Contemporary research suggests that voters often base midterm congressional election choices on their evaluations of the president. In short, midterm congressional elections are seen as referenda on presidential popularity. The referendum notion does not, however, adequately address nonpresidential variables, which may be key determinants of the vote. Employing a longitudinal analysis of House and Senate voting in the 1970, 1974, and 1978 elections, this study considers the relative, simultaneous effects of presidential popularity, congressional incumbency, party identification, and voters' perceptions of the candidates. The results indicate that presidential popularity is a relatively weak predictor of individual congressional voting decisions, when nonpresidential variables are also considered. The impact is slightly greater in 1970 and 1974 than in 1978. There also appears to be a somewhat larger effect for Senate than House voting. Yet in all instances, the nonpresidential variables are the major predictors of the congressional vote.
Read full abstract