We argue that conservative districts that go Democratic for the House shoul
d be likely to choose a Republican for president, while liberal districts r
epresented by a Republican should be likely to opt for a Democrat for presi
dent. We test these and related predictions about split-ticket voting with
election data from the eight presidential elections between 1964 and 1992.
We show that ideological differences in the estimated location of the distr
ict's median voter explains a substantial component of the systematic varia
tion in patterns of split outcomes in this period across districts, but tha
t other factors (e.g., an especially popular incumbent or a particularly po
or challenger, the magnitude of presidential election victory, region-speci
fic realignment effects) also play a role.