Measuring Constituency Opinion: The 1978 U. S. Congressional Election Survey

This paper discusses the major potential research advantages of using congressional districts as the primary sampling unit for election studies, as was done in the United States in 1978. This permits use of the district (rather than the respondent) as the unit of analysis. The paper describes how correlations between constituency opinion and roll-call behavior can be corrected for the reliability of the sample constituency characteristic. Data from the 1978 survey are analyzed, using this technique, and they show that roll-call behavior is quite responsive to constituency opinion.