40
This raises the question of whether the coefficients for the voting impact of
working welfare and national security in 1980, which are statistically significant, are also
seriously distorted by the absence of a measure for foreign engagement. The question is
not inconsequential, because 1980 is the first of the three successive elections in which
national security jumps up hugely as a voting influence, leading social welfare during this
period. There is no direct way to address this question, of course, since the relevant
measure is just absent. Yet it is possible within a comprehensive issue model to rerun the
analysis for the years immediately before and after 1980 without their measure of foreign
engagement, in order to see whether the working welfare and national security
coefficients change in those years under these altered conditions.
Table 4
Table 4 does this, and the answer is, straightforwardly, no. For 1976, the other
coefficients are little influenced by the absence of a measure for foreign engagement.
Moreover, 1976 is a particularly important year for this test, because foreign engagement
does achieve statistical significance. Without it, working welfare appears to pick up a
little of the variance in the presidential vote previously explained by foreign engagement,
but nothing is otherwise changed; the role of national security is not being distorted by
the presence or absence of a second foreign-policy measure. For 1984, where foreign
engagement does not begin with a statistically significant impact, differences are truly
inconsequential. Apparently, so far as one can judge, the absence of a measure of foreign
engagement in 1980 does not produce a misleading analysis in the other domains that are
measured.
Convention