twenty "dimensions of voter alignment," Lipset and Rokkan concentrate on social status, religion, urban-rural, and regionalism (1967, p. 527). Converse would add education to this list for its part as a "prime predictor for the whole class of dependent variables reflecting political interest, participation and mobilization" plus its "remarkable discriminating power as a status measure in predicting to variables on the other side of the watershed-ideology and party position" (1968, p. 4).
We have selected all six cultural differentiators cited here for our treatment of social attraction, concentration, and reflection. Specifically, we include (1) occupation (or social status or income in lieu of occupational data), (2) religion, (3) ethnicity/language/race, (4) region, (5) urbanization, and (6) education. Not all of these six major variables--which we hereinafter refer to as "cultural differentiators"--are present as forces for political division within each country. In particular, the religion and ethnicity / language / race differentiators are apt to be inoperative in highly homogeneous countries. Occupation as a differentiator, moreover, has less relevance for underdeveloped than developed countries. In addition to limits on the relevance or applicability of, these differentiators, the problem of missing data continually lurks in the background. Consequently, not all of our six variables apply to all party systems, and not all the parties are coded on all the differentiators, however appropriate they may be to the country in question.
If the applicability of the cultural differentiators varies from country to country, so too does the nature and number of groups within the differentiator. In the United States, for example, there are only two major racial groupings, whereas in Malaya, there are three major groups: Malay, Indian, and Chinese. In other cases, the underlying divisions are generally comparable, but the available data on party support use quite different breakdowns--for example, reporting only a crude urban-rural dichotomy in one case but detailed data by size or cities in another. Our data recording and data analysis procedures are quite flexible, and we accept whatever groupings the research literature identifies as relevant for the country and whichever appear to be supported with the best data.
In an effort to improve the quality of the data underlying our scoring of political parties on attraction, concentration, and reflection, we went outside the data available in our microfilm information base and sought original sample survey data to analyze ourselves or to have tabulated according to our specifications. Survey studies were obtained from the Roper Center, the Inter-University Consortium for Political Research, and the International Reference Service and Data Library at Berkeley, among other sources. Suitable survey data were located for about 15 of our 53 countries. For the remainder, we had to rely on tabulations reported in our information base or, in many instances, on researchers' personal judgments concerning the basis of party support.
According to practice within the ICPP project, the amount and character of the information underlying our assessments of party aggregation and articulation are reflected in the adequacy-confidence codes associated with each variable score. For these variables, however, we operationalize the AC codes differently from the procedure specified in Chapter 2. Our assignment of AC codes to scores for party attraction, concentration, and reflection follows the schema in Table 5.2.
In the ideal case, when scoring the parties in a given country on these variables, we were able to obtain two survey studies for the country--one for the first half of our time period and one for the second--with sample sizes large enough to produce adequacy-confidence codes of 9 for each party on each of the six cultural differentiators. The ideal case was almost never encountered, of course. For many of our countries, we considered it a stroke of good fortune to unearth any survey data. Often, these data did not include any item at all on party preference. While demographic variables pertaining to one or more of our cultural differentiators were present in nearly every study that possessed a party preference item, the available demographic variables and codes seldom satisfied all our needs. Moreover, sample sizes were often much smaller than necessary to make confident estimates, and sometimes the only available surveys were taken in the 1960s--after our time period.
If we had limited our scoring of social support only to those parties for which we had ideal data, we would have scored only a few parties and would have had mostly missing data for these variables. Alternatively, we tried to score as many parties on as many cultural differentiators as possible. If we had a survey for the second half of our time period but not the first, we generally used the data for the second as estimates of the first-revising our estimates if there were reason to suspect major shifts in party support between the first and second halves. If we had a survey dated two or three years past our time period, and party politics in the country were basically stable, we used the survey despite its date. If we had no survey data at all, we did not shun from hazarding guesses about party support based on statements in the literature, assigning our lowest adequacy-confidence score to any variables coded "impressionistically" in the absence of hard survey data.
Certain constraints were imposed in our impressionistic coding of party support to insure some logical consistency in our estimates. First, all the parties were coded on party support for a given cultural differentiator in relationship to one another. This was done by arraying the social groupings and the parties along the top and side of an empty matrix such as that in Table 5.1.