28
determining whether an individual is a policy winner tends to decrease the proportion of the
population who are deemed winners. In the nine domains, we find that only about 20% of the
GSS/Roper respondents overall are predicted winners. So, if in most domains African
Americans are about five percentage points less likely to be policy winners, we would expect
that about 15% of African Americans are policy winners, compared to about 20% of whites.
Finally, we test the robustness of these results to the effects of real world conditions.
As noted above, policy decisions may be responsive to the real world and citizen opinion may
be responsive to the real world in similar ways, creating a spurious correspondence between
citizen opinion and policy change. Moreover, if whites are more likely than African
Americans to be interested in and pay attention to real world events, this potentially spurious
relationship between opinion and policy will be greater for whites. So, the disparities in
influence we have found may be due to a combination of the effect of the real world and racial
differences in attention to the real world.
We take a different approach to testing for "real world" spuriousness here than we did
above. Specifically, we repeated our estimations in Table 3 for the subset of respondents who
reported not voting in the most recent presidential election.
10
The logic here is that nonvoters
tend to be less interested in and attentive to politics, and so presumably less likely to revise
their spending preferences on the basis of changes in the "real world.” If we find that even
among the subset of GSS respondents who are less interested in politics whites are more likely
to be policy winners, we will gain confidence that racial differences in attention are not
driving our results. As Table 5 shows, even among less politically interested citizens (i.e.
10
This approach excludes all of the Roper respondents because those surveys did not query about
turnout.
Convention