Using data on non-presidential-year elections for governor and U.S. Senators in eight southern states over the period 1922– 1990, we provide a rational-choice-inspired model of the factors that should be expected to affect the relative levels of turnout in primaries as compared to general elections. Both V.O. Key and Anthony Downs have argued that voters will be more likely to participate in the elections in which they can most expect to be decisive. V.O. Key (1949) proposed that when general elections are usually lop-sided because of one-party dominance of a state's politics the primary of the dominant party of the state should have a higher turnout than the general election. Downs argued that turnout should be higher in competitive elections. Our modelling combines these ideas. We use as our dependent variable the ratio of primary to general election turnout in each year. We posit that this ratio will increase (1) the greater the degree of within-party competition in the primary (especially that within the dominant party of a state, if there is one), and (2) the weaker the degree of between party competition in the general election. In addition to election-specific effects, we also posit long-run effects, such that the ratio for the offices of governor and U.S. Senator will be affected not merely by the degrees of competition within and between parties specific to any given election, but also by the long-run trends in party competition. This hypothesis leads us to expect that, (3) in the South, with the rise of the Republican party, the ratio of primary to general election turnout should decline over time. All of our expectations about the links between turnout and competition are strongly supported. We argue that rational choice models of turnout perform quite well when we view them in a comparative statics perspective, rather than using them to make predictions about who will and who will not vote in any given election.