The death of polling as anything more than a guess
by Jerome Armstrong, Tue Jan 08, 2008 at 06:27:39 PM EST
Clinton 39 Obama 36 Edwards 17 Richardson 5Every single one of them is wrong outside the margin of error.
Tags: 2008 election (all tags)
Comments
Re: The death of polling
Yeah I am wondering if this was the dreaded return of the race gap that people have been avoiding talking about. I am not an Obama supporter but it was a concern. I've been talking about this but during New Years a few of my friends myself were hanging out- all of us black professionals. One of which said point blank- "if you think they are going to vote for us behind closed doors you are crazy." I thought him too cynical even by my standard. There are many other reasons for the turn around than race,but I am curious about this issue and how it will play out in other states.
Re: The death of polling
I think we're going to have to study this more closely. So far all the buzz is about the huge shift in the female vote over the last couple days. I don't know about the male vote, but if it turns out that women account for the vast majority of the disparity, the question would be: is there a reason why women would be more prone to hidden racism?
I don't think the Bradley effect has shown up consistently over the years, but there's certainly individual election results that need to be explained somehow.
Re: The death of polling
Not that it counts for much- but I am certainly hoping the race gap theory isn't a factor here. I wish this site were a place where we could have deeper conversations about race during the primary,b but I am not sure it is without being accused on something for asking about it or bring it up.
Re: The death of polling
I think we all hope it isn't a factor. I think it would be a waste if we didn't try to discuss race issues here as, with all apologies to Francis L. Holland, we really do seem to have a diverse commentariat here.
Re: The death of polling
I dont think it is a factor. I put this in my latest diary. I dont think polling samples were representative of the actual turnout. Hillary's soft support became hard and there were a couple of factors that probably served as a rallying cry.
Here is why I am sure about my point. If the women demographics who were +3 for Obama in the advance polls lied, why wouldn't they lie in the exit polling? But they didn't. The exit polling said women went +13 for Hillary. I just think that the women who were more likely to vote for Hillary showed up in bigger percentages than the segment of women who were more likely to vote for Obama. And those women who were on the borderline between the two switchd to Hillary after all the media nonsense in the last two days.
Re: The death of polling
I think there will always be a race gap, but my gut feeling is that it was negigible in this primary. Less than 1% which won't account for the discrepancy. I think you will see more of a race gap if you asked people in the poll and preceded Obama's name with "the only African American candidate Barack Obama". Or if the poll had something like "would you vote for the African Ameican candidate in an election" and you might have a lot more people than reality saying that they would.
Re: Not here
I am not arguing anything. I am asking a question. after last night where ever was so certain in their assumptions I am glad yesterday on predicting this thing I said I have no clue. The same goes here with this. That's what I was responding to - not out of hand dismissing this. I don't think it is race, but pretending like I am certain what's what giving the changes and shifts so far seems a bit much. That's just me.
Re: The death of polling
Seems to me racism would have benefited Edwards more than Clinton, the bigotries of racism and sexism tend to go together. But Edwards took the biggest hit in NH.
I am sure racism is a factor, even among Democrats, as is sexism, but I don't think it is a determining factor in Democratic primaries. Obama got a hell of a lot of votes in NH, nearly the same percentage he got in IA.
Re: The death of polling as anything more than a g
I think those polls were right up until the tears. Edwards taking the bait was the worst thing possible for Obama i think.
Re: The death of polling as anything more than a g
truly unbelievable....just goes to show that no one knows ANYTHING about this race.
Think about it, they've all been at this for a year, and we still have no freaking clue...
Congrats to the Hillary team again. My hat goes off to ya
I don't agree.
Before we pile on polls, consider that (1) polls were mostly on the field over the week-end when the bounce was at its highest, (2) independents apparently massively switched their allegiance at the last minute from the Dem to the GOP race. We knew Obama and McCain couldn't both do well, and that's what happened. And (3) Rasmussen, ARG and Suffolk all suggested that the race had stabilized and that Clinton was slightly coming back. Remember, ARG said last night that Clinton had regained her lead among female voters.
Frankly there were signs in the polls that Clinton was very much alive. The media is trying to push the responsibility of their misguided coverage on polls, when frankly it's the way they portrayed them.
Re: I don't agree.
Has data been posted on how the indies voted?
If you look at latest CNN figures (85% and 83% D and R precincts reporting), D votes outnumber R votes. If I gross up for the 85% and 83%, I'm calculating that 59% of NH voters voted in the Democratic primary
If so, that's great news, and suggests the indies turned out for Democrats. I think the issue here is that hillary turned out a lot of dems and widened her lead among dems
Re: I don't agree.
My advice is when it becomes a race between a white and a non-white person, or a man and a woman, dont believe the polls. It is politically correct to say that they prefer the black over white or a woman over a man..and when they are in the privacy of the polling booth they do something else. Nearly a quarter century ago there was white candidate for governor of a state whose name I could hardly pronounce and there was a black candidate who was a mayor of large city. Polls, even exit polls showed the black candidate winning..when the actual results came the white candidate had one. The black candidate was none other than Tom Bradley. The year was 1982 long before anyone heard of even John McCain. Not the death of polling. I dont believe people told the pollsters the truth that they cannot bring themselves to vote for a black candidate.
Re: I don't agree.
CNN just posted this below. Per CNN, indies did turn out for obama, but clinton won among dems, particularly women. i would agree with you re white voters saying they vote black then voting white, but in this case, it was the dems and women who voted clinton. so i don't think it was race ... dems aren't more racist than indies, women not more racist than men. i think women rallied around hillary, and she did a good job turning out dems in sufficient numbers to blunt the indies.
---
Forty-three percent of self-styled independents said they voted for Obama, and 31 percent said they backed Clinton. Independents made up 43 percent of all voters polled.
But Clinton was ahead of Obama 45 percent to 34 percent among those who said they were registered Democrats. Those voters made up a majority -- 54 percent -- of all those respondents.
...
According to the exit polls, Clinton had a sizable lead over Obama among women, 47 percent to 34 percent. Analysts say that shift was crucial to the Clinton turnaround. "If I had a single word, the word would be 'women,' " said CNN political analyst Bill Schneider. "She got the women back."
Re: The death of polling as anything more than a g
so, maybe Michael Whouley is the political turnout god that he has been made out to be...i sure wouldn't want to go against him in the future...
Re: The death of polling as anything more than a g
I think whenever people feel like media is shoving a specific candidate down their throats they will pick the next available choice.
First they rebelled when Hillary was inevitable.
Now they rebel when Obama is inevitable.
Media should stick to reporting and not converting.
WE HAVE TO BEAT THEM
JUST LIKE WE HAVE TO BEAT THE GOP.
I LIKE Obama - but my biggest problem with him has been his buying into Dowd, Rich, Matthews, Russert, ARIANNA scripts and narratives bout the Clintons. these same jerks destroyed Gore and put bush in the White House . Screw them. They are NOT our friends...ever.
Read the daily Howler. Im meeting with him tommoroww. In coming weeks we will be announcing big news about MEDIA PUSHBACK, utilizing many of the strong voices on the Net. Maybe even the author of this diary's FUNNY, FUNNY headline.
Nice Eve Sir!
HA HA HA
NOW YOU KNOW WHY I WAS GETTING SO FRUSTRATED ALL THRESE MONTHS ABOUT SO MANY POLLING DIARIES. Discussing polling is fine if not overdone. So I was never against bringing up polling. It was even worse when you had new MYDD diarists act like polling experts and giving polling discussions a reverence it did not deserve.
You Jerome just made me laugh out loud!
That was a great headline! I will never forget it!
yeah and to the person above - the legend of Whouley's genius just grows and grows and grows...
Likely voter models ...
... probably were way off the mark on this one. My guess is that a lot of voters for Clinton were self-identified Dems who rarely vote in primaries. My further guess is that a lot of these Dems-who-don't-vote-in-primaries were working-class women (the very demographic comparatively less likely to attend an Iowa-style caucus due to scheduling concerns such as second-shift employment, child care etc.).
I'm not completely sold on the HRC's "showing emotion" moment as making the difference as I am on the possibility that the likely voter models under-counted a lot of HRC's support.
The "showing emotion" moment makes for a nice narrative, but the daily tracking poll missing a HRC "surge" suggests to me that the likely-voter models were a fundamental problem.
I'm no expert, though. Perhaps the polling gurus out there could weigh in?
Re: Likely voter models ...
i bet you're on to something. cnn just confirmed women turned out in big numbers for hillary. i think this shows hillary's campaign is able to bring in new voters (women), just as obama's campaign is able to bring in youth in bigger than expected numbers (or huck with evangelicals).
note that overall turnout in iowa and nh were much higher than before. so clearly, new voters are coming into the fold, perhaps throwing off the methodologies
by my calculation, i'm guessing 59% of nh voters voted dem. that's great news for us
Re: Likely voter models ...
I don't know if this unaccounted-for group are new voters in as much as they are new primary voters. Maybe they are new voters. I bet a lot of this group, however, are self-identified Dems or Dem-leaners who only vote in the general.
(Of course, this is further hypothesizing upon previous hypothesizing on my part ... I could be way off.)
Re: Likely voter models ...
... and as I sort of hinted at (and I believe HRC has said this herself), the Caucus system is the least-likely system to allow participation from a demographic such as working-class women.
First, there's the whole issue of being able to attend a caucus at a given time and location. Many working-class women are working a job at 7 in the evening, taking care of children, etc. In a primary, you have a 12-hour window to vote, so its less difficult to find a moment to head down to the polling place and vote.
Second, caucuses are kind of foreboding. They're often very "clubby" affairs dominated by strong personalities and/or party regular-types. If you don't think of yourself as "part of the club," you're less likely to want to attend. I'm guessing that a lot of this demographic don't think of themselves as "part of the club" (and often they're probably right to think that).
Third, caucuses can take a long time, or at least a lot of people think they take a long time. Usually it's all settled in 45 minutes or so, but you never know for sure. Like my first point, the caucus system's inflexibility keeps a lot of people away.
Re: Likely voter models ...
yes, good distinction
the theories on the table so far (from here, kos and andrew sullivan) are:
- people resented coronation and switched vote
- undecideds made late break for hillary
- the tears swayed people over
- independents voted for mccain b/c they thought obama had this wrapped up
- race factor
- gender factor 1 - lots of women came out for hillary, more than polls predicted
- gender factor 2 - a sort of reverse-race factor: women voted their gender in the secrecy of the ballot box; couldn't do this in iowa in the open
i'm sure the numbers will be parsed in the next couple days, and we can get the data to start narrowing it down
Found Pollster Average
http://www.pollster.com/SensNHDem.png
Pollster has Obama winning by 7.4 points. Don't leave out our guys in your criticisms Jerome.
You left a column off
UNDECIDED
None of those add to 100.
Here are the undecided counts. For each poll, there are 9-10 % left off.
I am a statistician. I work in the med school at a major research university in the midwest. I am not a poll specialist. I will say one thing: The standard formulae for standard errors are correct for SINGLE PROPORTIONS. With MULTIPLE CASES, it's much more complicated. And with if you consider it one v other, the standard errors get much much higher.
Re: The death of polling as anything more than a g
I'd never say one blown election makes for the death, but obviously a thorough reconsideration of methodolog is needed. It could just be timing - that late breaking events like Clinton's emotional softening had a huge impact, and no poll caught it, but Obama really would have won two days ago. It could be that the Independents weren't properly reflected, or that cell phones mattered sooner than pollsters expected - but we won't know about "death" for quite some time.
Even More Unusual
Questioning of the polling is necessary, of course. What I was thinking earlier and the table confirms, is that the polling was only (that's "only") off for Clinton.
Scan the table:
- Obama's actual % is pretty close, well within the margin of error, to the polls
- Edward's actual % is pretty close, well within the margin of error, to the polls
- Richardson's actual % is pretty close, well within the margin of error, to the polls
Only Clinton's % is WAY off. I've never seen a poll that was wrong, but only for one of the candidates. But, here we have seven polls with this very unusual pattern. Really unusual.
As I noted
each of the polls is missing between 7-9 % of undecided voters. None of the rows add to 100 %. SO, what has happened is that the undecideds broke for Hillary pretty solidly. That's unusual, but if they were women, I could see it.
Re: The death of polling as anything more than a g
Even Iowa wasn't all that accurate... We are getting record turnout and that's got to be fucking up the polls too... LV models are JUST NOT falling into line. Senateguru has an interesting theory as well.
Iowa is not consistent with polling models
Iowa is not consistent with polling. It is a two- or three-choice-point situation. First, you declare for your first choice. If your first choice is too unpopular, you can either go for the second or go home.
That's extremely difficult to model. Just a bear.
You can add Clinton's internal polling
Russert reports that Clinton's own internal polling had her trailing by 11.
A stunning poll failure
My guess is as the election came closer women rallied to Hillary with the nasty stuff in the media against her, and we misjudged how independents would vote in the Republican vs Democratic primaries. In anycase this is a day many candidates coming into election as the underdogs according to polls will point to to try to encourage their base.
Turnout numbers posted - good news
NYT reports. CAPS are my comments
"Exit polls suggested that there was a record turnout, with half a million voters -- 280,000 Democrats, and 230,000 Republicans."
THAT MEANS 55% OF VOTERS IN AGGREGATED VOTED IN THE DEMOCRATIC PRIMARY.
In the Republican primary, Mr. McCain got 38 percent of voters unaffiliated with either party, and the same proportion of registered Republicans, according to a surveys of voters leaving the polls conducted by Edison/Mitofsky for the television networks and The Associated Press. Such undeclared voters made up about a third of voters in the Republican primary.
SO MCCAIN GO THE SAME SUPPORT FROM REPUGS AS INDIES
It was different for the Democrats. Undeclared voters make up a larger portion of the voters in the Democratic primary -- about 40 percent. Mr. Obama got about 4 in 10 undeclared voters and Mrs. Clinton got about a third of their support. Mrs. Clinton got 45 percent of registered Democrats, and Mr. Obama got a third.
INDIES MADE UP 40% OF DEM VOTERS, A HIGHER PERCENTAGE OF A LARGER POOL OF VOTERS.
DOING THE MATH, THIS IMPLIES INDIES VOTED 60% IN THE DEM PRIMARY AND 40% IN THE REP PRIMARY. THESE ARE QUITE CONSISTENT WITH THE POLLS.
SO THESE FIGURES REFUTE THE NOTION THAT INDIES BROKE FOR MCCAIN OVER OBAMA IN LARGER THAN EXPECTED NUMBERS.
THE GOOD NEWS IN ALL THIS IS THAT NH VOTERS IDENTIFIED WITH DEMS.
Bury them altready ...PLEASE !!!!!!
What is so amazing to me is not how wrong the polls were, they are wrong far more often than they are even close to being right. What is amazing is reading so many comments below where people are claiming anomalies or some other bullshit trying to prove that polls really do work. I guess polls really do work as long as you never use them to try and guess what human beings will do.
I know some people here live and die by polls but it seems to me you are wasting your life on falsities.
Get a life, turn off the TV, stop reading the polls and get out on the streets.
Re: The death of polling as anything more than a g
Dis the sample of any of those polls include 57% women? (Pissed off women!)The Tweety syndrome:
Re: The death of polling as anything more than a g
The biggest lesson here:
Any poll prior to the day of the election that doesn't show "Undecided" with a sizable lead over everyone else is not to be trusted, and the result of the pollster pushing poll respondents to make a choice when they really haven't yet.
Another lesson:
Don't rely so heavily on the polls in the year-long buildup to the primaries. They are entirely meaningless until probably December, and even then they are only a rough approximation.
-----------
myDD - skin
-----------
60 Comments