The commentariat's topic du jour is this AP story
which cites a study conducted in conjunction with Yahoo!, Knowledge
Networks and Stanford University and which reports that "Statistical
models derived from the poll suggest that Obama's support would be as
much as 6 percentage points higher if there were no white racial
prejudice." Here are some thoughts I have on the matter:
1.
It is irresponsible to cite this study without fully disclosing its
methods or making it subject to peer review, particularly as it appears
to use a rather convoluted soup of statistical and inferential
techniques.
2. The study appears to be one of all adults, rather
than registered or likely voters. Expressions of racial prejudice have
a strong inverse correlation with education levels, and so do turnout
rates. Therefore, even if it is true that Barack Obama's race puts him
at something like a 6-point disadvantage with the population as a
whole, the margin is probably more like 4-5 points among likely voters.
3. A related and unresolved question is how many persons will vote for Barack Obama because
he is black. Such behavior would probably be more implicit and harder
to ascertain than voting against a candidate because of racial
prejudice. For instance, Obama's biography is significantly more
compelling because he is black (actually, bi-racial), and his change
message is probably somewhat easier to sell because he looks different
than other (e.g. white) politicians. If he were white, in other words,
Barack Obama would not be Barack Obama.
Moreover, there may be some people who explicitly vote for Obama
because they think it will advance a goal of racial equality, present a
different face to the world, and so forth. In the absence of sufficient
detail on the study's methodology, it is impossible to know whether
these things have been accounted for.
4. One should be very
careful not to confuse a study like this with the Bradley Effect. Of
course some people are racist, and will vote against Obama because he
is black -- I have met some of them. But the Bradley Effect concerns
something different -- whether such people are likely to lie about
their behavior to pollsters. There is simply no empirical evidence
that the Bradley Effect exists any longer. It did not exist in the
primaries, and it did not exist in the 2006 Senate race in Tennessee,
which was perhaps the most racially-tinged contest of the past decade
(in fact, Harold Ford slightly outperformed the late polls).
--Nate Silver