47% undecideds in NH ? NO, not at all, 3 polls and the real undecideds in NH

There is a buzz going around about the 47% undecideds in a NH poll. I will attempt to decipher this issue.

The CNN/WMUR poll conducted in NH Jan5th & 6th, indicates that 53% of those polled were "definitely decided". While 26% were "leaning towards someone" and 21% were still "trying to decide". These are really stunning numbers. I have always "heard" that within 72 hours of an election about 90% of voters were decided. If true this poll goes against the conventional wisdom.



Well it turns out there are 2 documents on the net concerning the CNN/WMUR poll, the first link above and this one, which indicates 5% of voters were undecided.



Which is very close to the Marist Poll conducted during the same 2 days:




Marist also had this:



This again suggests that by Jan 5th, 90%+ of NH voters had made of their minds.

I will now move on to the CBS NH poll, again conducted Jan 5th & 6th. THe CBS poll says 28% could change their minds, & 9% were undecideds.

Summary

At this point there is an obvious pattern, 3 polling firms state that the undecided vote is 4% to 9%. Its easy to see this as they all use the term "undecided".

Marist poll: Obama up by 8%, with 4% undecided. 4% MOE.

CBS poll: Obama up by 7%, with 9% undecided. 5% MOE.

CNN/WMUR poll: Obama up by 9%, 6% undecided. 4% MOE.

Only the CNN/WMUR comes up with this huge 47% number of waffling voters, while at the same time they say the more strictly labeled undecideds are @ 6%. Both the Marist & CBS do not have any comparative questions or numbers that seem to relate to the 47% found in the CNN/WMUR poll.

Now obviously if Obama ate a live baby on MSNBC the night before the NH election, that would have affected his numbers. Looking solely at the CBS results, it could be that if the undecideds break massively for Hillary she wins. When considering the MOE in these 3 polls it very well could be that Hillary's 2% win in NH comes in right at the edge of possibility.

What does it mean ? You tell me.

Found here

http://www.wmur.com/download/2008/0107/14995708.pdf




Where? Last I checked 21 plus 26 is 47... no ?

The 2nd CNN/WMUR pdf is here

http://www.unh.edu/survey-center/news/pdf/primary2008_demprim10708.pdf

the meaning of undecideds is moot. THusly the hazards of polling.

if its accurate, fine, we can move on. If not, investigate. If it comes to convictions, great, at least we'll have some closure.

They seem to know what they're doing, too. I hope we get updates on their thinking.

Clinton had much stronger support, even before the "humanizing" backlash. You also have to adjust those numbers for the turnout - which was 57% women. Since pollsters adjust their samples to reflect populations, and I'm pretty sure NH isn't 57% women, the candidate preferred by women would do much better.

I know what you are saying IIRC HRC support was more solid than Obamas. This is from the Marist poll



Which suggest HRC strength @ 88%, Obama @ 77%.

It's posted right above - the much higher percentage of "will vote fors" compared to "could be swayed" is right there.

"will vote fors" compared to "could be swayed"

SO, Iam not sure what you are saying.

How does one quantify their significance in the election results ? Hoes that translate into an HRC win ?

And back to the OP, whats your take on the 47%....

that far fewer Obama supporters were very strongly behind him. 23% thought he was the best choice when the poller called, but only weakly or flat out stating they were shopping. Only 12% - about half as much - of Hillary's support was that weak. That means in effect 88% of Hillary supporters were pretty much guaranteed voters if honest, and only 77% of Obama's were. This difference ALONE could easily explain the difference in the votes but tehn you have to consider:

That 23% were sort of supporting a candidate who looked like a guaranteed win. How many said "meh - I'll watch House instead"? Compared to how many of only 12% for Hillary who knew they were behind and had to pull out all the stops.

The Clinton and NH Dem establishment machines were set up to absolutely maximize their turnout better than Obama's. The Clinton's had the local powerbrokers sown up,and have long been evry popular there. Plus they were, a little at least, desperate. I bet they didn't leave any of that 12% without a babysitter and a ride, let alone that 88%.

The turnout was not representative of the population. More women voted than men by a third or more. Who has the advantage among women? Less dramatically, but the older voters turned out in greater ratio than the younger. That split works for her too.

The 47% is just the 26 and 21 from your recent poll. It's measuring those who are weak or fungible in support, doubtless using a less restrictive definition than the other polls cited. It's not like they all ask the same question - polls are far better compared to earlier polls of the exact same design for this reason, rather than other polls. It's not like measuring "how likely are you REALLY to not change your mind?" is an exact, repeatable, measurand.

Cant recall seeing that sort of poll questions before, so I lack the exposure/experience. But I do not think that alone comes close to explaining the results.

the words are plain enough obviously. I'm a bit of a stats guy but no pollster but it's obvious they are the strongest supporters - the ones most driven to vote for candidate X, agreed?

This is not an attempt to say "this is what happened" but simply an example of the rather deceiving difference between those numbers.

Let's say Clinton got 100% of her strong AND weak support, because of GOTV efforts, the perception she was behind and needed the help, and last minute sympathetic responses.

We know what her vote count was, right? 112166

So let's say that the last CNN poll was also spot on and that's 29% of the likely Dem caucus voters.

so that would put (again this is absolutely not what happened, since neither of course was 29% for Hillary, but at the time of the poll, it's what would be implied) the total Dem likely voter pool at 112166/.29 or 386,779.

Now at the time of that last CNN poll Obama was weell ahead with 39% of them, which means he had, again at teh time of the poll, an implied support of 150,843 likely primary voters.

But his support was weaker, and much less impetus to vote on their side as he looked like a gimme. His 77% should have given him 116K or so and a narrow victory but NOW you have to add in the male/female ratio with late break to the symapthetic HRC, and the independent drain, and the age differential, and it's not exactly a huge step to his real vote total.

In short, his support was weaker and less driven.

But I wouldn't try to quantify it. What I did try to quantify in the OP is far more manageable.

Any chance you can drop some links, you have covered a lot of ground here....

When JUST giving candidate numbers, you include leaners. Thus the 39/29 included these. Obviously what happened is a large number of the leaners went to Clinton instead, and Clinton's much stronger support remained constant. The heavy turnout of women, older voters and the possible "not worth bothering supporting Obama since he's a done deal" effect did the rest.

"Furthermore your own data shows"

Not my data....

"When JUST giving candidate numbers, you include leaners."

Again not my data, everything I posted on photobucket is from the Polling pdfs.

But back to the OP, whats your take on the 47%.....

Support is fungible. The undecideds broke for Hillary. The turnout favored Hillary. the last minute news favored Hillary. The independents who like Obama thought he was a given so either didn't show up or voted for a Republican they could stomach (note I'm not saying they chose wisely - but McCain still acrries a lot of "maverick" cred for some reason). It's not like everyone who is polled votes and everyopne who votes is polled and nothing changes between the two. They don't have to be "wrong" to disagree with the result. Despite what they are used for, polls are not predictive and definitely not "correct".

A large laundry list of things have to break to Hillary for her to have won.

Interesting the Marist poll says Obama got slightly better support from women 34 to 33.

and secure the ballots asap.

others also say this, but its not accurate.


http://www.dailykos.com/comments/2008/1/11/0240/54559/23#c23

From http://www.realclearpolitics.com/epolls/2008/president/nh/new_hampshire_democratic_primary-194.html#market

I added a column called "other" which means kucinich/gravel/undecided.
Larry Johnson (and others) have asserted "at least 17% undecided." I think that's stretching it.

Poll	                Date	        Sample	Obama	Clinton	Edwards	Rich	Total	Other	Spread
RCP Average	        01/05 - 01/07	-	38.3	30	18.3	5.7	92.3	7.7     Obama +8.3
Suffolk/WHDH	        01/06 - 01/07	500 LV	39	34	15	4	92	8	Obama +5.0
American Res. Group	01/06 - 01/07	600 LV	40	31	20	4	95	5	Obama +9.0
ReutersC-Span/Zogby	01/05 - 01/07	862 LV	42	29	17	5	93	7	Obama +13.0
Rasmussen	        01/05 - 01/07	1774 LV	37	30	19	8	94	6	Obama +7.0
CNN/WMUR/UNH	        01/05 - 01/06	599 LV	39	30	16	7	92	8	Obama +9.0
Suffolk/WHDH	        01/05 - 01/06	500 LV	35	34	15	3	87	13	Obama +1.0
Marist	                01/05 - 01/06	636 LV	36	28	22	7	93	7	Obama +8.0
Rasmussen	        01/05 - 01/06	1203 LV	38	28	18	8	92	8	Obama +10.0
CBS News	        01/05 - 01/06	323 LV	35	28	19	5	87	13	Obama +7.0
USA Today/Gallup	01/04 - 01/06	778 LV	41	28	19	6	94	6	Obama +13.0
Franklin Pierce	        01/04 - 01/06	403 LV	34	31	20	6	91	9	Obama +3.0
Strategic Vision (R)	01/04 - 01/06	600 LV	38	29	19	7	93	7	Obama +9.0
Reuters/CSpan/Zogby	01/04 - 01/06	844 LV	39	29	19	--	87	13	Obama +10.0
American Res. Group	01/04 - 01/06	600 LV	39	28	22	4	93	7	Obama +11.0
FOX News	        01/04 - 01/06	500 LV	32	28	18	6	84	16	Obama +4.0

now if Obama eats a baby on CNN, of course all bets are off.

For some reason I am reminded of Chuck Hagels big win coming from behind.....

Yeah 17% undecideds are a stretch.

Those are not similar polls measuring the same thing. And the 90% doesn't mean they CAN'T be persuaded to change their mind, only that they thought they wre set. You have to include weak supporters in that available to be persuaded pool.

Never said that, these polls were different, and should not be compared, thats why I dont look at real clear politics. So yes I have committed a cardinal sin.


"And the 90% doesn't mean they CAN'T be persuaded to change their mind,"

Maybe you missed the eating a baby comment..... I think you did.

;j;j

Let me add that the media narrative "the comeback kid" had its start before the NH election.

I'll tell you, if Obama had been ahead for weeks, then I think we'd really be in shock. As it is, we're confused -- we've never had so many polls, and therefore so many opportunities for the polls to give the same wrong answer -- but we knew going in that with just four polling days between the caucus and the primary, and over a weekend at that, the results could be confusing.

Anything is possible. (ETA: Well, not "anything," but some combination of "undecideds" breaking to Clinton, nominal Obama voters switching back, Clinton supporters turning out at a higher rate.... Everything was happening very fast.)

I agree that it isn't very meaningful to talk about "undecideds" in the 40s. Just understand that "undecided" has no fixed meaning.

But thanks for trying. :hi:

Dont ya think ?

Then someone on DU posted a chart of polling numbers, and it showed what people were thinking before the vote, at the exit polls and at the final vote tally.

And although CLinton started out low, at 30% *for her* according to this poll, there was a five % uncommitted factor no one discussed. Maybe Chris Matthews and the rest of the media did gang up on her to the point that Hillary got most of the undecided.

And in this election, Kucinich claimed 1% rather than having that 1% handed over to Obama.

I really don't know what to think at this point.