a Department of Psychology, Northwestern University, 2029 Sheridan Road, Evanston, IL 60208, USAb Department of Psychology, Wellesley College, Wellesley 02481, MA, USA
Available online 30 October 2003.
Journalists and authors of trade books increasingly assert a female advantage in leadership, whereby women are more likely than men to lead in a style that is effective under contemporary conditions. Contrasting our analysis of these claims with Vecchio's [Leadersh. Q. 13 (2002) 643] analysis, we show that women have some advantages in typical leadership style but suffer some disadvantages from prejudicial evaluations of their competence as leaders, especially in masculine organizational contexts. Nonetheless, more women are rising into leadership roles at all levels, including elite executive roles. We suggest reasons for this rise and argue that organizations can capture the symbols of progressive social change and modernity by appointments of women in key positions.
Author Keywords:Leadership style; Sex differences; Prejudice; Meta-analysis; Gender
After years of analyzing what makes leaders most effective and figuring out who's got the Right Stuff, management gurus now know how to boost the odds of getting a great executive: Hire a female. [Sharpe, 2000]
The idea that women are effective leaders has jumped from the writers of feminist trade books on management (e.g., [Helgesen, 1990 and Rosener, 1995]) to the mainstream press and is steadily making its way into the popular culture. Articles in newspapers and business magazines reveal a cultural realignment in the United States that proclaims a new era for female leaders. As Business Week announces that women have the “Right Stuff” [Sharpe, 2000], Fast Company concurs that “The future of business depends on women” [Hefferman, 2002]. Even more startling is Business Week's subsequent cover story on the “New Gender Gap,” maintaining that “Men could become losers in a global economy that values mental power over might” [Conlin, 2003].
The sharp edge of these female advantage articles must be quite baffling to the many academic leadership researchers who have argued that gender has little relation to leadership style and effectiveness (e.g., [Dobbins & Platz, 1986 and Powell, 1990]). They might be tempted to conclude that in our postmodernist world the voices of social scientists have not been accorded any special authority. However, to earn the trust of journalists and the public, leadership researchers must approach these issues with sophisticated enough theories and methods that they illuminate the implications of gender in organizational life. Toward this goal, we show that a careful sifting through social scientific evidence, separating wheat from chaff, suggests that contemporary journalists, while surely conveying too simple a message, are expressing some of the new realities associated with women's rise into elite leadership roles. 1
To address these issues, researchers must confront the perennially important issue of what behaviors characterize effective leaders. Is it the firm execution of authority over subordinates or the capacity to support and inspire them? More likely, as situational theories of leadership contend (e.g., [Chemers, 1997]), the effectiveness of leader behaviors depends on contextual variables, such as the nature of the task and the characteristics of the followers. Yet, historically, leadership has been construed as primarily a masculine enterprise, and many theories of leadership have focused on the desirability of stereotypically masculine qualities in leaders (e.g., [Miner, 1993]). Nevertheless, it is probable that stereotypically feminine qualities of cooperation, mentoring, and collaboration are important to leadership as well, certainly in some contexts and perhaps increasingly in contemporary organizations.
As we analyze these issues in this article, we contrast many of our views with those of [Vecchio, 2002], who recently provided a review of some of the research relevant to the debate about female advantage. The basic questions that we address in this article and that demand attention in relation to these arguments are (a) whether men and women behave differently in leadership roles, (b) whether women receive prejudiced evaluations as leaders and potential leaders, and (c) whether leadership by women might be more effective or better meet the needs of organizations than leadership by men. Although Vecchio addressed the first and third of these questions, our conclusions differ from his, and he gave insufficient attention to the second question—the crucial issue of female disadvantage from discriminatory processes.
2. The changing context of female leadership
Before delving into these issues of female advantage and disadvantage, we note that women's corporate and political leadership is on the rise. Whereas women held only 18% of managerial and administrative positions in the United States in 1972, by 2002 that percentage had increased to 46% [U.S. Bureau of Labor Statistics, 1982 and U.S. Bureau of Labor Statistics, 2002]. Moreover, in the Fortune 500, both the percentage of women among all corporate officers (15.7%) and CEOs (1.4%) are at all-time highs [Black, 2003 and Catalyst, 2002]. Likewise, although women constitute only 14% of the Congress of the United States and 12% of state governors, 42% of the women who have ever served in the Senate are in office now, as are 26% of the women who have ever served as governors [White House Project, 2002]. Also, 43 of the 59 women who have ever served as presidents or prime ministers of nations came into office since 1990 [Adler, 1999 and De Zárate, 2003]. Despite these changes, men, far more often than women, occupy positions conferring decision-making authority and the ability to influence others' pay or promotions [Smith, 2002].
The increase in female leaders has been accompanied by changes in theories and practices of leadership. Whereas in the past, leaders based their authority mainly on their access to political, economic, or military power, in postindustrial societies leaders share power far more and establish many collaborative relationships [Lipman-Blumen, 1996]. Therefore, contemporary views of good leadership encourage teamwork and collaboration and emphasize the ability to empower, support, and engage workers (e.g., [Hammer & Champy, 1994 and Senge, 1994]). Trade books urge managers to put people first by using “resonance-building styles…that support commitment, involvement, active pursuit of the vision, and healthy, productive work relationships” [Goleman et al., 2002].
These contemporary approaches to leadership not only recommend a reduction in hierarchy but also place the leader more in the role of coach or teacher than previous models of leadership. Although the specifics of these views vary, most such discussions emphasize that leader roles are changing to meet the demands of greatly accelerated technological growth, increasing workforce diversity, intense competitive pressures on corporations and other organizations, and a weakening of geopolitical boundaries. As [Kanter, 1997] wrote:
Managerial work is undergoing such enormous and rapid change that many managers are reinventing their profession as they go. With little precedent to guide them, they are watching hierarchy fade away and the clear distinctions of title, task, department, even corporation, blur. Faced with extraordinary levels of complexity and interdependency, they watch traditional sources of power erode and the old motivational tools lose their magic.
Is it possible that the changing nature of managerial work accords female leaders some advantages that they did not possess in the past? As we explain in this article, social scientists have often emphasized the prejudicial disadvantages that women face because of the construal of leadership in masculine terms. To the extent that modern characterizations of effective leadership have become more consonant with the female gender role, this female disadvantage may be eroding.
The gradual erosion of female disadvantage would be consonant with the emphasis of many popular mass-market management books on traditionally feminine communal behavior, involving creating a sense of community, empowering subordinates, and communicating and listening effectively [Fondas, 1997]. Indeed, writers of popular books on leadership have argued that effective leadership is congruent with the ways that women lead [Book, 2000, Helgesen, 1990 and Rosener, 1995]. For example, [Rosener, 1995] labeled women's leadership as interactive, involving collaboration and empowerment of employees, and men's leadership as command-and-control, involving the assertion of authority and the accumulation of power. Such authors construe men and women as quite different in the ways that they lead, with men relying on a somewhat antiquated leadership style that does not fit the needs of most contemporary organizations. These provocative descriptions of sex-typed leadership styles invite careful scrutiny from social scientists.
3. Methods of studying gender and leadership
Before reviewing research on how gender impinges on leadership, we give some consideration to the methods by which we draw our conclusions. Popular writing typically has relied on qualitative analyses or on surveys or interviews with select groups of women leaders (e.g., [Helgesen, 1990, Rosener, 1990 and Rosener, 1995]). Although such approaches can illustrate many of the concerns and experiences of women leaders, they do not allow systematic examination of sex differences and similarities. In contrast, social scientific research has explored gender effects on leadership through a wide variety of research methodologies in many hundreds of studies.
Making use of large numbers of studies is a formidable task. Using narrative methods, some reviewers qualitatively analyze groups of studies to identify common themes or patterns in the findings. Such reviews have value principally when there are few studies that have addressed a particular question. Alternatively, reviewers summarize studies using meta-analysis, which quantitatively combines the results of individual studies [Lipsey & Wilson, 2001]. This method is especially useful for integrating large numbers of studies, which would quickly overwhelm reviewers' abilities to produce accurate generalizations from more intuitive, narrative summaries.
In meta-analyses, the outcome of each study is computed using the common metric of effect sizes, which in reviews of gender and leadership generally take the form of a standardized difference (or d), defined as the difference between the mean scores of women and men (e.g., on a measure of leadership style) divided by the pooled standard deviation. The goal of meta-analysis is to combine study findings to yield measures of the average magnitude of an effect and, even more important, to statistically test whether variation in these findings can be accounted for by the characteristics of the studies themselves.
Given our and [Vecchio, 2002] reliance on meta-analyses, it is important to scrutinize them with care, especially because they purport to provide authoritative summaries of research literatures. To help readers of The Leadership Quarterly evaluate meta-analyses pertaining to gender and leadership, we offer a very short primer on how to judge their quality:
1. Evaluate whether a meta-analyst conducted an extensive enough search that virtually all relevant studies likely were identified. The search should encompass unpublished studies, to lessen publication bias [Sutton et al., 2000]. Multiple databases should be searched as well as the reference lists of existing reviews and all located studies. The exclusion of studies that tested the hypothesis of interest requires a defensible justification. Of course, the cardinal rule of meta-analysis is that all data sets that are included address the same hypothesis.
2. Determine whether at least two independent judges who achieved high interjudge reliability coded the studies extensively enough to represent the differences in studies' attributes, including features relevant to study quality.
3. Appraise whether a meta-analyst thoroughly analyzed the database of studies' effect sizes and coded attributes. Central tendencies of the effect sizes should be presented, accompanied by counts of studies producing findings in each direction. The meta-analyst should calculate all possible models relating study attributes to the effect sizes and report the significant models.
4. Consider whether a meta-analyst appropriately discussed and interpreted the size of effects, the relations between studies' attributes and the effect sizes, the strengths and weaknesses of the studies, and the theoretical and practical implications of the findings.
To illustrate these criteria, we compare [Dobbins & Platz, 1986] meta-analysis of sex differences in leadership style and effectiveness with those of [Eagly & Johnson, 1990] on leadership style and of [Eagly et al., 1995] on leaders' effectiveness. Dobbins and Platz located 17 documents, yielding 8 studies of leadership style and 16 studies of the effectiveness of leaders or satisfaction with leaders' performance. In contrast, Eagly and Johnson located 161 documents, yielding 162 studies of style, and Eagly et al. located 87 documents, yielding 96 studies of effectiveness (including satisfaction with leaders' performance). The discrepancy in the number of studies derives primarily from the far more thorough search procedures of Eagly and her colleagues.
In addition to using minimalist search procedures, [Dobbins & Platz, 1986] failed to code the included studies. They presented only means and variabilities of the effect sizes and dispatched discussion of these findings in three paragraphs, one of which called for a moratorium on research comparing male and female leaders. In contrast, the [Eagly & Johnson, 1990] and [Eagly et al., 1995] meta-analyses included: (a) extensive coding of the studies by two judges who achieved adequate intercoder reliability, (b) the computation of all possible models relating study attributes to the effect sizes and the presentation of the significant models, and (c) a discussion of the strengths and weaknesses and the theoretical and practical implications of their findings.
The most startling defect of the [Dobbins & Platz, 1986] review is its inclusion of studies with designs inappropriate to drawing conclusions about sex differences in leaders' style or effectiveness. Reviews of these questions should include only studies that allowed leader behavior to vary naturally and omit those that constrained or manipulated leader behavior. However, Dobbins and Platz included seven experimental studies featuring male and female behavior that had been made artificially equivalent by presenting participants with (a) standardized written descriptions of leader behavior ascribed to a man or woman [Bartol & Butterfield, 1976, Butterfield & Powell, 1981, Frasher & Frasher, 1980, Haccoun et al., 1978 and Rosen & Jerdee, 1973]; (b) male or female experimental confederates who had been carefully trained to lead in a particular style [Lee & Alvares, 1977]; or (c) videotapes of a man or woman portraying a leader by delivering the exact same script [Welsh, 1979].
These studies holding male and female behavior constant and varying only sex were variants of the Goldberg paradigm, designed to investigate biases in perceptions of equivalent male and female leader behavior [Goldberg, 1968]. Despite the fact that 41% of the included studies deliberately insured the actual equivalence of male and female behavior, [Dobbins & Platz, 1986] announced conclusions about sex differences in leadership style and effectiveness. The invalid conclusions from this meta-analysis spread like a virus through organizational and management science, as shown by the 78 citations of this project [ISI Web of Science, 2003]. [Vecchio, 2002] repeatedly cited this defective meta-analysis to support his views and even maintained that “Dobbins and Platz can be applauded for focusing their analysis on more rigorous, published studies” (p. 651). In this statement, he also wrongly maintained that published studies are more rigorous than unpublished ones, which primarily consist of dissertations. In contrast, meta-analysts judge study quality, not by publication status, but by coding of quality-relevant study attributes.
In this article, we rely on high-quality meta-analytic reviews because of their advantages in producing valid conclusions from large research literatures. Of course, questions concerning gender and leadership, as with any social science research question, are best derived from a variety of different and valuable methods: true experiments, quasi-experiments, organizational studies, survey research, and studies using aggregate social statistics such as wages. Both convergent and divergent findings across differing methods can yield important information.
Experiments can be useful for their excellent internal validity, especially when they are considered in the context of studies using methods with excellent external validity, albeit lesser internal validity (e.g., organizational studies, survey research). [Vecchio, 2002] suspicion of true experiments derives in part from his inaccurate opinion that they are conducted in laboratories with impoverished stimuli. Instead, experiments are defined only by manipulated independent variables and the random assignment of participants to conditions. The stimuli may be complex (e.g., realistic job résumés, videotaped interviews) or simple (e.g., brief descriptions of leaders). The participants can be managers or other nonstudent groups, and the sites include organizations and other nonlaboratory settings. In contrast to Vecchio's tilt toward organizational studies, we thus adopt an ecumenical approach toward research evidence.
4. Sex differences and similarities in leadership style
We first turn to the question of whether women and men differ in leadership style, an issue that many researchers have addressed, primarily with organizational studies. Although [Vecchio, 2002] questioned the value of this research because job descriptions do not ordinarily stipulate particular styles, we believe that this research is important, well beyond its potential to yield “developmental insights related to the understanding of others and one's self” [Vecchio, 2002]. Job candidates' leadership styles are surely among the attributes given special scrutiny in interviews of candidates for managerial positions, and managers fired from their positions are often critiqued for their leadership styles (e.g., faulty “top-down management style,” [Steinberg, 2003]). Moreover, the impetus of this research was, in some cases, to investigate whether the dearth of women in high-level positions could be explained by their leading with styles that are less effective than those of their male counterparts, and in other cases, to investigate whether women possessed superior leadership styles. To shed light on these issues as well as to participate in wider scholarly debates about the magnitude and implications of sex differences in behavior [Eagly, 1995], many investigators have compared the leadership styles of women and men.
4.1. Task-oriented, interpersonally oriented, and autocratic–democratic styles
In the long-standing tradition of studying leadership style (see [Bass, 1990]), most research conducted prior to 1990 distinguished between task-oriented style or initiation of structure and interpersonally oriented style or consideration. A somewhat less popular distinction was between leaders who (a) behave democratically and allow subordinates to participate in decision-making, or (b) behave autocratically and discourage subordinates from such participation. To examine sex differences and similarities in these styles, [Eagly & Johnson, 1990] reviewed 162 studies that yielded comparisons of women and men on relevant measures. Approaching this project in an exploratory frame of mind, Eagly and Johnson began their analysis by discussing both reasons to expect the absence of sex differences and reasons to expect their presence.
This synthesis found that leadership styles were somewhat gender-stereotypic in (a) laboratory studies generally conducted as experiments on group processes with student participants and (b) assessment studies using participants not selected for occupancy of leadership roles (e.g., samples of employees or students in university business programs). Specifically, in such research, women, more than men, manifested relatively interpersonally oriented and democratic styles, and men, more than women, manifested relatively task-oriented and autocratic styles. In contrast, sex differences were more limited in organizational studies, which examined managers' styles. Male and female managers did not differ in their tendencies to manifest interpersonally oriented and task-oriented styles. However, in these studies of managers, as in the laboratory and assessment studies, women manifested a somewhat more democratic (or participative) style and a less autocratic (or directive) style than men did.
This autocratic–democratic finding, which was based on 23 data sets and a heterogeneous set of measuring instruments, produced a relatively small mean effect size (d=0.22). Nonetheless, 92% of the available comparisons went in the direction of a more democratic or participative style among women. The diversity of measures of autocratic and democratic tendencies, regarded as undesirable by [Vecchio, 2002], can be regarded as a strength because the conclusion was not based solely on a particular measure and thus has potentially greater generalizability. Moreover, some confirmation of Eagly and Johnson's findings is contained in a later meta-analysis that surveyed studies published subsequent to their review [van Engen, 2001].
[Vecchio, 2002] would prefer that [Eagly & Johnson, 1990] had excluded studies that used certain measuring instruments, especially [Fiedler, 1967] measure (Least Preferred Co-Worker Scale [LPC]). Instead of excluding studies that used methods that some had criticized, Eagly and Johnson adopted the more informative strategy of including them but then disaggregating their meta-analytic data to test whether differing methods produced differing findings. In particular, studies using the LPC and similar one-dimensional measures of task-oriented versus interpersonally oriented style were aggregated separately from those that assessed interpersonal style and task style on two dimensions.
[Vecchio, 2002] view that studies using leaders' self-ratings should be deleted also violates the meta-analytic principle of including a wide range of methods and disaggregating based on method. Moreover, his opinion that leaders' self ratings have less validity than others' ratings of leaders appears to be debatable. When [Malloy & Janowski, 1992] applied the statistical analyses of [Kenny, 1991] social relations model to perceptions of leadership (e.g., amount of talking, friendliness, leadership), self and other perceptions appeared to be highly similar and in general quite accurate. It is apparently for strictly evaluative ratings of managers' effectiveness that self ratings may be more biased than others' ratings (e.g., [Atkins & Wood, 2000]). At any rate, the identity of raters was treated as a moderator variable in analyses of the effect sizes. In fact, consistent with meta-analytic practice, all coded study attributes were related to the effect sizes, and only the significant models were reported in the text.
Based on analyses of their large database, [Eagly & Johnson, 1990] concluded that gender-stereotypic sex differences in leadership behavior were less common in organizational studies than in other types of studies because male and female managers were selected by similar criteria and subjected to similar organizational socialization—forces that tend to equalize the sexes. Among managers, a sex difference was detected only in a fairly narrow range of leadership behaviors assessed by measures of autocratic–democratic (or directive–participative) tendencies, which relate primarily to the exercise of power. These findings illustrate the value of comparing studies of different types: Without the context provided by the laboratory and assessment studies of persons not occupying managerial roles, any conclusion about the lessening of sex differences among managers would not have been convincing.
This meta-analysis included some possible interpretations of the autocratic–democratic sex difference—specifically, (a) the greater social skills of women (vs. men) may have facilitated collaborative, democratic leadership behavior and (b) such behavior may have been especially advantageous for women because it placated subordinates and peers who might otherwise have been resistant to female leadership. Although we do not concur with [Vecchio, 2002] view that the demonstrated difference in autocratic–democratic tendencies is unimportant, we agree that this style's effectiveness depends on context (e.g., [Foels et al., 2000 and Gastil, 1994]), as did [Eagly & Johnson, 1990], who were “unwilling to argue that women's relatively democratic and participative style is either an advantage or disadvantage.” Nonetheless, consistent with [Fondas, 1997] analysis, democratic, participative leader behavior is more similar to the leadership styles advocated by contemporary managerial writers than is autocratic, directive behavior.
Another significant moderator analysis in [Eagly & Johnson, 1990] meta-analysis established that leaders of each sex were especially task-oriented relative to the other sex when their role was congruent (or congenial) with their gender—that is, defined in relatively masculine terms for male leaders or in more feminine terms for female leaders. This finding does not support [Vecchio, 2002] interpretation that “regardless of sex, individuals are inclined to be work-focused” (p. 651), but rather that leaders' work-related task behavior occurs more often when the leader role is perceived as congenial to their gender.
4.2. Transformational, transactional, and laissez-faire styles
Debates about the leadership styles of women and men gained momentum in the 1990s because of new research attempting to identify the styles that are especially attuned to contemporary conditions. The new emphasis was on leadership that is transformational in the sense that it is future oriented rather than present oriented and that strengthens organizations by inspiring followers' commitment and creativity. As initially described by [Burns, 1978] and elaborated by [Bass, 1985], transformational leadership entails establishing oneself as a role model by gaining followers' trust and confidence. Transformational leaders state future goals, develop plans to achieve those goals, and innovate, even when their organization is generally successful. By mentoring and empowering followers, such leaders help followers to develop their potential and thus to contribute more effectively to their organization.
Leadership researchers (e.g., [Avolio, 1999]) contrasted transformational leaders to transactional leaders, who appeal to subordinates' self-interest by establishing exchange relationships with them. Transactional leadership involves managing in the conventional sense of clarifying subordinates' responsibilities, rewarding them for meeting objectives, and correcting them for failing to meet objectives. In addition, researchers distinguished a laissez-faire style that is marked by an overall failure to take responsibility for managing. These distinctions between aspects of leadership style are commonly assessed by the Multifactor Leadership Questionnaire, known as the MLQ (see Table 1 and [Antonakis et al., 2003]).
Table 1. Study-level effect sizes for transformational, transactional, and laissez-faire leadership styles
Positive effect sizes (ds) indicate that men had higher scores than women on a given leadership style, and negative ds indicate that women had higher scores than men.
k=number of studies; d+=mean weighted effect size; CI=confidence interval; Q=homogeneity of ds.
Table adapted from [Eagly et al., 2003].
View Within Article Although [Vecchio, 2002] discounted the possibility that noteworthy gender effects might emerge in research on transformational and transactional leadership, researchers in this area have reasoned that transformational leadership might be particularly advantageous to women because of its androgynous qualities (e.g., [Yoder, 2001]) and, indeed, the substantial research literature comparing women and men on these styles has yielded interesting outcomes. Pursuing these ideas, [Eagly et al., 2003] carried out a meta-analysis of 45 studies that compared male and female managers on measures of transformational, transactional, and laissez-faire leadership styles. This meta-analysis encompassed a large study conducted to provide norms and psychometric standards for the MLQ [Center for Leadership Studies, 2000] as well as many studies conducted within specific organizations or groups of organizations, the majority of which were in business or educational domains. The measures of leadership style were completed by leaders' subordinates, peers, or superiors or by the leaders themselves.
In general, [Eagly et al., 2003] meta-analysis revealed that, compared with male leaders, female leaders were (a) more transformational (significant in general and on all but one subscale) and (b) engaged in more of the contingent reward behaviors (i.e., exchanging rewards for followers' satisfactory performance) that are one component of transactional leadership (see Table 1). Also, male leaders were more likely than female leaders to manifest two other aspects of transactional leadership: active management by exception (attending to followers' mistakes and failures to meet standards) and passive management by exception (waiting for problems to become severe before intervening). Men were also higher on laissez-faire leadership (exhibiting widespread absence and lack of involvement). These sex differences were small, but prevailed in the meta-analysis as a whole as well as in auxiliary analyses of (a) the MLQ norming study, (b) the other studies that used the MLQ, and (c) the studies that used other measures of these styles.
These provocative findings are unlikely to be an artifact of publication bias, given that the majority of the included studies were unpublished and that the authors of the published studies had often not included sex comparisons (but were willing to provide these data when asked for them). In addition, contrary to the view, endorsed by [Vecchio, 2002], that sex differences in style are an artifact of the placement of women and men in different leadership roles, which often cede men more power than women (e.g., [Kanter, 1977 and Ragins, 1991]), the sex effects did not differ between (a) studies assessing leaders who had the same specific role description (e.g., college hall directors) and (b) studies assessing leaders in a broad category within which men and women were likely concentrated in different roles (e.g., managers of various organizations). Also, these findings are unlikely to be a product of the application of different standards in judging men and women (see [Biernat & Kobrynowicz, 1999]) because the rating scales used in this research are not adjective rating scales but are behaviorally anchored scales assessing the frequency with which a leader engages in each of the relevant behaviors.
Interpretation of these findings included several possibilities [Eagly et al., 2003]: (a) the transformational repertoire (and contingent reward behaviors) may resolve some of the typical incongruity that exists between leadership roles and the female gender role because these styles are not distinctively masculine (see [Eagly & Karau, 2002]), (b) gender roles may influence leadership by means of the spillover and internalization of gender-specific norms, which could facilitate women's focus on the more feminine aspects of transformational leadership (e.g., individualized consideration), and (c) the glass ceiling and the associated double standard for the selection and promotion of managers may produce more highly skilled female than male leaders.
Sex differences in transformational and transactional leadership do have implications for female advantage arguments because researchers defined these styles in an effort to identify effective leadership. Substantiating these claims, a meta-analysis of 39 studies showed positive correlations between effectiveness and all components of transformational leadership as well as the contingent reward component of transactional leadership, the one aspect of transactional leadership on which women exceeded men [Lowe et al., 1996]. The norming study of the MLQ measure produced similar effectiveness findings [Center for Leadership Studies, 2000], and in addition, showed negative relations between leaders' effectiveness and two of the remaining measures: (a) passive management by exception, which is one of the components of transactional leadership, and (b) laissez-faire leadership.
In view of these findings, the tendency of women to exceed men on the components of leadership style that relate positively to effectiveness (i.e., transformational leadership and the contingent reward aspect of transactional leadership) and the tendency of men to exceed women on the ineffective styles (i.e., passive management by exception and laissez-faire leadership) attest to women's abilities. Thus, research on transformational, transactional, and laissez-faire leadership styles does suggest female advantage, albeit a small advantage.
5. Prejudice and discrimination against women as leaders
Any female advantage in leadership style might be offset by disadvantage that flows from prejudice and discrimination directed against women as leaders. Prejudice consists of unfair evaluation of a group of people based on stereotypical judgments of the group rather than the behavior or qualifications of its individual members. When people hold stereotypes about a group, they expect members of that group to possess characteristics and exhibit behavior consistent with those stereotypes. Perceivers then tacitly assimilate information to their gender-stereotypic expectations [von Hippel et al., 1995] and spontaneously fill in unknown details of others' behavior to conform to those expectations [Dunning & Sherman, 1997]. These stereotypic inferences yield prejudice against individual group members when stereotypes about their group are incongruent with the attributes associated with success in certain classes of social roles. This incongruity tends to produce discrimination by lowering evaluation of such group members as potential or actual occupants of those roles.
According to [Eagly & Karau, 2002], incongruity between expectations about women (i.e., the female gender role) and expectations about leaders (i.e., leader roles) underlie prejudice against female leaders (see also [Burgess & Borgida, 1999 and Heilman, 2001]). This explanation, based on Eagly's social role theory of sex differences and similarities in social behavior [Eagly, 1987 and Eagly et al., 2000] asserts that the activation of beliefs about women and men by gender-related cues influences people to perceive individual women as communal but not very agentic and individual men as agentic but not very communal.
Consistent with role incongruity theory, stereotype research reveals that people do consider men to be more agentic than women and women to be more communal than men [Deaux & Kite, 1993]. Also, the communal qualities that people associate with women, such as warmth and selflessness, diverge from the agentic qualities, such as assertiveness and instrumentality, that people perceive as characteristic of successful leaders. In contrast, the predominantly agentic qualities that people associate with men are similar to the qualities perceived to be needed for success in high status occupations, which would include most managerial occupations (see [Cejka & Eagly, 1999]).
Stereotypes about women and men, like other stereotypes of social groups, appear to be easily and automatically activated [Fiske, 1998]. Although stereotypes are not necessarily activated or applied to bias judgments of individuals, many circumstances do favor both activation and application [Kunda & Spencer, 2003]. Because such circumstances are common, we assume that, in general, perceptions of individual leaders reflect both beliefs about leaders and beliefs about gender. Therefore, it is likely that judgments of female leaders ordinarily manifest an amalgam of the communal traits associated with the female gender role and the agentic traits associated with leadership roles [Heilman et al., 1995]. The typical similarity of the male gender role and leader roles requires no such amalgamation.
Ample evidence exists that managers (and undergraduate and graduate business students) link management ability with being male and possessing masculine characteristics [Powell et al., 2002 and Schein, 2001]. Also, nationally representative survey data reveal greater preference for male than female bosses, although this preference has been decreasing over time [Simmons, 2001]. Because of doubts about women's leadership ability, they are generally held to a higher standard of competence than men are. For women to be considered as instrumentally competent as men, perceivers must be given clear evidence of women's greater ability or superior performance compared to their male counterparts [Biernat & Kobrynowicz, 1997, Foschi, 1996, Shackelford et al., 1996 and Wood & Karten, 1986].
Examples of the challenges women face in being perceived as competent leaders include a field study examining judgments of leadership ability among U.S. Army captains attending a leadership training course. This study revealed a bias favoring men in both self and other judgments, even when controlling for years as captain and other status factors, and particularly when women were in a solo or token situation in their groups [Biernat et al., 1998]. A field study of a university cadet corps also showed that judgments of competence and leadership qualities were biased against women, when evaluated against the criterion of objective military performance [Boldry et al., 2001].
Despite the external validity advantages of organizational studies, they remain vulnerable to the internal validity criticism that uncontrolled sex differences may account for apparently prejudicial effects. To address this concern, researchers have turned to Goldberg paradigm experiments that equate the objective characteristics of men and women other than their sex and are carried out in a variety of settings with diverse participants, including managers and other employee groups.
One popular variant of these Goldberg paradigm experiments involves presenting a résumé or application for evaluation, with half of the participants receiving it with a female name attached to it and half receiving it with a male name. Meta-analysis of these experiments showed that men were preferred over women for jobs rated as male sex-typed (d=0.34) and women over men for jobs rated as female sex-typed (d=−0.26; [Davison & Burke, 2000]). Given that leadership roles are usually sex-typed as masculine, these experiments suggest bias against female candidates for such positions. Other studies in this tradition include some more naturalistic but less controlled field experiments conducted by sending male and female job applications to companies (e.g., [McIntyre et al., 1980]) or having male and female confederates respond by telephone to advertised jobs [Levinson, 1982]. These studies also showed sex discrimination that depended on the sex-typing of jobs.
A meta-analysis of Goldberg paradigm experiments that investigated evaluations of male and female performances on a wide variety of tasks revealed an overall effect favoring male expertise [Swim et al., 1989]. This overall prejudicial bias was small (ds between −0.05 and −0.08, depending on the method of estimation). However, consistent with the role incongruity theory, this bias was larger for masculine and gender-neutral tasks (ds between −0.10 and −0.32) than feminine tasks (ds between −0.01 and −0.08).
Even in face-to-face interactions with no evidence of male superiority at the task [Wood & Karten, 1986] or no objective gender differences in the quality of performance [Carli, 1991 and Carli, 1997], undergraduate participants rated men to have performed more competently than women. Also, in an experiment manipulating managers' speech, union members rated female managers as less effective than their male counterparts when the managers spoke in a relatively uncertain manner, although they gave the female and male managers comparable ratings when they communicated more competently [Geddes, 1992]. In addition, women's task contributions in small groups are more likely to be ignored or to evoke negative reactions from others than men's contributions are (e.g., [Butler & Geis, 1990 and Propp, 1995]). Consequently, women, more than men, face a difficult challenge to prove that they have the high level of ability required to become a leader.
Despite doubts about women's competence as leaders, one might expect that highly agentic female leaders would be able to overcome such difficulties. However, people may perceive women who demonstrate clear-cut leadership ability as insufficiently feminine. Thus, a female leader can be rejected because people perceive her to lack the agentic qualities associated with effective leadership or because she possesses too many of them. This rejection as “too masculine” results from injunctive or prescriptive gender role norms—that is, consensual expectations about what men and women ought to do—that require women to display communal behavior and not too much agentic behavior [Eagly & Karau, 2002 and Fiske & Stevens, 1993].
As a result of these injunctive demands, female leaders often receive less favorable reactions than male leaders do for male-stereotypic forms of leadership. This generalization was confirmed in a meta-analysis of Goldberg paradigm experiments on the evaluation of male and female leaders [Eagly et al., 1992]. Although the overall bias in favor of men was small (d=0.05), women received lower evaluations than equivalent men for autocratic leadership (d=0.30) but comparable evaluations for democratic leadership. Also, women encounter more dislike and rejection than men do for showing dominance, expressing disagreement, or being highly assertive or self-promoting (e.g., [Carli, 1998, Copeland et al., 1995 and Rudman, 1998]). In addition, dominance lowers women's but not men's ability to influence others [Carli, 1998 and Ellyson et al., 1992]. Greater penalties against women than men for dominant and assertive behaviors reflect the constraints on women to avoid stereotypically masculine behavior.
The resistance to female leadership demonstrated by these findings is problematic for female leaders, especially because it appears that, with some exceptions (e.g., [Rudman, 1998]), men, who currently hold most positions of power and authority, generally find female agency and leadership more objectionable than women do. This generalization is consistent with the [Eagly et al., 1992] meta-analysis, which showed that male evaluators rated female leaders less favorably than equivalent male leaders, whereas female evaluators did not exhibit gender bias. For example, in one study, employees watched a videotape depicting an assertive, confident target employee who expressed interest in advancement and later disagreed with senior management [Geller & Hobfoll, 1993]. Results revealed that, whereas the male target received similar evaluations from the men and women, the female target received less favorable evaluations from the men. Such findings are consistent with evidence that, although both sexes view the social category of “successful managers” as more like men than women, men show this “think manager, think male” effect more strongly than women [Schein, 2001].
Male-dominated environments can be difficult for women. The [Eagly et al., 1992] meta-analysis of Goldberg paradigm experiments thus demonstrated that female leaders received less favorable evaluations than their equivalent male counterparts in male-dominated leader roles, but were equally evaluated in roles that were not male-dominated. Congruent evidence in survey data emerged from an analysis of the Panel Study of Income Dynamics [Hill, 1992], which showed that, despite controls on numerous variables (e.g., human capital, family characteristics, skill requirements of occupations), working in male-dominated occupations increased men's chances of promotions (as assessed by substantial wage increases in successive years), but increased women's chances of leaving their jobs [Maume, 1999]. Consistent with such findings, women appear to be disadvantaged in employment contexts in which advancement depends on sex homophilous male networks, which are especially prevalent in environments where men constitute a strong majority [McPherson et al., 2001].
In terms of everyday organizational behaviors, discrimination in male-dominated settings occurs through blatant and subtle stereotyping, questioning of women's competence, sexual harassment, and social isolation (e.g., [Collinson et al., 1990, Fitzgerald et al., 1997, Kanter, 1977 and Martin, 1992]). Stereotyping can produce its own reality through expectancy confirmation processes that can derail women's performance in the stereotypic domain [Geis, 1993]. In a demonstration of such processes, experiments in the “stereotype threat” paradigm made the female stereotype especially accessible to students by having them view female-stereotypic (vs. neutral) television commercials [Davies & Spencer, 2003]. Then in a subsequent procedure portrayed as an unrelated experiment on leadership, the women, but not the men, who had been exposed to the female-stereotypic portrayals, expressed less preference for a leadership role versus a nonleadership role.
How can women cope with disapproval elicited by their assertive, directive, or highly competent behavior (e.g., [Carli, 1990])? Not surprisingly, given the injunctive demand for female communality, one technique supported by research findings involves combining agentic behavior with warm, communal behavior, which seems to mitigate suspicion of agentic women. Women thus increase their likableness and influence by “feminizing” their behavior and displaying increased warmth or cooperativeness, whereas men's influence does not depend on displays of communality [Carli, 2001, Carli, 1998, Carli et al., 1995, Ridgeway, 1982 and Shackelford et al., 1996]. Because men are not penalized for communal behavior but can gain from dominant and assertive behavior, male leaders may enjoy easier access to a wider range of leader behaviors that can be tailored to fit the demands of the situation.
In summary, research provides ample demonstration of bias against women as leaders, despite the failure of some reviewers (e.g., [Arvey & Murphy, 1998]) to recognize important moderators of these effects. Our conclusions are strengthened by converging results from differing methods—specifically, organizational studies, survey research, and experiments that equate the objective characteristics of men and women. Women thus face discriminatory barriers mainly in male-dominated and masculine environments and with male evaluators. Because higher levels of authority and higher wages are concentrated in such environments and are controlled primarily by men, this prejudice is highly consequential for women's advancement.
Traditional arguments against prejudice as accounting for women's lower wages and lesser workplace authority ascribed these phenomena to women's lesser human capital in terms of education, training, and work experience. However, women's human capital investments have increased [Eagly & Carli, in press]. In addition, sex differences in human capital investments account for only a portion of the male–female wage gap (e.g., 21% in [O'Neill & Polachek, 1993]) and are clearly insufficient to explain women's lesser access to higher-level leadership positions in view of evidence that women receive substantially smaller gains in workplace authority than men do for similar human capital investments [Smith, 2002]. In addition, considerable evidence argues against the “female choice” proposition that family and domestic responsibilities cause employed women to avoid leadership responsibility (see [Smith, 2002]).
Paradoxically, the discriminatory disadvantage that women encounter in male-dominated environments can sometimes produce the appearance of a female competence advantage. Given impediments to achieving high-level leadership roles, those women who do rise in such hierarchies are typically the survivors of discriminatory processes and therefore tend to be very competent. This increment of competence, driven at least in part by a double standard, is no doubt one factor underlying social scientific evidence and journalistic claims of female advantage (e.g., [Eagly et al., 2003 and Sharpe, 2000]).
6. The effectiveness of male and female leaders
We have presented meta-analytic evidence that women are slightly more likely than men to lead in the ways that managerial experts consider particularly effective and that have been shown to be effective in research on transformational, transactional, and laissez-faire leadership [Lowe et al., 1996]. However, we have also presented evidence of prejudice against female leaders and potential leaders in masculine domains. Given both advantage and disadvantage, how do female leaders fare relative to their male counterparts? This question should be addressed with direct assessments of effectiveness on appropriate measures, including performance appraisals and objective outcome measures.
Given the findings we have presented on prejudice, it would be reasonable to expect that the relative success of women and men in leadership roles depends on context. In masculine contexts, prejudicial reactions not only restrict women's access to leadership roles, but also can reduce the effectiveness of women who attain these roles [Eagly & Karau, 2002]. Testing this role incongruity hypothesis, [Eagly et al., 1995] conducted a meta-analysis that included 96 studies that compared the effectiveness of male and female leaders. As required to properly assess reactions to actual leadership, the review contained only studies that did not artificially equate male and female behavior. In these 74 organizational and 22 laboratories studies, the male and female leaders held the same role, which was sometimes given a general definition (e.g., middle managers in one or more industries) and sometimes a specific definition (e.g., elementary school principals in a particular city). Most of these studies used subjective ratings of performance or effectiveness, and some included more objective measures. It is of course likely that subjective ratings of effectiveness were biased to some extent, given the demonstration of some gender bias in the evaluation of leaders in the [Eagly et al., 1992] meta-analysis of Goldberg paradigm experiments.
The results of these studies proved to be quite heterogeneous. Nevertheless, moderating variables successfully predicted the effect sizes, revealing support for the incongruity hypothesis that women are relatively less effective in leadership roles defined in especially masculine terms and more effective in roles defined in less masculine terms. The following findings were consistent with these expectations: (a) women were less effective than men to the extent that leadership positions were male dominated; (b) women were less effective relative to men as the proportion of male subordinates increased; (c) women were less effective relative to men the greater the proportion of men among the raters of leader effectiveness (see also [Bowen et al., 2000]); (d) women were substantially less effective than men in military organizations, a traditionally masculine environment, but modestly more effective than men in educational, governmental, and social service organizations; (e) women fared particularly well in effectiveness, relative to men, in middle-level leadership positions, as opposed to line or supervisory positions. This finding is consistent with the characterization of middle management as favoring interpersonal skills that are in the more communal repertoire (e.g., [Paolillo, 1981]).
To further explore the role congruity principle that the masculinity of leader roles affects whether men or women are more effective, [Eagly et al., 1995] ran six additional tests based on a group of respondents' ratings of each of the leadership roles in the meta-analytic sample of effectiveness studies. These ratings were correlated with the studies' effect sizes, which represented the comparison between male and female leaders' effectiveness. Although [Vecchio, 2002] criticized [Eagly et al., 1995] for using university students as raters, young adults have experience with managers in many educational and business settings. More important, Eagly et al. based their conclusions not merely on analyses using these ratings, but also on the relations between the coded study attributes and studies' outcomes (see prior paragraph).
These respondents rated the leadership roles on three types of measures: (a) how competent they thought they would be in each role and how interested they would be in performing each role, (b) how interested the average man and the average woman would be in occupying each role, and (c) how much each role required the ability to direct and control people and the ability to cooperate and get along with other people. Roles were considered congruent with the male gender role to the extent that the male respondents indicated more competence and interest in them, the roles were perceived as more interesting to the average man, and the roles were seen as requiring the ability to direct and control people. Roles were considered congruent with the female gender role to the extent that the female respondents indicated more competence and interest in them, the roles were perceived as more interesting to the average woman, and the roles were seen as requiring the ability to cooperate and get along with others. Using these several measures of leader roles' congruity with the male and female gender role, [Eagly et al., 1995] found that the relative effectiveness of female leaders compared with their male counterparts decreased for the roles rated as more congruent with the male gender role and increased for the roles rated as more congruent with the female gender role.
In view of the consistent results produced by these 11 moderating variables (five based on coded study attributes and six on ratings of the leadership roles), clearly the mean sex difference in effectiveness in the meta-analysis depends on the balance of masculine or feminine leadership roles that happened to be represented in the sample of studies. Overall, there was no difference in the effectiveness of male and female leaders (d=−0.02, indicating nonsignificantly greater female effectiveness). Nonetheless, consistent with the prejudice we documented in the prior section of this article, women do fare worse than men in masculine settings. However, perhaps because of an advantage accorded by their tendency toward a transformational (and contingent reward) leadership style, they fare somewhat better than men in less masculine settings. In general, gender appears to be consequential in relation to leaders' effectiveness.
Finally, people's perceptions of female advantage must be framed by the well-known history of concrete walls and glass ceilings that have restricted women from positions that carry substantial authority. Therefore, research findings or everyday observations suggesting equality between women and men in their effectiveness as leaders likely appear notable because men might be expected to be generally more able than women to justify their greater success in obtaining higher-level leadership roles. In addition, because social perceivers often believe that women in extremely male-dominated positions have had to overcome very strong barriers, they may occasionally accord such women competence beyond what they accord to equivalent men (e.g., [Heilman et al., 1988]). This effect presumably occurs because perceivers augment the causal importance of a force (i.e., task competence) that they believe has prevailed over a countervailing force (i.e., discrimination; [Kelley, 1972]).