Posted by Andy Herrity on March 12, 1998 at 09:50:37:
"Corporate Political Groupings: Does Ideology Unify Business Political Behavior?"
Alan Neustadtl and Dan Clawson, ASR, 1988, Vol. 53 (April: 172-190).
Summary and Comment by Andy Herrity
This article is chronologically the second in a series of three analyzing a data set of donations by 243 political action committees (PACs) to 1980 Congressional candidates (Mark Mizruchi also uses part of this data set in his analysis). In all three articles, the problem the varying authors address is whether business PACs pursued pluralistic (pragmatism) or ideological (conservatism) donation patterns in 1980. Pragmatism and conservatism are operationalized the same way in each study, but the articles have slightly different perspectives and analyze the data differently. Taken together, the three are instructive regarding useful possibilities for applying both network analysis and other techniques to the same data. The summary of, and comment on, the focal article is preceded by concise summaries of the other two.
THE 1986 ARTICLE
In the first article, Clawson, Neustadtl, and Bearden find PACs divided using simple linear regression, but note in concluding comments that business avoids direct political confrontation over pragmatism (donations to incumbents) versus ideology (donations to advance the cause of business as a whole).
THE 1989 ARTICLE
In the third article, Clawson and Neustadtl use multiple regression in an effort to explain differing corporate political strategies (an explanation is important because, in the late 1970s, large corporations generally became more conservative in their political stance). The authors found more highly interlocked firms to have less conservative donation patterns, and explain this by noting that highly interlocked firms tend to pursue their political agenda in ways other than donations through PACs.
THE 1988 ARTICLE
In the second article, Neustadtl and Clawson conduct a clique analysis in an effort to find pluralism and, instead, find conservative ideology. "A clique is a group in which every member is connected to every other member, and there is no other actor that could be added who would be connected to every member of the clique" (p. 175) and, for Neustadtl and Clawson, a clique does not require face-to-face connection (however, they acknowledge that this use differs from that used by most sociologists).
They develop a corporation-by-candidate matrix where cell entries equal the dollar amount each corporation contributed to each candidate. Then they correlate the donations of each corporation with those of all other corporations to construct a corporation-by-corporation matrix of correlation coefficients. They argue that "Each element in this "similarity" matrix is a measure of the degree of political similarity between two corporations," (p. 175) and then use various levels of degree of similarity as cutoffs in the identification of cliques. The initial findings are presented in Table 1 (p. 177), revealing a large number of cliques at each cutoff level.
CLIQUE ANALYSIS
Before discussing Table 1, Neustadtl and Clawson address concerns which might threaten the internal validity of their approach. First, as mentioned earlier, they point out that they are not using the generally sociologically recognized face-to-face meaning of clique. However, they attain validity because their "procedure returns to the mathematical conception of a clique in graph theory ..." (p. 175). Second, in defense of their use of correlation as a measure of network similarity, they argue that such correlations remove the impact of the dollar size of a corporation's donations. Third, they accede that "By allowing the size of individual contributions to influence the similarity measure, we diminish the ability to find groupings containing corporations with similar strategies but different total resources" (p. 176). In other words, use of this approach might cause similar resources to look like similar strategy, but the authors offer no mitigating factors for this threat.
COMMENT 1: as I see them, the second and third threats together seem quite serious. Because large corporation PACs tend to give more money, and vice-versa, surely the correlations are of both similarity in strategy and similarity in size. If I'm correct about this confounding in the correlations, the authors' ultimate findings are actually much more murky than they claim.
Having found the multiplicity of cliques identified in Table 1, they "needed a method to combine cliques that are similar to determine the number of groupings that differ from one another" (p. 178). The results of their combination method are depicted in Table 2: far fewer cliques (which the call subgraphs) and, most importantly, one very large clique consisting of 120 PACs.
STATISTICAL CHARACTERISTICS AND SIGNIFICANCE OF THE SUBGRAPHS
Table 3 presents the results of successful tests Neustadtl and Clawson apply to each of the subgraphs regarding the internal coherence (or completeness) and lack of connectedness to other subgraphs (centripetality).
COMMENT 2: I wondered why they did not apply these tests to their initial findings in Table 1. To have done so, and found neither completeness nor centripetality, would have strengthened their argument that they did not find pluralism.
COMMENT 3: They are silent on the matter of my musing in comment 2 but, even if they performed these tests unsuccessfully, it seems to me that to look for plurality on a national level in a Congressional race would be an exercise in futility. Surely PACs are favorably predisposed (albeit not exclusively so) to give to candidates running from their "home" state or region? I can't help but wonder what would have happened if Table 1 had been the result of several geographically sorted matrices. Perhaps they chose not to do this because it would have "polluted" a network analysis with an assumption of pre-existing categories.
COMMENT 4: For me, the unfortunate consequence of the concerns raised in comments 2 and 3 is the distinct possibility that the authors do not find pluralism in this study because they do not seem to look thoroughly for pluralism. This is unfortunate, because their major finding (described in the next section, below) is very important and, despite my concerns about the confounding of dollars and strategy, I believe it is valid.
UNDERLYING BASIS OF THE SUBGRAPH
As shown in Table 4, "The clear unifying basis of the large subgraph at both the .4 and .5 levels is conservatism" (p. 180). This is the key finding in the study and, answers their title question: "Does ideology unify business political behavior?"
INTERPRETATION AND DISCUSSION
The authors begin this section with an important tribute to their method: "Network methods have revealed characteristics of the data that could not have been established on a statistical footing by other approaches."
In Tables 5 and 6, they present economic characteristics of large subgraph members and nonmembers. Table 6 displays more statistically significant effects than table 5 and demonstrates that subgraph members are of much larger size than nonmembers (a finding the authors find especially interesting). However, as mentioned in COMMENT 1, I'm concerned that this size finding may be partially an artifact of their original correlations of dollar donations. What do you think? Am I making something out of nothing?
"See" you at seminar time.