Economic models of voting typically assume the transitivity of individ
ual-level preferences. Other conditions, such as single-peakedness or
dichotomization, are also sometimes postulated. Despite the ubiquity o
f these assumptions, there is a paucity of empirical tests using real-
world elections with mass electorates. Using CPS data, I address these
issues in the context of U.S. presidential elections from 1972 to 198
4. It is maintained that (a) the traditional assumption of transitivit
y is empirically plausible, even with a large number of alternatives,
and while preferences are (b) predominantly not dichotomous, they do (
c) tend to be single-peaked along a traditional ideological dimension.