Gender Differences in Cooperative Environments: Evidence from
the US Congress∗
Stefano Gagliarducci
M. Daniele Paserman
Tor Vergata University and IZA
Boston University, NBER, CEPR and IZA
June 2013
Abstract
This paper uses data on bill sponsosrhip, cosponsorship, and roll call voting behavior in the U.S. House of Representatives to estimate gender differences in a number
of different measures of cooperative behavior. Women of both parties are able to recruit a larger number of cosponsors on the bills that they sponsor. In the aggregate,
male and female representatives are equally likely to sponsor legislation that attracts
bipartisan support, and to offer support to bills sponsored by members of the opposite
party. However, this result masks important differences by party affiliation: female
Democrats are substantially less likely exhibit bipartisan behavior, while the opposite
is true for female Republicans. These differences vanish when one controls for selection
by gender into different congressional districts, by focusing on mixed-gender electoral
races decided by a narrow margin.
Keywords: US Congress, Cooperativeness, Bipartisanship, Gender
JEL codes: ???
∗
VERY preliminary and incomplete. We thank participants at the Boston University Empirical Microeconomics Workshop
for many helpful suggestions. Giacomo Brusco, Ying Lei and Nathaniel Young provided excellent research assistance.
1
Introduction
As the share of women in legislatures and executive positions continues to grow in the
U.S. and around the world, it is natural to ask how this change is likely to shape future
policy. It is not uncommon to read political commentators arguing that more women in key
posts in the executive and legislative branches of government would result in more desirable
policy outcomes. One of the most commonly voiced arguments is that womens style of
politics is more conducive to cooperation, and may contribute to breaking partisan gridlock
in Congress. A recent editorial by Swanee Hunt, a former US Ambassador to Austria, argues
that “women may be our best chance at breaking through disastrous partisan gridlock As
Congress operates in an increasingly acrimonious space, politicians on each side are locked
in a stare-down. Yet the other current and future women in the Senate nodded agreement
when Susan Collins told Diane Sawyer on ABC World News, ‘If [women] were in charge
of the Senate and of the administration, we would have a budget deal by now... With all
deference to our male colleagues, women’s styles tend to be more collaborative.’ ”
Yet, in the absence of credible quantitative measures of cooperativeness, many of these
arguments seem to be guided more by casual empiricism than by solid empirical evidence.
While some recent studies have found that female politicians are able to deliver better
policy outcomes, in a variety of different contexts and at different levels of government
(more transfers from the central government, more investment in public goods that address
womens concerns, less corruption), others have reached more mixed conclusions. A common
shortcoming of most of the existing studies is that they are only able to look at final outcomes,
with little understanding of the process and mechanism through which female representatives
may be able to achieve superior policy outcomes. Therefore, it is difficult to distinguish
whether women are able to achieve better outcomes because they are inherently more skilled,
hard-working or ambitious (as would be the case if a woman has to overcome discrimination
to be elected, so that only the most qualified women are observed in the legislative body),
or because they act more cooperatively.
This contribution of this paper is to create credible measures of cooperative and bipartisan
behavior, exploiting the large amount of information available for the U.S. Congress. The
paper thus sheds light on one important mechanism through which women may be able to
deliver better policy outcomes: the ability to build broad coalitions and generate bipartisan
support for policy initiatives.
We will focus in particular on the extent to which members of Congress are able to
recruit support for their policy initiatives and extend support to others’ policy initiatives.
We will focus both on the total amount of support as an indication of the ability of congress
1
members to build large coalitions; and on support across party lines, as an indication of the
congress member’s bipartisanhip. Our primary analysis will focus on bill sponsorship and
cosponsorship, but we will also supplement this with an analysis of roll-call voting behavior
[to be added].
Clearly, any differences in cooperative behavior among Congress members may reflect
the preferences of one’s constituents as well as any individual differences in cooperativeness.
Moreover, it may be that cooperativeness is driven by other personal Congress member
characteristics that are correlated with gender, rather than gender per se. To address these
potential confounding factors, we will control in our analysis for a large number of individual
and district characteristics, as well as for district fixed effects [to be added]. In addition, we
will also implement a regression discontinuity analysis (RD), where we will focus attention
on mixed gender races decided by a narrow margin. In these races, the gender of the
elected representative can be thought of as exogenously determined by idiosyncratic events
on Election Day, and therefore the RD design allows us to identify the causal effect of gender
on the outcomes of interest.
The raw data show that women of both parties on average recruit more cosponsors to their
bills, and engage in more cosponsorship activity. As for the measures of bipartisanship, male
and female representatives are equally likely to sponsor legislation that attracts bipartisan
support, and to offer support to bills sponsored by members of the opposite party. This result
masks important differences by party affiliation: female Democrats are substantially less
likely exhibit bipartisan behavior, while the opposite is true for female Republicans. However,
this gender gap vanishes when one controls for selection by gender into different congressional
districts, by focusing on mixed-gender electoral races decided by a narrow margin. We
conclude that much of the apparent gender differences in cooperativeness reflects mostly
differences in the types of congressional districts in which male and female representatives
are elected.
The rest of the paper is structured as follows: Section 2 discusses how the paper fits into
different strands of the literature. Section 3 gives some institutional background on the U.S.
Congress, and on bill sponsorship and cosponsorship. Section 4 presents the data. Section5
describes how we construct our different measures of competitiveness. Section 6 presents
the empirical methodology. Section 7 presents the main empirical results and Section 8
concludes.
2
2
Related Literature
This project is related to at least three strands of literature: 1) on gender differences in the
performance and behavior of political representatives; 2) on bipartisanship and cooperation
in Congress, as measured by roll-call votes, sponsorship and cosponsorship activity; and 3)
on gender differences in attitudes towards competition and cooperation.
Women in Politics. There is an emerging literature in economics on the choices and
performance of female politicians. Much of the literature has been concerned with identifying
gender differences in the way local fiscal policy is conducted. In an early and influential
paper, Chattopadhyay and Duflo (2004) use political reservations for women in Indian Village
Councils to show that female leaders invest more in public goods more closely linked to
women’s concerns. Using similar data from India, Clots-Figueras (2011) finds that that
female State Legislators who are elected in scheduled caste seats favor women-friendly laws,
such as amendments to the Hindu Succession Act that give women the same inheritance
rights as men, and affect the educational levels of individuals who grow up in the districts
where these female politicians are elected (Clots-Figueras, 2012).
However, in contexts in which there are no reserved quotas for female politicians, the
results have been more mixed. Gagliarducci and Paserman (2012) using a large sample of
Italian municipal governments, find little evidence of differences in policy outcomes such as
the size of the budget deficit or the allocation of expenditures between municipalities headed
by male and female mayors, but a large difference in the probability that a female mayor
survives until the end of her term. This result persists and becomes stronger when controlling for non-random sorting of women into municipalities using regression discontinuity
in gender-mixed electoral races decided by a narrow margin. Ferreira and Gyourko (2011)
conduct a similar analysis in a sample of U.S. municipalities, and find no effect of gender
of the mayor on policy outcomes related to the size of local government, the composition
of municipal spending and employment, or crime rates. In contrast, Brollo and Troiano
(2012), find that Brazilian municipalities headed by female mayors are awarded more discretionary federal transfers, are less likely to have administrative irregularities, and achieve
better health outcomes; however, despite these results, female mayors are less likely to be
re-elected, probably because they engage less in political patronage.
Other papers have looked instead at gender differences in the behavior of legislators, at
both the national and local level. Funk and Gathmann (2008), using data from direct democracy in Switzerland, find large gender gaps in the areas of health, environmental protection,
defense spending and welfare policy, but no difference in the overall size of government. Rehavi (2007), using data from state legislatures in the U.S., and exploiting the quasi random
3
variation in the number of female legislators created by the outcomes of extremely close
elections, finds that women contributed to a modest but significant increase in state health
spending and slowed the growth of corrections institution spending.
Most closely related to this project is the recent paper by Anzia and Berry (2011), who
find that women in the U.S. Congress are able to deliver more discretionary federal spending
to their districts, and are also more politically active in that they sponsor and cosponsor
significantly more legislation. Anzia and Berry interpret these results as evidence of bias on
the part of voters, which induces only the most talented and hard-working female candidates
to succeed in the electoral process. The paper, however, is silent as to the mechanism through
which congresswomen are able to deliver better policy outcomes to their districts, and does
not address at all the question of whether women engage in more bipartisan behavior.
The evidence above is in line with the rich literature that has looked at gender differences
in legislators voting behavior. Studies have found that women tend to be more liberal than
men (Welch, 1985; Norton, 1999), and more likely to support and promote womens issues
(Swers, 1998, Vega and Firestone, 1995). Another strand of the literature has examined
whether female legislators are as effective as their male colleagues in sponsoring bills that
are eventually turned into law (Bratton and Haynie, 1999; Jeydel and Taylor, 2003).
Other aspects of gender differences in politics include studies of gender stereotypes of
candidates (Huddy and Terklidsen, 1993; Sanbonmatsu, 2002), and their interplay with
campaign strategies (Herrnson, Lay and Stokes-Brown, 2003); gender differences in electoral
behavior (Sapiro and Conover, 1997), and the evolution of the political gender gap (BoxSteffensmeier, De Boef and Lin, 2004; Edlund and Pande, 2002).
Existing Research on Bipartisanship and Legislative Activity in Congress.
Much attention has been devoted in the political science literature to the issue of partisanship and polarization in the U.S. Congress. Using ratings issued by interest groups such
as the Americans for Democratic Action and the United States Chamber of Commerce,
Poole and Rosenthal (1984) found that beginning in the mid-1970s American politics became much more divisive, with more Democrats staking out consistently liberal positions,
and more Republicans supporting conservative ones. Building on these early findings, Poole
and Rosenthal (1985) developed a spatial measure of legislative ideology aggregating information from all roll-call votes in Congress. The NOMINATE score and its refinements
(D-NOMINATE and DW-NOMINATE) are now updated continuously on Poole and Rosenthals website (http://www.voteview.com) and confirm that party polarization has continued
to increase since the 1980s, and has even accelerated since the mid-1990s. Other scholars
have reached similar conclusions with different roll-call summary scores: (Rohde, 1991; Cole-
4
man, 1997; Stonecash et al. 2003; Brewer et al., 2002, Collie and Mason, 2000; Fleisher and
Bond, 2000; see also Theriault, 2008, and the references therein).
In the literature there is some debate about how to correctly interpret roll-call voting
behavior, and the extent to which it is governed by party discipline. The traditional notion,
originally articulated by Mayhew (1974) and later assessed in an empirical study by McCarty,
Poole and Rosenthal (2001), is that parties have generally had negligible influence on roll-call
voting behavior, the voting on the floor being mostly dependent on the ideological position
of each representative and their loyalty to the constituency. This view, however, contrasts
with other recent empirical studies, like Jenkins (1999), Cox and Poole (2002), Ansolabehere,
Snyder and Stewart III (2001) and Snyder and Groseclose (2000), which show instead that
parties largely affect the roll-call voting behavior, the incidence being highest on close votes
and key party issues.
Harbridge (2009, 2011, 2012), however, argues that focusing only on roll-call voting may
miss part of the picture, because the set of bills that actually reaches a roll-call vote is determined endogenously by the congressional leadership, which has an incentive to bring to
the floor only partisan bills that are not likely to create divisions within the party. In fact,
Harbridge develops a new measure of bipartisanship based on bill cosponsorship coalitions,
and shows that bipartisanship in bill cosponsorship continues at relatively high levels, even
as polarization has increased. The argument for using cosponsorship activity as a measure of
bipartisanship is that it is less likely to be subject to agenda control and is one of the most
independent activities of Congress members. While there is some debate about the exact
nature of cosponsorship (whether it is used as a low-cost way of taking position and communicate with constituents, or as a way to send signals within the legislature; see Krehbiel,
1995, and Kessler and Krehbiel, 1996), we agree with Harbridges position that cosponsorship
of bills that receive support from both side of the aisle is a credible measure of bipartisanship
and cooperation across party lines.
The project also plans to construct measures of cooperation based on the primary bill
sponsor. Primary sponsorship conveys very important information about individuals legislative behavior (Wawro, 2001). Schiller (1995) remarks that a “senators choice of bills
is a strong indicator of which issues he or she wants to be associated with.” In fact, the
interpretation of bill sponsorship by Wawro and others is remarkably close to the concept
of cooperative behavior in its common accession. “The tasks and responsibilities of the
primary sponsor typically involve entrepreneurial activities such as coalition building and
shepherding legislation through the House.” (Wawro, 2011, p. 27, our emphasis). Schneier
and Gross (1993) remark that the primary sponsor is responsible for discussing the bill with
5
proponents and opponents, and guiding the bill through the legislative process.
The number of cosponsors on a bill is indicative of a Congress member’s ability to build
coalitions (Wawro, 2001, p. 30). Campbell (1982) notes that members exert significant effort
to recruit members as cosponsors and use the number of diversity of cosponsors to make
claims about the support for the legislation. In fact, a number of studies have found that
the number of cosponsors on a bill indeed positively related to the bill’s passage probability
(Browne, 1985; Wilson and Young, 1997).
Therefore, the existing literature on bipartisanship and legislative activity confirms the
basic premise of this project, that sponsorship and cosponsorship activity, as well as roll-call
voting behavior, can be used as credible measures of cooperative behavior.
Gender and Competition/Cooperation. Recent research has highlighted substantial
gender differences in preferences for competitive environments and in performance in such
environments. Men are more likely to select into more competitive compensation schemes
(Niederle and Vesterlund, 2007; Dohmen and Falk, 2011; Booth and Nolen, 2012); they are
more likely to seek difficult challenges (Niederle and Yestrumskas, 2008); they tend to raise
their performance in competitive settings (Gneezy, Niederle and Rustichini, 2003; Gneezy
and Rustichini, 2004) and are better able to maintain high levels of performance in high
pressure situations (Lavy, 2008a; rs, Palomino and Peyrache, 2008). These differences may
play a role in explaining the substantial gender pay gap at the top of the income distribution
and the large underrepresentation of women in “power” professions (Bertrand and Hallock,
2001). On the other hand, Lavy (2008b) finds no gender differences in performance in
a tournament in which contestants have more time to prepare and plan their strategies,
and Manning and Saidi (2010) argue that gender differences in the incidence of pay-forperformance schemes can account for only a small fraction of the gender gap in the U.K.
Paserman (2010) finds that professional tennis players of both genders substantially reduce
their performance in high-pressure situations.
Some researchers have argued that women may have other characteristics that compensate for this lack of “competitive spirit.” For example, women may exhibit a cooperative
personality that gives them a comparative advantage in contexts (such as certain types of
negotiations) where such skills translate into superior outcomes for all parties (Babcock
and Laschever, 2003). Cooperativeness is usually defined, in contrast to competitiveness,
as the aptitude of working or acting together willingly for a common purpose or benefit.
The experimental evidence on this subject (based on public good games, ultimatum games
and dictator games) is mixed, with some studies showing that women are more concerned
for the outcome of the other party (Andreoni and Vesterlund, 2001; Eckel and Grossman,
6
1996 and 1998; Nowell and Tinkler, 1994), but others finding no effects (Croson and Buchan,
1999), or effects of the opposite sign (Brown-Kruse and Hummels, 1993). Croson and Gneezy
(2009) survey the experimental literature and suggest that the variance across studies can
be explained by a differential sensitivity of men and women to the social conditions in the
experiment.
It has been hard to test for gender differences in cooperative behavior outside of experimental contexts, because of the difficulty of measuring cooperativeness. One of the
contributions of this paper is to provide evidence on gender differences in cooperativeness
from an important real-world setting with high stakes.
3
Institutional Background
The US House of Representatives
• The House of Representatives is the lower chamber of the US legislative branch.
• There are 435 congressional districts, each state is represented in proportion to population.
• Elections for all districts occur every two years.
• Representatives draft legislation, which must pass the House, the Senate and be approved by the President before becoming law.
• There are different types of legislation:
– Bills
– Simple Resolutions
– Concurrent Resolutions
– Joint Resolutions
• We will focus on bills, since they have force of law and are the primary vehicle for
members’ legislative efforts.
Bill Sponsorship
• Each bill has one primary sponsor. The sponsor is not necessarily the sole or the most
important author of the bill, but he/she is is identified with the bill content.
7
• The sponsor’s activities:
– gathering and communicating information
– coalition building
– public relations
– shepherding legislation through House.
Cosponsorship
• Bill can be signed by any number of cosponsors.
• There is some debate in the literature about whether cosponsorship is just positiontaking or cheap talk.
• Wawro: “The number of cosponsors a member can [attract] ... is indicative of the member’s
ability to convince others that the bill is worthy of their support.”
• Attracting numerous cosponsors can keep a bill moving through the legislative process.
4
Data
We collect and link data from five different data sources: a) bill sponsorship and cosponsorship data; b) biographical information on members of congress; c) roll call voting data; d)
data on electoral results; and e) data on district demographic and economic characteristics.
These five data sources will be described in turn.
Bill sponsorship data. Using the Library of Congress data information system, THOMAS
(http://thomas.loc.gov), we retrieved information on all bills submitted from the 93rd Congress
(1973) onwards. The information includes the names of the sponsors and cosponsors of the
bill; the detailed legislative history of the bill, the committees of referral, reporting and
origin. We also cross-validated this data with other sources, such as Adler and Wilkersons Congressional Bills Project (http://www.congressionalbills.org) and Fowler, Waugh and
Sohns Cosponsorship Network Data (http://jhfowler.ucsd.edu/cosponsorship.htm). These
data sources also provided information on the major topic of the bill, and whether the bill
eventually passed in the House, passed in the Senate, and whether it was eventually enacted
into law.
Biographical Data. Biographical information on all members of Congress, up the
104th Congress (1995-1997) is available from the Inter-university Consortium for Political
8
and Social Research (ICPSR) at the University of Michigan. Information on subsequent
Congresses was retrieved from the Biographical Directory of the United States Congress
available online at the Library of Congress (http://bioguide.congress.gov).
Roll Call Voting Data. Roll call voting data has been widely used in academic research.
We will use as our basis the data compiled and made publicly available by Keith Poole and
Howard Rosenthal on the web site http://www.voteview.com. The advantage of using this
data base is that it easily allows linkages to ICPSR and other data sets.
Election Data. Election statistics for all house districts and senate seats are available
from the Office of the Clerk of the House of Representatives (http://clerk.house.gov).
Economic and Demographic Data. Demographic and economic information on congressional districts are available from the U.S. Census Summary Data Files, retrievable
through ICPSR.
5
Measuring Cooperativeness
A key part of our analysis involves the construction of measures of cooperative behavior
based on legislative activity. We should first clarify that, following Wawro (2001), we focus
exclusively on House Bills, therefore excluding other forms of legislative activity such as
simple and concurrent resolutions, which are often used more for pure position-taking and
can have little legislative impact.
We calculate several different measures of cooperative behavior, based on the bill sponsorship data and on roll call voting behavior. Our main measures are:
1. The total number of cosponsors on bills sponsored, which can be used as a measure
of coalition-building ability, both within and across party lines.
2. The total number of bills cosponsored, which tells us about the willingness of Congress
members to lend their support to initiatives sponsored by others, both within and across
party lines.
3. The number or fraction of cosponsors from the opposite party. The measure captures the degree to which congress members are able to generate policy initiatives that are
supported, at least initially, by members of the opposite party. It is worth remarking that,
regardless of ones position on why members choose to cosponsor bills, introducing legislation
that garners support from the other side of the aisle is undoubtedly indicative of cooperation
across party lines.
4. The number or fraction of bills cosponsored in which the primary sponsor belongs
to the opposite party. Harbridge (2009) uses cosponsorship by members of the opposite
9
party to that of the primary sponsor as her main measure of bipartisanship. Our measure
is constructed analogously at the individual congress member level.
We can extend these measures of cooperation to include also roll-call voting behavior,
and derive two additional measures:
5. The number/fraction of Yea votes coming from the opposite party in roll call votes
on bills sponsored by a member of Congress. This measure is analogous to measure 3, in
that it captures the extent to which sponsored legislation is able to garner support from the
opposite party. It is a somewhat stronger measure of bipartisanship, because roll-call voting
implies a stronger expression of support than simple cosponsorship.
6. The number or fraction of Yea votes in roll call votes on bills sponsored by members
of the opposite party. This is similar to measure 4, but stronger, because roll-call voting is
a stronger expression of support than simple cosponsorship.
Notice that some of the measures can be thought of as measures of proactive cooperative
behavior, in the sense that they capture the extent to which members of Congress initiate and
promote legislation that will receive broad bipartisan support. Others instead can be thought
of as measures of reactive cooperative behavior, in the sense that they capture the extent
to which members of Congress are willing to extend support to legislation initiated by the
opposite party. Some of these measures have already been used in the literature, although not
always in the context of cooperative behavior. Nevertheless, calculating all these measures
will give us a comprehensive view of legislative activity that can be interpreted under the
lens of cooperation.
In addition to these measures of cooperativeness, it will be also useful to study whether
differences in cooperativeness also translate into more legislative success (a higher fraction of
bills passing or becoming public laws, and the length “life” of a bill in the legislative process)
[To be added].
For each measure, where appropriate, we sometimes also provide a “gender” specification, i.e., we identify the percentage of female members of the opposite party who cosponsored/initiated a bill, or voted for a bill. This helps us understanding whether cooperative behaviors also depends on the gender of the members of the opposite party. Finally, in
some specifications we try to identify if there are differences in our measures depending on
whether the bill addresses female-friendly topics, and on the importance of the bill.
6
Methodology
Ordinary Least Squares and Fixed Effects. Consider first an OLS regression:
10
Yidt = α + βF emaleidt + γXit + δZdt + ηt + εidt ,
where Yidt is the measure of cooperativeness of Congress member i, from district d, in
Congress t; F emaleit is a gender dummy; Xit is a vector of control variables which include
other demographic characteristics of the member of Congress (age, seniority in Congress),
the member’s ideological position based on the DW-NOMINATE score, and on his or her
role in congressional committees; Zdt is a vector of demographic and economic characteristics
of district d at time t; and ηt is a Congress fixed effect.
The main coefficients of interest is β, which represents the gender difference in the measure of cooperativeness, holding constant other individual and district characteristics. The
large set of control variables in the OLS regression enables us to make a first assessment of
the effect of the member of Congress’ gender, after controlling for a host of other factors
that may be correlated with gender and also affect cooperativeness. The error term εidt can
also include district fixed effects, to control for fixed factors at the district level that may be
correlated with the gender of the representatives and may affect the outcome of interest.
Regression Discontinuity. The OLS specification captures whether there is an association between gender and cooperative behavior in Congress, once we control for the most
obvious confounding factors that are potentially correlated with gender. However, it may
be the case that women are more likely to be elected in districts that, at a specific moment
in time, are more distant from the political center. Presumably, representatives of districts
that are closer to the political center are more likely to engage in cooperative behavior. This
is true for all types of measures of cooperativeness. In terms of overall coalition building,
representatives from contested districts have stronger incentives to demonstrate to voters
that they have been active and that they have been able to garner broad suppport for their
policy initiatives; and in terms of bipartisanship, representatives from contested districts
must reach out to the opposite party to appeal to the key centrist voters.
Therefore, the simple OLS estimates may yield a spurious relationship between gender
and cooperativeness, which can be either positive or negative, depending on whether female
representatives are more likely to be elected in districts that are closer or farther away from
the political center.
To overcome this concern, we will employ a Regression Discontinuity (RD) design, in the
attempt to identify the effect of gender on cooperativeness off of close electoral races. The
identification strategy relies on the fact that in mixed-gender races decided by a narrow margin, the gender of the representative is likely determined by idiosyncratic shocks on election
11
day. Therefore, the RD design effectively randomizes the gender of the representative, and
allows us to estimate the causal effect of gender on the outcomes of interest.
Somewhat more formally, let M Vidt be the margin of victory (i.e., the difference in vote
share) for a female candidate i in district d in Congress t. In the RD literature, this is
often referred to as the forcing variable. As is well known, the goal of RD is to estimate the
boundary points of the conditional expectation of Yidt , conditional on M Vidt , on both sides
of the threshold value M Vidt = 0.
To estimate these conditional expectations, we apply two methods: local linear regression
proposed by Imbens and Lemieux (2008), and the split polynomial approximation used by
Lee, Moretti, and Butler (2004) and Lee (2008). The first method restricts the estimation to
a compact support, and fits linear regression functions to the observations within a distance
h on either side of the threshold. This method is particularly attractive because it is not
sensitive to outcome values for observations far away from the threshold. In other words, it
restricts the sample to an interval M Vdt ∈ [−h, +h] to estimate the model:
Yidt = α + βF emaledt + θ1 M Vdt + θ2 F emaledt × M Vdt + εidt ,
using OLS. The bandwidth h can be selected applying the cross-validation method proposed
by Imbens and Lemieux (2008). The second method uses the whole sample, choosing a
flexible specification to fit the relationship between Yidt and M Vdt on either side of the
threshold:
Yidt = α + βF emaleidt + (κ1 M Vdt + ... + κp M Vdtp ) +
(λ1 F emaledt × M Vdt + ... + λp F emaledt × M Vdtp ) + εidt ,
which is then estimated with OLS. In both specifications the coefficient β identifies the effect
of a female Congress member on the outcome variable at the discontinuity point.
Interpretation of RD estimates.
[To be completed.]
Which types of representatives are likely to be elected in close races? It depends on
whether voters value cooperativeness positively or negatively, and on whether candidates
can move closer or farther away from the median voter.
7
Results
[Incomplete]
12
7.1
Summary Statistics and Trends
Table 1 presents some basic summary statistics about our sample. The top presents summary
statistics for bills, and the bottom panel for Congress members. Women represent 12.3% of
Congress members during our sample period (first row of Panel B), but they sponsor 14.0%
of Congressional Bills (first row of Panel A). The distribution of the number of cosponsors
is heavily skewed to the right (Panel A). Female-sponsored bills tend to attract a larger
number of cosponsors (true for both the mean and the median). On the other hand, femalesponsored bills have fewer cosponsors of the opposite party. The bottom panel shows that
women are also more active as cosponsors, but they are slightly less likely to cosponsor bills
originating from the other side of the aisle.
Figure 1 shows trends in the fraction of female Congress members, overall and separately
by party. The fraction of women in the House has risen from about 7% to 17% over our
sample period, but there are important differences between the parties. In fact, while in
1988 Republicans had a slight advantage in the fraction of female representatives, by the
end of the sample period a substantial gap had developed in favor of the Democrats.
Figure 2 presents trends in our four measures of cooperative behavior. Figures 2a and 2b
look at the two measures of coalition-building and coalition formation: the mean number of
cosponsors (Figure 2a) and the mean number of bills cosponsored (Figure 2b). Both series
exhibit a drop between 1988 and 1994, a rise between 1994 and 1998, and then are fairly
constant. Figures 2c and 2d look at measures of bipartisanship: the fraction of cosponsors
from the opposite party (Figure 2c), and the fraction of cosponsored bills sponsored by a
member of the opposite party. Both series exhibit an overall downward trend, consistent
with notions of polarization in the U.S. Congress. The decline is sharper in the later years,
and also more pronounced for women.
7.2
Validity Tests
In order to implement the RD design, we must verify two conditions: a) that the density
of the forcing variable (the margin of victory of the female candidate) is continuous at
the discontinuity point; and b) that the and district and representative characteristics are
balanced at the discontinuity point.
We report the results of a McCrary density test (McCrary, 2009) in Figure 3 and Table
2. We restrict the sample to congressional districts that had a mixed gender electoral race.
Figure 3 shows the density of the margin of victory of the female candidate for our preferred
bandwidth parameter, h = 5 . Men are on average more likely to win mixed-gender elections
13
(probably because of the incumbency advantage), but the density appears to be continuous
at the threshold. This is confirmed in Table 2, where we report the McCrary test statistic for
a number of different values of the bandwidth, ranging from h = 1 to h = 9. In all cases the
test statistic fails to reject the null that the log difference in height around the discontinuity
point is equal to zero.
Table 3 and Figures 4 and 5 test whether covariates are balanced around the discontinuity
point. The figures fit a third order polynomial on the margin of victory on either side of the
discontinuity point. In the table, we report the height of the jump at the threshold and its
standard error for two different specifications: linear regression (i.e., a first order polynomial)
and a third order polynomial on the margin of victory.
Figure 4 and the top panel of Table 3 show balancing tests for a number of district
characteristics. Both the visual evidence and the statistical tests show that on the whole
district characteristics are balanced at the discontinuity point. Figure 5 and the bottom
panel of Table 3 show balancing tests for Congress member characteristics. Most of the
member characteristics appear to be balanced, but we do find some evidence of imbalance
in two variables: female who win close elections against a male opponent have on average
served in Congress for fewer terms (relative to men who win close elections against a female
opponent), and are much more likely to be Democrat. Nevertheless, we will be able to control
for these variables in the empirical implementation.
7.3
Gender and Cooperativeness
Visual Evidence. We first present in Figures 6-9 visual evidence of our main analysis.
All the Figures include three panels, for all Congress members and separately by party. In
Figures 6 and 8, the unit of observation is a Congressional bill, while in Figures 7 and 9 it is
individual Congress members. In all cases, the sample includes only representatives elected
in mixed-gender races.
Some interesting patterns emerge from the Figures, with many of the outcomes varying
with the margin of victory. For example, our main measure of bipartisanship, the fraction of
cosponsors of the opposite party, exhibits an inverted-U shape, meaning that there is more
bipartisan behavior in contested districts, as one would have expected. However, in none of
the graphs is there any strong evidence of a discontinuity at the threshold. To probe this
result further, we now move to regression analysis, where we can additionally control for all
other district and individual characteristics.
Regression Analysis. In Table 4 we report the coefficient estimates on the gender
dummy in a series of regressions where the dependent variable is the number of cosponsors
14
on a Congressional bill. The first row presents results for the sample as a whole, while rows
2 and 3 look at Democrats and Republicans separately. In the first column, we include in
the sample all available bills, and run a simple OLS regression of the number of cosponsors, a gender dummy, and additional individual and demographic controls. In the second
column, we restrict the sample to include only bills sponsored by representatives elected in
mixed-gender electoral races, and perform the same OLS regression. In columns 3 and 4 we
additionally control for the margin of victory of the female candidate (linearly in Column 3,
with a third order polynomial in Column 4), allowing for different slopes on the two sides
of the discontinuity point. Thus, the gender dummy captures the difference in height of the
regression function at the threshold, i.e., the effect of electing a female representative in a
tightly contested electoral race.
Even after controlling for a host of individual and district characteristics, we still find
evidence that female-sponsored bills are able to attract a larger number of cosponsors in
the full sample. This conclusion hold true even when we restrict attention to the subsample
of mixed-gender electoral races. However, the results become more mixed when we include
controls for the margin of victory (columns 3 and 4), with the actual magnitude and significance of the coefficients depending on the exact specification. Also, the estimated gender
coefficient is quite similar for Democrats and Republicans in the OLS specifications, with
the latter appearing to be also robust to the control for margin of victory.
Table 5 presents the results where the unit of observation is the individual Congress
member, and the dependent variable is the number of bills cosponsored. Again, we find
evidence of a positive gender gap in the OLS specifications (albeit mostly not significant),
which disappears and even becomes negative once we control for the margin of victory.
In Table 6 the unit of observation is a congressional bill, and the dependent variable is the
fraction of cosponsors from the opposite party, one measure of bipartisanship. When looking
at the sample as a whole, there appears to be no meaningful gender difference, neither in
the OLS specifications, nor in the RD ones. Interestingly, the split by party reveals some
interesting contrasts between Democrats and Republicans. In the OLS specifications, there
is strong evidence that female Republicans are more bipartisan, and female Democrats less.
The gender gap for Democrats becomes small and insignificant in the RD specifications,
while the one for Republicans remains at a roughly similar magnitude, but is not estimated
precisely in the RD specifications.
Table 7 looks at an alternative measure of bipartisanship. The unit of observation is the
individual Congress member, and the dependent variable is the fraction of cosponsored bills
that were sponsored by a member of the opposite party. We find a pattern that is broadly
15
similar to that of Table 6: no gender gap on average; more bipartisanship among female
Republicans and less among female Democrats in the OLS specifications, with the results
going away in the RD specifications.
Additional Results and Robustness Checks
• Tables 8-11: Do women recruit more female sponsors? Female sponsors of the opposite
party? Do they offer more support to bills sponsored by other women? By other women
in the opposite party?
• Tables 12-15: Repeat main analysis, but concentrate only on “women-friendly” bills:
civil rights, health, labor, education, law crime and family issues, social welfare.
• Tables 16-17: Female friendly bills and female cosponsors.
• Down the road:
– What happens to female-sponsored bills? Do they pass the House, the Senate,
become public laws? How long does it take for the bill to go through the various
stages?
– Look at roll-call voting behavior: fraction of opposite-party members that voted
in favor of bills sponsored; fraction of votes in favor of bills sponsored by opposite
party; fraction/number of roll-call votes with opposite party on divisive votes
(party unity score).
– Can also look at alternative measures of bipartisanship, i.e., various indexes scored
by different agencies (ACLU, NRA, etc.).
8
Conclusion
In this paper we have studied gender differences in cooperative behavior in an important
real-world setting with high stakes. In addition, we make an important contribution to
the understanding of cooperative behavior and bipartisanship in the US Congress. We find
that, after controlling for the characteristics of the district in which they are elected, and in
particular the electoral margin of victory, there is little evidence to suggest that women are
inherently more cooperative or bipartisan.
The paper can also have broader implications that go beyond the political arena. Understanding gender differences in cooperative behavior may have important implications for
organizations in a variety of different settings. An organizations success may depend both
16
on the workers incentives to excel and stand out individually relatively to their peers (as
highlighted by the vast literature on rank-order tournaments pioneered by Lazear and Rosen,
1981), but also on the workers propensity to cooperate and work together as a team towards
the common good. A growing recent literature on gender differences in competitiveness
has found that women shy away from competition, and tend to have lower performance in
competitive settings. Some authors have argued that these differences may play a role in
explaining the substantial gender pay gap at the top of the income distribution and the
underrepresentation of women in “power” professions. Is it possible, though, that women
compensate for this lack of “competitive spirit” with a more cooperative attitude? Understanding gender differences in the ability to cooperate may help us understand the factors
that shape the wage distribution, and may also yield insights on the optimal organization of
team production.
References
[1] Andreoni, James and Vesterlund, Lise. 2001. “Which is the Fair Sex? Gender Differences
in Altruism.” Quarterly Journal of Economics, 116, 293-312.
[2] Ansolabehere, Stephen, Snyder, James M. Jr. and Stewart III, Charles. 2001. “The
Effects of Party and Preferences on Congressional Roll-Call Voting.” Legislative Studies
Quarterly, 26, 533-572.
[3] Anzia, Sarah F. and Berry, Christopher R. 2011. “The Jackie (and Jill) Robinson Effect:
Why Do Congresswomen Outperform Congressmen?” American Journal of Political
Science, 55(3), pp. 478-493.
[4] Babcock, Linda and Laschever, Sara. 2003. Women Don’t Ask. Princeton University
Press: New Jersey.
[5] Bertrand, Marianne and Hallock, Kevin F. 2001. “The Gender Gap in Top Corporate
Jobs.” Industrial and Labor Relations Review, 5, 3-21.
[6] Booth, Alison and Nolen, Patrick. 2012. “Choosing to Compete: How Different are Girls
and Boys?” Journal of Economic Behavior and Organization, 81(2), February, 2012. p.
542-555
[7] Box-Steffensmeier, Janet M.; De-Boef, Suzanna and Lin, Tse-Min. 2004. “The Dynamics
of the Partisan Gender Gap.” American Political Science Review, 98: 515-528.
17
[8] Bratton, Kathleen A. and Haynie, Kerry L. 1999. “Agenda Setting and Legislative
Success in State Legislatures: The Effects of Gender and Race.” Journal of Politics, 61:
658-679.
[9] Brewer, Mark D., Mariani, Mack D. and Stonecash, Jeffrey M. 2002. “Northern
Democrats and Party Polarization in the U.S. House.” Legislative Studies Quarterly
27, pp. 42344.
[10] Brollo, Fernanda and Troiano, Ugo. 2012. “What Happens When a Woman Wins a
Close Election? Evidence from Brazil.” Mimeo., Universidad de Alicante.
[11] Brown-Kruse, Jamie and Hummels, David. 1993. “Gender Effects in Laboratory Public
Goods Contribution: Do Individuals Put Their Money Where Their Mouth Is?” Journal
of Economic Behavior and Organization. 22, 255-267.
[12] Browne, William P. 1985. “Multiple Sponsorship and Bill Success in U. S. State Legislatures.” Legislative Studies Quarterly , 10(4), pp. 483-488.
[13] Campbell, James E. 1982. “Cosponsoring Legislation in the U.S. Congress.” Legislative
Studies Quarterly, 7(3), pp. 415-422.
[14] Chattopadhyay, Raghabendra and Duflo, Esther. 2004. “Women as Policy Makers: Evidence from a Randomized Policy Experiment in India.” Econometrica, 72, 1409-1443.
[15] Clots-Figueras, Irma. 2011. “Women in Politics. Evidence from the Indian States.”
Journal of Public Economics, 95, pp. 664-690.
[16] Clots-Figueras, Irma. 2012. “Are Female Leaders Good for Education? Evidence from
India.” American Economic Journal: Applied Economics, 4(1), pp. 212-244.
[17] Coleman, John J. 1997. “The Decline and Resurgence of Congressional Party Conflict.”
Journal of Politics 59(1), pp. 16584.
[18] Collie, Melissa P. and Mason, John Lyman. 2000. “The Electoral Connection Between
Party and Constituency Reconsidered: Evidence from the U.S. House of Representatives, 19721994,” in David W. Brady, John F. Cogan and Morris P. Fiorina (eds.)
Continuity and Change in House Elections. Stanford, CA: Stanford University Press.
[19] Cox, Gary W. and Poole, Keith T. 2002. “On Measuring Partisanship in Roll-Call
Voting: The U.S. House of Representatives, 1877-1999.” American Journal of Political
Science, 46, 477-489.
18
[20] Croson, Rachel and Gneezy, Uri. 2009. “Gender Differences in Preferences.” Journal of
Economic Literature, 47(2), pp. 448-474.
[21] Dohmen, Thomas and Falk, Armin. 2011. “Performance Pay and Multi-dimensional
Sorting: Productivity, Preferences and Gender.” American Economic Review, 101(2),
pp. 556-590.
[22] Eckel, Catherine and Grossman, Philip J. 1996. “The Relative Price of Fairness: Gender
Differences in a Punishment Game.” Journal of Economic Behavior and Organization,
30, 143-58.
[23] Eckel, Catherine and Grossman, Philip J. 1998. “Are Women Less Selfish than Men?
Evidence from Dictator Games.” Economic Journal, 108, 726-735.
[24] Edlund, Lena and Pande, Rohini. 2002. “Why Have Women Become Left-Wing? The
Political Gender Gap and the Decline in Marriage.” Quarterly Journal of Economics,
117: 917-961.
[25] Ferreira, Fernando and Gyourko, Joseph. 2011. “Does Gender Matter for Political Leadership? The Case of U.S. Mayors.” Mimeo., The Wharton School, University of Pennsylvania.
[26] Fleisher, Richard and Bond, Jon R. 2000. “Partisanship and the Presidents Quest for
Votes on the Floor of Congress,” in Jon R. Bond and Richard Fleisher (eds.) Polarized
Politics: Congress and the President in a Partisan Era. Washington, DC: CQ Press.
[27] Funk, Patricia and Gathmann, Christina. 2008. “Gender Gaps in Policy Making: Evidence from Direct Democracy in Switzerland.” Mimeo., Universitat Pompeu Fabra.
[28] Gagliarducci, Stefano; Nannicini, Tommaso, and Naticchioni, Paolo. 2011. “Electoral
Rules and Politicians’ Behavior: A Micro Test.” American Economic Journal: Economic
Policy, 3(3), pp. 144-174.
[29] Gagliarducci, Stefano and Paserman, M. Daniele. 2012. “Gender Interactions in Hierarchies: Evidence from the Political Arena.” Review of Economic Studies, 79(3), pp.
1021-1052.
[30] Gneezy, Uri, Niederle, Muriel and Rustichini, Aldo. 2003. “Performance in Competitive
Environments: Gender Differences.” Quarterly Journal of Economics, 118, 1049-1074.
19
[31] Gneezy, Uri and Rustichini, Aldo. 2004. “Gender and Competition at a Young Age.”
American Economic Review, 94, 377-381.
[32] Harbridge, Laurel M. 2009. Bipartisanship in a Polarized Congress. Ph.D. Dissertation,
Stanford University.
[33] Harbridge, Laurel M. 2010. “The Elasticity of Partisanship in Congress: An Analysis
of Legislative Bipartisanship.” Mimeo., Northwestern University.
[34] Harbridge, Laurel M. 2011. “Congressional Agenda Control and the Decline of Bipartisan Cooperation.” Mimeo., Northwestern University.
[35] Herrnson, Paul S., Lay, J. Celeste and Stokes-Brown, Atiya Kai. 2003. “Women Running as Women: Candidate Gender, Campaign Issues and Voter Targeting Strategies.”
Journal of Politics, 65, 244-25.
[36] Huddy, Leonie and Terkildsen, Nayda. 1993. “Gender Stereotypes and the Perception
of Male and Female Candidates.” American Journal of Political Science, 37, 119-147.
[37] Imbens, Guido and Lemieux, Tomas. 2008. “Regression Discontinuity Designs: A Guide
to Practice.” Journal of Econometrics, 142, 615-635.
[38] Jenkins, Jeffrey A. 1999. “Examining the Bonding Effects of Party: A Comparative
Analysis of Roll-Call Voting in the U.S. and Confederate Houses.” American Journal of
Political Science, 43, 1144-1165.
[39] Jeydel, Alana and Taylor, Andrew J. 2003. “Are Women Legislators Less Effective?
Evidence from the U.S. House in the 103rd-105th Congress.” Journal of Politics, 56(1):
19-27.
[40] Kessler, Daniel and Krehbiel, Keith. 1996. “Dynamics of Cosponsorship.” The American
Political Science Review, 90(3), pp. 555-566.
[41] Krehbiel, Keith. 1995. “Cosponsors and Wafflers from A to Z.” American Journal of
Political Science, 39(4), pp. 906-923.
[42] Lavy, Victor. 2008a. “Do Gender Stereotypes Reduce Girls’ or Boys’ Human Capital
Outcomes? Evidence from a Natural Experiment.” Journal of Public Economics, forthcoming.
20
[43] Lavy, Victor. 2008b. “Gender Differences in Competitiveness in a Real Workplace: Evidence from Performance-Based Pay Tournaments among Teachers.” Mimeo., Hebrew
University.
[44] Lazear, Edward P. and Rosen, Sherwin. 1981. “Rank-Order Tournaments as Optimum
Labor Contracts.” Journal of Political Economy, 89(5), pp. 841-864.
[45] Lee, David S.. 2008. “Randomized Experiments from Non-random Selection in the U.S.
House Elections.” Journal of Econometrics, 142, 675-697.
[46] Lee, David S., Moretti, Enrico and Butler, Matthew J. 2004. “Do Voters Affect or
Elect Policies? Evidence from the U.S. House.” Quarterly Journal of Economics, 119,
807-859.
[47] Manning, Alan and Saidi, Farzad. 2010. “Understanding the Gender Pay Gap: Whats
Competition Got to Do with It?” Industrial and Labor Relations Review, 63(4), pp.
681-698.
[48] Mayhew, David. 1974. Congress: The Electoral Connection. New Haven: Yale University
Press.
[49] McCarty, Nolan; Poole, Keith T. and Rosenthal, Howard. 2001. “The Hunt for Party
Discipline in Congress.” American Political Science Review, 95, 673-687.
[50] Niederle, Muriel and Vesterlund, Lise. 2007. “Do Women Shy Away from Competition?
Do Men Compete too Much?” Quarterly Journal of Economics, 122, 1067-1101.
[51] Niederle, Muriel and Yestrumskas, Alexandra H. 2008. “Gender Differences in Seeking
Challenges: The Role of Institutions.” NBER Working Paper No. 13922.
[52] Norton, Noelle H. 1999. “Uncovering the Dimensionality of Gender Voting in Congress.”
Legislative Studies Quarterly 24: 65-86.
[53] Nowell, Clifford, and Tinkler, Sarah. 1994. “The Influence of Gender in the Provision
of a Public Good.” Journal of Economic Behavior and Organization, 25, 25-36.
[54] Örs, Evren; Palomino, Frdric and Peyrache, Eloc. 2008. “Performance Gender Gap:
Does Competition Matter?” CEPR Discussion Paper No. 6891.
[55] Paserman, M. Daniele. 2010. “Gender Differences in Performance in Competitive Environments? Evidence from Professional Tennis Players.” Mimeo, Boston University.
21
[56] Poole, Keith T. and Rosenthal, Howard. 1984. “The Polarization of American Politics.”
Journal of Politics, 46(4), pp.1061-1079.
[57] Poole, Keith T. and Rosenthal, Howard. 1985. “A Spatial Model for Legislative Roll
Call Analysis.” American Journal of Political Science, 29, 357-384.
[58] Rehavi, Marit. 2007. “Sex and Politics: Do Female Legislators Affect State Spending?”
Mimeo., University of British Columbia.
[59] Rohde, David W. 1991. Parties and Leaders in the Postreform House. Chicago, IL:
University of Chicago Press.
[60] Sanbonmatsu, Kira. 2002. “Gender Stereotypes and Vote Choice.” American Journal
of Political Science, 46, 20-34.
[61] Sapiro, Virginia and Conover, Pamela Johnston. 1997. “The Variable Gender Basis of
Electoral Politics: Gender and Context in the 1992 US Election.” British Journal of
Political Science, 27: 497-523.
[62] Sapiro, Virginia. 2003. “Theorizing Gender in Political Psychology Research,” in Sears,
David O., Huddy, Leonie and Jervis, Robert L. (eds.) Oxford Handbook of Political
Psychology, 601-34. New York: Oxford University Press.
[63] Schiller, Wendy J. 1995. “Senators as Political Entrepreneurs: Using Bill Sponsorship to
Shape Legislative Agendas.” American Journal of Political Science, 39(1), pp. 186-203.
[64] Schneier, Edward P. and Gross, Bertram. 1993. Legislative Strategy: Shaping Public
Policy. New York: St. Martins.
[65] Snyder, James M. Jr. and Groseclose, Tim. 2000. “Estimating Party Influence in Congressional Roll-Call Voting.” American Journal of Political Science, 44, 193-211.
[66] Stonecash, Jeffrey M.; Brewer, Mark D. and Mariani, Mark D. 2003. Diverging Parties:
Social Change, Realignment, and Party Polarization. Boulder, CO: Westview Press.
[67] Swers, Michele L. 1998. “Are Women More Likely to Vote for Womens Issue Bills Than
Their Male Colleagues?” Legislative Studies Quarterly, 23: 435-448.
[68] Theriault, Sean. 2006. “Party Polarization In The U.S. Congress: Member Replacement
and Member Adaptation.” Party Politics, 12(4), pp. 483-503
22
[69] Vega, Arturo and Firestone, Juanita M. 1995. “The Effects of Gender on Congressional
Behavior and the Substantive Representation of Women.” Legislative Studies Quarterly,
20: 213-222.
[70] Wawro, Gregory. 2001. Legislative Entrepreneurship in the U.S. House of Representatives. Ann Arbor (MI), University of Michigan Press.
[71] Welch, Susan. 1985. “Are Women More Liberal than Men in the U.S. Congress?” Legislative Studies Quarterly, 10: 125-134.
[72] Wilson, Rick K. and Young, Cheryl D. 1997. “Cosponsorship in the U.S. Congress.”
Legislative Studies Quarterly, 22(1), pp. 25-43.
23
.25
Fraction female
.15
.2
.1
.05
1990
2000
1995
Year Congress Elected
All
Republicans
2005
Democrats
Figure 1: Fraction Female Congress Members
Figures and Tables - Page 1
2010
24
Mean number of cosponsors
16
18
20
22
14
1990
2000
1995
Year Congress Elected
Total
Males
2005
2010
Females
Mean number of bills cosponsored
150
200
250
300
350
Figure 2a: Mean Number of Cosponsors
1990
2000
1995
Year Congress Elected
Total
Males
2005
Females
Figure 2b: Mean Number of Bills Cosponsored
2010
Fraction of Cosponsors of Opposite Party
.15
.2
.25
.3
.35
1990
2000
1995
Year Congress Elected
Total
Males
2005
2010
Females
.15
Fraction of cosponsored bills
sponsored by opposite party
.2
.25
.3
.35
Figure 2c: Fraction of Cosponsors of Opposite Party
1990
1995
2000
Year Congress Elected
Total
Males
2005
Females
Figure 2d: Fraction of Cosponsored Bills Sponsored by Opposite Party
2010
.02
0
.005
.01
.015
Bandwidth = 5
-100
-50
0
Figure 3: McCrary Density Test
50
100
Mid-West
-.5
-.5
0
0
.5
.5
1
1
North-East
-100
-50
0
50
Margin victory female sponsor
-100
100
0
50
-50
Margin victory female sponsor
100
West
0
-.5
.2
0
.4
.6
.5
.8
1
1
South
-100
-50
0
50
Margin victory female sponsor
100
-100
-50
0
50
Margin victory female sponsor
100
Predicted values - 3rd order polynomial
Observed values
Figure 4a: Balancing Tests - District Characteristics 1
% Over-65
% Urban
1
.8
.7
.6
.05
-100 -50
0
50
100
Margin victory female sponsor
% Black
Median Income
% Democratic t-1
Observed values
1
.8
.6
.4
.8
.6
.4
.2
0
-100
-50
0
50
100
Margin victory female sponsor
1.2
-100 -50
0
50
100
Margin victory female sponsor
20000 30000 40000
15000 25000 35000
-100
-50
0
50
100
Margin victory female sponsor
1
500000
.1
550000
.15
.9
.2
600000
.25
650000
Population
-100 -50
0
50
100
Margin victory female sponsor
-100 -50
0
50
100
Margin victory female sponsor
Predicted values - 3rd order polynomial
Figure 4b: Balancing Tests - District Characteristics 2
Age
.6
.8
Rookie
-.2
0
30
0
40
5
.2
50
.4
10
60
70
15
Terms in office
-100
-50
0
50
100
Margin victory female sponsor
Incumbent
Democrat
-100
-50
0
50
100
Margin victory female sponsor
0
0
.2
.5
.4
.6
1
.8
1
1.5
-100
-50
0
50
100
Margin victory female sponsor
-100
-50
0
50
100
Margin victory female sponsor
Observed values
-100
-50
0
50
100
Margin victory female sponsor
Predicted values, 3rd order polynomial
Figure 5: Balancing Tests - Member Characteristics
All
0
0
10
10
20
20
30
30
40
40
Democrats
-100
-50
0
50
Margin victory female sponsor
100
-100
-50
0
50
Margin victory female sponsor
100
0
10
20
30
40
Republicans
-100
-50
0
50
Margin victory female sponsor
Actual
100
Fitted values - 3rd order polynomial
Figure 6: Number of Cosponsors - RD Analysis
Democrats
0
100
100
200
200
300
300
400
400
500
All
-100
-50
0
50
Margin victory female cosponsor
100
-100
-50
0
50
Margin victory female cosponsor
150 250 350
100 200 300
Republicans
-100
-50
0
50
Margin victory female cosponsor
Actual
100
Fitted values - 3rd order polynomial
Figure 7: Number of Bills Cosponsored - RD Analysis
100
Democrats
0
0
.1
.1
.2
.2
.3
.3
.4
.4
All
-100
-50
0
50
Margin victory female sponsor
-100
100
0
50
-50
Margin victory female sponsor
100
.2
.3
.4
.5
.6
.7
Republicans
-100
50
0
-50
Margin victory female sponsor
Actual
100
Fitted values - 3rd order polynomial
Figure 8: Fraction of Cosponsors of the Opposite Party - RD Analysis
Democrats
.1
.1
.2
.2
.3
.3
.4
.4
All
-100
50
0
-50
Margin victory female cosponsor
100
-100
50
-50
0
Margin victory female cosponsor
.1
.2
.3
.4
.5
.6
Republicans
-100
-50
0
50
Margin victory female cosponsor
Actual
100
Fitted values - 3rd order polynomial
Figure 9: Fraction of Cosponsored Bills that Were Sponsored
by the Opposite Party - RD Analysis
100
Table 1: Summary Statistics
Panel A: Bill Characteristics
Number of bills
All
62911
Male sponsor
54105
Female Sponsor
8806
Number of bills with at least one
cosponsor
Number of cosponsors: mean (std.
dev.)
Number of cosponsors: median
42065
35725
6340
16.58
(35.58)
3
16.12
(35.15)
3
19.46
(37.97)
5
Percent cosponsors opposite party:
mean (std. dev.)
27.2
(0.307)
28.1
(30.8)
22.6
(29.4)
All
4885
Male sponsor
4273
Female Sponsor
603
Number of bills cosponsored: mean
(std. dev.)
Number of bills cosponsored: median
212.1
(114.8)
189
205.4
(111.3)
183
261.6
(125.7)
245
Percent of cosponsored bills from
opposite party
27.1
(16.9)
27.3
(17.0)
25.5
(15.6)
Panel B: Cosponsor Characteristics
Number of Congress Members
Table 2: McCrary Density Tests for Different Bandwidth Values
Discontinuity Estimate (log
difference in height)
h=1
0.588
(1.222)
h=3
1.414
(0.940)
Page 10 of 25
h=5
0.564
(0.563)
h=7
0.136
(0.428)
h=9
-0.200
(0.365)
Table 3: Balancing Tests
Type of Control for Margin of Victory
Linear
3rd Order Polynomial
A: District Characteristics
Northeast
0.035
[0.074]
-0.141
[0.092]
Midwest
-0.144*
[0.076]
-0.071
[0.099]
South
0.051
[0.079]
-0.127
[0.099]
West
0.060
[0.097]
0.348***
[0.122]
-7,694.21
[6,582.725]
-11,619.21
[8,481.225]
% Black
-0.006
[0.024]
-0.015
[0.026]
% Urban
0.046
[0.045]
0.016
[0.062]
% Foreign Born
0.002
[0.007]
-0.012
[0.009]
% over 65
0.002
[0.007]
-0.012
[0.008]
1,218.03
[1,630.972]
-1,181.69
[2,101.587]
0.033
[0.026]
0.026
[0.040]
0.860
[1.788]
1.232
[2.371]
-1.590***
[0.529]
-1.856**
[0.774]
Rookie
0.032
[0.053]
0.090
[0.100]
Incumbent
-0.009
[0.061]
-0.081
[0.120]
Democrat
0.302***
[0.089]
0.347***
[0.117]
Population
Median income
Lagged democratic share
B: Member Characteristics
Age
Terms in Congress
Notes: All specifications control for congress fixed effects.
Page 11
Table 4: Gender and the Number of Cosponsors
(1)
OLS - Full Sample
(2)
OLS- Mixed Gender
Elections
(3)
RD - Linear
(4)
RD - 3rd order
polynomial
All
2.335**
[1.186]
61,285
2.981**
[1.335]
12,532
3.591*
[1.878]
11,978
1.366
[2.989]
11,978
Democrats
1.731
[1.548]
33,032
2.609
[2.009]
6,729
2.290
[3.075]
6,186
-0.316
[4.831]
6,186
Republicans
3.095*
[1.707]
28,101
1.988
[1.721]
5,803
5.784**
[2.478]
5,792
5.327
[3.741]
5,792
Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the
individual-congress level, in parentheses. All regressions control for sponsor age, terms in congress, a dummy for
whether the sponsor is a rookie, party, district demographic characteristics (see Table 3) and congress fixed effects.
Table 5: Gender and the Number of Bills Cosponsored
(1)
OLS - Full Sample
(2)
OLS- Mixed Gender
Elections
(3)
RD - Linear
(4)
RD - 3rd order
polynomial
All
19.047**
[9.027]
4,859
15.586
[9.681]
930
8.344
[16.080]
905
-17.932
[21.277]
905
Democrats
16.827
[12.970]
2,552
15.157
[15.701]
472
25.951
[24.929]
448
-17.056
[36.503]
448
Republicans
13.49
[9.194]
2,297
7.661
[10.842]
458
7.646
[19.181]
457
5.129
[26.326]
457
Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the
individual-congress level, in parentheses. All regressions control for congress member age, terms in congress, whether
the congress members is a rookie, party, district demographic characteristics and congress fixed effects.
Table 6: Gender and the Fraction of Cosponsors of the Opposite Party
(1)
OLS - Full Sample
(2)
OLS- Mixed Gender
Elections
(3)
RD - Linear
(4)
RD - 3rd order
polynomial
-0.008
[0.015]
41,958
0.020
[0.017]
8,778
0.037
[0.028]
8,385
0.068
[0.043]
8,385
Democrats
-0.044***
[0.011]
22,856
-0.020
[0.015]
4,844
-0.020
[0.025]
4,458
-0.017
[0.041]
4,458
Republicans
0.063**
[0.025]
18,988
0.054**
[0.027]
3,934
0.054
[0.044]
3,927
0.078
[0.095]
3,927
All
Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the
individual-congress level, in parentheses. All regressions control for sponsor age, terms in congress, a dummy for
whether the sponsor is a rookie, party, district demographic characteristics (see Table 3) and congress fixed effects.
Table 7: Gender and the Fraction of Cosponsored Bills Sponsored by the Opposite Party
(1)
OLS - Full Sample
(2)
OLS- Mixed Gender
Elections
(3)
RD - Linear
(4)
RD - 3rd order
polynomial
0.002
[0.010]
4,847
-0.014
[0.015]
928
0.004
[0.025]
903
-0.017
[0.041]
903
Democrats
-0.029***
[0.007]
2,545
-0.021**
[0.009]
471
-0.014
[0.017]
447
-0.002
[0.025]
447
Republicans
0.039***
[0.014]
2,292
0.027*
[0.016]
457
-0.012
[0.043]
456
0.042
[0.032]
456
All
Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the
individual-congress level, in parentheses. All regressions control for congress member age, terms in congress, whether
the congress members is a rookie, party, district demographic characteristics and congress fixed effects.
Table 8: Gender and the Fraction of Female Cosponsors
(1)
OLS - Full Sample
(2)
OLS- Mixed Gender
Elections
(3)
RD - Linear
(4)
RD - 3rd order
polynomial
All
0.049***
[0.005]
41,992
0.033***
[0.008]
8,792
0.015
[0.015]
8,398
-0.003
[0.026]
8,398
Democrats
0.054***
[0.006]
22,874
0.034***
[0.009]
4,851
0.033**
[0.016]
4,464
0.002
[0.027]
4,464
Republicans
0.033***
[0.011]
19,004
0.027*
[0.014]
3,941
-0.007
[0.027]
3,934
0.012
[0.050]
3,934
Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the
individual-congress level, in parentheses. All regressions control for sponsor age, terms in congress, party, and congress
fixed effects.
Table 9: Gender and the Fraction of Cosponsored Bills sponsored by Females
(1)
OLS - Full Sample
(2)
OLS- Mixed Gender
Elections
(3)
RD - Linear
(4)
RD - 3rd order
polynomial
All
0.049***
[0.004]
4,851
0.038***
[0.005]
931
0.023***
[0.008]
906
0.013
[0.011]
906
Democrats
0.054***
[0.004]
2,546
0.039***
[0.006]
472
0.034***
[0.011]
448
0.019
[0.014]
448
Republicans
0.034***
[0.006]
2,295
0.030***
[0.008]
459
0.013
[0.011]
458
0.007
[0.013]
458
Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the
individual-congress level, in parentheses. All regressions control for sponsor age, terms in congress, party, and congress
fixed effects.
Table 10: Gender and the Fraction of Female Cosponsors of the Opposite Party
(1)
OLS - Full Sample
(2)
OLS- Mixed Gender
Elections
(3)
RD - Linear
(4)
RD - 3rd order
polynomial
All
0.009*
[0.005]
41,992
0.013**
[0.007]
8,792
0.013
[0.011]
8,398
-0.008
[0.017]
8,398
Democrats
0.001
[0.003]
22,874
0.004
[0.004]
4,851
0.004
[0.007]
4,464
-0.014
[0.014]
4,464
Republicans
0.029**
[0.011]
19,004
0.023*
[0.013]
3,941
-0.002
[0.023]
3,934
0.009
[0.037]
3,934
Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the
individual-congress level, in parentheses. All regressions control for sponsor age, terms in congress, party, and congress
fixed effects.
Table 11: Gender and the Fraction of Cosponsored Bills Sponsored by Females of the Opposite Party
(1)
OLS - Full Sample
(2)
OLS- Mixed Gender
Elections
(3)
RD - Linear
(4)
RD - 3rd order
polynomial
0.010***
[0.003]
4,851
0.008**
[0.004]
931
0.012**
[0.006]
906
-0.002
[0.007]
906
Democrats
0.002
[0.001]
2,546
0.002
[0.002]
472
0.002
[0.004]
448
-0.003
[0.006]
448
Republicans
0.024***
[0.006]
2,295
0.018**
[0.007]
459
0.008
[0.009]
458
-0.016*
[0.009]
458
All
Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the
individual-congress level, in parentheses. All regressions control for sponsor age, terms in congress, party, and congress
fixed effects.
Table 12: Gender and the Number of Cosponsors on "Women-Friendly" Bills
(1)
OLS - Full Sample
(2)
OLS- Mixed Gender
Elections
(3)
RD - Linear
(4)
RD - 3rd order
polynomial
All
3.217*
[1.704]
18971
3.737**
[1.830]
4263
4.626
[3.455]
4050
5.857
[6.111]
4050
Democrats
2.886
[2.038]
11061
3.811
[2.517]
2594
1.797
[4.788]
2381
3.337
[9.665]
2381
Republicans
4.712*
[2.643]
7853
2.756
[2.386]
1669
9.372*
[4.818]
1669
11.525
[8.582]
1669
Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the
individual-congress level, in parentheses. All regressions control for sponsor age, terms in congress, party, and congress
fixed effects. Bills coded as "women-friendly" if the major topic is one of the following: Health; Labor, Employment, and
Immigration; Education; Law, Crime, and Family Issues; Social Welfare.
Table 13: Gender and the Number of "Women-Friendly" Bills Cosponsored
(1)
OLS - Full Sample
(2)
OLS- Mixed Gender
Elections
(3)
RD - Linear
(4)
RD - 3rd order
polynomial
All
19.940***
[4.197]
4,863
17.035***
[4.893]
933
9.457
[7.839]
908
-13.073
[9.701]
908
Democrats
21.191***
[5.767]
2,553
18.569**
[7.270]
473
23.957**
[11.475]
449
-11.606
[16.379]
449
Republicans
12.739***
[4.487]
2,300
12.684**
[5.837]
460
8.234
[9.765]
459
0.749
[10.105]
459
Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the
individual-congress level, in parentheses. All regressions control for sponsor age, terms in congress, party, and congress
fixed effects. Bills coded as "women-friendly" if the major topic is one of the following: Health; Labor, Employment, and
Immigration; Education; Law, Crime, and Family Issues; Social Welfare.
Table 14: Gender and the Fraction of Cosponsors of the Opposite Party on "Women-Friendly" Bills
(1)
OLS - Full Sample
(2)
OLS- Mixed Gender
Elections
(3)
RD - Linear
(4)
RD - 3rd order
polynomial
0.017
[0.018]
14,329
0.051**
[0.024]
3,306
0.115***
[0.038]
3,156
0.133***
[0.050]
3,156
Democrats
-0.041***
[0.011]
8,208
-0.013
[0.018]
1,978
0.021
[0.032]
1,828
0.018
[0.050]
1,828
Republicans
0.134***
[0.036]
6,077
0.115***
[0.041]
1,328
0.136**
[0.053]
1,328
0.195*
[0.106]
1,328
All
Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the
individual-congress level, in parentheses. All regressions control for sponsor age, terms in congress, party, and congress
fixed effects. Bills coded as "women-friendly" if the major topic is one of the following: Health; Labor, Employment, and
Immigration; Education; Law, Crime, and Family Issues; Social Welfare.
Table 15: Gender and the Fraction of Cosponsored "Women-Friendly" Bills sponsored by the Opposite Party
(1)
OLS - Full Sample
(2)
OLS- Mixed Gender
Elections
(3)
RD - Linear
(4)
RD - 3rd order
polynomial
-0.01
[0.012]
4,845
-0.007
[0.017]
931
0.011
[0.027]
906
-0.038
[0.039]
906
Democrats
-0.062***
[0.009]
2,542
-0.036***
[0.012]
472
-0.040**
[0.018]
448
-0.018
[0.030]
448
Republicans
0.055***
[0.021]
2,293
0.039
[0.025]
459
0.016
[0.033]
458
-0.046
[0.037]
458
All
Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the
individual-congress level, in parentheses. All regressions control for sponsor age, terms in congress, party, and congress
fixed effects. Bills coded as "women-friendly" if the major topic is one of the following: Health; Labor, Employment, and
Immigration; Education; Law, Crime, and Family Issues; Social Welfare.
Table 16: Gender and the Fraction of Female Cosponsors on "Women-Friendly" Bills
(1)
OLS - Full Sample
(2)
OLS- Mixed Gender
Elections
(3)
RD - Linear
(4)
RD - 3rd order
polynomial
All
0.063***
[0.007]
14,329
0.053***
[0.012]
3,306
0.059***
[0.020]
3,156
0.053
[0.034]
3,156
Democrats
0.063***
[0.008]
8,208
0.039***
[0.011]
1,978
0.044*
[0.024]
1,828
0.039
[0.049]
1,828
Republicans
0.059***
[0.015]
6,077
0.063***
[0.019]
1,328
0.070**
[0.032]
1,328
0.079
[0.052]
1,328
Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the
individual-congress level, in parentheses. All regressions control for sponsor age, terms in congress, party, and congress
fixed effects. Bills coded as "women-friendly" if the major topic is one of the following: Health; Labor, Employment, and
Immigration; Education; Law, Crime, and Family Issues; Social Welfare.
Table 17: Gender and the Fraction of Cosponsored "Women-Friendly" Bills sponsored by Females
(1)
OLS - Full Sample
(2)
OLS- Mixed Gender
Elections
(3)
RD - Linear
(4)
RD - 3rd order
polynomial
All
0.074***
[0.005]
4,845
0.059***
[0.007]
931
0.047***
[0.012]
906
0.022
[0.017]
906
Democrats
0.077***
[0.005]
2,542
0.058***
[0.008]
472
0.067***
[0.014]
448
0.041*
[0.022]
448
Republicans
0.058***
[0.011]
2,293
0.049***
[0.012]
459
0.026
[0.018]
458
0.007
[0.026]
458
Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the
individual-congress level, in parentheses. All regressions control for sponsor age, terms in congress, party, and congress
fixed effects. Bills coded as "women-friendly" if the major topic is one of the following: Health; Labor, Employment, and
Immigration; Education; Law, Crime, and Family Issues; Social Welfare.