The Pro-Incumbent Bias in the 1978 and 1980 National Election Studies

This study demonstrates that there is a pronounced proincumbent response bias to the reported House vote in the 1978 and 1980 National Election Studies. The magnitude of the bias ranges from approximately 7 percentage points in 1980 to 14 percentage points in 1978. Only a small portion of the bias, less than 3 points, is attributable to the particular sample of congressional districts. We show that the major source of bias is the survey instrument itself. In the survey a number of questions before the vote question focus the attention of the respondent on the name and activities of the incumbent. Consequently, the survey instrument influences respondents who have little information and who are susceptible to prompting effects to identify the incumbent as the person for whom they voted. This means that studies using the 1978 or 1980 surveys run the risk of incorporating the bias into their findings. The nonrandom nature of the bias suggests that the impact of incumbency in determining the outcome of congressional elections has been overestimated.