Overreporting and electoral participation research

Do nonvoters who say they voted bias electoral participation research? Early studies said they do not, but Bernstein, Chadha, and Montjoy concluded that overreporters alter the effects of most independent variables. This research tests whether overreporters bias conclusions about additional untested variables and whether we obtain the same results as Bernstein and his coauthors when we test different models. Overreporters bias the coefficients of only a small proportion of independent variables in more typical models, although overreporting is more problematic than the early studies reported. Specifically, in the models this research tests, overreporters change the effects of African American race, marital status, Southern residence, and closing date on presidential election turnout and African American race, marital status, Hispanic ethnicity, occupation, and open seat races on midterm turnout. A warning about how overreporting bias affects, or may affect, their conclusions should accompany studies that include these, or other, untested variables.