Ms. Zogby
The blogosphere and the cable channels are filled with news of Barack Obama's "surge" especially in California. Although you would never know it to listen to MSNBC, almost without exception, the support for Obama taking California rests on the findings of a single pollster, the Zogby organization. All the other polls since Edwards dropped out range from a single point distinction to a twelve point advantage for Hillary Clinton. Why is Zogby so far from all the other polls?
Perhaps because the Zogby samples are 52% female.
The Democratic contests this year have averaged 58.5% female. There is some indication California will be even more female, possibly as high as 60%.
Women are a notoriously unreliable voting bloc, and at times polling has borne an ugly resemblance to the real estate market, but unless Senator Obama has succeeded in peeling off Senator Clinton's heavy voting white female support -- which has of late been pretty stable -- all the heavy breathing about his dominance is er premature.
Well, gosh. Where to start?
Let's begin with the fact that three polls released in the past couple of days show an Obama lead, not just Zogby. Suffolk (Obama +1) also uses a 52% female sample, and I didn't bother paying for Rasmussen's (Obama +1) crosstabs.
It's true that SurveyUSA, which produced the eye-popping 12 point Hillary margin, used a 59% female sample. On the other hand, the Field Poll, which is far more reputable and uses a more robust methodology, included a 56% female sample, but put Hillary's lead at only 2. That was similar to Mason-Dixon's 57%, but they put Hillary's lead at 9. In other words, there's more at play here than gender.
There are many reasons to discount Zogby's results, not the least because of his general unreliability. But he's been downright oracular compared to SurveyUSA, which produced the 12-point lead. The truth is that each of these polls has its flaws. Most under-represent voters younger than 35 by as much as they undercount women. Several are robo-polls. Throughout the cycle, black turnout and support for Obama has significantly exceeded the predictions in polls.
Here's the key thing to know, going into tomorrow's vote. In every single poll taken more than once in California, Obama's percent of the vote has increased significantly. That, not Zogby, accounts for talk of a surge. Over the past week, the pace of Obama's gains has increased, and Hillary's lead has eroded across the board. That's true of polls with a large sample of women, and those which undersampled. It's true of human polls and robo-polls, of out-of-state pollsters and experienced California pros. The average spread in RealClearPolitics' rolling tally is now just +2.6 for Clinton; that puts SurveyUSA further from the average than Zogby. It also means that we're now within the margin of error, and that tomorrow's result is unpredictable. No one, to the best of my knowledge, has proclaimed Obama's position in California to be dominant, or breathed heavily while they did it.
So let's not go hunting for anti-female conspiracies where they don't exist. I don't know who's going to win CA tomorrow, but I do know momentum when I see it. And it doesn't take Zogby to point it out.
February 4, 2008 9:15 PM | Reply | Permalink
I miss the ability to rate comments. And yours deserves a high rating. Once again, thanks for your excellent analysis and commentary.
February 4, 2008 10:42 PM | Reply | Permalink
58.5% women. why so high?
this time, because of clinton?
women vote more in primaries?
women vote more in general?
men are more likely to be republican?
are there just that many more women than men in the age ranges that vote?
February 4, 2008 11:11 PM | Reply | Permalink
Fly on the Wall said (where to start).
"HERE'S THE KEY THING. In every single poll taken more than once in California, Obama's percent of the vote has increased significantly." As it happens with a click of the mouse, this is embarrassingly easily demonstrated to be factually false: in Survey USA polls taken September to December, Obama's percent of the vote was: 27, 20, 25, 24.
It is also amazingly specious. Almost no polls have repeated for the time we care about, the last two weeks. In the ones that repeated, Rasmussen and Survey USA, BOTH candidates' percent of the vote increased, probably because Edwards dropped out.
For example, on 1/27, Survey USA reported Clinton ahead by 11. On 2/2--2/3, Survey USA reported Clinton at 53, up 4 points from the prior survey and ahead of Obama, up 3 points from the prior survey, by 12.
I cannot understand why the heat of the rhetoric is so seductive it leads people to write things that are so easy to rebut. I am looking at a video of Josh Marshall from today, Monday, 2/4 saying that the high lead number for Hillary in California (8) is old and the high number for Barack Obama (6) is new. At the time that video aired ARG showed her 8 points ahead in CA (2/1--2/2) and Barack Obama's highest lead was 4 points in CA (1/30-2/2). Where exactly did that newest 6 points come from?
And how about this from the Politico for more heavy breathing:
"Polls show Obama surge in California"
By: Chris Frates
Feb 4, 2008 11:21 AM EST
followed by a story mostly about the Zogby poll and that also mysteriously fails to mention the ARG poll.
You are correct: the polls are impenetrable and flawed, Survey USA is not the Delphic oracle, and doubtless many things besides gender are operating to make them unreliable, and maybe Barack Obama will win the CA primary overwhelmingly. But what you -- and Josh Marshall -- have said about the polling data you yourselves invoke is at best misleading and sometimes downright false. What is wrong with this picture?
February 4, 2008 11:49 PM | Reply | Permalink
Linda, Tough words for someone who doesn't seem familiar with the data. The 6 point Obama lead is today's Zogby poll. Please check. The big range, 9 points in M&D is also one of the two oldest. If people are not giving a lot of credence to the ARG poll that is likely because they have thus far compiled the worst record of any polling organization during the primaries. That is why have started giving them far less attention because they have consistently missed the mark in almost every contest. In any case, as I made clear in the video, it is over a very short period and thus it is difficult to draw too much from it. But Obama's lead are unmistakably clustered toward the most recent surveys, both in California and nationally. Please don't accuse others of peddling falsehoods when you don't even have your facts right.
February 5, 2008 12:27 AM | Reply | Permalink
Linda,
I guess there is a conspiracy against the Clintons inside their own political party. Our defense of the Clintons over the years has just been a ploy to get them to lower their guard- then wham!- hit them with some misleading poll numbers for the state of California.
Crying foul on poll numbers does not look good coming from a camp that has been throwing curveballs for the past few weeks.
February 5, 2008 1:31 AM | Reply | Permalink
Polls? We don’t need no stinking polls.
Let’s slowly get to a point here, as if we should tame our meandering spirits just enough to make a distinct, objective analysis. Well, it ain’t going to happen, simply because our times do not require it, and my mind this year is in core, instinctive mode. I’ve been ground level all my life, and have the scars to prove it.
I’ve learned enough about life to maintain my priorities. Like realizing that the innate strength of life as an American is in the very quality of our now seemingly lost, but retrievable continuity, that our leaders must be of a tested nature of layered experience, and that most importantly, they be charged with the ability and insight to lead a nation, not a party. I deeply believe that Senator Clinton possesses these qualities, in fact she appears to me to possess a heretofore sheltered ability to enable political communication, a personal attribute that resonants from her life as a young woman, and as a woman, has been further enhanced by her experience and involvement in one of the most incipient revolutions of our times.
She is the best we got, my faith on that point is simply irretrievable.
February 5, 2008 3:50 AM | Reply | Permalink
Linda,
How can you accuse me of failing to pay attention to the past two weeks, and then cite the SurveyUSA polls only until the beginning of December? No one talked about an Obama surge in California in early December. Let's complete that data series you present: 27, 20, 25, 24, 30, 35, 38, 41, 42. That last figure is from tonight, when even SurveyUSA found Clinton's lead continuing to shrink, from 12 down to 10.
But of course, you knew that - you cite Clinton's numbers and her margin from the some of the same polls. This is just fundamentally dishonest; you should be ashamed of yourself. Let's also extend the Clinton series backward in time, instead of cherry-picking two polls: 51, 57, 53, 50, 49, 50, 49, 53, 52. So since September, Clinton's numbers have clustered fairly tightly, without a significant trend upwards or downwards, while Obama has gone from polling in the twenties to polling in the forties. And you can't find a surge here?
A final note: Zogby has, in fact, run a tracking poll over the past three days. Clinton has sunk - 41, 40, 36. Obama has risen - 45, 46, 49. (He actually has Obama now leading among women, 45-42, which sounds crazy to me, but there you are.) Rasmussen polled twice in the past two weeks: Clinton - 43, 44. Obama - 40, 45. So did Field: Clinton - 39, 36. Obama - 27, 34. Noticing a trend? And while it's true that Rasmussen shows an increase for both candidates, that's one point for Clinton and five for Obama.
Again, I don't pretend to know what's going to happen tomorrow in California. But I can state with a high degree of certitude what the polls have shown over the past few weeks - and that's a surge for Obama.
February 5, 2008 4:04 AM | Reply | Permalink
Rank amateur here with a question:
I understand that independents are allowed to vote in the Dem. Primary in California. I am assuming that this factor is being weighted in all of the polls?
And do more men or women register as Independents in CA?
Just curious. Sounds like you guys know more than me on this score.
February 5, 2008 9:56 AM | Reply | Permalink
As of January 22, 19.37% of California voters were "decline to state" - what I'm terming independents. The Secretary of State doesn't offer a gender breakdown. The best indication I've found is an out-of-date Field Poll on California's political demography from 2002. Back then, decline-to-state and third party registrants composed 19% of registerered voters, and tilted male by a margin of roughly 60-40.
Of course, that doesn't answer the real question you're asking - what will the gender composition of independents who vote in today's Democratic primary look like? There's only one poll that I've found that attempts to answer that. Suffolk found them to be men by a margin of 57-43, even though its broader sample tilted female 52-48. Of course, there're a lot of problems with those numbers. For one thing, Suffolk also estimates them at a ridiculously high 25% of today's Democratic voters. And Linda's absolutely right that a 52-48 split likely undercounts women overall, and that bias may extend to the sub-sample of independents. (Of course, that works the other way, too - overestimating the number of independents accounts for much of the skew in the broader sample.) But the biggest problem with the Suffolk poll is that the sub-sample of independents is too small to allow us to draw meaningful conclusions.
Still, so far as I know, it's all we've got. And it does roughly jibe with the outdated Field numbers from 2002. So unless some more knowledgeable poster cares to chime in, I'd suggest that independents tilt heavily male in California, and that's one reason they're a group with which Obama runs well.
February 5, 2008 11:15 AM | Reply | Permalink
Josh,
I like facts, and here are the facts. You made your video reporting what must have been the just-released Zogby figure of +6 Obama at 2/1-2/3, and told us it showed Obama's momentum. At that moment there were two precisely contemporaneous polls on pollster.com showing Clinton up 8 (2/1--2/2)(ARG) and 12 (2/2--2/3)(SurveyUSA).
Now you say you didn't tell your viewers about those polls because you and the other cognoscenti don't think ARG is reliable. You are entitled to cherry pick the polls (it's your website), which you have now told me, but shouldn't you tell all your viewers?
And your aversion to ARG still does not explain the absence of reference to SurveyUSA as contemporaneous with the Zogby.
In the end all the up to date pro-Clinton data was left out. As I said you -- and Zogby -- may be right, but I have come to expect a much more careful version of the facts here on TPM than that late-breaking video reflects.
February 5, 2008 12:06 PM | Reply | Permalink
Linda:
I'm not really up on arguing which polling methodology is best, but I do think you have a compelling point that Zogby's numbers must be called into question because, in California for example, the percentage of females included in the poll was around 52 percent. That seems to be a glaring underestimation of the female vote. Is that Zogby's practice in all states?
Guess we'll see tonight which of the pollsters are most accurate, but to me it would seem to me that methodological disputes are something that can and should be discussed a bit more BEFORE the actual results are seen, particularly be folks who are posting polling data to their readers, listeners or viewers.
Bruce
February 5, 2008 12:41 PM | Reply | Permalink
Linda, If you like facts, these must be hard times for you, since you seem to have very few of them. When we taped the show, Zogby's numbers had been released, while SUSA's weren't released until later in the day. So you're wrong on that count. Second, we didn't ignore ARG. We said that the biggest margin difference was one of the oldest, i.e., Mason-Dixon. And that's true. What I mentioned with regards to ARG was to give you some background on our editorial focus. On both counts, you simply appear to be ignorant of key facts about the available polls. What we included in yesterday's show was 100% accurate at the time of taping. Subsequent information we posted on the site. I'm happy with that. Meanwhile, you have all the data available to you. And yet, you've gotten most of the facts wrong and used your errors to make all these silly attacks about how people are keeping the good news for Hillary secret.
February 5, 2008 12:42 PM | Reply | Permalink
No editing of typos with the new system? ERG!
February 5, 2008 12:43 PM | Reply | Permalink
How does early-voting in states that permit it figure into the polls? Are people who have already voted included in a survey of likely voters?
February 5, 2008 1:29 PM | Reply | Permalink
Josh,
This is probably part of your "tempers" sequence, but it's super tuesday and I have nothing to do but wait.
So in my Inspector Javert-like way, I nosed around and found out the fact that the video you made leaving off the SUSA numbers had a time stamp of 15:36 pm ET (3:36 pm) Feb.4. (Does that mean you started filming at 3:36 pm?) And the fact that the SUSA organization put up the numbers you did not cite on their website at 3 pm, the survey itself was released to its west coast media customers much earlier, at 10:50 A.M. ET and may have been public information any time after that. Maybe you were all locked up in the studio filming your video? But we poll obsessives sort of know to keep checking, no? Since they are coming out so quickly.
In fairness, the later posting about the SUSA alerted the site reader, although I don't think it offset the dominant narrative. A picture and a thousand words and all.
I thought the gender cross tabs on the Zogby were worth thinking about, regardless of the outcome tonight, and actually did not anticipate a pissing match until flyonthewall started with the "every single survey" business (turns out he didn't mean every single survey, just the recent ones and then only some of those).
I have always wondered why people say things like "every single" in online arguments. It just made this old trial lawyer's mouth water. Anyway, it's something to do while we all wait.
February 5, 2008 2:58 PM | Reply | Permalink
Linda, I've never had any beef with the arguments about whether Zogby is under-estimating the female vote. I'm not sure. And as you say, we'll find out soon enough. As for your sleuthing, I'm afraid you're going to keep coming up snake eyes on this one. The time stamp is likely when it was uploaded. So no, filming comes well before uploading the edited video. The video was filmed just after 1 PM, I believe, and then edited and uploaded roughly two hours later. It was posted on the site just before for. A little more sleuthing would show that SUSAs public release of the poll was just after 3 PM. The email they sent out was received by us around 3:30 PM. Anyway, probably more details than most readers are interested in knowing or care about. I think you're about 0 for 6 at this point. But if you want to keep raising hypotheses and incorrect data I'm game to keep batting them down.
February 5, 2008 3:50 PM | Reply | Permalink
Linda, I think I will be checking out your intuitions/prognostications in the future.
From exit polls, just announced a few minutes ago on MSNBC: women were 55% of California Dem primary voters, and they went for Clinton 57% and Obama 39%.
February 6, 2008 12:37 AM | Reply | Permalink
artappraiser:
As it happens, Linda's figures for the percentage of women (58.5% and 60%) were further from the actual percentage than Zogby's. It turns out that Obama did surge - from down by 20+ to a final 10 point gap - but that many polls overstated that surge, none more egregiously than Zogby. But if you're looking for a monocausal explanation of those errors, I'd suggest that race had a whole lot more to do with it than gender. Let me crib from Mark Blumenthal at pollster.com:
"Latinos are currently showing as 29% of the Democratic vote in California on the exit poll tabulations. In the final pre-elections polls they were:
26% on SurveyUSA
25% on Mason-Dixon
20% on Field
9% on Suffolk."
Those polls put Clinton at +10, +9, +2 and -1, respectively. And Latinos supported Clinton, 69-29%.
To be sure, polling is a complicated business, and many factors play a role in the results. But I haven't seen any evidence this morning that systematic underestimation of women voters was a controlling variable. Latino voters came out to the polls in unprecedented numbers in California, sharply increasing their percentage of the primary electorate; blacks did not. That, in the end, seems to account for much of the variation in the polls.
February 6, 2008 10:40 AM | Reply | Permalink