Partisan turnout bias in midterm legislative elections

Lower salience elections present greater opportunities for representational bias at the polls than do elections with higher levels of political interest. We hypothesize that turnout bias is most likely to occur during Midterm congressional elections in which there are clear short-term forces that exploit the low turnout setting. The effects of these forces are more likely to be observable among registered nonvoters than citizens who are not registered to vote because registrants have access to the polls and are likely to have voted in previous presidential contests. Using midterm National Election Study data from 1918 to 1998, we find that registered nonvoters ate frequently more Democratic than midterm election voters, particularly in 1994 and 1998. The historic 1994 congressional election seat losses for Democrats may be partially explained by the finding that voters going to the polls were clearly more conservative than registered nonvoters.