Tarrance numbers in the post, and they were the ones that got it spot on in '04, not Lake. I'm of course not saying they are right, but given they got it right last time, its worth noting what they are saying for the context of the follow-up after the result.

Peter Feld thought 2:1 for McCain, funny that TIPP said 4:1 for Obama. I have  no idea, and would just go with 1:1.

by Jerome Armstrong 2008-11-04 10:03AM | 0 recs
Tarrance and Lake have different projections

They conduct the poll together, but Lake has it a five point lead and Tarrance two points. The difference? Tarrance gives all the undecideds to McCain. Lake splits them.

TIPP, by the way, gives most undecideds to Obama.

The undecideds were demographically McCainish and politically Obamaish. They were rural, white and less educated. But they hate Palin, are moderate, want change and think the GOP is awful.

by elrod 2008-11-04 07:39AM | 0 recs
Seriously, some of you need to grow up

Sometimes I agree with Jerome. Sometimes not.

But what I find is if you disagree with someone articulating your point of view in contrast to his is more productive than trolling and name calling.

by Jon Niola 2008-11-04 07:59AM | 0 recs
Re: Seriously, some of you need to grow up

Yea, its frustrating, I would like to have the time to build a new blog platform that solves the problem by thresholds of participation. Its on my to-do list.

by Jerome Armstrong 2008-11-04 10:08AM | 0 recs
Re: Tarrance projects a toss-up

Getting it right in 2004 is exactly why they won't be right this time. Even if Obama loses it won't be because the dynamics of this race are even close to 2004.

by attorney at arms 2008-11-04 08:02AM | 0 recs
if your 3-year-old

is like my 3-year-old, I wouldn't take her anywhere that might involve standing in line for an hour or more.

by desmoinesdem 2008-11-04 08:25AM | 0 recs
Re: if your 3-year-old

Me neither. I've stood in an hour-plus line with two 3-year-olds (and two parents; my wife was there too) for something they actively wanted to do and it was a nightmare.

However: figure a lot of polling places will have a number of parents with children. What's to prevent voters from organizing ad hoc childcare, with one or two parents' places in line being held while they watch a number of kids (outside, in full view, mind you :))? That's how I'd do it, were it my polling place.

by Texas Gray Wolf 2008-11-04 09:55AM | 0 recs
Re: if your 3-year-old

heh, yea you got a mover there. But no, our lines are now slimmed down, no one there right now.

by Jerome Armstrong 2008-11-04 10:09AM | 0 recs
Re: Tarrance projects a toss-up

Are you ignoring big advantage for McCain in places like Texas, AR,and other very red states.

I had a feeling your earlier posts' positive attitude would not stand.

by IowaMike 2008-11-04 08:45AM | 0 recs
Re: Tarrance projects a toss-up

This poll tickles my funny bone.  Lunch is done.  I'm going to finish this knock list, then go get another one!


by lojasmo 2008-11-04 08:51AM | 0 recs
Re: Tarrance projects a toss-up

I won't bash you Jerome, I will just note that that is a ridiculous redistribution of the undecided vote by Tarrance, the Republican here. A 60-40 pro-McCain split I could see, but an 80-20? No freaking way.

by Mullibok 2008-11-04 09:07AM | 0 recs
Re: Tarrance projects a toss-up

I think the final numbers would be closer to Lake's polling and perhaps a four million vote margin for Obama. However, the electoral college could be a lot closer with Obama getting about 320 electoral votes. Would like to see a category five anti-GOP hurricane strike New York, New Jersey, Illinois, Iowa, New Mexico, California and Washington. It would likely carry a large number of Dem candidates into the House and perhaps even threaten Bilbray and Rohrbacher in California?

by Boilermaker 2008-11-04 09:35AM | 0 recs


Advertise Blogads