CHAPTER 10

Gender Inequality in School Boards

ON JANUARY 5, 2012, the school board of Londonderry, New Hampshire, gathered for its first meeting of the new year. Founded in 1722, Londonderry is a small town in the southern part of the state that boasts of its high quality of life, considerable open space, and excellent schools. The January school board meeting was attended by five board members—four men and Nancy Hendricks. Over the course of the two-hour and ten-minute meeting, Nancy made three comments, no motions, and offered one second to a motion made by another board member. The four men at the meeting accounted for all the other actions the board secretary recorded. Although Nancy Hendricks was 20% of the board, her motions and comments accounted for less than 9% of the speaking turns that evening.

Just a few days earlier, on December 21, the school board of Wakefield, New Hampshire, held one of its last meetings of 2011. Wakefield, like Londonderry, can boast of its long history. Founded in 1774, it advertises its “New England Town Charm” and high quality of life. Its school board is also made up of five members, but unlike Londonderry, four of the elected officials are women. Over the course of the three-and-one-half-hour meeting, the women together accounted for 84% of the speaking turns and 88% of the motions made—both numbers slightly higher than their proportions at the meeting.

Our aim in this chapter is to explore the experiences of men and women in Londonderry, Wakefield, and many other towns and cities across the country. To this point, our focus has been almost entirely on our experimental data. We have done so for good reason: with their random assignment to group conditions and carefully controlled lab setting, the experimental data give us high levels of internal validity about the causes of women’s and men’s behavior. They also allow us to explore systematically what transpires during discussion and to measure the perceptions and views of the participants before and afterward. We explored a large number of variables, and did so with high confidence that we are tapping into causal relationships in a meaningful way. But we also care about external validity. That is, we want to be sure that the patterns we found in the lab can also be seen in the real world of deliberating groups, including groups discussing issues other than the redistribution of income.

We have reason to believe that the gender gaps we found in our experimental groups are likely to be found outside the lab. First, natural settings vary in women’s level of speech, one of our key dependent variables. In his study of New England town meetings, for example, Frank Bryan found that the percentage of female speakers ranges from 5% to 68% and that the ratio of women’s percentage of speaking turns at the meeting to male percentage of turns ranges from 0.23 to 3.37. Second, many natural settings seem to produce a significant participatory disadvantage for women. On average, in Bryan’s study, women comprise 47% of meeting attenders, but only 36% of the speakers and only 28% of the speaking turns.

This chapter thus has several goals. First, it seeks to show that the findings we obtained in the lab, under controlled but artificial conditions, replicate with actual deliberators and in the real world, where some groups meet many times and the members get to know one another, and where people make statements and reach decisions free of the obtrusive gaze of an experimenter. Second, we wish to see if women obtain more substantive representation when the agenda deals with a women’s issue. Women have long been active in education institutions, and their levels of engagement and knowledge on school board matters equal men’s, as we detail. Therefore, examining the effect of gender composition within school boards constitutes a hard test for our hypothesis that women need particular circumstances to participate and influence as effectively as men do. In addition, our experiment examined groups without a formal leader. But in the real world, many groups assign formal leadership positions to their members. Our school board data allow us to examine whether women’s formal leadership position improves women’s participation and influence, either directly or in combination with their descriptive representation. Our school boards study thus introduces a factor we did not vary in our experiment—the gender of the group’s formal leadership—and allows us to test the limits of our findings by examining settings focused on children’s education, a topic of distinctive concern to women.

There are two things that this chapter does not do. First, it does not directly compare the effect of the decision rule within our sample of school boards because all the boards use majority rule. We thus test our hypothesis about numbers under majority rule only, though we turn to the question of unanimous rule with a second data set of community meetings. Second, we cannot provide a rigorous causal analysis of these meetings. This chapter aims at external validity; strong causal inference and high internal validity was the aim of the experimental chapters.

Strikingly, these new data sets reveal precisely the same trends we found in the lab. Women who serve on a school board where men outnumber them participate far less than their already low proportion of board members. When women comprise a majority of the board, their participation comes close to equality, though it takes a supermajority of women before the voices of men and women are balanced. Thus even among an arguably elite sample of women who were elected in order to speak up for their constituents, we find that when women’s authority is low, women are the silent sex.

SCHOOL BOARDS IN AMERICA

We have argued from the outset that the sorts of small group discussions that were at the heart of our experimental work are ubiquitous and important in American political life, from the local to the national level. Perhaps nowhere is that better evident than in the regular meetings of school boards held all over the United States. At these meetings, Americans gather, discuss issues of importance to the local educational system, and make collective decisions that are binding for those within the school district.

School boards are notable for our purpose for several reasons. First, there are lots of them. To be more specific, there are over 14,000 in the United States. This large number means that boards are a common venue for local decision making. It also provides us with many potential cases. The plethora of boards helps with our particular purpose, which is to study what happens inside boards that differ in their gender composition. With so many boards, we are likely to find enough variation all the way from 0 to 100% female.¹ Then we can say what difference composition makes. As Reingold notes in her review of the literature on women in office, few studies of women’s representation examine variation across decision-making bodies that vary in gender composition (2008, 140). The variation that 14,000 potential units afford is far higher than the variation afforded by the fifty state legislatures, the most commonly studied setting by scholars interested in gender composition (Reingold 2008, 140). While we cannot analyze the contents of 14,000 board meetings, we can leverage this high number for a sample with adequate variance in the independent variable.

Second, school boards are among the most important public institutions in a community. Boards are authorized to and in fact direct and oversee the main educational institutions in the community. Their budgets are often substantial by local standards. And they are quite powerful. They have the ability to enact budgets, enter into contracts, issue and campaign for bond issues, handle all manner of constituent needs and complaints, and generally provide leadership on educating the young (NSBA).² In other words, the decisions they make involve real power, with important implications for children, parents, and the larger community. Boards are active and meet often; over 90% of boards meet once or twice a month (NSBA table 48).

Third, a large majority of boards have either five or seven seats (NSBA table 39), making boards a good place to see if our five-member lab group findings apply in real settings. Board meetings are one example of the sort of group deliberation we had in mind when designing our experiment, and the number of board members parallels well the lab setting we constructed.

DO WOMEN PARTICIPATE MORE WHEN WOMEN’S ISSUES ARE ON THE AGENDA?

Our board study also differs from our lab investigation in one important way: the subject of discussion is always, in some sense, the welfare of children in the community. Whereas our lab groups could choose whether to frame redistribution as about children, families, the poor, and the needy, or instead about taxes and the work ethic, board members are nearly always dealing with what might broadly be considered issues of distinctive concern to women: education and the needs of children. As we have presented our lab findings to other political scientists around the United States, one frequent response is to assert that the dynamics we found in the lab would have been different if we had picked a different topic, one that was more centrally focused on the issues women care most about. A focus on school boards allows us to examine this proposition.

The education of children has long been viewed as women’s purview. Women were allowed to teach Sunday school in churches even as they were barred from voting or even discussing church matters (Burns, Schlozman, and Verba 2001). Women compose the majority of PTA groups.³ Women thus are likely to view the topic as a close fit with their traditional gender role. They may feel qualified and confident in speaking up in school board meetings.

Consistent with this notion, there is no place where women find higher representation: about 40% of board members are women (Burns, Schlozman, and Verba 2001, 104; Donahue 1997; NSBA). That is about double the female percentage in city councils, state legislatures, or Congress (Crowder-Meyer 2010). Thus women may view themselves, and may be viewed by voters, as qualified to serve on boards, to judge by their relatively greater numbers there. Furthermore, board members are well educated and possess financial means, so the women who serve on boards are not likely to suffer from a deficit of participatory resources (NSBA). If women participate equally, and have equal authority and influence, anywhere in elected office in the United States, we will see it here.

Women may engage in public discussion when it deals with the spheres implicating women’s role as women. Consistent with the notion that women engage when the issues speak to them, Delli Carpini and Keeter found that the sizable and consistent gender gap in political knowledge shrinks considerably or disappears—and even reverses—for some issues. These are issues that reflect women’s “special concerns” (2005, 40). And perhaps for the same reasons that women are far more likely to serve on school boards than in any other legislative setting, women are more likely than men to know who their school superintendent is (Burns, Schlozman, and Verba 2001). Women close the knowledge gap with men on issues of local politics (Delli Carpini and Keeter 2005). When discussion is more focused on women’s distinctive issues, then women may accelerate their participation in it.

Will the fact that the board is always talking about one giant women’s issue—education—mean that gender differences on school boards are muted? Some argue exactly that. On this view, local politics is different because the topics of discussion and decision making tend to consist of some of the issues that women often prioritize, such as education. This in turn means that women and men are equally likely to care, to feel qualified, and to participate. Tolleson-Rinehart finds in her study of mayors that at the local level gender differences are in fact muted (in Carroll 2001). Similarly, Donahue studied school committees and came to that same conclusion (Donahue 1997).

Others claim that gender differences at the local government level will be apparent, but in women’s favor. Janet Boles found that among county and municipal legislators, women were more likely to take leadership roles than men on women’s issues and establish new programs on these issues. Susan Beck studied women and men in local town councils and found that women tend to be more responsive to constituents and more persistent in their questioning.⁴ But no one has looked at, or reported, the effect of gender composition in local settings (see Hannagan and Larimer 2011a, 2011b).

If gender differences are minor when the agenda is fixed on a women’s issue, then by implication gender composition may matter little. Do women participate equally to men regardless of their numbers when the agenda is permanently fixed on a women’s issue? We don’t think so. Bryan finds that women’s speech participation in Vermont town meetings is no greater when the meeting agenda includes school issues (Bryan 2004, 220). Not surprisingly, there are some conflicting studies. Rosner found more women’s participation in kibbutz meetings when the topic is education (Bryan 2004, 220, note 11). And Bryan suggests that women may participate more in discussions of education when it is not formally on the agenda. However, the evidence is nuanced and mixed (Bryan 2004, 250–52), and the evidentiary base is thin.⁵

In any case, training our analytic focus on the ways in which the gender composition of school boards affects the participation of women allows us to examine this question directly. Perhaps more importantly, the analysis of school boards is, we assert, a tough test for our theory. Not only is the topic of discussion women’s issues, but also the women who comprise the school boards are elected officials. They have put themselves forward as representatives of their communities and are, presumably, willing to speak up when issues of importance to them are on the agenda. These women are well educated and above the median income (NSBA). If we continue to find a relationship between women’s participation in group discussion and group gender composition, even among a group of elected officials discussing issues of special importance to women, then we will count this as important evidence on behalf of our theory.

POLITICS AS A MAN’S GAME—OR A WOMAN’S

Burns, Schlozman, and Verba’s PROPA offers another compelling hypothesis that we could not test in our experiment but can test on school boards. PROPA provides a terrific analysis of the mobilizing, enlightening, and empowering effect that female leaders have on ordinary women. Other studies second this finding and extend it: the more women run for a high-visibility office, the more that women voters in that election learn about their leaders, the more that they feel efficacious, and the more that they attempt to converse with and persuade others (Hansen 1997). There is something empowering for women about other women playing what is widely still perceived as a “man’s game” (Burns, Schlozman, and Verba 2001, 334). What happens when women find themselves in a situation with visible, powerful women—that is, when politics becomes more of a woman’s game?⁶

One thing that does not happen is a simple role model effect. It is not the case that any woman on the ballot moves women. The females running for office have to be visible and viable. This can happen because they get a lot of publicity, or because the office they seek to obtain is visible. Either of these may be needed before the election can send a signal that politics is a woman’s game. When women run for office in a widely hailed “year of the woman,” as they did in 1992, they have a big impact on other women. But when they run in other years for lower-visibility offices, such as the US House of Representatives, their effect on women’s engagement, such as women’s attempt to “politically proselytize”—that is, to persuade someone about politics—is null (Hansen 1997).⁷ As Atkeson summarizes, “Viable symbolic representation enhances political engagement and may increase substantive representation, but token symbolic representation does not” (Atkeson 2003, 1053).⁸

In other words, what women may be doing by running for or holding important offices is resetting the norm about who is entitled and qualified to lead the polity. This is a lesson that women take in and act on. We may find a similar effect in small groups. When a woman occupies a visible powerful position in the group, such as chair or president, she may be implicitly signaling to other women that the meeting is not a man’s game.

Just such evidence comes from village meetings in India. A study by Beaman and colleagues (2010) takes advantage of the fact that some local leadership positions are reserved for women on a random basis.⁹ The study finds that a woman attending the meeting is 25% more likely to speak at the meeting when the local leader position is reserved for a woman. Whether this effect obtains because women see female leaders at the meeting or because of broader or indirect changes brought about by instituting quotas for female leadership is unclear. Regardless, this finding supports the notion that a visible female leader elevates women’s participation.

Further support for the notion that authoritative female figures affect women’s engagement comes from Atkeson and Rapoport’s study of women’s opinionation (2003). As we noted in chapter 2, they found that women are less likely than men to respond with an opinion, or to respond with as many opinions, when asked what they liked or disliked about the parties and candidates of the day. But the gender gap is smaller for women whose mothers were more interested in politics (at least, as best as they could recall, though recall is not a great measure). No such effect obtains for men or from fathers. In other words, politically interested women may confer on their daughters a socialization effect—perhaps, a sense that they can, and are expected to, be politically active (see also Delli Carpini and Keeter 2005). Again, this finding prompts us to ask what expectations and sense of entitlement to participate may be set for women by authoritative women around them.

THE SCHOOL BOARD STUDY

We sampled board meetings widely in order to minimize a correlation between gender composition and other variables. For example, in a recent national survey of members, small districts are more likely to have male members (two-thirds of members in small districts are men), while other districts are as likely to have men as women (NSBA table 1). Small districts are also more likely to have conservative members, white members, and slightly less wealthy members (though the central tendency is the same as in other districts—members in all districts are likely to be nonconservative, white, and wealthy). So it is important for us to look at equal numbers of small, medium, and large districts within each gender composition to avoid confounding gender composition with these other characteristics of boards.¹⁰

In addition to size and rural character of the district, gender composition may be correlated with the general conservatism of the district. Also, the effects of gender, and of gender composition, may differ in conservative and liberal settings. For example, Dolan and Ford (1995) found that female state legislators in the South were less “feminist” but more devoted to the issues traditionally viewed as women’s province and those we have identified as “women’s issues” in an earlier chapter—children and family.

Our sampling strategy is based on access to minutes from board meetings. To that end we first searched all fifty states, using the National Association of School Boards and the state Departments of Education to produce a comprehensive list of districts within each state. We then focused on states that provided links to district web pages. We found that thirty-nine of the fifty states offered such links.¹¹ To further supplement our population of available boards, we contacted BoardDocs, a private company that provides website and document hosting services for government entities and school boards all over the country. BoardDocs provided us with a list of all school boards who use their service. Using both the state school board information and the BoardDocs list, we undertook a search for school boards that post their minutes online.¹²

We then identified school boards whose electronic minutes included sufficient detail and information on our key variables.¹³ We sought boards with meeting notes that were detailed enough to record the presence of board members; the speaking turns for each board member; and the motions, seconds, and votes that occurred during the meeting. Fortunately, most boards do post minutes online, though there is considerable variation in the level of detail of the meeting minutes. Our data set includes only meetings with sufficiently detailed official minutes. All told, we have eighty-seven meetings from twenty different states. Nearly all boards in the sample are elected, not appointed (97%), and nearly all of those elections are nonpartisan (95%), meaning that candidates for office did not label themselves as Republicans or Democrats. Full descriptive statistics for the boards in our sample can be found in the online appendix table C10.1. Our sample of meetings contains school boards located in all different regions of the country, including the South. However, because our aim was to find sufficient variation in the gender composition of the boards we analyzed, we are mindful that our sample is not randomly drawn and may not be representative.

Our data collection efforts involved careful attention to the gender composition of the meetings. As we identified school boards with useable minutes, we sorted them into gender composition categories—0% female, 1–20% female, 21–40% female, 41–60% female, 61–80% female, 81–99% female, and 100% female. Within each category, we randomly sampled meetings with minutes that were sufficiently detailed to code. Our procedure resulted in at least eight meetings in every category, with a larger number of boards in the middle categories of the distribution.¹⁴ In the analysis that follows, we use the exact percentage of women attending the board meeting; our five categories were only for purposes of collecting a diverse sample.

As Reingold notes, scholars of women’s representation have been especially interested in settings where women are a small minority. For some, that is operationalized as less than 15%. In some studies, token women are actually more likely to actively pursue women’s interests and carry more influence in doing so than are more numerous women (Bratton 2005; Crowley 2004). Bratton argues that in legislative settings, unlike other settings, women may be viewed as experts on women’s issues and gain authority as such. Our own lab findings suggest that women’s influence will depend on their percentages in ways that are not necessarily monotonic. The 50% mark is important for some measures of substantive representation. A supermajority is needed for yet others. Our sample includes forty-two boards where women are a minority, thirty-nine boards where they comprise a majority, and six boards that are evenly divided. And finally, we need enough enclave groups to gauge the effects of gender-homogenous dynamics. To that end, we sampled enough groups to fill finely sliced categories of gender composition, including nine all-male and nine all-female groups.

Existing studies also show that token women may have a higher level of perceived responsibility for acting on women’s issues and think that if they don’t, no one else will do so, while more numerous women have a more diffused and weaker sense of personal responsibility (Carroll 2002; Reingold 2000). We are specifically interested in a subset of this “small minority” category—tokens. As we found in our lab study, being the lone member of one’s gender can affect a person even more than being a small minority with company. Our sample includes twenty-seven boards with gender tokens—fifteen males and twelve females. Another threshold may be around a quarter. Reingold puts it this way: women will “avoid acting for women until they are surrounded by a critical mass of female colleagues,” and the “transformation” of the institution and its politics will be “unlikely until women constitute at least a substantial minority (20–30%) of officeholders” (Reingold 2008, 140).

Along with numbers, we examine the official role or position within the board—something that was not possible with our experimental data, where no group members were given a formal leadership position of any sort. As scholars of women’s representation note, women’s power within the institution depends on occupying the formal roles that grant that power. Cindy Rosenthal, for example, finds that as women’s numbers increase in state legislatures, male committee chairs become “less inclined toward such integrative behaviors of leadership as collaboration, inclusiveness, and accommodation,” while women chairs become “more likely to embrace these integrative strategies” (in Reingold 2008, 141). Rosenthal’s online survey of state legislators found that most women believed that men in leadership positions “forget to include women” and “discount women’s advice” (2005, 211, cited in Reingold 2008, 141). In other words, when it comes to school board meetings, women’s increasing numbers may make the dynamic less friendly to women if men run the meeting and more friendly to women if women are chairing it. In any case, we want to attend carefully to how the dynamics of participation change when women run the meeting, as compared to boards where men are in charge.

For each board meeting in our sample, coders recorded the presence or absence of each board member, the board member serving as chair or president, and each speaking turn, including motions, seconds, and other recorded comments, identified in the official minutes. Our focus is on the participation of elected board members, not any other speakers, including school superintendent, staff, or citizens. Clearly, our data set is not the same as the careful recording of every single utterance that occurred in our lab setting. We expect that board clerks or secretaries tend to focus on major interventions from board members, not small asides or comments that our recording equipment caught in the lab. But we do find that a sufficiently large number of boards record a great deal of detail in their minutes, and a few boards make full transcripts of their meetings available. Thus for every board meeting, we know the gender composition of the meeting attenders, the identity of formal board leaders, and several types of verbal participation. For most boards meetings, the length of the meeting was also recorded. A sample of such detailed minutes is included in online appendix H.

The official minutes typically reference three forms of verbal participation: motions, seconds, and recorded comments. Motions are formal interventions that move the board members to a vote or propose some other official action. In one sense, these are the most authoritative forms of verbal participation, as they are the first step needed for the board to make a collective decision. They are also easy for those keeping the meeting minutes to record, which means that this measure is likely to be the most reliable. Recorded comments are general speaking turns about a topic under discussion by the board. Because our data are culled from the official board minutes and we have only a small number of full transcripts, we rely on the board clerk or secretary for this measure, and we are unable to tie the clerk’s record of the comment to the length of each speaking turn. And comments during executive sessions—portions of the meeting that are not open to the public—are not recorded in any of our boards. Nonetheless, these are verbal interventions in the public portions of the meeting rising to a level of prominence that the clerk felt merited inclusion in the official minutes. The final measure is seconds to a board member’s motion. Like motions, these are easy to identify and record, but they are less authoritative than motions in that they simply express agreement or endorsement of what another board member has done. Two of our three measures—motions and comments—thus represent more authoritative and demanding forms of verbal participation, with seconds scoring substantially lower on both counts.

THE EFFECTS OF GENDER COMPOSITION

We expect that the gender composition of the meetings will matter in ways that are largely consistent with the patterns we saw in the majority-rule groups in the lab. This means that women’s participation should increase as the percentage of female board members present at the meeting increases. But our argument in the preceding chapters is not simply that women participate more, on aggregate, when there are more women in the group. Instead, we have argued that women’s voices are disadvantaged when they are the gender minority in a deliberating group that makes decisions by majority rule. That is, in such contexts, women’s voices will be underrepresented relative to their presence at the meeting. When women comprise only 20% of the group, they are significantly less likely to take up even 20% of the conversation in groups deciding by majority rule. Equality, on this understanding, is defined as voice proportional to representation in the group.

To test our expectations about the relationship between equality and gender composition, we develop a new measure: the ratio of women’s proportion of speaking turns to their proportion of attendees. A score of one on this “equality ratio” measure would mean that women spoke in exact proportion to their presence in the group: in a group comprised of 20% women, women would account for 20% of the speaking turns; if women were 40% of the group, they would take 40% of the speaking turns, and so on. Scores less than one signify underrepresentation for women, relative to their presence in the group; scores above one mean overrepresentation.

In our experimental data, all groups had five members, so we could always use a 20% standard to measure equal participation. Our school board data, however, include a greater variety of group sizes. While about half the boards in our sample are comprised of five members, another third of the sample includes six or seven members, and a few have as little as three or as many as twelve members. Thus the ratio of proportion of speaking turns to proportion of attendees who are women allows us to standardize across board size and is ultimately analogous to the approach we used in chapter 5. The models we report below also add a control for board size in case the overall number of people at the meeting matters.

Table 10.1 shows the relationship between the proportion of women at the meeting and the equality ratio in mixed-gender groups. The dependent variables include the two most demanding sorts of verbal interventions, motions and recorded comments. Bivariate relationships are shown in columns one and three, and full models with a variety of controls for various attributes of the board or meeting can be found in columns two and four. All models include cluster robust standard errors, with the clustering accounting for differences across states.¹⁵ As in our experimental data set, we have one significant outlier in our data set, and we control for this outlier in our models, just as we did in their earlier analyses of speaking time.¹⁶

With both dependent variables, motions and recorded comments, we find a significant relationship between the board’s gender composition and the extent to which women’s voices are equal to their numbers.¹⁷ When women are the minority in the group, they are much less likely to take speaking turns equal to their presence in the group. Put differently, women on boards with few women are disadvantaged, even relative to their minority status in the group.¹⁸ As the percentage of women in the group rises, however, so too does women’s participation, and as women come to outnumber men, women’s voice comes much closer to or even exceeds equality. These relationships hold in the presence of controls for district size, budget, location, or other attributes, such as the length of the meeting, whether the superintendent is female, or whether the meeting included public comment from citizens.¹⁹

The magnitude of the effect can be dramatic. Figure 10.1 shows the predicted values that emerge from the models with controls (Models 2 and 4 in table 10.1). Notably, the point estimates only rise to the level of full equality for motions when the board is composed of at least 60% women, and boards only approach equality for the recorded comments when at least 70% of the members are women.²⁰ When the dependent variable includes motions only, predicted values for the equality ratio move from 0.48 in boards with only 10% women to 1.29 in boards with 90% women, and in the model that includes recorded comments, the estimates run from 0.60 to 0.99. The confidence intervals that surround these point estimates (not shown) only overlap one when women form a large portion of the group—again above 60%. As in the experimental data, a supermajority of women is required before women reach equality of participation.

A somewhat different look at the results, using raw data and collapsing the gender compositions into minority or majority female, shows the same pattern. The mean equality ratio for recorded comments when women are the gender minority is 0.72—that is, women participate at a little less than three-fourths of their presence in the group. When women are the gender majority, the average equality ratio rises to 0.92. Average equality ratios for men, by contrast, never dip below 1, whether they are the gender minority or the gender majority.

In addition, we can combine the measures of comments and motions and calculate the percentage of groups in which women face a severe disadvantage, defined as an equality ratio of less than 0.5 (that is, when women contribute less than half of the motions and speaking turns that their proportion in the group would predict). We find a severe disadvantage in equality ratios for 24% of the groups in which women are the gender minority. When they are the gender majority, women experience a severe disadvantage in only 2% of boards. Men face a severe disadvantage relative to their numbers in only two groups in our entire sample, and in only eleven of the sixty-nine mixed-gender boards do men have an equality ratio for motions and comments of less than 0.9. In other words, women face a persistent deficit in verbal participation relative to their presence in the group when they are in the minority, rarely coming close to levels of participation that match their numbers, and it takes more than a majority before women come close to equality. Men routinely participate at levels equal to or much higher than their numbers in the group might anticipate. Only when women are a supermajority are the equality ratios of men and women roughly comparable.

Our analysis to this point reflects exactly the same patterns we saw in our majority-rule experimental groups, providing helpful validation of the results we found in the lab. But the school boards are different from our lab groups in at least one important respect. In our experimental groups, no member of the group was given any sort of formal power to lead the meeting or guide the decision making. Unlike our lab experiment, however, most school boards choose a formal leader or board chair who might have special responsibilities to conduct the meeting or otherwise speak up. What happens when we account for this formal position of leadership in our models?

Table 10.2 highlights these results by adding to the models in table 10.1 a control for whether or not the board chair is a woman. Model 1 of the table shows that this formal leadership position makes no difference to the pattern for motions. Women’s equality ratio is still strongly influenced by the proportion of women attending the board meeting, even in the presence of controls for women’s leadership responsibilities. In the vast majority of meetings in our sample, the board chair does not make any motions whatsoever, so this increase in the likelihood of making motions comes primarily from the other, nonchair women in the group.

Model 2 of Table 10.2 shows, though, that formal female leadership makes an important difference to the equality ratio for recorded comments. Formal authority exerts a large and positive effect, swamping the effect of the board’s gender composition. Closer examination shows that this effect holds for boards with few women and boards with many women.²¹ This is good news for minority women: the equality ratio for comments is 0.57 without a female chair—about half of what we would expect given their numbers on the board—but 1.20 with a female chair.²² In groups with a majority of women, the equality ratio is 0.71 in boards without a female chair and grows to above the line of equality (1.08) with a female chair (see chapter appendix table A10.1).²³ Regardless of the gender composition of the board, women are much more likely to reach our standard of equality, and even exceed it, if the board chair is a woman.²⁴ We interpret this finding as evidence that formal authority in the form of a leadership position in the meeting can be a critical factor for women’s equality of voice.

Does this result mean that formal authority is all that matters for women’s speaking turns and that gender composition is unimportant? Our answer to that question is no, for two distinct reasons. First, gender composition is strongly related to the likelihood of having a female chair. It is rare for a woman to occupy a position of leadership when the gender composition of the board tilts toward men. Only 19% of mixed-gender groups with a majority of men in our sample have a female chair, while 67% of groups with a majority of women have a female leader. This strong relationship holds up in models that control for a variety of other important board-level factors, as shown in chapter appendix table A10.2.²⁵ The predicted probability of having a female chair is less than 13% in boards comprised of 20% women, but rises to 65% for boards where 80% of the attending members are women.²⁶ In other words, gender composition still matters for verbal participation in part because it is strongly correlated with the likelihood of having a board chair who is a woman, and female board chairs—like all board chairs—tend to participate actively in the meeting. Put differently, the effect of gender composition on equality of speaking behavior is mediated by the gender of the official leader of the group. Formal tests of mediation yield strong evidence for just such a relationship.²⁷

Table 10.2: Effect of Female Board Chairs on the Ratio of Proportion of Turns Taken by Women to Proportion of Women, Mixed-Gender Groups Only

	(1) Motions	(2) Recorded Comments
Proportion Women	1.08***	0.02
	(0.40)	(0.21)
Female Chair	–0.08	0.51***
	(0.15)	(0.13)
Size of Board	–0.05	–0.03
	(0.10)	(0.04)
Board Experience (median years)	–0.03	0.01
	(0.02)	(0.02)
No. of College Graduates on Board	0.06	0.08**
	(0.05)	(0.04)
Public Comment at Meeting	0.08	0.09
	(0.14)	(0.19)
Urban Area	0.26*	–0.16*
	(0.18)	(0.10)
Superintendent is Female	–0.16*	–0.05
	(0.12)	(0.14)
No. of Students Enrolled	–0.12**	0.00
	(0.05)	(0.05)
Budget per Student	–0.24	0.12
	(0.23)	(0.15)
Length of Meeting	0.03	0.07
	(0.05)	(0.06)
South	–0.15	0.05
	(0.33)	(0.29)
West	0.17*	0.10
	(0.12)	(0.12)
Midwest	–0.07	0.08
	(0.13)	(0.15)
Constant	3.43	–0.97
	(2.09)	(1.54)
Observations	63	63
R-squared	0.46	0.40
Control for Outlier	Yes	Yes

Note: Group-level analysis. Cluster robust standard errors in parentheses (cluster by state). *** p < 0.01, ** p < 0.05, * p < 0.1, one-tailed test.

But gender composition matters for a second reason as well: it may also affect the remaining board members who do not hold leadership positions. To explore their behavior, we turn from the board-level data to an individual-level analysis. This move allows us to separate the board members who are chairs from those who are not and to control for other individual-level attributes, such as the member’s education or length of tenure on the board, that may affect a willingness to make comments during the meeting.²⁸ In this analysis, we subtract all comments the board chair made and look closely at how all other comments were distributed among the remaining board members. The dependent variable is the proportion of all nonchair comments the individual board member made. In table 10.3, we examine women only, and we find that as the proportion of women among the nonchair members of the board rises, so too does the proportion of speaking turns made by the average woman. The effect holds in the face of controls for the number of years served on the board as well as for a variety of other attributes of the board, including its size, location, and even whether the chair is a woman (see Models 2 and 3 of table 10.3).²⁹

The effect of gender composition on nonchair women is substantial, as can be seen in figure 10.2. Where 20% of the nonchair members are women, the average nonchair woman accounts for a little less than 15% of the nonchair recorded comments. For boards where 80% of the nonchair members are women, the average woman accounts for about 21% of the conversation.³⁰ This six-point change in average participation may at first glance seem small, but it is not. The ratio between the predicted value for women and the prediction for men when women comprise 20% of the group is 0.65, meaning that women are speaking at a rate of about two-thirds that of men. When women comprise 80% of the group, the ratio rises to 1.13—a little higher than the rate for men. Nonchair women’s voices thus move from severe underrepresentation to slight overrepresentation relative to men as the gender composition of the nonchair board members changes.

Restricting the analysis to boards that have a female chair only, we find that the patterns are exactly the same as with the full sample: gender composition is a powerful determinant of the participation of nonchair women (p < 0.05, one-tailed for the Proportion of Nonchair Women coefficient). More to the point, we can directly test the effect of a female board chair on women who do not hold this formal leadership position by regressing the proportion of comments from nonchair women on a dummy variable for whether the board has a female chair. This analysis shows that on boards where women are the minority, the presence of a female chair has no significant effect on the participation of the other women at the meeting.³¹ Thus having a female chair does not dramatically elevate the talk of nonchair women on boards with few women; among the women who do not occupy formal positions of leadership, we continue to see that participation is strongly influenced by the gender composition of the group, not by the presence of a woman in a position of power. These results put the findings from table 10.2 in context. They show that female chairs affect women’s average comments because female chairs speak a great deal, not because the other nonchair women change their patterns of participation in groups with female leaders.

Figure 10.2 also shows that the participation of nonchair males appears to fall as the gender composition changes—the more women, the lower the proportion of comments made by the average man—but these trends fall short of statistical significance in models with controls (online appendix table C10.3).³² In addition, even when women heavily outnumber them, men’s participation never sinks to the exceptionally low levels we find among women in groups with many men. In groups where men are heavily outnumbered (80% women and 20% men), for example, men still account for more than 19% of the conversation. By contrast, when men similarly outnumber women, they account for only 14.7% of the conversation. In other words, the same gender imbalance affects female board members more deeply than male board members. Finally, just as we saw above, figure 10.2 shows that nonchair women do not reach or exceed equality of voice—in the sense of equal participation by the average man and the average woman—until the board’s gender composition moves heavily in favor of women. Men’s advantage continues until they find themselves in groups where more than 60% of nonchair board members are women.

As in chapter 5, we also examined how equally the board members distributed their speaking turns by computing a Gini coefficient for the comments board members offered. In part because chairs, both male and female, tend to offer many more comments than most other members of the board, we do not find any relationship between the Gini coefficient and the percentage of women present at the meeting when board chairs are included in the analysis. However, when board chairs are removed from the analysis and we look at how speaking turns are distributed among the nonchair members only, we again find a strong relationship between the number of women present at the meeting and speaking turns. When more women are present in mixed-gender groups, speaking turns among nonchair board members are distributed more equally.³³

In sum, when female chairs are present, the board’s overall equality ratio—the relationship between the speaking turns of the average man and the average woman on the board—improves because the chairs themselves make a large number of comments. This mutes the relationship between the equality ratio and the board’s gender composition. But when we examine women who do not hold these leadership positions, we continue to find evidence of a relationship between gender composition and verbal participation. Nonchair women speak less and comments from board members are distributed less equally when women’s numbers are low; women speak more and speaking turns are more equally distributed when women’s numbers increase. While female chairs take an especially active role in the meeting, the mere presence of a female leader does not shield the nonchair women on the board from the influence of the board’s gender composition.

Of course, a more complete test of these findings awaits a larger sample of school boards (including a larger sample with female leaders) and a more detailed list of comments made during the meeting. The latter is probably only available with full transcripts of meetings where we could also explore for the length of speaking time for each board member. As we indicated already, the measure of comments from detailed meeting minutes is the most subject to the preferences of the board clerk in terms of what to record in the official report of the meeting. Perhaps, for example, we find such a powerful effect of board chairs in part because clerks are especially attentive to the ways in which the formal leaders manage the meetings and move through the agenda items.

Those caveats aside, our results for both motions and comments are quite consistent in both direction and magnitude with our findings from the lab. The fact that motions—a form of power leading to actual decision making in formal groups—depend so heavily on women’s numbers in the group is especially important evidence, in our view. Women are simply much less likely to attempt to influence the group with formal proposals when their numbers are few. They do offer such interventions when they gain additional power in the form of increased presence in the group.³⁴ With respect to comments, we find a similar story, though this is complicated somewhat by the effects of alternative sources of power such as a formal leadership position in the group.

In our experimental data, we paid special attention to the speaking behavior of gender tokens—lone men or women surrounded by the opposite gender. This condition, we found, brought significant disadvantage to women, as compared to men in similar circumstances. Our sampling strategy with the school boards, designed to ensure equal representation across different gender compositions, allows us to explore gender tokens in these groups as well. We have fifteen token men and twelve token women in our sample. Figure 10.3 shows the mean equality ratio for tokens of both genders. The figure includes the equality ratio for recorded comments and for motions. In both cases, we find a dramatic disparity between the speaking behavior of token men and token women. Token men still average at or above the equality ratio of one; they average 1.00 for comments and 1.07 for motions. Token women, on the other hand, have predicted equality ratios of about 0.6 (0.61 for comments and 0.59 for motions).³⁵ Women’s equality ratios hover well below our standard of equality, and they represent significant underrepresentation of women’s speech relative to their already-small position in the group. Once again, these patterns mirror our findings from the experimental data quite closely, with the possible exception of the fact that token men on school boards were, if anything, less disadvantaged than those in our experimental groups. Token men in our sample of school boards simply do not speak less than their proportion in the group, while token women often do—and often substantially less.

Our data set also includes one additional form of verbal participation: board members speaking up to second a motion made by another member. These are expressions of agreement with the proposal put on the table by the motion, and they do not contain any additional content or substance beyond the desire that the proposal should move to a vote or some other form of action. When it comes to this form of participation, the board’s gender composition has essentially no effect (see online appendix figure C10.1, which shows the relationship between the equality ratio for seconds and the proportion of women attending the meeting). Even when women make up a small portion of the board, many are willing to speak up to offer seconds. In one group comprised of 20% women, for example, the equality ratio nearly reaches three, a rate that well exceeds any result for comments or motions. It seems that women participate by seconding even when their numbers are low, though they do not offer more demanding and authoritative forms of comment. This pattern fits what we saw in chapter 2 of women’s participation in American society more generally; women volunteer, and act, when the act is viewed as a form of help to the group, but shy away from acts viewed as more assertive or opinionated.

While we want to tread carefully, given that our measure of comments is simply a summary from the board clerk or secretary, we find this pattern to be substantively meaningful in addition to providing reassurance about the wider applicability of our lab results for the majority decision rule. In conditions in which women are less empowered because of the group’s gender composition, women are less likely to engage in the more authoritative and substantive forms of speech, but such reticence does not apply to the less authoritative seconding of others’ motions.

Our analysis so far has focused almost entirely on the behavior of men and women in mixed-gender groups. We have explored the extent to which both genders intervene in board meetings at rates consistent with their proportions in the group. Such measures do not apply to gender enclaves, where by definition women will always take 100% of the conversation in all-female groups, and the same applies to men in all-male groups. What can we say about the nature of conversation in gender homogeneous groups? Unlike with the experimental data, where we have a rich set of variables to examine, the indicators from the board minutes are more limited. Nonetheless, the school board minutes do allow us to explore at least one aspect of gender homogeneous meetings: the duration of the meeting.

In our analysis of the experimental data, we found that women’s average talk time was longest in gender enclave groups. While we do not have an individual-level measure of talk time, we do have data on the overall length of nearly every board meeting.³⁶ And here we find that meetings in which the board members are exclusively women last more than an hour longer than meetings in which board members are exclusively men. All-female meetings take, on average, about three hours ten minutes, while all-male meetings in our sample last about an hour less (p-value of the difference = 0.06, one-tailed t-test).³⁷ That is, all-female meetings last about 50% longer than all-male meetings. As figure 10.4 shows, meetings of all-male boards are, on average, shortest; all-female board meetings tend to be longest; and mixed-gender groups are between these two extremes (2.5 hours).³⁸ These results are only suggestive, however, because our sample of enclave groups is relatively low—only seventeen total. In regression models that include controls for other board-level factors, the coefficient indicating the difference between male and female enclaves is large and in the expected directions, but falls short of statistical significance.

Notably, we also find no statistically meaningful difference between men and women in their respective enclaves in the number of turns (motions, comments, or seconds) taken. Total meeting time can be affected by many factors, including the issues under consideration, the level of public input, the size and location of the school district, or even the region of the country. Given these potential confounds, we are hesitant to draw any firm conclusions about the relationship between overall length of meeting and group gender composition. Nonetheless, the rough trends in our school board data are consistent with our lab findings of increased talk time in all-female groups. While much remains to be learned about the determinants of meeting length, our findings—however preliminary—point to the possibility that all-female boards have a different style of public meeting, one that is more inclusive or at least more willing to gather and discuss issues for longer periods of time. It is certainly not the case that female board members have less to say. In fact, they may have more to say than men do. But the conditions of deliberation dramatically affect their likelihood of saying it.

Our analysis in this chapter has focused primarily on whether women reach our standard of equality, which we define as a proportion of speaking turns equal to women’s proportion on the board. But when do men and women achieve a balance in terms of total participation? To answer that question, we examine the gap between the proportion of total turns taken by all men and those taken by all women on the board. Figure 10.5 presents this gap when the two most important forms of verbal participation in our data set, motions and comments, are combined. The y-axis in the figure is the gender gap at the meeting, defined as the proportion of motions or comments taken by women minus the proportion for men. Thus a meeting at which only women spoke and made motions would be at 1 on the scale and a meeting completely dominated by men would be at -1. The horizontal dashed line indicates a perfectly even balance in verbal participation, and the vertical dashed line indicates a meeting in which half the board members are men and half are women. Because of our sampling strategy, exactly half of mixed-gender boards in our sample have more men than women, and exactly half are evenly split or have more women than men.

The figure shows that the board’s gender composition profoundly influences the overall balance of voices found at the meeting. When women are the gender minority, their voices are heard less often than men’s about 88% of the time, and when women are the gender majority, female voices are less prominent than men’s only 27% of the time. In other words, descriptive representation helps women’s total participation. However, descriptive representation does not translate into total participation in a commensurate way; it takes boards that are heavily populated by women before women achieve an equal balance or better of total participation. Of the fifteen boards in our sample where women comprise 50 to 65% of board members, women’s total participation lags behind men’s in over half of them. These can be seen in the lower right quadrant of figure 10.5. Only when women’s descriptive representation exceeds 65% do they regularly attain or exceed equality of total participation. That is because even where women are a majority, men tend to speak more than their numbers. We can see this by comparing minority women and minority men. While women speak less than men in 88% of the boards where women are the minority, men speak less than women in only 73% of the boards when they are the minority.

And what about the size of the gender gap in total participation? As figure 10.5 indicates, the size of the gap changes with the gender composition of the board, but on the mixed-gender boards where women are the minority, the average gap is -0.48 on the -1 to 1 scale. Men outnumber women on these boards, but even so, men take an average of 50% more turns than their already large proportion of the board membership would predict.³⁹ When women are the majority, any overrepresentation of participation relative to their numbers on the board is much smaller than what we find among men.⁴⁰ Put simply, the measure of total participation is a second indicator that men talk more than one might expect even accounting for their overrepresentation on the board.

Thus when women are the minority on the board, men’s comments and motions heavily outnumber women’s, and they do so far beyond the imbalance in numbers. When women are in the majority, their voices predominate, though their dominance is not as dramatic as that of the male majority. These findings hold in a more rigorous analysis that includes the long list of controls we have used in previous models (see online appendix table C10.4). In sum, the measure of total participation confirms the problems we saw when focusing on average participation: when women are outnumbered, boards have a deep imbalance of voices—deeper even than gender imbalance in board membership would predict—and it takes a supermajority of women before women’s voices achieve balance or better.

WHAT ABOUT GROUPS THAT DO NOT DECIDE BY MAJORITY RULE?

To this point, we have focused exclusively on settings in which school boards make decisions by majority rule, which is the rule typically employed by elected bodies throughout the country. Yet our interaction hypothesis holds that the dynamic will be very different in groups that use unanimous rule. Accessing a sufficiently large data set of verbal participation in nonlab groups using that rule is difficult. For example, juries often decide by unanimous rule, but their deliberations are private, making systematic analysis of the dynamics impossible.

Our solution is to look to the world of deliberation and civic forums, where there has been a concerted effort to provide new opportunities for public dialogue and discussion. Often, deliberative groups ask participants to achieve consensus or pursue other forms of decision making than simple reliance on majority rule (Walsh 2007). Still other groups eschew formal group decision making altogether, preferring to focus on community building through public discussion. The civic dialogue groups studied by Katherine Cramer Walsh (2007) are a good example of groups brought together to explore issues of public concern in an atmosphere of inclusion. As the Everyday Democracy website puts it, the goal is “to help create communities that work better for everyone because all voices are included in public problem solving.”⁴¹ The groups Walsh studied were convened specifically to discuss the potentially thorny issue of race relations. But the important aspect of these groups from our perspective is that they set as their explicit goal a high level of unity through explicit instruction that participants should involve everyone and attend to “all voices.”

If these intergroup civic dialogues do not make collective decisions, how are they relevant to our interaction hypothesis? In our view, our argument about group dynamics is not necessarily about the need for a formal rule of unanimity, but rather about creating a norm of inclusion, solidarity, and support—a place where all members of the group are valued and encouraged to participate. These are precisely the goals of the civic dialogue groups Walsh studied. The instructions to participants in these dialogue groups include, for example, an encouragement to “speak your mind freely, but don’t monopolize the conversation,” to “really try to understand what others are saying and respond to their ideas, especially when their thinking is different from yours,” and to “listen carefully” (quoted in Walsh 2007, 42–43). The idea is to create a “safe space” where all can be heard, where inclusiveness is explicitly valued, and where the group focuses on the goals of discovering a common good (Walsh 2007, 40).

Intergroup civic dialogue thus attempts to generate many of the group dynamics that, we have argued, are fostered by unanimous rule. In this sense, it offers us a unique opportunity to explore patterns of participation in groups organized around norms of inclusion as they naturally occur. The focus of Walsh’s study were twenty-two meetings held by five groups in various parts of Illinois and Wisconsin, each of which met multiple times to discuss issues of race and ethnicity in their communities. Group sessions included as few as six participants and as many as sixteen, with the average group numbering about nine participants. All groups included both men and women. Happily for our purposes, Walsh carefully documented the number of speaking turns and the gender of each participant at each meeting (2007, 187–91). We analyze these data to explore how the gender composition at the session is related to patterns of verbal participation for men and women.

Just as we did with our school board data, we compute an equality ratio for each civic dialogue session. This is the proportion of speaking turns from women divided by the proportion of participants who are women. If we are correct in our assumption that these civic dialogue groups were likely to develop the same norms of inclusion we would expect from a formal rule of unanimity, then the relationship between the equality ratio and the group’s gender composition should be very different from what we found among school boards. Our interaction hypothesis predicts that unlike in the majority-rule school boards, the equality ratio should decrease as the percentage of women at the session increases.

And that is exactly what we find. Figure 10.6 presents the relationship between the equality ratio and each session’s gender composition for each of the five dialogue groups in Walsh’s data set, along with the fitted regression line for all sessions combined. The pattern is distinctly downward. Although the group gender composition is limited—most groups included a majority of women—the equality ratio is highest in sessions with a minority of women and declines as the proportion of women attending the session increases. When women are a minority, the equality ratio slightly exceeds the line of equality (1.04), and when women comprise 80% or more of the group, the ratio falls to 0.89. The magnitude of this change is not large, but dramatic inequalities are probably not to be expected when the group makes inclusion such an explicit goal. Moreover, this negative effect of gender composition is strongly significant (p < 0.001), even when we control for the total number of speaking turns in the session and whether or not the group facilitator was a male (see chapter appendix table A10.3). In fact, that analysis shows that male facilitators are correlated with a small but statistically significant decrease in the equality ratio. This is further evidence that facilitators trained in general dialogue techniques that do not specifically account for gender are not a panacea for the gender dynamics we have described throughout the book.

While a limited data set cannot ultimately be a definitive test of our hypothesis, these findings are fully consistent with our interaction hypothesis. Walsh’s data were collected for different purposes and without our interaction hypothesis, or gender dynamics, in mind. Unlike our experiment, the groups met repeatedly, not just once. The topic was race, not economic redistribution. Nonetheless, the patterns echo those we found in the lab. The results tell us that when groups develop a norm of inclusion and mutual support, the dynamics can be wholly different from what we find under majority rule. That pattern works to the benefit of women when they are fewer and to the detriment of women when they are many.

CONCLUSION

In her review of the literature on gender and office holding, Reingold concludes:

It is at the level of institutions … that some of our most challenging and creative efforts could be directed. We could capitalize on the most recent critiques of critical mass theory and consider the numerous ways in which the gender—and racial and ethnic—diversity of a political institution night (or might not) affect individual and collective behavior on behalf of women and women’s interests. Beyond questions of numbers … we should be open to considering the gender dynamics of … procedural rules.” (2008, 145)

We have attempted to engage in just such a “challenging and creative” effort. Our focus here, as it has been throughout the book, is on the intersection of numbers and procedural rules. While we cannot look at variation in formal rules, we were able to examine variation across numbers within a rule—the most common rule in politics, and the defining institutional feature of democracy: majority rule. We were also able to examine a different type of institutional feature—the official leadership of the board. And we explored the other side of our interaction hypothesis in naturally occurring midwestern groups that met on repeated occasions.

While our school boards data set is culled from official meeting minutes and thus has some limitations, we regard the findings we have reviewed here as a tough test of our expectations for majority-rule meetings. Not only are these settings ostensibly (though perhaps not always in actuality) focused on the needs of children—an issue highly important to women—but they are also comprised of women who are elected representatives. These are women with sufficient efficacy and confidence to have successfully run for elected office in the places where they live, women whose official position involves speaking up on behalf of their constituents. In that sense, these women are among the more privileged and powerful people in the country.

And yet even with this group of citizens successful and ambitious enough to become local officials, and even when the discussion deals with matters of importance to women, we find evidence of disparities of voice. While women in minority female groups tend to participate substantially less than their proportion on the board, minority men never face this participatory deficit. Token women are especially disadvantaged, and in our sample, minority women face severe disadvantage (verbal participation at rates of less than half of their proportion in the group) about one-quarter of the time. Women come closer to equality of participation—defined as voice in proportion to attendance—as the proportion of women on the board increases, but it typically takes a supermajority of women before they reach our definition of equality. These findings are most robust with respect to motions, the most assertive form of speech, and disappear as we might expect on seconds, the least opinionated and proactive form of participation. With respect to comments recorded by the board clerk, the trend is in the direction we expect.

We also find that women’s disadvantage can be somewhat offset when women occupy the formal positions of power, such as board chair, but these leaders are the only women who benefit from their formal role. Although female leaders themselves talk more, their presence does not appear to help the remaining women who do not serve as board chairs. Absent the responsibility of formal authority or the power that comes from increased numbers, women’s voices are, quite simply, more silent.

 

---

¹ By contrast with the average 40% female composition of school boards, corporate boards are far less diverse, composed on average of 15% women. In addition, 87% of the Fortune 500 companies have at least one female board director but less than 20% had three or more women (Pande and Ford 2011, 27, citing Catalyst Census: Fortune 500 Women Board Directors, available at http://www.catalyst.org/file/320/2009_fortune_500_census_women_board_directors.pdf). Female percentages in European boards are also very low, only improving recently as a consequence of laws mandating minimum quotas. See, for instance, http://www.nytimes.com/2011/01/27/world/europe/27iht-women27.html?_r=1&pagewanted=all.

² Superintendents report that boards have a substantial degree of autonomous authority and are not required to seek approval for their most important actions. For example, city or county councils have to approve school board budgets in only 9% of districts (NSBA table 49). Nearly two-thirds of boards have the authority to levy taxes, although such levies frequently require voter approval (NSBA tables 50–51). In 79% of cases, boards are solely authorized to call bond elections (table 52).

³ It is one of the organizations listed as gender segregated in Burns, Schlozman, and Verba (2001).

⁴ Both the Boles and Beck studies are referenced in Carroll (2001).

⁵ He also suggests, with anecdotal evidence, that men participate more when discussion is focused on road machinery (Bryan 2004, 249).

⁶ A similar effect obtains in seventeen Latin American countries—the presence of women among a country’s elected officeholders increases women’s political involvement (Desposato and Norrander 2009).

⁷ Based on the NES question: “During the campaign, did you talk to any people and try to show them why they should vote for or against one of the parties or candidates?”

⁸ “Women are more internally efficacious, more likely to discuss politics, discuss politics often, convince others, and comment on the political parties and less likely to say ‘don’t know.’ The fact that this finding holds true across multiple dependent variables, with controls for psychological, situational, and structural factors as well as the power gained from multiple election years makes these results robust and powerful” (Atkeson 2003, 1053)

⁹ We note as a caveat the criticism of this natural experiment, discussed in chapter 4.

¹⁰ In small districts 43% are conservative; in large districts, 27% are liberals and 22% conservatives. African Americans constitute 21.8% and Latinos 6% in the large districts. Small districts include only 6% African Americans and 1% Latinos (table 2 in the NSBA report). Large districts report somewhat higher incomes but even in small districts, 42% have incomes over $100,000 (NSBA).

¹¹ States without web directories included Arkansas, California, Georgia, Indiana, Kansas, Montana, Nebraska, New York, North Carolina, Tennessee, and Vermont.

¹² Between these sources we located electronic minutes for boards in forty-eight states.

¹³ The NSBA survey finds that 56.2% post board minutes and supporting documents online. Almost all large districts post minutes online; most other types of districts do too, but less uniformly. Eighty-four percent of large districts do so; 54 percent of the smallest districts do. We also found that school boards in some states were less likely to post minutes online; these were often in the South.

¹⁴ Chapter appendix figure A10.1 presents the distribution of our sample by the gender composition of the board members attending the meeting.

¹⁵ Because education is primarily a state-level responsibility, clustering by state in the group-level data sets is warranted.

¹⁶ This outlier is over four standard deviations away from all other groups in the sample with respect to the equality ratio for motions and nearly three standard deviations above groups in the 95th percentile of our sample.

¹⁷ The relationship is also statistically significant if we combine motions and recorded comments together (not shown).

¹⁸ We only have one board with seven or more women that has a majority of women, so we cannot test the interaction of gender composition and board size.

¹⁹ Six mixed-gender groups did not record the length of their meetings, but the patterns are identical if we exclude the meeting length variable to allow the inclusion of those groups.

²⁰ Those are the points at which the 95% confidence intervals for the point estimates in figure 10.1 cross the line of equality (a ratio of 1).

²¹ If we separate meetings where a high proportion of the attending board members were women from meetings with a low percentage of women, the effect of having a female chair on the equality ratio for recorded comments is positive and statistically significant (not shown).

²² This difference is significant at p < 0.01, one-tailed t-test (t = 3.39).

²³ This difference is significant at p < 0.01, one-tailed t-test (t = 2.95).

²⁴ The effect of gender composition is the wrong sign and is not significant for boards with a majority of women or for boards with a minority of women (analyzed separately).

²⁵ We note, for example, that boards with a female superintendent are also more likely to choose a female chair, even after controlling for the gender composition of the board.

²⁶ These predicted probabilities are computed from Model 2 of chapter appendix table A10.2.

²⁷ Both a Sobel test and the approach of Imai, Keele, and Tingley (2010) yield statistically significant evidence of an indirect relationship between gender composition and the equality ratio for recorded comments (p < 0.01).

²⁸ We were also able to collect demographic data on 420 of our 510 board members, and not surprisingly, the board members are a well-educated group. Only 8% of the board members for whom we could find information stopped at a high school education, and another 18% had some college or technical training, while nearly 40% had completed some form of postgraduate education. We control for education with a dichotomous variable indicating whether or not the member was part of the 74% of the sample that had graduated from college. The patterns we report below are very similar if we drop this control and use the full data set. The primary difference in the full data set is that the effect of gender composition on men is even smaller.

²⁹ In this individual-level analysis, standard errors are clustered by board.

³⁰ These are predicted values computed from Model 3 of table 10.3.

³¹ The coefficient indicating the presence of a female chair is small and does not come close to significance (p = 0.42, one-tailed). This result comes from analysis in which we restrict the sample to boards with less than 50% women. The model includes the same set of controls used in table 10.3, though results are the same without controls. In addition, the presence of a female chair has no effect on the participation of nonchair women when women comprise 50% or more of the board.

³² Readers may wonder why the decrease for men is not commensurate with the increase we see for women. If we were analyzing the total proportion of comments by men and the proportion of comments by women, these two would move in tandem and always sum to 1. But here, the predicted values from the regression are essentially showing the average proportion of comments made by men and the average proportion for women, similar to the individual-level analysis of verbal behavior found in chapter 5. These averages are sensitive to the overall number of board members and the number of men or women on the board, and the movement we see in the predicted values may not be perfectly parallel.

³³ In a board-level regression analysis in which the dependent variable is the Gini coefficient for speaking turns, the relationship between the percentage of nonchair women present and the Gini coefficient is strongly significant (b = -19.7, SE = 6.5, p < 0.01, one-tailed). Analysis is restricted to mixed-gender groups only. The Gini coefficient runs from 0 to 100, so this means the level of inequality in the distribution of speaking turns declines by about 20 percentage points as gender composition moves from its minimum to its maximum.

³⁴ We do not analyze the relationship between motions and passage rates because in our sample, very few of the issues being debated were highly controversial. We do not find enough variation in passage rates of motions to analyze this relationship fully.

³⁵ These equality ratios are predicted from OLS regressions of the equality ratio for comments and motions on an indicator for token women, with a control for our outlier group. In both regressions, token women are significantly different from token men (p < 0.01, one-tailed). Raw means tell an essentially identical story.

³⁶ Six mixed-gender boards and one gender homogenous board did not record total meeting time in their minutes.

³⁷ Both averages are inflated by one outlier meeting—a six-hour meeting in the case of women and a five-hour meeting for men, but if those outliers are removed, the difference is still substantial and statistically significant: all-female meetings last 2.7 hours and all-male meetings last 1.8 hours (p-value of the difference = 0.03, one-tailed t-test).

³⁸ If we exclude the outlier meetings that last more than five hours, all-male meetings last 1.8 hours, mixed-gender groups last 2.3 hours, and all-female groups last 2.7 hours.

³⁹ On average, the gap between men and women’s proportion of the board when women are the minority is 0.32. The gap of 0.48 in total speaking time is thus 50% greater than the gap in seats on the board.

⁴⁰ Only when we restrict the analysis to boards comprised of more than two-thirds women do women’s voices exceed their proportion of the board. In boards where women heavily outnumber men, the gap in total speaking time is, on average, 15% greater than the gap in seats on the board.

⁴¹ This was one of the organizations involved with the dialogue groups; it was formerly known as the Study Circles Resource Center. http://www.everyday-democracy.org/en/Page.WhatWeDo.aspx.