was more ambiguous than desired,, the empirical results for this cluster of variables support the initial conceptualization quite strongly.

Centralization of Power. We view "centralization of power" in terms of the location and distribution of effective decision-making authority within the party. A centralized party is one that features the concentration of effective decision-making authority in the national party organs, with a premium placed on a smaller number of individuals participating in the decision. We sought to tap the locus of power within a party with eight basic variables:

9.01	Nationalization of Structure. The more hierarchically structured the party, with the national organs having direct influence on local units, the more nationalized the structure and thus the greater the centralization of power.
9.02	Selecting the National Leader. The fewer the number of party members involved in selecting the leader and the more "closed" the process of selection, the more centralized the party.
9.03	Selecting Parliamentary Candidates. The greater the role of the national party organs in selecting candidates for the legislature or parliament, the greater the centralization of power.
9.04	Allocating Funds. The greater the role of the national party organs in formulating policy and the smaller the number of individuals who participate at the national level, the greater the centralization of power.
9.05	Formulating Policy. The greater the role of the national party organs in formulating policy and the smaller the number of individuals who participate at the national level, the greater the centralization of power.
9.06	Controlling Communications. The greater the control of the national party organs over party communications media and the more influential those media, the greater the centralization of power.
9.07	Administering Discipline. The more differentiated the party machinery for administering discipline and the more this machinery is controlled by the party executive committee, the more centralized the party.
9.08	Leadership Concentration. The fewer the number of individuals with the power to bind the party to courses of action, the greater the centralization of power.

Once again, these eight variables were expected to intercorrelate positively with one another, and once again the expectation was confirmed. The mean correlation was .38, and the first factor accounted for 46 percent of the variance among the variables, all of which loaded above .60 on the first unrotated factor. The "centralization of power" scale produced by these eight variables had a reliability of .83. The validity of our measurement can also be tested by checking the data concerning another expectation of the conceptual framework. Centralization of power was proposed as a dimension of internal organization that was distinct from degree of organization. This distinction allows for parties to be highly organized but decentralized and parties to be highly centralized but not highly organized. The expectation inherent in these concepts was strikingly confirmed by the absence of correlation (r = .03) between the centralization of power and degree of organization scales. Obviously, these two highly reliable scales are measuring two distinct aspects of party organization, as proposed.

Even greater confidence in the validity of the measurement of both scales can be gained by pushing further into the measurement of the centralization of power. Although all eight proposed variables loaded higher on the first unrotated factor than on any other, rotating the factors uncovered subclusters among the variables. Two sets of three variables each defined two different factors, with the remaining two variables divided almost equally between them. One set of variables appeared to signal the presence of a "structural" or "organizational" aspect of party power. These variables were 9.01, "nationalization of structure"; 9.06, "controlling communications"; and 9.07, "administering discipline." The other set suggested the existence of a "personalized" type of party power. The high loading variables were 9.02, "selecting the national leader"; 9.05, "formulating policy"; and 9.08, "leadership concentration." For each of these three variables, the highest score was attached to the exercise of power in the hands of a single individual, while, for each of the other variables, the highest score was linked with the exercise of power by the national committee. The reliability of a three-item "structural power" scale was .77; that of a three-item "personal power" scale was .75.

This distinction among items within the centralization of power scale was not anticipated in the original conceptual framework, and thus the determination of these subclusters is the second example of an a posteriori rather than a priori distinction which cannot count directly in the validation of the conceptual framework. But, if these inductively derived scales were to stimulate new expectations which were subsequently confirmed, then some claim for validation can be entered. Recall that the full centralization of power scale was completely uncorrelated with the degree of organization scale. Of course, the structural power and the personal power subscales from the centralization of power scale are related to some degree, but their correlation is only .25. If the structural power scale is really a measure of struc-

go to page 154