Voter Turnout and the National Election Studies

Though the overreporting of voter turnout in the National ElectionStudy (NES) is widely known, this article shows that the problemhas become increasingly severe. The gap between NES and officialestimates of presidential election turnout has more than doubledin a nearly linear fashion, from 11 points in 1952 to 24 pointsin 1996. This occurred because official voter turnout fell steadilyfrom 1960 onward, while NES turnout did not. In contrast, thebias in House election turnout is always smaller and has increasedonly marginally. Using simple bivariate statistics, I find thatworsening presidential turnout estimates are the result mostlyof declining response rates rather than instrumentation, questionwording changes, or other factors. As more peripheral votershave eluded interviewers in recent years, the sample becamemore saturated with self-reported voters, thus inflating reportedturnout.