Gender Differences in Cooperative Environments? Evidence from the U.S. Congress Stefano Gagliarducci and M. Daniele Paserman University of Rome Tor Vergata and Boston University Preliminary: suggestions welcome April 2015 Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the April U.S. 2015 Congress1 / 44 Introduction Main question: Do women in the U.S. Congress engage in more cooperative behavior? Specifically: do they build larger coalitions? Coalitions that cross party lines? Question is relevant for at least three literatures: Gender differences in the performance of elected politicians. Cooperation and partisanship in Congress. More broadly: gender differences in cooperative behavior in a real-world setting with high stakes → implications that go beyond the political arena? Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the April U.S. 2015 Congress2 / 44 Motivation 1 - Gender Differences in Cooperativeness Commonly held view that women are better at building compromise (Pew Research Center, 2015): Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the April U.S. 2015 Congress3 / 44 Motivation 1 - Gender Differences in Cooperativeness Recent flurry of work in economics (mostly experimental) on gender differences in psychological traits and preferences. Niederle (2015) summarizes experimental (lab and field) literature on three traits: Competitiveness: Women shy away from competition. Risk aversion: Most studies find women to be more risk averse, but evidence more mixed with newer elicitation methods. Altruism or cooperative attitudes: evidence very mixed, essentially no field evidence. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the April U.S. 2015 Congress4 / 44 Motivation 1 - Gender Differences in Cooperativeness Can these differences explain differences in labor market outcomes? Much focus on competitiveness, but it is not the only trait employers care about. In some contexts, maybe more important to be able to reach out to others, work together toward the common goal, build coalitions. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the April U.S. 2015 Congress5 / 44 Motivation 2: Bipartisanship in Congress Large literature in political science on partisanship and polarization in Congress. Poole and Rosenthal’s NOMINATE score (1985) and its refinements (D-NOMINATE, DW-NOMINATE), based on roll-call voting behavior (http://www.voteview.com). Party polarization has been on the increase since 1970s, acceleration since the 1990s. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the April U.S. 2015 Congress6 / 44 Motivation 2 - Bipartisanship in Congress Partisanship is often viewed as a major obstacle to good government. Common view that more women would increase bipartisanship: Women may be our best chance at breaking through disastrous partisan gridlock in Congress...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.” [Swanee Hunt (former US ambassador), Boston Globe, January 2, 2013]. Is there substance to these claims? Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the April U.S. 2015 Congress7 / 44 Motivation 3 - Gender Differences in Policy Outcomes? In the US Congress: Anzia and Berry (AJPS 2011): female congress members secure more discretionary federal spending to their districts. They interpret this as evidence of bias on the part of voters, who elect only most talented and hard-working women. Similar finding for Brazilian municipalities in Brollo and Troiano (2012). In other contexts, evidence more mixed: Female leaders choose more female-friendly laws (Chattophadyay and Duflo, 2004; Clots-Figueras, 2011, 2012; Rehavi 2007; Funk and Gathmann 2008). Gagliarducci and Paserman (2012), Ferreira and Gyourko: no evidence that female mayors (in Italy and the US) achieve better policy outcomes. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the April U.S. 2015 Congress8 / 44 This paper Focus on cosponsorship of congressional bills in the US House of Representatives, 1989-2010. Two measures of cooperative behavior: Number of cosponsors on bills sponsored (coalition-building). Number/fraction of cosponsors from opposite party (bipartisanship). Also: Type of coalitions formed (with women, with more senior legislators...) Topic of the bill (”women-friendly” topics and not) Outcomes of the bills (passed House, became public law) Methodology: First show OLS with varying sets of controls. How to address selection of districts in which women are elected? RD (with caution) and propensity-score weighting. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the April U.S. 2015 Congress9 / 44 Unique contribution: Test of gender differences in cooperative behavior using real-world setting (US Congress), high stakes decisions. There may be more evidence of bipartisanship in bill sponsorship and cosponsorship: Polarization literature has focused on roll call voting behavior. But sample of bills that reach the floor misses part of the picture, party leadership has interest in pushing forward divisive agenda (Harbridge, 2009). Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 10 / 44 A note about cooperativeness Three possible interpretations of cooperativeness: 1 Altruism: Legislator 1 wants to pass a bill, needs support from legislator 2. Legislator 2 is indifferent to content of bill, offers support. 2 Public Good: Same as above, but legislator 2 cares about bill, but would rather free ride on someone else’s work. 3 Battle of the Sexes: Tax Hike Spending Cut Tax Hike 2,1 0,0 Spending Cut 0,0 1,2 In all three cases, we can think of acting cooperatively or inducing cooperation. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 11 / 44 Preview of the Findings Coalition-building: Women have higher number of cosponsors on bills they sponsor. Differences almost disappear after controlling for selection. Bipartisanship: No gender difference in number cosponsors of opposite party. But large differences by party: female Republicans have more opposite-party (and female) cosponsors, female Democrats fewer. All effects amplified on female-issues. Female-sponsored bills no more likely to become public law. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 12 / 44 Outline Introduction and Motivation Institutional Framework Data and Methodology Results Conclusion Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 13 / 44 Institutional Background House of Representatives: lower chamber of US legislative branch. 435 congressional districts, each state is represented in proportion to population. Elections for all districts every two years. Representatives draft legislation, must also pass Senate and be approved by President before becoming law. Different types of legislation: Bills Simple Resolutions Concurrent Resolutions Joint Resolutions Will focus on House bills, have force of law and are primary vehicle for members’ legislative efforts. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 14 / 44 Institutional Background - Sponsorship Each bill has one primary sponsor. Not necessarily sole or most important author of bill. However, sponsor is identified with bill content. Sponsor’s activities: gathering and communicating information coalition building public relations shepherding legislation through House. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 15 / 44 Institutional Background - Cosponsorship Bill can be signed by any number of cosponsors. Some debate about whether cosponsorship is just position-taking. Wawro (2001): “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. 0 cosponsors 1-10 cosponsors 11+ cosponsors Gagliarducci and Paserman Passed the House 15.6% 20.6% 21.0% Became Public Law 4.2% 5.4% 5.3% Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 16 / 44 Data Focus on Congresses 101-111 (1989-2010) Information on bill sponsorship, content, outcome, number and identity of cosponsors from THOMAS (Library of Congress). Information on individual congress members (age, gender, tenure in congress, committee membership) from ICPSR, Biographical Directory of Congress. Information on electoral outcomes (votes by party, identity of losing candidates): House Clerk, Congressional Quarterly. District demographics: US Census Summary Data Files. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 17 / 44 5 10 Pct. female 15 20 25 Results - Descriptive Analysis 1990 1995 2000 Year Congress Elected All Republicans 2005 2010 Democrats Figure 1: Fraction of Female Congress Members Figure 1: Fraction of Female Congress Members Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 18 / 44 Results - Descriptive analysis Table 1: Summary Statistics Panel A: Sponsor Characteristics All Sponsors Democratic Sponsors Republican Sponsors All 4,778 Male 4,188 Female 590 All 2,505 Male 2,099 Female 406 All 2,273 Male 2,089 Female 184 12.8 (10.5) 12.5 (10.4) 14.5 (11.1) 13.2 (11.1) 12.9 (10.9) 14.8 (12.0) 12.4 (9.7) 12.2 (9.8) 14.0 (8.8) All 61,161 Male 52,577 Female 8,584 All 33,043 Male 27,027 Female 6,016 All 28,118 Male 25,550 Female 2,568 Number of cosponsors: median 17.0 (35.9) 3 16.5 (35.5) 3 19.9 (38.3) 5 17.6 (35.8) 4 17.0 (35.6) 3 19.9 (36.8) 6 16.3 (36.0) 3 15.9 (35.4) 3 19.8 (41.7) 4 Percent cosponsors opposite party: mean (std. dev.) 15.0 (21.3) 15.2 (21.4) 13.5 (20.6) 11.7 (18.1) 12.4 (18.6) 9.0 (15.32) 18.8 (24.0) 18.3 (23.6) 23.9 (26.8) Number of Sponsors Number of bills sponsored: mean (st. dev.) Panel B: Bill Characteristics Number of bills Number of cosponsors: mean (std. dev.) Notes: In Panel A the unit of observation is an individual-congress member, while in Panel B is a bill. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 19 / 44 14 Mean number of cosponsors 20 22 16 18 24 Results - Descriptive analysis 1990 1995 2000 Year Congress Elected Total Males 2005 2010 Females ite Party 0 35 Figure 2a: Mean Number of Cosponsors Gagliarducci and Paserman Figure 2a: Mean Number of Cosponsors Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 20 / 44 Males Results - Descriptive Analysis 15 Pct. of Cosponsors of Opposite Party 25 30 35 20 Figure 2a: Mean Number of Cosponsors 1990 1995 2000 Year Congress Elected Total Males 2005 2010 Females Figure 2b: Percent Cosponsors of the Opposite Party Figure 2b: Fraction Cosponsors of Opposite Party Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 21 / 44 OLS Analysis Table 2a: OLS Regressions (1) (2) (3) (4) (5) A: Number of Cosponsors All s.e. n. bills 3.308 *** (0.667) 61,334 2.103 *** (0.635) 61,334 2.737 *** (0.650) 60,670 1.450 ** (0.651) 61,334 2.004 *** (0.659) 60,670 Democrats s.e. n. bills 2.675 *** (0.836) 33,043 1.686 ** (0.805) 33,043 3.012 *** (0.772) 32,847 0.840 (0.841) 33,043 1.925 ** (0.789) 32,847 Republicans s.e. n. bills 4.082 *** (1.135) 28,118 3.174 *** (1.046) 28,118 3.791 *** (1.127) 27,671 2.623 *** (1.000) 28,118 3.162 *** (1.068) 27,671 Year effects Yes Yes Yes Yes Yes Bill category fixed effects No Yes Yes Yes Yes Sponsor charactersistics No No Yes No Yes District characteristics No No No Yes Yes Notes: Each entry in the table represents the coefficient on the female dummy from separate regressions. Robust standard errors, clustered at the individual-Congress level, in parentheses. The unit of observation is a bill. All estimates include Congress fixed effects. Bill characteristics include 20 dummies for the major topic. Sponsor characteristics include: age, tenure in Congress, a dummy for whether the sponsor is a rookie, a committee leader (chair or ranking member) or black, a party dummie, 5 occupational dummies, a dummy for whether the sponsor has an Ivy League college degree, a dummy for whether the sponsor was born in the state of election, and the total number of bills sponsored within the congress. District characteristics include: 3 macro area dummies, the percentage of black, over-65, foreign and urban residents, the logarithm of the median income, and the logarithm of the population density. Table 2a: OLS Analysis - Number of Cosponsors Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 22 / 44 OLS Analysis Table 2b: OLS Regressions (1) B: Number of Cosponsors of Opposite Party (2) (3) (4) (5) All s.e. n. bills -1.539 ** (0.626) 61,331 -1.894 *** (0.615) 61,331 0.047 (0.524) 60,667 -0.767 (0.577) 61,331 0.347 (0.522) 60,667 Democrats s.e. n. bills -2.996 *** (0.417) 33,042 -2.800 *** (0.411) 33,042 -2.460 *** (0.406) 32,846 -1.757 *** (0.411) 33,042 -1.681 *** (0.424) 32,846 Republicans s.e. n. bills 5.847 *** (1.177) 28,116 4.669 *** (1.053) 28,116 5.830 *** (1.017) 27,669 3.560 *** (0.941) 28,116 4.345 *** (0.972) 27,669 Year effects Yes Yes Yes Yes Yes Bill category fixed effects No Yes Yes Yes Yes Sponsor charactersistics No No Yes No Yes District characteristics No No No Yes Yes Notes: Each entry in the table represents the coefficient on the female dummy from separate regressions. Robust standard errors, clustered at the individual-Congress level, in parentheses. The unit of observation is a bill. All estimates include Congress fixed effects. Bill characteristics include 20 dummies for the major topic. Sponsor characteristics include: age, tenure in Congress, a dummy for whether the sponsor is a rookie, a committee leader (chair or ranking member) or black, a party dummie, 5 occupational dummies, a dummy for whether the sponsor has an Ivy League college degree, a dummy for whether the sponsor was born in the state of election, and the total number of bills sponsored within the congress. District characteristics include: 3 macro area dummies, the percentage of black, over-65, foreign and urban residents, the logarithm of the median income, and the logarithm of the population density. Table 2b: OLS Analysis - Percent Cosponsors of the Opposite Party Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 23 / 44 OLS Analysis Number of cosponsors: Women attract higher number of cosponsors. In all specifications, female coefficient attenuated with inclusion of district characteristics. Number of cosponsors opposite party: Democratic women less cooperative, Republicans more. Coefficients attenuated with inclusion of district characteristics. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 24 / 44 Methodology From OLS analysis, two main facts emerge: 1 2 Controls matter, especially district characteristics. Important differences by party. Districts that elect female representatives different in terms of observables. Perhaps also unobservables? Solution: RD Analysis? In mixed-gender races decided by narrow margin, gender of elected representative essentially randomized. But, in RD analysis separate by party, races on two sides of the threshold come from very different pools: Compare close races in which female R narrowly beats male D to races in which male R narrowly beats female D (and vice versa for D). Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 25 / 44 RD analysis - Distribution of Margin of Victory Kernel Density of the Running Variable 0 .005 .01 .015 .02 Party = All, bandwidth = 15 -100 -50 0 50 100 Kernel Density of the Running Variable Kernel Density of the Running Variable Party = Republicans, bandwidth = 15 .04 0 0 .01 .02 .03 .005 .01 .015 .02 Party = Democrats, bandwidth = 15 -100 -50 0 50 100 -100 -50 0 50 100 Figure 3: McCrary Density Tests Figure 3: Distribution of Margin of Victory: All, Democrats and Republicans Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 26 / 44 Within party, substantial discontinuity in density of the running variable at threshold. Within party, electoral races on two sides of threshold come from different pools, not necessarily similar in terms of observables or unobservables. RD is very specific form of matching estimator, where we try to match on a single value of a single variable. Calculate standardized difference in covariates (Imbens and Wooldridge, 2009), for full sample and RD sample: does RD achieve better balance? Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 27 / 44 Balancing Tests Table 3A: Balancing Tests (1) OLS - Full Sample ALL (2) (3) RD PS matching optimal bandwidth Full Sample (4) OLS - Full Sample Democrats (5) (6) RD - optimal PS matching bandwidth - Full Sample Republicans (7) (8) (9) OLS - Full RD - optimal PS matching Sample bandwidth - Full Sample A: District Characteristics Northeast -0.038 -0.003 0.035 -0.114 -0.354 * -0.028 0.051 Midwest -0.053 -0.218 -0.017 -0.022 0.242 0.003 -0.097 -0.655 * 0.061 South -0.204 -0.152 -0.020 -0.304 * -0.413 * 0.030 0.003 0.120 -0.033 0.005 0.418 * 0.525 * -0.022 -0.005 0.048 0.563 * -0.004 % Black 0.094 -0.077 -0.030 -0.172 0.017 -0.200 -0.271 * -0.120 % Urban 0.429 * 0.094 0.036 0.435 * 0.216 0.016 % Foreign Born 0.421 * 0.050 -0.002 0.435 * 0.164 -0.179 -0.209 0.013 -0.308 * 0.121 0.093 0.024 0.207 Log(Population Density) -0.048 0.025 -0.003 -0.066 Lagged Democratic Share 0.248 0.087 0.016 0.010 Campaign Expenditures Ratio (D/R 0.227 0.054 0.053 -0.033 0.000 -0.059 West 0.293 * % Over 65 Log(Median Income) Margin of victory 0.445 * -0.036 0.057 0.260 * -0.240 0.027 -0.026 0.241 -0.225 0.003 -0.260 * 0.037 0.013 -0.276 * -0.023 0.363 * 0.014 0.070 -0.205 -0.001 0.345 * 0.058 -0.091 0.025 -0.075 -0.228 -0.069 0.122 0.369 * 0.018 0.044 -0.001 0.047 0.019 0.113 -0.021 0.000 -0.054 -0.149 0.000 -0.040 0.089 Notes: Each coefficient represents the standardized difference between treated and control units. All specifications control for Congress fixed effects. The unit of observation is an individual-congress member. Leader is a dummy for being a committee chair or ranking member. Camp. Exp. D/R is the percentage of total campaign spending by the Democratic candidate. In column (4) the optimal bandwidth is 25. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 28 / 44 Balancing Tests Table 3A: Balancing Tests (1) OLS - Full Sample ALL (2) (3) RD PS matching optimal bandwidth Full Sample (4) OLS - Full Sample Democrats (5) (6) RD - optimal PS matching bandwidth - Full Sample Republicans (7) (8) (9) OLS - Full RD - optimal PS matching Sample bandwidth - Full Sample B: Member Characteristics Democrat 0.276 * 0.617 * 0.042 -0.304 * -0.283 * -0.307 * Rookie 0.110 0.181 0.089 0.155 Age 0.139 0.128 0.099 0.124 -0.173 Leader -0.204 -0.079 -0.261 * -0.197 -0.208 Born in State -0.363 * Ivy League College Tenure in Congress -0.367 * - - - - - -0.541 * -0.348 * -0.265 * 0.156 -0.330 * 0.400 * 0.133 0.053 -0.494 * 0.029 0.015 0.129 0.395 * -0.291 * -0.222 -0.082 -0.457 * 0.112 -0.234 -0.370 * -0.330 * -0.308 * -0.263 * -0.137 -0.497 * -0.124 -0.100 -0.157 -0.168 -0.195 -0.177 -0.094 No occupation 0.147 -0.050 0.164 0.126 -0.069 0.157 0.107 0.380 * Education 0.254 * 0.165 0.169 0.225 0.115 0.165 0.281 * 0.301 * 0.101 Lawyer -0.436 * -0.176 -0.430 * -0.455 * -0.191 -0.458 * -0.518 * -0.380 * -0.502 * Professional -0.110 -0.065 -0.135 -0.048 -0.100 -0.042 -0.267 * -0.251 * -0.275 * Business -0.175 0.178 -0.213 -0.007 0.636 * -0.097 -0.299 * 0.127 -0.313 * 0.532 * -0.252 * Other 0.283 * -0.079 0.380 * 0.170 -0.387 * 0.264 * Black 0.209 -0.009 0.059 0.189 -0.078 0.065 -0.058 0.063 0.000 -0.550 * -0.132 0.104 0.676 * -0.063 Notes: Each coefficient represents the standardized difference between treated and control units. All specifications control for Congress fixed effects. The unit of observation is an individual-congress member. Leader is a dummy for being a committee chair or ranking member. Camp. Exp. D/R is the percentage of total campaign spending by the Democratic candidate. In column (4) the optimal bandwidth is 25. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 29 / 44 Solutions RD augmented with inverse propensity score weighting to account for differences in covariates on two sides of the threshold. Inverse propensity score weighting on full sample (including non mixed-gender races), controlling also for margin of victory of winning candidate. Propensity score can be based on district characteristics only, or on both district and individual characteristics. Different interpretation of the results. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 30 / 44 Results - Coalition Building Table 4: Gender and the Number of Cosponsors (1) OLS - Full Sample (2) RD - optimal bandwidth (3) RD - 3rd order polynomial (4) (5) (6) RD - optimal PS matching - Full PS matching - Full bandwidth with PS Sample Sample matching All s.e. n. bills n. individuals optimal bandwidth 2.004 *** 0.659 60,670 4,746 2.308 2.597 4,871 403 25 0.575 3.290 11,749 883 1.173 2.434 4,871 403 25 Democrats s.e. n. bills n. individuals optimal bandwidth 1.925 ** 0.789 32,847 2,492 2.182 3.609 2,343 193 30 1.065 4.684 6,141 441 -1.201 3.441 2,343 193 30 Republicans s.e. n. bills n. individuals optimal bandwidth 3.162 *** 1.068 27,671 2,244 6.101 4.796 1,227 100 13 0.709 4.442 5,608 442 4.768 4.203 1,227 100 13 1.290 * 0.696 55,672 4,403 0.972 0.758 55,008 4,358 1.809 1.116 29,560 2,278 2.090 * 1.150 29,364 2,265 1.825 ** 0.898 26,089 2,121 3.116 *** 1.171 23,818 1,953 Sponsor characteristics Yes No No No No No Bill charactertistics Yes No No Yes Yes Yes District characteristics Yes No No No No No Propensity Score Distr. Distr.+MV Distr.+MV+Spon. Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the individual-Congress level, in parentheses. The unit of observation is a bill. All estimates include Congress fixed effects. Bill characteristics include 20 dummies for the major topic. Sponsor characteristics include: age, tenure in Congress, a dummy for whether the sponsor is a rookie, a committee leader (chair or ranking member) or black, a party dummie, 5 occupational dummies, a dummy for whether the sponsor has an Ivy League college degree, a dummy for whether the sponsor was born in the state of election, and the total number of bills sponsored within the congress. District characteristics include: 3 macro area dummies, the percentage of black, over-65, foreign and urban residents, the logarithm of the median Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 31 / 44 Results - Bipartisanship Table 5: Gender and the Percent Cosponsors of the Opposite Party (1) OLS - Full Sample All s.e. n. bills n. ind. optimal bandwidth 0.347 0.522 60,667 4,746 Democrats s.e. n. bills n. ind. optimal bandwidth Republicans s.e. n. bills n. ind. optimal bandwidth (2) RD - optimal bandwidth (3) RD - 3rd order polynomial 0.822 3.876 2,781 232 16 0.222 3.445 11,748 883 -1.681 *** 0.424 32,846 2,492 -3.351 * 1.910 1,978 167 24 -0.907 1.907 6,140 441 4.345 *** 0.972 27,669 2,244 12.514 * 6.673 1,043 88 11 5.819 7.133 5,608 442 (4) (5) (6) RD - optimal PS matching - Full PS matching - Full bandwidth with PS Sample Sample matching 0.332 3.540 2,781 232 16 -6.885 *** 2.115 1,978 167 24 7.533 5.769 1,043 88 11 1.181 0.789 55,670 4,403 0.730 1.010 55,006 4,358 -0.960 0.634 29,559 2,278 -1.332 ** 0.550 29,363 2,265 3.374 *** 0.931 26,088 2,121 3.400 *** 1.072 23,817 1,953 Sponsor characteristics Yes No No No No No Bill charactertistics Yes No No Yes Yes Yes District characteristics Yes No No No No No Propensity Score Distr. Distr.+MV Distr.+MV+Spon. Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the individual-Congress level, in parentheses. The unit of observation is a bill. All estimates include Congress fixed effects. Bill characteristics include 20 dummies for the major topic. Sponsor characteristics include: age, tenure in Congress, a dummy for whether the sponsor is a rookie, a committee leader (chair or ranking member) or black, a party dummie, 5 occupational dummies, a dummy for whether the sponsor has an Ivy League college degree, a dummy for whether the sponsor was born in the state of election, and the total number of bills sponsored within the congress. District characteristics include: 3 macro area dummies, the percentage of black, over-65, foreign and urban residents, the logarithm of the median Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 32 / 44 Basic Results - Summary Women of both parties recruit more cosponsors: Strong in OLS (+5-20%, w.r.t. 16). Not precise in RD or matching. On average, no evidence that women are more bipartisan. But large differences by party: OLS: Female Republicans recruit more cosponsors (+25%, w.r.t. 16), and are more bipartisan (+30%, w.r.t. 5) RD + matching: similar qualitatively to OLS. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 33 / 44 Additional Analysis Do women recruit more female cosponsors? Any differences on female-relevant issues? Definition 1: Major Topic: Health, Labor/Employment/Immigration, Education, Law/Crime/Family, Social Welfare. Definition 2: Rank minor topics by fraction of female sponsors, take top 25%. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 34 / 44 Results - Coalition Building with other Women Table 6: Gender and the Percent Female Cosponsors and Female Cosponsors of the Opposite Party (1) OLS - Full Sample (2) RD - optimal bandwidth (3) RD - 3rd order polynomial (4) (5) (6) RD - optimal PS matching - Full PS matching - Full bandwidth with PS Sample Sample matching A: Percent Female Cosponsors All s.e. n. bills n. ind. optimal bandwidth 2.772 *** 0.303 60,667 4,745 2.926 ** 1.454 3,311 271 18 Democrats s.e. n. bills n. ind. optimal bandwidth 2.977 *** 0.402 32,845 2,492 Republicans s.e. n. bills n. ind. optimal bandwidth 1.485 *** 0.406 27,670 2,243 2.053 1.436 11,748 883 1.899 1.328 3,311 271 18 2.414 *** 0.306 55,670 4,402 2.880 *** 0.383 55,006 4,357 2.186 1.615 1,795 152 22 3.208 1.969 6,140 441 2.190 1.768 1,795 152 22 2.460 *** 0.396 29,559 2,278 3.033 *** 0.469 29,363 2,265 2.355 2.576 1,157 95 12 -0.189 2.741 5,608 442 1.302 2.201 1,157 95 12 1.621 *** 0.421 26,088 2,120 1.902 *** 0.480 23,817 1,952 Sponsor characteristics Yes No No No No No Bill charactertistics Yes No No Yes Yes Yes District characteristics Yes No No No No No Propensity Score Distr. Distr.+MV Distr.+MV+Spon. Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the individual-Congress level, in parentheses. The unit of observation is a bill. All estimates include Congress fixed effects. Bill characteristics include 20 dummies for the major topic. Sponsor characteristics include: age, tenure in Congress, a dummy for whether the sponsor is a rookie, a committee leader (chair or ranking member) or black, a party dummie, 5 occupational dummies, a dummy for whether the sponsor has an Ivy League college degree, a dummy for whether sponsor was born Gender in the state of election, the total number of bills sponsored within congress. District35 / 44 Gagliarducci and the Paserman Differences in and Cooperative Environments? Evidence fromthe the April U.S. 2015 Congress Results - Coalition Building with other Women Table 6: Gender and the Percent Female Cosponsors and Female Cosponsors of the Opposite Party (1) OLS - Full Sample (2) RD - optimal bandwidth (3) RD - 3rd order polynomial (4) (5) (6) RD - optimal PS matching - Full PS matching - Full bandwidth with PS Sample Sample matching B: Percent Female Cosponsors of the Opposite Party All s.e. n. bills n. ind. optimal bandwidth 0.408 *** 0.142 60,669 4,746 -0.398 0.978 3,144 258 17 -1.025 0.974 11,748 883 -1.084 1.074 3,144 258 17 0.799 *** 0.198 55,671 4,403 0.650 *** 0.202 55,007 4,358 Democrats s.e. n. bills n. ind. optimal bandwidth 0.096 0.107 32,846 2,492 -0.410 0.535 1,927 163 23 -0.462 0.580 6,140 441 -0.463 0.622 1,927 163 23 0.340 *** 0.119 29,559 2,278 0.287 ** 0.137 29,363 2,265 Republicans s.e. n. bills n. ind. optimal bandwidth 1.367 *** 0.328 27,671 2,244 1.740 2.128 1,140 94 12 0.290 2.144 5,608 442 0.639 2.108 1,140 94 12 1.323 *** 0.321 26,089 2,121 1.414 *** 0.366 23,818 1,953 Sponsor characteristics Yes No No No No No Bill charactertistics Yes No No Yes Yes Yes District characteristics Yes No No No No No Propensity Score Distr. Distr.+MV Distr.+MV+Spon. Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the individual-Congress level, in parentheses. The unit of observation is a bill. All estimates include Congress fixed effects. Bill characteristics include 20 dummies for the major topic. Sponsor characteristics include: age, tenure in Congress, a dummy for whether the sponsor is a rookie, a committee leader (chair or ranking member) or black, a party dummie, 5 occupational dummies, a dummy for whether the sponsor has an Ivy League college degree, a dummy for whether the sponsor was born in the state of election, and the total number of bills sponsored within the congress. District characteristics include: 3 macro area dummies, the percentage of black, over-65, foreign and urban residents, the logarithm of the median Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 36 / 44 Results - Coalition Building with other Women Women attract a higher number of female cosponsors (+25%, w.r.t. 2.6). Both R and D women more likely to attract female cosponsors of the opposite party (+60%, w.r.t. 0.7), but not estimated precisely. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 37 / 44 Results - Coalition Building on Female Issues Table 7a: The Number of Cosponsors on Women's Issues (1) OLS - Full Sample (2) RD - optimal bandwidth (3) RD - 3rd order polynomial (4) (5) (6) RD - optimal PS matching - Full PS matching - Full bandwidth with PS Sample Sample matching A: Number of Cosponsors All s.e. n. bills n. ind. optimal bandwidth 3.921 *** 1.021 20,072 1,504 9.013 6.260 1,058 97 18 4.945 6.107 4,240 328 4.046 5.366 1,058 97 18 Democrats s.e. n. bills n. ind. optimal bandwidth 4.223 *** 1.162 11,721 836 -0.705 8.821 613 61 21 -0.104 8.839 2,524 189 Republicans s.e. n. bills n. ind. optimal bandwidth 4.365 ** 1.912 8,292 665 14.172 9.161 508 44 16 11.176 8.900 1,716 139 1.955 * 1.055 18,136 1,383 3.567 *** 1.096 17,973 1,367 -2.458 8.338 613 61 21 1.395 1.343 10,316 749 4.475 *** 1.603 10,277 745 11.981 7.815 508 44 16 1.714 1.696 7,811 631 3.631 * 2.045 6,980 572 Sponsor characteristics Yes No No No No No Bill charactertistics Yes No No Yes Yes Yes District characteristics Yes No No No No No Propensity Score Distr. Distr.+MV Distr.+MV+Spon. Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the individual-Congress level, in parentheses. The unit of observation is a bill. All estimates include Congress fixed effects. Bill characteristics include 20 dummies for the major topic. Sponsor characteristics include: age, tenure in Congress, a dummy for whether the sponsor is a rookie, a committee leader (chair or ranking member) or black, a party dummie, 5 occupational dummies, a dummy for whether the sponsor has an Ivy League college degree, a dummy for whether the sponsor was born in the state of election, and the total number of bills sponsored within the congress. District characteristics include: macro area dummies, the percentage of black, over-65, foreign andEnvironments? urban residents, theEvidence logarithm of the median Gagliarducci and 3Paserman Gender Differences in Cooperative from the AprilU.S. 2015 Congress 38 / 44 Results - Bipartisanship on Female Issues Table 7b: Percent Cosponsors Opposite Party on Women's Issues (1) OLS - Full Sample (2) RD - optimal bandwidth (3) RD - 3rd order polynomial (4) (5) (6) RD - optimal PS matching - Full PS matching - Full bandwidth with PS Sample Sample matching B: Percent Cosponsors of the Opposite Party All s.e. n. bills n. ind. optimal bandwidth 2.012 *** 0.721 20,070 1,504 2.964 4.630 1,710 156 27 5.089 6.246 4,240 328 Democrats s.e. n. bills n. ind. optimal bandwidth -0.967 ** 0.485 11,721 836 -2.014 2.623 710 71 25 0.069 2.796 2,524 189 Republicans s.e. n. bills n. ind. optimal bandwidth 7.750 *** 1.600 8,290 665 24.394 ** 10.462 383 36 12 24.395 ** 11.033 1,716 139 1.317 4.582 1,710 156 27 -5.309 ** 2.622 710 71 25 16.527 * 9.574 383 36 12 4.066 *** 1.509 18,135 1,383 3.708 * 1.995 17,972 1,367 -0.268 0.743 10,316 749 -0.891 0.582 10,277 745 7.350 *** 1.245 7,810 631 7.059 *** 1.570 6,979 572 Sponsor characteristics Yes No No No No No Bill charactertistics Yes No No Yes Yes Yes District characteristics Yes No No No No No Propensity Score Distr. Distr.+MV Distr.+MV+Spon. Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the individual-Congress level, in parentheses. The unit of observation is a bill. All estimates include Congress fixed effects. Bill characteristics include 20 dummies for the major topic. Sponsor characteristics include: age, tenure in Congress, a dummy for whether the sponsor is a rookie, a committee leader (chair or ranking member) or black, a party dummie, 5 occupational dummies, a dummy for whether the sponsor has an Ivy League college degree, a dummy for whether sponsor was born in the state of election,inand the total number of bills sponsored withinfrom the congress. District Gagliarducci and the Paserman Gender Differences Cooperative Environments? Evidence the AprilU.S. 2015 Congress 39 / 44 Female Issues On female-issues, R women recruit more cosponsors (+20-80%, w.r.t. 24), more of the opposite party (+100%, w.r.t. 6), more female (+50%, w.r.t. 3.8), and more female of the opposite party (+200%, w.r.t. 1). Democrats: some evidence of more cooperation with other women on female-relevant issues, but not as robust. Large effect on R women possibly explained by absence of women within their party, or R men reluctant to cooperate on women issues. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 40 / 44 Results - Bill Outcomes Table 10: Gender and Bill Outcomes (1) OLS - Full Sample (2) RD - optimal bandwidth (3) RD - 3rd order polynomial (4) (5) (6) RD - optimal PS matching - Full PS matching - Full Sample Sample bandwidth with PS matching A: Passed the House All s.e. n. bills n. ind. optimal bandwidth -0.014 ** 0.007 60,670 4,746 0.023 0.025 3,457 282 19 -0.018 0.028 11,749 883 0.007 0.027 3,457 282 19 -0.032 *** 0.007 55,672 4,403 -0.035 *** 0.007 55,008 4,358 Democrats s.e. n. bills n. ind. optimal bandwidth -0.013 * 0.008 32,847 2,492 0.014 0.039 1,393 117 18 -0.003 0.037 6,141 441 0.066 * 0.039 1,393 117 18 -0.035 *** 0.007 29,560 2,278 -0.032 *** 0.008 29,364 2,265 Republicans s.e. n. bills n. ind. optimal bandwidth 0.002 0.012 27,671 2,244 0.038 0.055 1,353 108 14 -0.016 0.041 5,608 442 0.026 0.057 1,353 108 14 -0.012 0.012 26,089 2,121 -0.001 0.013 23,818 1,953 Sponsor characteristics Yes No No No No No Bill charactertistics Yes No No Yes Yes Yes District characteristics Yes No No No No No Propensity Score Distr. Distr.+MV Distr.+MV+Spon. Notes: Entries in the table represent the coefficient on the female dummy. Robust standard errors, clustered at the individual-Congress level, in parentheses. The unit of observation is a bill. All estimates include Congress fixed effects. Bill characteristics include 20 dummies for the major topic. Sponsor characteristics include: age, tenure in Congress, a dummy for whether the sponsor is a rookie, a committee leader (chair or ranking member) or black, a party dummie, 5 occupational dummies, a dummy for whether the sponsor has an Ivy League college degree, a dummy for whether the sponsor was born in the state of election, and the total number of bills sponsored within the congress. District Gagliarducciinclude: and Paserman Differences Cooperative Environments? Evidence April U.S. 2015 Congress 41 / 44 characteristics 3 macro area dummies,Gender the percentage of black,inover-65, foreign and urban residents, the logarithmfrom of thethe median Bill outcomes No effect of gender on bill outcomes, if not negative. The advantage for women in recruiting cosponsors not enough for a bill to move further in the legislative process. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 42 / 44 Conclusion We have studied cooperative behavior in a real-world setting with high stakes. Findings: After controlling for the characteristics of the district in which they are elected, not much evidence that women are inherently more cooperative/bipartisan. However, fairly robust evidence that female Republicans build larger coalitions and attract more cosponsors from the opposite party, especially on female issues. ⇒ Cooperation driven by common interest, rather than gender per se. Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 43 / 44 Thank you! Gagliarducci and Paserman Gender Differences in Cooperative Environments? Evidence from the AprilU.S. 2015 Congress 44 / 44
© Copyright 2026 Paperzz