on party support was intended. Perhaps the presumption was that these three questions all produced much the same answers, and thus any one was an alternative (albeit imperfect) indicator of any other. Indeed, this was the assumption underlying the measurement of party support in the ICPP project, and it largely--but not completely- withstood empirical testing.

Six dimensions of social cleavage were selected for assessing the bases of party support. These were (1) socioeconomic status (usually occupation), (2) religion, (3) ethnicity (including language and race), (4) region, (5) urban-rural, and (6) education. The particular subgroupings within each of these dimensions designated for coding necessarily varied from country to country, depending on both the historical circumstances of the country and the availability of data resources for coding the parties. Thus, in a Western country the "socioeconomic" subgroups might be based on as many as eight categories of occupations, while in a Third World country it might stem from a crude "traditional elite," "urban worker," "farmer/peasant" classification. For all parties within a given country, however, the subgroupings for any given cleavage dimension were identical.

The formulas for calculating the attraction, concentration, and reflection scores operated on the values in a cross-classification of party preference by social subgroups for each dimension of cleavage (see Table 5.3). Where relevant, these formulas adjusted for differing numbers of subgroups across countries. The resulting social attraction score revealed the extent to which the party attracted supporters evenly from each significant subgroup within a given cleavage dimension. The social concentration score measured the extent to which party supporters were concentrated in specific subgroups within a cleavage dimension. The social reflection score expressed the extent to which the composition of the party's supporters accurately reflected the social composition of the population along the cleavage dimension.

Except at the extremes of certain limiting conditions, these measures are free to vary independently of one another, thus offering three different perspectives from which to evaluate the social bases of party support.

Both social attraction and social reflection can be conceived as alternative measures of party "heterogeneity," and parties high on these concepts can be regarded as socially "diverse" or "catchall" in nature. Social concentration, on the other hand, is a measure of party "homogeneity," and parties high on concentration can be regarded as socially "cohesive" (Rose and Urwin 1969). Obviously, one expects positive intercorrelations between paired measures of attraction and reflection, and both of these are expected to be negatively related to concentration when all the measures are applied over common subgroupings within a given dimension of social cleavage.

The actual intercorrelations among these measures are reported in Table 14.3 along with the reliabilities that result when the three measures are combined into scales of social diversity.

TABLE 14.3: Analysis of the Social Bases of Party Support
Cleavage Dimension	Attraction with Concentration	Attraction with Reflection	Reflection with Concentration	Scale Reliabilities
Socioeconomic	-0.38	0.87	-0.08	0.71
Religion	-0.64	0.93	-0.45	0.86
Ethnicity	-0.60	0.87	-0.33	0.83
Region	-0.88	0.97	-0.82	0.96
Urban-rural	-0.61	0.94	-0.45	0.86
Education	-0.36	0.86	-0.08	0.69

The data in Table 14.3 display the predicted relationships, in that attraction and reflection measures are positively intercorrelated and both are negatively related to concentration. The very high correlations between attraction and reflection demonstrate a close relationship between the evenness with which parties draw support from social subgroups and the representativeness of the parties. Parties that draw equal proportions of support from each subgroup tend also to represent the social composition of the society along that dimension in their membership.

But Table 14.3 makes clear that social concentration is not simply the inverse of attraction and certainly not of reflection. In every case, the correlations between attraction and concentration are higher than those between reflection and concentration. Specifically in the cases of socioeconomic status and education, there is almost no relationship between the extent to which party supporters reflect the composition of society and their tendency to be concentrated among certain subgroups. If this sounds impossible, an example should explain the discrepancy. Consider the case of a party whose members are equally divided between Catholics and Protestants in a country that is 90 percent Catholic. Because the religious composition of the party does not reflect the religious composition of the society very well, the

go to page 147