Variability in Electoral Behavior: Longitudinal Perspectives from Causal Modeling

Voting theory has been vulnerable to limitations in synchronic survey data analysis. These limitations include lack of longitudinal perspective, overgeneralization from single, unique elections, and the conjoining of new methodologies with synchronic data. To promote comparative longitudinal research using survey data, causal models of the 1956, 1964, and 1972 presidential elections are developed. Methods used include a seven-variable causal model and cohort analysis. The 1956 model supports hypotheses which emphasize electoral stability. Party identification is the dominant influence on the vote, with issue preferences found to be weakly related to party identification or the vote. The 1964 and 1972 models show considerably greater voter responsiveness to political issues. Actual political and behavioral variability over time, rather than the electorate's inherent limitations, or methodological inadequacies, are seen as responsible for previous divergent findings.