Obama-Clinton and racial resentment

Sunday May 25, 2008

Filed under: politics,statistics — jackman @ 5:53 pm

I caught a snippet of Meet The Press this morning, where the Newsweek guy made a reference to a poll they commissioned (from Princeton Survey Research Associates). They looked at racial resentment as a predictor of Obama-vs-Clinton support. I thought I’d post some similar findings that Lynn Vavreck and I saw in the March wave of our Cooperative Campaign Analysis Project, a 6 wave panel study we’re running through YouGov America (Polimetrix). I presented these findings at a talk at Princeton in early May and thought I’d share here.

Bottom line: racial resentment is a big and powerful predictor of what separates Obama supporters from Clinton supporters, and indeed, is one of the few attitudinal predictors of the Obama/Clinton break (Obama and Clinton supporters are otherwise pretty much identical on a whole bunch of political issues).

First, a definition of racial resentment. We use a battery of four items that is now fairly standard in the political science literature: see the following slide from my talk.

We create an index from the four items, which has the following marginal distribution (again, click on the thumbnail). There is a cluster of respondents scoring high on racial resentment at the right end of the histogram; these are the respondents who consistently gave the most “resentful” responses on the four item battery.


Racial resentment cleaves Clinton and Obama supporters as shown in the next picture. The vertical axis shows the proportion of voters at a given level of racial resentment (horizontal axis) who support Obama over Clinton, with the analysis restricted to survey respondents who told us that that voted for either Obama or Clinton in the Democratic primaries/cacuses (or will vote for one of those two candidates).
Finally, some multivariate statistical analysis, summarized in the following table. I let a bunch of demographic predictors enter the logistic regression predicting Obama vs Clinton support, with racial resentment entering the model and generating a considerable lift in model fit. In short, racial resentment is a fairly powerful predictor of the way Democratic primary voting has been breaking.

We also have some data suggesting that about 31% of Clinton primary voters would support McCain in a Obama/McCain November match-up. I don’t put a lot of stock in that number, since its a long road from March to November, and at some point the nomination battle will be behind us, and partisans of both sides will tend to revert to type. Nonetheless, 31% is a big number. Even if the real number if half that, its still of no small political significance.

I’ve been asking people to tell me the states they think Obama will take from McCain, that Kerry and Gore couldn’t take from Bush. That is, lets talk Electoral College. The conversation usually goes “Virginia and Colorado will probably go blue this cycle”, which could well happen (and look out for a Webb VP nomination). I follow up by asking who last won the presidency without winning Ohio? To this end, also keep an eye on the possibility of a Strickland VP nomination.

Comments (11)

Brad Efron’s 70th birthday celebrations

Sunday May 18, 2008

Filed under: statistics — jackman @ 10:35 pm

I wonder if Brad wanted to be reminded of this milestone, but Brad Efron’s 70th birthday was celebrated with an afternoon of short talks by various luminaries. I caught talks by Carl Morris and Trevor Hastie.

The Morris talk was neat, a brief recap of some ideas to do with shrinkage estimators, going back to the genesis of Morris’ famous work with Efron on empirical Bayes, the Stein phenomenon and those famous baseball data. Charles Stein was in the front row (as he almost always is), and for a while it seemed we were fêting him! What was especially cute is that I’d just taught Stein’s result as a frequentist rationale for hierarchical Bayesian modeling that same morning in my Bayesian statistics class, using the baseball data.

I got a few photos here.

Comments (2)

Powered by WordPress

Bad Behavior has blocked 13600 access attempts in the last 7 days.