Get yer polls, bets, numbers & pretty graphs! Elections 2008

Obama looks like he's in trouble in NH. Still predicting a comeback?
Btw, here's his and Hill's recent charts. Don't look promising.


By the time I prepared this; Hill jumped to 99%

I believe the fat lady has sung.

AP says she sang.

I want to post a snippet of analysis from pollster.com from last night: it's interesting for those who are interested in polls, but also for Obama supporters looking for ways to understand his last moment-loss.

Key phrase: the "critical importance of college educated women as swing voters".


Source

Hmm. That's sounding like women reacting to the reaction to the tears (if you follow). What Butrflynet said on the Obama thread, and I get what she meant -- some of what was being said about Hillary was so over the top, and then the guys with the "Iron my shirts," and... I'm firmly in the Obama camp (obviously) but I did get that "take that, sexist idiots!" urge myself. If enough people were wavering... (sigh).

Now here's a headline that one wouldnt have expected..

Troublesome 'graph for Obama supporters: "Turnout broke records [in Iowa], and the influx of caucus-goers favored Obama. But high turnout in the caucuses is still low turnout, compared with a primary in New Hampshire. Tonight's election saw, in effect, a general election turnout, and it appears the party regulars in the state's most heavily populated areas more than made up for whatever new voters Obama brought out."

Obama's bulwarks remain the college towns - the same ones Howard Dean won.. "Regular folk" towns went for Hillary.

Interesting.

Still wondering if it could have something to do with that last run of polls... less urgency. ("Obama's gonna win...")

I was thinking that the fundamental difference between a caucus and an election has something to do with it. You don't have wasted votes in a caucus because you get to change your mind. Not so in a vote. Also, if Obama needs crossover and independent support he's going to have a tough fight in closed primary states.

Hey poll dude (that's you, nimh ;-)), what do you make of the discrepancy between the last batch of polls and the results?

1/7/2008


1/8/2008

Hillary Clinton: 39%
Barack Obama: 36%
John Edwards: 17%

Zogby's analysis -- makes sense to me:

http://www.huffingtonpost.com/john-zogby/polling-the-new-hampshire_b_80657.html

Especially this one:



I would add 2.5:

Many people, especially women, were extremely turned off by the coverage of Clinton in the last few days before the primary and reacted by shifting their allegiance to her. I agree with this letter, for example.

An even better letter to Sullivan, same link as above:



Couldn't agree more.

Cycloptichorn

Yep, I posted that one somewhere else (and apologized for the Sullivan binge. He's had a lot of good stuff lately though).

The pollsters certainly will have to do some searching as to what they missed. One of my customers, without any angst towards Clinton, commented that voters may say they could vote for a black person, but once in the voting booth, will quickly vote for the non-black.

I'll retake a few strictly polling comments.

1. A problem when calculating and merging diverse polls is the quality of the data. One must separate the pros from the non-so-trustable from the quack, "to separate the beans from the pebbles", as we say here. Mexican pollsters fit all in one big room and they (we) know each other too well. But the US has tons of pollsters and it's not easy to tell.

2. The problem with the late NH polls was more the interpretation than the polls themselves. We saw a swing and supposed the [Obama] momentum effect was still on the voting day. No wonder they call it "swing": it was a temporary shift. Plus there is the huge possibility that, even if voters were in "record numbers" (talk about commonplace), several pollsters may have overestimated the turnout (thus, overestimating independent participation).
Nevertheless, yesterday I saw the NYT poll, noticed that the sample was relatively small and said to myself: "Hey, Obama and Hillary's vote intentions cross below the margin of error!".
There we have two problems: one is that people -and media, mostly- tend to look at straight figures, not ranges. A poll gives you a range. The other one is that, anyway, it was odd that most polls were "accurate" in the edge of the range. Means there was a non-house effect: a true voter shift.

Well, in the end, voters are humans and the ways of humans are deep and mysterious. Or superficial and mysterious, maybe. :-)

OK, I'll give it a shot. OK, some random shots.

I dont think there is a single reason for Hillary's surprise surge in the actual vote. I dont buy that it was just "the tears", or "Bill's speech".

But here's a smart observation by commenter Petey on Matt Yglesias' blog:

"Folks tend to forget how much of an election gets decided in the last 48 hours, which is after most to all of the polling gets done. You can have a perfectly accurate poll which is upended by the late surge. Elections tend to break at the end in one direction rather than the other. Between Hill's tears and Bill's rant, the Clintons owned TV coverage in the last 48 hours. That matters."

Now there's an essential thing to realise here about the polls. In the other thread, Butrflynet asked rhetorically, "Have the polls ever been so wrong?" But what did the polls get so wrong? Hillary's support and Hillary's support only, Matt Yglesias himself points out in that blog entry. And he's got a graph to argue it with:


Interesting, no? The comparison you made above with the table already showed the same thing: the polls were actually very close to both Obama's and Edwards' support. It's only Hillary that remained completely underreported.

Obama getting as much as he was polled for certainly seems to belie, at first blush, the argument that Obama suffered from many of his potential voters staying home, or voting McCain in the Republican race instead, out of a sense of "less urgency" being involved in the Democratic race.

(Another problem with that theory that Obama was leading Hillary in the polls by roughly the same margin as McCain led Romney, so it's not immediately clear why those polls would have made people think McCain needed their vote more.)

So if polls were "only" off on Hillary, where did Hillary get that extra 10% from? If neither Obama nor Edwards got much less than they had been polling for? The undecideds, say Yglesias and Brian. The polls still had about 10% undecideds on average, and since Hillary got 10% extra and noone else moved, she must have mopped up that 10%.

That's too facile, I think. Reality doesn't conform to such straightforward equations. If you see +10% or -5% on the last day, it's just the sum of people both coming in and going out, there'll be some of both. It's not like every last undecided voter really went to Hillary that last day - far from it, as Cyclo pointed out.

But still, if you'd go just on the media's talking points, you'd think that massive numbers of would-be Obama voters switched to Hillary at the last moment. But considering that Obama actually got pretty much what he was polling for, that doesnt seem so logical. Some of Hillary's surge must have come from the undecideds instead. And you could also read the exit polls as confirming this. I mean, among those who decided sometime last week, Hillary got just 28%; among those who decided in the last three days, she got 34%; and among those who decided just today, 39%.

But there's another dimension still. From what I understand, one of the major uncertainties in a poll's set-up, the single main one even, is not necessarily in how it pegs the proportions between the candidates among the likely voters it identifies; it's in whether its model of who is a "likely voter" is correct.

Like, almost every pollster uses a "likely voter" model to filter the results it gets - fine, they say, the raw data says Hillary has X%, but are all those people actually likely to turn out? And different pollsters use different methods to estimate whether someone is likely to actually vote or not. And that's perhaps where the biggest gamble is.

Part of the method might be based on whether people voted before, but what if today's candidates mobilise a distinctly different selection of possible voters than yesterday's? Part of the method might be based on how sure people say they are to go vote, but what if some last-moment development spurs one or the other demographic more to vote than others?

That's what commenter Dan Kervick posited: "Matt is wrong. Undecideds broke equally for Clinton and and Obama. The polls underestimated the extremely high level of Democratic voter turnout from Clinton's urban base." Along the same lines, commenter Brian wrote that "the polling shows Obama won all his traditional areas of college educated, wealthy, etc, [and] Hillary won her demographics" - so if you want to explain the results, he implies, you gotta look at which of those demographics turned out more strongly. Figure out why women suddenly turned out for Clinton in such large numbers, for example.

Now I dont know - to check out whether that's true you'd need to compare the turnout in, say, Manchester (city with blue-collar districts) vs that in Keene (college town), and then compare both with what turnout was in those locales in previous elections. That way you could see whether the proportion in turnout numbers between the two towns has shifted starkly. Or, in the same vein, you'd need to look at how turnout among college-educated women now compares to previous election cycles, in comparison with how turnout among other demographics increased. If it increased clearly above average, then it might be Hillary mobilising her basis more successfully than Obama rather than the switching of voters between camps that played the biggest role. I aint gonna do that tho.. ;-)

There is a depressing explanation in this context, which which Rjb just mentioned, and which Yglesias hints at: "note the relevance of this to Wilder/Bradley effect speculations". Eg - the theory that white people dont want to vote for a black man, but dont want to admit this even when questioned anonymously by a pollster. They'd answer - say - oh I have no preference, I dont know yet, I'm not going to vote. And then in the shelter of the voting booth, vote for the white candidate after all. That might have played a role among those older / blue-collar / urban voters.

Could be. It would be a way to explain the difference between Iowa, where Obama did much better than he'd been doing in the polls in the course of those very public caucuses, and New Hampshire, where he didnt do any better than he'd done in the polls in a secret vote.

But then, what weighs against that again is that you'd expect that Hillary, as a woman, would suffer herself from a similar effect as well (men who wouldnt admit that they could not vote for a woman, but go for a male candidate in the voting booth after all). And that the kind of person who would refuse to vote for a black man would hardly rush to support Hillary instead..

Regarding "why women voted so overwhelmingly for Hillary in the last 24 hours," posters like Kervick point to the tears; to the "ganging up" on Hillary by Edwards and Obama in the debate; the harsh reaction by Edwards to Hillary's tearful moment; and to fliers suggesting Obama wasn't 100% pro-choice that the Clinton campaign apparently spread widely at the last moment

Might all well be. But digging in all such detailed motivations for why that 5% or 10% of the voters switched between candidates or changed their mind about (not) turning up, the way I see it, does carry the risk of not seeing the forest for the trees anymore. Of looking right over the more fundamental things playing a role here.

Recap: Obama got 35% of the first preferences in Iowa (before the reallocation of the support for non-viable candidates etc). He got 36% of the vote in New Hampshire. In Iowa it was enough for a big win, in NH it wasnt enough for any win, but the number is pretty much the same. You gotta wonder - might there just be a ceiling to Obama's support, at least in overwhelmingly white states like IA and NH?

He's done overwhelmingly well in rallying a movement of enthused voters around himself, for sure. But that movement has a clear demographic profile. Of course there are older Obama supporters too, and yes, in Iowa he did as well as Hillary or better in every income group, for example. But he did much better than her in the higher income groups. Basically, in both states, he's had a specific set of groups to thank for getting a total score well in the 30s, rather than, say, in the high 20s. The young; the well-educated; the Indies rather than registered Democrats. The higher-income voters, the non-union households - things like that.

In SC he can focus on keeping black voters heartened enough to turn out for him regardless; adding up those and Obama's core white support among more upmarket Democrats should still give a very good shot at a win. But beyond SC, Obama faces the fundamental problem of increasing his appeal, his reach, to those other demographics. To the less-educated, the older, the rural and working class urban, to union households, to those who are most concerned and insecure about the economy (and their place in it). Among whites at least, there are still simply too many people in those groups who dont trust him enough.

Here's something from the exit polls that should make Obama worried. Among the 1 in 6 New Hampshire voters in the Democratic race who said that out of four qualities, "Cares about people like me" was the most important one, Obama got a measly 19% of the vote. He was trounced by both Clinton (41%) and Edwards (37%).

And the same thing happened in Iowa. Among the one in five voters there who listed "Cares about people like me" as the most important quality in a candidate, Obama got a mere 24% to Edwards' 44%.

In New Hampshire, among those who thought the economy is in a "poor" state, Obama lost to Hillary 31% to 44% - whereas he cleaned up among those who thought the economy is in a good state (46% to 32%). Among those who say their family's financial situation is "getting ahead", he cleaned up too (48% vs 31%); but among those who thought their family's financial situation is "falling behind", he lost to Hillary 33% to 43%.

In short, his soaring and inspirational, but rather abstract and high-minded rhetorics might just not be enough to win over what I'll call, for lack of a more comprehensive label, the white working class. His message of hope and confidence in the future might inspire many; but might also seem out of touch to the more acute concerns of the some who feel dread and insecurity about their economic future. Might seem simply too intangible.

His focus on process issues probably doesnt help. See Saturday's debate. Edwards defined how he was about change and effecting change by referring to health care and the people who cant afford it, to the patients' bill of rights he helped create. Obama jumped in eagerly and proudly segued into his goal to bring honesty back to DC, his anti-lobbying bill.

Obama's problem, way I see it, is that these are all the very same things that lift him out above the field, that have yielded him such devotion and such broad appeal (especially across the aisle). His inspirational speeches, his focus on issues of ethics, his appeal to hope and confidence. So changing course to win over the more sceptical white, urban, blue-collar vote could just lose him just as many Indies on the other side, who might switch to McCain for example.

Plus, this is what he's good at, it's his thing. But it might just do better among the Sozobes than the Bipolar Bears of America. ;-) And that might just not be good enough. Especially since in the 19 races coming up, Indies are only allowed to vote in half.

OK, that was a lot of rambling.

For clarity's sake: I guess the first post was more interesting in terms of reading the polls and testing the various interpretations, without ever reaching a straightforward explanation. While the second post is more about the meat of Obama's dilemma now, the way I see it anyhow.

Wow, fbaezer! I know you from that other extremely intellectual thread we post on together. Welcome, here.

He's the real expert here.. I wish he wrote posts as overlong as mine.. :wink: