Exploring the Asymmetrical Effects of Gender Tokenism on Supervisor–Subordinate Relationships

Abstract

Drawing from social identity theory, this research examines scarce gender representation as a contextual condition that inhibits same-gender supervisors' support. Survey results in Study 1 found that when women were proportionally underrepresented, they reported feeling less supported by female supervisors than male supervisors. Study 2 showed that women who perceived they were gender tokens in their organization were less likely to support an outstanding female subordinate than an identical male. Study 3 experimentally tested social mobility as a mechanism for the effects of tokenism on same-gender supervisor support. Results suggest that social mobility and group composition jointly affect ratings of same-gender targets. Perceptions of gender-based social mobility appear to be one mechanism through which tokenism influences same-gender relations at work.

Now, more than ever, women are advancing into a greater number of leadership roles (Bureau of Labor Statistics, 2009; Eagly & Carli, 2007). Despite the rising tide of equality for women in the workplace, obstacles still hinder their ascent. Interestingly, some of the forces impeding women's advancement are unintuitive and paradoxical. Anecdotal evidence (Dellasega, 2005; Heim, Murphy, & Golant, 2001), along with some emerging empirical work (Parks-Stamm, Heilman, & Hearns, 2008; Rosette, Tost, Hernandez, & Sitkin, 2007), has suggested that one surprising challenge facing women in the workplace is other successful women.

The notion that successful women would somehow hinder their successors seems both counterintuitive and counterproductive to the advancement of the female gender in the context of work. Not only does research show that people generally tend to favor others in their same demographic group (Brewer, 2007; Dienesch & Liden, 1986; Kraiger & Ford, 1985; Tsui & O'Reilly, 1989; Turban & Jones, 1988), it stands to reason that historically disadvantaged minority groups would be supportive of their in-group members as they struggle against the obstacles imposed on them by society (i.e., majority groups). However, anecdotal evidence, like that described in a New York Times article, points out that women can be “their own worst enemies at work” (Klaus, 2009, p. BU2). Why might a woman fail to favor her fellow females, actually withholding support from other women? Moreover, why, as popular culture and anecdotal evidence suggest, would this phenomenon occur among women, but not among men? The purpose of this paper is to explore scarce gender representation as a contextual condition under which this phenomenon emerges among women.

Background

Researchers agree that women have more difficulty reaching senior levels despite having similar educational levels, years of service, and job performance (Nelson & Burke, 2002; Powell, 1999; Tharenou, 1999). The Bureau of Labor Statistics (2009) reported that women occupy 51% of management, professional, and related occupations, yet only 14.8% of the nation's 500 largest companies have female corporate officers (Catalyst, 2007). Social identity and tokenism theories point to a combination of scarce proportional representation and lower societal status of women, which creates a potent social context for women.

Female Tokenism: By the Numbers

Organizational demography (Pfeffer, 1983) and tokenism (Kanter, 1977) theories point to the critical nature of demographic representation in the work setting (Vecchio & Bullis, 2001). Underrepresentation has several key effects on how minority-group members experience a given social context: higher visibility (and thus increased scrutiny), exaggeration of differences from majority-group members, and the assimilation of personal characteristics to fit into existing stereotypes. Kanter's study also revealed an unanticipated consequence of tokenism: strained relations among the women themselves. Whereas in-group favoritism (Hewstone, Rubin, & Wills, 2002) predicts that women should be more positive toward other members of their gender group, Kanter observed that token women attempted to dissociate from the collective group of female tokens and experienced strained in-group relations. In other words, proportional gender representation within an organization may have unexpected consequences in terms of same-gender relations, particularly for women.

Female Tokenism: Low Gender Status

Empirical research has shown that simple numerical underrepresentation (i.e., tokenism “by the numbers”) is not sufficient for triggering the adverse experiences of tokenism, but rather involves the overall social context in which underrepresentation occurs. A number of authors have agreed that social status significantly moderates the effects of tokenism (McDonald, Toussaint, & Schweiger, 2004; Sekaquaptewa & Thompson, 2003; Williams, 1992; Yoder, 1991; Yoder, Schleicher, & McDonald, 1998). When men enter into a predominantly female field or organization, they do so with the privilege associated with their dominant social status (Williams, 1992). Empirical tests have supported the fact that token men do not share the negative experiences of token women (Budig, 2002; Cognard-Black, 2004; Fairhurst & Snavely, 1983; Johnson & Schulman, 1989; Kadushin, 1976; Ott, 1989; Simpson, 2004; Unger, 1978; Wagner, 1988). Moreover, research has pointed to instances in which scarcity is actually advantageous for upper status tokens, such as men (Crocker & McGraw, 1984; Grimm & Stern, 1974; Heikes, 1991; Yoder & Sinnett, 1985; Williams, 1992). Women are subject to negative consequences of tokenism by virtue of their double deviance: They are members of a group that is both proportionally scarce and of lower status (Laws, 1975).

It may also be the case that a skewed gender context activates status-related gender stereotypes, which are disadvantageous to women, especially those in leadership positions. Inherently status-oriented gender-role stereotypes lead a man alone in a workgroup composed of women to be perceived as a leader; thus, he may be more influential than the women in his group, despite their numerical advantage (Craig & Sherif, 1986; Crocker & McGraw, 1984; Hewstone et al., 2006; Williams, 1992). In other words, men are evaluated favorably when they represent the minority because a positive gender stereotype is activated (Fuegen & Biernat, 2002). Their stereotypical social status then becomes an advantage, despite the fact that they are outnumbered (Ott, 1989; Sekaquaptewa & Thompson, 2002), whereas this effect is inverted for women when they enter a predominantly male field.

Moreover, tokenism findings with racial minorities mirror those with female tokens. The results support the key role of status in experiences of tokenism and, again, illustrate negative in-group relations experienced by token individuals who are doubly deviant, or contrast with the numerical majority and the higher status group (Goldberg, 2005; Laws, 1975; Niemann & Dovidio, 1998; Pollak & Niemann, 1998; Sackett, DuBois, & Noe, 1991; Sekaquaptewa, Waldman, & Thompson, 2007; Yoder, Aniakudo, & Berendsen, 1996). Regardless of the specific group, the experiences of lower status group members with one another in environments where they are underrepresented have the potential to result in substantially less positive interactions than those between higher status group members in similar contexts (Blanz, Mummendey, Mielke, & Klink, 1998; Chattopadhyay, George, & Lawrence, 2004; Goldberg, Riordan, & Schaffer, 2003; Goldberg, Riordan, & Zhang, 2008).

In this paper, we argue that the doubly deviant status of female tokens (i.e., low proportional representation and low societal status) serves to explain why women, but not men, might experience strained same-gender relations in token contexts. Using social identity theory, we also explore the motivation driving such strained same-gender relations among women. It is our contention that their lack of support for other in-group members is a manifestation of women's desire to distance themselves from a negatively valued in-group, thereby enhancing their own social identity, but at the expense of their fellow in-group members. We also posit that this individualistic social identity enhancement strategy is particularly likely to emerge in token contexts because of the limited and ambiguous nature of professional opportunities for members of token groups.

Social Identity Enhancement: In-Group Relations in the Context of Tokenism

In her foundational study, Kanter (1977) observed that token saleswomen attempted to dissociate or differentiate themselves from the collective group of female tokens. Ellemers, van den Heuvel, de Gilder, Maass, and Bonvini (2004) found that women in a field where they are historically underrepresented (i.e., academic positions in scientific fields) were more likely than were men to engage in gender stereotyping in order to contrast themselves with the rest of their lower status gender. In other words, evidence has suggested that women's strained relations with each other stem from a desire to escape the collective fate of their gender.

Social identity theory is founded on the notion that individuals use group memberships to positively enhance their own sense of self (Tajfel & Turner, 1979). However, when those groups are negatively distinctive (e.g., lower status), group membership hurts rather than helps one's positive sense of self, and people become motivated to improve their social identity (Ellemers, Spears, & Doosje, 2002). In order to achieve this goal, members of lower status groups can strive to improve their social identity by using one of three strategies: social mobility, social creativity, and social change (Tajfel & Turner, 1979; van Knippenberg, 1989). It is the first strategy—that is, social mobility—that would serve to explain the finding that doubly deviant tokens (i.e., women) experience negative in-group relations. A context of scarce representation encourages these double-deviants to use an individualistic social identity enhancement strategy and actively dissociate from members of their own in-group. Notably, men are not members of a historically lower status group; therefore, they should be less motivated to withhold support from other men in the context of underrepresentation.

A social mobility strategy is an individual strategy (indicated by an individual's attempt to leave or dissociate himself or herself from the in-group) for improving one's social identity; whereas collective strategies, such as social creativity and social change, are intended to enhance the status of the collective in-group as a whole (Jackson, Sullivan, Harnish, & Hodge, 1996). Social identity research has generally revealed that social mobility (the individual strategy) is the dominant strategy for achieving positive social identity (Tajfel & Turner, 1979; Taylor, Moghaddam, Gamble, & Zellerer, 1987; van Knippenberg, 1984; Wright, Taylor, & Moghaddam, 1990), and that individuals will first engage in social mobility strategies before resorting to collective strategies (Taylor & McKirnan, 1984).

We posit that tokenism makes an individualistic social-identity-enhancement strategy especially likely because advancement opportunities for token in-group members are limited and ambiguous. The individualistic strategy is more likely because of the asymmetrical limitations for collective enhancement that are inherent in such a gender-skewed social context. The existence of a few successful tokens can be taken as a symbol of merit-based opportunity (Taylor & McKirnan, 1987), encouraging female tokens to distinguish themselves from other in-group members (i.e., females; Chattopadhyay, 2003; Chattopadhyay et al., 2004; Williams & Giles, 1978).

Under conditions of rare proportional representation, only some token individuals are allowed entry into the higher status group, causing people to pursue individual advancement, even when they know that the possibility of access to a higher status position is very low (Wright, 1997, 2001; Wright & Taylor, 1998). Given such limited advancement opportunities, token female supervisors may dissociate from their same-gender subordinates (Chattopadhyay, 2003; Chattopadhyay et al., 2004; Ely, 1994), rather than investing in the advancement of the collective and supporting their female subordinates. In other words, to be viewed as distinct and especially worthy of inclusion into the dominant male group, a token female leader may be motivated to withhold support from members of her same gender to achieve maximum differentiation between her own exceptional status and the lowly status of the token group. Men, however, would be less likely to experience this same motivation to dissociate from their male in-group because, as mentioned previously, they experience gender tokenism as an advantage (e.g., Crocker & McGraw, 1984; Williams, 1992). Scarce representation and the lower status of the female gender (i.e., their double deviance) combine to create an individualistic token context that encourages women to dissociate from members of their own gender.

Building on Prior Research: Focus on Supervisor Support

Unfortunately, only limited existing research has directly tested the notion that women who work in token contexts may experience negative relationships with other women. One notable exception by Ely (1994) explored the influence of proportional representation of women in high-status positions (partners in law firms) on same-gender (female–female) relations within organizations. Ely found that “compared with women in firms with many senior women, women in firms with few senior women were less likely to experience common gender as a positive basis for identification with women” (p. 203). She also examined the attitudes of subordinate women toward more senior women, finding that when few females were in senior positions, women were less likely to perceive them as role models with legitimate authority.

In the current study, we build on Ely's (1994) findings in two ways. First, we propose that women do not simply fail to use their gender as a positive basis for social identity, but, rather, subtly dissociate themselves from other females by failing to support members of their in-group. Second, we build on Ely's illustration of a lack of support within token female peer relationships and investigate the effect of tokenism on how female superiors act with respect to supporting their same-gender subordinates. Presumably, if women in positions of power (e.g., supervisors) do not support women under them, the professional consequences for female subordinates would be far greater than if the reverse were true (i.e., subordinates who have negative attitudes toward supervisors; Ely, 1994). This is because excelling in an organization and thus ascending to higher levels of leadership is already difficult for many women to achieve. Combining this with a lack of supervisor support can cause the situation to become even bleaker for women's advancement.

Indeed, organizational research has repeatedly illustrated the beneficial effects of social support, especially from supervisors, on worker performance and well-being (Ganster, Fusilier, & Mayes, 1986; Karasek, Triantis, & Chaudhry, 1982). Supervisor support has been empirically linked to a wide range of meaningful individual- and organizational-level outcomes, such as stress, job satisfaction, turnover, and productivity (Baruch-Feldman, Brondolo, Ben-Dayan, & Schwartz, 2002; Viswesvaran, Sanchez, & Fisher, 1999). As discussed previously, women are still a marginalized group in higher levels of organizations; and it may be that the career and psychosocial benefits that women receive from supportive supervisors are especially important.

Our first study captures the perspective of the subordinate. Utilizing archival data from a nationally representative survey, we tested the hypothesis that, from the subordinate's perspective, female supervisors would be less likely than male supervisors to support same-gender subordinates under conditions of tokenism. In Study 2, we investigated this phenomenon from the perspective of the supervisor. We hypothesize that when supervisors hold token status, female (but not male) supervisors will be less likely to provide social support to same-gender subordinates. We tested this prediction with a quasi-experiment where working adult participants assumed the role of “supervisor” and made assessments regarding a fictitious job applicant. In Study 3, we examined social mobility as the explanatory mechanism. We hypothesize that under impermeable boundary conditions, female tokens will withhold support from subordinate women as an attempt to improve their individual status.

Study 1

Purpose

The purpose of the first investigation is to examine the relationship between proportional gender representation and supervisor support. More specifically, our aim is to explore the expectation that under conditions of gender tokenism in real-world settings, women (but not men) will be less supportive of same-gender subordinates. In accordance with tokenism theory, we hypothesize that token female subordinates of women supervisors will report receiving less supervisor support than will non-token female subordinates with women supervisors. In other words, women will be less likely to feel supported by their female supervisors when they work in an organizational context of gender underrepresentation (i.e., gender tokenism). Since token status is not combined with lower gender status for males, we do not expect that gender token status will elicit a significant difference in same-gender supervisor support for men.

• Hypothesis 1. Female subordinates of token women supervisors will receive less supervisor support than will female subordinates with non-token women supervisors.

Method

Participants and Procedure

Study 1 participants included 3,504 working adults from the 2002 National Study of the Changing Workforce (NSCW), which is a privately supported, nationally representative survey. Using a computer-assisted telephone interviewing system by Harris Interactive, the participants completed a questionnaire that was developed by the Families and Work Institute. Interviews were conducted between October 2002 and June 2003. Participants were restricted to individuals over 18 years of age and those currently working as paid members of the continental U.S. workforce. Eligible participants were offered $25 to complete the questionnaire. The overall response rate was approximately 52%, after accounting for ineligible households.

The sample was 53.5% male and 46.5% female. Participants' mean age was 42.3 years (SD = 13.3). With regard to ethnicity, the participants were 78.6% Caucasian; 9.7% African American; 1.5% Asian; 1% Native American; and 8.1% who self-identified as “other.” The majority of participants were married and living with their spouses. The mean number of children under 18 years and living in the home was 1.68 (SD = 1.51).

Measures

Gender

The participants indicated their own gender (1 = male, 2 = female). They also indicated the gender of their supervisor.

Token status

Participants were asked to indicate the percentage of their coworkers that were of the same gender. This was measured on the following 6-point scale: 6 = 100% of coworkers; 5 = 75%–99%; 4 = 50%–74%; 3 = 25%–49%; 2 = less than 25%, but more than 0; or 1 = 0%. The scales were then dichotomized to reflect scarce gender representation or tokens (< 25%) and non-tokens (> 25%).

We point to both Kanter's (1977) original theory and definitions, as well as recent findings targeting numerical cutoffs versus subjective-perceptions tokenism (King, Hebl, George, & Matusik, 2010) to support the appropriateness of our measurement strategies. While Kanter originally proposed that the skewed groups in which tokens reside are “those in which there is a large preponderance of one type over another, up to a ratio of perhaps 85:15” (Kanter, 1977, p. 966, emphasis added), subsequent research has suggested that tokenism is a categorical experience of severe underrepresentation of one's gender, rather than strict numerical representation. Note that in its original form, Kanter's 15% definition suggests an approximate cutoff point under which underrepresentation represents the categorical experience of tokenism. It is also worth noting that according to Kanter's original paper, not until the ratio of dominants to tokens gets to “perhaps 65:35” (p. 966, emphasis added) does the distribution become less extreme and lead to less exaggerated effects where dominants are just a majority and tokens are just a minority. We argue that when using 25% as the boundary point under which the tokenism category is indicated, exaggerated effects of severe underrepresentation remain. Although past research has used 15% representation as the boundary point under which tokenism is indicated, the current study relies on a cutoff value of 25% or less, which is supported in both Kanter's original theory and in recent research.

Moreover, in support of this measurement strategy, recent research measuring gender representation along a continuum has found that the relationship between gender representation and female managers' perceptions of inequity exponentially decreased, such that the relationship was strongest only at the lowest levels of representation. Furthermore, in their explicit comparison, tokenism represented as a token/non-token dichotomy accounted for significantly more variance than did tokenism represented as a continuous variable (see King et al., 2010, Study 2, p 495). We hold that 25% still represents a severe enough underrepresentation to create a context that is skewed (as opposed to “tilted”), wherein members of underrepresented groups are treated as symbols of their category, experience increased scrutiny and social isolation, and find it difficult to generate alliances as a result of strained in-group relations (as opposed to simply members of a minority group).

The use of coworkers as the reference group in Study 1 served to capture gender salience in respondents' immediate work context. Presumably, the immediate work context reflects the domain in which the supervisor and subordinate interact most frequently. Asking participants to report on the gender of their coworkers represents an assessment of the most proximal environmental conditions for influencing behavior at the individual (interpersonal) level. Although recent research has shown that the two do overlap (King et al., 2010), this is not to say that gender tokenism among coworkers must necessarily relate to token status in the organization as a whole. A female accountant, for instance, could feel like a token in her immediate work context (i.e., the accounting division), even though her company as a whole employs predominantly females. In our example, we predict that this female will be less likely to support a female subordinate in the accounting department (her immediate context).

Supervisor support

We assessed supervisor support with four items, using a 4-point Likert-type scale ranging from 1 (strongly disagree) to 4 (strongly agree). A factor analysis on these items suggests that they comprise a single factor (eigenvalue = 2.50; 62.6% of the variance). The items include “My supervisor keeps me informed of things I need to do my job well” (factor loading =.85); “My supervisor has realistic expectations of my job performance” (factor loading =.82); “My supervisor recognizes when I do a good job” (factor loading =.76); and “My supervisor is supportive when I have a work problem” (factor loading =.72). Coefficient alpha for this measure was.80.

Controls

Respondent characteristics were measured and controlled for in the analyses. These characteristics include race of the respondent (1 = White, 0 = other) and whether the supervisor is of the same racial–ethnic background (1 = Yes, 2 = No). We controlled for race of the respondent because, as Yoder and colleagues have illustrated (Yoder, 2002; Yoder & Aniakudo, 1997; Yoder et al., 1996; Yoder & Berendsen, 2001; Yoder & Kahn, 1993), gender and race may interact in their effects on the social atmosphere and colleagueship among tokens. Moreover, we controlled for subordinate race and supervisor–subordinate racial match in accordance with research regarding ethnic minorities and support perceptions (Ensher & Murphy, 1997). It should be noted that the pattern of findings reported in the Results section is consistent with or without these controls.

Results and Discussion

Descriptive statistics, correlations, and reliability estimates are presented in Table 1. The only significant correlation was between respondent gender and supervisor gender (r =.39, p <.01). The correlation between respondent gender and composition of the workforce was not significant, but was in the expected negative direction (r = −.03, p >.05).

Table 1. Intercorrelations and Descriptive Statistics for Study Variables: Study 1

Variable	M	SD	1	2	3
Note. Gender: 1 = male, 2 = female. **p <.01.
1. Gender	1.47	0.50	—
2. Supervisor gender	1.37	0.48	.39**	—
3. Composition of workforce	0.86	0.34	−.03	−.02	—
4. Supervisor job-related support	3.49	0.60	.03	.03	.01

We conducted a 2 (Participant Gender) × 2 (Supervisor Gender) × 2 (Token Status: token vs. non-token) ANCOVA on participant ratings of supervisor support. For this analysis, work composition was collapsed into two categories, reflecting token and non-token status (see Table 2). Additionally, Table 3 presents the results of the ANOVA.

Table 2. Cell Descriptive Statistics: Study 1

	Male supervisor						Female supervisor
	Token			Non-token			Token			Non-token
	M	SD	N	M	SD	N	M	SD	N	M	SD	N
Male participant	3.42	0.56	48	3.46	0.60	988	3.50	0.63	47	3.47	0.56	190
Female participant	3.55	0.54	77	3.49	0.66	462	3.27	0.55	39	3.53	0.52	659

Table 3. ANCOVA for Supervisor Support: Study 1

Source of variation	SS	df	MS	F	Partial η^2
Note. SS = sum of squares; MS = mean square. Overall R^2 =.01. †p <.10. *p <.05.
Control variables
Supervisor race	1.15	1	1.12	3.04	.001
Participant race	0.24	1	0.24	0.65	.00
Main effects
Participant gender	0.00	1	0.00	0.01	.00
Supervisor gender	0.24	1	0.24	0.65	.00
Tokenism	0.53	1	0.53	1.45	.00
Two-way interactions
Participant Gender × Supervisor Gender	1.28	1	1.28	3.50†	.00
Participant Gender × Tokenism	0.46	1	0.46	1.25	.00
Supervisor Gender × Tokenism	0.74	1	0.74	2.02	.00
Three-way interactions
Participant Gender × Supervisor Gender × Tokenism	1.93	1	1.93	5.25*	.01

The main effects were nonsignificant for participant gender, F(1, 2772) = 0.01, p >.05; supervisor gender, F(1, 2772) = 0.65, p >.05; and token status, F(1, 2772) = 1.45, p >.05. The interaction effects between participant gender and token status, F(1, 2772) = 1.25, p >.05; and between supervisor gender and token status, F(1, 2772) = 2.02, p >.05, were also nonsignificant. While the two-way interaction between participant gender and supervisor gender was in the expected direction and approached significance, F(1, 2772) = 3.50, p <.10, this was qualified by a significant three-way interaction, F(1, 2772) = 5.25, p <.05, indicating that the relationship between participant gender and supervisor support varied, depending on both the gender of the supervisor and the composition of the workforce (see Figures 1 and 2). This significant three-way interaction was probed by conducting separate ANOVAs on male and female participants. These analyses yielded a significant interaction between token status and supervisor gender for female participants F(1, 1498) = 8.75, p <.01, but not for male participants (p >.10).

Figure 1. Interaction of supervisor gender and token status on ratings of supervisor support (as depicted on the y axis) for female participants.

Figure 2. Interaction of supervisor gender and token status on ratings of supervisor support (as depicted on the y axis) for male participants.

Subsequent t tests supported Hypothesis 1, revealing that women reported receiving less support from female supervisors when they were tokens (M = 3.30, SD =.60) than when they were not tokens (M = 3.54, SD =.69), t(696) = −2.33, p <.05. As expected, the same pattern did not emerge for men with male supervisors, t(1032) = −0.36, p >.00. For men, there was no significant difference between the support they received from male supervisors, regardless of whether or not they were tokens.

Overall, the results suggest that when women represent the minority in their working environment, they receive less support from female supervisors than do men or women working in a non-token environment. The results of Study 1 support our expectations and demonstrate that the context of rare proportional gender representation influences female managers' support of subordinate women. The findings of this study illustrate that women who work in an environment where females are extremely underrepresented reported feeling less supported by female managers. Gender token status did not influence men's reports of support from male supervisors. According to these findings, one explanation for why women would be less likely to support same-gender subordinates resides in the context in which evaluations are made. The token women who doubly deviated from the male-dominated context were less likely to support same-gender subordinates, whereas the same was not true for the men who are simply proportionally unrepresented (but not of lower gender status).

Study 2

Purpose

Since our first study's results suggest that women feel less supported by their female supervisors when females are proportionally underrepresented, we sought to replicate those findings in Study 2 by examining whether supervisors would be less likely to support same-gender targets under conditions of tokenism. First, since our focus is on the effects of gender composition (tokenism), we sought to control for the complex dynamics surrounding established relationships between supervisors and subordinates (Jeanquart-Barone, 1996; Turban & Jones, 1988; Wayne & Ferris, 1990; Xin, 2004) by asking participants to evaluate a fictional subordinate (as depicted in a résumé). Second, by asking for the supervisors' reports, we intended to clarify whether subordinates' perceptions of supervisor support (as measured in Study 1) represent genuine differences in the support women and men are likely to provide. Additionally, given that research shows that tokenism goes “beyond the numbers” to the subjective experience of being a token (King et al., 2010; Yoder, 1994, 2002; Yoder & Sinnett, 1985), we chose to use subjective perceptions of tokenism in our second study, rather than strict numerical representation.

We predict that, for women, the relationship between target gender and participants' likelihood of providing support will be moderated by perceptions of gender tokenism. In other words, as female supervisors' perceptions of tokenism increase, they will be less likely to support female than male subordinates. We do not expect perceived tokenism to moderate this relationship for men: Male supervisors' likelihood of supporting their subordinates will be the same, irrespective of whether they feel they are gender tokens at work.

• Hypothesis 2. Among female supervisors, tokenism will moderate the relationship between subordinate gender and supervisor support. The extent to which women experience tokenism will be positively related to their support for male subordinates, whereas there will be a negative relationship between tokenism and their support of female subordinates.

Method

Participants and Procedure

Working professional adults were recruited by eight undergraduate psychology majors (6 female; all Caucasian) who volunteered to assist with this study for research experience. In public locations in the business district of a large southern metropolis, these research assistants were told to ask men and women to participate in a 5-min study. The respondents were recruited to participate without incentive and completed the surveys in those same locations immediately upon agreeing to participate.

Out of the total of 116 people who were approached, 33 females and 42 males agreed to participate, reflecting a response rate of approximately 65%. Individuals who agreed to complete a short study on the spot were told to imagine that they were hiring a direct-report (who would work below them) in their organization and were presented with the “résumé” of this fictitious applicant. Résumés were developed by downloading several résumés from a résumé template software program. The outstanding nature of the résumés was evidenced by a high-quality educational background (a degree from Stanford), high GPA (4.0), and impressive professional experience (e.g., job title of “project manager”). The “male” and “female” versions of the paper-and-pencil survey (résumé evaluations) were combined in a nonsystematic fashion, and research assistants distributed whichever questionnaire was at the top of the stack of papers without knowing whether the target was male or female.

The study was a 2 (Participant Gender) × 2 (Target Gender) between-subjects design. The experience of tokenism was treated as an additional continuous independent variable. The researcher then presented the participant with the résumé of a fictitious applicant who had ostensibly graduated Phi Beta Kappa from Stanford, had a wide range of internship experience, and was seeking “to obtain an entry-level position with advancement potential requiring strong analytical and organizational skills.” The content of male (N = 34) and female (N = 41) applicants' résumés were identical, and the gender of the applicant was manipulated by using a feminine or masculine name (Katherine or Kevin, respectively) and the corresponding proper pronouns.

Participants rated their own self-esteem, and indicated the gender tokenism they had encountered in their organization. With regard to ethnicity, most participants were Caucasian (76.0%), while 11.0% were African American, 4.2% were Asian American, and 5.6% were Hispanic. The median age of participants was 33 years, and the range of self-reported job titles included, architect, attorney, controller, and store manager.

Measures

Social support for the target

Participants indicated the strength of their agreement with five items that were developed for the purpose of the present study to assess the degree to which participants were likely to support the applicant should the applicant be hired. The items were rated on a 7-point Likert-type scale ranging from 1 (strongly disagree) to 7 (strongly agree). Sample items include “I would show this applicant the ‘ins and outs’ of the company,” “I would help this applicant succeed in the company,” and “I would like to be a mentor to this applicant.” The internal consistency reliability of these items was.83, and a composite was created to represent social support.

Tokenism

Our Study 1 findings showed that being a token among one's coworkers elicits reversals of in-group favoritism among female supervisors. In the present study, we sought to build on those findings and asked participants to report their perceptions of gender tokenism within a larger context (their organization as a whole). We also chose to build on Study 1 by employing a subjective assessment of tokenism in Study 2, rather than simply depending on numerical reports of gender composition. As the recipients of perceived negative outcomes of tokenism, tokens may be more aware of this skewed social context than observers and dominants (Kanter, 1977). Research has suggested that token composition (i.e., numbers) alone is not enough to bring about the negative consequences associated with being a token (Yoder & Sinnett, 1985). Furthermore, recent findings (King et al., 2010) have illustrated that the relationship between token status and psychological climate of gender inequity is mediated by the subjective experience of tokenism. In other words, perceptions of tokenism are a key component to the effects of strictly numerical tokenism on the gendered context in which they work. Therefore, tokenism must be perceived (by the tokens themselves) in order to affect interpretations of and behaviors within the social context of their work environment.

Using a four-item scale, the participants reported the extent to which they felt that they had been a token in their organization because of their gender. Items were rated on a 7-point scale ranging from 1(strongly disagree) to 7 (strongly agree). The scale was derived from items used in a paper evaluating a range of token processes (Yoder, 1994) and has shown acceptable reliability (.70) in other studies of tokenism (King et al., 2010). The scale was not tailored to the specific gender of the participants, but, rather, required them to reference their own gender (male vs. female) when reporting their experiences as a token member of their gender.

A principal components factor analysis on these items yielded a single factor that accounted for the majority of the variance (eigenvalue = 1.87; 46.8% of variance explained). The items include “I feel that I am a ‘token’ representative of my gender in my company” (factor loading =.67); “I have to explain the perspective of my gender to others in my organization” (factor loading =.65); “People in my company look to me as a representative of my gender” (factor loading =.62); and “I feel I have to represent the perspective of my gender in my organization” (factor loading =.79). An additional factor analysis combining tokenism and evaluation items, which was conducted to ensure discriminant validity across constructs, supports a two-factor solution wherein the factors of Social Support and Subjective Tokenism emerged (eigenvalues > 1; cross-loadings <.40). The internal consistency reliability for this scale was.63. A composite was created to represent subjective tokenism.

Self-esteem

Parks-Stamm et al. (2008) found evidence for the effects of self-esteem on reversals of in-group favoritism such that women with lower self-esteem were more likely to denigrate other women as a way to invalidate the ego threat of an outstanding comparison target. In Parks-Stamm et al.'s study, women penalized highly successful female exemplars in order to salvage their own self-views regarding competence. These findings help to explain why a female of lower rank would be threatened by and thus would penalize a more successful female through the process of upward comparison.

In the current study, we are interested in a downward comparison process whereby a female supervisor should not be threatened by an already lower-ranked female. Although this downward-comparison process should not provide the same motivation to invalidate a threat by denigration or penalization of the comparison other (Wills, 1981), we chose to control for the effects of self-esteem in order to focus on the contextual factors (rather than the individual factors) that are the focus of the current research. To control for such individual differences in positive views of the self, we assessed self-esteem with Rosenberg's (1965) Self-Esteem Scale (RSES). Participants indicated their agreement with 10 statements (internal consistency reliability =.76) on a 7-point scale ranging from 1(strongly disagree) to 7 (strongly agree). A sample item is “I feel that I am a person of worth.”

Results and Discussion

Descriptive statistics and intercorrelations for all variables are reported in Table 4. We hypothesized that, for women, perceived tokenism would moderate the relationship between subordinate gender and supervisor support. In other words, we expected that women who perceive they are gender tokens in their work environment (under conditions of high gender tokenism) would be more likely to support a male target than a female target; but that when perceptions of gender tokenism were low, women would provide the same level of supervisor support to both male and female targets. We expected that male supervisors, on the other hand, would be equally likely to support a male or a female target, regardless of whether or not they felt like tokens in their organization.

Table 4. Intercorrelations and Descriptive Statistics for Study Variables: Study 2

Variable	M	SD	Range	1	2	3	4
Note. Gender: 1 = male, 2 = female.
1. Participant gender	1.44	0.50	1–2	—
2. Target gender	1.55	0.50	1–2	−.00	—
3. Tokenism	2.64	1.21	1–6.5	.01	−.14	—
4. Social support	5.04	1.01	1.8–7	−.14	.00	.02	—
5. Self-esteem	6.16	0.71	3.67–7	−.11	−.09	−.19	.18

To test our expectations, we conducted a moderated hierarchical regression analysis predicting supervisor support (see Table 4). After controlling for individual differences in self-esteem, we entered the main effects of gender, target gender, and perceived tokenism; then, the two-way interactions; and finally, the three-way interaction, in the regression model.

No significant main effects or two-way interactions emerged. However, supporting our expectation that women may withhold support from same-gender targets when they perceive that they are gender tokens, a significant change in the amount of variance explained from Step 3 to Step 4 (ΔR² =.07, p <.05) and a significant weight for the interaction term (β = −.27, p <.05) demonstrates that participant gender, target gender, and subjective tokenism interacted to predict supervisor support (see Table 5). This interaction can be interpreted graphically (see Figures 3 and 4; standardized coefficients and values at ±1 SD) and provides support for our predictions.

Figure 3. Interaction of target gender and tokenism on likelihood of providing supervisor support (as depicted on the y axis) for female participants.

Figure 4. Interaction of target gender and tokenism on likelihood of providing supervisor support (as depicted on the y axis) for male participants.

Table 5. Summary of Moderated Hierarchical Regression Analyses Predicting Supervisor Social Support and Professional Evaluations: Study 2

	Social support
	β	ΔR^2
Note. N = 75. *p <.05.
Step 1		.03
Self-esteem	.18
Step 2		.02
Participant gender	.06
Target gender	−.12
Tokenism	.06
Step 3		.02
Participant Gender × Target Gender	−.06
Participant Gender × Tokenism	.01
Target Gender × Tokenism	−.14
Step 4		.07*
Participant Gender × Target Gender × Tokenism	−.27*	.14**

In addition, this effect was probed with separate models for male and female participants. As expected, a significant two-way interaction between subjective tokenism and target gender for female participants suggests that the effect of target gender on support was moderated by tokenism among female participants (β = −.43, p <.05, ΔR² =.18). We predicted that women who work in environments with few other women (i.e., tokens) would be more likely to support a male subordinate than a female subordinate, but that women who are not gender tokens would provide the same level of supervisor support, regardless of their subordinates' gender.

In line with these expectations, we found that the extent to which women experienced tokenism was positively related to their reported likelihood of supporting opposite-gender (male) targets (r =.49, p <.05). The relationship between tokenism and the likelihood that women would support same-gender (female) targets was not significant, but was in the expected negative direction (r = −.30, p >.10). In other words, the extent to which women felt like tokens was positively related to their support for male targets and negatively related to their support for female targets. Moreover, the two-way interaction between tokenism and participant gender was not significant for male participants (β = −.29, p >.10). As expected, the extent to which men experienced tokenism was not significantly associated with their reported likelihood of supporting either male or female targets (rs = −.13 and.21, ps >.10).

The results of this study suggest that when women perceive that they are gender tokens in their organization, they report that they would be less likely to support outstanding targets of their own gender than targets of the opposite gender. Illustrating the gendered (and asymmetrical) nature of this phenomenon, the results suggest that male tokens did not seem to be as influenced as female tokens by their lower proportional representation when offering support to same-gender subordinates. Men rated their likelihood of supporting male targets similarly when they experienced token and non-token contexts. These findings provide support for our hypotheses and suggest that scarce gender representation leads women—but not men—to withhold support from same-gender subordinates.

We acknowledge that there are several limitations to this study. First, the sample size is rather small, but even so, we were able to detect a significant three-way interaction of Supervisor Gender × Tokenism × Subordinate Gender on supervisor support. Second, the subjective tokenism scale had a relatively low internal consistency. This may be a result, in part, of our small sample; or it may be that there was ambiguity in the ideas represented in the items that were adapted to capture both men's and women's general perceptions and experiences, rather than their specific status as a gender token. We find it encouraging that in a larger sample (N = 196) King et al. (2010) tailored this scale to capture the experiences of female managers as gender tokens and found it to be acceptably reliable (.70). Third, a manipulation check was not included, making it difficult to confirm that participants were cognizant of the gender of the applicant they evaluated. While no study can be expected to capture the full spectrum of well documented complexities involved in supervisory relationships, a fourth limitation is that we did not control for participants' prior work experience with respect to familiarity with supervisory relationships or experience evaluating employees. Finally, neither Study 1 nor Study 2 directly explored the psychological processes that give rise to reversals of in-group favoritism among token women. In order to address these aforementioned limitations, we conducted a third study in which we (a) sought a larger sample; (b) included a manipulation check; (c) manipulated token status and measured subjective perceptions of tokenism; (d) controlled for prior supervisory experience; and (e) directly explored social mobility as an underlying psychological process that could explain the results of our first two studies.

Study 3

Purpose

The main purpose of Study 3 is to experimentally test the operation of social mobility as an explanatory factor for the effects of gender tokenism on reversals of female in-group favoritism. According to social identity theory, beliefs about the feasibility of change are central determinants of the extent to which members of disadvantaged groups pursue achievement opportunities (Tajfel & Turner 1979). Perceptions of group boundary permeability (i.e., chances of advancing into an upper-status group) affect whether individual position improvement is feasible (Taylor & McKirnan, 1984). We predict that tokens who have few opportunities for upward mobility may be most likely to enact reversals of in-group favoritism because of the fact that threat of permanent (low-status) group membership is more salient when boundaries are less permeable; and, therefore, distancing oneself psychologically from the in-group is more important for achieving positive identity (Jackson et al., 1996).

In addition, we anticipate that the underlying status differences between men and women in society will yield disparate patterns for men and women who perceive fewer opportunities for advancement. Whereas men are positively stereotyped and rarely discriminated against in (job-related) performance contexts, women are more often exposed to negative stereotyping and discriminatory treatment, which tends to make them more aware of the relevance of gender categorization across contexts (Branscombe, Schmitt, & Harvey, 1999; Simon & Brown, 1987). Thus, women may respond differently (than men) to the same objective information about the possibility of upward social mobility and draw different conclusions about their own opportunities for advancement, given the disadvantaged social history of the female gender (Barreto, Ellemers, & Palacios, 2004). We predict that token women who believe that there are scarce opportunities for women to advance will be most likely to enact reversals of in-group favoritism because they may perceive their gender group membership as a liability (Ely, 1994) and become particularly motivated to distance themselves from it. Thus, we propose the following:

• Hypothesis 3. Social mobility (permeability of higher-status group) will moderate the relationship between subordinate gender and supervisor support provided by token female supervisors such that token women who perceive there are plentiful opportunities for female advancement will be more likely to support their female subordinates than token women who perceive there are scarce opportunities for women to advance.

Method

Participants and Procedure

Study 3's design was a 2 (High-Status Group Mobility: low permeability vs. high permeability) × 2 (Participant–Target Gender Match: male or female) × 2 (Peer Group Composition: all male vs. gender-balanced) between-subjects design. A total of 290 adult undergraduate students (190 female, 100 male) were recruited from the subject pool of a large mid-Atlantic university to take part in an online study. After removing participants who responded only partially and those who failed the manipulation check (which required recall of the target gender and peer group composition), the final sample consisted of 137 women and 96 men. With regard to ethnicity, most of the participants were Caucasian (41.9 %), with 9.4% African American, 25.1% Asian American, 6.0% Middle Eastern, and 11.0% Hispanic. The majority of participants (79.7%) were currently working for pay outside of school.

Participants learned that the experiment was designed to test their leadership skills and that participants in the study would be divided into two groups on the basis of their ability as leaders. All participants were informed that they would begin the experiment in the position of “low-level manager” and that their performance on an initial assessment exercise would determine whether they would complete the rest of the experiment as a member of the “high-potential manager” group or remain in the low-level manager group. Participants were given information about their chances of advancing into the high level (1 out of 10, or 5 out of 10), the gender composition of their fellow low-level managers (gender-balanced, all male, or all female), and the gender composition of the high-status group (expressed as a list of 4 male names). If they were accepted into the high-potential group, they were told,

You will be grouped with High Potential others, who have been recognized as superior leaders. As a member of the High Potential group yourself, you would to have access to greater rewards and responsibilities. High Potential group members will receive more exciting and challenging tasks in the latter part of the experiment. In addition, only High Potential Managers will be entered into a grand prize drawing for gift certificates to a national coffee shop chain.

After completing a brief “leadership assessment exercise,” the participants were presented with a message from a fictitious participant matching their own gender. The experimenter explained that the new participant wanted to join the study, thereby joining their low-level manager group. Participants also received a summary of the ostensible new participant's work experience. This same-gender applicant had a wide range of internship experience that was especially relevant to the stated purpose of the study (i.e., leadership and business decision making). The content of male and female applicants' experience summaries was identical, and the gender of the applicant was expressed by using a feminine or a masculine name (Katherine or Kevin, respectively) and the corresponding pronouns.

Manipulations

Permeability

In order to manipulate the perceived permeability of the upper-status group, the participants were informed that “For every 10 participants, there is a slot for 1 person [5 people] to get into the High Potential Manager group.” Telling participants the odds of advancing from one group to another has been an effective way to manipulate permeability in prior studies of social mobility (e.g., Barreto et al., 2004; Charlton & Bettencourt, 2001).

Gender composition of peer group

Participants were told that they would have the opportunity to view the profiles of participants in their (current) low-level manager group. They were presented with a list of four people who were currently in the low-level managers' group and reminded that they were competing with them to become a member of the high-potential manager group. The gender composition of the low-level manager group was manipulated by using either all masculine names or an equal number of feminine and masculine names.

Measures

Responses to the target

A scale was developed for the purpose of the present study to assess the degree to which participants evaluated the applicant favorably overall, as well as the extent to which the participants were likely to support the applicant should they be allowed to participate in the study. Similar to the measure used in Study 2, participants indicated the strength of their agreement with 18 items on a 7-point scale ranging from 1 (strongly disagree) to 7 (strongly agree).

A principal components factor analysis suggests that 12 of the items loaded cleanly on three factors that accounted for the majority of the variance (and for which no cross-factor loadings exceeded.40). The first factor (eigenvalue = 5.27; 29.0% of variance explained) is labeled Social Support and includes the following items: “I would keep this person informed of things they needed to do this experiment well” (factor loading =.86); “I would help this person do well in this experiment” (factor loading =.82); “I would support this person if they had a problem while participating in this study” (factor loading =.78); “I would show this person the ‘secrets to success’ in this study” (factor loading =.78); “I would like to be a mentor to this person for the remainder of the study” (factor loading =.66); and “I would recommend this applicant for promotion to the High Potential managers group” (factor loading =.63). The internal consistency reliability of these items was.85, and a composite was created to represent social support.

The second factor (eigenvalue = 2.71; 15.0% of variance explained) includes the following items: “This person has the potential to move up in into the High Potential managers group” (factor loading =.86); “This person is well qualified” (factor loading =.89); and “This person will perform well on the leadership assessment exercise” (factor loading =.76). The internal consistency reliability of these items was.83, and a composite was created to represent professional evaluations.

A third factor had an eigenvalue greater than 1 (1.52; 8.5% of variance explained). This factor includes the following items: “This person is intelligent” (factor loading =.70); “This person would work well by himself or herself” (factor loading =.68); “If accepted in this timeslot, this person will probably be selected as a High Potential manager” (factor loading =.55); and “This person would work well in groups” (factor loading =.46). The internal consistency reliability of these items was.61. Thus, a composite of these items was created to reflect general reactions.

Self-esteem

Identical to Study 2, we assessed self-esteem with the RSE (Rosenberg, 1965). Participants indicated their agreement with 10 statements (internal consistency reliability =.76) using a 7-point scale ranging from 1(strongly disagree) to 7 (strongly agree). A sample item is “I feel that I am a person of worth.”

Tokenism

As in Study 2, we assessed the experience of tokenism with five items on a 7-point scale ranging from 1 (strongly disagree) to 7 (strongly agree). A sample item is “I felt like a token representative of my gender.” The internal consistency of these items was.66.

Results and Discussion

Descriptive statistics and intercorrelations between Study 3 variables are presented in Table 6. Before directly testing our hypotheses, we examined scores on the tokenism measure to determine the effectiveness of the manipulations. Female participants in an all-male group reported higher levels of tokenism (M = 2.77, SD = 0.71) than did those in a balanced group (M = 2.41, SD = 0.74), t(137) = 3.09, p <.05. Although the difference was not significant, male participants tended to report somewhat higher levels of tokenism in the balanced group (M = 2.90, SD = 0.65) than in the all-male group (M = 2.78, SD = 0.68), t(96) = −0.70, p >.10.

Table 6. Intercorrelations and Descriptive Statistics: Study 3

Variable	M	SD	1	2	3	4	5	6	7	8	9
Note. Work status: 1= currently working full-time, 2 = currently working part-time, 3 = not working. Race: 0 = White, 1 = non-White. *p <.05. **p <.01.
1. Gender	1.66	0.48	—
2. Gender composition	1.63	0.48	.02	—
3. Mobility	1.59	0.49	.12*	.20**	—
4. Tokenism	2.67	0.71	.22**	−.11	−.02	—
5. Support	3.26	0.72	.01	.12*	.10	−.03	—
6. Professional evaluations	2.98	0.70	.15**	−.10	.07	.20**	.09	—
7. General reactions	3.49	0.50	.15**	−.05	.09	.10	.19**	.52**	—
8. Work status	2.36	0.63	−.08	−.05	−.01	−.16**	.07	−.15*	−.08	—
9. Race	0.45	0.50	.03	.03	.05	−.20**	.00	−.09	.12*	.12	—
10. Self-esteem	3.94	0.43	.00	.11	.01	−.03	.04	−.19**	−.06	.01	−.07

To test the hypotheses, we conducted 2 (Gender Match: male or female) × 2 (Composition of Peer Group: all male vs. gender-balanced) × 2 (Mobility: low permeability vs. high permeability) ANCOVAs on the social support, professional evaluations, and general reactions measures. As in Studies 1 and 2, participant ethnicity was entered as a covariate. In addition, we also included work experience as a covariate. Finally, consistent with the previous study and in line with the findings of Parks-Stamm et al. (2008), we controlled for participants' trait self-esteem. In the case of the general reactions variable, no significant effects emerged.

In the case of social support (see Table 7 for cell means), a marginally significant two-way interaction between mobility and group composition emerged, F(1, 178) = 3.22, p =.06 (see Table 8). When high-status groups were permeable, participants indicated that they would give more support to targets when groups were gender-balanced (M = 3.38, SD = 0.64) than when groups were all male (M = 3.13, SD = 0.87), F(1, 152) = 2.50, p <.05. In low-permeable conditions, support did not depend on group composition: Targets received comparable support in gender-balanced groups (M = 3.13, SD = 0.89) and in all-male groups (M = 3.26, SD = 0.71), F(1, 169) = 0.51, p >.10. Participants indicated the greatest likelihood of supporting same-gender targets when in gender-balanced groups in contexts wherein mobility into high-status groups is feasible.

Table 7. Cell Means for Social Support Variable: Study 3

	All-male group						Balanced-group composition
	Male			Female			Male			Female
	M	SD	N	M	SD	N	M	SD	N	M	SD	N
Low permeability	3.17	0.89	8	3.20	0.76	31	3.07	0.76	22	3.21	0.65	29
High permeability	3.15	0.78	11	3.00	1.06	22	3.43	0.60	34	3.57	0.63	21

Table 8. Results of ANCOVA on Social Support: Study 3

Source of variation	SS	df	MS	F	Partial η^2
Note. SS = sum of squares; MS = mean square. Overall R^2 =.07. †p <.10. *p <.05. **p <.01.
Control variables
Work status	1.60	1	1.60	2.82†	.01
Participant race	0.01	1	0.01	0.01	.00
Trait self-esteem	0.31	1	0.31	0.55	.00
Main effects
Mobility	0.59	1	0.59	1.04	.00
Group composition	1.46	1	1.46	2.57	.01
Gender	0.09	1	0.09	0.15	.00
Two-way interactions
Mobility × Group Composition	1.83	1	1.83	3.22†	.01
Mobility × Gender	0.01	1	0.01	0.02	.00
Group Composition × Gender	0.52	1	0.52	0.91	.00
Three-way interactions
Mobility × Group Composition × Gender	0.14	1	0.14	0.25	.00

The ANCOVA on the professional evaluations dependent variable (see Table 9 for cell means) suggests a more complex pattern (see Table 10). A main effect of group composition was qualified by a significant three-way interaction, F(1, 177) = 9.14 p <.01, η² =.05. We expected that conditions of gender underrepresentation would lead to lower ratings of same-gender targets when high-status groups were low in permeability; or in other words, that group composition would affect ratings of same-gender targets when social mobility is constrained. Consistent with our expectations, in all-male contexts, women indicated higher ratings of female targets when boundaries were permeable than when they were not F(1, 52) = 2.47, p <.05. Men showed a similar pattern in gender-balanced contexts, rating male targets more positively when it was more likely that they could gain access to the high-status group, F(1, 57) = 2.23, p <.05.

Table 9. Cell Means for Professional Evaluations Variable: Study 3

	All-male group						Balanced-group composition
	Male			Female			Male			Female
	M	SD	N	M	SD	N	M	SD	N	M	SD	N
Low permeability	3.33	0.87	8	3.01	0.75	31	2.67	0.59	22	3.13	0.67	29
High permeability	2.91	0.60	11	3.30	0.74	22	2.89	0.70	34	2.86	0.75	21

Table 10. Results of ANCOVA on Professional Evaluations Variable: Study 3

Source of variation	SS	df	MS	F	Partial η^2
Note. SS = sum of squares; MS = mean square. Overall R^2 =.14. †p <.10. *p <.05. **p <.01.
Control variables
Work status	1.56	1	1.56	3.33†	.02
Participant race	0.29	1	0.29	0.61	.00
Trait self-esteem	3.89	1	3.89	8.32**	.05
Main effects
Mobility	0.18	1	0.18	0.38	.00
Group composition	2.33	1	2.33	4.98*	.03
Gender	0.28	1	0.28	0.60	.00
Two-way interactions
Mobility × Group Composition	0.10	1	0.10	0.22	.00
Mobility × Gender	0.18	1	0.18	0.38	.00
Group Composition × Gender	0.25	1	0.25	0.54	.00
Three-way interactions
Mobility × Group Composition × Gender	4.27	1	4.27	9.14**	.05

We also explored ratings of same-gender targets in contexts in which underrepresentation would be less salient (balanced groups for women and all-male groups for men). The results suggest that women in balanced groups tended to evaluate other women more positively when in conditions of low permeability than high permeability, F(1, 49) = 1.90, p <.05. In all-male groups, men's ratings of male targets did not vary as a function of permeability, F(1, 18) =0.83, p >.10. Together, these results suggest that mobility and group composition jointly affect ratings of same-gender targets (see Figure 5).

Figure 5. Interaction of gender, permeability, and tokenism on professional evaluations of same-gender subordinates.

We expected that women would have been less positive toward same-gender targets across all dependent variables when there were fewer opportunities for advancement, but the effects only emerged on professional evaluations. Unlike Studies 1 and 2—where women withheld social support—female supervisors denigrated same-gender targets on professional qualifications (the domain on which their evaluations were based). The finding that reversals of in-group favoritism among women manifest themselves in professional evaluations shows that these token gendered dynamics may operate differently, depending on the specific situation. In Study 1, participants were going to be direct supervisors of the same-gender applicants they were evaluating. In Study 3, the “applicants” were going to be potential direct competitors (entering the peer group of low-level managers). Future research should investigate the potential moderators that would explain the differential emergence of lack of social support versus professional denigration in token contexts.

Unexpectedly, it seems that proportional representation and perceptions of social mobility (i.e., permeability) not only affect same-gender evaluations among women, but among men as well. We found that men's reactions to other men in gender-balanced contexts were similar to the pattern found with women in token contexts: Lower representation produced lower ratings of same-gender targets when there were fewer opportunities for advancement. When men were not the dominant gender group (i.e., in gender-balanced contexts), male supervisors were less positive toward other men (i.e., they gave less positive professional evaluations of other men) when social mobility seemed less feasible. Parallel to the effects found with women, men seem to enact an individualistic social mobility strategy (i.e., professionally denigrating other men) as their proportional representation decreases and opportunities for advancement seem scarce.

These findings should be interpreted in light of the methodological limitations, including the potential ecological validity of the experiment (particularly given the lack of face-to-face interaction between participants and targets); small sample sizes in some conditions; and low reliability of the subjective tokenism scale, which could point to the potential variability in the subjective experiences that were created by the experimental manipulation of token status for men and women. Nevertheless, this study addressed many of the limitations in the previous two studies, and the results of Study 3 specifically suggest that social mobility processes likely affect gendered relations (among both men and women).

General Discussion

The present research sought to examine the influence of proportional gender representation on support that supervisors provide to same-gender subordinates. The results of the first two studies suggest that when women hold token status, they will be less likely to support their female subordinates than their male subordinates. That is, when working in environments in which women were scarce, women tended to show less support to subordinates of the same gender than when they worked in more gender-balanced contexts. The results of Study 3 point to a specific mechanism—social mobility—through which representation affects same-gender relations in organizational settings. Therefore, this research contributes to an understanding of relationships among and between genders at work and organizational demography.

First, the situational influence of gender composition is clearly important in gender relations in the workplace. More specifically, organizational demography seems to be a critical factor in the emergence of differences in same-gender supervisor support. At first glance, as in Studies 1 and 2, it seems that lower proportional representation combined with the historically lower status of women in the workplace to create a social environment in which token females are doubly removed from the dominant male group (Yoder & Sinnett, 1985). However, the similar patterns found for both men and women in Study 3 point to perceived opportunities for advancement, rather than gender status as an explanation for the negative in-group relations that emerge.

Gender underrepresentation and ambiguous or limited status-advancement opportunities seem to create a particularly potent social context for same-gender relations. The appearance of asymmetrical effects on same-gender relations for women versus men (Studies 1 and 2) may be explained by the historically fewer opportunities for advancement for women (compared to men). Although underrepresentation of one's gender does not necessarily correspond with the number of opportunities for advancement into higher level positions, perhaps token females are less likely to support same-gender subordinates (as in Study 2) because they believe that the general underrepresentation of females serves as a signal that there are fewer opportunities for women to advance. In other words, women may enact individualistic social mobility strategies and be less positive toward same-gender subordinates because they interpret their opportunities through the lens of their historically lower gender status.

In both of our samples of working adults, it appears that the majority of women and men did not perceive that they were gender tokens. This finding might provide an encouraging illustration of gender equality and the general advancement of women since the publication of Kanter's (1977) foundational work on the experiences of female tokens, or it could be indicative of continued gender segregation of work (Bureau of Labor Statistics, 2009; Valian, 1998).

We also found support for the suggestion of Yoder and colleagues (Yoder, 1991; Yoder & Kahn, 1993) that tokenism can go beyond the numbers. In Study 2, instead of simply using numerical proportions as in Study 1, we used a subjective assessment of perceived tokenism in order to replicate our findings from Study 1. Moreover, by investigating the effects of tokenism on both social support and professional evaluations, Study 3 provides support for the importance of perceptions of opportunities for advancement, rather than sheer proportional representation for meaningful work outcomes. Our results provide one more piece to the puzzle in attempting to explain the relationship between organizational demography and meaningful work outcomes.

Second, our findings point to an individualistic social identity enhancement strategy (i.e., social mobility) as another important influence on same-gender relations at work. They also illustrate the critical influence of gender status on reactions to tokenism. Presumably, men are unaccustomed to being in an underrepresented minority, given their historically higher status gender group and majority representation employment contexts. If men experienced proportional (under)representation in the same way that women appeared to in Studies 1 and 2, we would expect token male supervisors to withhold support from their male subordinates, thereby distancing themselves from their in-group in an attempt to indirectly improve the esteem they derive from their social identity. In fact, Study 3's results seem to suggest this type of pattern with men.

Such results point to the critical influence of perceived opportunities for advancement on reversals of in-group favoritism. In other words, as opportunities for advancement seem less likely and proportional representation decreases, people tend to utilize individualistic social identity enhancement strategies at the expense of their in-group (professionally denigrating members of their same gender). Women may have fundamentally different interpretations and attributions regarding an identical social context, given that women have historically experienced prejudice and discrimination (and limited opportunities for advancement) on the basis of their gender, particularly at work (Branscombe et al., 1999; Simon & Brown, 1987). They may believe, whereas men may not, that scarce proportional representation is evidence of gender-biased context where opportunities for advancement may be limited (as history would predict) on the basis of gender.

As such, we used Study 3 to directly assess how perceived likelihood of advancement would affect whether tokens create (psychological) distance from members of their same gender (by withholding support) in order to position themselves as an exception to their in-group and improve their social identity. As Studies 1 and 2 suggest, when women are in a token situation, they may want to make sure they are not seen as just another member of the disadvantaged group, thus dissociating from their in-group by reporting they would be less likely to support other females. By specifically manipulating likelihood of advancement in Study 3, we found support for the detrimental effects of tokenism on same-gender relations.

Third, and unexpectedly, tokenism effects seemed to emerge for both men and women, such that token-gender supervisors were less positive toward same-gender subordinates when perceived opportunities for advancement were scarce. It seems that the experience of a token may depend more on the perceived opportunities for advancement than on the status (e.g., gender) of the token. In light of these findings, the results of Studies 1 and 2 (reversals of in-group favoritism among women) could also be explained by the mechanism of social mobility, rather than the double deviance of token women (low-status gender and scarce proportional representation). In other words, when women (who typically encounter fewer opportunities for advancement) are also tokens (lower status group proportionally), they tend to seek individual advancement at the expense of their in-group (withholding social support, as in Studies 1 and 2; or professionally denigrating other women, as in Study 3).

Fourth, our results have implications for minority supervisors and their supportiveness toward other minorities. Our findings support the key role of perceived opportunities for advancement, suggesting that negative in-group relations may also be experienced by token minorities. Similar to the pattern suggested for women, the historically limited opportunities for advancement among minority group members may affect minorities' interpretations of proportional underrepresentation, such that they see proportional underrepresentation as a signal for limited opportunities for advancement of their in-group. This has implications, not only for gender relations, but also for intra-relations of racial–ethnic groups that experience a similar combination of proportional underrepresentation and historically low-status group membership (e.g., Sackett et al., 1991; Yoder et al., 1996), which, together, seem to be interpreted as signaling fewer opportunities for advancement. In other words, tokenism may lead members of any historically disadvantaged social group to become “their own worst enemies” since low-status social group membership could be interpreted as an illustration of fewer opportunities for advancement.

Furthermore, it seems likely that those who are simultaneously members of two historically disadvantaged social identity groups (e.g., Black women) could (a) suffer doubly negative in-group relations; or (b) benefit from a protective factor of one social identity category (e.g., race) over another (e.g., gender). Some research by Yoder and colleagues (e.g., Yoder & Aniakudo, 1997) has suggested that Black women suffer more negative intragroup relations (e.g., get support from other Black women) as members of both a gender token and a racial token than would, say, White women who are only gender tokens. However, other research has indicated that Asians, Blacks, and Hispanics evidence a more collectivist–cooperative orientation (as opposed to individualistic orientation; Cox, Lobel, & McLeod, 1991) and that those who are highly tied to their in-group are more likely to employ strategies to improve the treatment of their group (Branscombe & Ellemers, 1998). It may be that women who become successful leaders in spite of a combination of gender and racial barriers (e.g., Asian women, Black women) could be more inclined to support similar others to follow in their path. Future research should specifically examine the effects of tokenism on social mobility processes and within-group support, and evaluations of racial–ethnic minority women and other employees whose group memberships represent some combination of lower-status identity groups.

Practical Implications

The notion that successful women would somehow hinder their successors is in stark contrast to the well documented phenomenon of in-group favoritism, whereby people generally tend to favor others in their same demographic group (Dienesch & Liden, 1986; Kraiger & Ford, 1985; Tsui & O'Reilly, 1989; Turban & Jones, 1988). The most troubling practical implication of these findings is that in settings where historically lower status females must struggle for representation (and, ultimately, for success), female leaders may not provide their female subordinates as much support as their male counterparts. In male-dominated organizations and industries, women already face substantial challenges, prejudices, and gender barriers without having to struggle against the biases of female leaders (Goldenhar, Swanson, Hurrell, Ruder, & Deddens, 1998; Kanter, 1977). Importantly, when women hold biased evaluations of other women, this is usually not recognized as a form of gender discrimination (Baron, Burgess, & Kao, 1991; also see Petty, Fleming, & White, 1999), but is taken at face value instead. It is critical that awareness and, hopefully, remedies are advanced in organizations in light of these findings in order to promote the continued professional advancement of traditionally underrepresented groups.

The finding that supervisor support was impacted by representation is particularly problematic for several reasons. First, people's perceptions of supervisor support have an important influence on meaningful work-related outcomes (e.g., stress, job satisfaction, turnover, productivity, performance; Baruch-Feldman et al., 2002; Viswesvaran, et al., 1999). Second, in contrast to “official” managerial duties, supervisor support may be viewed as somewhat discretionary. Research (Barreto et al., 2004) has revealed that women suffer most when they encounter gender sexism by other women, as their difficulty in recognizing this as a form of gender discrimination impairs the use of adequate coping responses. Third, subtle displays of in-group distancing, such as withholding supervisor support, may not only be difficult to define and substantiate in terms of legal recourse, but may also be particularly destructive to one's psychological well-being (Cortina, 2008). Finally, it may be the case that withholding supervisor support gives way to even more damaging supervisor behaviors, such as social undermining and workplace bullying of same-gender subordinates (Duffy, Ganster, & Pago, 2002).

Moreover, our Study 3 findings portray a certain context (skewed gender representation and few opportunities to advance) that may create increased susceptibility to gender-based favoritism or conflict. Perhaps more troubling is that reversals of in-group favoritism also emerge on professional evaluations, rather than simply discretionary behaviors (e.g., supervisor support). A more encouraging implication of these findings, however, is that reversals of in-group favoritism may be avoided to the extent that opportunities for advancement are made apparent. The results of Study 3 illustrate that plentiful advancement opportunities seem to alleviate some of the negative consequences of group minority status.

Limitations and Future Research Directions

Like any research endeavor, the current investigation is not without its limitations. The realism of the experimental manipulation and sampling methods are potential limitations that should be addressed by further research. Although our hypothesized effects were supported across complementary samples in field, quasi-experimental, and experimental settings, it would be instructive to investigate the processes involved in supervisor–subordinate dynamics in ongoing work relationships. Whereas people may be more likely to rely on stereotypes in experimental settings, there may be a heightened threat in the workplace where individuals must maintain interpersonal relations with each other.

Also, interpersonal consequences are more meaningful in real-world work situations where one's performance appraisal, promotion opportunities, and even compensation depend on other people's—for example, supervisors'—support and professional opinions. According to a recent critique of laboratory studies of intergroup dynamics (Landy, 2008), one of the primary differences between real-world workplaces and those constructed in the laboratory is that participants in lab studies typically have no history or expected future relationships. The presence of actual relationships could change the dynamics that occur when people interact. Future research should investigate intact dyads of ongoing supervisor–subordinate relationships.

Future research should include multilevel approaches that allow for consideration of unique organizational factors (e.g., climate, job type, industry) in order to clarify our findings. Although research has suggested that males employed in traditionally female jobs (e.g., nurses) do not experience the negative effects of being tokens to the same extent as do women in traditionally male jobs (Fairhurst & Snavely, 1983), additional factors surrounding gender tokenism deserve further attention (e.g., occupational deviance, job prestige). Furthermore, investigations of the influence of multiple in-group identities (e.g., Black women) should be pursued. There is a relative lack of research on the intersectionality of multiple (minority) group identities, especially as it pertains to the experience of tokenism at work (for exceptions, see Yoder, 1991).

In Study 3, we found that, regardless of work experience, token representation at the peer level produced lower ratings of same-gender targets when there were fewer opportunities for advancement. This may suggest that lower level women are threatened by talented same-gender individuals who are ranked closely in the hierarchy. However, further research is needed to determine whether, for example, lower ranking female managers feel more threatened by closely ranked female colleagues than, perhaps, well established female leaders. No research has explicitly examined the top–down direction of same-gender relations in conjunction with degree of hierarchical rank difference within a token environment.

In her study of a large law firm, Ely (1994) found that the presence of token high-status women negatively impacted peer relationships among lower status females in that organization, but did not investigate higher ranking females' reactions to lower ranking females, nor degree of difference between high and low ranking. Moreover, Parks-Stamm et al. (2008) found that (lower ranking) women penalized highly successful female leaders in order to minimize self-evaluative threat and salvage their own self-views regarding competence. However, they did not examine the impact of rank difference, nor did they investigate the opposite comparison of higher ranking females' reactions to lower ranked females. Duguid (2011) found that competitive threat and collective threat shaped female tokens' negative responses to allowing other women into a high-prestige work group, but those “threatening” women were peers with no measure of the impact of rank similarity (or difference) on these same-gender interactions.

Finally, the literature on the queen-bee phenomenon (Staines, Tavris, & Jayaratne, 1974) suggests that token females denigrate same-gender subordinates because they believe there is limited space for females in the work environment, and other females are in competition for limited spaces for females within the male in-group (Mathison, 1986). While directionality is implied, research on the impact of the degree of difference between high-status and subordinate ranking women is lacking in the queen-bee literature. Further research is needed to explicitly test the impact of the degree of rank difference on same-gender relations in token environments. Also, tenure within those higher ranks could be an important factor that may impact same-gender support in gender token environments.

It may be the case that a woman who has securely held a leadership position for many years would be more likely to support an up-and-coming woman than would a new female manager. Research has suggested that in later career stages, individuals who have successfully overcome stereotypes and related obstacles may be more resilient to stereotypical biases and more likely to try to improve the treatment of their social identity group in the workplace (Block, Koch, Liberman, Merriweather, & Roberson, 2011). Moreover, new perspectives of positive leadership also suggest that effective real-world leaders may be more focused on the good of the whole, rather than on personal gain (e.g., May, Chan, Hodges, & Avolio, 2003). Additional examinations of the impact of leadership tenure on same-gender support would improve our understanding of same-gender relations in token environments.

Specific individual differences and interpersonal processes should also be investigated in order to shed light on the potential mechanisms underlying the lack of same-gender supervisor support, especially among women. In particular, perceptions of subordinates as threats to the self or the ego and supervisors' responses to such threats may be a promising avenue of research (Parks-Stamm et al., 2008). Future studies should investigate the role of self-esteem/ego threat in relation to gender-based status as well as status within an organizational hierarchy. It may also be that threats to mobility (possibility for advancement) are related to perceived threats to self-esteem/ego.

Moreover, future research should examine the impact of the difficulty of the leadership journey on same-gender support within token contexts. It may be the case that women who have paved the way for females to overcome gender barriers and become successful leaders are resentful of emerging female leaders who “have it easy” and can just climb the corporate ladder without having to blaze the trail. A study of female faculty members' attitudes toward same-gender doctoral students (Ellemers et al., 2004) found that bias against female doctoral students was limited to the older generation; that is, to female faculty with an age above the median (47 years). These women (who were born between 1921 and 1949) decided to pursue their own academic careers at a time when this was still exceptional for women. The authors suggest that the younger generation of female faculty were perhaps less biased toward female graduate students because, in the modern academic environment, “It no longer seems necessary to distance oneself from other gender group members in order to prove one can be successful at work” (p. 334). Additional research on the perceived severity of gender or racial barriers along a leadership journey would contribute to our understanding of individual leaders' support of same-gender or same-race subordinates.

In our studies, the reduced likelihood of same-gender supervisor support among tokens is evidenced in self-reported perceptions and likelihood of support behaviors, but these could very well be only one aspect of the potential negative outcomes for tokens at work. Consistent with the findings of Study 3, reversals of in-group favoritism may also emerge on professional evaluations that have large symbolic and material consequences. The reduced support and apparent professional denigration between women in particular could be a symptom of a more disturbing phenomenon whereby successful professionals subtly hinder, or even sabotage, the professional advancement of same-gender subordinates.

The results of the present studies suggest that when they are proportionally scarce in a workplace setting, managers may withhold support from same-gender subordinates in order to psychologically distance themselves from their in-group and improve their own individual social identity. This phenomenon may be driven, at least in part, by perceptions of social mobility whereby fewer perceived opportunities for members of one's gender result in decreased support for same-gender subordinates. These findings highlight the impact of gender demography in organizations on same-gender relations. Most notably, though, these findings fly in the face of popular wisdom regarding in-group favoritism. Our results point to a new avenue for research in organizational demography: the role of social mobility perceptions in inter- and intra-group relations. Only with a greater understanding of these phenomena can we make significant strides in ensuring the equitable treatment of all employees.