On the representativeness of voter turnout

Prominent theory research on voting uses models in which expected pivotality drives voters' turnout decisions and hence determines voting outcomes. It is recognized, however, that such work is at odds with Downs's paradox: in practice, many individuals turn out for reasons unrelated to pivotality, and their votes overwhelm the forces analyzed in pivotality-based models. Accordingly, we examine a complementary model of large-N elections at the opposite end of the spectrum, where pivotality effects vanish and turnout is driven entirely by individuals' direct costs and benefits from the act of voting itself. Under certain conditions, the level of turnout is irrelevant to representativeness and thus to voting outcomes. Under others, \anything is possible"; starting with any given distribution of preferences in the underlying population, there can arise any other distribution of preferences in the turnout set and thus any outcome within the range of the voting mechanism. Particular skews in terms of representativeness are characterized. The introduction of noise in the relationship between underlying preferences and individuals' direct costs and benefits from voting produces, in the limit, fully representative turnout. To illustrate the potential disconnect between the level of turnout (a focus of much empirical literature) and representativeness, we present a simple example in which, as noise increases, the turnout level monotonically falls yet representativeness monotonically rises