major source of the party's voting strength, the extensiveness of its organizational network, its success in winning control of the government in various areas, and its participation in national level politics. The one judged most important for our concept is its participation in national level politics. We grant the party's national orientation if it participates in national level politics, but we also give weight to the success of its participation and to the distribution of its support among the major governmental subdivisions.
The definition of "major governmental subdivisions" depends on the country in which the party operates. For federal systems of government, the major subdivisions are likely to be the states that comprise the federal system. However, the relevant subdivisions may also consist of more than one such state. Moreover, the notion of major governmental subdivisions could be applicable to unitary states. The test here is whether the literature identifies "major governmental subdivisions" within a country.
Operational Definition. "National participation" is coded in two steps. Initially, we evaluate each party according to the following rough scale:
Where the data permit a relatively elaborate analysis, we compute the mean deviation of the regional distribution of party support from the regional distribution of population, relying on the following expression:
Where N is number of regions, xi is the proportion of the party's national support that derives from region i, and pi is the proportion of the national population that resides in region i.
Support (xi) is most easily measured by means of electoral statistics where they are available. Where they are not, as well as in the case of countries in which elections are not meaningful indicators of party support (as, for example, in most single-party states), proportionate party membership figures are used instead. The census data necessary to compute p (percentage of the population contained in each region) were often difficult to obtain, and we occasionally substituted data on proportionate voter turnout -- that is, the percentage of the national vote (or, in the single-party states, party membership) cast by region.
When the coder was in doubt as to which of the scores 4-6 to assign to a party and the appropriate data were available, he computed D and coded the party according to the following scale:
Coding Results. Because the bias in the ICPP definition worked against the inclusion of regional parties, scores for this variable tended to cluster at the high end of the scale. In particular, there were difficulties in distinguishing between codes 5 and 6, which required judging whether a party's success was "highly variable" or "rather uniform" across regions. About halfway through the coding, therefore, we introduced the calculation of the D statistic to provide a more precise quantitative measure for distinguishing between the higher codes on the scale. In retrospect, perhaps the D statistic itself should have constituted the basic operationalization of this variable, for it is a more sensitive measure of regional variation. Nevertheless, the original operationalization was retained, with the computation of D only serving to select which of the higher codes should apply. The results of the coding are contained in Tables 4.4a and 4.4b for BV204.
More than 100 of the parties in the study were characterized as "primarily national" in their orientation, taking codes 5 and 6 together. Reflecting the pattern shown for each of the preceding three indicators of governmental status, the BV204 scores were higher in the second half of the period than the first. Of course, both distributions are highly skewed. If the operationalization had been more sensitive to variations at the "primarily national" end of the continuum, the skewness would have been reduced. The lower mean adequacy confidence scores reveal that this variable was more difficult to code than "governmental leadership" and "cabinet