Normal Vote Analysis: Sensitivity to Change Over Time

Abstract

In the last two decades normal vote analysis has become an accepted analytic tool of students of electoral behavior. But the parameters of the normal vote model are context dependent and may be affected by political change. Much evidence suggests that the politics of the 1960s and 1970s differ substantially from those of the Eisenhower era on which the parameters estimated by Philip Converse were based, implying a need for their reassessment. This report presents newly calculated turnout and partisanship parameters for the normal vote model based on data from the Center for Political Studies six national election surveys from 1962 to 1972. The theoretical significance of the new parameters and the methodological and statistical consequences of applying the old rather than the new parameters to normal vote analysis of recent elections are discussed. Among other results, the research reveals a systematic overestimation of the long-term component, and the reciprocal underestimation of the importance of the short-term component, when normal vote analysis of recent election data is performed with the fifties parameters rather than with the new parameters.