ARG has Hillary Clinton up by 26 points in South Dakota. The only other poll of the state had Obama ahead by 12. But that poll was conducted two months ago and had an extremely small sample size, so ARG is essentially flying blind here.
This much for certain: Clinton's South Dakota stock is a ridiculous bargain on Intrade right now. But South Dakota isn't all that idiosyncratic a state, and 26-point win just doesn't make any sense in the context of what we know about the demographics of this race. The national tracking polls are fairly stable, and the campaigns aren't behaving like South Dakota is a 20-30 point Clinton win.
As far as methodological nitpicks go, I don't really have any, because ARG does not disclose all that much about their methodology. It's considered less than ideal to poll entirely over a weekend, but that's something that happens fairly routinely (by ARG and most everyone else).
It's either a genius call or ... something the opposite of genius. You know on which side my bread is buttered.
Monday, June 2, 2008
On which reputations are made (or lost)
-- Nate Silver at 12:54 PM
Labels: arg, south dakota
31 comments
Didn't they have Clinton winning Iowa by 13? Maybe they want to finish as well as they started 5 months ago.
ARG's largest misses thus far in the primaries are: a 30-point miss in the Nevada Republican caucus, a 25-point miss in the South Carolina Democratic primary, a 20-point miss in the Illinois Democatic primary, and 16-point misses in both the Connecticut primary and the Iowa caucus.
Why don't you run a regression?
That does not include, by the way, polls where ARG refreshed/revised its result with a last-minute poll that differed substantially from its previous polls. They had a poll a week in advance of Wisconsin that missed by 20+ points.
What would be hilarious is if ARG overpolled HRC supporters this weekend, because the Obama supporters were at his rallies...
(Ok, I know that's not mathematically possible, but...)
ARG didn't poll SD, they averaged KY, WV and PR and guessed. Seriously, if Obama wins SD, no one should ever EVER link to ARG again. This could be the death of them. This one will stand out as being the end of their outfit... which is essentially a drunk monkey and a bango anyway.
Nate: I think you've got to run your model here. It would be an excellent test for your model: can it beat a reputable pollster (well, sort of reputable) when either the pollster or the model generates surprising results. So far, your model has had mixed success -- incredibly good for Indiana, NC, but not good at all for Kentucky. This would be the tie breaker!
I doubt that'll be the result, but it may show the trend. The Clintons have been down on the ground in SD alot more these last couple weeks than Obama has. They're a small state, who don't get paid much attention, so actually getting to meet these people might influence them alot. Not enough to swing the result by 50 points, but maybe enough to squeak out a win for Hillary, where Obama may have otherwise won it handily.
sigh...i hope it is wrong =(
he needs to come away tuesday night with combined wins that give him at least 300,000 more popular votes in those two contests than hillary gets from them, that way he takes that argument away from her altogether--otherwise there will remain a whisper from now until the convention that he didn't really win it, that more people wanted hillary, that he stole it, that he is not the legitimate candidate. it doesn't matter that it won't be true--he has to be able to silence that claim altogether. by my count if he can walk away tuesday night with 300,000 more votes than hillary gets, he can do that effectively, because that erases her "lead" and gives him a few votes to spare (fingers crossed)
(and toes too)
There's no way there's going to be more than 400,000 total VOTES in Montana and South Dakota combined, Mr. Anonymous.
As the great Jerry Seinfeld says, ARG, Good luck with aallll that.
Here's a "for what it's worth" comment. I'm being vague on identifying details because I don't want to get a staff member in trouble.
I was an Obama volunteer in a state with a primary last month. An Obama field staff member that has moved on to SD started sending out urgent emails last week looking for volunteers to head up to SD to help with GOTV. (Odd as we are a loooong way from SD.)
They forwarded an article about the Clintons spending lots of time in the state and perhaps being able to pick up enough votes there to end up ahead in popular votes. (This was all before Michigan/FL decision and before PR.)
Personally, I don't put any stock in the popular vote argument... also I took the whole episode with a grain of salt as it is probably hard to keep field staff focused when the race has been essentially over for weeks.
Nonetheless, in the multiple emails it implied campaign polls showing a Clinton lead at a level not being picked up by the national press.
Maybe ARG is playing Intrade? (You were looking for their business model, weren't you?)
It's a pity you do not publish a projection, really. It's unfortunate that readers look at this as if a projection is a game in which Nate may "win" or "lose" and so Nate is now led to project only where he believes he will "win". Projections are best considered as experiments, in which any outcome is of interest, "predicted" or otherwise (indeed, in an experiment, an unpredicted result is often more interesting, in that it reveals some new reality).
Anyway, even if we don't have Nate's MT-SD Clinton/Obama projection, we do have his against-McCain 538 regression, which is at least an indication of the demogaphics. The gap against McCain is identical, an Obama 7 points advantage (i.e. Obama loses in both cases, but by about 7 points better than Clinton does), which agrees well with the 10-20 Obama victory that everyone expects. However, heavy campaigning is worth about 5 points by Nate's model, i.e. this can well be very close in SD. (In other words, MT-SD can be a very nice experiment to study the impact of campaigning, clearly the one factor distinguishing the two).
Hard to believe that this poll is accurate given the results in neighboring states:
IO - O+8
MN - O+34
ND - O+25
NE - O+35
WY - O+24
Admittedly, all of these are caucus states. But, even if you assume an Obama caucus advantage of 15 or so points (about the difference between the Texas primary and Texas caucus results), Obama would still seem a solid favorite. As Nate notes, there also hasn't been any significant movement in the national polls. Thus, if this ARG poll is accurate, it would seem the Clinton SD campaign has really moved the electorate somehow.
ARG also says Obama's only up 4% in Montana.
http://americanresearchgroup.com/
First of all, Nate, might I say this site is one of the greatest developments in the history of wonking.
Second, might I beg that you consider taking on the BCS next.
Third, regarding the above post, have you any thoughts on the ARG Montana poll? Montana is, as you know, one of the whitest and blue-collar-est states in the Union. So it was somewhat surprising to see last week that Obama was polling (Lee Newspapers) almost 20 points ahead of Ms. Clinton. But now we see ARG suggest that the two are locked in a statistical tie. What gives? Sadly, neither one of the polls is by what I would consider a very reputable source, so I'm not sure we can really draw any solid inferences from them. What do you think?
First of all, Nate, might I say this site is one of the greatest developments in the history of wonking.
Second, might I beg that you consider taking on the BCS next.
Third, regarding the above post, have you any thoughts on the ARG Montana poll? Montana is, as you know, one of the whitest and blue-collar-est states in the Union. So it was somewhat surprising to see last week that Obama was polling (Lee Newspapers) almost 20 points ahead of Ms. Clinton. But now we see ARG suggest that the two are locked in a statistical tie. What gives? Sadly, neither one of the polls is by what I would consider a very reputable source, so I'm not sure we can really draw any solid inferences from them. What do you think?
How the hell did that post twice? I swear I only clicked once. Really!
Nate, will you post projections for SD and/or MT? I will certainly be looking forward to them if the answer is "yes."
Yes, we will definitely have projections up later.