Obama '08?

www.startribune.com

Underestimate Oprah's influence if it makes you feel better but...

Miller is drawn to black men. It frightens her.

Fox News

Strong words from someone who got booed yesterday in Iowa.

Cycloptichorn

"Who's doing it for political reasons, and who has a lifetime of conviction and commitment?"


I agree with her. She should have a conviction on her record and be committed to a lifetime of actually working.

Per Pollster.com



It's pretty clear that Obama is hoovering up some of the support from Edwards and Richardson, in Iowa at least.

Nationally:



Hillary is still way ahead on name recognition.

Cycloptichorn

I believe woiyo recognizes her name.

I wonder what happened in early June/ late May to cause that dive from Edwards? (A debate?) Obama didn't really start picking up until mid-August or so.

I read it as end of april beginning of may. First dem debate was april 26.

adding... I don't recall Edwards doing poorly. I recall Hillary doing very well, followed by a consensus that she was, followed by a growing notion of inevitability. My guess is that that is what hit Edwards there. Obama seems to have a separate and quite loyal following less susceptible to wind change, if you will.

I don't think so.



The rectangles are quarters. Quarter 1 -- Jan, Feb, March. Quarter 2 -- April, May, June. Edwards flatlines April-ish but then doesn't start the drop until about 2/3rds through the quarter, which would make it the end of May or beginning of June.

Could just be a delayed reaction though I guess.

Yup, you are right. Not sure what other factor might account for it but we're working from memory, of course. A trend that strong (edwards downward) seems likely to have an event at source which is then made stronger through some building of consensus.

Edwards started going Neg against Hillary around that time, IIRC. Met with mixed results.

Also - how often do you hear about the Edwards campaign? They've dropped off of the media map completely.

Cycloptichorn

Edwards had practically been campaigning in Iowa for years, long before anyone else was thinking of these elections; they joked that he lived there. He also enjoyed residual support from his 2004 run. Those two factors explained why, while he was trailing far behind nationally, he started out well positioned in Iowa when the campaign began, in first place even.

Soon as the campaign did start for real however, in April-May-June, the other candidates started moving in. Debates were held. National publicity picked up, with most all of it focusing on Hillary and Obama. And so Edwards slowly had to concede ground, increasingly so as his competitors established field offices, travelled across the state, and as Cyclo points out, the national media narrative pretty much dropped Edwards from view altogether.

Of course Edwards also travelled the state - but he was basically playing defense, trying to hold on to an anomalous lead that was based on his prior activities there.

Thats an overreading of the data. The line is merely a trendline - a running average of sorts (just a very, very cautious one). The pink dots represent Edwards' results in the individual polls, which were still rather thin on the ground -- there wasnt the kind of polling going on like you have now. As you can see they varied greatly from poll to poll at the time. He was scoring anything between 21% and 34% in individual polls round that time!

Now the trendline tries to make sense of all that and at least roughly pick up where the overall movement was going, but thats all. In short, you can deduce from the graph that somewhere between March and June, Edwards reached his "peak" and then started sliding down; but thats about it. There's little sense in going deeper and trying to define the exact month or week in which something must have happened.

And not just because the variation in the polls was far too great to be able to isolate any one such point in time. Public attention, even among Iowan primary voters, was still so scarce that there couldnt have been one defining event, simply because no one event ever caught enough public attention to create some kind of turnaround. That includes the debates - viewing numbers were (and are!) just too low for any one of them to suddenly trigger a turnabout.

So you have to look at larger things: what kind of developments were taking place around that time? Started taking place around that time? The debates do start having an effect cumulatively, of course, especially as media coverage of them rolls one layer of impressions over another. And you get to the things I mentioned above: it's no longer just Edwards travelling all over the state. Obama was only just introducing himself to the locals, most of whom didnt know much more than his famous speech. Hillary slowly started doing the retail politics. Offices were set up in local communities.

And you probably cant stress enough the impact, over time, of media coverage that completely focused on Hillary vs. Obama and all but ignored Edwards.

You got to note also that the pollster.com trendlines are designed explicitly to roll over any sudden movement in the polls, out of a wariness to pick up too much on momentary variation: it's really intended to recognize long term trends only (something that's a little frustrating sometimes).

In this case, those long-term trends were Edwards sliding down a bit; Hillary going up and picking up speed especially over summer, when her inevitability status seemed to be rooting, and slowing down a bit again in the last couple of months; and Obama being 'in the lift' since summer. If you want to tease out more short-term trends and the impact of specific events (a debate, a revelation, etc), than the trendlines aren't of much use. Thats why I impatiently started making my own graphs.. I'll post some later.

Those come with a caveat, though, because the design of pollster.com's trendlines implicitly present a warning: the reason those dont provide such detail is because the polling expert behind the design doesnt believe that opinion polls can track and signal developments in more detail or on shorter terms - at least not until they become much thicker on the ground, like they're getting only now.

Nope, I'd disagree with that - see the post above.

For one, we're not talking about "that strong" a trend; we're talking about Edwards, on average, going from 27% to 22% - in half a year's time! Meaning that individual polls had him anywhere between 21% and 34% in June, and anywhere between 19% and 25% in November. Be a smart man to identify any drastic trigger event in that .. just some gradual sliding over time.


Its true that CW has it that Iowans respond badly to negative campaigning, but - well, I'm also just going on memory here - but I think Edwards only became really strident in the last couple of months...

He's not actually doing too badly with it, either, tho its obviously hasnt gotten him the much-needed breakthrough; the pollster.com trendline isnt picking up on it yet, but if you look at the dots you can see that there appears to be a distinct uptick in his support again since September or so. Same in New Hampshire, where Hillary appears to be declining quite notedly since September (though its not been long enough yet for it to show up in the pollster.com trendline), and Obama hasnt had the bounce he clearly seems to be having in Iowa this month. Wait and see - tho time's running out...

Isn't that about the time that the story about Edwards' $400 hair cut surfaced? Maybe that's what caused the dip.

I doubt it was any one single thing, but it certainly wont have helped..

OK, so I've been improvising a bit...

What I've been doing is taking the individual poll results (on both Iowa and NH, but right now I'm just talking Iowa), as registered on pollster.com, and calculating month-to-month averages for the period since June.

Well, as close to month-to-month averages as you can responsibly get, because I didnt want to base any averages on less than 5 polls at a time, so what you get is an average for June + July, one for August + most of September, one for late September + October, one for the first half of November, and one for the second half of November.

The result is a bit more sensitive to short-term trends than the pollster.com trendlines -- but also more vulnerable to the impact of outliers and other statistical "noise".

I've plotted the results out in this graph, on the right, with for comparison's sake the corresponding bit (1 June - 31 December) from the pollster.com graph on the left.



As you can see, the additional graph I made yields some extra stuff to speculate on. For example, it appears to show that in the first half of November, Hillary's rapid rise was suddenly reversed and she started losing support. At first, it was not Obama who was benefiting, however - he was stagnating, in fact. It seemed like Edwards and Richardson were picking up support instead. But then came the last two-three weeks, and now it's Obama who seems to be positively surging.

Are those short-term trends that the cautious pollster.com trendline doesnt pick up on (yet)? Or is it just random statistical variation, caused by the differences in methodologies of different pollsters, margins of error, etc? Quite possibly, a lot of it is the latter.

On Obama's steady rise in the Iowa polls, in any case, both graphs agree. They also agree that Richardson's best days are over, for now. The graph on the right does seem to show, at least, two additional developments that don't show up in the Pollster trendlines (yet?): Hillary's rise being actually reversed, not just weakened; and Edwards' fall being reversed as well, not just slowed down.

In addition, they may be showing that Obama now really has the momentum, with his rise in the polls escalating. But we'd better wait to see what, if any, impact the Hillary "hostage crisis" and her health care attacks on Obama will have - we should know more in a week or two!