Huckabee, Obama** Lead DMR Poll

The newest numbers from the BEST public pollster in Iowa, Ann Selzer, who conducts the Des Moines Register poll.

Obama: 28 Huckabee: 29
Clinton: 25 Romney: 24
Edwards: 23 Giuliani: 13

** Clinton loses 12 pts among union households; running roughly equal with Obama among women (he leads within the subset margin). But 7 in 10 women are not choosing Clinton.
** Clinton most presidential, knowledgeable and ego-driven
** Obama most principled, likeable
** Huckabee shoots up among self-described born agains
** Romney seen as most presidential

** The margin of error is +/- 4.4%, so Obama technically does not lead outside the margin of error. Huckabee does.

Comments

Am I missing something? With Huck only ahead of Romney by 5, and the MOE at +- 4.4, isn't he within the MOE as well?

Posted by justinb | December 1, 2007 10:18 PM

justinb,

The formula for calculating the MOE of the difference between two poll numbers is actually pretty complicated as I recall, but you are almost certainly correct.

Posted by DTM | December 1, 2007 10:28 PM

Last time I checked 5 was still greater than 4.4

Posted by simple math | December 1, 2007 10:32 PM

IIRC, the likelyhood of the "correct" number being at either end of the MOE range becomes vanishingly small as one approaches the edge, but by definition, a 29 to 24 race with a 4.x% MOE is within the margin, AFAIK.

Posted by justinb | December 1, 2007 10:34 PM

Last time I checked 5 was still greater than 4.4

Yes, but if the MOE is 4.4, and Huck's at 29, then he could *really* be anywhere between 33 and 25 (or so). Romney could likewise be between 28 and 20. Of course, the chances of Huck being 25 and Romney being 28 are much less than Huck being 29 and Romney being at 24, but by definition these two guys are within the MOE.

Posted by justinb | December 1, 2007 10:36 PM

justinb,

Usually the confidence level is set at 95%, which means there is around a 2.5% chance of the true number being beyond the given range in a particular direction.

But the reason this gets complicated is that the numbers are not independent (meaning if the numbers for Candidate A are wrong, that also will affect the numbers for Candidate B).

Posted by DTM | December 1, 2007 11:04 PM

When you're leading in the Poll, you don't care about the MOE. But when you're not...the MOE becomes important...

Posted by Mil MSNZ911 | December 1, 2007 11:25 PM

Usually the confidence level is set at 95%, which means there is around a 2.5% chance of the true number being beyond the given range in a particular direction.

I know there's an decreasing probability as you get away from the median of the range, but I wasn't sure how it was calculated.

Are you saying there's a 95% confidence that the true number for any such number lies within 4.4 points of the stated number? Doesn't the probability of accuracy go down the farther you get to the edge?

I know what you mean about one candidate's numbers affecting the other(s), makes me wish I hadn't forgotten everything I tried not to learn in those stats classes :)

Posted by justinb | December 2, 2007 12:00 AM

Selzer just missed the Indianapolis mayoral race by 8 points last month. If they are the BEST then I'd hate to see who is in second place.

Posted by Joel | December 2, 2007 12:39 AM

No one saw the Indy Mayor's race coming.

Posted by Beb | December 2, 2007 12:48 AM

Speaking on behalf of liberal arts majors everywhere, my head hurts.

Posted by Paul | December 2, 2007 1:08 AM

I agree with the earlier commenters Marc, this notion that just because something is w/in the MOE means that you have to qualify "Obama Leading" with an asterisk is off.

Good primer on this here.

http://www.washingtonmonthly.com/archives/individual/2004_08/004536.php

Posted by Tom B. | December 2, 2007 1:13 AM

In fact, using the excel sheet that Kevin Drum posted, there is an 82% probability that Obama is leading. I'd say that means it's a lead.

And I'm not quibbling that you're misrepresenting anything, but putting the asterisk up there calls into question the lead.

Posted by Tom B. | December 2, 2007 1:20 AM

Rumor has it (or at least this Des Moines-based blog has it) that Ted Kennedy will endorse Obama this Sunday: http://www.intothevortex.net/2007/12/kennedy-to-endo.html

Posted by dry_fish | December 2, 2007 4:55 AM

A Kenndy endorsement would certainly be helpful, so I hope it happens, but the blogger you linked to is also predicting a Biden surge -- displacing John Edwards for third place finish in Iowa.

???

Posted by donna L. | December 2, 2007 10:28 AM

Wow, Donna. Hadn't heard that before. Biden is an interesting candidate, in the sense of why he is so unable to gain some traction. Hadn't heard anyone saying they thought he would.

For Edwards, it doesn't matter. Third, fourth... either way it would kill his candidacy. He had the largest head-start in Iowa, and has focused his efforts on Iowa more so than anyone. If he doesn't win there, he's done.

And that could be the worst news for Clinton. Her best chance of pulling off this nomination is to have others split the "anyone-but..." vote. Right now we have that split between Obama and Edwards. If Edwards drops, it isn't clear either Biden or Richardson could move up into a new three-way dynamic.

Maybe Hillary ought to give Edwards a few additional million with which to really blitz in Iowa.

Posted by Paul | December 2, 2007 10:58 AM

It will not matter. No matter (how much I dislike HRC), Obama cannot win IA or NH.

HRC will win. NYT supports her. The media shows (like Hardball, NPR, and CNN) cover her favorably.

The Clinton machine has not yet gone into a full gear. When that happens, even GOP/Bush fail.

Clinton Machine > (ALWAYS) Bush/GOP Machine.

Obama (will get my vote) has to work harder to win. He needs Edwards to go after Clinton, too.

Without Obama and Edwards working harder, harder, and harder, HRC will win.

No one can beat the Clintons. No one. Has anyone ever done that? No. No one can beat the Clintons. The cemetry is littered with GOP dead white guys who tried to challenge Clintons.

Clintons forever (the message from media, NYT, Hollywood, etc.)

What can an ordinary voter do? NOTHING.

Remember: In the end, you will vote for HRC no matter what. Is that the coolest thing from the Clinton Machine? Never doubt the Clintons. NEVER.

Posted by eorse | December 2, 2007 12:39 PM

Glad someone else noted the MOE is plus OR minus 4.4%. The innumeracy pervading journalist's interpretations of polls is maddening. How about before you write something you're not sure of, you check with someone who is? That used to be called reporting out a story; I'm sure it's now called an overabundance of caution.

Posted by Michael's Mom | December 2, 2007 1:46 PM

There's seems to be a lot of confusion about this, but it's really quite simple. The point values are means in distribution curves. You can calculate the joint distribution curve from the standard error of the difference, which is sqrt((p(1-p)+q(1-q)+2pq)/n), which given N=500, p=.28 and q=.25 (p=Obama, q=Hillary) we get a standard error of difference of .0326, roughly. Then, we fill that into the normal curve function with a p-q=mu=.03, and we get our joint normal, roughly, as exp((-1/2)((x-.03)/.033)^2)/(.033*sqrt(2pi)), which gives us the probabilities of each possible comparative outcome between Hillary and Obama. So we integrate from 0 to +infinity over 1 (the value of the full normal from -infinity to +infinity) and get a definite integral. That looks like this: 7.97*integral from (0,+inf) of (exp(27.55x-459x^2) dx, which gives us the proportion of the curve that is greater than 0. That turns out to be .816 according to my trusty TI-89, which means that, given the sampling size and responses, assuming a dyadic choice (an approximate and fair assumption) that there is an 82% chance that, at the time of the poll, Obama was and is ahead of Hillary in Iowa.

Such is math.

Posted by Vincent | December 2, 2007 2:34 PM

Just a small addition to my point above calculating the likelihood of an Obama lead puts the likelihood of Huckabee leading Romney (se_d=.056 and m=5) under this poll at 81%.

Posted by Vincent Vecera | December 2, 2007 2:57 PM

I call into question any poll that has 7 in 10 women not choosing Clinton, especially if that's moved in the last month or so. Has the narrative really been that wrong all along? Is Iowa that much different than everywhere else?

Posted by james | December 2, 2007 6:10 PM

playing the gender victim card did go well with Iowa women who have never been supportive of women canididates for congress anyway. Women don't like to be charartized as victims. Hillary is in full panic mode with today's attacks on Obama's character. It's so laughable.

Posted by sonya | December 2, 2007 8:50 PM

Everyone seems to think that Edwards is a dead letter, but he is still in the "post position" in this race coming into the final lap and turn. I am highly dubious that Joe Biden will surpass John Edwards, As for Ted Kennedy endorsing Obama, it might do Obama some good, but Kennedy endorsed Kerry and we all know how well that turned out for the democratic party in the general election. Take a look at the incredible speech Edwards gave at the DNC (http://www.democrats.org/a/2007/11/john_edwards.php) if you want to see how a strong, committed campaigner looks. He's Bill Clinton at his best without the baggage. If the past six years have taught us anything it is that Edwards' message that the democrats need a fighter to get their legislation passed is much more accurate than Barack Obama's wishful thinking about working collaboratively with entrenched special interests on big issues like healthcare or environmental policy. Having grown up in Oklahoma I firmly believe that with his rural and southern roots Edwards would be the strongest candidate for a true national campaign, ceding as few electoral states to the Republicans at the outset, and bringing the largest Democrat majorities to the Senate and House along with him. I hope he wins Iowa, for the sake of seeing real progressive policies enacted over the next four years.

Posted by popudem | December 2, 2007 9:49 PM

Eorse, most people I know are as damned tired of the "Clinton machine" as they are of the Bushes. Obama IS going to win the Democratic nomination for President. You can already see the momentum for his candidacy building and Hillary Clinton's star fading. She is a TERRIBLE potential candidate, with MORE of her husband's baggage, in the form of post-Presidential sex scandals waiting to be revealed by the GOP. Many, if not most Democrats who've become aware that the Republicans are salivating to run against her are also aware that Obama CAN win the White House against the candidate the GOP is now most likely to nominate, Huckabee.

Posted by digbydolben | December 2, 2007 11:35 PM

With a +/- 4.4 points, Huckabee does not lead (factoring in the margin of erro results in Romney 28.4%; Huckabee, 24.6%).

Posted by frank | December 3, 2007 6:37 AM

We not only need to think about who can win, but who can govern. Im giving Obama a second look. He might be the best candidate to have AFTER the election.

Posted by Kev2112 | December 4, 2007 7:15 AM

AGGRESS AutoPost Test

Posted by asdf | April 16, 2008 7:36 AM

http://Good.com

Posted by Good | April 16, 2008 10:19 AM

Çàõîäèòå è Âû íå ïîæàëååòå!

Posted by Athanasios | May 8, 2008 3:33 AM

Posted by Atnas | May 9, 2008 3:36 AM

Hello

Posted by Brutello | May 28, 2008 9:28 AM