I just wanted to point out how worthless the Newsweek poll is that everyone is talking about. The only polls that are worth anything are ones that are restricted to likely voters (or at least registered voters). This is the description of the methodology used in the Newsweek poll:

For this NEWSWEEK poll, Princeton Survey Research Associates interviewed 1,006 adults aged 18 and older Jan. 22 and Jan. 23 by telephone. The margin of error is plus or minus 3 percentage points.

This poll was for any adult over 18 which means at least 50% of these people wouldn’t even bother to vote. The NYT does these types of misleading polls too. Of course the reason they do this isn’t necessarily bias, they just don’t want to spend the money required for an accurate poll.

Posted by: hugehawk at January 26, 2004 09:36 AM

hugehawk…ridiculous…so you are saying the poll is not valid because it is not biased towards young people?

The poll is good to go…it is the same newsweek method they always use…

Posted by: what2 at January 26, 2004 11:19 AM

I don’t think you understand what I am saying. My point has nothing to do with a persons age. My point is that the survey is of anyone of voting age. But as we all know, most people do not vote. The only polls that are accurate are ones that limit respondents to actual voters. The best polls are from likely voters but at the very least they should be from registered voters. Asking “adults” who they will vote for is userless since the majority of them won’t bother to vote anyway.

And as far as being the same newsweek method they always use, that is my point. These newspapers and magazines dont want to pay the money necessary to get an accurate poll of likely voters. Candidates internal polling is much more accurate because they pay more money to get more accurate polls. The newspapers and magazines don’t care about that kind of accuracy, they just want a headline.

Posted by: hugehawk at January 26, 2004 11:34 AM

what2,
Hugehawk’s point was not about the age - polls should only ask those 18 and up since those younger than 18 cannot vote. Newsweek did this and properly just polled registered voters.

The point was that, historically, polls that merely interview registered voters rarely accurately capture the opinions of those people who actually vote. Experienced pollsters who make a living on their polls (vs. just selling stories about the results) give much greater weight to voters who say they are likely or very likely to vote. It appears Newsweek did not even ask this question.

Bottom line: While the poll is a good indicator of how registered voters currently feel, it is a poor indicator as to how those who will actually vote feel.

Posted by: Admiral Quixote at January 26, 2004 11:39 AM

Admiral,

You are on target, but from what I could see Newsweek didn’t even limit themselves to registered voters. When they say “adults” it means they also included people who are unregistered which makes it even more useless than a poll of registered voters.

Posted by: hugehawk at January 26, 2004 11:47 AM

I’m not sure the distinction between registered and non-registered voters is that important, but it may be. I have the impression that most non-registered voters don’t see an important difference between either party, thus they don’t bother registering, but I have never seen a survey that would verify or refute it.

However, before writing my first comment, I went to the link you posted. It was poorly written, but there are several references to Newsweek polls of registered voters in the story itself. So I assume this is their standard methodology - although the story is written is such a way your interpretation could be correct as well.

Posted by: Admiral Quixote at January 26, 2004 12:21 PM

What2, how old are you?

Posted by: johnnymozart at January 26, 2004 04:57 PM

VERY SORRY! I just put up this diatribe as a Comment to the Wrong Post (the Kucinich one). Like I say way below: “Dummy”. Apologies. Webmaster, feel free to delete it from the Kucinich-says-deadlocked convention thing. (Well, from here too of course, if ya want to. :) - TP
******************************************

All national polls are either worthless, or at best misleading — but not necessarily for the reasons discussed discussed.

Good pollsters endeavor vet their samples not only for voting eligibility* (including registration), but also for voting *likelihood — thus reducing (not eliminating) the error-margin re those variables. Newsweek doesn’t make clear whether they do that or not. (They refer to “registered and likely Democrats”. What the Hell is a “likely Democrat”?)

But national polls ARE off-base because of polling the wrong voters: they poll Us. (Statistical samplings of us.) Yeah like We, nationally, nominate the Presidential candidates & then elect the President.

Bulletin: state Delegates — 4,321 of them for the Dems, 801 of whom are “unpledged Superdelegates” not allocated by primaries or caycuses by the way — are the Voters who nominate the candidates.

AND — sorry but it’s true; boys & girls can you say “2000”? — real live warmbody human state Presidential Electors, all 538 of them, are the Voters who elect the President at our nation’s Presidential Election held in mid-December. 270 wins it.

No, these points aren’t just Technical. Not at all. They render national popular-vote polls, even the ones with good methodology, invalid.

Especially so the party-nomination-preference polls nationwide. Democratic primary & caucus voters do allocate most — 81.5% actually — of their delegates to candidates via their votes. BUT THEY DO SO ONLY STATE-BY-STATE. And candidate support varies widely, by state. So what is needed & would be meaningful is individual state-by-state Dem-voter tracking polls. 50-plus of them (D.C., Puerto Rico, American Samoa, U.S. Virgin Islands, “Democrats Abroad”, etc. etc., have Delegates too, in the Inclusivity-minded Democratic Party :). Then you plug each state’s results into the Dems’ uniform equations for modified-proportional-distribution of that state’s allotted number of “pledged” delegates, amongst only those candidates who made the 15% popular-vote threshold, and — Bingo! NOW you’ve got youself a POLL.

Newsweek sayeth:

“Dean’s support among registered and likely Democrats, meanwhile, has been cut in half, to 12 percent. That puts him in three-way tie for second place in the Jan. 27 New Hampshire primary…”

Assuming — as appears to be the case though again it’s foggily described — that their sample of “registered and likely Democrats” [sic] was (as in their Dem- vs - Bush matchups) a national one: that is sheer gibberish. How does a national poll put Dean in a tie for 2nd, or anywhere else, IN NEW HAMPSHIRE?

Comparably to, but more simply than, the nomination process: We the People, state-by-state, Elect our Electors in November. All the way from California’s humble 55 of them, to the District of Columbia’s exalted 3.

Therefore only a rolling series of state-by-state polls looking toward November, can meaningfully forecast the outcome of the election to be held in 51 separate state-capital meetings in December. (Especially when the nationwide popular-preference stats are, as now, Close. ) At least we don’t need a fancypants allocation formula: the Electors are winner-take-all, per state. Well. Except for Maine & Nebraska, but never mind about them now. :)

I suppose such a full-fledged centralized ongoing state-by-state polling methodology, for nomination & for election, would be very expensive, perhaps prohibitively so. Certainly it would be Complex & Annoying for the media who like to Keep it Simple, Stupid.

But at least let us Beware the Jubjub Bird of these nationwide popular vote polls. So what? It’s the Delegates, Dumbass. It’s the Electors, Eejit.
:)

Posted by: The Pooka at January 26, 2004 06:15 PM

Did they record respondents’ party preferences either (it should be obvious why this matters)? If they did I missed it.

Posted by: DrSteve at January 27, 2004 09:34 AM