"For the Article provided : "

WHEN IS FEMALE LEADERSHIP AN ADVANTAGE? 1157 Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. 36, 1153–1175 (2015) DOI: 10.1002/job male counterparts at creating cohesion and at facilitating cooperative learning and participative communication. Hence, Hypothesis 1: As functional diversity increases, female-led teams will report (a) more cohesion, (b) more cooperative learning, and (c) more participative communication as compared with male-led teams. Larger and geographically dispersed teams are also likely to benefit more from female leaders than from male leaders in developing communal team outcomes (e.g., cohesion, cooperative learning, and participative communication), because of the tendency for women to have more of a relational self-concept than men. Creating team and mutually empowering members—which are critical relational activities (Fletcher, 1998)—should help larger and geographically dispersed teams overcome the communication, cooperation, and coordination challenges that larger and geographically dispersed teams tend to encounter. In their study of over 7000 scientific teams in the biotech industry Tzabbar and Vestal (In press) suggest that trust, familiarity, and shared understanding help reduce coordination and cooperation costs associated with team dispersion. Leaders on globally dispersed teams who develop high quality relationships with their subordinates and also communicate with them frequently enable a higher level of member participation in team decisions (Gajendran & Joshi, 2012). When team members feel known, they report a higher level of interpersonal trust, which, in turn, is associated with higher levels of personal learning (Purvanova, 2013). And, inspirational leaders of geographically dispersed teams, because their communications focus on the collective (rather than on themselves), appear especially apt at eliciting team trust and commitment, both of which facilitate effective group functioning (Joshi, Lazarova, & Liao, 2009). Because female leaders are more likely than men to practice relational leadership (due to the higher frequency of relational self-construal among women relative to men), I argue that female leaders may, therefore, also be more successful in mitigating the coordination requirements that larger and geographically dispersed teams pose to the quality of the relationships between team members and the team and to team interaction norms. Thus, as illustrated in Figure 1, I anticipate that larger teams and geographically dispersed teams will report more cohesion, cooperative learning, and participative communication when they are led by women as compared with when they are led by men. Hence, Hypothesis 2: As team size increases, female-led teams will report (a) more cohesion, (b) more cooperative learning, and (c) more participative communication as compared with male-led teams. Figure 1. Hypothesized relationship with statistically significant results in bold 1158 C. POST Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. 36, 1153–1175 (2015) DOI: 10.1002/job Hypothesis 3: Among geographically dispersed teams, female-led teams will report (a) more cohesion, (b) more cooperative learning, and (c) more participative communication as compared with male-led teams. Method Sample and procedure The study’s data collection was initiated through a research alliance with the Industrial Research Institute (IRI), a prominent professional association of industrial R&D executives representing over 200 major industrial firms. Member companies of the IRI participated in a study about diversity and innovation by volunteering teams to complete an online survey. A committee of senior IRI scientists and engineers provided input into the development of the survey instrument to ensure that it would not compromise the sensitive nature of the work being carried out. In addition, a pilot study with 28 teams was conducted to identify the most reliable survey items for inclusion in the final survey to ensure that the time required to take the survey did not exceed the 30-min time limit that participating organizations had requested for the survey. The teams that participated in the survey were tasked with product, service, or process innovation, were cross-functional, had spent at least three months together, and were either still operating or had disbanded no more than 60 days prior to the study. In total, 86 cross-functional innovation teams representing 837 individuals from 29 organizations participated in the study. To ensure the highest level of reliability in the aggregated data, I excluded teams from the study (i) if fewer than 70% of their members responded to the survey or (ii) if leaders’ gender was not identifiable. This resulted in the loss of four teams. Thus, the final sample comprises 82 teams. Approximately 30% of the teams had female leaders. The average number of functions represented on a team was 3.7, the average team size was 11 members, and the average tenure was one to two years. Twenty-seven percent of the teams were geographically dispersed. For each team, all members were asked to complete a 30 -min online survey that provided information for the study. On average, per team, 93 percent of team members responded to the survey. Unless otherwise indicated, the response range is 1 = “Strongly disagree.” to 7 = “Strongly agree.” Measures Dependent variables Because participating organizations imposed strict limitations on the survey length, in many cases, it was impractical to include intact scales developed in peer-reviewed work. Personal judgment and results from a pilot study guided the selection of the items for the survey. Consistent with Bollen and Hoyle’s (1990) conceptualization of the social dimension of cohesion as both a feeling of belonging to a particular group and a feeling of morale associated with group membership, the measure of cohesion consists of two items from Bollen and Hoyle’s (1990) scale, one from each the two dimensions of their construct: “I feel a sense of belonging to [this team]” and “I am enthusiastic about being a member of [this team].” Factor analysis at the team member level showed a single-factor structure with a Cronbach alpha of 0.76. Hence, I averaged the items into an individual level measure of cohesion. The measure of cooperative learning was derived with items adapted from the two dimensions of Janz and Prasarnphanisch’s (2003) scale of cooperative learning that best capture the relational aspect of team learning: promotive interactions and positive interdependence. My factor analysis of pilot survey data identified two items pertaining to promotive interactions and one item related to positive interdependence as providing the most parsimonious and reliable measure of cooperative learning. The three items loaded onto one factor and were averaged into a scale of cooperative learning (Cronbach a = .86). Given the lack of a previously established measure of participative communication, Jassawalla and Sashittal’s (2002, 2006) conceptual work informed the development of the participative communication measure. Twelve items relating to participative communication derived from Jassawalla and WHEN IS FEMALE LEADERSHIP AN ADVANTAGE? 1159 Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. 36, 1153–1175 (2015) DOI: 10.1002/job Sashittal (2002, 2006) and also drawn from Carson et al. (2007) were used in a pilot survey, conducted in conjunction with the present survey, with responses from 308 individuals representing 28 teams. I conducted factor analyses of the pilot survey data, and they suggested a more parsimonious set of items for the present study, which also helped address the survey length concerns of participating organizations. Thus, the measure of participative communication used for the entire sample consists of five items. Two items from Carson et al.’s (2007) voice scale capture participation and input: “People on this team are encouraged to speak up and test assumptions about issues under discussion,” and “This team encourages everyone to actively participate in decision making.” The remaining three items are derived from Jassawalla and Sashital’s (2002, 2006) conceptual work to capture transparency (“Team members actively share their special knowledge and expertise with one another;” “Team members have access to all information available to the team.”) and mindfulness (“Team members are listened to and taken seriously.”). All items loaded onto a single factor (Cronbach a = 0.81) and I averaged them into an individual level scale. To gain greater confidence in my measures, I took several steps. First, I conducted a confirmatory factor analysis on all ten items constituting the three dependent variable measures. This validated the three factor structure and showed good fit with the data (comparative fit index = .97, root mean square error of approximation = .06 (Hu & Bentler, 1999), relative chi square index = 4.187 (Schumacker & Lomax, 2004), standardized root mean square residual = .035, and GFI = .965.) Second, I assessed the correlation among the three dependent variable measures. At the level of the team members, the correlations among the three dependent variables ranged from 0.45 to 0.65, suggesting positive, but moderate relationships between the measures. To determine if team-level aggregation of each dependent variables was empirically justifiable, I took two precautions. First, I evaluated within-group agreement to determine the amount of consensus among team members in their evaluations of the team environment (Klein, Conn, Smith, & Sorra, 2001; Kozlowski & Hattrup, 1992). I assessed within-group agreement using Klein et al. (2001) rwg(J) procedure. To further evaluate whether aggregation of the dependent variables to the team level was empirically justifiable, I performed the ICC(1) intra- class correlation coefficients test (Raudenbush & Bryk, 2002), which provides an indication of the amount of variance attributed to group membership (Bliese, 2000). Although no absolute standard cut-off values for rwg(J) or ICCs have been established, aggregation is deemed warranted for rwg(J) values at or above 0.70 and ICC(1) values above .05 (Bliese, 2000). The median rwg(2) for cohesion across teams is .83, with a range from .07 to .97. For nine of the teams, the rwg (2) fell below .70, necessitating robustness checks. For cohesion, the ICC(1) value of 0.09, F (82,738) = 2.01 and p<0.001 shows sufficient between-group variance. The median rwg(3) for cooperative learning across teams is .82, with a range from .36 to .99. For seven of the teams, the rwg(3) fell below .70, necessitating robustness checks. For cooperative learning, the ICC(1) value of 0.05, F(82,723) = 1.62 and p<0.001 shows sufficient between-group variance (Bliese, 2000). The median rwg(5) for participative communication across teams is .90, with a range from .19 to .99. For three of the teams, the rwg(5) fell below .70, necessitating robustness checks. The ICC(1) value for participative communication was 0.07, F(82,754) = 2.36 and p<0.001. Based on these indicators, I aggregated the individual-level dependent variable measures into team-level variables (Bliese, 2000). Leader gender Participating organizations provided team member lists that identified the formal team leaders. I ascertained leader gender using the first name of the identified leaders (if unambiguous) or their survey response to the question about their gender. The leader gender dummy variable is equal to 1 for teams with female leaders, and zero otherwise. Functional diversity Team members self-reported their functional area by selecting from a predetermined list the function that best matched their primary job responsibility. These functions included accounting, auditing and finance; general management and human resources; computer and information services; engineering; marketing, sales, and customer service; production, manufacturing, and supply chain; basic and applied research; development; legal; and others. Drawing on Harrison and Klein’s (2007) diversity typology, I conceived of functional diversity as indicative of 1160 C. POST Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. 36, 1153–1175 (2015) DOI: 10.1002/job variety and computed it with a Blau’s index (1977). Because functional area includes 10 categories and because teams in the sample differ in size (with some having fewer than 10 members), there is a risk of bias in the Blau’s index because “maximum possible variety increases with unit size” (Harrison & Klein, 2007: 12). Therefore, following established procedures, I standardized the Blau’s index by dividing it, for each team, by its theoretical maximum (Biemann & Kearney, 2010; Harrison & Klein, 2007). Among the teams in this study, the mean score of functional diversity was 0.69 with a range from 0 to 1. Team size Team size is the number of members on each team. To measure team size, I counted the number of members on team lists supplied by participating organizations. Geographic dispersion Geographic dispersion is a dummy variable that captures whether a team is located in one area or in multiple geographic areas. For each team, individual members responded to the question: “Which country best describes your current workplace location?” by selecting from a predetermined list of 193 countries, the one that best matched their current work location. When a team had members whose current workplaces were located in two or more countries, I coded that team as geographically dispersed. To ensure the reliability of this measure, I checked team members’ selfreported work location against information about geographic team dispersion provided by the participating organizations. The geographic dispersion dummy variable is equal to 1 for geographically dispersed teams and zero otherwise. Control variables All analyses controlled for team tenure (i.e., the average team member tenure) to account for positive performance outcomes associated with the length of time that members spend working together (Hackman, 2002). Team tenure was assessed by asking respondents to indicate how long they had been part of their team. Six response options were provided: “less than three months” (1), “three to six months” (2), “six to twelve months” (3), “one to two years” (4), “two to four years” (5), and “less than four years” (6). For each team, I averaged the tenure responses into a measure of team tenure. All analyses controlled for gender diversity using a Blau’s index (1977) to account for the previous findings that leader effectiveness may be contingent on team gender composition (Eagly et al., 1995). Because gender has only two categories and because all teams have at least two members, the Blau’s index for gender does not require standardization (Biemann & Kearney, 2010; Harrison & Klein, 2007). The maximum range for a Blau’s index when there are only two categories is from 0 to 0.5 (Harrison & Klein, 2007). Among the teams in this study, the mean score of gender diversity was 0.26 with a range from 0 to 1. Results The descriptive team-level statistics and inter-correlations presented in Table 1 indicates that teams led by women are larger (r = .24, p<.01) and more gender-diverse (r = .39, p<.01). Teams with longer tenure tend to be more cohesive (r = .29, p<.01). More functionally, diverse teams tend to be less cohesive (r =.23, p<.01). Team cohesion is positively associated with evaluations of cooperative learning (r = .55, p<.01) and of participative communication (r = .62, p<.01); and cooperative learning and participative communication are also positively associated (r = .74, p<.01). Because, in several cases, multiple teams from a single organization participated in the study, I computed intraclass correlations to evaluate the extent to which teams are more similar within, as compared with across, organizations (Snijders & Bosker, 2012). Organizations accounted for 9.4 percent of inter-team differences in cohesion and for 7.3% of inter-team differences in participative communication and did not seem to account for variation in cooperative learning across teams. When organizations account for non-trivial variation in team differences (as they WHEN IS FEMALE LEADERSHIP AN ADVANTAGE? 1161 Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. 36, 1153–1175 (2015) DOI: 10.1002/job do here for two of the dependent variables), “the hierarchical linear model is a better method of analysis than OLS regression analysis because the standard errors of the estimated coefficients produced by ordinary regression analysis are not to be trusted” (Snijders & Bosker, 2012: 52). I, therefore, tested all the hypotheses with hierarchical linear modeling (HLM) (Raudenbush, Bryk, & Congdon, 2000). HLM allows proper modeling of non-independent data by partitioning variability in the dependent variables between the organizational and team-level variables. Because the team-level phenomenon is the focus of this study, a random intercept model with level-1 covariates only was constructed for each dependent variable, with team-level variables centered around the grand mean (Raudenbush & Bryk, 2002; Snijders & Bosker, 2012). For each dependent variable, the main effect model includes the controls and the main effects of team tenure, gender diversity, functional diversity, team size, geographic dispersion, and leader gender. Models 1, 2, and 3, respectively, test the hypothesized interaction effects of gender with functional diversity, team size, and geographic dispersion. Because the relatively small sample size limits the statistical power in the analyses, I tested each interaction effect with separate HLM models (Aguinis, Beaty, Boik, & Pearce, 2005). Finally, to ensure that the interactions I observed were not spurious, I included the three interaction terms together in model 4. To test the significance of the moderator effect, I computed the difference between the deviance score of the main effect model (without the interaction term) and the deviance score of the model with the interaction term. The deviance score difference is a test statistic following a chi-squared distribution with a single degree of freedom (Snijders & Bosker, 2012). A statistically significant deviance score indicates that the slope difference between men and women is significantly different, from a statistical perspective. The proportional reduction in mean squared predictor error provides a meaningful way to express R2 for multilevel models (Snijders & Bosker, 2012). Following Snijders and Bosker (2012), I calculated this pseudo level-1 (i.e., team-level) R2 for the main effect models by calculating the proportional reduction in the value of the level 1 (team-level) and level 2 (organization-level) variance components after introducing the control and main effects, compared with the null model. For each interaction model, I calculated the pseudo level-1 R2 by calculating the proportional reduction in the value of the level 1 and level 2 variance components after introducing the interaction term, compared with the null model. The HLM results are summarized by dependent variable in Table 2 for cohesion, in Table 3 for cooperative learning, and in Table 4 for participative communication. Hypothesis 1 predicted that, as functional diversity increases, female-led teams report more (H1a) cohesion, (H1b) cooperative learning, and (H1c) participative communication than teams led by men. The results suggest that as functional diversity increases, female-led teams report more cohesion than male-led teams (? = 1.79, p = .001), as shown in Table 2, model 1, thus supporting Hypothesis 1a. To assess the robustness of these results, I conducted the HLM analysis with the reduced set of teams for which the team rwg(J) for cohesion, cooperative learning, and participative communication were above 0.70. The robustness analyses show a similar pattern of results, with female-led teams reporting more cohesion than male-led teams (? = 1.68, p = .01) as Table 1. Means, standard deviations, and bi-variate correlations among team-level variables. Mean SD (1) (2) (3) (4) (5) (6) (7) (8) (1) Team tenure 3.83 0.864 (2) Gender diversity 0.29 0.180 .06 (3) Functional diversity 0.69 0.209 .11 .02 (4) Team size 10.82 6.388 .03 .21 .04 (5) Geographic dispersion 0.27 0.445 .14 .04 .07 .07 (6) Leader gender (1 = female) 0.30 0.463 .13 .39** .11 .24* .08 (7) Cohesion 5.71 0.494 .29** .13 .23* .01 .04 .04 (8) Cooperative learning 4.98 0.486 .03 .00 .20 .10 .02 .06 .55** (9) Participative communication 5.55 0.448 .18 .04 .20 .20 .04 .01 .62** .74** *Correlation is significant at the 0.05 level (two-tailed). **Correlation is significant at the 0.01 level (two-tailed). 1162 C. POST Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. 36, 1153–1175 (2015) DOI: 10.1002/job functional diversity increases. The results fail to support Hypothesis 1b and c, as the coefficient correlates for the interaction terms between leader gender and team size are not statistically significant in explaining team variance in cooperative learning (? =.09, p>.10), and cooperative participation (? = 0.23, p>.10) as shown in Table 3 (model 1) and Table 4 (model 1), respectively. To further examine the interaction between leader gender and functional diversity as it relates to cohesion, I plotted the relationship between leader gender and cohesion at high and low values of functional diversity (one standard Table 2. Random intercept model of cohesion, with level-1 covariates. Main effect model Model 1 Model 2 Model 3 Model 4 ßs ßs ßs ßs ßs Intercept 5.73*** 5.71*** 5.71*** 5.74*** 5.71*** Team tenure 0.18*** 0.17*** 0.17** 0.18*** 0.17† Gender diversity 0.34 0.20 0.43 0.39 0.33 Functional diversity 0.43† 0.76*** 0.45† 0.41 0.77*** Team size 0.00 0.00 0.02† 0.00 0.02† Geographic dispersion 0.10 0.08 0.09 0.17 0.13 Leader gender (1 = female) 0.07 0.04 0.05 0.00 0.04 Leader gender * functional diversity 1.79*** 1.82*** Leader gender * team size 0.03* 0.03** Leader gender * geographic dispersion 0.22 0.18 ?2 25.0 28.2 23.9 23.6 25.8 Model deviance 100.9 92.4 98.1 100.0 88.1 Decrease in deviance 13.8* a 8.5** b 2.8† b 0.9b 12.8** b Pseudo level-1 R2 0.16 0.10 0.03 0.01 0.14 Note: for teams, n = 82, for organizations, n = 29. Entries corresponding to the predicting variables are estimations of fixed effects. ***p<0.001. **p<.01. *p<.05. †p<0.1. aDecrease in deviance in comparison to null model. bDecrease in deviance in comparison to main effect model. Table 3. Random intercept model of cooperative learning, with level-1 covariates. Main effect model Model 1 Model 2 Model 3 Model 4 ßs p-value ßs p-value ßs p-value ßs p-value ßs p-value Intercept 4.94*** 4.94*** 4.92*** 4.98*** 4.96*** Team tenure 0.01 0.01 0.01 0.01 0.01 Gender diversity 0.04 0.04 0.08 0.06 0.18 Functional diversity 0.50† 0.49† 0.53* 0.47 0.49* Team size 0.01* 0.01* 0.03** 0.01* 0.03** Geographic dispersion 0.01 0.01 0.01 0.16 0.14 Leader gender (1 = female) 0.13 0.13 0.10 0.02 0.04 Leader gender * functional diversity 0.09 0.06 Leader gender * team size 0.04* 0.04* Leader gender * geographic dispersion 0.44* 0.42* ?2 20.1 20.1 19.0 19.2 16.7 Model deviance 108.6 108.6 103.7 105.6 100.7 Decrease in deviance 5.2a 0.0b 4.9* b 3.0† b 7.9* b Pseudo level-1 R2 0.06 0.00 0.06 0.04 0.09 Note: for teams, n = 82, for organizations, n = 29. Entries corresponding to the predicting variables are estimations of fixed effects. ***p<0.001. **p<.01. *p<.05. †p<0.1. aDecrease in deviance in comparison to null model. bDecrease in deviance in comparison to main effect model. WHEN IS FEMALE LEADERSHIP AN ADVANTAGE? 1163 Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. 36, 1153–1175 (2015) DOI: 10.1002/job deviation above and below the mean), following the procedures described by Preacher and colleagues (2003). As shown in Figure 2, among more functionally diverse teams (as compared with more homogeneous teams), those led by women report more cohesion than those led by men. Hypothesis 2 predicted that as team size increases, female-led teams report more (H2a) cohesion, (H2b) cooperative learning, and (H2c) participative communication than teams led by men. Supporting Hypothesis 2a, b, and c, the results indicate that among larger teams (as compared with smaller ones), women-led teams report more cohesion (? = .03, p = .05), more cooperative learning (? = .04, p = .05), and more participative communication (? = .03, p = .05) than men-led teams, as shown in the model 2 columns of Tables 2–4, respectively. To evaluate the robustness of these results, I conducted the HLM analysis with the reduced set of teams for which the rwg(J) for cohesion, cooperative learning, and participative communication were above 0.70. The robustness analyses show a similar pattern of results, with female-led teams reporting more cohesion (? = .03, p = .01), cooperative learning (? = .05, p = .01), and participative communication (? = .04, p = .001) than male-led teams as team size increases, thus strengthening the support for Hypothesis 2a, b, and c. To further examine the interaction between leader Table 4. Random intercept model of participative communication, with level-1 covariates. Main effect model Model 1 Model 2 Model 3 Model 4 ßs ßs ßs ßs ßs Intercept 5.54 *** 5.54*** 5.52*** 5.60*** 5.58*** Team tenure 0.08 0.08 0.08 0.09† 0.09† Gender diversity 0.10 0.12 0.02 0.10 0.15 Functional diversity 0.46** 0.51** 0.47** 0.35* 0.40* Team size 0.02** 0.02* 0.03*** 0.01* 0.03** Geographic dispersion 0.06 0.05 0.05 0.26* 0.25† Leader gender (1 = female) 0.07 0.07 0.05 0.12 0.14 Leader gender * functional diversity 0.23 0.18 Leader gender * team size 0.03* 0.02* Leader gender * geographic dispersion 0.63** 0.62** ?2 36.0 36.6 37.5 27.3 26.9 Model deviance 90.3 90.1 87.5 82.9 80.5 Decrease in deviance 10.4† a 0.2b 2.8† b 7.4** b 9.8* b Pseudo level-1 R2 0.11 0.01 0.03 0.10 0.12 Note: for teams, n = 82, for organizations, n = 29. Entries corresponding to the predicting variables are estimations of fixed effects. ***p<0.001. **p<.01. *p<.05. †p<0.1. aDecrease in deviance in comparison to null model. bDecrease in deviance in comparison to main effect model. Figure 2. Leader gender differences in mean team cohesion as function of team functional diversity 1164 C. POST Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. 36, 1153–1175 (2015) DOI: 10.1002/job gender and team size, I plotted the relationships between leader gender and cohesion, cooperative learning, and participative communication at high and low values of team size (Preacher, Curran, & Bauer, 2003). As shown in Figure 3, on large teams (as compared with small teams) women-led teams report more cohesion (panel A), cooperative learning (panel B), and participative communication (panel C) than men-led teams. Hypothesis 3 predicted that among geographically dispersed teams (as compared with co-located teams) femaleled teams report more (H3a) cohesion, (H3b) cooperative learning, and (H3c) participative communication than teams led by men. The results in Table 3 (model 3) and Table 4 (model 3) show that among geographically dispersed Figure 3. Leader gender differences in team outcomes as function of team size. Male and female means for cohesion (A), cooperative learning (B) and participative communication (C) WHEN IS FEMALE LEADERSHIP AN ADVANTAGE? 1165 Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. 36, 1153–1175 (2015) DOI: 10.1002/job teams, those led by women report more cooperative learning (? = .44, p = .05) and more participative communication (? = .63, p = .01), supporting Hypothesis 3b and 3c. The results fail to support Hypothesis 3a, as the coefficient correlate for the interaction between team geographic dispersion and female leadership did not explain a statistically significant amount of variance in cohesion across teams (? = .22, p>.10). To assess the robustness of these results, I conducted the HLM analysis with the reduced set of teams for which the rwg(J) for cohesion, cooperative learning, and participative communication were above 0.70. The robustness analyses show a similar pattern of results, with female-led teams reporting more cooperative learning (? =.51, p= .01) and participative communication (? =.64, p= .01) than male-led teams among geographically dispersed teams. To further examine the interaction between leader gender and team geographic dispersion, I plotted the means, for female-led and male-led teams, in cooperative learning and in participative communication for collocated and geographically dispersed teams (Preacher et al., 2003). As shown in Figure 4, among dispersed teams, those led by women teams report more cooperative learning (panel A) and more participative communication (panel B) than those led by men. Discussion, Limitations, and Implications The main purpose of this paper was to understand to what extent and in what contexts female leadership may be advantageous for teams. This study contributes to the debate on the female advantage in leadership, not only Figure 4. Leader gender differences in team outcomes as function of team geographic dispersion. Male and female means for cooperative learning (A) and participative communication (B) 1166 C. POST Copyright © 2015 John Wiley & Sons, Ltd. J. Organiz. Behav. 36, 1153–1175 (2015) DOI: 10.1002/job by identifying characteristics of teams (functional diversity, size, and geographic dispersion) for which the relationship between female leadership and team outcomes is more positive but also by examining the quality of the relationships between team members and their team and team interaction norms (rather than team performance) as outcomes of leader gender. Drawing on the research on gender differences in relational self- construal (Cross & Madson, 1997; Eagly & Wood, 1999; Eagly et al., 2000), I proposed that leader gender interacts with team coordination requirements (i.e., functional diversity, size, and geographic dispersion) to affect the quality of the relationship between individual team members and their team and team interaction norms. I argued that, as team coordination requirements increase (e.g., with functional diversity, team size, and when members are geographically dispersed) teams led by women experience more cohesion and report more cooperative learning and participative communication than those led by men. I reasoned that female leaders, because they are more likely than male leaders, on average, to have a relational self-construal (Cross & Madson, 1997; Eagly & Wood, 1999; Eagly et al., 2000), are therefore more likely to successfully attend to the coordination challenges presented by more functionally diverse teams, larger teams, and teams with geographically dispersed members. Results from the analysis of 82 innovation teams provide support for this contextual model of a female leadership advantage: they show that (i) as functional diversity increases, teams led by women report more cohesion compared with similar teams led by men; (ii) as team size increases, female-led teams report more cohesion, cooperative learning, and participative learning as compared with similar teams led by men; and (iii) geographically dispersed teams led by women report more cooperative learning and participative communication than similarly dispersed teams led by men.