The Origins of Opinion of American Party Activists

I would like to thank Peter Esaiasson, John Geer, Cheryl L. Herrera and Carolyn Warner for their helpful comments on earlier drafts of this paper. Some data were made available by the Inter-university Consortium for Political and Social Research. Funding for the survey of delegates attending the 1992 conventions was provided by the National Science Foundation, grant SBR-9310507.

1 See Norris (1995) for an analysis of May's law in British party politics.

2 See Wattenberg (1990) for a discussion of the term `perceptual screen' with regard to the effects of partisan identification.

3 The 1988 and 1992 Convention Delegate Studies are the fourth and fifth surveys in a series of delegate surveys begun in 1972. The population for the studies are all delegates attending the Democratic and Republican national nominating conventions. See Appendices 1 and 2 for a list of the items used in the analysis from the two surveys.

4 As I suggested on p. 235, I modify the analysis in a couple of important ways from what typical Downsian analysis would suggest (May, 1979). First, the prediction by rational-choice models that the positions of activists will be most similar to those held by party voters as tested by a simple comparison of positions held by each strata is insufficient. Comparing the positions of activists and voters does not adequately address the underlying theory. The linchpin of this application of rational-choice theory is that the motivations of the party activists (i.e. electoral, candidate-focused) cause variability in the degree of proximity between them and voters. Given this premise, it may be misleading to only use the relative positions of activists vis-a-vis voters. Activists may misperceive the views of the voters, which may result in unintended differences between activists and voters. That is, activists may be unable to ascribe accurately the views of those partisans. Differences between the positions of activists and regular partisan voters, then, may not be due to the activists' lack of concern for matching the positions of voters but rather to inaccurate placement of regular partisans' positions on issues. Unless activists have perfect information about the spatial locations of their parties' voters, they cannot be expected to choose positions that are congruent with those of the rank-and-file, except by chance. We need, therefore, to control for that uncertainty about where voters are situated along the continuum of positions on issues. We can do that with data that provide not only the views of the activists on issues but their perceptions of what positions they believe partisans to hold on those same issues.

Second, I take into account the motivations of the activists when examining their views on politics. The adaptation of rational-choice models to the opinions held by partisans of various activity levels suggests that the motivations of the activists at those levels should matter. As I elaborate in the next section, each level of activist contains members who are motivated for more than one reason. That is, even though I look at activists from one of the higher rungs on the ladder, there are surely those who are motivated not so much by winning, but by other concerns.

5 A reference to wanting to get the party and its candidates to support particular policies was also frequently given. I associate that response with activists being interested in policies as well as candidates and winning elections. It does not, alone, indicate that issues are more important than candidates or winning, however. Among other reasons given are wanting to make business contacts, socializing with politicians, seeking careers in politics, and enjoying the fun and excitement of campaigns, but none of these reasons rivaled the frequency of the reasons listed in the tables.

6 I used the activists' perceptions of where their own partisan voters from their state were located. I also used the activists' perception of where their partisan voters across the nation were located and, in 1992, their perception of the national electorate's views. The results were similar to those presented here. In most cases the differences were larger than those shown here. In no instances were they smaller.

7 Appendix 2 summarizes the variables in the OLS equation that are used, and their accompanying labels.

8 Clearly, partisan affiliation and positions on issues explain much of the variance in activists' ideological self-designation. In the following analysis I control for the effects of partisan affiliation and issue positions, leaving, I expect, the two hypotheses as the sole explanatory variables. I do this by examining the activists from each party separately. In so doing, I control for partisan affiliation and positions on issues. Since the level of constraint found among these highly active partisans is quite high (Jennings, 1993), I can be fairly sure that the homogeneity of the party delegates will control for their positions on issues as well. I have therefore controlled for most of the variance to be found in activists' ideological self-placement. I should not be surprised, then, if the model does not perform well in term of R² values since I am only explaining what is left after partisanship and issues are removed.

9 Why the fourfold difference? I suspect that the degree of affect toward a candidate has much to do with the magnitude of the candidate's effect. To check this notion, I compared the feeling thermometer scores for the two groups of candidate supporters. As anticipated, the average feeling thermometer score for Jesse Jackson by Jackson supporters was 10 points higher than the average score for Dukakis by Dukakis supporters (93 and 83 respectively).

10 Additional analyses were also conducted in which supporters of Jack Kemp were substituted for those of Robert Dole. This change resulted in near-identical coefficients across the board. Also, bivariate regressions were run as a check on the multiplicative variables included in the models presented. Controls were made for those activists who preferred candidates and/or were motivated to win elections. In all cases, the results supported those reported here.

11 I also conducted aggregate-level analysis and obtained further support for the candidate-motivation hypothesis.