OK

Until recently, the conventional wisdom had been that by remaining in the race, John Edwards was taking more votes away from Barack Obama than Hillary Clinton, and thereby helping Hillary Clinton.  However, over the course of the past week or so, we have begun to see some shifts in this narrative.  The excellent Chris Bowers has gone so far as to say:

Obama's only chance in this campaign is if Edwards stays in the race through February 5th.

And an article in today's New York Times suggests:

But some political strategists say Mr. Edwards also has another compelling reason to stay in, at least in South Carolina. He could end up sharing the white vote with Mrs. Clinton, thus helping Mr. Obama, whom Mr. Edwards has signaled he favors.

This shift in narrative is hardly isolated to these sources.  I have also seen it everywhere from supporters of all candidates here at Daily Kos, to the talking heads at MSNBC.  However, the shift in narrative does not appear to have been prompted by any shift in the underlying evidence.  On the contrary, the evidence still seems to suggest that a majority of John Edwards supporters would break for Barack Obama.

Before we go any further, let me get a couple of things out of the way.

  1. I am not trying to sway anybody's vote.  People have every right to vote for who they want to vote for, and I'm fine with that.  Candidates have the right to stay in the race for as long as they want to stay in the race, and I'm fine with that.  In this diary, I'm merely trying to analyze the numbers.  
  1. There are multiple definitions of "help".  I'm concerned only with where Edwards' current votes would wind up.  For example, even if John Edwards is taking somewhat more votes from Obama than from Hillary Clinton, he could wind up helping Obama if he determined at some later date to pledge his delegates to Obama.  Also, by remaining in the race, Edwards could affect the media narrative in a way that his favorable to Obama.  These sorts of things are harder to determine, and so I'm concerned with the simpler question of where Edwards votes are most likely to end up.

We would have a pretty good idea of whom Edwards voters' second choice would be if we simply asked them.  Unfortunately, most pollsters have not done this.  Quite a few pollsters do ask about second choices -- but they haven't been breaking the second choices of the Edwards supporters out from the second choices of the Clinton and Obama supporters.  So mostly what we're getting are the second choices of the 80% of the electorate that is now voting for Obama or Clinton -- which tells us next to nothing about the group that we're interested in, Edwards voters.  (FWIW, the average of second choice numbers in four recent national polls that did ask about them are Obama 30, Clinton 28, Edwards 26).

However, there is one exception.  A CNN poll conducted January 9th and 10th asked people who they'd vote for if the nomination were between only Clinton and Obama (this is actually not the most recent CNN poll, but it is post-New Hampshire).  Those results looked as follows:

	    All Candidates    Only Clinton/Obama      Net
Clinton 	 49		     53 	      (+4)
Obama		 36		     44 	      (+8)
Edwards 	 12		     --
Kucinich	  1		     --

That is, if voters were not able to vote for John Edwards (or Dennis Kucinich), Barack Obama would pick up 8 percent, and Hillary Clinton would pick up 4 percent.  That suggests that about 2/3 of Edwards voters would break to Obama.

While that may tell us something, we need to be cautious, because the sample size is very small.  There were 443 registered Democrats interviewed for this survey; of those about 12% had Edwards as their first choice.  So that's a subsample of only about 55 voters -- not very many.

There is one (and only one) source I am aware of that contains a decently-sized sample of Edwards votes.  This is the CNN Exit Poll from New Hampshire, which interviewed 1,955 respondents, of which about 17% (around 330 voters) voted for Edwards.  This is a respectable sample size, producing a margin of error of +/- 5.4 percentage points; not fantastic, but consistent with what we're used to seeing in primary polls.

Unfortunately, this survey did not directly ask Edwards supporters about their second choice candidate.  But it did ask Edwards supporters about their opinions of the other candidates, from which we can make some strong inferences.  From doing a little algebra, we find that Edwards supporters had the following opinions of Hillary Clinton and Barack Obama.

Edwards New Hampshire voters -- opinion of Hillary Clinton (CNN exit poll)

11.9% strongly favorable
51.5% somewhat favorable
21.4% somewhat unfavorable
15.2% strongly unfavorable

Edwards New Hampshire voters -- opinion of Barack Obama (CNN exit poll)

34.7% strongly favorable
51.7% somewhat favorable
10.1% somewhat unfavorable
3.5% strongly unfavorable

These are fairly large differences.  86% of Edwards' New Hampshire voters have a favorable opinion of Barack Obama, as opposed to 63% of Hillary Clinton.  Moreover, 35% of Edwards supporters have a strongly favorable view of Barack Obama, as opposed to just 12% for Clinton, and 15% have a strongly unfavorable view of Hillary Clinton, as opposed to just 3-4% for Obama.

If we assume that the Edwards voters' opinions of Barack Obama and Hillary Clinton are independent from one another -- this may or may not be a good assumption, but it's the only one we can make given the data we have -- we can cross-tabulate their preferences as follows:

Note that I've color-coded this data to attempt to estimate the Edwards' supporters second choices.  For example, if a voter has a strongly favorable opinion of Obama and a somewhat unfavorable opinion of Clinton, I assume that's a safe Obama vote.  On the other hand, if a voter has a strongly favorable view of Obama and a somewhat favorable view of Clinton, that situation is more ambiguous, and so I call this a "leans Obama" vote.  And naturally, if the voter has the same opinion of Hillary Clinton and Barack Obama, I assume she is undecided.

Let's assume that a candidate gets 100% of their safe supporters, and 75% of their leaners.  If we do that, we get the following:

That is, about 63% of Edwards supporters would break to Obama, and 37% to Clinton.  Interestingly, this estimate is pretty consistent with the data from the CNN national poll above, which suggested that 2/3 of Edwards supporters would break to Obama.  Although none of this data is definitive, the best available evidence suggests that somewhere between 3 in 5 and 2 in 3 Edwards supporters would break to Obama, at least on the national level.

What about South Carolina?

John Edwards is not going to discontinue his campaign before South Carolina, so this is largely an academic question.  However, one of the arguments that Bowers and the New York Times have advanced is that Edwards helps Obama in South Carolina by splitting the white vote with Hillary Clinton.  This seems like a rather undetermined hypothesis.  Let's look at the current breakdowns among white voters in four current South Carolina and national polls that categorize support based on racial classifications:

In South Carolina, where Edwards has 27 percent of the white vote, Clinton leads Obama by 25 points among white voters.  But nationally, where Edwards has the support of 17 percent of white voters, Obama trails Hillary by "only" 17 points among this subgroup.  Rather than suggesting that Edwards is particularly helpful to Obama among white voters in South Carolina, the data would seem to suggest the opposite.

Of course, there may be other things that are affecting these numbers.  Hillary, via her ties to Arkansas, can claim some sort of Southern heritage.  And there seems to be an implicit assumption that there is more racism in the South.  I have no doubt that some Southerners are racist -- but I don't know that the proportion is higher in the South than it is elsewhere in the country (I see plenty of racism here in Chicago), and I don't know how much it affects the subsegment of the population that intends to vote in the Democratic primary.

At the very least, I suspect that Edwards is about neutral to Obama in South Carolina -- and probably hurts him more then helps him elsewhere in the race, at least in terms of where his supporters would end up.  Certainly, it is not as simple as there being a "change vote" or an "anyone but Hillary" vote -- voter preferences are complicated, and a substantial number of Edwards supporters would wind up voting for Hillary if that's what it came to (including many here at Daily Kos).  A 65-35 or 60-40 margin is not overwhelming by any means.  However, the evidence suggests that Obama would close his national margin with Hillary Clinton by 2-4 points if Edwards decided to exit the race.

Originally posted to poblano on Mon Jan 21, 2008 at 04:29 PM PST.

EMAIL TO A FRIEND X
Your Email has been sent.