of subgroups within the party. Moreover, comparing the party's support drawn from different groups (its social composition) with the composition of society offers an opportunity to measure social reflection. Precise conceptualizations and operationalizations of each of these measures will be discussed in turn.
"Social attraction" is defined as the extent to which the party attracts its supporters evenly from each significant subgroup within any dimension of social cleavage. For the present, "party supporters" can be interpreted as those who express a preference for a given party in answer to direct questions about their partisan feelings. In the case of the United States, for example, we can count as party supporters those who "think of themselves" as Republicans or Democrats in response to interviewers' questions about their party identifications.
Under our conceptualization of social attraction, it makes little difference what overall or mean level of support is enjoyed by a party. We are interested instead in the evenness of its support from social groups. Thus we seek to measure social attraction in terms of absolute deviations from the mean level across all groups (percentages calculated by column in Table 5.1) for a major social variable. The average amount of deviation (sum of absolute deviations divided by the number of groups) is then divided by the mean to express the deviations as a proportion of the mean. The average deviation is normed by dividing by the mean because an average deviation of 1.0 percentage points is judged to be relatively small for a mean support level of 50 percent, but relatively large for a mean support level of only 10 percent.
The value resulting from these operations is then divided by the maximum deviation that could be obtained for a specified number of groups. This maximum is achieved when any single group gives a party 100 percent of its support and the party gets no support from any other group. These several concerns are included in our formula for measuring social attraction:
where k is the number of subgroups within the cleavage dimension included in the analysis; Xj is the proportion of the jth group's support given to the party; and is the mean proportion of support for the party, calculated over all social groupings, k.
The social attraction values produced by this formula range from 0 to 1.0, with higher values meaning higher attraction. A score of 1.0 is achieved only if there is no variation in the percentages of support received by the party from the different social groups in the analysis. A score of 0 results only if a party receives all the support of one group while winning no support from any other. This formula has an operational interpretation as the proportion of evenness of support enjoyed by the party--measured against the situation of equal support from each social group.
"Social concentration" is defined as the extent to which party supporters are concentrated in specific subgroups within any dimension of social cleavage. In this concept, the focus is on the pattern of party composition, with the pattern based on the percentages of the party's strength received from each group (i.e., percentages calculated by rows in Table 5.1).
There is considerable literature on measuring concentration in economics (Singer 1968, esp. ch. 13) and international relations (Ray and Singer 1973). A complemen-