In the past decade, a substantial amount of scholarly work has
examined the effect of the degree of diversity within an organization upon
the experience of women working in that organization. Using as a case the
“Industrial Supply Corporation,” an anonymous corporation,
Rosabeth Kanter (1977) argued that token women
within an organization (that is, women who make up less than about 15% of
an organization's membership) are subject to performance pressures,
boundary heightening, and role entrapment. Token women in the Industrial
Supply Corporation reported that they were more visible than men; they
were “always viewed by an audience” (1977, 973). Moreover, co-workers and superiors viewed
these women on the basis of gender stereotypes and evaluated their actions
as evidence not only of their worth as sales personnel but also of their
worth as women. Token women in the corporation were evaluated in part on
the basis of their physical appearance, and they had to work harder to
have their achievements noticed. Gender differences were consistently
highlighted and exaggerated; furthermore, the acts of women tended to have
consequences for how other women were viewed and treated. Tokens were also
aware of a need to avoid placing members of the dominant group (that is,
men) in a negative light. The most common response of women to their
status was to minimize gender differences, keep a low profile, and let
others assume leadership positions or take credit for the token's
accomplishments. In general, in response to their status as token members
of a group, Kanter argued, women may downplay group differences, trying to
“blend unnoticeably into the predominant male culture” (1977, 973). In more equitable contexts, on the other
hand, women may become less isolated and may affect group processes and
the organization's culture.
Political scientists have applied Kanter's work to the political
setting, examining the behavior, treatment, and success of female state
legislators (e.g., Saint-Germain 1989; Thomas
1991, 1994). These
scholars generally argue that gender differences in legislative agenda
setting will be less pronounced in overwhelmingly male-dominated
legislatures, and that women will be most successful in relatively
balanced contexts. In this article, I examine distinctive agenda-setting
behavior by women; I also examine gender differences in legislative
success in terms of bill passage and appointment to leadership positions.
I argue that women do not have an incentive to avoid focusing on
women's issues, that there is a link between descriptive and
substantive representation, and that this link between descriptive and
substantive representation may be particularly evident in skewed
legislatures.
THE COMPOSITION OF THE LEGISLATURE AND THE
REPRESENTATION OF WOMEN'S INTERESTS
Several scholars have applied this work to the legislative setting,
arguing that gender differences in legislative policy priorities and
agendas will be minimized in homogeneous legislatures. Michelle
Saint-Germain (1989) found that significant
gender differences in the introduction of traditional women's
interest measures were evident once the percentage of women reached
approximately 15% (that is, in the last three sessions she studied); prior
to this, when women served as “tokens,” gender differences in
bill introduction were muted.1
Saint-Germain
(1989, 958–59) suggested that women in
less diverse legislatures would be more likely to propose feminist
legislation, but less likely to pass that legislation. The formal
hypotheses that she lists, however, make no reference to the gender
composition of the legislature (959). In her discussion of her findings,
she notes that “when women served in relatively few numbers they
addressed traditional women's issues to no greater extent than their
male colleagues; when women surpassed 15 percent of the Arizona state
legislature, they increased their attention to these issues, and the gap
between the importance of these issues for men and women legislators
became statistically significant” (963).
Using a 12-state
survey of state legislators, Sue Thomas (
1991,
1994) presented evidence that gender differences
in the prioritizing of legislation involving women, children, and families
were least marked in states with low percentages of women, and most
evident in states with high percentages of women. Sandra Grey (
2002) found, in New Zealand, that female politicians
were more actively involved in debates regarding feminist issues as they
approached a “critical mass” of 15%. Yet other scholars who do
not base their work on Kanter find little evidence that women avoid
engaging in substantive representation in settings where there are fewer
women. Irene Diamond's (
1977) study of New
England state legislators indicated that women expressed different
attitudes than did men on the issues of day care, abortion, and
protectionist labor laws. Evidence at the congressional level indicates
that gender differences in policy interests and behavior exist even when
women make up less than 15% of the legislature (
Welch
1985;
Wolbrecht 2000), and actually
become less evident as the legislature becomes less skewed (
Vega and Firestone 1995).
What accounts for these different empirical findings? One explanation
points to the different approaches that scholars have taken. Saint-Germain
used sponsorship as a dependent variable and based her conclusions on five
sessions of the Arizona state legislature; gender differences in
sponsorship did not emerge in the first two sessions (when there were very
few women in the legislature), but they did emerge in the last three
sessions (when the legislature was at least 15% female). It is difficult
to generalize from two sessions of the Arizona state legislature to other
skewed environments, particularly given that Saint-Germain found gender
differences only in the sponsorship of traditional women's interests,
and not in the sponsorship of feminist measures.2
Traditional women's interest measures generally focus on
policies that are broadly of interest to women, regardless of ideology,
such as pregnancy. Feminist measures focus on promoting equality or
battling discrimination.
Thomas (
1994)
conducted a much broader study, surveying legislators in 12 states. Yet
her conclusion regarding the relationship between the degree of gender
balance and the likelihood of distinctive behavior on the part of female
legislators is largely based on a lack of gender differences in two skewed
states: Mississippi and Pennsylvania. Moreover, as Thomas notes, other
states in her study do not conform to expectations. In South Dakota, women
composed almost 19% of the legislature, yet neither men nor women listed a
women's interest measure as a priority; in Georgia, with women
composing 13.3% of the legislature, women on average listed more
women's interest bills as a priority.
A second set of explanations for the different empirical findings on
critical mass stems from the problems that emerge from applying
Kanter's work to the legislative behavior of women. Kanter
acknowledged that tokens could respond in two possible ways to
their status: Tokens could promote themselves and their achievements, or
they could minimize differences with the dominant group. Kanter's
assertion that the latter was the more common response (1977, 974) may be less valid in the legislative
setting. It is reasonable to expect that many of the women who served in
“skewed” legislatures were politically active and voluntarily
entered a competitive political arena (Maxine Waters, Pat Schroeder,
Corrine Brown, Patsy Mink, and Lucille Roybal-Allard come to mind).3
Maxine Waters served from 1977 to 1991 in the
California legislature, and then from 1993 to the present in the U.S.
House. Pat Schroeder served from 1973 through 1997 in the U.S. House.
Patsy Mink served in the Hawaii legislature from 1956 to 1959, and then in
1962–64; she then served in the U.S. House from 1965 to 1977, and
then from 1990 to 2002. Corrine Brown served from 1983 to 1993 in the
Florida legislature, and then from 1993 to the present in the U.S. House.
Lucille Roybal-Allard served from 1987 to 1992 in the California
legistature, and then from 1993 to the present in the U.S. House.
It is very likely that many of these women saw their role as “acting
for” women. Indeed, research indicates that female legislators see
themselves as representing women (
Reingold 1992)
and are more likely to report being proud of accomplishments relating to
bills dealing with women (
Thomas 1994, 70).
The corporate setting differs markedly from the legislature in that
legislators respond not only to other legislators but also to voters.
Kanter argues that token women will be more visible, but additional
visibility can often benefit an elected official. Little evidence suggests
that voters are generally hostile to legislators—or to women in
particular—who focus on either feminist interests or on traditional
women's interests. Female legislators may instead be seen by voters
as better able to handle traditional women's interests (Dolan 2004; Huddy and Terkildsen
1993; Sanbonmatsu 2002), and women may
therefore have an electoral incentive to focus on those interests.
Finally, Kanter's work does not allow for the possibility that as
women increase their presence within legislatures, they begin to influence
men. If men become more likely to support women's interests,
then gender differences should narrow, not widen. Thomas (1991, 962) acknowledges this alternative possibility
when she hypothesizes that as the number of women increases in the
legislature, “the more likely it will be that women's attitudes
permeate the wider legislative atmosphere.”
Scholars applying Kanter's work to both the behavior and success
of female state legislators are putting the cart before the horse: One
needs to demonstrate that relevant performance pressures exist before one
can argue that women respond to performance pressures. Although
it is likely that discriminatory treatment does exist in the legislative
setting, it is unclear that it would discourage “token” women
from focusing on women's interests. Kanter argues that token women
face retaliation for doing too well, and therefore find themselves in a
double bind: They have to work harder than men in order to be seen as
competent but then face reprisals for their success. But are token female
legislators less successful than men, and do they face
retaliation for emphasizing gender differences in their legislative
agendas?
THE COMPOSITION OF A LEGISLATURE AND THE
SUCCESS OF FEMALE LEGISLATORS
Research investigating the effect of the composition of the
legislature on the success of female state legislators has produced mixed
results. Saint-Germain (1989) found little
association between the percentage of women in the legislature and the
success of women in passing the legislation they introduced. Women were
significantly more likely than men to pass women's interest
legislation in the 1969, 1973, and 1985 sessions of the Arizona state
legislature, when they comprised 12.2%, 13.3%, and 20% of the legislature,
respectively. No significant gender differences were found in the 1977 and
1981 sessions, when women comprised 16.7% and 18.9%, respectively. Thomas
(1991, 1994) found
that women were most successful in passing such legislation in relatively
gender-balanced legislatures. She concluded that “reduced conformity
pressures and increased support for distinctive behavior were advanced in
places where the proportion of women was highest” (Thomas 1994, 104). Drude Dahlerup (1988) applied Kanter's work to the Scandinavian
setting, arguing that resistance to women elected officials will generally
diminish as their numbers grow. At the same time, she acknowledged that
there may be examples of women who were more successful because
of their token status (Dahlerup 1988, 295).
Cindy Simon Rosenthal (1998) and Mark Considine
and Iva Ellen Deutchman (1996) argue that while
token legislators may indeed be subject to increased visibility and
performance pressures, the likelihood of increased influence depends on a
host of factors other than the number of women in the legislature. Given
the mixed results, it is likely that such factors as a sponsor's
majority party status and seniority have a pronounced effect on outcomes
like bill passage (Ellickson 1992; Moore and Thomas 1991), more pronounced than that of
the sex of the measure's sponsor. Moreover, prior research indicates
that legislative effectiveness is enhanced when legislators sponsor bills
on salient policy issues (Weissert 1991); gender
issues can certainly be regarded as salient in the time period under
consideration.
There is little evidence that a focus on women's interests in
particular makes female legislators less effective, even when they are
“tokens.” Thomas (1994, p. 97), for
instance, found that women generally had higher passage rates than men for
legislation dealing with women, children, and family; large gaps between
men and women were found both in states with relatively few women in the
legislature (such as Pennsylvania) and states with a relatively high
percentage of women in the legislature (such as Washington). Likewise,
Rosenthal (1998) speculates that although female
committee chairs exercise different styles of leadership than their male
counterparts, their styles may facilitate getting their agenda enacted
into policy.4
Kathlene (1994) found that women were not disadvantaged in terms
of passage rates. She noted, however, that measures sponsored by women
faced obstacles. Women-sponsored bills were significantly more likely to
be assigned to multiple committees, were debated longer in committee, and
received more hostile witness testimony. More research is needed to
determine whether such obstacles are more or less evident for
“women's interest” measures sponsored by
women.
Indeed, Kanter (1977, 981–85) argues
that token women will encounter “role entrapment.” Her own
words illustrate the theory most effectively: In skewed groups,
“tokens are often treated as representatives of their
category, as symbols rather than individuals” (1977, 966, emphasis mine). In the corporate setting,
the treatment of women as “representative of their category”
may lead them to downplay gender differences. There is, after all, no
obvious avenue for female sales personnel to profit from their experience
on women's issues. In the legislative setting, however, women may be
regarded as experts on political matters of relevance to women, and may be
encouraged to focus on women's issues. Where representation is part
of the job description, the treatment of women as “representative of
their category” may encourage them to behave distinctively from men,
to focus on legislation that is of particular interest to women. Michele
Swers (2002) notes that in the U.S. Congress,
Republican congresswomen are more likely to focus on women's issues
than are their male counterparts, while expressing concern about being
expected to be experts on traditional women's issues. If
female legislators are treated as representatives of “their
category,” there is little reason to expect that they would have an
incentive to downplay women's issues and minimize gender
differences.
Moreover, the effect of increasing numbers of women may not
necessarily be positive. As Karen Beckwith (2002) observes, critical mass can also explain
negative outcomes; as the number of women increase in an organization,
they may be less successful in achieving their goals. Janice Yoder (1991) argues that the increased presence of a group in
an organization may actually invite a backlash effect; similarly, Lyn
Kathlene (1994) finds that female legislators
meet with resistance in legislative committee debate, and that resistance
is particularly pronounced on committees that approach gender balance.
Even Kanter, in the 1993 edition of Men and Women of the
Corporation, acknowledged that women and minorities may be more
threatening to a dominant group when their numbers are higher. She
observes that “tokenism is hard on the sole representative but poses
no threat to the majority group,” and “research shows that
dissatisfaction and tension are greatest in groups in which there are
several women or minorities, but not enough to fully balance the
numbers” (Kanter 1993, 316).
Notwithstanding the mixed empirical record, as Donley Studlar and Ian
McAllister (2002) point out, the notion of a
“critical mass” necessary for changes in behavior and
influence has gained increasing currency in scholarly literature (e.g.,
Broughton and Palmieri 1999; Jaquette 1997) and even in writings designed for a
broader audience (e.g., Baer 1999). 5
Political scientists have often referred to the
“critical mass” thesis, and generally cite Thomas (1994) as the primary source. Dolan and Ford (1998, 77) write that “there is a variety of
evidence to support the ‘critical mass’ thesis—that
women act more distinctively once their numbers reach a certain
threshold.” Broughton and Palmieri (1999,
29) write that critical mass theory suggests that “[o]nce
women reach somewhere between 15% [Kanter
1977] and 30% [Thomas
1994] of elected politicians in effective democracies greater
representation and articulation of women's interests and concerns
will follow [Sapiro 1983; Phillips 1993].” Baer (1999, 9) writes that “research has shown that
when the proportion of women in the legislature falls below 15%, their
legislative performance is constrained.” Swers (2001, 217) notes that “scholars found that the
sex differences in the policy priorities of legislators intensified as the
proportion of women in the legislature approached a ‘critical
mass.’ ” See also Sam Howe Verhovek, “Record for Women
in Washington Legislature,” New York Times, 4 February
1999, sec. A.
DATA
The data used in this paper were gathered for a larger project taking
a historical perspective on female representation in state legislatures.
Information from the lower houses of state legislatures was gathered from
state legislative documents for the years 1969, 1979, 1989, and 1999 in
the states of California, Illinois, and Maryland. Because of the range of
female representation within these states and years, these legislatures
provide an excellent opportunity to test whether women respond to their
token status by altering their agendas, and to test whether an increased
proportion of women facilitates women's success. The percentage of
women within the legislature ranges from approximately 2% in Illinois in
1969 to approximately 32% in Maryland in 1999. Illinois and California
were Republican controlled in 1969; the remaining legislatures were
controlled by Democrats. In 1969, all three of the legislatures under
consideration were “skewed” according to Kanter's
typology; by 1999, all three had become “tilted” (that is,
women composed more than 15% of the legislature). Descriptive information
regarding the gender composition of the legislatures under consideration,
as well as the composition of legislatures in the United States overall,
is presented in Figure 1.6
As Figure 1 demonstrates, the state
legislatures selected are representative in terms of the trend of
increasing gender diversity. However, the state legislatures selected are
not entirely representative of the nation at large in terms of the
absolute percentage of women; they generally range from average to high on
that dimension. They are also not entirely representative in terms of
state ideology; on the basis of measures of state elite and mass ideology
developed by Berry et al. (1998), Illinois is
relatively “average” in ideology across the time period
studied, California is somewhat liberal, and Maryland consistently ranks
in the top 20% of liberal states, in terms of both mass public ideology
and government ideology. The reason for this nonrepresentativeness is that
only states that had at least enough women in the legislature in 1969 to
conduct empirical analyses were chosen for analysis. However, data exist
for Arkansas from 1969 through 1989; Arkansas is a relatively homogeneous
state (it is consistently ranked among the 10 states with the lowest
percentage of women in the legislature) and a relatively conservative
state. The low number of women makes obtaining reliable parameter
estimates in a multivariate regression difficult, particularly in 1979.
However, in 1969, when there were only three women (that is, 3% of the
chamber) serving in the statehouse, women sponsored significantly more
women's interest legislation. In 1979, when there were still only
three women in the legislature, women were not active in sponsoring
women's legislation. In 1989, the number of women had increased to 8%
of the chamber, and women introduced significantly more women's
interest legislation than men.
Trends in gender diversity in state legislatures
1969–1999.
Each bill introduced in the lower house of these legislatures was
coded according to its substantive content. Only substantive proposals for
new laws were counted as bill introductions; nonbinding resolutions and
memorials were excluded from the analyses. Introducer lists found in the
various state legislative documents were used to match legislators with
their proposals. When a distinction was made between primary and secondary
sponsors, only primary sponsors were included (though bills could have
multiple primary sponsors).
The focus of this study is women's interest legislation.
Political scientists have defined such legislation in a variety of ways;
Table 1 summarizes the recent approaches of five
scholars in the field, including Thomas (1991,
1994), Beth Reingold (2000), Swers (2002),
Chistina Wolbrecht (2000), and Kathleen Bratton
(2002).
Recent approaches of political scientists to defining women's
interests
There is a large degree of overlap among these approaches; all define
as “women's issues” a core set of issues that directly
and almost exclusively affect women. There are, of course, some
differences across the definitions. Some scholars include in their
definition of women's issues a broader set of issues with which women
have traditionally been associated, such as welfare, education, and
health; others focus on such issues as categories for analysis separate
from the analysis of women's issues. Some scholars merge what are
seen as traditional concerns of women with feminist concerns. Most but not
all scholars include measures or policies that are designed to protect
multiple groups (including women) from discrimination. The definition of
women's issues used in this study captures the core definition
reflected in these five approaches; women's interest legislation
includes bills that directly address and seek to improve women's
economic, political, and social status. Some specific examples are an
equal pay act, a bill that provides or regulates day care services, a bill
that requires insurance coverage of mammographies, and a bill establishing
affirmative action programs for women. Thus, women's interest
legislation is coded from a feminist perspective. It should be noted that
measures placed in broader categories such as “education” or
“health” were not included in the definition of women's
interests, unless these measures directly focused on women, because such
broader definitions, though certainly part of women's traditional
interests, do not quite as obviously run counter to the “predominant
male culture.”7
An example of a
women's interest measure relevant to education would be a measure
that banned gender discrimination in college sports. An example of a
measure relevant to education that was not considered a
women's interest measure would be a measure that increased funding
for textbooks in public schools. Three supplementary analyses were
performed on three broader categories of measures: education measures,
health measures, and measures focused on poverty. No evidence was found
that sex differences in sponsorship became more pronounced as the
percentage of women in the legislature approached or moved beyond a
“critical mass.”
If token women do downplay gender
differences, one of the most obvious ways to do this is by avoiding the
sponsorship of measures that directly and almost exclusively affect women.
This somewhat narrow definition of women's interests matches well
with Kanter's argument that tokens would downplay group differences
and, as noted, blend unnoticeably into the predominant culture (
Kanter 1977, 973). Moreover, given that this
definition is relatively stringent, it sets a relatively high standard for
finding gender differences. More information regarding the coding of
women's interest legislation can be found in the Appendix.
METHODS
The experience of legislators is examined at two points in the
legislative process: agenda setting and success. Agenda setting is
measured as the sponsorship of measures involving women's interests.
Success is measured in two ways: as bill passage within the chamber and as
appointment to leadership positions. The agenda-setting stage of the
legislative process provides an opportunity to evaluate whether women in
homogeneous settings downplay gender differences in response to their
token status; the passage stage of the legislative process provides an
opportunity to evaluate whether the legislative success of
“token” women is limited.
Legislative Agenda Setting
The unit of the sponsorship analyses is the individual legislator; the
dependent variable in these analyses is the number of women's
interest bills sponsored. Twelve negative binomial regression analyses
(one for each state and year) are performed to test the effect of gender
on the sponsorship of women's interest legislation.8
Because event-count data tend to be strongly skewed to the
right (that is, legislators who introduce a women's interest bill
tend to introduce only one such measure, fewer legislators introduce two,
even fewer legislators introduce three, and so on), ordinary least squares
is an inappropriate method to analyze such data. Negative binomial
regression or Poisson regression analysis is appropriate when the
dependent variable is an event count (King
1988). In this case, the negative binomial distribution is the
appropriate choice, because a legislator who sponsors one women's
interest measure is more likely than other legislators to sponsor other
such measures. That is, negative binomial regression is appropriate in
cases such as this where there is overdispersion (that is, the variance of
the distribution is greater than the mean).
Democrats and
Republicans tend to focus on different policy issues in general (
Jacobsen 1992); specifically, Republicans may be less
likely to sponsor feminist legislation (
Bratton and
Haynie 1999;
Swers 2002). Therefore,
party is included in the analysis as a control variable. Seniority may
also affect the sponsorship of women's interests; John Hibbing (
1991) demonstrates that increased tenure in office is
accompanied by increased legislative expertise. Seniority (measured as the
number of consecutive years in the legislature) is therefore included as a
control variable. Previous research indicates that women are more likely
to be elected from urban districts, and that gender differences in
legislative behavior narrow once the urbanness of the district is taken
into account (
Welch 1985); thus, I control for
urbanness, measured as the size of the largest city in the district (
Herring 1990). As an additional measure of district
demographics, I control for the percentage Black in the district, as
individuals elected from districts with a high proportion of minority
voters may be relatively liberal and thus relatively likely to focus on
women's interest measures. Finally, because service on a relevant
committee likely affects sponsorship choices (
Bratton
and Haynie 1999,
Swers 2002), I control
for membership on relevant committees. “Relevant committee”
was defined relatively broadly to include committees focusing on
education, health, children, aging, welfare, or human
services/resources.
Legislative Success: Passage of Legislation
The unit of analysis here is the bill, and the dependent variable is
coded 1 if the bill passed in the lower house, and 0 otherwise. Female
legislators have the most direct influence over passage in the chamber (as
opposed to passage into law). Twelve logistic regression analyses (one for
each state and year) are performed to examine the effect of the gender of
the sponsor on whether a bill passes in the lower house. The two
independent variables of primary interest are the percentage of sponsors
that are female, and whether the measure is a women's interest
measure, coded as described here. Controls are included for several other
likely influences on bill passage. Drawing on previous research that
indicates that legislation sponsored by African American legislators is
under some conditions less likely to pass (Bratton and
Haynie 1999), I control for the percentage of sponsors who are
black. It is likely that legislators who have served in the institution
for a relatively long period of time are more adept at steering
legislation toward passage; thus, I control for the average seniority of
the sponsors. Sponsors in the majority party and legislative leaders are
more likely to gain support for their sponsored legislation; thus, I
control for the percentage of sponsors that are affiliated with the
majority party, as well as the percentage of sponsors that are leaders.
Leadership positions include the Speaker of the House, the floor leaders,
assistant and deputy floor leaders, whips, assistant and deputy whips, and
committee chairs. Drawing on previous research that indicates that fiscal
measures are often more or less likely to pass (Bratton
and Haynie 1999), I include controls for whether the bill is a
taxation or appropriations measure.9
Taxation
measures generally included those involving changes in tax levels, and the
establishment of or changes in tax exemptions and credits, as well as
measures regulating tax collection and protecting taxpayer rights.
Appropriations measures were generally spending measures.
Previous research indicates that women are particularly successful when
they focus on women's issues (
Saint-Germain
1989; Thomas
1991,
1994); therefore, I include an additional variable
that takes the value of the percentage of female sponsors if the measure
is a women's interest measure, and is coded 0 otherwise; the
parameter estimate for this variable measures any additional likelihood of
passage of a women's interest bill if sponsored by women. Finally,
because legislators from predominantly black or relatively urban districts
may be more likely to sponsor legislation at odds with the preferences of
other legislators, controls are included for the average percentage Black
(logged) and the average size of the largest city (logged) in the
sponsors' districts.
10 I expect that
any effect of the percentage Black in the district and the size of the
largest city in the district will be nonlinear. For instance, an increase
of 5,000 in the population of the largest city will have a bigger effect
if the largest city is less populated than if it is more populated. The
values of these two variables are therefore logged to discount large
values.
Legislative Success: Appointment to Leadership
Positions
Of course, women could be very successful in terms of passing
legislation—but still be very much marginalized. In a second set of
analyses, I examine whether token women are less likely to hold leadership
positions, particularly if they focus on women's interests. Holding a
leadership position is an excellent measure of one's status within a
legislature; it is the legislative analogy to promotion within the
corporate setting. The unit of analysis here is the legislator; the
dependent variable is coded 1 if the legislator holds a leadership
position, and zero otherwise. Leadership positions included the Speaker of
the House, the floor leaders, assistant and deputy floor leaders, whips,
assistant and deputy whips, and committee chairs. I present two logistic
regression analyses, one for relatively homogenous legislatures, and one
for legislatures with a relatively balanced proportion of men and women.
The independent variable of primary interest is the sex of the legislator.
To examine whether individuals who focus on women's issues are
particularly marginalized, I control as well for the percentage of all
measures introduced by that legislator that are women's interest
measures. To examine whether women who focus on women's
issues are particularly marginalized, I include a variable that takes the
value of the percentage of women's measures introduced if the
legislator is female, and is equal to 0 for men. The parameter estimate
for this variable measures the additional effect for women of a
relatively heavy focus on women's issues on attainment of leadership
positions. As in the passage analyses, in addition to the sex of the
legislator, I control for a number of other influences on the likelihood
that a legislator will hold a leadership position, including the race,
seniority, and majority party status of the legislator; the district
demographics of the legislative district (including, as before, the
percentage Black in the district [logged] and the size of the
largest city in the district [logged]). I also include dummy
variables to control for year and state effects.
FINDINGS
Gender Differences in Bill Sponsorship
The results from the analyses of bill sponsorship demonstrate that
women consistently focus on women's interests more than do men.
In California, women are generally more likely than men to sponsor
women's interest legislation, although the differences are
statistically significant only in 1969 and 1989.11
Supplementary analyses across the 12 cases indicate that
there is no pattern of significant gender differences in the total number
of bills sponsored. In California in 1969, women sponsored significantly
fewer bills than men; in Maryland in 1979 through 1999, women sponsored
significantly more bills than men.
In Illinois and Maryland,
gender differences are more marked. In both states, significant gender
differences exist in all four time periods. There appears to be little
evidence that token women are constrained in their sponsorship
behavior.
12 Several scholars have recognized
the importance of partisanship in the representation of women and of
women's issues (Bratton 2002; Bratton and Barnello 2002; Sanbonmatsu 2002; Swers
2002). Supplementary analyses within parties indicate that gender
differences are fairly consistent across Republicans and Democrats. I also
performed supplementary analyses on four subsamples of legislators: those
who served as tokens in the majority party, those who served in the
majority party in relatively diverse chambers, those who served as tokens
in the minority party, and those who served in the minority party in
relatively diverse chambers. In all four contexts, there were substantial
gender gaps in bill sponsorship.
The results across the states
with token women are strikingly consistent, particularly given that the
analysis focuses on relatively small numbers of women within these
chambers.
Tables 2, 3, and
4 present no evidence that women become more
likely to behave distinctively as the legislature becomes more gender
balanced. Indeed, gender differences in both Illinois and Maryland
gradually diminish as the proportion of women in the legislature
increases. One potential explanation for this narrowing is that men are
sponsoring more of these measures; more research should be done to
investigate to what degree and under what conditions men focus on
“women's issues.”13
Some
preliminary research does exist. Across six states (12 chambers) in 2001,
Bratton and Barnello (2002) find that as the
percentage of Republican women in a legislature increases, the likelihood
that a Republican sponsoring a women's interest measure drops
substantially, but no similar effect exists among Democrats. Factors such
as partisanship and age had much more pronounced effects on the behavior
of men than women. In fact, young, African American, married male
Democratic legislators with more seniority are actually as likely as
comparable women to focus on women's interests.
A second
potential explanation for the narrowing of the gender gap is that multiple
women are sharing the role of representing women's interests. The
establishment of women's caucuses in 1969 in Maryland and in 1979 in
Illinois may have helped to coordinate the sponsorship of measures; future
research should investigate this possibility as well.
14 In their discussions of the representation of women and
women's interests, both Beckwith (2002) and
Childs (2001) focus on “newness” in
addition to numbers. Supplementary analyses performed on the skewed
legislatures in this database indicate no pattern of significant
sponsorship differences between recently elected female legislators and
female incumbents.
CALIFORNIA. Influences on sponsorship of women's interests:
Parameter estimates, negative binomial regressions (robust standard errors
in parentheses)
ILLINOIS. Influences on sponsorship of women's interests:
Parameter estimates, negative binomial regressions (robust standard errors
in parentheses)
MARYLAND. Influences on sponsorship of women's interests:
Parameter estimates, negative binomial regressions (robust standard errors
in parentheses)
The findings in these analyses support the expectation that women are
perceived as “representatives of their group as a whole”
(Kanter 1977, 976) and thus may be encouraged to
create distinctive policy agendas. One reason that women are not
constrained in their sponsorship behavior may be that even in relatively
homogeneous settings, the measures that women sponsor and measures that
focus on women's interests are as successful as other measures. The
next set of analyses examines that question.
Gender Differences in Bill Passage
The question here is whether women and/or those who introduce
women's interest measures are disadvantaged in skewed settings. That
is, is there discrimination in bill passage in skewed settings that might
provide an incentive for token women to downplay gender differences? The
results of the logistic analyses on bill passage, presented in Tables
5, 6, and 7, offer strong evidence that “token”
women do not face relatively high obstacles at the passage stage of the
legislative process.
CALIFORNIA. Influences on passage of legislation: Parameter estimates,
logistic regression analysis (robust standard errors in
parentheses)
ILLINOIS. Influences on passage of legislation: Parameter estimates,
logistic regression analysis (robust standard errors in
parentheses)
MARYLAND. Influences on passage of legislation: Parameter estimates,
logistic regression analysis (robust standard errors in
parentheses)
Two observations are in order. First, the results suggest that women
are often even more likely than men to pass the legislation that they
sponsor. Second, when women are advantaged, the gender gap is most
substantial in earlier years, when they comprise a smaller percentage of
the institution. In California, women were significantly more likely to
pass the legislation they sponsored in 1969, 1979, and 1989; this is
particularly true in 1969 and 1979, when women comprised less than 15% of
the legislature. In Illinois, women were most advantaged in 1969, when
they comprised only 2% of the legislature. Only in Maryland in 1989, where
women filled about a quarter of the legislative seats, were women
significantly less likely to pass the legislation they sponsored.15
If the analysis is restricted to legislation
with only female or only male sponsors, the pattern of significance and
direction is generally the same with few exceptions; the substantive
conclusions remain the same. Supplementary analyses were also performed
regressing the vote margins of bills that reached a floor vote onto the
independent variables; the findings are substantively the same as those
presented in this article.
Thus no evidence is found that women are disadvantaged in bill passage
by their token status—indeed, token women may even have an
advantage. Some marginal support is found for increased resistance to the
measures that women sponsor. As women grow in number in California and
Illinois, they lose an advantage in passing legislation; this may indicate
more resistance to women, although there is no evidence of a disadvantage
relative to men. In Maryland, there were no significant gender differences
in passage between women and men until 1989, when women comprised about
25% of the legislature; at this point, they were significantly less likely
to pass legislation than were their male counterparts. In 1999, however,
when women comprised about 32% of the legislature, gender differences in
passage disappeared.
Interestingly, the results suggest that the sex of the legislator
matters more than the content of the legislation. In each case,
women's interest legislation is no more or less likely to pass than
are other measures, and no trend over time is discerned. In Maryland in
1999, women's interest bills were more likely to pass if sponsored by
women; in Illinois in 1969, women's interest bills were less likely
to pass if sponsored by women. Generally, however, women are no more
likely to achieve passage of women's interest bills than they are to
achieve passage of other measures.
Gender Differences in Leadership Positions
Are women in token settings less likely to hold leadership positions?
The percentages presented in Table 8 suggest
that this might be the case. In token settings, a moderately sized gender
gap exists; only 14% of women hold leadership positions, compared to about
20% of men. This gap shrinks to only 2% in more heterogeneous
settings.16
Recall that leadership positions
were defined to include the Speaker of the House, the floor leaders,
assistant and deputy floor leaders, whips, assistant and deputy whips, and
committee chairs. When leadership was defined somewhat more narrowly, not
including committee chairs, gender differences virtually disappeared; men
and women were about equally likely to serve as leaders.
Percentage of women and men holding leadership positions, homogeneous
versus heterogeneous chambers
However, the results of the two logistic regression analyses
(presented in Table 9) suggest that once
controls are included, gender does not have a significant effect on the
likelihood of holding a leadership position in either homogeneous or
heterogeneous legislatures.
Influences on holding leadership positions, token versus non-token
settings: Parameter estimates, logistic regression analyses (robust
standard errors in parentheses)
Part of the gender difference apparent in the simple percentages
(Table 8) is actually due to gender differences
in seniority; in legislatures where women hold less than 15% of the seats,
the average seniority for women is about 4.3 years, whereas the average
seniority for men is about 6.2 years. These differences shrink markedly in
the nontoken legislatures (to about 6.8 years for women, compared to about
7.2 years for men). There are also some differences in majority party
affiliation that contribute to the gender gap demonstrated in Table 8; for instance, in legislatures where women
hold less than 15% of the seats, about 60% of men (compared with 50% of
women) are affiliated with the minority party. As the percentage of women
in the legislature reaches parity, there are virtually no gender
differences in majority party status.
Moreover, the results indicate that male or female legislators who
sponsor a relatively high percentage of women's interest measures do
not seem disadvantaged. There appears to be no incentive for women in
token settings to downplay gender differences by avoiding the sponsorship
of women's interest measures.
Gender Diversity and Legislative Outputs
What can we conclude from these findings? If an increase in gender
balance is not accompanied by an increase in distinctive behavior, and if
women are less likely relative to men to pass legislation, is there any
reason to advocate increasing the percentage of women in legislatures? The
answer, unequivocally, is yes. First, a legislature that closely reflects
the diversity of the society that it represents provides a host of
benefits: It may be seen as a sign of a legitimate government, it provides
role models, and it more effectively utilizes the talents of all the
citizenry (Darcy, Welch, and Clark 1987; Kelly, Saint-Germain, and Horn 1991; Mandel 1988; Mandel and Dodson
1992; Norris 1996).
Second, it is extremely important to distinguish changes in policy
outcomes from changes in the success of female legislators at the
individual level. Even if the distinctive behavior and legislative success
of female representatives do not depend on increasing the percentage of
women within the legislature, an increase in female legislators who focus
on policies relevant to women may change policy outputs. Although research
regarding policy outcomes is less plentiful than that regarding individual
behavior, some scholars have found evidence that increased gender
diversity does have the potential to influence the day-to-day lives of
female citizens. For instance, Sarah Poggione (2001) found that more gender-diverse state legislative
committees were more likely to pass relatively generous welfare benefits.
Jocelyn Crowley (2004) finds that tokens can
make a policy difference, and the potential to make a difference increases
along with their numbers. Likewise, Bratton and Leonard Ray (2002) found that increased female representation in
Norway was associated with increased child-care provision. The data
presented in Table 10 support this
possibility.
Overall number of women's interest measures passed, by state and
year, lower chamber
As the number of women in each legislature increased, the number of
women's interest bills passed generally increased. The only exception
to this trend is California in 1999; the number of women's interest
bills that passed declined from 1989. It is interesting to note that the
number of bills passed overall declined; the recently instituted term
limits in California may have had some influence here. This overall trend
in passage of women's interest measures appears to be due to the
activity of both men and women. In both California and Illinois, as the
number of women in the legislature increased, the number of women's
interest bills sponsored by women increased. In California, as the number
of women in the legislature increased, the number of women's interest
bills sponsored by men increased. In Maryland, where cosponsorship is much
more common, the increase was seen in bills sponsored by both men and
women.
CONCLUSIONS
This research demonstrates that women in very homogeneous settings do
not react to their token status by minimizing gender differences in agenda
setting. Indeed, gender differences in agenda-setting behavior in some
states narrow as the percentage of women in the legislature increases. I
find as well that women are generally at least as likely as men to pass
legislation even when they make up a very small minority of the
institution. In two states (California and Illinois), token women are
actually more successful relative to men in achieving passage of the bills
they sponsor. Perhaps most intriguingly, these findings replicate what
Saint-Germain (1989) found but did not call
attention to: Women serving in legislatures with little gender balance are
actually more successful relative to men than their counterparts
in more equitable settings. Dahlerup (1988)
observed that such cases did exist; these results suggest that the
phenomenon may be more common than previously thought. These findings also
lend credence to the notion that female legislators might respond to their
status by overachievement.
There are several reasons why numbers may have little effect on the
behavior of women. First, many women are interested in women's issues
(hence, the name of the category), and elected officials are less likely
to object to the sort of visibility that Kanter described. Second, there
are a host of other factors that influence legislative agenda setting,
including the mass public, institutional climate and culture, individual
partisanship and ideology, and majority party status. Numbers are likely a
very small part of the equation, and that may be particularly true as
women's numbers increase and as multiple individuals are interested
in sharing the role of representing women.
Of course, women do face discriminatory treatment in the legislative
process. There is really little doubt that female legislators suffer from
much of what Kanter described: loyalty tests, isolation, evaluation not
only as a legislator but as a woman, exaggeration of differences, and
mistaken identifications as wives or secretaries rather than as
legislators. As Kathlene (1994) demonstrated,
the composition of the legislature may influence the interaction between
male and female legislators (as opposed to the behavior of individual
women); this would be a promising avenue of future research. However, the
likely response of women to discriminatory treatment is different
from that of the sales personnel in Kanter's study. Women serving as
tokens in a political setting are quite likely to emphasize gender
differences and successfully bring a different perspective to
lawmaking.
What can we expect from the future? Women in more diverse legislatures
may share the role of bringing these issues to the agenda, and the number
or proportion of women's interest measures sponsored by individual
women may drop slightly. But given that there is little incentive to avoid
these measures, and given that many female legislators have expressed
interest in these issues, many women will continue, overall, representing
women in substantive legislation. Men may also increase their focus on
women's interests. That is, descriptive representation may make a
difference not only through the distinctive behavior of female legislators
but also through their impact on the behavior of their male
colleagues.
In sum, the aggregate results indicate that as the number of women in
a legislature grows, the potential for changes in the day-to-day lives of
female citizens increases. That is, even if the individual
behavior and success of women does not depend on the gender composition of
the legislature (indeed, even if individual women sponsor fewer
women's interest measures), an increase in the number of women
legislators over time may well be accompanied by changes in the
institution itself and in the policies it produces.
APPENDIX
According to Gelb and Palley (1982, 2),
feminism is “a movement seeking to operationalize self-determination
for women in political, economic, and social roles.” In this study,
women's interest legislation is defined as legislation that would
decrease discrimination or counter the effects of discrimination, or would
improve the social, economic, or political status of women. These
generally involved three overlapping categories: measures that addressed
the health concerns of women; measures that addressed the social,
educational, and economic status of women; and measures that addressed the
political and personal freedom of women. Some examples of measures that
addressed the health concerns of women are those focusing on pregnancy,
childbirth, breast cancer, osteoporosis, and so on. Some examples of
measures that addressed the social, educational, and economic status of
women are those encouraging women's participation in male-dominated
sports and academic programs, those establishing education programs in
women's prisons, measures protecting property rights and pensions for
divorcees or widows, measures establishing displaced homemaker programs,
measures providing quality child care, and affirmative action measures.
Some examples of measures addressing the political and personal freedom of
women are measures supporting reproductive rights, measures addressing
violence against women, and measures addressing sexual harassment.
