Voting cycles and the structure of individual preferences

Empirical studies have shown that cyclical preferences are infrequent, but they have been less clear about why. Using thermometer ratings from nationally-representative samples of the U.S., we examine preferences for presidential candidates in order to determine what it is about them that leads to few cycles. Single-peaked preferences as usually construed (meaning that all of a set of preferences satisfy single-peakedness criteria) are, of course, rare. Yet we find a high degree of unidimensionality in the sense that for any given set of preferences, a relatively high proportion of the preference orders are consistent with single-peakedness. We also find that the highest amounts of unidimensionality often do not occur along partisan or left/right lines. Strong feelings for or against candidates, often not derived from an issue base, form the basis for the dimensionality discovered.We would like to thank Daniel Severance for programming the random model calculations and three anonymous referees, whose comments were very helpful in clarifying the presentation in this paper.