Social Forces, University of North Carolina Press

Social Forces, University of North Carolina Press Power and Social Structure in Community Elites Author(s): Roger V. Gould Source: Social Forces, Vol. 68, No. 2 (Dec., 1989), pp. 531-552 Published by: University of North Carolina Press Stable URL: http://www.jstor.org/stable/2579259. Accessed: 19/02/2011 17:00 Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at. http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at. http://www.jstor.org/action/showpublisher?publishercode=uncpress.. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. Social Forces, University of North Carolina Press and University of North Carolina Press are collaborating with JSTOR to digitize, preserve and extend access to Social Forces. http://www.jstor.org

Power and Social Structure in Community Elites* ROGER V. GOULD, Harvard University Abstract Although there has been considerable theoretical interest the relationship between pouver and position in a social structure, few researchers havexplored this issuempirically. This paper develops a quantitative measure of an individual's structural capacity for "brokerage" in a social network to test hypotheses abouthe effect of structural position on power in community elites. Data on two community elites are used to show that (1) position in a social network has an effect on perceived influence independent of other influence resources, and (2) power derived from structure and power derived from resources undermine rather than enhanceach other. Moreover, this relationship between power and structural position is dependent on the degree of factionalization and opposition in th elite. Following a series of empirical studies (e.g., Galaskiewicz 1979; Laumann & Pappi 1976) which revealed a connection between position in a social structure and various forms of power or perceived influence, a number of theorists have attempted to explicate the exact nature of this connection. In particular, sociologists have devoted considerable attention to the process by which structural position in exchange networks can become a source of power. Cook et al. (1983) conducted experiments which demonstrated that strategically located individuals negatively connected networks are able to exploithose who are dependent on them for the exchange of resources. In a similar vein, Marsden (1982, 1983) has developed sophisticated mathematical models which predicthe emergence of price-making and brokerage behavior in networks with restricted access. *This material is based on work supported under a National Science Foundation Graduate Fellowship. I thank Peter S. Bearman, Roberto M. Fernandez, Neil Fligstein, Harrison C. White, and two anonymous referees for their helpful comments and suggestions. In addition, I would like to thank Edward 0. Laumann for permission to use his data, and Ronald L. Breiger for providing it. Direct correspondence to the author, Department of Sociology, William James Hall, Harvard University, Cambridge, MA 02138. X The University of North Carolina Press Social Forces, December 1989, 68(2):531-552 531

532 / Social Forces Volume 68:2, December 1989 Finally, in a series of task-oriented (rather than exchange-oriented) experiments, Freeman, Roeder & Mulholland (1980) found that individuals with a high level of control over communication in a group were perceived as relatively powerful by other members of the group. In a variety of ways, real or perceived power is seen to be tied to centrality in social networks. Despite the refinements formal and theoretical treatments of the subject, however, few attempts have been made to reapply these concepts in real-world settings (exceptions include Boje & Whetten 1981; Mintz & Schwartz 1985). In particular, ideas involving the emergence of power or influence in social structures characterized by restricted access have received little attention in the field of community elite research. This may be in part due to the difficulty of adapting pristine mathematical formalizations to typically messy empirical situations. Nevertheless, community elite networks are a natural setting in which to explore these issues: as small, relatively closed systems of actors who exhibit dense social relations, elites can provide-and often have provided-significant insights into the complex connection between structure and influence. Therefore, this paper will use data on two community elites collected by Laumann and Pappi (1973, 1976) and Laumann, Marsden, and Galaskiewicz (1977) to bridge the gap between the theories and the realities of power. The section that follows will develop a theoretical conception of brokerage which is applicable to community elites with oppositional structures. Building on the graph-theoretic notion of betweenness, I develop a measure of the potential for brokerage between rival subgroups in a social network and show that this structural property is an important determinant of influence rank in an oppositional community elite even when other influence resources are controlled. A comparison of two community settingshows, moreover, that the importance of brokerage is diminished when consensus within the elite increases. I conclude by calling for a recognition of the fundamental theoretical and empirical distinction between power based on resources and power based on structural position. Power and Brokerage THEORETICAL PERSPECTIVE A common theme in the theoretical and experimental work noted above is the idea that some form of centrality is a crucial factor in determining which members of a social network will emerge as the most powerful or influential. Building on earlier small-group research (Bavelas 1950; Leavitt 1951), Freeman (1977, 1979) showed that "betweenness," or the ability to control communication (by altering or witholding information, for instance) is an important component of any intuitive notion of centrality.

Power & Social Structure / 533 Thus individuals who lie on communication paths between other individuals have the potential to exert power over them. While the work on exchange networks cited above is not strictly concerned with betweenness in Freeman'sense, it nevertheless involves closely related concepts. Most notably, Marsden (1982) developed a mathematical model of exchange in restricted-access networks in which "brokerage behavior" was found to be a source of power. That is, individuals who can act as intermediaries exchanges between otherwise unconnected actors will be able to charge commissions, converting their structural position into a resource of monetary value. This research has made significant steps toward an understanding of the role of brokerage in exchange and communication systems. However, the near ubiquity of brokerage and intermediaries a wide variety of historical and ethnographic literature suggests that an exclusively exchange-based view of brokerage roles is too limiting; a richer theoretical conception of brokerage is needed. Accordingly, therefore, it is essential to establish a definition of brokerage behavior that is operational and at the same time sufficiently general to embrace instances of the phenomenon in disparat empirical settings. Following Marsden (1982, p. 202), brokerage will be understood as a process "by which intermediary actors facilitate transactions between other actors lacking access to or trust in one another." To avoid an unnecessarily restrictive interpretation of this definition, the term "transaction" should be taken to mean any form of instrumental relation, not simply those involving direct exchange of resources. Another important feature of such a definition, noted by Gould and Fernandez (1989), is that it does not require such mediation to be rewarded by a commission; this is only one way in which brokerage behavior can be compensated, if it is compensated at all. This permits us to count as brokerage behavior the mediation functions of Sicilian mafiosi (Blok 1974), South American caciques (Friedrich 1968), and other types of power brokers whose reward for acting as intermediaries is untied to any specific transaction. Political anthropologists have documented behavior corresponding to brokerage as defined here for over three decades. In an early discussion of the issue, Wolf (1956) noted that the rise of centralized governments in predominantly rural nations had created both problems and opportunities for "middlemen" or "brokers" mediating relations between local and national interests. Friedrich (1968) developed some of these observations further in his study of Mexican caciques, whose legitimacy as authority figures stemmed from their ability to act as intermediaries between their communities and the national political arena. Similarly, Adams (1970) argued that "power brokers" were prevalent in Guatemalan politics prior to the 1944 54 revolutionary period, and declined in importance after barriers to social mobility between local and nationalevels broke down. Lest it be thought

534 / Social Forces Volume 68:2, December 1989 that the phenomenon is peculiar to Latin America, it is worth noting that similar patterns have been observed in India (Tinker 1968), Africa (Fallers 1955), and Western Europe (Tilly 1964). In probably the best-known such study, Blok (1974) attributed the long-standing power of Sicily's mafiosi to their monopoly of linkages between Sicilian peasant communities and the central Italian government. A common factor in all these studies has been the observation that brokers or intermediaries are most likely to be influential in situations where the levels or groups they mediate between are separated or segmented by barriers of culture, language, distance or mistrust. Indeed, Wolf (1956, pp. 1075-76) notes that "power brokers" may not only exploit but actually encourage such barriers, "maintaining the tensions which provide the dynamic of their actions." This suggests that, for brokerage to be a source of power, the restricted access included in the definition must take the form of a more or less institutionalized "gap" (Weingrod 1968) or "synapse"(wolf 1956) in the network of transaction relations. Transient or easily bridged junctures can undermine the power of the broker by providing principals with alternate means of conducting transactions. This pattern is clearly seen in Adams's (1970) study of Guatemala, cited above: brokers between local communities and the central government were largely unable to maintain their influential position after 1954, as increasing mobility made it easier for rural groups to communicate directly with national elites. The emphasis that many scholars have placed on vertical segmentation between community and national levels should not be allowed to obscure the fact that brokers play an important role in bridging horizontal gaps as well. For instance, Evans-Pritchard's (1940) study of the Nuer tribe in the Sudan observed that one of the activities of the leopard-skin chief was to mediate between kinship groups feuding over a murder or the theft of cattle. Because Nuer kinship structure was of the segmentary type, and because alliances and conflicts were always reckoned in terms of kinship, these feuds generally arose between clans or lineages of equal size and stature, rather than hierarchically. It was the role of the leopard-skin chief to negotiate a truce between the feuding clans without either party appearing to have won. Consequently, the mediation role of the leopardskin chief both depended on, and helped to preserve, the horizontal nature of cleavages in Nuer society. Similar mediation patterns have been observed among the Ila of Zambia (Tuden 1966) and the Swat Pathans in Pakistan (Barth 1959). In all of these and in other cases, a common feature of the broker or mediator role is that it is generally situated outside the institutionalized power structure, or at least outside the power relations involved in the conffict. Thus the mushimo spokesman who attempts to settle conflicts between Ila clans is a spiritual, apoliticaleader, and is not perceived to have a direct

Power & Social Structure / 535 interest the outcome of the dispute. Likewise, the Saints or mullahs who generally mediate between rival factions in the Swat valley do not participate in the political competition associated with local chiefs and their "men's houses." Nuer leopard-skin chiefs possessed no actual authority either: Barth notes that the threats and imprecations they uttereduring negotiations had a distinctively ritual character, and thathe primary motivation to resolve the conffict was the desire on the part of the disputing parties to save face without having to fight. In general, then, brokers who bridge horizontal cleavages appear to distance themselves from the power struggle itself; to become a contender would compromise the appearance of impartiality which is essential to mediation. In the context of community elite politics, these observations have a number of theoretical implications. First, they suggesthat a conceptualization of elite politics as a series of dyadic or mediated exchanges may obscure important aspects of the community decision-making process. While some interaction among community influentials may in fact follow Coleman's (1973) model of exchange of control over events (also the point of departure for and a pivotal element of Marsden's model), this may only be one part of community decision-making. The anthropological literature on brokerage shows that mediators frequently do not engage in exchange themselves, but merely establish the situation in which exchanges or settlement of disputes can take place; this implies that negotiation over the possibility and terms of exchange is an analytically distinct process. Brokers should be powerful in community elites, therefore, because of their ability to exercise control over this negotiation process. Stable cleavages arise in elites when positions on community issues are correlated with each other (Laumann& Pappi 1976; Laumann & Marsden 1979), creating the "synapses" which make brokerage not only possible but also necessary for the resolution of conflict. Without the intervention of individuals who maintain relations with both sides of a dispute, negotiation and compromise-and consequently exchange-are impossible. Thus actors who by virtue of their position in community elite networks are able to facilitate negotiation between members of rival groups will be perceived as important influential even when they make no overt attempto "win" on an issue. On the other hand, when conflicts do not consistently pit the same groups against each other, stable factions do not emerge (Laumann& Marsden 1979), making the broker less important to the decision-making process. In such settings, where enduringaps or "synapses" are not present, actors who occupy brokerage positions will be less influential. Insofar as the brokerage role enables an actor to determine what issues are discussed and resolved, the power accorded to brokers corresponds to the kind of 'agenda-setting" power discussed by Bachrach and Baratz (1962) and by Crenson (1971). A crucial difference, however, is that

536 / Social Forces Volume 68:2, December 1989 brokers may not simultaneously influence the agenda and openly pursue their own goals, since to do so would be to undermine their status as impartial mediators; this pattern was observed in the ethnographic literature cited above. There is therefore an inherent tension between the "'structural" power of a broker and power in the conventional sense of the "capacity to mobilize resources for the accomplishment of intended effects" (Walton 1968). Even though brokers may be able to determine the outcome of issue conflicthrough their power over the negotiation process, they will not be able to supplementhis power with traditional influence resources. Conversely, elite members who pursue their goals by mobilizing influence resources will not be able to act as mediators and consequently will not derive power from mediation even if they occupy potential brokerage positions. These empirical predictions can be summarized in the following three hypotheses: I. Members of community elites who occupy brokerage positions between rival factions are perceived as powerful, independent of other influence resources they possess. II. The effect of brokerage position on influence is greater in elites characterized by stable factional structures than in elites that are not divided by long-term cleavages. III. Occupants of brokerage positions are inhibited in their ability to derive power from other influence resources, since the use of such resources to achieve goals would render them untrustworthy as mediators. For these hypotheses to be applicable, one further condition must hold: the degree of polarization or factionalization within the elite must not be so great that pragmatic negotiation cannot occur. Alternative forms of conflict resolution, such as public mobilization or appeal to other sources of power, cannot (according to the theory outlined here) be used to translate structural position into influence. According to Laumann and Pappi (1976) and Laumann, Marsden, and Galaskiewicz (1977), both the West German city of Altneustadt and the American community of Towertown meet this requirement: while Altneustadt exhibited a polarized structure, and the study of Towertown revealed a community elite with much more diffuse and shifting political cleavages, both elite groups were frequently able to resolve conflicts through pragmatic negotiation. MEASURING BROKERAGE Existing measures of centrality in social networks based on the concepts of degree, closeness and betweenness (see Freeman 1979) are not appropri-

Power & Social Structure / 537 ate for the kind of analysis required to test the hypotheses outlined above. While degree and closeness are related to intuitive notions of "good connections," often though to be an important source of community influence, neither addresses the issue of control over communication which is crucial to the idea of structural power. Betweenness, on the other hand, is directly tied to this issue; nevertheless, it will not be useful in this study for two reasons. First, such measures do not differentiate between communication among allies and communication among opponents; thus an individual who lies on many paths between friends but no paths between enemies-the critical factor, I argue, in determining structural power-will receive a high betweennesscore despite his or her complete inability to facilitate negotiation. Second, betweenness measures take into account paths of any length (up to N- 1, where N is the number of nodes in the network). It is difficult to see how paths of more than two or three steps could possibly be of use to community influentials attempting to resolve conflicts: in any situation where conflict arises, it is quite probable that the interested parties should be relatively close to each other in path-distance terms. If social structure has any relationship at all with community decision-making, it should certainly manifest itself at the level of two- and three-stepaths before it appears at the rather unmanageable level of four steps and beyond. The measure developed for this study avoids these problems by ignoring all paths between pairs of individuals in the same network subgroup as well as all paths of length greater than two. Thus an individual's raw score on this measure, which might be called "capacity for brokerage" or, more concretely, "inter-clique betweenness," is simply the total number of two-step directed paths between members of rival groups that he or she lies on. Note that in a network with no clusters or subgroups, every path counts; that is, an individual'score takes into account all the twostep paths on which he or she lies, regardless of their endpoints. On the other hand, in a network where everyone belongs to the same group (a monolithic or "ruling elite" structure), every individual's score will be zero. One refinement is added to the measure. In cases where there is more than one two-ste path between two points, the incremento each intermediary's score is lip, where p is the total number of two-ste paths between the pair of endpoints. The logic behind this procedure is comparable to that of partial betweennes scores: the value l/p corresponds to the probability that communication will pass through the point in question (see Freeman 1979). For example, if j lies on the only two-ste path between i and k, the incremento js score resulting from this path is 1. If, on the other hand, there are three two-step paths between i and k, the incremento j's score is only 1/3 or.33. More formally, an actor j's interclique brokerage bj can be defined as:

538 / Social Forces Volume 68:2, December 1989 O b _ N b(ik) mi (p(ik) *0, bl= w4p(ik) i# j -k) where b(ik) equals 1 if actor i sends a tie to j, j sends a tie to k, and i does not send a tie directly to k, and 0 otherwise; p(ik) is the number of twostep paths between i and k; and m(ik) equals 1 if i and k are in different subgroups, and zero otherwise. (For a detailed discussion of similar measures of brokerage, see Gould & Fernandez 1988.) The use of the measure is illustrated in Figure 1, which contains a hypothetical network of eight actors linked by directed (nonsymmetric) social relations. The network is arbitrarily partitioned into two non-overlapping subgroups containing four actors each; Figure 1 depicts one set of actors as squares on the left and the other set as circles on the right. Brokerage scores for each actor based on this partition are included in the figure. Clearly, actors A and B play the most prominent brokerage roles linking these two groups. Actor A's score is based on five brokered relations: H-A-C, F-A-C, G-A-C, H-A-B, and F-A-B. The firstwo paths count for one point each, since no other actor links H or F to C; on the other hand, the other three paths count for only.5 each, since G shares in linking H and F to B, and B shares in linking G to C. (Note that, even though A lies on a two-ste path between G and B, the path G-A-B does not count toward A's brokerage capacity because G is tied directly to B.) Similarly, B's score contains two full points from the paths G-B-D and G-B-E, and three half-points from G-B-C, A-B-D and A-B-E. C and G, finally, each have a brokerage capacity of 1 based on two half-points: A-C-D and A-C-E for C, and H-G-B and F-G-B for G. The other actors in the network have brokerage scores of 0 because, even though they lie on a variety of twostep paths, none of these paths crosses the boundary between the two subgroups. Thus the measure identifies two actors, A and B, who provide most of the linkage between two largely unconnected "cliques." The other two actors with non-zero scores have some capacity to link members of the two cliques, but their brokerage roles are not as pronounced. Data and Methods TWO COMMUNITY ELITE STUDIES Studies of community elites by Laumann and his colleagues (Laumann& Pappi 1973, 1976; Laumann, Marsden & Galaskiewicz 1977) have avoided many of the pitfalls of earlier elite studies (Dahl 1962; Hunter 1953) by employing a combination of the reputational, issues-oriented, and positional approaches to the identification of community influentials Alt-

Power & Social Structure / 539 FIGURE 1. Hypothetical Network with Two Subgroups (brokerage scores in parentheses) (O) (3.5) (1 ) H A (O)~~~~~~~~~~~~~~0 neustadt and Towertown (respectively a West German and an American city). After identifying a group of 51 influentials in Altneustadt, Laumann and Pappi (1973) successfully interviewed 45 of them and collected data on three relational networks: each informant was asked to name three persons on the list with whom he or she had informal social ties; three with whom he or shg had business or professional contacts; and three with whom the informant discussed community affairs. These choices could, of course, overlap. Analysis of the network of social ties by means of a graph-theoretic procedure revealed the cleavage of the elite into two major "cliques" as well as a smaller group (the "county businessmen") and eight individuals in a residual category. The term "clique" was used to denote a strong component of the directed graph representing informal social ties; by definition, then, two individuals are part of the same clique if they are mutually reachable by some path. The split into two major subgroups corresponded roughly to a division along political lines: one group was associated with the generally conservative and-in Altneustadt-more influential Christian Democratic Union (CDU), while the other consisted of the more left-wing and less powerful Social Democratic Party (SPD). Multidimensional scaling of the political views of these groups tended to confirm the impression of an "oppositional" structure in the Altneustadt elite (Laumann& Pappi 1976, p. 122). Finally, there is clear evidence that the groups are differentiated in terms of religion and socioeconomic background. The SPD group contains Protestant scientists and clerics, and is generally seen to represent what Laumann and Pappi call the Neubuirger, the Protestant middle- and upper-

540 / Social Forces Volume 68:2, December 1989 middle-class citizens of Altneustadt who arrived in the city with the construction of the large Natural Science Research Center in the 1950s. The CDU group, on the other hand, is largely composed of Catholic businessmen and local officials, including the city's Biirgermeister or mayor; they are, naturally, more closely aligned with the Catholic working-class and middle-class Altbiirger population of the city, characterized by the authors as native-born residents with largely traditional and conservative values. Breiger's (1979) reanalysis of the data using blockmodeling techniques reproduced this "oppositional" structure, providing further support for the conclusion that Laumann and Pappi's method of mapping Altneustadt's community elite network was not subject to the same bias as earlier sociometric methods. Furthermore, the correspondence of qualitative perceptions of the community elite to the more objective results generated by the data indicated that these forms of network analysis were fruitful for understanding community decision-making. Laumann and Pappi (1976) further divide the CDU group into two "sub-cliques." One of these is headed by the mayor, Herr Koenig, and based in the city; the other is led by Koenig's chief personal rival, Herr Jung, and is oriented more toward county affairs. Since this division corresponds to a difference in focus rather than conflicting interests, however, these two subgroups are treated here as belonging to the same clique. In a replication of the first study, Laumann, Marsden, and Galaskiewicz (1977) examined two mid-western American cities, called "Towertown" and "River City," which were of comparable size and class composition to Altneustadt. In contras to the German city, these communities were found to have much less clearly defined oppositional structures; that is, elite members who allied with one another on one community issue were not necessarily allied on other issues, so that lines of confrontation in local politics tended to cross-cut each other. A later study which introduced the concept of the "collective actor" (Laumann& Marsden 1979) confirmed these impressions of Towertown and Altneustadt by showing that the latter corresponded to a "unidimensional" or two-faction model of elite opposition, while the former adhered closely to a "multidimensional cleavage" model. Although Laumann, Marsden, and Galaskiewicz did not use the same graph-theoretic technique in their study of the American cities as they did in Altneustadt, a strong component analysis of the informal social ties network in Towertown yields an informative result. Instead of reflecting a two-faction social structure, which in the German case showed that informal social contacts mirrored a cleavage of political alliances into two opposing groups, the Towertown social ties network contains only one strong component which incorporates over two-thirds of the 77 elite members. Moreover, this strong component includes members of both the Democratic and Republican parties as well as independents. This result

Power & Social Structure / 541 tends to confirm Laumann, Marsden, and Galaskiewicz's (1977) and Laumann and Marsden's (1979) conclusion that Towertown does not exhibit a stable factional structure; it also shows that the opposition that does exist in the political arena is not sufficiently sharp or confrontational to translate into cleavages in the social realm. In the context of the theoretical framework outlined above, then, Altneustadt is a setting in which political brokerage is expected to contribute significantly to perceived influence; conversely, since Towertown does not exhibit an institutionalized factional structure, we should not expect brokerage to be as important as a determinant of influence. Results To compute scores for inter-clique betweenness or "brokerage potential," each member of the two elite groups was assigned a subgroup value. The individuals in Altieustadt's elite identified by Laumann and Pappi's strong component analysis as belonging to the CDU group were given one value, and those identified as the SPD group were assigned another. The eight individuals not collapsed by Laumann and Pappi into any group and those in the three-person county business "cique" were assigned memberships based on party affiliation; the two who had no such affiliation and for whom there were no other empirical grounds for assigning a group value were assigned to a residual group. Paths to and from these individuals therefore counted as significant negotiation; however, alternative assignments of this residual category had negligibl effects on the results.' In Towertown, the same criterion for assigning individuals to subgroups could not be used because, as noted above, the American city did not exhibit a stable factional structureither in political activities or in the network of informal ties. To render the Towertown analysis comparable to the analysis of Altneustadt, elite members were assigned to subgroups on the basis of party affiliation: members of the Republican party constituted one group, Democrats a second, and independents were placed in a third group regardless of whether they leaned toward the Democratic or Republican party, or toward neither. Again, alternative assignments of borderline cases did not influence the results. Brokerage potential scores were calculated for the Altneustadt elite using the community affairs network for the 45 influentials who were interviewed in the study. In Towertown, network data exist both for this general community affairs discussion network and for discussion of five specific community issues: the construction of a private health services center; the question of whether the city or the county should administer a new airport; the imposition of a curfew; the location of a new post office;

542 / Social Forces Volume 68:2, December 1989 and the closing of an experimental school (for more detailsee Laumann & Marsden 1979; Laumann, Marsden & Galaskiewicz 1977). Since the issue networks representedata concerning communication over actual events, these seemed preferable to the more general community affairs network; thus a proxy for the community affairs network was generated by taking the union of the five issue-specific networks. In other words, an elite actor was treated as sending a tie to another actor if he mentioned that actor in response to at least one of the five issue questions. Brokerage scores were then calculated for this composite network using the party membership partition just discussed.2 (Brokerage capacities for all actors with non-zero brokerage scores are reported for both communities in the Appendix; the pseudonyms correspond to those used in the original studies to permit comparison of results.) To test the hypothesis that high structural capacity for brokerage would contribute to power, these scores were used as an independent variable in OLS regressions predicting influence rank in each community elite. This rank was based on the number of nominations as "very influential" each elite membereceived from other members (see Laumann & Pappi 1976; Laumann, Marsden & Galaskiewicz 1977).3 Because standard sociological accounts of community power place a heavy emphasis on resources as a source of influence, a variable reflecting the total number of "influence resources" an individual possessed was included as a control (see Laumann & Pappi 1976). Following the original studies, each individual was coded as possessing a particular type of resource if three or more respondents named that elite member as possessing the resource. The resource variable is simply the number of distinct resources the individual was found to possess. (In Altneustadt, attributed resources ranged from 0 to 7, with a mean of 1.49 and a standar deviation of 1.63; in Towertown, the index ranged from 0 to 8 with a mean of 2.73 and a standard deviation of 2.4). Zero-order correlations among these variables appear in Table 1; regression results appear in Table 2.4 Resources are, clearly, a good predictor of influence rank in both communities:. the higher one's number of resources, the more influential one tends to be (since a low rank indicates high influence, coefficients to variables with positiveffects have negative signs). This result is consistent with most past research, including that of Laumann and Pappi (1976). When score for brokerage potential is added to the equation for Altneustadt, both variables continue to be strong predictors, although each is weaker than when it is alone in the equation; resources by themselves explain about 60 percent of the variance in influence rank, while brokerage potential alone accounts for about 30 percent. We can conclude with considerable confidence that, in Altneustadt, a structural position which permits an individual to act (or not to act) as intermediary between competing factions contributes to perceived influence independently of

Power & Social Structure / 543 TABLE 1. Zero-order Correlations for Influence Rank, Resources, and Potential for Brokerage (Altneustadt above diagonal, Towertown below) Influence Resources Brokerage Influence -.78.55 Resources.88 -.49 Brokerage.30.24 TABLE 2. Regression Coefficients with Influence as the Dependent Variable (Standard Errors in Parentheses) Variable Model I Model II Model III A. Altneustadt Resources -6.859** -5.874** -7.781** (0.852) (0.935) (1.090) Brokerage -1.637* -4.492** (0.754) (1.214) Resources X 1.121** Brokerage (0.391) Constant 36.696 36.705 39.143 R2.601.642.701 B. Towertown Resources -8.351** -8.145** -8.694** (0.502) (0.512) (0.522) Brokerage -0.054-0.371** (0-033) (0.112) Resources X 0.071** Brokerage (0.024) Constant 62.593 62.844 64.678 R2.787.794.816 *p <.05 **<.01 Note: Influence rank ranges in Altneustadt from 1 (highest influence) to 50 (lowest influence), and in Towertown from 1 to 78. Thus negative coefficients signify positiveffects on influence.

544 / Social Forces Volume 68:2, December 1989 the effect of resources. This constitutestrong support for Hypothesis I above. In contrast, the brokerage variable is only marginally significant as a predictor of perceived influence in Towertown, providing support for Hypothesis II, which posits that brokerage is a source of power only in communities with clearly defined oppositional structures. If we view brokerage potential as a specialized form of centrality, then the result for Altneustadt demonstrates a much strongerelationship between structural centrality and influence than that revealed in past elite research. Laumann and Pappi (1976) find a correlation of only.30 between their measure of centrality in the community affairs network and influence rank (compared with.55 for the brokerage measure), and the highest correlation across all three kinds of ties is.40 for the business/professional network. There are two probable reasons for this difference. First, their measure is computed as the distance from the centroid of a smallest-space depiction of the network (the lower the distance, the higher the centrality), and thus involves an extra level of abstractionot required in the present analysis-that is, their measure is taken from an interpretation of the data whereas the measure used here is computed directly from the data. Second, the smallest-space analysis Laumann and Pappi use is itself based on path distances between points in the network. A measure of centrality grounded on the concept of control over negotiation is intuitively more relevanto the study of influence than is a measure grounded on the concept of closeness; and in this case, empirical research confirms intuition. The final model presented here (Table 2) includes an interaction term for resources and brokerage score to test Hypothesis III, which predicts that these two sources of influence will tend to interfere with each other. In Altneustadt, the coefficient for this interaction term is significant beyond the.01 level using the standard t-test; the partial F-test was also significant beyond the.01 level (F= 8.23, d.f. = 1, 41). Note that the interaction coefficient has a positive sign (negativ effect) in contrasto the other two coefficients. Thus the interaction suppresses the effects of resources and brokerage potential: the magnitude for each increases when the interaction is included in the model. What this suggests is that possession of many resources undermines the contribution of structural position to influence, while occupancy of a position with high potential for brokerage blunts the effect of resources. Indeed, inspection of the regression coefficients yields a remarkable result: for very high values of one variable, the effect of the other variable reverse sign. For example, if an individual has a brokerage score of eight (the highest score in the sample is 9.5), the total effect of resources will be 1.121(8)-7.781=1.187. This positive slope on the number of resources means that, for this brokerage score, having more resources actually reduces one's influence. The implication of this result is clear: actors occupying brokerage positions cannot supplement the power they derive from this role with power derived from traditional

Power & Social Structure / 545 influence resource such as money, official authority, control over jobs or land. Indeed, actors in brokerage positions may actually lose power when they are perceived to have such resources. In Towertown, inclusion of the third term reveals that brokerage does indeed have a significant effect on perceived influence, and that this effect was suppressed by the omission of the interaction. Thus a similar process is in fact observed in Towertown as well: both brokerage and resources have a positive effect on perceived influence, but they tend to dampen each other. However, consistent with Hypothesis II, the effect of brokerage is significantly weaker in Towertown than in Altneustadt: comparing the main effects, t= -2.971, p<.01; comparing the interaction terms, t=2.357, p<.05. (The t-test for comparing regression coefficients across the two cities is equivalento pooling the samples and including in the model interactions between a dummy variable for city and each other term.) On the other hand, the coefficients for the resources variable are not significantly different in the two cities (t =.684). This suggests that, while the contribution of brokerage to influence is substantially greater in the community with a factionalized structure, the contribution of resources is not dependent on the degree of opposition in the elite. Moreover, the extent to which resources and brokerage interfere with each other in producing influence is greater in the city where opposition between elite subgroups is more salient.5 This result is exactly contrary to what one would predict from Marsden's (1982, 1983) models of brokerage and price-making behavior in restricted access networks. In his formulation, the ability to control other individuals' exchange opportunities should enhance the value of one's resources. For instance, Herr Schiitz, the city civil administrator, should be able to use his advantageoustructural position (his score on brokerage is 9.5, the highest in the elite group) to augment his capacity to influence community decisions through the exchange of his resources: his "commission" for acting as intermediary would be the increased willingness of his clients to accept his terms of trade in bargaining over other community issues. The opposite signs of the main effects and the interaction effect do not support the hypothesis based on Marsden's model of restricted-access exchange networks. Instead, the result show that elite members who derive their power from resources may be limited in their ability to use structural position as a source of power; conversely, individuals in positions of high potential for brokerage are generally unable to marshal their resources as a base for community power. This conclusion confirms Hypothesis III: to the extent that elite members attempt to influence the outcomes of community decision-making through the direct exchange of resources, they will be incapable of acting as brokers and therefore will not derive power from structural position. To test whether a simpler form of structural centrality might have

546 / Social Forces Volume 68:2, December 1989 the same effect on influence as the measure used here, influence was also regressed on degree (the total number of choices received) in place of the brokerage score. As Table 3 indicates, the effect of degree in Altneustadt is not significant at even the.1 level when resources are controlled. This result is largely due to the high collinearity (r=.83) between resources and relative degree; by contrast, capacity for brokerage is correlated.49 with number of resources. In Towertown, on the other hand, degree is a significant predictor of influence rank; thus the special status of brokerage disappears when the elite group does not exhibit a stable factional structure, and other kinds of centrality may be more important in predicting perceived influence. Finally, the importance of the concept of brokerage across subgroups, which lies at the heart of this paper, was subjected to scrutiny by examining the same measure without dividing the network into "cliques." In other words, the same brokerage measure was computed (using the community affairs network in Altneustadt and the union of the issue networks in Towertown), except that this time all two-step paths were counted: each person's score now included paths between members of the same clique as well as paths between members of rival cliques. Once again, the data provide strong support for the theory outlined above: the "null" brokerage measure, in which subgroup effects are ignored, cannot predict influence in Altneustadt when resources are controlled. In contrast, the effect of the null measure in Towertown is nearly identical to the effect of the measure using the party-affiliation partition, suggesting once again that opposition between elite subgroups is not an important feature of community politics in the American city. Summary and Conclusion This paper has argued for an analytical distinction between two kinds of power or influence. One type, thoroughly treated in the literature on community leadership, is based on resources which can be mobilized to bring about particular outcomes on community issues; the other type is the product of a position in the elite social structure which allows its occupant to facilitate negotiation between rival factions. While both sources of power can be used to determine outcomes-in the case of brokerage power, primarily through control over the issue agenda-they cannot be used together, since the mobilization of influence resources erodes the image of impartiality which is crucial to the brokerage role. In this study, the idea of brokerage was associated with a mathematical measure of "inter-lique betweenness" which locates individuals with high potential for brokerage in social networks. This measure differs fundamentally from other measures of structural centrality in that it focuses on paths of length

Power & Social Structure / 547 TABLE 3. Regression of Influence on Two Alternate Measures of Structural Centrality, Controlling for Resources (Standard Errors in Parentheses) Variable Model I Model II A. Altneustadt Resources -6.545** -7.635** Degree (1.539) -0.160 (1.469) (0.645) Brokerage without 0.197 subgroups (0.302) Constant 36.530 37.223 R2.602.605 B. Towertown Resources -6.289** -8.108** (0.653) (0.526) Degree -0.818** (0.188) Brokerage without -0.039 subgroups (0.023) Constant 63.992 62.844 R2.830.795 p <.05 p <.01 Note: Influence rank ranges in Altneustadt from 1 (highest influence) to 50 (lowest influence), and in Towertowni from 1 to 78. Tlhus negative coefficients signify positive effects on influence. two-that is, on triadic brokerage-and takes into account divisions of the network into empirically meaningful subgroups. The results reported here lend strong supporto the hypotheses set out in the theoretical section. Brokerage potential was positively related to influence an elite group characterized by bargaining between rival factions; this relationship persisted even when the model controlled for level of resources. The same relationship was observed in a community with a less oppositional structure, except in this case the strength of the relationship was significantly diminished. This supported the hypothesis that brokerage tends to be more important in determining influence when actors can be divided into clearly defined oppositional groups.

548 / Social Forces Volume 68:2, December 1989 Moreover, the brokerage measure was the only centrality measure that had an independent effect on influence in Altneustadt; other centrality measures were too closely related to the number of resources to exhibit an independent effect. In the less factionalized community of Towertown, however, other forms of centrality not dependent on subgroup partitions appeared to be as important as inter-clique brokerage. Finally, structural position and resources were shown to undermine rather than enhance each other as sources of influence: ambiguity about goals appears to be a prerequisite to structural power. This finding militates against current theoretical work rooted in an exchange perspective, which tends to predict a positive interaction between resource-based and position-based power. This research has significant methodological implications as well. The success of an emphasis on divisions within networkshows that it may be misleading to analyze social structures under the assumption that all social ties have the same analytical status. Communication across subgroups-whether these are defined in terms of cohesion, structural equivalence, or some other criterion-may have profound effects on the relative power of individuals in social networks, while communication within such groups may be so frequent or unproblematic that its structure affords no insight into social processes. In other settings, of course, the reverse might be true: links between individuals in the same subgroup might be decisive while ties across group boundaries might be irrelevant. Finally, both kinds of tie might be important, but in dramatically different ways. This paper has succeeded in showing that the pattern of cross-group communication is a critical determinant of power in oppositional community elites; furtheresearch is needed to identify the social contexts in which other patterns of communication are significant. Notes 1. It was pointed out above that both Towertown and River City were shown to have nonfactionalized elite structures, so that neither would be expected to be a setting in which brokerage was particularly important in determining influence. While both cities could have been included in this study, only Towertown was examined because it would not be interesting to presen two negative cases alongside a single positive case. Aside from Altneustadt, there are no network data known to the author for a community elite characterized by a stable oppositional structure. 2. Ideally, one would want to test the brokerage hypothesis with respecto each of the issues separately, partitioning the elite according to their positions on the particular issue in question. This strategy would circumvent the problem of a lack of a factionalized structure in Towertown, since each issue by definition divided the elite into opposing groups. Unfortunately, the issue networks taken individually are far too sparse to generate meaningful results, since almost every elite member's brokerage score is zero. The next best solution is to union the five networks and partition the elite according to party membership on the ground that when cleavage does occur, it would most likely follow party lines. On the other hand, for purposes of comparison, it is desirable to use the same sociometric question that was used in the Altneustadt study. I therefore computed brokerage scores using the community

Power & Social Structure / 549 affairs network in Towertown as well; this yielded nearly identical results to the analysis based on the union of the five issue networks. 3. The dependent variable in these regression analyses is, strictly speaking, an ordinal rather than an interval variable. Nevertheless, it is treated here as though it were an interval variable for two reasons. First, the rank variable is based on an interval-level variable, namely the number of nominations as "very influential" each elite member received from other elite members. The original data for this variable in Altneustadt were not available. Second, because there are many categories on the rank variable (influence ranks range from 1 to 50 in Altneustadt, 1 to 78 in Towertown), least-squares estimateshould be robust. 4. Note that this is not a sample of community elite members in Altneustadt or Towertown, since the observations in each case constitute the city's entirelite group. However, because I think that the model of influence and structural position studied here can be generalized to members of other community elites, I use statistical inference to estimate the likelihood that the observed relationships among the variables hold in other similar settings. Readers who are concerned that the observations are not a random sample, and who therefore prefer to view the Altneustadt elite as a population, may disregard the standard errors reported here and trea the regression coefficients as population parameters. 5. An alternative test of this negative interaction could be performed by examining whether the effect of brokerage on influence decreases when elite members use their resources to produce desired outcomes on local issues; the theory outlined above predicted that elite members with high capacity for brokerage would be perceived as influential only when they appear to be impartial. Unfortunately, the data on Altneustadt do not provide good information on elite actors' positions on community issues: in all but one of the four issues studied, fewer than one fifth of the elite members reported that they had publicly voiced an opinion, while in the fourth issue (the establishment of a secular school) only about one fourth of the elite acknowledged that they had taken a position. Almost everyone in the elite group would therefore appear to have been impartial on most issues, leaving little room for variation on this variable. Thus data on issue positions in Altneustadt are not adequate for the kind of test described, and a comparison between Towertown and Altneustadt on this question is not possible. TABLE Al. Altneustadt Elite Members with Non-zero Brokerage Scores ID Pseudonym Brokerage Influence Rank 4 Bode 3.00 S 6 Blendermann 4.83 12 7 Bunck 1.00 28 10 Chelius 1.50 13 13 Ewig 2.00 7 15 Franke 2.33 8.5 20 Hilgers 1.00 32 25 Kiel 1.00 42 27 K Koenig 5.33 1 34 Nickel 6.00 4 39 Rudolf 3.00 19.5 43 Schiiller 4.00 16 44 Schiitz 9.50 2.5 48 Stahl 1.00 32 50 Zacharias 0.50 14

550 / Social Forces Volume 68:2, December 1989 TABLE A2. Towertown Elite Members with Non-zero Brokerage Scores ID Pseudonym Brokerage Influence Rank 1 Aldworth 6.00 43.0 3 Bell 15.92 10.0 5 Biewald 2.98 29.5 7 Cantwell 25.96 26.0 8 Cater 53.94 2.0 9 Collins 26.52 22.0 12 Crane 75.21 6.0 14 Dickens 74.61 24.0 15 Drew 0.60 19.0 16 Evans 41.81 12.5 17 Fields 3.48 67.0 18 Flagg 62.77 1.0 19 Fricke 37.77 14.0 20 Fulton 1.15 45.5 21 Gale 2.69 43.0 23 Giles 38.15 51.5 24 Gould 3.33 51.5 25 Haber 0.35 22.0 26 Heller 4.34 26.0 27 Iverson 2.05 9.0 32 McDaniel 3.06 7.0 33 Menard 6.19 38.0 35 Natkin 11.93 11.0 36 Olson 15.28 16.0 38 Pines 29.29 34.0 39 Rennels 0.58 17.0 40 C. Roland 17.88 8.0 41 Ross 6.65 34.0 42 Ryan 0.33 22.0 43 Seeley 6.24 12.5 45 Simmons 2.56 36.0 46 Solar 12.23 51.5 47 Stanton 41.90 25.0 48 Thayer 0.10 43.0 49 Tunney 0.50 67.0 51 Whitney 12.76 55.0 52 Williams 0.33 74.0 53 Wirth 43.27 37.0 56 Kassel 2.80 5.0 57 Edwards 0.59 4.0 58 Jenson 286.05 20.0 59 Kasper 12.25 18.0 60 Naughton 17.80 32.0 61 Jones 5.80 39.5 62 Magnuson 117.00 41.0 65 Brinkman 9.51 29.5 66 Cardinal 19.11 34.0 67 Jarrett 0.94 48.0 68 Snyder 0.45 58.0 73 Deering 1.11 55.0 74 Manley 0.89 74.0

Power & Social Structure / 551 References Adams, Richard N. 1970. "Brokers and Career Mobility Systems in the Structure of Complex Societies." Southuestern Journal of Anthropology 26:315-27. Bachrach, Peter, and Morton Baratz. 1962. "Two Faces of Power." American Political Science Review 56:947-52. Barth, Fredrik. 1959. Political Leadership among Swat Pathans. Athlone. Bavelas, Alex. 1950. "Communication Patterns in Task-oriented Groups." Journal of the Acoustical Society of America 22:725-30. Blok, Anton. 1974. The Mafia of a Sicilian Village. Harper & Row. Boje, David A., and David A. Whetten. 1981. "Effects of Organizational Strategies and Contextual Constraints on Centrality and Attributions of Influence in Interorganizational Networks." Administrative Science Quarterly 25:378-95. Breiger, Ronald L. 1979. "Toward an Operational Theory of Community Elite Structures." Quality and Quantity 13:21-57. Breiger, Ronald L., and Philippa Pattison. 1978. "The Joint Role Structure of Two Communities' Elites." Sociological Methods and Research 7:213-26. Coleman, James S. 1973. The Mathematics of Collective Action. Aldine. Cook, Karen S., Richard M. Emerson, Mary Gillmore, and Toshio Yamagishi. 1983. "The Distribution of Power in Exchange Networks: Theory and Experimental Results." American Journal of Sociology 89:275-305. Crenson, M. 1971. The Un-Politics of Air Pollution. Johns Hopkins University Press. Dahl, Robert A. 1961. Who Governs? Yale University Press. Evans-Pritchard, E. E. 1940. The Nuer. Oxford University Press. Fallers, Lloyd. 1955. "The Predicament of a Modem African Chief: An Instance from Uganda." American Anthropologist 57:290-305. Freeman, Linton C. 1977. "A Set of Measures of Centrality Based on Betweenness." Sociometry 40:35-41.. 1979. "Centrality in Social Networks I: Conceptual Clarification." Social Netzvorks 1:215-39. Freeman, Linton C., C. D. Roeder, and D. L. Mulholland. 1980. "Centrality in Social Networks II: Experimental Results." Social Networks 2:119-41. Friedrich, Paul. 1968. "The Legitimacy of a Cacique." Pp. 243-69 in Local-level Politics: Social and Cultural Perspectives, edited by Marc J. Swartz. Aldine. Galaskiewicz, Joseph. 1979. Exchange Networks and Community Politics. Sage. Gould, Roger V., and Roberto M. Fernandez. 1989. "Structures of Mediation: A Formal Approach to Brokerage in Transaction Networks." Forthcoming Sociological Methodology 1989, edited by Clifford C. Clogg. Hunter, Floyd. 1953. Community Power Structure. University of North Carolina Press. Laumann, Edward O., and Franz U. Pappi. 1973. "New Directions in the Study of Community Elites." American Sociological Review 38:212-30.. 1976. Networks of Collective Action. Academic Press. Laumann, Edward O., and Peter V. Marsden. 1979. 'The Analysis of Oppositional Structures in Political Elites: Identifying Collective Actors." American Sociological Review 44:713-32. Laumann, Edward O., Peter V. Marsden, and Joseph Galaskiewicz. 1977. 'Community-elite Influence Structures: Extension of a Network Approach." American Journal of Sociology 83:594-631. Leavitt, Harold J. 1951. "Some Effects of Certain Communication Patterns on Group Performance." Journal of Abnormal and Social Psychology 46:38-50. Marsden, Peter V. 1982. "Brokerage Behavior in Restricted Exchange Networks." Pp. 201-18 in Social Structure and Network Analysis, edited by Peter V. Marsden and Nan Lin. Sage. 1983. "Restricted Access in Networks and Models of Power." American Journal of Sociology 88:686-717.

552 / Social Forces Volume 68:2, December 1989 Mintz, Beth, and Michael Schwartz. 1985. The Power Structure of American Business. University of Chicago Press. Tilly, Charles. 1964. The Vendee. Wiley. Tinker, Hugh. 1968. "Local Government and Politics, and Political and Social Theory in India." Pp. 217-26 in Local-level Politics: Social and Cultural Perspectives, edited by Marc J. Swartz. Aldine. Tuden, Arthur. 1966. "Leadership and the Decision-making Process." Pp. 275-283 in Political Anthropology, edited by Marc J. Swartz, Victor W. Turner, and Arthur Tuden. Aldine. Walton, John. 1968. "Differential Patterns of Community Power Structure: An Explanation Based on Interdependence." Pp. 441-59 in Community Structure and Decision-making: Comparative Analyses, edited by Terry N. Clark. Chandler. Weingrod, Alex. 1968. "Patrons, Patronage, and Political Parties." Comparative Studies in Society and History 10:377-401. Wolf, Eric. 1956. "Aspects of Group Relations in a Complex Society: Mexico." American Anthropologist 58:1065-78.