8
white male candidates when asked to vote. As an experiment, it is still questionable whether
some of the responses indicating a willingness to vote for a non-traditional candidate would
actually transition to votes in the poll. The tendency to report favorably for non-traditional
candidates may continue when respondents realize that their votes are merely for an experiment,
and not for true political office.
A third option for examining the role of race in elections is to use actual elections that
involve a minority candidate. This is typically done by gathering exit poll data or looking at
public opinion surveys conducted during the campaign (Citrin, Green and Sears 1990). The
benefit of this method is that it permits an analysis of real-world conditions, comparing how
respondents actually evaluate real candidates, and then ultimately how that results in support in
the election. The correspondence to the real world that this provides is tempered by the lack of
control over the election, and a loss of causal attribution that can be made regarding the
significance of race. A persistent problem is that respondents, especially white respondents,
typically over-compensate in reporting their support for minority candidates (and race-coded
issues), but do not actually provide support in the voting booth (Berinsky 1999, Berinsky 2002).
While this provides evidence to the persistence of self-correction among white voters, it tells us
little about what drives the evaluation process. In a real world situation, how do whites evaluate
black candidates?
A previously unrealistic prospect would be to contrast two non-white male candidates in
a large-scale election. By removing the prospect of voting for a white male candidate,
respondents (and voters) lose the option to revert to the traditional choice. Instead, they are
forced to make candidate evaluations and vote decisions that provide less ability, or motivation,
to flee from in the real world vote (though they could simply stay home). As recently as five
Convention