Saturday, February 02, 2008

Galluping Along the Wrong Track

I'm always amused at how poll readers can't look at the data set in front of their noses and do a little serious thinking about the events that inspire the numbers. I'm looking at the Jan. 30 - Feb 1st Gallup poll information. Across the blogosphere, disappointed ObamaBots just can't believe that Hillary has gone up sharply in the polls. With Edwards dropping out, everyone knows that those supporters will go to The Golden One because everyone knows that everyone Hates Hillary! Let's examine the data. EDIT: Also, the Gallup poll is a three day tracking poll, used to smooth out highs and lows. Thus, the numbers are really trends of voters, and lag the daily indicators. A high point and a low point must be viewed against the subsequent data point. Abrupt changes indicate an even stronger reaction in the direction of the trend than is being reported.

The key here is to look at the poll movement along two dimensions - the first being the events happening in the days immediately before the poll result, and then look at the pattern of movement between the three candidates. My argument is that while Hillary and Barry were the ones making the big news and setting the trends, it was the behavior of Edwards supporters who shifted things one way or the other.

The numbers start on Janury 10th at the left hand side. What was going on at that time? HRC had won the New Hampshire primary big time about a week before, and there was the start of an anti-HRC vote drive in Michigan. You can see she goes up for three days and Obama drops sharply, losing the last of his bounce from Iowa. It would be interesting to see the full poll results from January 1. My own guess is that Obama may have polled ahead of HRC immediately after Iowa, and probably was at 39-40% - approximately where he is now. Edwards is unaffected.

The NH buzz peaks on the 13th, which is about when the first accusations of race baiting were being thrown at Hillary and as Barry's side makes hay over the CWU endorsement. There is a decline for a few days and then the Nevada debate on the 15th. What is interesting to note is that Edwards goes up during this period of time almost the same amount as HRC declines. Also to remember is that there are a bunch of undecideds swirling around because Biden, Dodd and Richardson have left the race, so there is substantial volatility as their supporters decide where else to go.

HRC begins upwards again within two days of the Nevada debate (following the first news cycle after the debates) and peaks the day of the Nevada caucuses. Obama has a sharp drop, probably due to the abruptly nasty turn his campaign takes and word that his CWU endorsement may not translate to actual support. Edwards also declines the day of and the day after the caucuses.

Then begins the long slog towards South Carolina, with the media, the blogosphere and Obama's campaign screaming at every turn that the Clintons are racists. The low point in the polls for Hillary is the day of the Florida primary itself, with Obama riding a huge surge of primary success and slobberingly sycophantic media promotion. Edwards sees an upturn just before South Carolina, but has it peak just before Florida, the same time that HRC begins to go up again.

The day after Florida, HRC sees a bump up in the poll, plus Edwards leaves the campaign. The next day, she goes up again and the LA debate is held. The day after the debate, she bounces up by 4 points. To me this marks three things:
  1. The end of the South Carolina bump for Obama. I also think this marks the end of any gains he can get from pushing the racism line. People aren't buying it anymore, not after Florida. It's also becoming clear that there isn't anything like that coming from the Clinton campaign, Harold Meyerson's pathetic and shameful lies not withstanding.
  2. The departure of Edwards is making a large group of voters have to make a final choice between HRC and Obama, and they are going to HRC. Edwards' numbers stopped at 4. The next day, Hillary picks up four. No, I don't think there was a simple transfer. Part of that were Obama fence-sitters swayed by the debate, but I'm willing to bet the bulk were people who had been Edwards supporters or leaners.
  3. But something else had to happen to catalyze that movement, and I think it was the LA debate. She came across better than Barry did, and (with Edwards out), people could unequivocally ask "Why not both of these candidates?" When asked (as it was all over the blogosphere and I suspect all over TV viewing America) in the context of the very strong, very enjoyable presentations by both candidates, there was the recognition that you put your strongest person in the top slot. And that's HRC.

Now, study the graph carefully. People tend to look at the top two lines, Hillary and Obama, and draw foolish conclusions that somehow a gain in support for him is a direct loss of support for her and vice-versa. It's not completely wrong, but it's wrong in terms of causality. Look at Edwards' lines and compare to HRC. Notice that when he is increasing support, hers decreases, though the two are not perfectly correlated.

It is a reasonably well known phenomenon in political science that voters do not generally jump directly from one candidate to another if they have had any investment in the first candidate. You usually see them move to an "undecided" or "leaning" position first, then they see or hear something about the second candidate that makes it OK to commit, such as a great deabate performance or hearing that someone said something inflammatory. Loss of support for one candidate shows up a few cycles before transfer of that support to another. Since people don't all make up their minds at the same moment, you see patterns that indicate trends which are always lagging signifiers of final opinion. (Hint, this is why Hillary in NH and Obama in SC exceeded the declared preference in the polls just before the respective votes. The polls didn't catch the final movement from the uncommitted bucket.)

Though blogospheric conventional wisdom holds as an article of faith (it must, since no facts support this claim) that the patterns of support for the candidates are H---E-O or else H---O-E, meaning that there is a bigger distance between Hillary and the voters next choice (be it Obama or Edwards) than between Obama and Edwards themselves, voting patterns actually reinforce that, controlling for class (and isn't that ironic!), the patterns is actually H-E---O, or even E-H---O. Thus, looking at the Gallup trend lines, when Edwards loses support, it tends to go to Hillary and vice-versa. Their defecting supporters who move all the way to another candidate tend to move to each other (H=>E or E=>H), and the amount of switching between them is probably higher than the switching between either of them and Obama. The one demographic group where this is probably not true, where Hillary and Obama are likely to exchange support more than H-E or O-E is with Black voters (regardless of class).

Thus, HRC's long decline after SC (which was no worse than her sharp decline after New Hampshire, btw - everything is relative) should be see as a combination of defections: Weak support leaving for undecided, weak support leaving for Edwards, weak support leaving for Obama, and weak Edwards support leaving for Obama. This allows Edwards to maintain position even if he loses support to Obama because he recovers the losses because of Clinton defections. HRC is back up to where she was before the race-baiting tactics, Obama has probably collected as many defections from the HRC camp as he is going to get before Super Tuesday, and what remains is to see how the Edwards support will go.

The blogosperic screamers need to understand that they are a very unique slice of the American demographic and are disproportionately Obama supporters because they hate HRC. In this, they resemble the MSM more than any ordinary voter demographic. (C'mon, do you think Barry would get the time of day from the likes of Tweety if he wasn't competitive against Hillary?) I remember the slams and shit thrown at Barry by the very people now on their knees worshipping him. The anti-Obama spew at FireDogLake, for example, after his speech about religion and the public sphere was so vitriolic and, yes, racist, that I stopped going to the site.

Barry is rightly worried that Hillary will strongly appeal to former Edwards supporters who are very policy oriented and want to see those kinds of policies implemented, especially healthcare (Prime example - Paul Krugman. Not an HRC fan, but big on progressive policy), and are not so inflamed about Iraq as long as the candidate has a decent plan for leaving. It is not a mistake that the current round of attacks are focused on healthcare, though it is offensive that they are doing a pure Karl Rove FUD maneuver.

I could be wrong, but I think that people who supported Edwards are more likely to move to Hillary than to Obama, and I think this is what the Gallup poll, taken in the context of other polls, is telling us.



ccp said...

Are Zogby polls reliable? Right now most of the MSM is reporting Zogby which has Obama in the lead in CA and both "virtually tied" in Missouri and New Jersey. I find this a bit hard to believe because I thought HRC had a comfortable lead in CA and NJ. Are absentee ballots in CA included in the Zogby results?

Daryl said...

There's another number I'm keeping my eye on and that is whether Obama this time around can get more than 25% of the white democratic vote. If he cannot then no matter how much money he has left, he's finished.

FreakyBeaky said...

I think it is safe to ignore Zogby. On the other hand, HRC has never had a comfortable lead in CA, i.e., she had a lead that was much smaller than the combined undecideds and might-change-minds. She hasn't significantly lost support, but she hasn't gained. A spot-check of today's CA polls says that HRC and Obama are separated by less than 2x the margin of error (i.e., the margin of error of the difference between them) in all. It's a tossup. Nobody has any idea what is going to happen here. 'cept maybe that it is likely to be close.