It will be a pleasure to read intelligent commentary about polls.  TNR has been, to say the least, VERRRY disappointing on this score.

For those who don't know, Nate Silver also writes for Baseball Prospectus, which should be required reading for baseball fans.  My employer thanks you for continuing to make my work days wildly unproductive.

Welcome, Nate!  Great addition to the Plank.  Now I'm really getting my subscriptions' worth.

Great!  The cross-posting, that is.  The polling data is nice too, but when I saw the WSJ-MSNBC poll last night I checked the official NateSite anyway.  

I saw more than a bounce though, especially in the internal numbers. Females, Hispanics and a few other demos flipped to Obama more rapidly than I expected. There seemed to be a dramatic shift to Obama (from those groups) and in that regard their support may be more significant than a typical bounce.

The one question I have or maybe it is a two parter:

1. Does the polling methodology account for the number of new or newly registered voters and if so what 'number' would the use since Obama seems to be piling up big numbers in his registration drive?

1b. How would a young, previously unaffiliated or any new group be targeted let alone be reached?  Students aren't in school, they may not even have land lines and by definition any first time voter doesn't even have a profile so how are pollsters weighing them even if they can track them down?

I doubt new-first time voters won't impact the final vote and wonder if we will know anything about them before the election. Since more than a few states were lost by Gore and Kerry by just a few points & the 'new' voters can wipe that out, it would be nice to know if this seemingly invisible group is being measured and/or counted in the polling data.

Great feature.  I keep meaning to add fivethirtyeight.com to my daily rounds but there's too much to keep track of at this point.  I would love to hear Nate's opinion on the commonly espoused theory that as soon as McCain displays a weakness in polls the race will turn into a self-perpetuating landslide.

fivethirtyeight is a great site for exhaustive information, less so for accurate predictions. He blew the turnout in lots of the late states by wide margins and had Obama winning South Dakota by five when Clinton won by almost ten. I'm not really criticizing the guy, there's a lot of great information there, but poll prediction is a fool's errand and I wish he'd just stick to presenting the information.

No man who contributes to Baseball Prospectus should ever have to do anything else, ever. (Gals, before you slam me for sexism, understand that I think you're all already perfect and thus don't have to do *anything at all, ever,* so back off.)

That Nate has also graced us with FiveThirtyEight, which some nefarious TNR poster alluded to ("Go ahead, kid--try a hit. It won't hurt you.") and which I'm now shooting up daily, elevates him in my opinion to the upper pantheon of humankind--those who have negotiated ends to wars, who have discovered disease-thwarting vaccines, who invented that wooden backscratcher gizmo shaped like a little hand that exterminates even the hardest to reach itch.

[Servants standing on either side of the seated Nate Silver gently fan the Blessed One with giant ostrich-feather fans as williamyard bows deeply while slowly backing out of the room.]

Nice.  Any chance Mr. Silver can be persuaded to cross-post some Baseball Prospectus stuff?  

Now, seriously, I'm wondering about the New Jersey remark.  One normally hears of a presidential candidate's "coattails" lifting down-ballot candidates.  Are Lautenberg's struggles really enough to suggest he could have a depressing effect on Obama at the top of the ticket?

I'm still suspect of any polling done this far out, but at least as someone supporting Obama I'm encouraged that he's getting a bounce instead of a dip.  I seem to remember much polling data showing Clinton with a decent lead here in the Badger State with just 3 weeks left until the primary took place.  By the eve of the vote, the polls here showed Obama with as much as a 6 point advantage and after the votes were tallied he'd won by around 16 points.  While I was really encouraged by the outcome, I was caught off guard at the wide margin.  In a state that is loaded with traditional Democratic voters where women usually vote in greater numbers than men, I thought Clinton would have had a stronger showing.

I'm sure Obama must have scored pretty heavy gains from the Dane County region and I know he and his people really concentrated on registering new voters at the two biggest college campuses in Madison and Milwaukee, but those are pretty much the only Dem strongholds in a state that runs pretty Republican in nearly every county north of I-94.  Maybe with all the flooding problems these past two weeks, no McCain supporters have any phone service up north and can't participate in the polling.  I think Kerry might have hung on to win here by maybe just 5,000 votes in '04, so I'm hoping these latest polls become a trend.

If I really want polling data at this point-- in other words, if I'd like even more noise in my daily internet diet-- I can go to realclearpolitics.com. I think the analysis of such data's pretty meaningless at this point and will be for at least another three months, if not four. Dukakis was up by s.t. like 20 points at this stage.

I'm happy about the cross-posting (not that it affects me much personally, since I check FiveThirtyEight frequently anyway).  Nate does a fantastic job with this stuff.

Peter, the primary predictions that you refer to weren't based on polling at all.  They were based purely on a regression model that made predictions based on the underlying demographics and the performance of similar places that had already voted.  Turnout turned out to be pretty harder to predict than Nate thought, but overall his predictions did remarkably well, despite the miss in SD (e.g. he pretty much nailed NC and IN, when polls had been showing Clinton doing much better than the final in both states).  He also uses a regression for the states in the general election model, so again, it's not just based on polls.  The methodology's up on his website if you want a more detailed account.

tep,

A few points.  First, even if you just wanted the polling data, both FiveThirtyEight and pollster.com are better for that purpose that RCP, in my opinion.  Second, as I noted above, the analysis isn't just based on polling.

Finally, I think those Dukakis numbers are frequently misused by people who want to say polls are meaningless.  If all you're interested in is predicting a winner based on polling, then yeah, check back in 3 months.  But polls are useful for more than that.  Because of those Dukakis numbers, we know what the trajectory of the race was.  It illustrates, for example, how brutally effective the Republican campaign was that year.  Polls are also useful for other purposes, like helping campaigns and other groups (e.g. the DNC) choose where to allocate their resources.  I think the argument that the data is meaningless takes an overly narrow view of the matter.

tep,

If it's noise you're after, may I suggest the Library of Congress? I'm talking about the brick-and-mortar, not the dot-gov. Talk about TMI: once I was lost for a week under a pile of books a page delivered to my carrel. In the noise business, LofC is basically the Mormon Tabernacle Choir on PCP with the amps turned up to 11.

Speaking of noise, I'm at the moment listening for the first time to a Miles Davis concert recorded at Carnegie Hall, circa 1974. This is no throbbing comfort as with "Kind of Blue." For starters, there are three electric guitars, and it devolves from there. I have no idea what these cats were trying to do, and I'm not sure Miles did, either.

But like all righteous noise it keeps serendipitously relocating me like a puppy mindlessly licking his butt one minute only to be snatched up by a proximate human and deposited for no apparent reason on the back porch. Wow! Now what?!

Calm, soothing silence has its place, but Buddha, as he would be the first to tell us, is overrated.

"Peter, the primary predictions that you refer to weren't based on polling at all.  They were based purely on a regression model that made predictions based on the underlying demographics and the performance of similar places that had already voted."

To follow up on AlanSP's post. Nate's model did quite well. I believe that over all most poll accuracy this season has been between 6-8%. Nate's models are right there (6%) and beat some professional pollsters like ARG.

Link: www.fivethirtyeight.com/.../call-for-pollster-poblano-unity.html

Not bad if you ask me.

(BTW, Great to have him here)

Well golly Tep, nobody's forcing you to read it!

I can't believe you Obamabotphiles.  Now that you have one Messiah Annointed, you - especially WilliamYard - are about to annoint another Annointed One, the Bringer of Good News, the Master of Numbers.  Your Messiah complex is simply nauseating.  Wake up and smell the Arugulatte: the man is an empty suit, there is no there there, he is all speechifying idolatry (like Che and Ceaucescu and Fidel and Genghis Khan and Nero and Incitatus), the slightest scandal and he will fall over, you'll see.  McCain will win by a landslide as soon as he manages to utter a single sentence that does not land him in hot water.  Oh yeah, and all of these polls were done before the Michelle tapes hit the stands.  Just you wait, 'Enry 'Iggins, just you wait.

But then, Tep was right writing about Wright.  So who knows.

icarusr:  brilliant.

Peter1943; You make a point that I make all the time: poll predictions are a fool's errand. The general campain themes of Obama and Mac have not even solidifed yet and so much is yet to happen. It is interesting to follow the polls but it is truly amazing how much emphasis many people - and even some political writers - place on every twist and turn in the polls.

[obstructive sleep apnea halts suddently]

Somebody said something about williamyard? W--what was the question?

Man, I could go for an arugulatte right about now for some reason. First, though, I better put more ointment on this Incitatus...damn hooker shoulda told me she was having a flare-up...

Well, this must be some kind of coup for The Plank. I would have expected to find your dailies with  Obsessives Crowley, Fairbanks, and Scheiber, but what do I know?

Wow, Icarusr -- I haven't thought about Incitatus in years. And Arugalatte -- perfect description of the Obama crowd. And of the whole polling obsession, here and elsewhere....

Thanks....

There is no god but Thompson. Yard is his prophet.

or is it the other way around?

I know the polling currently shows Montana going for McCain, and I know it's only three EC votes, but smart money would place a bet on this state landing in the Obama column come November.

The very popular Democratic governor is up for re-election, Max Baucus' opponent is a clown, and people are livid about how the Republicans have conducted themselves at the state level.  There's no cultural wedge issue on the ballot, so the Rove tactic of driving the base to the polls isn't in play.  And the Obama disciples are highly organized and motivated, and will ensure a large turnout of younger voters who normally either ignore or despise politics; polling under-represents this group because they only have cell phones.

It will be close, but Montana will be a blue state in 2008.