Conclusion

’Tis woman’s strongest vindication for speaking that the world needs to hear her voice.

—ANNA JULIA COOPER¹

THE MEETING HAS LONG BEEN VIEWED as an essential practice of democracy. Centuries ago the French observer Tocqueville wrote: “Town-meetings are to liberty what primary schools are to science; they bring it within the people’s reach, they teach men how to use and how to enjoy it” ([1835] 2006, book 1, chapter 5). Tocqueville’s words apply today, not only to town meetings, but also to formal settings of all kinds where people talk about matters of common concern. Public talk is the lifeblood of democracy and of community. And meetings remain an essential way for people to make collective decisions that matter.

This faith in public talk has prompted impassioned arguments in favor of deliberation in civic and political life (Gutmann and Thompson 1996). Some supporters of deliberation ground their advocacy in part in the notion that the people of a democracy have an obligation to become adequate citizens of it, so they can fulfill their function in the political system and thus sustain that system. An adequate citizen, on this view, has the basic knowledge needed to hold officials accountable and to decide matters of common concern (Converse 1964; Delli Carpini and Keeter 1996). Deliberation properly constructed can allow people to arrive at more considered, authentic views free of manipulation (Fishkin 2009). In addition, the adequate citizen is one who is willing to engage respectfully with views opposed to their own (Mutz 2006). Thus, some observers are excited about the potential of deliberation to raise the public’s level of information and reasoning about politics and its willingness to listen to “the other side” (Walsh 2007; Fishkin 2009; Mutz 2006; Myers and Mendelberg 2013; Rosenberg 2007a). Deliberation also has the potential to lay bonds of mutual trust and cooperation, and even enlarge citizens’ interests beyond their selfish confines, to the benefit of the communal enterprise (Walsh 2007; Gastil et al. 2010; Macedo et al. 2005; Myers and Mendelberg 2013). As a bonus, deliberation can boost the legitimacy of the system, strengthening it not only by creating strong citizen building blocks but also by granting it an aura of integrity (see, for example, Gastil et al. 2010, 19–21; Myers and Mendelberg 2013). All these are ways in which deliberation could serve the needs of the political community.² Put simply, deliberation can lay the foundation of citizenship and thereby secure the health of democracy.

But to participate in deliberation about matters of common concern is more than an opportunity for learning, competence, and enlightenment. The reason that meetings are so important to democracy goes beyond their ability to educate the masses in the skills and knowledge they need to be competent, as some might be tempted to interpret Tocqueville’s “school for democracy” passage. That phrase is about education, to be sure, but it also serves as a ringing endorsement of liberty. Meetings are good because they allow the people to “use” and “enjoy” liberty—that is, agency. Discussion in meetings is a basic necessity and a fundamental right of the citizen because it reflects, and shapes, the citizen’s agency. And when it comes to agency over public affairs, citizens can only carry influence if their authority is as high as others’. The democratic citizen, by our definition, is not merely an individual in need of education and enlightenment, or a cog in the machinery of government, she is first and foremost an equally authoritative member of a political community. Therefore, exercising equal participation and influence is a crucial way for people to reap the benefits of citizenship, to achieve the same high standing in the self-governing community that the highest governor enjoys. Meetings are not merely medicine for the body politic, taken to fix its ills and get it to do its proper work. They enable status in the community of self-rulers and authority in society at large.

Put differently, meetings matter because meetings grant the citizen equal standing in the political community. As a British juror remarked after serving for the first time, “I’ve been approached after 35 years and this has now made me feel that I’ve been treated like a British citizen” (quoted in Gastil et al. 2010, 22). It is this sense of worth, or value in a community—including how individuals see themselves and the authority and respect given by others—that this book has sought to understand.³ As Supreme Court justice Anthony Kennedy has argued, the Fourteenth Amendment protects “equal participation in civic life” and thus the right to jury service because to be a juror is not only a “duty” but also an “honor” (Holland v. Illinois 1990, 488–89). We agree, but take a still broader view, beyond the jury setting. Basic honor, which we call worth, standing, and status, is essential to any person, because it is the currency of society and a fundamental social need.

We have shown that standing, worth, and status are heavily shaped by the interaction among people at the public meeting. The meeting is thus not only necessary for the empowerment of the people to govern but also can create equality among the people as they attempt to self-govern. That is the promise of the public meeting. But the dark side of the public meeting is that equality is too rare and the reality too distant from it. We have documented just how far discussion lies from the standard of equality, in the common circumstances of politics. The conditions of deliberation in many political settings produce a marked disadvantage for one-half of the human species.

Women are no longer officially restricted from full participation in public gatherings. On paper, they have equal access to juries, university seminars, voluntary association boards, government committees, and even work teams. And so, in theory, we would expect to find that they participate and exercise power at the same rate as men do at the meetings that govern our society in these influential spaces. Yet women are still underrepresented in the ranks of active and influential participants in many of these settings. Consequently, women do not benefit as men do from meetings as a school for democracy. Nor do they gain the recognition and standing that men do as valued members of the community. Writing about class differences, Tocqueville wrote, “The humblest individual who is called upon to co-operate in the government of society acquires a certain degree of self-respect; and as he possesses authority, he can command the services of minds much more enlightened than his own” ([1835] 2006, chapter 14). In most political settings, ordinary women do not reap these rewards of self-respect, authority, liberty, and influence as much as their male counterparts.

In turn, women’s continued lower overall position in politics, and their continued disadvantages in many areas of society, may be reinforced by their continued devaluation as persons of equal worth in the consequential setting of the public meeting. Despite gaining resources and making enormous advances toward equality in American society, women are frequently still expected to be less suited to the exercise of authority. Politics continues to be a “man’s game” that generates continued gender gaps in participation and influence. As Nannerl Keohane has written, “Very few women have exercised authority in institutional settings over men and women of comparable social and economic status” (2010, 125). Wielding power is still deemed an activity that is not fully appropriate for women’s feminine gender roles. To sum it up, gender continues to be a dimension of political underrepresentation today, all the more so when authority is at stake. And meetings are a site of authority.

While some women may respond that they are content with this state of affairs, and they choose not to participate in public discussion in the ways that men do, there are real problems with this response. To be sure, women do tend to provide worthwhile contributions via their traditional feminine role, by working behind the scenes. Often women function quietly as the backbone of an organization, the mainstay of the community, and they may be appreciated for the important service they provide. While there is value in serving in this sort of understated way, when women disproportionately occupy this role, they are unlikely to be given their due in important modes of collective decision making. As we found, by speaking less, women exercise less influence, are viewed as less authoritative, and their distinctive priorities and perspectives are less likely to be aired and heard. As William Lloyd Garrison Jr. said in the nineteenth century of the abolitionist movement, women are an “army of silent workers, unknown to fame” (quoted in Jeffrey 1998, i). If women restrict themselves primarily to these behind-the-scenes roles and forego opportunities to discuss their interests or concerns more openly, they deny themselves the full standing as members of the community. No matter how much they are valued as workers, they are not valued as citizens.

Another counterargument posits that women accomplish as much, or more, by speaking seldom. The power of the quiet, restrained, and judicious approach may be understated by our findings. Or so the argument goes. However, the findings here, and in other studies, are clear: people who speak little are perceived as less influential, tend to perceive themselves that way, and in fact are less influential. Women incur this penalty from relative quiescence even more than men do.

A related counterargument is that by lamenting women’s underparticipation relative to men, we have set up the talkative masculine style as the desirable normative standard and devalued a quieter style perhaps more typical of women. On this view, women tend to talk less than men, and it is men’s tendency to talk too much that is the problem. Rather than find ways to elevate women’s talk time, we should find ways to depress men’s volubility. To that we respond that our baseline for comparison is not only men’s volume, but also women’s own volume when men are not around. Our findings show that women talk less with men than they do with women. In other words, women in fact talk more than men do when they do not face the deleterious interactions of the typical mixed-gender group. By implication, the quiescence we have found among women is not what women do when free of constraints and is not a desirable normative standard for women.⁴

Granted, speaking a great deal, and doing so with a great deal of assertiveness, is not always advantageous. The masculine style can backfire on the speaker and surely is maladaptive to the group if exercised too aggressively. Dominance can turn into clueless failure to read the subtle cues of other members and slide into norm-violation that triggers social sanctions.⁵ Extremely hypertalkative men may lose influence if they deviate radically from the social script of the give-and-take and forego the codes of minimal politeness that discussion presumes.

However, there is little evidence that many highly engaged, talkative people, men included, use overly dominant styles that carry a large net cost. In the most common settings of public meetings, women speak at two-thirds the rate of men, and in doing so, they are not exercising influence by virtue of avoiding the pitfall of assertion run amok. Instead, they are doing themselves, the meeting, their gender, and their society a disservice by holding back too much.

SUMMARY OF THE FINDINGS

If public meetings are crucial to democracy, and are in fact widely practiced and attended, then we need to know what goes on in them. Do women exercise their voice where all voices should be heard? Under what conditions do they do so equally, and as effectively as men?

To generate expectations about circumstances that help or prevent women from speaking and carrying influence, we began with the dominant framework in social science—gender role theory. This theory predicts that the more women in the group, the more the average woman participates and gains substantive and symbolic representation in it. That is, the power of numbers will cause women to speak more; to do so with fewer negative interruptions and with more positive affirmations; to articulate the issues of distinctive importance to women; to state their individual prediscussion preferences about the group decision; to be perceived, and perceive themselves, as efficacious and influential; and to have confidence in their own opinion. In this view, it is difficult for women to gain authority and power when they are few. Conversely, they have all they want when they are many.

We found that this view is partially true, but very incomplete. The effect of women’s numbers depends on the group’s procedures. Low numbers do not doom women to powerlessness, and that is the good news. But high numbers do not always solve women’s problems, and there lies some unexpected bad news.

The procedure we examined in depth is the decision rule. We picked the decision rule because theories of behavior directed our attention to its power to set group norms of interaction. And in our study, too, we found evidence that the decision rule affects how members converse with each other. When we assigned a unanimous rule rather than a majority rule, we found several indicators of an inclusive, consensual norm. The group talked longer. More members articulated their preference over the group decision. The members expressed their preferences more often. Unanimous rule tends to create a more inclusive norm of discussion.⁶ To be sure, level of inclusion varies tremendously, as we explain below. But still, there is sufficient evidence that the rule in fact begets norms of interaction, as we claim.

There are other reasons to study the effect of procedural rules such as the decision rule. The decision rule is a feature of the group that is easy to control and simple to implement. That means that groups are likely to adopt it once they understand its benefits, and it can effectively serve the goal of reform. Finally, the rule is a clean experimental treatment in the study of groups and allows us to draw clear conclusions about the effects of procedure. Yet unanimous rule is often not practical. We will address this problem, and others, later in this chapter. For now, the point is that unanimous rule can have dramatic democratic effects, and that it—and more practical procedures that set in motion the same equalizing norms—can be leveraged into a major force for social equality. However, our key argument is that the rule does not operate alone, and neither do women’s numbers. Unanimous rule helps women when they are few, while majority rule helps women when they are many.

Many studies have suggested that unanimous rule helps the minority, but they focus on preference minorities.⁷ We went a step further and generated hypotheses about how unanimous rule could help a very different kind of minority—a social identity minority. We reasoned that unanimous rule creates norms of inclusion, and that these norms raise the participation and influence of both men and women when they are the numerical minority in the group. And that is generally what we found, but with some important wrinkles.

Because women start at a lower level of participation than men, the elevating effect of unanimous rule carries a paradoxical consequence for gender equality. When women are the numerical minority, the unanimous process raises their voice and authority. Unanimous rule is an aid to democracy, giving more authority to a social identity group whose status remains low in society and politics. But the rule’s equal effect means that when men are the numerical minority, men participate and influence even more than they do otherwise. That in turn increases the gender gap in participation within the group. So unanimous rule is good for women when women are the minority, as the gender gap in their group shrinks, but it is bad for women when they are the majority, when the gender gap in the group grows as men—the numerical minority—increase their participation.

We found consistent support for the notion that unanimous rule protects women in the minority, on a variety of measures of participation and representation. We also found that unanimous rule elevates the participation of numerical minority men, and this means that women in the majority do not benefit under unanimous rule. However, unanimous rule does not consistently affect other forms of men’s representation, such as talk of issues of distinctive concern to men. Unanimous rule also does not appear to make minority men more “masculine” or assertive—they do not engage in unusually high levels of negative interruptions, for example. In fact, unanimous rule with any gender composition prompts members to provide women with a higher ratio of positive to negative engagement while they speak than does majority rule with few women. Thus on the whole, relative to majority rule, unanimous rule acts to include not only preference minorities but also identity groups who find themselves a numerical minority in a deliberation. That boost helps minority women considerably. The same boost for minority men carries negative consequences for majority women’s participation, but not consistently so for their substantive representation. Thus the rule matters a great deal, in combination with the gender composition of the group. Women’s numbers alone do not tell the story of women’s equal representation.

Where gender role theory wins strong vindication is under conditions of majority rule. Here we found that the more women, the more the average woman participates and obtains substantive and symbolic representation. That representation consists of several elements: women gain perceived influence, raise their self-assessment of efficacy in the group, are more likely to introduce into the conversation issues of distinctive concern to women and to mention these issues more frequently, are more likely to experience a warm reception while they speak, are more likely to articulate their own predeliberation preferences, and are better able to enact their preferences in the group’s decision. Why does majority rule boost a woman’s authority when she is surrounded by many women? Because majority rule signals that the more numerous groups—whether it is the preference majority or the social identity majority—are entitled to exercise power. Women benefit from this signal to exercise power.⁸

But again, there are some wrinkles to the happy story of majority rule. Women require a supermajority for some of the benefits of increasing descriptive representation to materialize. By contrast, men benefit from their majority status at a bare majority. In addition, when these majority women advocate for their positions during the discussion, they encounter more resistance from men in the group. These findings qualify the gender role hypothesis somewhat. Majority status is good for women but not in a simple or unqualified way. Put differently, majority rule is good for women when women are the majority, but not as good as it is for men when they are the majority.

We also explored several subsidiary hypotheses of gender role theory. Most important is the enclave hypothesis. We found that all-female groups operate largely as gender role theory expects—as environments that nurture women. These environments allow them more full expression as individuals, they generate longer talk times, they prompt women to engage in mutually supportive interjections, and in these groups women reach more generous decisions that help the worst off. Female enclaves, that is, generally create warm, cooperative, and nurturing interactions and outcomes. However, rarely do these groups display a qualitative difference from majority-female groups (especially compared with largely female groups under majority rule). Enclaves are the best setting for women’s full participation and representation, but not because the absence of men makes them an alternate universe. Rather, they help for the same reason that a large increase in women’s numbers helps. All-female groups, for the most part, merely represent the final increment in women’s rising numbers.

All these effects are relative, so we should be clear about what they are relative to. The central comparison point for the conditions of deliberation is the setting that is most prevalent in the world of politics—majority rule with few women. Recall from chapter 1 that women often tend to compose about a fifth of the members of a deliberating body. That number is found around the world, across the American states, and at nearly all levels of American government. Moreover, even in the instances when women exceed this percentage, they are very rarely anything but a small numerical minority.

In these common settings with conflictual norms and few women, we found a litany of ills. Women speak far less than men, they are almost never viewed as the most influential member of the discussion group; they view themselves as powerless; they very rarely introduce topics of distinctive concern to women or dwell on them; their references are rarely picked up in the conversational thread; they receive little positive reinforcement while speaking; men interrupt their speech with hostile remarks, which in turn depress women’s chances of discussing children and lower their sense of efficacy in the discussion; and finally, women are less likely to advocate for generous measures of redistribution for the group and for society at large. It is little surprise that these settings produce an outcome far out of line with these women’s specific preferences about redistribution, and with the general orientation of most women to help those in need.⁹

Especially badly off in our experiments are women who enter the deliberation with low levels of confidence. In general, women tend to walk into deliberative settings with lower levels of confidence than men, and this confidence disparity is especially harmful in the common setting of majority rule with few women. There, confidence plays a large role in women’s willingness to participate actively, and those with the least confidence rarely speak up at a rate anywhere close to women’s proportion of the group. The combination of low confidence and negative feedback about competence is especially damaging to women’s participation, and women who experience both in majority-rule groups with few women barely join the conversation at all. These results apply to women of all different backgrounds—if anything, the best-educated women participate the least in the conditions where women’s standing is lowest, suggesting that they may be the most sensitive to the signals of women’s disempowerment.

These confidence patterns are not immutable, however. Where women have greater standing—either because of rule or numbers—the effects of confidence on women’s participation evaporate. In those settings, women are also likely to be bolstered by more positive interruptions, and such affirming feedback is uniquely important in building the efficacy of low-confidence women. Women’s predeliberation confidence disparity is not destiny, therefore; when women’s standing increases and when the conversational dynamic is more supportive, both low- and high-confidence women speak up.

Because these findings come from experiments, we provided validation from real world settings. We chose school boards, with their traditionally feminine mission of educating children, and their higher-than-average female representation. This was a hard test for our hypotheses. But we expected that because these are public meetings dealing with matters of common concern to the community, and because they use majority rule, we would find the same pattern as in the experiments—the average woman would only speak equally to the average man when women composed a significant majority of the members. And that is exactly what we found. In addition, when women hold official positions of leadership, those women become a more active part of the conversation, though this beneficial effect does not alter the participatory patterns of the other women on the board.

Gender, and not alternative factors, is at work in these results. The evidence is indirect, because the experiments only took us so far—while we created gender composition and assigned rules, we cannot experiment with the individual’s gender. But the available methods of analysis nevertheless suggest that the differences we observed between men and women are due to being a man or a woman—that is, to gender itself. We ruled out the main alternatives to gender—the person’s and the group’s egalitarianism, liberalism, and predeliberation preferences about income redistribution.¹⁰ We also rule out other possible confounding variables that may be correlated with gender, especially income (more on that below). When it comes to talk time, we located much of women’s quiescence in conditions in which they are disempowered in attributes that are closely connected to gender, such as their confidence in their speaking ability and in their adverse reactions to negative feedback about their expertise. A general discomfort with disagreement had little effect, however. Women appear to be affected by conflict not because it carries disagreement, but because it signals their low relative value in the group.

Furthermore, these differences are produced specifically by gendered norms of social exchange, which affect women and men differently, and not by pure, gender-free strategic considerations that people pursue regardless of their own sex and of the gendered norm around them. Being a woman rather than a man means not only being more egalitarian, liberal, and favoring generous income redistribution to assist the vulnerable, on average. It means something beyond the orientations to politics that characterize many men as well as many women. It means the tendency to be less confident about one’s competence (often unrealistically) and thus one’s lower authority in society (often realistically), to be more sensitive to negative engagement with one’s speech, to avoid conflict or competition, and to be affected by the lack of warm social bonds in the group. We directly examined the extent to which the effects of gender composition and decision rule depend on these core features of the person’s gender and found that they do indeed. And so we can point to gender itself, and not its more peripheral attachments or confounding correlates.

In addition, we can point to the gendered norms, and their differential effects on the average man and woman, rather than merely gender-free strategic motivations as an important cause of the patterns we observed. If people were simply operating as strategic actors, they would seek a group decision matching their prior preference and change their behavior according to the preferences prevalent in the group. But that does not square with a number of the results we found.

Put differently, the strategic model yields predictions that do not find support and cannot explain our results. In the strategic model, gender does not matter in itself, only preferences over outcomes do. For example, it predicts “cheap talk”—the average man, because he is likely more economically conservative, accelerates his talk of women’s distinctive concerns in majority-female groups simply because he wishes to convince the female majority to choose the low amount of a minimum guaranteed income he most prefers. It also predicts that the average woman speaks less in groups with many women and unanimous rule because there, the majority shares her views, and she need not take the trouble to talk. And by the same token, it predicts that the average woman does not divulge her prediscussion preference over the group’s decision simply because she figures out that the majority-male group, being also majority conservative, will never consent to it. Could these strategic motivations to accommodate to the preference majority, or to free ride on its coattails, or to capitulate to it, explain the results?¹¹

The evidence clearly answers in the negative. First, the analyses control on the member’s attitudes toward equality between groups in society, on their liberal or conservative political orientation, and on their preference for a group decision for more rather than less redistribution. Furthermore, the analyses also control on the number of members with these views and on the match between the member and the group’s preferences regarding income redistribution. So the results cannot be accounted for by the notion that the average man or woman changes what they say simply to adapt to the change in the preference majority. Neither can the model account for the interaction of rule and gender composition, or the further interaction of these factors with individual gender. That is, this gender-free strategic model cannot explain why women speak, agenda-set, and influence less than men in their group—and less than minority men do in other groups—but do so only under majority rule. Neither can it explain why female majorities are more powerful under majority rather than unanimous rule, or why majority women are less powerful than majority men. Only a model that accounts for the different tendencies of individual men and women to respond to gendered norms can do so.

Thus the bottom line is that who is present matters, and the norms governing the interaction matter—but most of all, the two matter together. Equal representation for people defined by their social identity comes from either a critical mass where the mass matters or on inclusive norms actively applied to women. These findings shed light not only on gender composition but also on how rules work; the rule depends on who the people are. Rules are not universal in their effect despite being neutral on their face. To know what a rule will do to a group, we must understand the identities of the members.

IMPLICATIONS FOR WOMEN’S PARTICIPATION AND REPRESENTATION

Gender Norms and Gendered Institutions

The fact that women underparticipate men in public meetings is at odds with women’s strides in American society and politics. Women now have more education and have made significant advances in employment and the resources it provides—resources that prompt and enable participation in politics, including attendance at meetings. And so, women’s dearth of participation and influence in meetings presents a puzzle.

The solution to this puzzle is that women continue to be socialized into subordinate positions of authority in society. Women only equal (or exceed) men in modes of participation that follow their traditional gender roles, such as dutiful community member. Women continue to underparticipate men in modes of participation that are not primarily duty driven and that entail leadership over men (see chapter 2). And so, just as women are less likely to run for elected office than comparable men, to follow news about politics, and to develop into opinionated citizens, so are they less likely to speak up on matters of common concern when interacting with men. Even though the gender gap in political participation and engagement tends to be modest relative to the inequalities produced by age, race, and class, in some areas it rivals them (see chapter 2). Even when women have high qualifications, they are much less likely than men to put themselves forward in ways that appear to violate gender roles. Women still earn less personal income, train in less prestigious occupations, and are out of the labor force for longer periods than men. But the continued gender gap in the structural causes of participation, such as full-time employment, occupational status, and income, only accounts for part of the continued gender gap in political participation and engagement. Something besides women’s continued material or concrete disadvantages is at play. That something is gender itself—the same things that constitute people as women also constitute them as unequal participants in politics.

An important component of gender is found in norms, and forms, of interaction. Norms of interaction can be inclusive of women, or exclusive of them. In addition, they can signal a cooperating group or a conflictual group. Women are more likely to attempt to participate in groups where the norm is inclusive and cooperative. In this sense, gender works as a characteristic of interaction.

One implication of our argument that gender works as a feature of social interaction is that gender is more than a trait of the individual. As West and Zimmerman stated succinctly: “a person’s gender is not simply an aspect of what one is, but, more fundamentally, it is something that one does, and does recurrently, in interaction with others” (West and Zimmerman 1987, 140). The context has much to do with whether, and how, gender is manifested; as Bohan argued, “in particular contexts, people do feminine; in others, they do masculine” (quoted in Aries 1998, 77). While we do not view gendered traits as quite so variable as that, our findings confirm that context shapes levels of inequality between people defined by gender categories. We spell out one neglected aspect of that context—procedures. And we show that the group context and composition can build particular gendered interactions, or undermine them.

Our findings are also inspired by and support the theoretical concept of gendered institutions (Thomas in Tolleson-Rinehart and Josephson 2005, 253; see also Duerst-Lahti 2002a, 2002b; Duerst-Lahti and Kelly 1995; Kathlene 1994, 1998; Kenney 1996, cited in Thomas 2005; Ridgeway 2001). If gender is not just an individual difference, a quality of individuals, then it can also be a characteristic of settings, procedures, and norms. Gender composition is exactly such a characteristic. As the sociologist Cecilia Ridgeway put it:

More than a trait of individuals, gender is an institutionalized system of social practices. The gender system is deeply entwined with social hierarchy and leadership because gender stereotypes contain status beliefs that associate greater status worthiness and competence with men than women. Gender status beliefs create a network of constraining expectations and interpersonal reactions that is a major cause of the “glass ceiling.” In mixed-sex or gender-relevant contexts, gender status beliefs shape men’s and women’s assertiveness, the attention and evaluation their performances receive, ability attributed to them on the basis of performance, the influence they achieve, and the likelihood that they emerge as leaders. Gender status beliefs also create legitimacy reactions that penalize assertive women leaders for violating the expected status order and reduce their ability to gain compliance with directives. (Ridgeway 2001, 637)

In this sense the decision rule is itself gendered, meaning that it creates modes of interaction that are associated with feminine or masculine norms of behavior, and that it can produce or mute gender differences and inequalities. That is, a decision rule can implicitly signal to participants the expectation that they engage in modes of behavior that society defines as feminine or masculine. Most importantly, unanimous rule signals the more “feminine” norm of inclusion—all voices should be part of the conversation and the decision-making process. Our finding that unanimous rule aids numerical-minority women makes sense in light of this notion.

A final implication of our findings is for the concept of representation. Scholars have established the utility of different types of representation for disadvantaged groups (for example, Mansbridge 1998). Descriptive representation refers to the physical presence of a social group in the setting of decision making. Substantive representation occurs when the concerns, values, sensibilities, or interests of that group are expressed, acted upon, and ultimately influence the outcome in some way. Symbolic representation is defined as the notion that the group is capable of governing, that its exercise of power is legitimate. Our findings suggest an additional type of representation. In order to obtain substantive and symbolic representation, a subordinate group requires that the conditions of discussion provide the group with authoritative representation. That is, the setting, structure, and interaction norms of decision making must grant group members equal status during the decision-making process. In our view, such authoritative representation is a key aspect of deliberative equality.

Descriptive representation can build substantive and symbolic representation for women, but only if it grants women authoritative representation, that is, equal status in the group. Similarly, procedures grant women the capacity to produce substantive and symbolic representation but only if they equalize women’s status in the group, that is, build authoritative representation. To obtain substantive and symbolic representation—the outputs of the process of representation—representatives must operate with authoritative representation. That type of representation is only loosely connected to descriptive representation, but can build upon it.

The Effects on Men

While we have focused on the representation of women, that is not to deny the legitimate contributions of men, or the value of differences between men and women.¹² Discussion groups may benefit from both masculine and feminine perspectives and modes of action, and in our view, deliberation functions best when both genders contribute fully. However, the problem is that groups are commonly structured in ways that privilege the voices, perspectives, and interaction styles of men, to the detriment of women.

The loss is not only women’s. First, some men share the views that tend to be distinctive to women. Not all men prioritize financial issues above care for human beings, for example. Men are moving partway into traditional feminine roles in society, whether by significantly investing time and effort in the care of their children, taking more responsibility for domestic chores, or even holding traditionally feminine jobs. Economically developed societies are becoming a bit less rigid about gender roles, and if that trend accelerates, views and preferences about politics may follow. Importantly, there is variance within each gender, in the content of people’s views and in their interaction styles. Some women are masculine; some men are feminine. If women are included in the conversation, men may feel freer to express their individual preferences, or to adopt the interaction style they personally find the most comfortable, less constrained by the straitjacket of the masculine gender role. Some studies are already finding that larger numbers of women may shift men’s actions in legislatures, including men’s willingness to advocate legislation related to issues of care (Bratton 2005).

In fact, we find that men behave in less stereotypically masculine ways in the conditions where women are more empowered. When women have greater standing, men share the floor more equally, adopt the language of care for children more often, endorse more generous safety net support for the poor, are less likely to interrupt women in hostile ways, and provide more positive forms of support and encouragement to female speakers. At the same time, men do not give up on their own distinctive preferences when women are empowered—they become more likely to oppose ideas or principles that they did not favor prior to deliberation, and the most talkative men resist women’s more generous preferences. Still, women’s voice may free or encourage men to articulate perspectives that would be viewed as not gender normative and thus stigmatizing for them to voice when women play a smaller role in the group discussion.

Second, the loss of women’s voice is not only women’s loss, because that voice can benefit the group. Of course, there is a great deal of variability within each gender. And in political institutions, women often tend to conform to the expectations of the roles they occupy within an organization, which can wash out much of what is distinctive about women’s preferences. However, because women tend to arrive at the table with some perspectives, values, and vocabularies that are distinct from men’s, when they are silent and those perspectives are not articulated, the group as a whole is disadvantaged. When one gender is disproportionately silent and ultimately regarded as less authoritative, then the group is less than the sum of its parts; moreover, it pays the opportunity cost of failing to become more than the sum of its parts. In other words, the role of gender in group discussion is worthy of study because the voices of both women and men are equally valuable to the common good.

What Do Women’s Voices Say?

What do women’s voices say when they speak? In our experimental data and more broadly, what many of women’s priorities have in common is an orientation toward community. A community is a society of mutual assistance, a stable entity of cooperation. The stability of cooperation depends on a standing decision to cooperate—a norm. The norm is a widely shared standard of behavior by which a person’s eligibility to belong to the group is judged.¹³ Women tend to lean toward this norm in all its permutations. This explains why women are more likely than men to be “civic specialists” even though they are less likely than men to be “electoral specialists,” to use the labels of Zukin and colleagues (2006). Women in American state legislatures tend to use a more consensual legislative style (Thomas 2005). On average, women are more oriented than men toward connection and social responsibility.

The orientation toward community takes various forms. Some forms are geared toward helping others. This help may extend to various vulnerable populations, as we saw in our analysis of the issues that women tend to prioritize more than men. Other forms are geared toward upholding moral standards of conduct. Concerns about slavery in the early nineteenth century, the squalid lives of immigrants in the late nineteenth century, alcohol abuse in the 1920s, and more recently, abortion, share in common not only the needs of the vulnerable but also the notion that people are obligated to act morally. Women were at the forefront of the effort to abolish slavery and the settlement movement to aid the poor concentrated in cities, both movements with strong moral claims about proper behavior (Jeffrey 1998; Sklar 1985). Women’s long-standing involvement with education in and outside the home is a reflection in part of their desire to inculcate standards of behavior in others that foster the health of the community. For these women, education was not only a means to personal achievement or enlightenment, as we tend to regard it today, but a method of cultivation that feeds the common good (Kelly 2006).¹⁴ In addition, women’s efforts to help themselves have often been wrapped up with their quest for moral uplift, justified in terms that appealed to women’s needs and to their obligations to reject moral “vice” (Pascoe 1993, 32). Women have been “key contributors” in efforts to reform government so that it ceases to serve personal, narrow ends and orients to the collective as a whole (Norton et al. 2012, 509). What do prostitution, government, and sewers have in common? All are causes that women took up in the broader mission to clean up the moral standards of the community. In fact, woman suffrage campaigns made explicit rhetorical statements about women’s distinctive inclinations in this direction (Pascoe 1993). Thus over the course of American history, women have taken on causes that not only help others but also tie more tightly the fabric of community around each of its members, including themselves. These ties may bind and constrain, but women are less bothered by this than men, because for women, bindings enable community.¹⁵

Because women are more likely to be communitarians, the groups that include them produce communitarian decisions. Groups where women speak up as much as or more than men are more likely to protect the needs of everyone in the community, including the most vulnerable. As part of their tendency toward communitarianism, women are comparatively cool to the free market and have an affinity for strong centralized government. Even our youngest cohort exhibits a gender gap on the basic notion of government regulation. Women ages fifteen to twenty-five are more likely than men in this cohort to indicate that “government regulation of business is necessary to protect the public interest.” Women are far more averse than men to a hands-off approach that allows each person to keep what they earn without regard for the needs of others. Women are not averse to the notion that people need incentives for hard work—in a community, everyone should do their fair share—but they do not believe that hard work is incompatible with support for the vulnerable or concern for the broader public good. A purely individualistic or libertarian approach to freedom undermines community in ways that are less likely to be embraced by many women.

Women’s Communication Styles

The same orientation to community that shapes women’s views in a communitarian direction explains why women use more prosocial cues in communication, and why they are affected by these signals. A desire to function as an integrated, mutually interdependent unit shapes how people behave as they interact with others. It leads to interaction styles that invite others to participate and to exchange. It is a style that can accommodate differences but works hard to avoid hostility and destructive conflict—the kind of conflict that could distance the members of the group from one another. It emphasizes inclusion and rapport. Women are socialized to be the grease that allows the gears of the group to turn. Because women tend to be communitarian, they tend to do best in communitarian settings. A group where members function as an integrated unit is a group in which women are likely to feel comfortable and content. This is why women tend to participate more, and more freely, in groups where people act “nice.”

Our finding that the level of rapport in the group affects women’s participation and representation in deliberation squares with other studies of political discussion. Political scientist Diana Mutz studied what happens when people talk to a friend, relative, or acquaintance about a political subject on which they disagree (Mutz 2006). Two key ingredients emerge in predicting whether exposure to this political disagreement translates into the benefits deliberation advocates seek. One is the extent to which the person is aware of the rationales for the other’s position. The other ingredient is a focus on social connection. But importantly for our argument, the cognitive side—the awareness of rationales for the conversation partner’s political views—depends on the social side. A person comes to know the other’s reasons primarily by having a “civil orientation” to conflict. What exactly does that mean? It means combining “an acknowledgment of the importance of expressing dissenting views with an emphasis on social harmony” (Mutz 2006, 75). That is, people exposed to uncongenial arguments defensively forget those arguments unless they not only believe in dissent but also care about social harmony. These results are consistent with our conclusion that deliberation is fed by rapport.

Not only do citizens need social rapport in order to learn, but also they need it in order to tolerate. An important benefit that deliberation advocates envision for deliberation between people who disagree is that exposure to disagreement will produce tolerance of that disagreement. Mutz finds, though, that those exposed to disagreement become more politically tolerant in general only if the social connection is there. That social connection can happen if the person is aware of the rationales of their disagreeing discussant—which we have just noted is in turn rooted in a civil orientation to conflict. Or it can happen if the person is oriented toward intimacy with the conversation partner (Mutz 2006, 76, 78). In other words, if people prioritize social harmony or care about maintaining a personally close relationship, then they can develop the kinds of attitudes that liberal democracy wishes to see among its citizenry—an exchange of views in a general atmosphere of respectful toleration. Democracy rests on social connection.

All of this tells us that social factors are crucial in determining the meaning of political conversation. Many people who are protective of their social ties are likely to shy away from disagreement, and this is a tough problem for deliberation.¹⁶ Nevertheless, desiring social harmony does not shut down the exchange of reasons and evidence; in fact, a sense that harmony and disagreement can coexist may represent the most beneficial combination of all.

Most important for our argument, it is a combination that affects women more than men. Settings that promote social ties are especially helpful to women. In these settings, the bonds of community welcome women into full participation in the conversation, enabling both men and women to articulate their views fully, and thus to exchange perspectives fruitfully.

However, gender is not only, or even primarily, about differences in the level of communitarian orientation between men and women. It is also about status and authority. Settings that equalize participation and representation do so in part by elevating women’s status in the group. Because women occupy a lower status in society by virtue of their gender, women tend to enter into discussion with less confidence than men and to be more affected by low confidence. Even more important, they are more sensitive to discussion dynamics that fail to provide positive reinforcement and to negative feedback about their competence. Women’s lower confidence comes into play more powerfully when they interact with men, and all the more so when there are many men.

Is Gender Inequality Bad If It Reflects Differences in Relevant Expertise?

We have been arguing that the gendered patterns of participation and representation in common types of public discussion are problematic for the democratic standard of equality. However, democratic theories often leave room for legitimate inequalities based, say, in relevant expertise, knowledge, or experience. More precisely, liberal theories tend to emphasize expertise over strict equality, because they are more concerned with the true value and rationality of collective decision making. So if women were on average less educated, less knowledgeable about politics, or less expert at the issue under discussion, then liberal theories may not find it so troubling that women speak less.

Such are not the only democratic theories, of course. Group-rights theories hold the reverse order of priorities. For this type of democratic theory, equality is more important to achieve than rational decision making. The good of the average person is a less relevant marker of the good of society than is the good of the average disadvantaged member of the society. In addition, postmodern theories tend to argue that what counts as expertise is itself shaped by societal structures and ideologies that privilege the knowledge and beliefs of the dominant groups. If women are less expert than men, postmodern theories would question why women’s expertise is not considered general expertise and given equitable weight as such. What do we make of these theories, and how do they apply to our findings?

The most problematic theory for our argument is liberal theory, so we address it directly.¹⁷ In our experiment, there is no meaningful difference between men and women on the expertise specific to the issue under discussion. We provided each person with the same detailed, relevant information about the decision they were charged with making—that is, on principles of income distribution. We gave each person this information before they began deliberating. We then gave each person a quiz to measure how well he or she had learned the material (see chapter 4). Men and women scored at nearly the same high level on the quiz. We then gave each person the correct answer to any quiz questions they got wrong, further ensuring that each person began with the same information and was able to draw the same valid inferences from that information. Thus men and women came to the discussion with training that should have generated a level playing field. And yet women underparticipated relative to men.

In addition, the problem of women’s quiescence is not a felicitous result of poorly educated people getting out of the way of their more competent fellow deliberators. Overall in our experiment, education had no meaningful effect on talk time, for men or for women. Perhaps that is because nearly everyone in our experiment had some college education. And that renders most of our deliberators, men as well as women, minimally qualified to talk. And crucially for our argument, in the condition in which women are least empowered—majority rule with numerical minority women—the more educated women actually speak less than women with lower education (see chapter 6). In our sample of school boards, too, a woman’s expertise (in the form of years of service on the board) and formal education had little effect; women do not speak more when they have served longer or are more educated. By contrast, board gender composition played a very large role in women’s patterns of verbal participation, even after controlling for that board’s experience and education. In sum, women’s quiescence does not come predictably from more educated men speaking more than less educated women, and it therefore does not function as the price we must pay for the active involvement of the well educated.

It is not the case that the gender inequalities we observed are produced by their coincidence with competence. The gender gap in participation and influence is not an expertise gap. Men are not the more educated sex, either in our sample or in the nation. Inequality is not a necessary price groups must pay for the rule of knowledge, education, and expertise. There is no good defense of gender inequality in meetings.

HOW DOES OUR RESEARCH DESIGN AFFECT THE INTERPRETATION OF OUR FINDINGS?

Our study draws much of its evidence from our controlled randomized experiment. That choice of research design provides advantages but predictable disadvantages. While we undertook an extensive discussion of these issues in our method chapter, we summarize the main points here in order to review them and address the primary criticisms of our choice. The experiment at the core of our study has some important advantages that go beyond the existing scholarship on public deliberation. It used a larger number of groups than is typical. It recorded individual participants and linked each participant’s utterances to that person’s attitudes and characteristics, allowing us to control on predeliberation opinions and to examine postdeliberation perceptions. We examine situations that we create and over which we have a much higher level of control than other existing studies of political discussion. This allows us to examine the pure effect of gender composition and of institutional rules, something that is very tough to do with observational studies. We can have high confidence in the causes of the effects we observed because we exercised a high level of control over the conditions of deliberation and randomly assigned individuals to these conditions without the possibility that they would choose to select either into or out of their assigned condition. We have created conditions that vary across the full range of descriptive representation, from 0 all the way to 100% women. This also allowed us to see if there are specific breakpoints or thresholds for the effects of gender composition and to locate them precisely. Finally, we have created conditions that simulate a variety of real world situations and examined several types of designs that can be used in structuring actual deliberations.

While the experiment we used allows us much greater control than usual, it is far from perfect. As we discussed in chapter 4 and in further detail in the online appendix, we can control the number of women, but obviously we cannot assign a person to a gender. In addition, when we draw conclusions about the mechanisms responsible for the effect of the conditions, we lack the strong causal inference that we have for the direct effects of the conditions. For some experimentalists, any attempt to explore mechanisms outside of an experimental framework in which our preferred mechanisms are experimentally manipulated is probably misguided (Bullock, Green, and Ha 2010). We agree that the only definitive test of a mechanism is to manipulate it exogenously. However, that does not mean that we must abandon any attempt to say anything about mediators or mechanisms. Our strategy has been to look at different measures of mechanisms that are produced by the experimental treatment. Together, those variables make the case about the likely mechanisms even if we cannot experimentally manipulate those mechanisms. We see that speaking time, influence, conversational dynamics (interruptions and such), and content (words and expressions of preferences) are all affected by our experimental manipulations in ways that are consistent with our theory about status (and not with other potential stories). We see this as good, if not definite, evidence for the mechanism we propose. It is an advance over the existing state of scholarship in the field, and we regard it as valuable for that reason.

While the control afforded by our experimental design is an important advantage, there are predictable and well-known shortcomings from that design as well. Our study has the virtues and liabilities of high control and thus strong internal but weaker external validity. Most importantly, the results come from artificial settings and populations. However, we have taken large steps toward addressing these deficiencies.

A number of similarities between our experiment and actual settings are important to note. The task resembled the task in many deliberative settings in that the members were making decisions about the distribution of resources to themselves and, simultaneously, to others in society. Examples include: town planning (Karpowitz and Mansbridge 2005); school budgeting (Myers 2011); designing a new electoral system for a state or a province, as in British Columbia (Warren and Pearse 2008); reforming the country’s resource allocation (Humphreys, Masters, and Sandbu 2006); allocating local services (Fung 2004; Fung and Wright 2003); and so on. In that sense the task facing our participants is not dissimilar from what real world deliberating citizens may do. While our group decisions were nonbinding outside of the experimental setting, so are the recommendations of many actual citizens’ deliberative bodies. Neither is our degree of control unusual; in many real world settings deliberations are structured and directed by officials or authorities. They take place in formal settings that are not always familiar, much as ours did. In addition, while we assembled people unfamiliar with each other to avoid the confounding effects from familiarity, so do many real world settings. As Jacobs, Cook, and Delli Carpini find, in general, meeting attendees are unlikely to know each other (2009, 72). In sum, our experiment resembled the “real world” in several important ways.

In addition, most importantly, we reported a replication of some key results in actual school board meetings. There we find that the average woman on boards with very few women is disadvantaged, especially with respect to authoritative moves like making motions. This disadvantage dissipates as women make up a larger percentage of the board. To find this effect even among accomplished women who have been elected to speak for others in the town is substantial reassurance that our lab findings are picking up dynamics that can also be found in the real world.

Nevertheless, below we take some pains to think through how various conditions of actual deliberation probably do, or don’t, change the main findings.

HOW REAL WORLD CONDITIONS, AND ADDITIONAL VARIABLES, MAY ALTER THE FINDINGS

Our group size is not uncharacteristic in real world deliberations (for example, Esterling, Fung, and Lee [2009]), but worth further study as a possible contingent factor. A small group size may be more inclusive of women, which may make the rules and composition less influential. But it may also make little difference if the key factor is a formal setting where women’s act of speaking takes on an association with leadership. Because the expectations of it are ambiguous, and yet it holds the potential to play an important role in muting women’s disadvantage, group size is a candidate for further study.¹⁸

A similar question arises about the size of the community in which the discussion takes place. Our results may apply in most community sizes except perhaps the very smallest. We conclude this from Bryan’s finding that women’s participation is much higher in communities of 250 registered voters or less.¹⁹ This suggests that the character of the community as a place where people know each other and, perhaps, interact in more sociable ways, matters to women’s equal participation. That is, the high level of women’s share of participation may be caused by the change from an impersonal to a personal community. This is in line with our interpretation of decision rules. Perhaps the same mechanism we believe is created by a unanimous decision rule is at work in small communities, and in both cases, helps women participate more. More study of the sociable nature of interaction, and specifically an exploration of whether sociability rises predictably with particular characteristic of the community whose members are invited to deliberate, would be a fruitful direction for future research.

Would the dynamics change over repeated opportunities for the same individuals to interact? In other words, do women have more influence and standing when they are part of a group that meets repeatedly? There is no reason to think the results go away with repetition, and if anything, norms grow stronger with more interaction (Cialdini and Trost 1998, 158). Evidence from Bryan’s study of town meetings helps: there is almost no correlation in women’s speaking participation across back-to-back meetings within a town;²⁰ the first meeting may have a high level of female participation while the next one will have a low level, and vice versa (2004, 218–19).²¹ And as we have already indicated, in our school boards data set length of service on the school board had little effect on women’s patterns of participation. All this suggests that repeated interactions with the same group does not increase or decrease women’s equality in participation.

The groups we study make decisions, but not all discussion groups do so. Women’s inequality may be muted when the stakes of the group interaction are low, when the group goal is exploratory rather than decisive or directive, and thus when a member’s authority is less implicated in the interaction.

Some groups are charged with discussing matters that involve opinion or values, while others deal merely with the determination of fact. Groups are prone to various kinds of cognitive biases and suboptimal sharing of information even when the mission is to reach an objectively accurate decision. Socially informed expectations and stereotypes are not held at bay merely because values are bracketed off in the group’s mission. So there is no strong reason to expect our findings, or gender inequality, to differ depending on this aspect of the group’s mission.

Would gender gaps mute online? It is possible that online discussions may mute gender differences, if they mute cues to gender or gender roles. There are too few studies yet to be able to draw conclusions, and very few Americans participate in deliberations online (4% of a weighted national sample, according to Jacobs, Cook, and Delli Carpini [2007, 40]). Contrast this low percentage to the 25% who report participating in a face-to-face forum in that study. Nonetheless, as online communities become more prevalent, online deliberation may be a promising venue for gender equality, and worth further exploration.

APPLYING THE RESULTS ACROSS LINES OF CLASS, RACE, AND AGE

Education and Class

Our sample is highly educated, and we found that the women most affected by women’s low status in the group are highly educated. We also found that all the key results held even when we controlled on the person’s education and income and measures of the group’s class or educational composition. We found that these demographic variables do not account for the results—that is, they are not responsible for the effects of gender, and controlling on them does not alter the effects of gender or rule.²² So in all, the evidence implies that class does not explain away the effects of gender, and that while educated women are more vulnerable, all women are negatively affected.

Still, it is appropriate to ask if the findings from our experiment apply where participants are less educated. One possibility is that we found that women underparticipate relative to men because we examined middle-class women who tend to be socialized to especially feminine norms of behavior. Perhaps working-class women are less deferential to working-class men than middle-class women are to middle-class men. While this is a plausible hypothesis, there is no evidence for it. In fact, existing studies pull in the opposite direction.²³ For example, in Bryan’s study of New England meetings, the share of women’s participation was often far from equal regardless of the town’s occupational, income, or educational profile (Bryan 2004, 217, figure 9.1). Neither does the town’s percentage of working women associate with women’s share of participation at the meeting (Bryan 2004, 224). In other words, there is no evidence that working-class settings produce different patterns of gender inequality.²⁴

Race

Our sample consisted entirely of non-Hispanic whites, for a reason. We expected that nonwhite women and nonwhite men behave differently from white women and men. We mean this in the sense of the overall gender gap, which we expect is smaller among African Americans in particular, and in the sense of the reactions of each gender to gender composition. African American women may be less deferential to African American men than is the case for white women and men. Gimpel, Lay, and Schuknecht (2003, 87, 92) found that teen girls are actually more politically knowledgeable and internally efficacious than teen boys—among African Americans.

Our analysis of Walsh’s data on race dialogue groups supports the findings from our lab study. Although her groups were racially diverse, they behaved very similarly to our unanimous-rule groups, and we found the same pattern of declining average female speech participation there as we did with our experiment.

A more thorough and systematic study, however, is warranted. Unfortunately, even our school board data is very short on nonwhite members. How segregation shapes racial composition and thus the discussion dynamics is a worthy question for future research.

Age

A final demographic of possible relevance for the results is age. The mean age in our experiment was twenty-eight (SD 12). Several implications follow. First, the results tell us that younger women do not escape the ill effects of gender inequality. Gender inequality is not a problem confined to older cohorts. It will not automatically disappear with the passage of time. We saw this when we reviewed the general problem of women’s continued lower participation in politics, in chapter 2. Similar findings from mock jury studies also suggest that the gendered dynamics we found apply to younger people (for example, Golding et al. 2007, who studied college students). So our findings about younger women are no surprise.

Second, it is not the case that the women in our experiment were easily cowed or silenced because they are young. Our standard deviation is 12, which means that a considerable minority of participants were well into their adult lives.

Third, when we control on age in our analysis, the findings hold. That is, the results are not restricted to young women or to groups with a preponderance of young women.

Mistaking Average Female Representation for the Representation of All Women

We have, for the sake of economy, been talking about women as a whole. Yet women are far from monolithic in their perspectives, opinions, and interests. In no way do we mean to discuss women as if they are unvarying and unified. In fact, doing so carries the risk that the representation of more privileged women will serve as a mistaken signal that underprivileged women are adequately represented (Cornwall 2003). Clearly, raising the number of women as women risks short-changing the representation of women from disadvantaged groups (Dunning and Nilekani 2013; see also Strolovitch 2008). Future work should prioritize the question of how demographic and ideological diversity among women shapes the patterns we have documented here. If women with lower socioeconomic status or members of racial and ethnic minority groups are especially badly off in conditions where women are underrepresented, for example, then remedies may need to take these dimensions of inequality into consideration. Or if measures that aid women as a whole also aid more privileged women disproportionately, then additional or different remedies may be needed.

DO THE RESULTS APPLY IN OTHER COUNTRIES, CULTURES, AND PLACES?

Culture

Do norms of discussion vary with culture? How? How would that affect the effects of decision rule and gender composition? How might these effects vary when we move to societies that are more gender egalitarian than the United States? That are less?

First, we begin by noting again that we conducted our experiment in two very different cultural milieus in the United States—one a socially conservative, highly religious community in Utah, and the other a liberal, secular, and wealthy community in New Jersey. We controlled for this variable in all our analyses, and the results we reported thus are not peculiar to one place. That is, the effects of rule and composition are not unique to a particular type of cultural milieu. In places with more and with less gender equality, the same factors shape women’s relative participation and representation.

In addition, our analysis of school board meetings took place in a set of highly varied communities across the United States and replicated the findings from our experiment. This supports our ability to generalize widely within the country about the factors that shape women’s authority in discussion.

Nevertheless, it is worth thinking further about the role of culture, with its behavioral scripts and expectations, when it comes to gender inequality. Let us move for the moment to a very different setting from the United States: the slums of Nairobi, Kenya. There, government services are minimal; the large majority of the population lives without electricity, running water, or other basic necessities; and unemployment is extremely high. But many people, especially women, are members of civic associations that attempt to fill this gap (Kilavuka 2003, cited in Greig and Bohnet 2009).

The behavioral economists Greig and Bohnet conducted an economic “game” style experiment with residents of these Nairobi neighborhoods, gauging their willingness to cooperate with others by the extent to which they choose to contribute private resources for the benefit of everyone in the game. In the study, participants are instructed to decide how much to contribute of the private resource given to them by the investigators, and their sacrifice is rewarded to the extent that other players also contribute their own private resource for the mutual benefit of the players. The researchers randomly assigned women to mixed-gender groups or to all-female groups. They found that in general, women tend to view other women as far more likely than men to cooperate in donating private resources for the common good of the players. Consequently, the women in all-female groups are more likely than women in mixed-gender groups to (1) believe that the people with whom they are making joint decisions will behave more cooperatively, and (2) behave more cooperatively themselves (Greig and Bohnet 2009). In other words, this study replicates our finding that relative to mixed-gender groups, female enclaves are more focused on women’s issues, more generous to the disadvantaged, and more cooperative in their interaction dynamics.²⁵

This and similar replications suggest that the United States is not the only country where women are less likely to participate in deliberation, where women and men have distinct preferences in politics, where women articulate a different voice in deliberation, and where voice matters to the outcome only when women are numerous. Replication in countries with a lower level of social, economic, and political development suggests that the patterns we found are not restricted to highly developed countries.²⁶

Does Gender Inequality in Discussion Vary across Places?

While we have reason to believe that women’s participation and representation respond similarly to the conditions we identify as important regardless of the place, a related but separate question is whether some places are better than others.

We can get a systematic look at the effect of the community’s level of development on women’s participation by looking at Bryan’s analysis of Vermont towns (2004). Women’s level of participation during Vermont town meetings is not affected by the degree to which the town is “modernized” or follows a traditional pattern of social and economic relations. To measure the level of development of the town, Bryan constructed variables such as population growth, percentage of native Vermonters in the population, loss of agricultural lifestyle, isolation from larger towns, and sense of the community as bounded. It turns out that none of these indicators of a traditional community matter for women’s level of participation during discussion.²⁷ This suggests that towns that are more traditional in their politics and economics will not necessarily display higher levels of gender inequality. In our study, too, the political milieu of the town made no difference. We already noted that the factors we varied had similar effects in both liberal and conservative locations. Women on the whole participated no more in Princeton (a more gender-egalitarian place) than Provo (a more gender-traditional place). Conversely, more developed communities and perhaps societies do not automatically leave gender inequality behind. As one review of women’s presence and influence in government and corporate boards concludes, “economic development does not beget female leadership” (Pande and Ford 2011, 5).

What does seem to matter for the level of women’s participation is the general standing of women in the society. Among Bryan’s most powerful findings is that the passage of time, from 1970 to 1998, substantially increased women’s share of attendance, speakers, and speaking turns (2004, 226–27). While the passage of time could indicate many disparate causes, women’s general status in society is among the plausible ones.

More evidence in line with this notion that women’s general status in society affects women’s participation is found in a clever study using a controlled experiment in two societies with significant differences in women’s status. The Masai in Tanzania represent a highly patriarchal society, while the Khasi in India are matrilineal. In the patriarchal society, men choose to enter a competition twice as often as women. In the matrilineal society, the gender gap reverses, and women are more likely to choose to compete than men (Gneezy, Leonard, and List 2009). While this study does not look at discussion, it does suggest the possibility that women’s status in a society has profound consequences for women’s proclivity to participate in situations where status is on the line, perhaps including meetings.²⁸

More generally, what these studies suggest is that the society’s fundamental norms about gender matter. In places and times where women are granted more authority, women participate more. Perhaps they matter because they permeate within the society broadly and trickle down to the interactions that take place between the genders.

DO THE RESULTS APPLY TO ELITES?

We have designed our study to comment on discussion among ordinary people, which raises the question: do the results apply in all levels of government, including the highest?

For example, Senator Patty Murray was the only woman among the twelve members of the so-called supercommittee charged with deficit reduction in fall 2011. The committee used majority rule, the common decision rule in governmental settings. As we noted in the introductory chapter, women are a small minority in the vast majority of legislative settings, and majority rule is the common rule (with variations, such as a supermajority rule). Thus female senators often face the circumstances that we identify as especially deleterious to women’s voice. Does it matter if Senator Murray is alone or with other women on a committee? Does the heralded historic rise of women to 20% of the US Senate matter? More generally, do elites differ from citizens in how they react to gender composition or to procedures and rules?

At first it may seem that the answer is yes—elites are different. Especially at the higher levels, it takes an unusual woman to seek the office in the first place, given the paucity of women in that arena, and an even more unusual woman to do what it takes to obtain it. The women who seek leadership—an activity viewed as masculine—may be highly masculine individuals. They may be interested in the issues men are socialized to prioritize, such as finances or hardware. For example, while a barrister, Margaret Thatcher specialized in taxation—a traditionally male priority. Such pathbreaking women may be comfortable with conflict at the start and relish engaging in it and winning. They may be inclined to use an assertive communication style. Thatcher, for example, tended to use the “language of conflict and confrontation” (Campbell, quoted in Keohane 2010, 133). When pressed to do a “U-turn” on policy, she once famously responded: “You turn if you want to. The lady’s not for turning!” These ambitious women appear to know their mind and stick to it. And they may value their agency as individuals and their ambition to govern above co-operation with others. After all, the vagaries of personality, the idiosyncrasies of personal experience, and the tendencies built in by genetic codes all generate significant individual variation within women, and within men. In other words, some women may be inclined toward views and interaction styles that characterize the male central tendency. Accounts of Thatcher, Indira Gandhi, Golda Meir, and Angela Merkel sustain this view. In that sense, not all women fit our story.

Moreover, once this tough-as-nails, superconfident, highly qualified woman achieves a leadership role, she may not be much influenced by what the men around her do—because she already acts in traditionally masculine ways. These rare women must have talked a fair amount and been perceived as highly influential to get where they did, and they did so in a nearly all-male environment. So in this sense, clearly we can say that there are outliers among women who are not deterred by gender roles, or shaped much by them.

However, in a number of ways even women in the high positions of government may exhibit the patterns we have documented. First, the fact that these women adopt a highly masculine style is consistent with one of our claims. These women behaved in a masculine way in a heavily masculine environment. As we have shown, in conflictual settings in which women are few, those women who do speak tend to articulate masculine issue priorities and views. Relative to women in any other circumstance, women in groups with majority rule and only 20% females speak to the care of others only once for every four times, and speak to financial topics about half again as often as other women. Unlike other women, these women never initiate a discussion of children, families, or the vulnerable.²⁹ They are about 50% more likely than other women to advocate for principles of redistribution that they did not prefer—an indicator of conformity to the masculine group environment.³⁰ And they speak for the same ungenerous safety net that the men in these groups do.³¹ In fact, these women are so eager to conform that they become hypermasculine in their terms of speech; their ratio of care to financial words is about half—or less—of men’s ratio in any condition. Because of these women’s hypermasculine emphasis, the overall ratio of care to financial references in these groups is no higher than in male enclaves; that is, in highly masculine settings, women’s presence has no feminizing effect at all.³²

Consistent with this notion that predominantly male, conflictual environments prompt women to talk in masculine ways and to align their public speech with men’s, elite women in those settings seem aware that their unusual gender is an inescapable part of their public persona. They behave as if they must operate in a masculine fashion, and they attach masculine traits and styles to that persona.³³ The nicknames by which they are known testify to the salience of their female sex and to the juxtaposition of masculine traits. These monikers include phrases such as “Iron Lady” (Thatcher), “Iron Frau” (Merkel), “Ellen-she’s our man” (Ellen Johnson Sirleaf), and “the only man in the Cabinet” (Golda Meir).³⁴ Similarly, Sarah Palin, the 2008 Republican candidate for US vice president, came to be called—and described herself as—a “pitbull with lipstick” (Carlin and Winfrey 2009, 338).³⁵ These examples are consistent with our argument that when women are few and men many, women tend to adopt masculine styles as an adaptation to the predominantly male, masculine environment (see also Dodson 2006; Lovenduski 2005). These pathbreaking women behaved as if their very political survival dictated that they shun femininity, at least in style—do not give in, do not show weakness, do not compromise, do not appear to fail.³⁶ Each one conforms to our story about the effects of gender composition on the contents of women’s issue agenda and policy goals.

Interview evidence specifically supports our argument that many women at the highest levels of politics believe that they cannot get far with a feminine style. Interviews of members of the US Congress reveal a pervasive sense by women that they operate in a man’s world and must conform to its expectations in order to succeed: “The women most admired by congressmen were those who had adjusted most effectively to the workways [sic] and habits of the House” (Gertzog 1995, 65, quoted in Dodson 2006). In her interviews with most of the new members in the burgeoning ranks of female MPs of Britain’s Labour Party, Childs found that feminine styles were perceived as illegitimate in the predominantly masculine environment of the House of Commons. She concludes:

Many of the Labour women MPs considered that the House of Commons was not conducive to women acting in a feminised way. They talked about their perceptions of how their style of politics was considered less legitimate and less effective and they discussed the pressures they experienced to conform to the traditional norms of the House. There was also an acknowledgement of the costs associated with acting like (and for) women. (Childs 2004a, 14)

That is, even highly ambitious and successful women are shaped by the gender dynamics around them.

Furthermore, even high-achieving political women may initially experience such masculine settings as more difficult or hostile than their male counterparts do. That too conforms to our argument—masculine settings are tougher for women, even women inclined in a masculine direction. When two women joined the formerly all-male Senate Judiciary Committee after the Senate election of 1992, widely hailed as the “year of the woman” in American electoral politics for electing unprecedented numbers of women to the Senate, they asked not a single question of the witnesses in the Supreme Court confirmation hearings that this committee held (Mattei 1998, 446–47, note 6). Similarly, as we noted in an earlier chapter, US Secretary of State Madeline Albright wrote that at the beginning of her career, as the only woman in many committee meetings, she was reluctant to speak for lack of confidence. It took time for her to realize that what she had to say was in fact quite valued by the group, and to determine how to “interrupt at the right moment.” Albright says that she “learned that you shouldn’t wait to speak” (emphasis added).³⁷ That is, even at the very highest levels of government, at least some women must go through a process of adapting to masculine environments before they feel equally able to participate and assert their authority. Similarly, Angela Merkel “learned to use ‘hard power’ to consolidate her political position” (emphasis added; Thompson and Lennartz 2006).³⁸ Eleanor Roosevelt put it this way: “all women in public life needed to develop skin as tough as rhinoceros hide” (quoted in Keohane 2010, 144). The process is one of learning—learning how to interrupt, when to take the floor, to brush off hostile reactions, and generally, learning the specific forms of masculine speech acts that instantiate influence when people interact. That even tough, otherwise masculine women at the highest level of achievement and ability feel that they must learn to adapt to masculine environments suggests that some of the patterns we identified here may apply in a variety of settings, from the lowest to the most rarefied levels of public life.

Although women in elite masculine settings may adapt to masculine styles of interaction and decision making, and may shift their agenda toward the issues of priority to men, some may seize whatever small opportunity comes their way to articulate or act on a somewhat more care-oriented perspective than their male counterparts. Socialization into care may exert a sufficiently powerful force on most women’s values that it could set women on a higher level of cooperative or nurturing preferences and priorities, relative to otherwise comparable men.

For example, female senators on the Armed Services Committee appear to place a higher priority on issues of care than male counterparts. The first woman to serve on the committee, Senator Margaret Chase Smith, “crusaded during her House career for sexual equality in the military” and “continued that path when she joined the Senate committee in 1953.”³⁹ In 2013, many of the women on the committee focused on eliminating sexual assault against female soldiers, more so than their fellow males.⁴⁰ Although these women are in a predominantly masculine setting (composing 29% of that committee as of 2013),⁴¹ and even though the committee is defined by the highly masculine task of overseeing the military, they tend to bring a more care-oriented set of priorities than do their male colleagues. As Senator Kirsten E. Gillibrand said to the New York Times, “The men asked all the questions about ships, hardware, that sort of thing. We asked why divorce and suicide rates were so high.”⁴² This is not an artifact of the preponderance of Democrats among the female senators. Republicans have prioritized care issues too: Senator Kay Bailey Hutchison of Texas “focused on military family issues,” and Senator Susan Collins was “an early voice on the sexual assault issue,” a point validated by Carroll (2001).⁴³ Women may not be articulating that agenda as consistently and influentially as they would in a majority-female or in a unanimous-rule setting; for example, Senator Gillibrand’s bill, intended to crack down on sexual violence against military women, was replaced by the male chairman of the committee, who crafted a weaker version supported by three of the seven female senators.⁴⁴ But women may still on average demonstrate a higher level of commitment to care issues than their male counterparts.⁴⁵ These are merely anecdotes, but they are consistent with our systematic findings. Women are usually socialized to care for others, and even comparatively “masculine” women, if to a small extent, may tend to act on this agenda more than comparable men do, when the opportunity to do so opens up.

In addition, when more women enter the political setting, women may act in somewhat more feminine ways, at least with one another. This pattern parallels the behavior of women in our enclave groups. For example, after the number of women in the British Labour Party increased in the House of Commons, women reported a sense of sisterhood and solidarity with one another as women. They tended to provide the supports of friendship to one another. In language that fits our findings well, one MP said that “it does help to have other women who will sympathize with you and understand the predicament you are facing” (Childs 2013, 137). These women tended to regard their interactions with female colleagues as different from the parliamentary culture, which they view as “tough and hard.” Some women said they could “reveal” themselves “emotionally” when interacting with other women and not with men, and felt that the men but not the women regarded them as a “threat” (136). Finally, when women reach very large numbers in committees, the tenor of interaction can change, as we found in our study. A study of German and Swiss parliamentary discourse found that “it is mainly groups with female dominance that are conducive to higher respect levels” (Grünenfelder and Bächtiger 2007, 16). That is, women tend to behave in women’s enclaves within masculine institutions along the lines we found in our experiments—in a more mutually supportive and warm way.

Findings from the realm of the judiciary are also consistent with our argument about the effect of procedural norms on women’s representation. The findings come from rigorous, large-scale studies of US Court of Appeals panels, which consist of three judges.⁴⁶ The vast majority of these panels—68%—are all male.⁴⁷ However, when a female judge is assigned to sit on a panel, the panel issues more pro-plaintiff decisions, even after accounting for the judge’s liberal or conservative ideology.⁴⁸ The effects are not due to logrolling across cases, that is, they are not produced by a bargaining process in which the male judges acquiesce to the female judge in order to gain her vote elsewhere. Further evidence that the effects are due to gender differences between male and female judges is that these decisions occur only in particular cases—criminal procedure, civil rights, employment discrimination, or sexual harassment. The specific nature of these cases is telling—they focus on vulnerable populations with whom women tend to concern themselves. That is, women carry influence, and do so as women, in areas of distinctive concern to women.

This is what we would predict under a unanimous process. Sean Farhang and Gregory Wawro, who helped to pioneer this literature, speculate as much, pointing to the norm of unanimity on courts of appeals. About 95% of cases before these courts are decided unanimously, despite the official majority rule. But how can we know if the unanimous process is at work? While we have no majority-process equivalent in hand, we do know one other fact. Recall that under unanimity, increasing the number of women does little, nothing, or even backfires, because unanimity enhances the power of women when women are few. Importantly for our argument, these studies find that adding a second woman does nothing more. If the norm on these panels were one of de facto majority rule, we would not see this influence by lone women, but we would see women’s influence when they are the two-thirds majority. Again, consistent with our argument, unanimity helps women when they are few but does not allow women to leverage the power of large numbers.⁴⁹

Women in high positions of power may be an exception to our general pattern, but in some ways their actions and experiences confirm our findings. Not all women are feminine—there is much variety within gender. The rare women we have considered here—women who are unusually confident and masculine in their general leaning—are a testament to the variation surrounding averages. In addition, women tend to conform to the role requirements of leadership positions in political and formal settings, which tend to be masculine. Here, their performance is often judged based on masculine criteria of achievement, such as financial prosperity, efficiency, status, or victory over enemies. How-ever, even if these elite women are inclined toward masculinity as a general matter, and even more so when occupying roles judged by masculine criteria, they are also affected by the forces we identified as powerful: gender composition and norms of interaction. As the women in our study did, elite women also seem to adopt more masculine ways of acting and interacting when they travel in predominantly male, highly conflictual, and agentic masculine settings.⁵⁰ In that sense their cases support our argument about the power of the setting and its gender signals to influence individuals’ gendered behavior and the gendered nature of interaction. The institution’s norms of interaction, and its social composition, partly determine whether women face barriers, participate, articulate distinctive priorities and views rooted in care for others, and ultimately, govern.

IMPLICATIONS FOR CHANGE

What Does Not Help

As we have argued throughout, simply adding a few women to primarily male groups is not the answer. There are other, less obvious implications of our work about what does not work, and we take them up here.

It is clear that broad democratic procedures or reforms do not suffice. In Indian villages as in Vermont town meetings, women’s participation remains unequal even as other measures of democracy improve. For example, villages with higher literacy rates pull up the participation rates of some disadvantaged social groups, as one study concludes: “Illiterate, landless and [Scheduled Caste or Tribe] individuals … are more likely to participate in higher literacy villages”—“but not women” (Besley et al. 2005a, 654). These disadvantaged groups not only attend meetings at higher rates than do other groups, but they reap the rewards of doing so—they are more likely to be the beneficiaries of public policy at villages that use these meetings than at villages that do not hold them (Besley et al. 2005a, 655).⁵¹ By contrast with the beneficial effects of democratic meetings on illiterate, landless, and Scheduled Caste members, “it is clear that Gram Sabhas are not a forum for women in their current form” (Besley et al. 2005a, 656). That is, merely holding meetings may not suffice.

In addition, simply increasing the proportion of female local leaders does not increase the level of participation by female residents in local meetings. These already abysmally low rates of participation do move up a bit, but remain extremely low. In West Bengal, for example, they move from 7% to 10% of meeting attendees when villages are randomly assigned to have female quotas for local leaders (Chattopadhyay and Duflo 2004, 982).⁵² Our findings from the school boards underscore this conclusion. Having a female chair of the meeting increases the overall participation rate of women because female chairs, like all chairs, tend to talk more as they lead the meeting. In that sense, female chairs make a significant difference. But the mere presence of a female chair does not transform the participatory patterns of the other, nonchair women, and the likelihood of having a female chair changes substantially with the overall gender composition of the group. While the presence of a female chair enables the participation of the chair herself and is, no doubt, symbolically important, having a woman in a position of leadership does not, by itself, solve the problem among the other women at the meeting or render moot the need to attend to the larger gender composition and decision rules of the group.⁵³

Unanimous Rule

Our results do have two clear implications for remedies. One is to raise the number of women, which we address below. The other is to adopt unanimous rule—or procedures that achieve the inclusion that this rule produces—in situations where women constitute a minority of members. However, there are some difficulties in doing so, which we address here.

One difficulty with our suggestion to adopt unanimous rule is the perception that unanimous rule is uncommon. And this perception raises the question of whether it is practical and desirable. If it were so, then wouldn’t many more instances of it show up? Our response is that the dearth of unanimous rule is exaggerated, and that in any case, its dearth would not necessarily mean that it could not be fruitfully used, in some variation or other. The setting we examine is one where groups are sufficiently small to allow an exchange between individuals. While unanimous rule is obviously impractical in large democracies, it becomes much more realistic where groups are sufficiently small to allow interaction among the members. The jury system is the best-known setting for official unanimous rule. But there are other settings where unanimous rule is used, and/or where it is a good rule to use.

Unanimous rule has been used in a variety of official settings. Prominent international organizations, such as NATO, WTO, and Mercosur, make rules for their members by unanimity rule (Maggi and Morelli 2006). The European Union and the International Standards Organization did so for a time as well, with the EU deciding by unanimity on “sensitive” issues, and by majority rule on “technical” issues (Maggi and Morelli 2006, 1138). In these settings, unanimous rule actually makes more sense than majority rule, because the organization lacks effective enforcement capacity. A recent formal model of voting by the economists Maggi and Morelli “yields unanimity as the optimal system for a wide range of parameters” when the members interact repeatedly and when they cannot be forced to go along with a majority decision (2006, 1138). Many groups hold discussions in which the group outcome is not enforceable, or perhaps enforceable on paper but in reality quite difficult or costly or sensitive to carry out. In fact, some scholars suggest that unanimity is the typical rule in a common type of situation: discussion groups populated by representatives of organizations. While majority rule is typical in committees populated by individual members of the same organization, it may be rare in many cases where members represent diverse organized interests (Scharpf 1989, 154). In addition, rational choice scholarship suggests that in single-shot exchanges where the members have no expectation of future interaction, unanimous rule is the optimal one. That rule is “likely to maximize individual liberty and to increase allocative efficiency” (Scharpf 1989).

The biggest problem with unanimous rule is that it is sensitive to the “reversion point” or “default condition” (Ostrom 1986). The default condition is what ends up happening if the group fails to reach a decision. Some members may like the default condition and thus have little incentive to reach an agreement that leaves them worse off than the default.⁵⁴ Olson (1982) argued that in many cases, unanimity protects entrenched interests, and no decision ends up benefiting them at the expense of the common good. In other words, unanimity rule is biased toward the status quo, even if that status quo is dysfunctional or disproportionately benefits certain interests.

Let us take a closer look at the objections. They can be decomposed into two parts. First, the objection to unanimous rule is that it is a bad rule because it is prone to deadlock. Second, a member or a faction of members who are willing to cause deadlock can essentially use a veto by threatening to deadlock the group, preventing it from reaching a decision, and leading the group to revert to their preference, the status quo. If these minority members prefer the status quo to the majority’s preferred outcome, then the minority can get significant concessions by threatening to refuse to reach agreement. That is, unanimous rule may be better for minority women, but it may be worse for the preference majority, which is subject to the hijacking power of the preference minority.

However, we have no evidence that people operate this way. First, in our experiment, we had no deadlocked groups, although one might argue that the stakes were not high enough that we can learn much about deadlock. Second, an instructive place to look is the place where unanimous rule is used widely—the American jury system. While there are infamous cases of jury deadlock, in fact, juries hardly ever deadlock. And those that do, do not show evidence of hijacking by a preference minority.

The most systematic and rigorous study of hung juries is by Hannaford-Agor and colleagues (2002). These scholars examined official records from nearly two decades, starting in the 1980s, and determined that the rate of deadlocked juries in federal cases was between 1% and 2% per year (Hannaford-Agor et al. 2002, 22). The yearly rates are extremely stable over time and consistent across circuits. Most relevant to the type of discussion we are interested in are civil cases. Rather than determining the narrow question of what the relevant facts are and what objective conclusion about truth they support, as in criminal cases, the jury’s task in civil cases, particularly in applying punitive damages, is closely related to the kinds of discussions about values and political priorities that we are most interested in. In federal civil cases, the hung jury rate is only 1% to 1.5% per year.⁵⁵ One might wonder if these results are unusual because they come from the federal criminal justice system. The authors were able to obtain good data in only thirty jurisdictions for state criminal trials. They found that the rate in these state criminal trials was higher than in federal cases, but still only 6% (Hannaford-Agor et al. 2002, 25, 83). That figure is very close to the estimated 5% provided by the classic study of Kalven and Zeisel, who studied an earlier period yet came up with a nearly identical estimate (1966). Even when we look for the jurisdiction with the highest rate, we find that it peaks at 15% (Los Angeles County).⁵⁶ The “hung” jury, that is, is a rare—and in many places extremely rare—occurrence.⁵⁷

In addition, the fear that a preference minority will hijack the group is also unfounded in the American jury system. Hannaford-Agor and colleagues found that the large majority of hung juries are evenly split at the start of deliberations, while the large majority of nonhung juries had a clear leaning toward one side or the other side (2002, 66). In addition, the rate of hung jury is much higher among evenly split or small-majority straw poll cases than among large majority cases (2002, 66). That is, when a jury has a majority preference, it almost never hangs, which means that the minority rarely hangs a jury. What hangs a jury is the lack of a majority, not the obstinate behavior of a small minority.

We see little evidence that leads us to worry that preference minorities sabotage the group or exert disproportionate influence over it as a result of unanimous rule. It may be that the ability to walk away and prevent the group from making an official decision is easier under unanimous rule, but it does not seem to pose significant problems for the group or for the majority opinion, at least in a jury system.

We did find that unanimous rule is better for minority men than for minority women. That is because when they are the gender majority, women defer to men under unanimous rule—and not because minority men aggressively dominate the conversation. At least, we found no evidence of hijacking attempts by minority men under unanimous rule, either in our analysis of interruption patterns or in the preferences that people articulated for the group decision.⁵⁸ We have been cautious throughout the book to explain that unanimous rule is not a panacea, specifically because it does not always help majority women. However, minority men are evidencing no power plays, and the disadvantages majority women experience with unanimous rule are self-imposed.⁵⁹ Finally, unanimity on the whole helps women without exacerbating male dominance because women are in the minority in so many political settings.

The last point of objection is that unanimous rule is inefficient—it takes too many evenings, to paraphrase Oscar Wilde’s comment about socialism. Efficiency may well be a trade-off with unanimous rule, as the classic work of Buchanan and Tullock (1962) argued. Unanimous rule is less efficient in the general sense—it requires more work to negotiate a decision that everyone can endorse. But Buchanan and Tullock argued that this cost is worthwhile if the members have a need to protect themselves from adverse decisions. When the group could reach a decision highly adverse to members, the members may reasonably choose to use unanimous rule. A loss of efficiency thus is (narrowly) rational under these circumstances.

It is up to each group to decide whether gender equality is worth an increase in tedium. Deliberative democracy certainly prioritizes the former above the latter.⁶⁰ We would argue that any model of robust democracy should attempt to reduce social inequality in participation and representation even if it leads to longer meetings.

However, our argument is not so much about the power of, and need to adopt, a formal rule of unanimous vote. Rather, an official rule is merely one way to achieve the desired process: a conversation based on inclusion. That process is what we are fundamentally concerned with. As Scharpf puts it, aside from formal decision rules, there are “decision styles”: “cognitive and normative patterns characterizing the way in which interests are defined and issues framed and resolved” (1989, 159). The styles Scharpf defines includes the following:

Confrontation refers to competitive interactions in which winning, or the defeat of the other side, has become the paramount goal, and in which the battle can typically be decided only by superior prowess or force. Problem-solving … implies the pursuit of common goals and the cooperative search for solutions that are optimal for the group as a whole. (Scharpf 1989, 159)

While confrontation maximizes the individual’s payoff, problem solving focuses on the collective good. We have argued that decision rules can signal the expectation that a particular decision style is appropriate or will be used by the members. Unanimous rule is likely to signal that members should all cooperate in the outcome and thus generate a cooperative conversational dynamic. A majority rule carries the implicit expectation that the exchange is about who has more votes, and thus those with more power will win.⁶¹

However, in some cases, a style can come about even without an explicit rule as a way of solving problems that would be difficult to solve otherwise. In line with this reasoning, in a number of settings majority rule is the official voting rule, but in practice the group has developed an unofficial norm that comes closer to consensus. In these cases, the decision is reached with the votes of many more than the necessary “minimum winning coalition,” and even with complete unanimity.⁶² As we pointed out earlier, American courts of appeals operate on a strong norm of consensus even though their official rule is majority vote (Farhang and Wawro 2004; 2010). The variety and number of settings where consensus operates implicitly under the shadow of majority rule—US congressional committees and appeals courts, within and among European countries, and some international governing bodies—suggests that our recommendation to institutionalize norms of consensus is not unrealistic or overly onerous. While in many cases the consensus comes about because of a “bargaining” decision style, it is possible that in some cases the underlying decision style is more like “problem solving.” It is this problem-solving style that we are interested in, because it characterizes the kind of interaction process that we believe aids gender equality in deliberating groups.

WHY DISCUSSION FACILITATORS SOMETIMES GET IT WRONG, AND HOW TO GET IT RIGHT

An objection to our basic finding that women are often disadvantaged in common settings is that deliberations are often moderated by experienced facilitators. A good moderator will pursue fairness rather than passive neutrality (Walsh 2007, 303). As Siu and Stanisevski put it, moderators are “important to umpiring the deliberative process and maintaining equal opportunity for the involvement of all participants” (2012, 92). In other words, the objection goes, the presence of discussion moderators may mitigate deep gender inequalities of the sort we have documented.⁶³

We have several responses to this view. First, our findings are meant to apply most directly to groups that tend to lack facilitators trained in inclusivity. This is the type of group that predominates in virtually all political settings. The local school board, the town or neighborhood council, the PTA committee, the legislative committee, the party caucus meeting, the church council, the jury, or even the Supreme Court—none of these deliberative bodies are facilitated by trained moderators. Even when these groups have a formal leader, the leader’s role often does not extend to actively managing the discussion, and certainly not with participatory equality in mind, and the leader is typically not trained in the use of inclusive facilitation.

Second, even trained moderators and highly facilitated discussions can fall prey to gender inequality. Some doubt that there is much of a problem to preempt in the first place. As Collingwood and Reedy put it in their sanguine review of deliberative practices, “critiques of power imbalances [between groups in society] caused by deliberation have little empirical support” (2012, 234).

But mostly, the problem is a widespread sense among advocates of facilitated deliberation that their standard procedures already preempt social inequalities. While many trained moderators believe that they already prevent the deleterious effects of gender without the need to pay explicit attention to gender, there is reason to question this complacency. It is unlikely that most moderators in fact equalize discussion by gender. As Mansbridge and colleagues found, moderators are often not focused on assuring equal floor time for disadvantaged populations, and are not focused at all on gender equality; they are focused on airing various perspectives, but they may not tune in to inequalities by social identity (Mansbridge et al. 2006).

Our findings from dialogue groups second this conclusion. Our reanalysis in chapter 10 of the speech patterns of racial dialogue groups Walsh studied suggests that even when groups are moderated by a trained facilitator, and even if the groups signal their aim to be inclusive while airing differences and making a safe place for all perspectives to be heard, women will speak proportionately less than men if they are a majority. While the gender gap in the dialogue groups we reanalyzed is not large, it is not trivial either, with some majority-female groups exhibiting an inequality ratio of 80%.⁶⁴ That is, instructing participants to air all views and listen respectfully solves the problem of gender inequality only when women are not a clear majority of the group. Thus trained facilitators do not solve the problem with their standard instructions and procedures.⁶⁵

Paradoxically, it is precisely in forums that emphasize cooperation and inclusion that women may experience gender disadvantage from high numbers. Unlike many of the settings in politics, which are characterized by the conflictual dynamic captured by majority rule, dialogues, civic forums, and other settings that emphasize civic duty, inclusion of various views, or cooperation among individuals may attract large numbers of women. After all, as we discussed repeatedly, women tend to prioritize these considerations or interaction styles more than men do. If women speak less than men in majority-female groups that stress inclusion and the need for individuals to be civic-minded during discussion, as our results suggest, then gender inequality is likely a problem in many settings that emphasize this theme.

To illustrate the difference between our argument and the conventional wisdom among practitioners, consider the Journal of Public Deliberation. This journal is a prominent outlet for academic-minded writings on how to conduct deliberations. We searched the journal for any article that focused on the topic of social inequalities within deliberation. We found only one (Kadlec and Friedman 2007). It seems that social inequality is not a problem that the world of applied deliberation spends much effort thinking through. Moreover, in that article, the practitioner-authors write a rebuttal of Young, Sanders, and other theorists who criticize deliberation for neglecting social inequality. They argue that experienced moderators are already able to avoid the problem of social inequality in deliberation through their standard procedures. Their way of neutralizing this problem is to use the approach they have developed to deal with conflict in general. That approach consists of working with existing organizations to set up the list of participants in order to ensure adequate representation of the various perspectives, providing nonpartisan guidebooks that present multiple views on an issue, and conducting the meeting with an orientation of “confluence.” Confluence means to “explore multiple perspectives by focusing together on the examination of an issue from as many vantage points as possible … [it] seeks ongoing input and insight from the range of possible stakeholders in a process that clarifies serious differences as well as potential common ground” (Kadlec and Friedman 2007, 13–14). Thus the approach common to practitioners of facilitated deliberation is focused on airing and encouraging the process of listening to various views by all relevant stakeholders for a given issue or conflict.⁶⁶

The problem with this approach is that it conflates a stakeholder, or a deliberator, with a gender. Airing a variety of perspectives relevant to a given issue will not in itself air the distinctive perspectives of women if the issue does not explicitly deal with women. A group may articulate a variety of views yet still not focus on the needs that women tend to prioritize. For example, groups may articulate both liberal and conservative positions on an issue, thus appearing to satisfy the “various views” criterion, yet spend little time on the issues that women may care more about, thus failing the agenda-setting criterion of equal gender representation. When women are not stakeholders, focusing on including and airing the views of stakeholders will not solve the problem of gender inequality. Nor will that procedure lead to equal rates of speech participation and influence by men and women when women are the majority. In ignoring the gender basis of women’s tendency to speak less and to avoid asserting their views, this approach allows the gender inequalities we documented to flourish. Moderators would need to be trained specifically to address issues of equal social participation and influence.

Our recommendations can thus be useful even for moderators who believe that they already know how to avoid the problems of social inequality. Unanimous rule and the elements of a general inclusive, cooperative process help women when they are a gender minority. But otherwise, it is not enough for moderators to attempt to get a variety of views on the table. Moderators should thus not only monitor the assembly on their “various views” radar but also through the lens of gender inequality (and by extension, other salient social inequalities). They should ask themselves whether women are speaking at rates similar to men, in terms of floor time and speaking turns. They should attend to agenda setting—whether the issues, topics, and populations that women tend to prioritize have received as much focus as those that men tend to prioritize. They should attend to the “warmth” of the group dynamic, and they can monitor the patterns of active support granted to female speakers while they are speaking—avoiding hostility is necessary but not sufficient. And they need to ask themselves whether women articulated the preferences that they walked in with as much as men did (this requires findings out individuals’ private, pre-discussion views).

We have not tested these specific practices, but our expectation, based on the effects of unanimous rule and the processes it sets in motion, is that they can help. We summarize them as follows:

1.  Do not rely on random selection, or allow self-selection, into discussion groups. A virtue of deliberative polls and similar events is that they randomly draw from the population, creating a microcosm within the room of the broader society outside. But while random selection at this phase is beneficial for adequate representation of various social groups, the process should not stop there. And using random assignment into discussion groups, or alternatively, allowing people to choose their group, means that gender composition cannot be managed. It is just such management that designers need in order to avoid the deleterious effects on gender inequality. We recommend that discussion groups be assembled with the purpose of fitting the gender composition to the procedures: majority-female groups for majority-rule and its related process of agency, and minority-female groups for a consensus process.⁶⁷

2.  Make access to the floor easy. Do not put it on a speaker to take the floor. The more assertive a person has to be to take or keep the floor, the higher the level of gender inequality that develops in a mixed-gender group. However, this procedure is not sufficient, because merely removing obstacles will likely not be powerful enough to overcome the strong patterns we documented.

3.  Require each person to speak some minimum number of times. One can invite this type of participation with a ritualized going around the room to hear what each member thinks the discussion should be focusing on, or to articulate what concerns they currently have, or simply to say what’s on their mind at the moment. This has the side benefit of providing a check of the agenda and focusing the meeting on the concerns that will need to be articulated at some point in the process. This “go around the room” can be instituted repeatedly in a given meeting, say, at the beginning, middle, and end.

4.  Enforce turn taking. A moderator’s basic role is to ensure that once a person has the floor, there are no interruptions. Negative interruptions tend to be directed by men at women under conditions of majority rule and few women, and they carry especially harsh consequences for women.

5.  To ensure that no one person monopolizes the floor, the moderator can set an upper time limit on any turn, and a cumulative bound on number of turns. Men are more prone to monopolize in conditions of majority rule with many women or unanimous rule with few women. This technique may alleviate this problem.

6.  Invite members to indicate explicitly when they agree with a statement immediately after a relatively quiet person talks, and why they do so. We found that positive interruptions matter to women in particular. Women benefit from hearing others supporting them while they speak. However, if a moderator prohibits interruptions, this supportive function can be met in other ways. One such way is to build in support immediately after a speaking turn by a woman, especially if she has not spoken much yet, and especially early in the discussion. Of course, members should not be invited to express agreement when they do not actually agree, and establishing group rapport does not mean ignoring or discouraging authentic differences of opinion. The invitation to articulate support for a quiet speaker must not overtake the expression of legitimate disagreement. But asking whether anyone agrees with the statement is a way to invite the expression of agreement where it exists without implying that it should exist. Furthermore, specifically inviting agreement that elaborates on the prior speaker is a way to further the dialogue. Elaborated agreement functions to support women in particular without interfering with meaningful dissent and disagreement, which are crucial elements of high-quality deliberative exchange.

7.  Monitor women’s speaking time, number of turns, and participation early and at decisive moments in the discussion, even if the topic has nothing to do with gender on its face.

8.  Assess each individual’s perception of what the main issues are, what are the more urgent problems for the group to discuss, and what their preferences and attitudes are on these issues and problems. Do so before the group assembles. Then monitor that women are as likely as men to express their predeliberation views.

Raising the Number of Women

Given that many official settings use majority rule, the most obvious remedy is to increase the number of women in decision-making positions. However, for the same reasons that women don’t speak, women also “don’t run” (Lawless and Fox 2010). We reviewed the relevant literature in earlier chapters, and it is clear that women are less likely to choose to run for office despite having identical qualifications to men. Women now make up almost 60% of college graduates and receive slightly more than half of the advanced professional degrees in the United States, yet approximately one-third of that number makes it to the US House of Representatives and to American state legislatures (Crowder-Meyer 2010). The problem is a closed circle—women do not participate because they do not feel capable and valued, and the discussion does not center on issues they care about, but they do not feel capable and valued, and their issues are absent, when they do not participate. How can we break this closed circle? Can the percentage of women running be raised?

Studies of candidate recruitment suggest that there are some negative practices that decrease women’s chances of running and winning, and which could potentially be reformed. An important place to examine is the local level, because people who hold higher office often begin their career at the local level. That is, the local level is a feeder for higher levels. In her large survey of local parties in the United States, Crowder-Meyer found that female party chairs are more likely than male chairs to recruit female candidates for local offices. Local parties could therefore increase the number of women in local leadership positions by appointing more female party leaders who in turn are more likely to select female candidates. In addition, the party leaders’ own social networks tend to be composed of more women than their party networks; by turning more often to their own social network, male party leaders can elevate the proportion of women running and winning office (Crowder-Meyer 2010).⁶⁸

One fact working in favor of the increase in the number of women in leadership positions is that voters are no longer much biased against them, at least not in any direct way. Unfortunately, local party leaders still tend to believe that women are less electable than men, even though that belief is largely erroneous (Crowder-Meyer 2010). Perhaps some headway could be made by informing and educating party leaders about women’s electability.

Finally, male party chairs recruit more female candidates when they look outside the narrow network of party officeholders, which is predominantly male, and turn to networks with high proportions of women. These networks are found in organizations that tend to be composed of many women, such as schools, teachers’ unions, parents’ organizations, and women’s groups such as women’s business or professional associations. Here again we see the potential value of women’s enclaves, or groups consisting of heavy majorities of women and designed to serve the distinctive concerns of women. By implication, local party leaders can increase the number of women in local office by seeking out recruitment pools composed of many women, thereby significantly filling this crucial first section of the pipeline to higher office. Perhaps most practical would be to launch an effort to inform groups in which women predominate of their potential role in supporting and encouraging women to get involved in politics. In that way, women’s groups may help foster in potential candidates the motivation to seek leadership positions, not only by running for elected office, but also in the many civic spaces where people come together for collective decision making.

In the near term, women are unlikely to occupy far more than the current 20% or so of seats in official decision-making bodies. Still, the incremental rise of women from 20% to 40% is not impractical over the next couple of decades. And our experimental findings indicate that the gender gap in participation narrows as the number of women moves above the 20% threshold. Although it often takes a supermajority of women to fully erase gender disparities, the biggest disadvantages can be avoided when women account for more than one-fifth of the group. Should women achieve that rise, their substantive and symbolic representation is likely to improve, even if moderately.

Additional Remedies

If more women obtain political leadership positions, they will change the balance of voices heard at meetings. We saw in our school boards chapter that when a woman chairs a board, that woman plays a meaningful role in the meeting and changes the overall balances of voices heard. Although this benefit does not extend to the participatory behavior of the other women present, women can still help to close the overall gender gap by assuming leadership positions when women are scarce. This will be difficult to achieve, to be sure. As we saw in chapter 3, even in settings where women predominate or that are associated with feminine qualities, such as teaching or clerical work, men are more likely than women to obtain leadership positions. And in our school boards data, we saw that the likelihood of a female chair is closely tied to the overall gender composition of the board. So in a world in which women are still underrepresented on many decision-making bodies, the chances that a woman will obtain a position in charge of a discussion are lower than the chances of a man doing so. Still, lower chances do not mean impossible chances.

In addition, as we saw in earlier chapters, women’s engagement rises with the salience of women’s issues or viable female candidates. This suggests two remedies. One is to increase the salience of powerful female role models. We saw in the earlier chapters that no simple role model theory can explain the pattern of women’s participation. But women who run for visible office and are lauded as such do prompt women’s engagement (Burns, Schlozman, and Verba 2001; Hansen 1997). Similarly, as we saw in chapter 2, women who recalled having politically interested mothers were more engaged. Mothers may be able to make a difference to their daughters’ engagement and leadership activities.

A second possible remedy is to bring women’s interests and issues into the agenda. We saw that women are more engaged and knowledgeable on issues of special concern to them. Perhaps changing the issue agenda will engage women in discussion.

A third way is to integrate enclaves into the process. Enclaves could be assembled before the mixed-gender interaction begins and operate in parallel with mixed-gender deliberations (Young 2004, 50, cited in Siu and Stanisevski 2012, 96; Sunstein 2002b; Dryzek 2000; Mansbridge 1996). Scholars of gender and development have been advocating such remedies for some time. For example, Cornwall identifies enclaves as the single most important measure in raising women’s level of influence within deliberative settings: “The presence of a gender-progressive nongovernment organization (NGO) or women’s organization is a major factor: membership makes women more self-confident, assertive and vocal in mixed gatherings” (2003, 1330, citing Agarwal 1997). Enclaves are not only a remedy for women in developing areas where they tend to be heavily marginalized and subordinated; these can be useful in the United States and other countries where women enjoy a far higher status (Cornwall 2003, 1330; Karpowitz et al. 2009). Enclaves could also be built into the broader setting and not only into the deliberating situation. Women’s groups where women gather regularly to hold conversations where participation is active and influence is more equally distributed could be a partial solution as part of the general life of the institution where deliberation occurs. Our results suggest that enclaves can help, though they rarely make a dramatic difference beyond mixed-gender groups where women predominate.

Of course, enclaves could carry their own downside and potential pitfalls. First, there is the risk of essentializing gender by marking it as socially salient. Second, there is the potential for heightened tension or conflict between the sexes as a result from men’s perception that women are developing in opposition to men. Third, women’s (and girls’) interactions with each other are not all supportive and cooperative. We have emphasized the norm that develops when females interact with each other, and our evidence and others’ does suggest that his norm leans toward mutual support. However, that norm can also coexist with more subtle ways in which females can exercise power over other women. The argument is that because women are socialized to be nice to others, and as a result of a norm or gendered subculture of avoiding overt assertion among girls, when women do feel aggression toward others, they hide it in the guise of niceness.⁶⁹ Although there is not much systematic study of this dark side of women’s relationships, we do want to flag its possibility as a factor to look for in the future. In the final analysis, however, we surmise that these potential problems can be overcome. Certainly, the benefit is worth the attempt to implement enclave-like spaces.

It is also important to carve out some remedies that individuals can undertake to help themselves. Our most important finding regarding the individual-level mechanism behind women’s relative silence is that confidence matters. We found that women low in confidence tend to fare worst in the most common conditions of deliberation.⁷⁰ By implication, women can help themselves by joining programs and undertaking activities that build up their sense of confidence. The intuitive place to start is with simple measures that can increase competence in public speaking and in domain-relevant knowledge. For example, women tend to have less knowledge of politics; this can easily be remedied by occasional news consumption. Some television shows provide the kinds of entertainment that women tend to enjoy and also some informational content as a by-product (Baum and Jamison 2006). Women can raise their knowledge without putting unrealistic efforts into becoming a supercitizen. In addition, women can join organizations that empower women, by intent or as a side benefit. As we saw in chapter 3, such organizations are associated with higher meeting-relevant skills for women, especially all-female groups. Here again, enclaves can be a benefit.

However, while it can be empowering to women to know that they can affect their own circumstances, we do not wish to lose sight of the more important implication of our study: the conditions matter. As Wolbrecht concludes, “the persistent lesser influence and power of women thus draws our attention not to deficiencies of women as political actors but to the constraints of the social, economic, and political structures in which they act” (Wolbrecht 2008, 5).

Men can also do a great deal to hinder or help gender equality. We saw in the analysis of interruptions that men’s behavior is the primary moving part in the dynamics of interaction. By elevating their positive reinforcement of women and decreasing their negativity, men could make a significant difference in creating egalitarian conditions. Even a few positive interjections can make a difference to women’s experience. The remedy does not require an onerous change of habit.

TAKING STOCK OF DELIBERATIVE DEMOCRACY

Long ago, Tocqueville placed civic and town meetings at the heart of American democracy. An important element of “the right of association,” he wrote, “is the power of meeting” ([1835] 2006, chapter 12). By thinking about deliberation in terms of equality, we can come to see it not only as an avenue for rationality but also as an avenue for justice. Many scholars and advocates of deliberation are interested in it because they believe that deliberation can reveal correct solutions to collective problems. Put differently, they believe that deliberators should, and do, seek the truth value of language. They ask questions such as: Is the speech accurate? Is it logical? Is it well informed? Does it lead the group toward the optimal decision for its needs? These are worthy questions, but they are not the only ones we can and should pose. The social uses and meanings of language matter too. We asked not whether speech is informed, accurate, and logical, but whether it excludes or includes, whether it creates social inequality or equality, whether it privileges one gendered mode of interaction and perspective or reflects them both.

When we began our project, there was almost no large-scale study of what talk is actually like in deliberating groups. In his review of the scholarship, Ryfe summed up the situation as follows:

Deliberation … can lead to decisions that not only conflict with expert opinion but also conflict with subjects’ own opinions. … Why should this be the case? We simply do not know. And we won’t know, I think, until we learn more about how people actually deliberate with one another. Surprisingly, this issue remains something of a void in the literature. (Ryfe 2005, 54)⁷¹

We have set out to fill this gap, and in the process, understand why deliberation can fall so short of its ideal of equality.

In addition to a gap in our understanding of the content of deliberation, we have attempted to fill a gap in our understanding of how group-level factors operate on deliberation. Empirical scholarship on deliberation has tended to view deliberation as a collection of individuals rather than as a process shaped importantly by group characteristics such as composition. For example, Luskin, Fishkin, and Jowell argue that the composition of groups does not matter in any significant way (2002; see also Fishkin et al. 2010). And when scholars do look for the effects of group-level factors, they do not always find them. For example, Price, Cappella, and Nir (2002) find no meaningful effects on individuals randomly assigned to one of sixty heterogeneous or homogenous groups. What the findings in our book suggest is that the composition of the group does matter, but in ways that depend on other factors. The key is to understand what those factors are and how they work jointly with the characteristics of the group.

What then are we to make of exhortations to deliberate? Should a society attempt to institute and promote deliberation as a means to enhance democracy? On the one hand, as we saw in chapter 1, deliberation holds out the hope of vital, real democracy—true self-rule by the people, the transformation of narrow interests, the enlargement of civic capacity, the enlightenment of the individual, and equality of voice. On the other hand, these lofty and admittedly idealized goals are obviously impossible to achieve in the real and messy world of politics (Thompson 2008). Our findings show that in conditions resembling many of those in the real world, women do poorly in important respects. It would be plausible to conclude that the public meeting “accentuates rather than redress[es] the disadvantage of those with least power in society” (Mansbridge 1983, 277). As Mansbridge concludes, “participation in face-to-face democracies is not automatically therapeutic: it can make participants feel humiliated, frightened, and even more powerless than before” (Mansbridge 1983, 71).

Accordingly, some scholars categorically reject deliberative democracy. Hibbing and Theiss-Morse put it this way:

Deliberation will not work in the real world of politics, where people are different, and where tough, zero-sum decisions must be made. Democracy in authentic, diverse settings is not enhanced by town-meeting-style participation; it is probably diminished. Given the predilections of the people, real deliberation is quite likely to make them hopping mad or encourage them to suffer silently because of a reluctance to voice their own opinions in the discussion.” (Hibbing and Theiss-Morse 2002, 207)

Hibbing and Theiss-Morse conclude that in place of meetings, democracy should rest on representative assemblies: “representative democracy at least affords representation to those who shy away from the give and take of politics” (Hibbing and Theiss-Morse 2002, 207). For this reason, while they certainly have their differences, Mansbridge and Hibbing and Theiss-Morse advocate a significant role for representative institutions.

Our position is different. Our remedy to the ills of deliberation in practice is to use the power of institutions to address those ills. We have specifically argued that institutions can be used to address discursive inequality. Representative democracy does not cure the ills of deliberative democracy, because representatives themselves must deliberate. One cannot have a representative assembly without assembly. As Hibbing and Theiss-Morse recognize, “deliberation is going to have to be a part of any realistic democratic polity” (2002, 208). So we are back to the same problem we started with. If democracy entails interaction among people, whether among citizens or representatives, then the question is how to design it so that it minimizes social inequality.

One problem that deliberation does not in fact face is conflict avoidance. We found that people who tend to avoid political conflict are no more quiescent than people who seek it. This is not as surprising as it might seem. Political scientist Diana Mutz found that people seek to avoid political conflict only when it comes to the people they know and care about (2006). Many of the settings we are talking about are not restricted to interactions among intimates. Similarly, while Hibbing and Theiss-Morse identify conflict avoidance as a mainspring of Americans’ willful apathy and withdrawal from politics, at the end of the day, they move away from this claim, concluding that “most people are actually drawn to conflict” and “people crave picking a side and being part of a group with a shared concern” (2002, 224). The problem is not that people hate conflict, but that they view their political system as biased or ineffective, and they do not understand the relevance of most issues and the need to attend to the details of policy, or the legitimacy of political perspectives different from their own. But if that is the problem, then the remedy is to design institutions that can teach people and motivate them to think about policy alternatives and the legitimacy of diverse points of view. Deliberative settings could become part of this solution. But again, the question is how to design these settings.

And that is our main contribution to the debate: deliberation can work extremely well—in fact, by some measures gender inequality can be completely erased—under the right institutional conditions. Our findings vindicate the most optimistic advocates of deliberation—but the conditions must be right, and creating them takes thought, and effort.

 

---

¹ Quoted in Weekes (2007, 403).

² Normative advocates and their arguments are reviewed in chapter 1.

³ The Supreme Court specified and affirmed this worth in rulings on jury service (Gastil et al. 2010, 7–10).

⁴ The fact that women also get their way more, and are rated more influential, when they speak less, is further evidence that women’s relative quiescence is bad for women.

⁵ For example, see Ridgeway and Diekema 1989.

⁶ See chapters 5 and 9.

⁷ In addition, as we reviewed in chapters 3 and 4, some studies contradict this conclusion for preference minorities, and the literature on rules tends to suffer from various methodological problems.

⁸ We were able to find some of the breakpoints in the effects of descriptive representation under majority rule. In chapter 5, the breakpoint for the ratio of female to male Proportion Talk lies at 80% women. In chapter 7, women’s frequency of care words increases dramatically above 20%, and again after 40% (and the first mention breakpoint is at 40%). From that point, we mostly run linear models in subsequent chapters after the raw data suggest that the breakpoint is between minority and majority female groups. In addition, in chapter 9 we see that endorsement of nonpreferred principles decreases dramatically above 40% women. Finally, we found that enclaves set dramatically higher floors than all other groups.

⁹ See also the detailed summary of effects on page 334, the section on elite women.

¹⁰ We used statistical controls to test these alternative explanations; future work could also experimentally manipulate these aspects of groups.

¹¹ Here we are addressing some stylized predictions from a game theory literature on information exchange by strategic actors motivated to implement their preferences over a group decision, such as Ban, Jha, and Rao (2012), who predict that in Indian village meetings with diverse preferences, the meeting focuses most on the median voter’s top priorities; that is, the person most likely to decide the vote is the person whose preferences are discussed the most and who then influences the decision the most. We control on preferences in order to address this model. These and related predictions are based on theoretical models such as Austen-Smith and Feddersen (2006) and Meirowitz (2006).

¹² We focus on women because they are the group that suffers most often and most acutely from participatory deficits.

¹³ Definitions of norm tend to miss this part but it is crucial. A norm is not merely a widely shared standard of behavior, it is the standard by which membership is conferred, or withdrawn. This judgment can operate in degrees; membership grows stronger, and the person is deemed more and more a group member, the more they conform to and in turn uphold the norm with respect to others’ behavior.

¹⁴ That is not to downplay other motives, such as the agency of entry into civil society and participating in the public sphere (Kelly 2006). But part of the motive and the rhetorical justification to the entry of women into the public sphere was their responsibility for “superintending the physical energies of children, the development of their moral habits … and the forming of their religious character,” in the words of a prominent supporter of female education, George Emerson (quoted in Kelly 2006, 27).

¹⁵ Women are more moralistic and religious perhaps because they are more communitarian. Women are far more likely to be formally affiliated with a religious institution, to attend services frequently, and to participate in other activities in it (Burns, Schlozman, and Verba 2001, 89). A large majority of women say “religion is very important” and support prayer in schools (106). Women and men join religious associations in equal rates, but among the members, women are far more likely to attend a meeting of these groups (78, table 3.3; also 84–85). Even when women were excluded from deliberations of their congregation and denied a vote in its affairs, they were active as moral agents in it, teaching Sunday school during the nineteenth century and later on serving as missionaries (88). Women’s religiosity is a facet of their general orientation to upholding the community, which to them means upholding its moral standards. Among those active in a religious institution, the most common reasons mentioned for their activity, other than affirming their religious faith (80%), are to “lend a hand to people in need” (77%), and “make the nation better” (67%) (106). Even at a young age and in the most recent decade, girls are more likely to indicate that religion plays a “very important role” in their lives (see http://www.childtrendsdatabank.org/?q=node/302).

¹⁶ And as Mutz (2006) argues, for political participation more generally.

¹⁷ In national samples there is a statistically significant but small gender gap in knowledge (see chapter 2). Not only is the gender gap small, but also the gap in political knowledge is shaped by motivation that, as we showed in chapter 2, is deeply rooted in gendered practices. By implication, changing the gender dynamic can completely equalize political expertise, neutralizing this potential justification for gender inequality.

¹⁸ In Vermont town meetings, women’s share of speakers and of speaking turns moves up with a decrease in the meeting’s size, though the effect is small (Bryan 2004, 223).

¹⁹ Versus those of 250–500 registered voters (Bryan 2004, 227). See online supplementary figures at http://www.uvm.edu/~fbryan/newfig%20X-N.pdf. In the smallest towns (250 registered voters or less), women’s share of speakers and turns is 80–90%; in the biggest towns (above 2,500 registered voters), the share is 59% and 73% respectively (Bryan 2004, 227). Bryan (2004) finds almost no association between community size and percentage of women attending town meetings from 1970 to 1998. That is, small communities elevate women’s discussion participation without elevating their numbers. http://www.uvm.edu/~fbryan/newfig%20IX-E.pdf. It is also possible that town size generally decreases participation; it decreases the overall rate of attendance at town meetings measured as a percentage of the registered voters, as discovered by Zimmerman for meetings taking place in the 1990s, in a study of six New England states (Zimmerman 1999, 165). Town’s socioeconomic status or diversity have either a negative or, when controls are included, no relationship with overall rate of meeting attendance, and the same is true of individual-level socioeconomic status (Bryan 2004, 114–22). When both town and meeting size are examined simultaneously, it is town and not meeting size that matters for women’s share of participation (Bryan 2004, 227–28). In our school boards data set we do not have a good measure of town size, in part because school districts are often not identical to town boundaries; some districts span across multiple towns, while others include only parts of cities or towns. In our sample of school boards, female board members speak somewhat more in rural districts.

²⁰ See figure IX-C, plot 1: http://www.uvm.edu/~fbryan/newfig%20IX-C.pdf.

²¹ Bryan also found that length of residency by the town’s residents also matters little.

²² See results of models that include income and education in chapter 6, tables 1–3 and discussion section, and in chapters 5 and 7–9 we indicate in footnotes that key models are unchanged when education and income are controlled. In chapter 10 we do not have a measure of age for a sufficient number of board members to include it as a control, but we do control for their education and years of experience on the board and find these do not change the effect of board gender composition.

²³ Bryan finds very little association between a town’s education or educational diversity (or any of several other measures of socioeconomic status, or of diversity of socioeconomic status, or of economic situation) and its women’s share of meeting attendance or share of speakers or of speaking turns (2004, 205). This is despite a strong association between the town’s percentage of college graduates and the vote for Vermont’s Equal Rights Amendment, controlling for the Democratic voters in the town (Bryan 2004, 204). See also http://www.uvm.edu/~fbryan/newfig%20IX-H.pdf.

²⁴ A cross-country analysis found that as levels of women’s education rise, and as the community’s overall level of education rises, so does women’s status in politics (Fish 2002). Electoral support for the Equal Rights Amendment granting women equal status in the United States increased with the community’s level of education (Bryan 2004). Relatedly, in a study of city councils, the evidence suggests there is little relationship between a city’s socioeconomic status and the share of female council members (Welch and Karnig 1979; the cities all had populations over 25,000).

²⁵ Further replication in Africa comes from an experiment conducted by Humphreys, Fearson, and Weinstein in Liberia. There, each person had to decide how much of their personal earnings from the study they would contribute to a project to aid their local community (Humphreys, Fearson, and Weinstein 2011). Individuals made the decision independently, but if enough members of the group individually decided to contribute their private resource toward a common good, then the investigators would add a reward, and the whole group—and its village—would benefit. The investigators introduced two treatments: gender composition (individuals were assigned to all-female or mixed-gender groups) and democratic leadership (instituting elected democratic councils for the village). The councils’ treatment strengthens the capacity of village leaders to hold meetings and otherwise coordinate the members’ actions so they choose to contribute their private resource. The study found that in groups not assigned to the democratic leadership condition, all-female groups generated nearly 10% higher collective income for their group and donated 12% more of their private income than did mixed-gender groups. As in our study, there is no effect of individual gender; women contribute more than men only when their group consists of other women. In addition, the democratic reform lifts the mixed-gender groups up to the level of public goods contribution shown by the all-female groups. Mixed gender groups can make high contributions to the public good, but they rely on mechanisms of coordination from elected leaders. By contrast, women’s groups are more generous without the top-down efforts of leaders. These results parallel ours in highlighting the role of gender composition, in finding that all-female groups are particularly prosocial and cooperative, and in drawing attention to the way institutional procedures and leadership, broadly conceived, can overcome the problems of gender composition and lead to positive outcomes.

²⁶ Our results are also consistent with some patterns of women’s participation in Indian village meetings (Ban and Rao 2009). Ban and Rao’s study found that in villages where women’s representation quotas are mandated and where women thus have higher descriptive representation in meetings, “within women’s talk the preferences of women take up more time.” In addition, the Ban and Rao study replicates our finding regarding the importance of talk for power over group decisions; the more that land-owning, privileged villagers talk, the more the outcome matches their preferences. That is, powerful groups—in this case, the landed—dominated meetings, and did so mostly through voice. A similar study of villages found that women and illiterate men are “less likely to both hear of and attend these meetings” (Besley et al. 2005a, 652). See the chapter 4 appendix for more details.

²⁷ When all of these measures are included simultaneously, only out-of-town employment has an effect, but not a large one (Bryan 2004, 225).

²⁸ We do not wish to overstate the case, and there is a dearth of evidence. Furthermore, attitudes about women’s place in society, and indicators of women’s integration (such as women’s labor force participation and education), are only weakly related to each other and to women’s level of descriptive representation in government. See Pande and Ford 2011, 5, and http://www.pewglobal.org/2007/12/05/how-the-world-rates-women-as-leaders/. Women may be educated and employed, and the society generally approves of equal gender roles, and yet women’s leadership status may remain far behind unless women’s authority in society becomes nearly equal to men’s. An additional factor is gender quotas; to circumvent the dearth of women in leadership positions in order to comply with international pressure and norms, even countries with traditional gender arrangements may adopt gender quotas (Pande and Ford 2011).

²⁹ Specifically, see chapter 7, figures 1, 2, and 3.

³⁰ Chapter 9, figure 4.

³¹ Chapter 9, figure 2.

³² Chapter 7, figures 4, 5, and 6.

³³ Salient exceptions such as Sarah Palin may prove the rule—while some commented favorably on her feminine self-presentation, Palin was subjected to a high degree of negative coverage that highlighted her self-presented feminine traits and roles in a negative way (Carlin and Winfrey 2009, e.g., 338; see also the case of France’s Segolene Royale, discussed by Campus 2013, 6). However, some women, such as Hillary Clinton, also include cues to the mother stereotype to soften the perception that they are hypermasculine and thus overly “calculating” or cold, as with the famous “3 a.m.” ad (Carlin and Winfrey 2009, 335–37). Media commentators were not entirely persuaded; as one remarked, “Hillary Clinton didn’t figure it out. She didn’t put a skirt on!” (quoted in Carlin and Winfrey 2009, 338). See the discussion of the “double bind” in Jamieson (1995).

³⁴ Some terms opponents used are not flattering or are meant to evoke derision or hostility, such as “Attila the Hen” for Thatcher or “Old Witch” for Indira Gandhi. For nicknames, see Keohane 2010, 133 and 137–38. Many of these names conform to the “iron maiden” category in Kanter’s typology of sexist stereotypes of corporate women (1977a). Media coverage can feed these perceptions, which sometimes reflect fears of female domination. For example, MSNBC’s Tucker Carlson remarked about Hillary Clinton that “when she comes on television, I involuntarily cross my legs”; Chris Matthews referred to her male supporters as “castratos in the eunuch chorus” (quoted in Carlin and Winfrey 2009, 338).

³⁵ The masculine terms (“iron,” “man”) are contrasted in these nicknames with the feminine (“lady,” “frau,” “she”).

³⁶ To be sure, many also used feminine tropes and appealed to feminine icons, when it served them. Some of these are stereotypically feminine: the nurturing mother (Gandhi, Sirleaf); the moral person untainted by masculine corruption or infighting (Thatcher, Merkel). Some are feminine versions of otherwise masculine traits of valor and fierceness (a mythical Warrior Queen or goddess, in the case of Gandhi or Thatcher). See Keohane 2010, chapter 4 for a concise discussion.

³⁷ http://www.realsimple.com/work-life/life-strategies/inspiration-motivation/how-to-tactfully-speak-your-mind-00100000081879/index.html.

³⁸ Powerful women outside official positions of government testify to a similar evolution. The publisher of the Washington Post during its Watergate heyday, Katharine Graham, wrote that she had to “outgrow” her habit of breaking into tears at work. She “outgrew” a debilitating lack of self-confidence and “deep feeling of uncertainty and inferiority and a need to please” rooted in the assumption that “we [women] were not capable of governing, leading, managing” (quoted in Keohane 2010, 136).

³⁹ New York Times, June 2, 2013.

⁴⁰ Jennifer Steinhauer, “Women in the Senate Confront the Military on Sex Assaults,” New York Times, June 2, 2013, A1. That women prioritized the issue more than the men on the committee was highlighted by Senator Bill Nelson: “When I raised the issue of rape in the military seven years ago, there was dead silence. Clearly they [that is, the female senators] are changing things around here” (ibid).

⁴¹ http://www.nytimes.com/2013/06/03/us/women-in-the-senate-gain-strength-in-rising-numbers.html?pagewanted=2&partner=rss&emc=rss&utm_source=feedly&pagewanted=all&_r=0. Accessed June 21, 2013.

⁴² The senator was referring to her experience on the House Armed Services Committee, but drew the same conclusion about the Senate committee (New York Times, June 2, 2013).

⁴³ http://www.nytimes.com/2013/06/03/us/women-in-the-senate-gain-strength-in-rising-numbers.html?pagewanted=2&partner=rss&emc=rss&utm_source=feedly&pagewanted=all&_r=0.

⁴⁴ http://www.armed-services.senate.gov/press/releases/upload/SASC-RCVs-ON-FY-2014-NDAA-MARKUP.pdf. The committee did not hold a roll call vote on Gillibrand’s proposal, but instead on a weaker rival proposal, with the female vote breakdown as follows: In favor: McCaskill (D), Ayotte (R), Fischer (R); Opposed: Hagan (D), Shaheen (D), Gillibrand (D), Hirono (D). See also http://www.politico.com/story/2013/06/senate-armed-services-committee-sexual-assault-prosecutor-amendment-92674.html. Accessed June 21, 2013.

⁴⁵ Swers’s recent book-length study of the US Senate validates this conclusion (2013). Additional anecdotes abound. Even Thatcher herself took some care-oriented positions at odds with many in her party: she voted for an early bill to decriminalize homosexuality (Thatcher 1995). Similarly, when one of the first female presidents of elite universities, Judith Rodin of the University of Pennsylvania, explained her decision to help a local neighborhood in trouble, she said, “Perhaps I was more determined to fix the neighborhood because I was a woman and a mother” (quoted in Keohane 2010, 132).

⁴⁶ The findings in this section come from Massie, Johnson, and Gubala 2002; Boyd, Epstein, and Martin 2010; Farhang and Wawro 2010; and Peresie 2005.

⁴⁷ This figure was calculated based on the descriptive statistics reported in table 1 in Farhang and Wawro’s 2010 working paper. Of the 881 cases in Farhang and Wawro’s sample, only 278, or approximately 32%, included female judges.

⁴⁸ These effects extend not only to the panel’s decision, but also specifically to male judges’ votes, indicating that the female judge’s influence extends to both male judges, not just to one male judge who votes with her and against the remaining male judge. The assignment process simulates random assignment conditional on judge and case factors, according to Boyd, Epstein, and Martin (2010).

⁴⁹ As in our experimental findings, elite women may benefit from particular procedures and rules. For example, Pearson and Dancey (2011) analyzed floor speeches in the US House and found that women actually spoke much more often than men. We interpret this finding in light of the specific rules that govern speeches on the floor of the US House of Representatives. Turn taking in this setting is highly regulated and controlled. When men do not have the opportunity to interrupt women, and when access to the floor is not determined by the individual’s proclivity to initiate speech, then we see that women may actually out-participate men.

⁵⁰ A telling testimony is Keohane’s, which emphasizes her stereotypically feminine mode of decision making while in a women’s enclave setting and her shift to a masculine mode only slightly distinguishable from her male counterparts when she was the chief executive of a university (2010, 153).

⁵¹ It is not clear if the decision to hold the meeting is the causal factor or if instead there are other correlated factors that are the real cause; these results are suggestive but not definitive about the causal effect of holding the meetings (Besley et al. 2005a, 656).

⁵² The danger of assuming that descriptive representation equals substantive representation is high. As Mohanty writes, “‘the mere presence of women in the decision making committees with-out a voice can be counter-productive in the sense that it can be used to legitimize a decision which is taken by the male members’” (2002, 1, quoted in Cornwall 2003, 1330).

⁵³ As we explained above, simply using the general techniques of facilitation is inadequate. Majority-female groups, which are frequently found in situations where the group attempts to use a norm of inclusion, tend to respond to such norms by producing gender inequality.

⁵⁴ An example from the criminal jury system in the United States may help. If I serve on a jury whose majority wishes to convict, and I wish not to convict, all I have to do is refuse to reach agreement and I get my wish—the jury is deadlocked, the trial is over, and the probability of a retrial in which the accused is found guilty is small. That default outcome is almost as good as the jury voting to acquit.

⁵⁵ We do not wish to make too much of the contrast, because there are other important differences between civil and criminal cases, such as six rather than twelve jurors, and lower standards of evidence in civil cases (Hannaford-Agor et al. 2002, 23). However, Mendelberg (2012) finds a similarly low rate of hung juries in a large set of experimental mock juries using civil cases.

⁵⁶ In addition, data for all sixty-two counties in New York for an eighteen-month period in the late 1990s shows a statewide deadlock rate of 2.8% (Hannaford-Agor et al. 2002, 26).

⁵⁷ There is no relationship between the gender or other composition of the jury and its likelihood of deadlock (Hannaford-Agor et al. 2002, 57).

⁵⁸ What evidence we have appears to be consistent with a pattern of women being more deferential to men (including a willingness to endorse principles other than their most preferred) in unanimous groups with many women. But of course, this could also be women deferring to other women.

⁵⁹ As we noted in earlier chapters, what is most likely happening in this condition is that women read the meaning of unanimous rule to be: do not take too much of the conversation or impose your will. Men may be likely to read it as: each person should maximize their individual agency—note that this is not “dominate the group” but rather “express yourself.”

⁶⁰ Is use of rule endogenous to number women at the meeting? In other words, are groups consisting predominantly of women more likely to adopt unanimous rule, either formally or by using informal procedures that mimic it? Are women less likely to attend meetings with majority rule and secret ballot? That is, it may be that gender composition shapes the rule. If women tend to prefer consensus in order to avoid conflict, then groups composed predominantly of women may be more likely to adopt unanimous rule. Hannagan and Larimer (2010) suggest as much when they argue that female groups are more likely than male groups to converge on the median. Inconclusive evidence comes from Bryan’s study of town meetings (2004). He found that there is no association between the percentage of women attending and the meeting’s use of ballots (a more adversarial procedure than voice vote). The question remains open to investigation.

⁶¹ Some procedures or practices are likely to produce the kind of inclusion that we have identified as a key beneficial process of unanimous rule. One nonobvious way that groups use procedures resembling unanimous rule is the voice vote. The rules that govern Vermont meetings allow decisions by voice vote, standing vote, or Australian (secret) ballot (Bryan 2004). By law most elections of officers must take place by ballot, but most of the decisions the town meetings make are not the election of officers, leaving considerable room to use voice vote. The decision to use a voice vote indicates the expectation that the gathering is largely in agreement and thus approximates the expectation of consensus. The decision to use a ballot may be an indicator of conflict, and its use may thus be associated with the kinds of majoritarian dynamics that go hand in hand with majority rule (Mansbridge 1983). So here we have an approximation of our rule variable—procedures that either resemble unanimous rule in signaling the expectation that broad agreement is afoot, versus procedures that signal the opposite—conflict rather than cooperation. Consistent with our notion that women do not do well under majority rule unless they are a clear majority of the participants, Bryan found that “women did significantly better on verbal participation when the Australian ballot was not in use,” although this result is far from definite given that it goes away when other controls are included (Bryan 2004, 219). We take it as a possibility consistent with our findings that the use of the secret ballot indicates a meeting norm of conflict as opposed to consensus, and the use of voice vote indicates a meeting norm of cooperation and inclusion. As Mansbridge found, the less formal the voting procedure, the more the group operates with a norm of consensus (1983).

⁶² See Shepsle and Weingast (1981) for US congressional committees, the “consociational democracies” studied by Lijphart (1969), and federalism in Germany (Scharpf 1989).

⁶³ Indeed, moderators can exercise a big influence over members’ attitudes in general (Humphreys, Masters, and Sandbu 2006).

⁶⁴ That is, the average woman’s speech is 80% of the average man’s speech.

⁶⁵ This discussion also applies to the findings of Jacobs and colleagues, who report that most of their deliberating respondents indicate that their moderator attempted to ensure that all opinions were heard, and that the goal of the meeting was to reach agreement (2009, 75–77). Attempting to air all views and to reach consensus are characteristic of settings where we find a disadvantage for women when they are the gender majority.

⁶⁶ A concrete case of well-meaning but insufficient instructions is the “Friendly Advices on the Conduct of Quaker Meetings for Business” (Gastil 1993, 51). Quakers are exhorted to practice humility and refrain from dominance. “Let not certain Friends be known for their much speaking. Brevity is desirable in meetings for business as in meetings for worship.” “If thou art tempted to speak much and often, exercise restraint … having spoken on a matter of business, it is well for thee to refrain from speaking again till after others have had full opportunity to voice their concerns … should thy concern not meet with the general approval of the meeting, in common courtesy and in true humility withdraw thy concern that the meeting may act in some measure of unity” (quoted in Gastil 1993, 51). These are classic elements in the consensus process. And the typical assumption is that a group that follows these precepts will automatically protect any type of minority, because it protects any given individual. Our results give a resounding endorsement to that assumption—but only when women are few. Our findings imply that when the members are predominantly female, the female members may hold back far more than male members. Consequently, men are likely to be “much speaking,” “tempted to speak much and often” and not exercise much “restraint” or “refrain” from speaking again,” thereby denying the women “full opportunity to voice their concerns.” Whether this inequality in fact obtains in Quaker meetings, we do not know, but our results from moderated dialogue groups point to the need to closely examine the effects of procedures on social identity minorities.

⁶⁷ Although majority women with majority rule is the combination most likely to achieve gender equality across the various forms of representation we examined, minority women in a unanimous group have trouble moving the full group in the direction of women’s preferences (see chapter 7).

⁶⁸ Similarly, women’s scarcity on corporate boards may be caused in part by reliance on predominantly male networks from which members are recruited (Pande and Ford 2011, 8).

⁶⁹ See Simmons 2002. The aggression then comes out in ways that are no less damaging than overt expressions, but difficult to detect and to combat. The point is not to disagree with our claim that women’s interactions tend to be more mutually supportive; in fact, this alternative view endorses that claim. The point is that the norm of niceness does not mean that women are not pursuing aggression in other, more insidious ways.

⁷⁰ Also, women who received positive signals about their aptitude and had high levels of confidence actually participated at far above the rate of equality.

⁷¹ Similarly, Laura Black recently wrote that “very few studies provide detailed descriptions and depictions of the actual communication that occurs in deliberative civic engagement events … studies could explore some of deliberation’s social dimensions such as … respect, to see how these are communicated and what happens when they do no occur (2012, 77).