¹ For a brief description of the study see PS, Winter (1980), pp. 20–1.Google Scholar A complete detailed description is available from the Inter-university Consortium for Political and Social Research, in the introduction of the codebook for the early release of the Integrated Data File.

² The basic concepts involved in defining alternative criteria for relative ‘importance’ – and in using those criteria to assess competing explanations of the 1980 election – have somewhat complex intellectual roots. In particular, the disaggregation of a national electoral result into distinct components, based on the multiplication of regression slopes by the average scores for each predictor was introduced by Donald Stokes, in research that appeared first in Stokes, Donald E., Campbell, Angus, and Miller, Warren E., ‘Some Dynamic Elements of Contests for the Presidency’, American Political Science Review, LX (1966), 19–28.CrossRefGoogle Scholar The explicit contrast between a correlational criterion for a variable's ‘importance’ and the same variable's contribution to an (aggregate) electoral result appeared in Miller and Levitin's analysis of the Center for Political Studies' 1972 and 1976 election studies (Miller, Warren E., and Levitin, Teresa E., ‘Ideological Interpretations of Presidential Elections’, American Political Science Review, LXXIII (1979), 751–71Google Scholar) as a distinction between ‘theoretical’ and ‘political’ importance. During the same period, Shanks also developed a more comprehensive set of definitions and alternative statistical criteria for relative importance (Shanks, J. Merrill, ‘The Importance of Importance’, Survey Research Center, Berkeley, Working Paper, 1981Google Scholar). In addition to presenting a more general discussion of the interpretative issues involved in any assessment of quantitative ‘importance’, this later paper develops the specific computational algorithms used in calculating the ‘types’ of importance used in our current 1980 analyses.

³ Wattenberg, Martin P. and Miller, Arthur H., ‘Party Coalitions in Decay: Processes of Regional Partisan Change’, in Lipset, Seymour Martin, ed., Party Coalitions in the Eighties (San Francisco: Institute for Contemporary Studies, 1981).Google Scholar

⁴ For the most authoritative discussions of prior changes in party identification see Converse, Philip E., The Dynamics of Party Support: Cohort-Analyzing Party Identification (Beverly Hills, California: Sage Publications, 1976).Google Scholar Also, see Converse, Philip E., and Markus, Gregory B., ‘Plus Ça Change …: The New CPS Election Study Panel’, American Political Science Review, LXXIII (1979), 32–49.CrossRefGoogle Scholar

⁵ Brody, Richard A. and Page, Benjamin I., ‘Comment: The Assessment of Policy Voting’, American Political Science Review, LXVI (1972), 450–8.CrossRefGoogle Scholar

⁶ The principal architect of the measures of candidate attributes was Professor Donald R. Kinder, Center for Political Studies, The University of Michigan, who was a member of the 1980 NES planning committee.

⁷ Converse, Philip E., ‘The Concept of the Normal Vote’, in Campbell, Angus, Converse, Philip E., Miller, Warren E. and Stokes, Donald E., eds, Elections and the Political Order (New York: Wiley, 1966).Google Scholar

⁸ Jackson, John E. and McGee, William H. III, ‘Election Reporting and Voter Turnout’Google Scholar, paper prepared for and supported by a grant from the John and Mary R. Markle Foundation, with support from ABC News (October 1981). See also Wolfinger, Raymond and Linquiti, Peter, ‘Tuning In and Turning Out’, Public Opinion (02/03 1981), pp. 56–60.Google Scholar

⁹ It should be noted that thirty-four cases were omitted from our analysis of voters because they were not asked the question about ideology. The original number of voters to be analysed was 877. The thirty-four cases that were dropped consisted of reported voters whose ideological positions were not ascertained. Among the remaining 843 voters there were other instances, on virtually every variable, in which a few voters' positions were not ascertained. For all such instances we arbitrarily scored such voters as zero (‘0’) to guard against interpretations valid only for the subset of voters (the better educated, more politically involved, more articulate) who provided full information on all variables.

¹⁰ From a statistical point of view, it may help to describe the ‘incremental predictive power’ coefficient as the standardized regression coefficient which would be produced by a conventional regression programme if each independent variable were replaced by the residual from a regression in which it is the dependent variable, and all variables before it are predictors. To assist readers who are more familiar with conventional standardized regression coefficients, Table 33 also includes such a coefficient for each original variable in the prediction equation for the final vote. Note that these two kinds of ‘predictive’ coefficients can differ markedly, for the ‘incremental predictive power’ coefficient includes both direct and indirect effects, but counts only the explanatory power of that portion of each variable not explained by its predecessors – while the conventional standardized coefficient represents only direct effects, but includes the predictive power that should be attributed to the indirect influence of prior variables.

¹¹ Miller, and Levitin, , ‘Ideological Interpretations of Presidential Elections’.Google Scholar