22
Administration and Policy in Mental Health Vol. 26, No. 3, January 1999 TEAM LEADERSHIP: NETWORK DIFFERENCES IN WOMEN'S AND MEN'S INSTRUMENTAL AND EXPRESSIVE RELATIONS Cynthia Webster, Oscar Grusky, Deborah Podus, and Alexander Young ABSTRACT: Recent studies have emphasised differences in leadership styles between women and men. Women have an "interactive" leadership style while men are more "direc- tive" and "authoritative." Social network analysis is used to examine differences in eight mental health case management teams, half formally supervised by women and half by men. The techniques used are graphical displays and measures of centrality. Results show male leaders as the most central team member for both instrumental and expressive rela- tions. Female leaders, however, do not adhere to a single leadership style. Team centralisa- tion also differs with gender composition of teams influencing leadership differences. Interdisciplinary teams have become a prevalent form for accomplishing work in many areas such as business and health care. Teams are attractive because they are seen as an efficient way to divide the labor among a number of individual specialists who offer multiple skills and perspectives. The sharing of information and knowledge concerning specific issues not only reduces the risk associated with having only one person responsible for a client, project, or investment, but also potentially results in more Cynthia M. Webster, Ph.D., is with the Department of Anthropology and Sociology, University of Queensland, Brisbane, Australia. Oscar O. Grusky, Ph.D., is Professor in the Department of Sociology; Deborah Podus, Ph.D., is with the Drug Abuse Research Center, Department of Psychiatry & Bio- behavioral Sciences; Alexander Young, Ph.D., is also with the Department of Psychiatry & Bio- behavioral Sciences, all at the University of California, Los Angeles. This research was supported in part by Grant T32 MH-14583 (Oscar Grusky, Program Director) from the National Institute of Mental Health. The authors are grateful to the county mental health director, the head of evaluation research, other county officials, and the team staff members, for their participation and assistance. Address for correspondence: Cynthia M. Webster, Ph.D., Department of Anthropology and Soci- ology, University of Queensland, Brisbane, QLD 4072, Australia. E-mail: c.websterAmailbox.uq.edu.au. 169 C 1999 Human Sciences Press, Inc.

Team Leadership: Network Differences in Women's and Men's Instrumental and Expressive Relations

Embed Size (px)

Citation preview

Administration and Policy in Mental HealthVol. 26, No. 3, January 1999

TEAM LEADERSHIP: NETWORK DIFFERENCESIN WOMEN'S AND MEN'S INSTRUMENTALAND EXPRESSIVE RELATIONS

Cynthia Webster, Oscar Grusky, Deborah Podus,and Alexander Young

ABSTRACT: Recent studies have emphasised differences in leadership styles betweenwomen and men. Women have an "interactive" leadership style while men are more "direc-tive" and "authoritative." Social network analysis is used to examine differences in eightmental health case management teams, half formally supervised by women and half bymen. The techniques used are graphical displays and measures of centrality. Results showmale leaders as the most central team member for both instrumental and expressive rela-tions. Female leaders, however, do not adhere to a single leadership style. Team centralisa-tion also differs with gender composition of teams influencing leadership differences.

Interdisciplinary teams have become a prevalent form for accomplishingwork in many areas such as business and health care. Teams are attractivebecause they are seen as an efficient way to divide the labor among anumber of individual specialists who offer multiple skills and perspectives.The sharing of information and knowledge concerning specific issues notonly reduces the risk associated with having only one person responsiblefor a client, project, or investment, but also potentially results in more

Cynthia M. Webster, Ph.D., is with the Department of Anthropology and Sociology, University ofQueensland, Brisbane, Australia. Oscar O. Grusky, Ph.D., is Professor in the Department of Sociology;Deborah Podus, Ph.D., is with the Drug Abuse Research Center, Department of Psychiatry & Bio-behavioral Sciences; Alexander Young, Ph.D., is also with the Department of Psychiatry & Bio-behavioral Sciences, all at the University of California, Los Angeles.

This research was supported in part by Grant T32 MH-14583 (Oscar Grusky, Program Director)from the National Institute of Mental Health. The authors are grateful to the county mental healthdirector, the head of evaluation research, other county officials, and the team staff members, for theirparticipation and assistance.

Address for correspondence: Cynthia M. Webster, Ph.D., Department of Anthropology and Soci-ology, University of Queensland, Brisbane, QLD 4072, Australia. E-mail: c.websterAmailbox.uq.edu.au.

169 C 1999 Human Sciences Press, Inc.

170 Administration and Policy in Mental Health

informed decisions (Rosener, 1990). Such advantages have led to what Sin-clair (1992) has called the "tyranny of team ideology," where teams areviewed as a panacea for solving many of the complex issues faced by orga-nizations.

The demands of a team approach require coordination and leadership.Ideally, work gets done and decisions are made democratically; all teammembers participate, make suggestions, and take initiative. For democraticdecision making to occur, leaders refrain from exercising their legitimatepower and seek information and opinions from other team members.Rosener (1990) points out the risks involved in being inclusive. Solicitinginformation and ideas from others not only takes time and requires surren-dering some control, but is often interpreted as not having needed an-swers and being incompetent. Consequently, many teams function auto-cratically, even when this structure is not efficient (Fiorelli, 1988).

Recent studies have emphasised the differences in leadership styles be-tween women and men. Helgesen (1990) suggests that a team approach ismore consistent with a female leadership style. Women in leadership posi-tions are described as having an interactive, democratic leadership style.They tend to be people-oriented, encouraging the development of rela-tionships with and among team members. Women are also seen as moreconcerned with the expressive, socioemotional needs of team members.Men, on the other hand, are portrayed as having an autocratic leadershipstyle. They are more task-oriented, directive and authoritative, makingmost of the decisions with little sharing of authority.

These two contrasting leadership styles, along with their associated orien-tations, while highlighting general differences in leadership style, also touchon the relational nature of leadership. Viewing leadership as inherent in therelations among individuals not in the individuals themselves, grants a struc-tural approach to the study of leadership behavior first suggested over 20years ago. Salancik and colleagues (1975) noted that because leaders areembedded in a network of other individuals, their behavioral styles areconstrained. They suggested that leaders do not simply choose a leadershipstyle to suit their personalities or to match the situational demands of thetasks at hand. Instead, leaders develop behavioral styles that are consistentwith the relational demands put on them by the individuals with whom theyinteract. If this is correct, then the distinctive leadership styles of women andmen would also be displayed as structural differences.

SOCIAL NETWORK PERSPECTIVE

In the past few years a number of researchers have discussed the valueof network analysis in relation to organizations, in general (Salancik,

A social network approach forces the separation of leadership style andrelational orientation. While convention combines democratic leadershipwith expressive inclinations and autocratic leadership with instrumentalpreferences, this is not required. Autocratic leaders may control the instru-mental relations of the team leaving the expressive side to some othermember, or they may take control of both instrumental and expressiverelations. Democratic leaders may attempt to include all team members inboth task and social concerns or may be inclusive for only instrumentalones. A number of combinations are possible. A team could have a for-mally designated leader who fulfils an administrative role, an instrumentalleader who has a democratic leadership style, and an expressive leaderwho controls socioemotional needs.

The aim of this paper is to determine whether and how structural differ-ences in the leadership relations of women and men are represented.First, we review research that examines gender differences in leadershipbehavior. We identify factors that influence leadership and present a num-ber of propositions. Finally, we introduce relational data from eight mental

Cynthia Webster, Oscar Grusky, Deborah Podus, and Alexander Young 171

1995), and to leadership, in particular (Krackhardt & Brass, 1995; Sparrow& Liden, 1997; Webster, Grusky, Young, & Podus, in press). Employingsocial network analysis, leadership structures can be modeled on multiplerelations utilizing the judgments of all individuals involved. Leaders' ac-tions and subordinates' perceptions need not be uniform. Also possible isthe comparison of formal and informal leadership structures which, inturn, can uncover systematic differences in the ways in which formalleaders are positioned. Network analysis not only allows for an examina-tion and understanding of structure, it permits the analysis of the relationsthat exist and those that do not.

The most theoretically developed set of network measures for the studyof leadership are measures of centrality (Bavelas, 1950; Beauchamp, 1965;Bonacich, 1987; Freeman, 1979; Knoke & Burt, 1983; Leavitt, 1951; Sabi-dussi, 1966). Both individuals and groups can be considered in terms ofcentrality. Central individuals are those who have relations with manyothers within the group. Centralised groups are groups that are dominatedby a single or very few individuals. Because democratic leaders encouragethe development of relationships with and among team members, theseteams have a more decentralised structure and democratic leaders are notnecessarily the most central team member. With autocratic leaders, littlesharing of authority occurs. These teams have a highly centralised struc-ture centered around the leader.

A team approach requires leadership and coordination.

172 Administration and Policy in Mental Health

health case management teams that explore gender differences using anumber of network analytical techniques.

GENDER DIFFERENCES IN LEADERSHIP BEHAVIOR

Leadership Style and Role

Studies that explore gender differences in leadership behavior comefrom different disciplines, each with their particular perspective, and focuson different aspects of leadership. It is not surprising, therefore, that dis-crepant findings exist. Much of the organizational research focuses onleadership style and some report no differences between women and men(Nieva & Gutek, 1981; Kanter, 1977). Bass states, "The preponderance ofavailable evidence is that no consistently clear pattern of differences canbe discerned in the supervisory style of female as compared to maleleaders" (1981, p.499).

The experimental and communication research efforts have tended toexamine differences in instrumental and expressive behaviors. Again, anumber of studies have found no differences in these behaviors for fe-males and males (Winther & Green, 1987; Mabry, 1989; Hawkins, 1995).Operating from an Expectation States/Status Characteristics approach,Johnson, Clay-Warner, and Funk (1996) found that women and men insame-sex groups do not differ in their rates of active task behaviors, direc-tive behaviors, positive socioemotional behaviors, and passive behaviors.Others also find no significant gender differences (Stewart, 1988; Meeker& Elliott, 1996).

Numerous studies, however, have reported stylistic differences betweenwomen and men. Organizational studies by Helgesen (1990), Rosener(1990), Stanford, Oates, and Flores (1995) focus on women's leadershipstyle. These studies found that women leaders emphasise the principles ofinclusion and connection which is an unusual practice. Women encourageparticipation and are willing to share power and information. They de-emphasise formal authority status, which results in a loosely structured net-work, where lines of authority are less defined.

In a meta-analyses of 162 studies on leadership, Eagly and Johnson(1990) obtained the strongest evidence for gender differences in leader-ship style, with women tending to be more democratic and men moreautocratic. Gender differences were detected in task versus social orienta-tion held in laboratory experiments and assessment studies, but not inorganizational studies. In a second meta-analysis, Eagly and Karau (1991)examined gender differences in emergent leadership. They found thatmen emerged more frequently than women on the task leadership mea-

Cynthia Webster, Oscar Grusky, Deborah Podus, and Alexander Young 173

sures, whereas women emerged more frequently than men on social lead-ership measures.

Although to date there have been no network studies that have explicitlyexamined leadership style, research on gender differences in organizationshas been conducted. Lincoln and Miller (1979) took into account genderwhen they compared work and friendship networks within five organiza-tions. In two of the five, they found that males occupied a more centralnetwork position for work contacts. Findings for friendship networks werenot clear. In one organization, females were more central, and in one,males were more central. Ibarra's (1992) study of sex differences in anadvertising firm supports Lincoln and Miller. She found that men had sig-nificandy higher centrality scores across five relational networks, includingadvice and friendship. Differences were most marked in the advice net-work and least in the friendship network.

If work and advice relations are taken as evidence of task orientationand friendship as an indication of social orientation, the results from thenetwork studies are consistent with the sex differences found in organiza-tional research and in the two meta-analyses. Two propositions follow:

Proposition 1: Male leaders have autocratic leadership styles while femaleleaders have democratic leadership styles.

Proposition 2: Male leaders perform more instrumental behaviors whilefemale leaders perform more expressive behaviors.

Mediating Factors

A number of factors have been shown to influence gender differences inleadership style and orientation. Legitimation issues dealing with contextand competence are discussed first. Social distance and gender composi-tion of the group are reviewed next.

Legitimation of Context. Eagly and Johnson (1990) tested the hypothesisthat gender differences in leadership style are less stereotypic in studiesconducted in more natural settings. They found that gender differences insocial and task-related behaviors became smaller in organizational settingsthan in laboratory studies. Eagly and Karau (1991) also looked at contextand found that males emerged as leaders to a lesser degree in naturalgroups and decreased to the extent that leadership was assessed after alonger period of social interaction. These findings lead to the third propo-sition:

Proposition 3: In natural settings, where interaction among group mem-bers has occurred for an extended period of time and le-gitimate work is performed, there are fewer differences in

174 Administration and Policy in Mental Health

males' and females' leadership behaviors for both instru-mental and expressive interactions.

Legitimation of Competence. Researchers employing Expectations States/Sta-tus Characteristics Theory (EST) have identified legitimacy as having asignificant impact on leadership behavior. Lockheed and Hall (1976) ex-amined task-related behaviors in playing a board game under three experi-mental conditions. The first condition consisting of inexperienced, mixed-gender groups resulted in males being seven times as likely as females toemerge as leaders. The second condition examined single-gender groupsand found no differences in the average amount of task-oriented acts byfemales in all-female groups compared with males in all-male groups. Inthe third condition, experienced females from the all-female conditionwere assigned to play the game again in a mixed-sex group. These experi-enced females showed a significantly higher number of task-oriented ac-tions. These results indicate that women are not necessarily less task-ori-ented than men, but that experience or competence increases women'sinstrumental actions.

Eskilson and Wiley (1976) found that females appointed to leadershippositions showed little involvement in task leadership behavior; however,females who were told they had achieved the leadership position showedintense involvement with the instrumental aspect of the group task. Differ-ence in task leadership behavior was explained by the absence for appointedfemale leaders of legitimizing recognition of their personal competence.The legitimation provided by the achievement of the leader role was nec-essary to expand the range of role expectations for female leaders to in-clude those behaviors.

Pugh and Wahrman's (1983) experiments indicated that in the absenceof some form of legitimizing intervention, females in mixed-gender groupswill defer to the judgment of men and men will refuse to be influenced bythe judgment of women. They were able to produce greater equality ofinfluence in mixed-gender groups by providing evidence to both malesand females of female competence on a single task. Furthermore, theyshowed that demonstrating female superiority is more effective than dem-onstrating the equal competence of women. Women, however, nevergained a significant advantage over men even after a demonstration oftheir superiority. They concluded that being as good as a man is notenough to enable a woman to succeed.

A study by Walker, Ilardi, McMahon, and Fennell (1996) showed a morecomplex outcome when an assigned leader is not legitimated. In 19 mixed-gender groups, one member of each group was randomly selected to serveas leader (9 females, 10 males). Designated leaders were no more likelythan any other group member to serve as influence leaders. Groups with a

Cynthia Webster, Oscar Grusky, Deborah Podus, and Alexander Young 175

designated female leader enhanced the other female members' influencepotential. The authors concluded that a formal female leader only needs tobe present for other females to attempt to attain a leadership position. Thisis consistent with female leaders' democratic style in that they are morewilling to allow others to attain influential positions within the group.

The above four studies lead to the fourth proposition:

Proposition 4: The legitimation of females in leadership positions, eitherby designating females to leadership positions or by dem-onstrating females' competence, increases instrumentalbehaviors of female group members.

Social Distance. Organizational research shows that leaders tend to so-cially distance themselves from their subordinates in order to maintaintheir authority (Homans 1974; Fernandez 1991). When leaders' relation-ships with subordinates become too familiar, subordinates tend to disre-gard directives. Taking the gender of the leader into account, ExpectationsStates/Status Characteristics Theory (EST) predicts that female leadersmust be well-intentioned and competent for others to accept their positionof authority (Meeker & Weitzel-O'Neill, 1977; Fairhurst & Snavely, 1983;Ridgeway, 1988). Women must provide encouragement and approval toothers and show that their instrumental acts are due to their concern forthe group, not to promote their control.

The experimental evidence on gender differences, however, does notsupport the EST line of reasoning. Bradley (1980) found that while highlycompetent females exercised more influence over others, they were lesswell liked than their male counterparts. Wood and Karten's (1986) resultsshowed that members' perceived competence was significantly and pos-itively related to their task behavior, but negatively related to their socio-emotional behavior. Studies on emergent leadership (Owen, 1986; Bor-mann, Pratt, & Putnam, 1978) confirmed these findings. Females whoevidenced task-relevant competence emerged as group leaders; however,those who did attain leadership positions tended to be disliked for theirefforts.

Although EST recommends that women attempt to be both the instru-mental and socioemotional leaders of their groups, the empirical evidencesuggests that the acceptance of female leaders in this dual leadership roleis rare. This leads to the following:

Proposition 5: To be perceived as competent by other group members,both male and female leaders socially distance themselvesfrom the group members, which results in female leadersbeing disliked by other group members.

176 Administration and Policy in Mental Health

Gender Composition. Studies also show that the gender composition ofgroups influences the interactions among group members. Eskilson andWiley (1976) conducted experiments specifically to examine the effects ofsex composition on leadership. They found that male leaders received themost cooperation and were most effective in the all-male group situationand least effective in the mixed-sex situations. Female leaders were mosteffective when leading mixed-sex groups. In groups consisting of only malefollowers, female leaders performed few instrumental acts and receivedvery little response from males. Even though female leaders performedmore task-oriented acts in all-female groups, these groups were least effec-tive in completing the task.

Congruous results are obtained from the network studies that havelooked at gender composition within organizational settings. In his studyof a newspaper publishing company, Brass (1985) reports that the largestdifference between women and men was based on the gender compositionof the work group. Compared to women in all-female groups, women inintegrated groups were more central, had greater access to powerful indi-viduals, and were rated as more influential by their supervisors. Thesewomen also were not significantly different from their male counterparts.In a comparative study, Aldrich, Reese, and Dubini (1989) discovered thatalthough the overall size of networks did not differ for women and men,women had significantly more cross-sex ties than men.

Ibarra's (1992, 1993a) research conducted in an advertising firm ex-pands on the above findings. Women had a greater proportion of men asadvice and influence ties, an equal amount of women and men as commu-nication and support relations, and a greater proportion of women asfriends. Women obtained social support and friendship from ties to femalecoworkers, but relied on relationships with males to gain access to instru-mental resources. Although the men in the study displayed a greater degreeof homophily across all five relational networks, they did select some womenas friends and support ties. Ibarra concluded that preferences for homo-phily and status tend to coincide for men and exist in competition forwomen. Furthermore, women's functionally differentiated networks resultin an excess of uniplex ties which are more time consuming to maintain.

The above findings lead to the final three propositions:

Proposition 6: Gender composition of the group influences the instru-mental and expressive interactions among group mem-bers.

Proposition 7: For instrumental relations, both females and males ap-proach males for information: males' relations adhere tothe principle of homophily; females' relations are hetero-philous.

Cynthia Webster, Oscar Grusky, Deborah Podus, and Alexander Young 177

Proposition 8: For expressive relations, females choose other females forsocial contact; males choose other males, but includesome women: both females' and males' relations are ho-mophilous, females' are more so than males'.

METHOD

Mental Health Case Management Teams

A county adult mental health system in California served as the site forthis study. The system included 10 case management teams, consisting of112 clinical staff. Each of the teams had a formally assigned supervisor.Two of the supervisors had been in their positions for less than 6 months,these teams were excluded from this study. The eight supervisors rangedin age from 44 to 50 years (average age was 47), had been mental healthprofessionals for an average of 22 years (range from 8 to 30 years), andhad worked for the county's Department of Mental Health for an averageof 16 years (range from 6.5 to 24 years). Half of the supervisors were fe-male and two were of Hispanic background. All were well-educated withone having a doctorate, three master's and four with bachelor's degrees.Four were registered nurses, one was a doctorate psychologist, and threewere trained in social work.

Data used in this analysis were obtained from all clinical staff who hadbeen a member of their respective team for at least 1 year prior to datacollection. Clinical staff included psychiatrists, psychologists, social work-ers, nurses, psychiatric technicians, rehabilitation therapists, and commu-nity service workers. Although teams differed in their staff make-up, allteams had psychiatrists, social workers, and psychiatric technicians. Theeight teams examined ranged in size from 8 to 11 individuals. Three teamsconsisted of 8 members, two of 9 members, one of 10, and two of 11members. Half of the teams were predominantly female with the averagemake-up 54% female, ranging from 33% to 75%.

Social Network Data

Team members were interviewed individually using a structured inter-view protocol. During the interview they completed a number of self-re-port tasks concerning their relationships with other team members. Onetask had team members focus on their instrumental relations and theother inquired about their expressive affiliations within the team. Bothtasks were presented in a checklist format with individually randomizedlists of names of all team members. The instrumental task asked infor-mants to rank order each team member "according to how often youSEEK ADVICE from them." The expressive task had members rank order

178 Administration and Policy in Mental Health

other members "according to how often you SPEND TIME with them dur-ing work hours."

Both tasks resulted in a single vector for each informant for the tworelations. The vectors for the advice relationships were combined to form asquare actor-by-actor matrix of ADVICE relations. The vectors for the so-cial affiliations were combined to form a square actor-by-actor matrix ofSOCIAL relations. Both matrices were dichotomized to record only the topthree choices sent by each team member. Since direction of choice wasnoted, both advice and social relationships were recorded as non-sym-metric relations. Moving across the rows in these matrices choices sent byan informant are displayed and going down the columns indicates choicesreceived.

Centrality and Centralization as Measures of Leadership Structure

Since we are interested in examining both the centrality of specific teammembers as well as the centralization of the teams themselves, we use themeasures of degree, closeness, and betweenness. Freeman (1979) clarifiedthe meaning of actor centrality based on these three measures and thenderived network centralization measures for each. He discussed how thesethree measures relate to different aspects of centrality. Degree centrality ofa point, which is the sum of all other points directly connected to it, signi-fies activity level. Degree centrality can handle directed ties, where thenumber of ties reported are not the same as the number of ties received.The sum of the ties reported are the actor's outdegrees while the numberof ties received are the actor's indegrees. Closeness centrality is based onthe notion of distance. If an actor is close to all other actors, a distance ofno more than 1, for example, that actor is not dependent on any other toreach everyone in the network. Closeness measures centrality as indepen-dence or efficiency. Betweenness is a measure of potential for control. Anactor who is the only link between pairs of other actors is in the position tofacilitate interaction or to complicate it.

HYPOTHESES

From the propositions, we have established a number of individual-leveland group-level hypotheses for instrumental and expressive relations.

Instrumental Relations

Hypothesis 1a: Both male and female supervisors are the most central mem-ber of their team.

Hypothesis 1b: Male supervisors are more central than female supervisors.

Cynthia Webster, Oscar Grusky, Deborah Podus, and Alexander Young 179

Hypothesis 2a: Teams with male supervisors are more centralised than teamswith female supervisors.

Hypothesis 2b: Teams with male supervisors, but a predominance of femaleteam members, are the most centralised

Hypothesis 2c: Teams with female supervisors, and a predominance of fe-male team members, are the least centralised.

Hypothesis 3a: Male supervisors on predominantly female teams are themost central supervisors

Hypothesis 3b: Female supervisors on predominantly female teams are theleast central.

Expressive Relations

Hypothesis 1a: Neither female nor male supervisors are the most centralmember of their team.

Hypothesis 1b: Female supervisors are more central than male supervisors.Hypothesis 2a: Teams with female supervisors are more centralised than

teams with male supervisors.Hypothesis 2b: Teams with female supervisors, and a predominance of fe-

male team members, are the most centralised.Hypothesis 2c: Teams with male supervisors, and a predominance of fe-

male team members, are the least centralised.Hypothesis 3a: Female supervisors on mixed-gender teams are the most

central supervisors.Hypothesis 3b: Male supervisors on predominantly female teams are the

least central.

SOCIAL NETWORK ANALYSES AND RESULTS

Instrumental Ties

To determine whether there is any evidence to support hypotheses laand 1b, we examine the differences in centrality between male and femalesupervisors. The three instrumental centrality scores for the ADVICE rela-tions are displayed in Table 1. Teams with male supervisors have beenassigned odd numbers (1, 3, 5, 7) and teams with female supervisors evennumbers (2, 4, 6, 8). The scores are normalised so that they can be com-pared. Notice that the indegree and the closeness scores for two of themale supervisors are 100%, meaning that every member of their respectiveteams included them in their choices of ADVICE seeking (Table 1).Clearly, male supervisors on average are substantially more central for in-strumental relations than female supervisors on all three centrality mea-sures.

The column labelled "Rank" in Table 1 shows the relative positioning forthe supervisors on their respective teams. Two of the male supervisors arethe most central member of their team for all three centrality scores andthe two others share the most central position with one other team mem-ber or are the second most central for at least one of the measures. Noneof the female supervisors are exclusively the most central member for allthree measures. Two share the most central position or are the secondmost central member of their team while the other two are not the mostcentral on any of the measures.

For hypotheses 2 and 3, we examine the network centralization scoresand gender composition (Table 2). Not surprisingly, teams with male su-pervisors are on average more centralised than teams with female super-visors. The most centralised team, using all three centrality measures, isTeam 1 with a male supervisor and a majority of female team members(75%). This Team 1 supervisor is also the most central supervisor, better-ing the male supervisor on Team 3 on his betweenness centrality score.Team 3 is the second most centralised team with the second most centralmale supervisor, yet the team does not have a majority of female members

180 Administration and Policy in Mental Health

TABLE 1Supervisors' Network Centrality Scoresand Rankings for ADVICE Relations

Team

Males1357

X =

Females2468

X =

Indegree

Score Rank

Closeness

Score Rank

Betweenness

Score Rank

100.00100.0087.5060.00

86.875

1112

100.00100.0088.8976.92

91.45

1121.5

71.4342.8619.6428.52

40.61

1121

77.7870.0057.1442.86

61.945

11.534

81.8276.9270.0070.00

74.685

1.5255.5

15.9716.671.590.95

8.795

2256.5

Note. UCINET IV (Borgatti, Everett, & Freeman, 1992) was used for the centrality analyses (Tables 1,2,3,& 4).

(44%). The least centralised teams with the least central male supervisorsalso contain a majority of male team members.

Of the teams with female supervisors, all but Team 4 have a majority offemale members. Team 6, with the highest percentage of females (75%), isthe most centralised team for two of the three measures, yet the Team 6supervisor is at best the third most central member of her team and thethird most central of the female supervisors. Team 2, with the lowest per-centage of females (42%), is the second most centralised. This female su-pervisor, however, is the most central member of her team and is the mostcentral of the four female supervisors. The least centralised, Team 8, doesconsist of a majority of female members (62.5%) and has the least centralof the four female supervisors.

These results for instrumental ties can be visually represented in theform of digraphs (Figures 1 and 2). In a digraph, actors are pictured aspoints (called nodes) and the relations between actors are shown as di-rected lines (called arcs) connecting the nodes. Unreciprocated ties aredirected by a single arrow whereas mutual ties are headed by arrows onboth ends. To emphasize the mutual ties, these ties also have been drawnwith a thicker line. Since digraphs may be oriented in any number of ways,Figures 1 and 2 have been arranged to incorporate the indegree centralityscores. The most central actors are located at the top of the digraphs andperipheral members are towards the bottom.

Cynthia Webster, Oscar Grusky, Deborah Podus, and Alexander Young 181

TABLE 2Gender Composition and Team Centralization

Scores for ADVICE Relations

Team

Males1357

Females6428

% Female Indegree Closeness Betweenness

75.0044.4433.3345.45

85.71485.71466.07153.333

X = 72.708

81.82075.26071.90034.080

65.765

71.43040.18027.90023.590

40.775

62.5045.4570.0062.50

57.14355.55656.94442.857

61.12052.67032.65036.280

18.14039.22012.50021.540

X = 53.125 45.680 22.850

Digraphs allow for more of the structural details to be revealed. Forexample, the digraph of Team 1 clearly shows why the male supervisor isso central. The supervisor is the one team member who connects threedisjointed components. Of the two male supervisors and the one femalesupervisor who share the most central position (Teams 5, 7, and 4), all

182 Administration and Policy in Mental Health

FIGURE 1Digraphs for Male Instrumental Relations

have mutual ties to the other central team member. In all three of thesecases the other central actor happens to be a male team member. For thetwo teams in which the female supervisors are not the most central teammember (Teams 6 and 8), both female supervisors are mutually connectedto the most central team member. In both of these cases another female

Cynthia Webster, Oscar Grusky, Deborah Podus, and Alexander Young 183

FIGURE 2Digraphs for Female Instrumental Relations

184 Administration and Policy in Mental Health

team member is in the most central position. These mutual ties linking thesupervisors with other highly central team members not only indicate thatthe supervisors are aware of the informal, instrumental leaders on theirteams, they also show that the informal leaders acknowledge the formalinstrumental position of the supervisors.

Expressive Ties

We now turn to the first set of hypotheses for the SOCIAL data. Table 3shows the centrality scores for both male and female supervisors. Thesescores for expressive relations are lower than those for instrumental rela-tions. Contrary to the hypotheses, male team supervisors are on averagemore central than are female supervisors, as they are for the ADVICE rela-tions. Comparing the relative ranking of the supervisors shows the malesupervisors, unlike the female supervisors, remain ranked among the mostcentral team members in both ADVICE and SOCIAL relations on theirrespective team. Three of the four male supervisors have the highest rank-ing centrality score on their teams for at least one of the centrality mea-sures. None of the centrality scores for the female supervisors are in thetop place. In fact, the female supervisors rank about sixth on their teams,averaging across all the measures, while the male supervisors rank second.

TABLE 3Supervisors' Network Centrality Scores

and Ranking for SOCIAL Relations

Team

Males3571

X =

Females4628

X =

Indegree

Score Rank

Closeness

Score Rank

Betweenness

Score Rank

50.0050.0040.0028.57

42.14

21.52.53

80.0057.1466.6741.18

61.25

1421

51.7925.0023.7040.48

35.24

1211

20.0014.2911.110.00

11.35

56.597

58.8253.8556.5163.64

58.20

2.5854.5

18.002.380.690.00

5.27

37.57.56

Cynthia Webster, Oscar Grusky, Deborah Podus, and Alexander Young 185

Differences in which of the supervisors is the most central for expressiverelations do emerge. The male supervisor for Team 1 was the most centralin the ADVICE data, but he is the least central male supervisor in theSOCIAL data. The female supervisor for Team 2 also moves from beingthe most central for ADVICE to second to last for SOCIAL relations. Team8's female supervisor, however, remains the least central supervisor forboth relations.

Table 4 displays the centralization scores for the SOCIAL data. The in-degree and closeness centralization scores show that on average the teamsheaded by female supervisors are more centralised than those with malesupervisors. The betweenness centrality scores, however, show that two ofthe female headed teams are less centralised than any of the male ledteams. Comparing the SOCIAL centralization scores with the ADVICE cen-tralization scores indicates that the female led teams are slightly more cen-tralised in their SOCIAL scores than in their ADVICE, whereas all of themale led teams are less centralised in their SOCIAL relations. Of the teamswith female supervisors, Team 4, which is the only team without a predom-inance of females, is the least centralised but has the most central super-visor. The one team with a male supervisor and a predominance of fe-males, Team 1, is one of least centralised teams. Differences also exist inthe teams' centralization scores for instrumental and expressive ties. Team8 with a female supervisor is the least centralised team for ADVICE rela-

TABLE 4Gender Composition and Team Centralization

Scores for SOCIAL Relations

Team

Males3517

Females8264

% Female Indegree Closeness Betweenness

44.4433.3375.0045.45

32.14333.92928.57130.000

X = 31.161

49.37045.07021.26018.760

33.615

46.65031.47031.47017.190

31.695

62.5070.0062.5045.45

92.85758.33338.09536.667

X = 56.488

77.69056.51033.01014.740

45.487

47.62040.69013.61014.190

29.027

186 Administration and Policy in Mental Health

tions but is the most centralised team for SOCIAL relations. Team 1 with amale supervisor is the most centralised team for ADVICE relations andone of the least centralised for SOCIAL ties.

Again, we turn to the digraphs in Figures 3 and 4 for a more detailedpicture of the expressive relations. Although the female-led teams are

FIGURE 3Digraphs for Female Expressive Relations

more centralised than are the male led teams, they are not dominated bythe female supervisors. In fact, not only do the female supervisors hold aless central position, none has a mutual relation with the most centralmember on their respective teams. This is not to say that none of theseteams has a dominant member. Team 2 is centralised around a male team

Cynthia Webster, Oscar Grusky, Deborah Podus, and Alexander Young 187

FIGURE 4Digraphs for Male Expressive Relations

188 Administration and Policy in Mental Health

member and Team 8 around a female member. The two female super-visors of these team do choose the most central team member; however,neither of their ties are reciprocated. Of the teams with male supervisors,two supervisors share the most central position. The supervisor on Team 5has a mutual tie with the other central member, and the Team 3 supervisorchooses one of the other central members but his tie is not reciprocated.Three of the four male supervisors choose the most central member ontheir respective teams. Unlike the female supervisors, the male supervisorscompete for the expressive leader position. The male supervisor on Team1 has a high betweenness centrality score because he connects members 3and 4 to the others.

DISCUSSION AND CONCLUSION

A social network approach to the study of gender differences in leader-ship both confirms and extends previous research findings.

For instrumental relations: Male leaders tend to have an autocratic leader-ship style. Their teams are highly centralised with the male leader as themost central member of the team. Female leaders do not adhere to asingle leadership style. Their teams are not as centralised as teams led bymales, and they may or may not be the most central member. Gendercomposition of teams influences leadership differences. Teams with a ma-jority of female members and a male leader are highly centralised with themale leader in the most central position. Teams with a majority of femalemembers and a female leader are more decentralised with the femaleleader not necessarily in the most central position.

For expressive relations: Female leaders tend to have a democratic leader-ship style. Their teams are more centralised than teams led by males, yetfemale leaders are not the most central team member. Male leaders tendto have an autocratic leadership style. While their teams are not as central-ised as teams led by females, male leaders tend to be more central than arefemale leaders. And as described above under instrumental relations, gen-der make-up of teams influences leadership differences. Teams with a ma-jority of female members and a female leader tend to be more centralisedwith the female leader not in the most central position. Teams with a ma-jority of female members and a male leader tend to be less centralised withthe male leader not in the most central position. Female leaders on teamswith a majority of males tend to be more central than female leaders onpredominantly female teams.

These findings are merely suggestive due to the small number of teamsexamined and the limited operationalization of instrumental and expres-sive leadership. In this research, traditional measures of autocratic and

Cynthia Webster, Oscar Grusky, Deborah Podus, and Alexander Young 189

democratic leadership styles were not directly assessed. This needs to berectified. Also, there are multiple ways in which both leadership behaviormay be defined, as deference relations or respect relations, for example(Fernandez, 1991), and these require investigation. Finally, as noted byIbarra (1993b), and reiterated by Salancik (1995), further research em-ploying network analysis should focus on the consequences of structuraldifferences. Empirical work is needed to determine whether certain net-work structures are necessary for achieving successful outcomes or whetherdifferent structural arrangements can produce similar results.

REFERENCES

Aldrich, H., Reese, P.R., & Dubini, P. (1989). Women on the verge of a breakthrough: Networkingamong entrepreneurs in the United States and Italy. Entrepreneurship E Regional Development, 1, 339-356.

Bass, B.M. (1981). Stogdill's handbook of leadership: A survey of theory and research. New York, NY: The FreePress.

Bavelas, A. (1950). Communication patterns in task-oriented groups. Journal of the Acoustical Society ofAmerica, 22, 271-282.

Beauchamp, M. A. (1965). An improved index of centrality. Behavioral Science, 10, 161-163.Bonacich, P. (1987). Power and centrality: A family of measures. American Journal of Sociology, 92, 1170-

1182.Borgatti, S. P., Everett, M.G., & Freeman, L.C. (1992). UCINET IV Version 1.05. Columbia, MD: Analytic

Technologies.Bormann, E.G., Pratt, J., & Putnam, L. (1978). Power, authority and sex: Male response to female

leadership. Communication Monographs, 45, 119-155.Bradley, P.H. (1980). Sex, competence and opinion deviation: An expectations states approach. Com-

munication Monographs, 47, 101-110.Brass, D.J. (1985). Men's and women's networks: A study of interaction patterns and influence in an

organization. Academy of Management Journal, 28, 327–343.Eagly, A.., & Johnson, B.T. (1990). Gender and leadership style: A meta-analysis. Psychological Bulletin,

108, 233-256.Eagly, A.H., & Karau, S.J. (1991). Gender and the emergence of leaders: A meta-analysis. Journal of

Personality and Social Psychology, 60, 658-710.Eskilson, A., & Wiley, M.G. (1976). Sex composition and leadership in small groups. Sociometry, 39,

183-194.Fairhurst,G.T., & Snavely, B.K. (1983). Majority and token minority group relationships: Power acquisi-

tion and communication. Academy of Management Review, 8, 292-300.Fernandez, R.M. (1991). Structural bases of leadership in intraorganizational networks. Social Psychology

Quarterly, 54, 36-53.Fiorelli, J.S. (1988). Power in work groups: Team member's perspectives. Human Relations, 41, 1-12.Freeman, L.C. (1979). Centrality in social networks: I. Conceptual clarification. Social Networks, 1, 215-

239.Hawkins, K.W. (1995). Effects of gender and communication content on leadership emergence in

small task-oriented groups. Small Group Research, 26, 234-249.Helgesen, S. (1990). The female adavantage: Women's ways of leadership. New York, NY: Doubleday/Cur-

rency.Homans, G.C. (1974). Social behavior: Its elementary forms (rev. ed.). New York, NY: Harcourt, Brace.Ibarra, H. (1992). Homophily and differential returns: Sex differences in network structure and access

in an advertising firm. Administrative Science Quarterly, 37, 422–447.Ibarra, H. (1993a). Power, social influence, and sense making: Effects of network centrality and prox-

imity on employee perceptions. Administrative Science Quarterly, 38, 277-303.Ibarra, H. (1993b). Personal networks of women and minorities in management: A conceptual frame-

work. Academy of Management Review, 18, 56-87.Johnson, C. (1993). Gender and formal authority. Social Psychology Quarterly, 56, 193-211.

190 Administration and Policy in Mental Health

Johnson, C., Clay-Warner, J., & Funk, S.J. (1996). Effects of authority structures and gender on interac-tion in same-sex task groups. Social Psychology Quarterly, 59, 221-236.

Kanter, R.M. (1977). Men and women of the corporation. New York, NY: Basic Books.Knoke, D., & Burt, R.S. (1983). Prominence. In R.S. Burt & M.J. Minor (Eds.), Applied network analysis

(pp. 195–222). Newbury Park, CA: Sage Publications.Krackhardt, D., Blythe, J., & McGrath, C. (1994). KrackPlot 3.0: An improved network drawing pro-

gram. Connections, 17, 53-55.Krackhardt, D., & Brass, D.J. (1994). Intraorganizational networks: The micro side. In S. Wasserman &

J. Galaskiewicz (Eds.), Advances in social network analysis (pp. 207–229). Thousand Oaks, CA: SagePublications.

Leavitt, H.J. (1951). Some effects of communication patterns on group performance. Journal of Abnor-mal and Social Psychology, 46, 38-50.

Lincoln, J.R. & Miller, J. (1979). Work and friendship ties in organizations: A comparative analysis ofrelational networks. Administrative Science Quarterly, 24, 181–199.

Lockheed, M.E., & Hall, K.P. (1976). Conceptualizing sex as a status characteristic: Applications toleadership training strategies. Journal of Social Issues, 32, 111-124.

Mabry, E.A. (1989). Some theoretical implications of female and male interaction in unstructuredsmall groups. Small Group Behavior, 20, 536-550.

Meeker, B., & Elliott, C. (1996). Reward allocations, gender, and task performance. Social PsychologyQuarterly, 59, 249–301.

Meeker, B., & Weitzel-O'Neill, P.A. (1977). Sex roles and interpersonal behavior in task-orientedgroups. American Sociological Review, 42, 91–105.

Nieva, F., & Gutek, B.A. (1981). Women and work: A psychological perspective. New York, NY: Praeger.Owen, W.F. (1986). Rhetorical themes of emergent female leaders. Small Group Behavior, 17, 475-486.Pugh, M.D., & Wahrman, R. (1983). Neutralizing sexism in mixed-sex groups: Do women have to be

better than men? American Journal of Sociology, 88, 746-763.Ridgeway, C. (1988). Gender differences in task groups: A status and legitimacy account. In M. Web-

ster & M. Foschi (Eds.), Status generalization: New theory and research (pp.188–206). Stanford, CA:Stanford University Press.

Rosener, J.B. (1990). Ways women lead. Harvard Business Review, 68(6), 119-125.Sabidussi, G. (1966). The centrality index of a graph. Psychometrika, 31, 581-603.Salancik, G.R., Calder, B.J., Rowland, K.M., Leblebici, H., & Conway, M. (1975). Leadership as an

outcome of social structure and process: A multidimensional analysis. In J.G. Hunt & L.L. Larson(Eds.), Leadership frontiers (pp. 81-101). Kent, OH: Kent State University Press.

Salancik, G.R. (1995). Wanted: A good network theory of organization. Administrative Science Quarterly,40, 345-349.

Sinclair, A. (1992). The tyranny of a team ideology. Organization Studies, 13, 611-626.Sparrow, R.T., & Liden, R.C. (1997). Process and structure in leader-member exchange. Academy of

Management Review, 22, 522-552.Stanford, J.H., Oates, B.R., & Flores, D. (1995). Women's leadership styles: A heuristic analysis. Women

in Management, 10, 9–16.Stewart, P. (1988). Women and men in groups: A status characteristics approach to interaction. In M.

Webster & M. Foschi (Eds.), Status generalization: New theory and research, (pp.69–85). Stanford, CA:Stanford University Press.

Walker, H.A., Ilardi, B.C., McMahon, A.M., & Fennell, M.L. (1996). Gender, interaction, and leader-ship. Social Psychology Quarterly, 59, 255-272.

Webster, C.M., Grusky, O., Young, A., & Podus, D. (in press). Leadership structures in case manage-ment teams: An application of social network analysis. In J.P. Morrissey (Ed.), Research in communitymental health. Stamford, CT: JAI Press.

Winther, D.A., & Green, S.B. (1987). Another look at gender-related differences in leadership behav-ior. Sex Roles, 16, 41-56.

Wood,W., & Karten, S.J. (1986). Sex differences in interaction style as a product of perceived sexdifferences in competence. Journal of Personality and Social Psychology, 50, 341-347.