Gender Stereotypes and the Policy Priorities of Women in Congress

Transcription

1 Gender Stereotypes and the Policy Priorities of Women in Congress May 25, 2016 Abstract We argue that the need to overcome gender stereotypes on the campaign trail leads female legislators to develop diverse legislative portfolios that can be used to demonstrate competence on salient, masculine issues. We analyze a comprehensive database of all bills introduced in the U.S. House between 1963 and 2010 and find that female MCs propose more bills, spread across more issues, than do men. Further, electorally vulnerable women focus more on defense-related bills than do their male colleagues. Our findings speak to the link between legislative agendas and campaign agendas, and the policy implications of increased female representation. Keywords: Women s Representation Legislative Agendas Descriptive Representation

2 Introduction Many factors shape the policy agendas legislators pursue in office, including the interests of constituents, electoral coalitions and organized interests within their districts, and the partisanship, ideology, and interests of the legislators themselves. Here, we consider an additional factor the need to overcome gender stereotypes with demonstrations of expertise. In this paper, we bridge two existing branches of literature on women in politics to provide a fuller picture of the impact of gender on policymaking. The first finds that female legislators are more active than their male colleagues on issues related to women s rights and women s health (e.g. Dodson 2006; Gerrity, Osborn and Mendez 2007; MacDonald and OBrien 2011; Osborn and Mendez 2010; Swers 1998, 2002a, 2002b). The second shows that the public views women as less capable of handling stereotypically masculine issues, like defense, foreign affairs, crime, and macroeconomics topics voters prioritize when evaluating candidates (e.g., Huddy and Terkildsen 1993; Kahn 1996; Druckman and Ostermeir 2004). Together, these two observations suggest women should be seriously disadvantaged at the polls, especially in years when defense, crime, and the economy are highly salient. And yet, when women run for public office, they are just as likely to be elected as men are. We seek to explain why, and contend that women are successful in part because they craft large, diverse legislative portfolios that include women s issues and masculine issues. This allows female lawmakers to demonstrate competence in dealing with high salience topics like war and peace, while simultaneously offering substantive representation to women nationwide. On the campaign trail, female lawmakers can highlight their legislative accomplishments to quell concerns stemming from gender stereotypes. 1

3 As a first step in evaluating the theory we develop, we analyze a comprehensive database of all bills introduced in the U.S. House between 1963 and Unlike prior studies that focus primarily on the extent to which female lawmakers prioritize women s issues, we examine three key dependent variables: the number of bills introduced by MCs, the degree of concentration in lawmakers agendas, and the propensity of legislators to sponsor bills in 19 different topic areas that cover the full range of issues considered by Congress. These are all, respectively, modeled as a function of legislator characteristics (including sex), electoral safety, institutional position, and district level factors. This makes the analysis the most thorough examination to date of sex differences in sponsorship behavior. We find that female MCs propose more legislation than do men and that the legislation they introduce is distributed across more policy topics meaning women are less likely to craft specialized policy agendas than are men. Instead, they distribute their attention across a range of topics that includes civil rights, health care, social welfare, international affairs, and defense. We further find that electorally vulnerable women, who were recently reelected by a slim margin, are the most active sponsors of defense-related bills. This finding, in particular, supports our theory showing that female lawmakers ramp up their attention to this salient, masculine topic in response to electoral competition. Our findings help to explain the success of female politicians at the ballot box while also speaking to the policy implications of increased female representation. Legislatures across the globe have seen the number of female lawmakers increase in recent decades. In many countries this growth has been aided by the implementation of gender quotas. The idea behind such policies is that increased descriptive representation for women will result in increased substantive representation for women. We find support for this assertion in the context of the U.S. Congress. But our findings also 2

4 shed light on one of the many additional ways that increased female representation matters. We find that women are adding their voices to the debate over topics long considered the province of men. We further discuss the implications of this finding in the paper s concluding section. Combating Stereotypes with Expertise There is an important and self-reinforcing relationship between legislative agendas and campaign agendas. Members of Congress introduce bills as a way of demonstrating their commitment to and expertise on various issues. Doing so allows members to develop and maintain relationships with interest groups, attentive publics, and reelection constituencies. By introducing bills on the topics that are salient to these groups, MCs can claim credit for being active on these topics (Mayhew 1974a). Not surprisingly then, Sellers (1998) finds that senators run campaigns that are centered on their legislative agendas. Reciprocally, legislators are active sponsors of legislation on the issues they campaign on (Sulkin 2009). Members of Congress also integrate the campaign issues raised by challengers into their legislative portfolios (Sulkin 2009). They hope that by doing so, they will inoculate themselves against similar challenges in the future. This ability to claim credit and demonstrate expertise may be particularly important for female lawmakers who must combat gender stereotypes on the campaign trail (Windett 2014). Due to such stereotypes, Hernson, Lay, and Stokes (2003) argue that male and female candidates are respectively able to own divergent sets of issues. Women are seen as credible on topics related to care and compassion, while men own those related to toughness, leadership, and aggression. When women s issues are salient, as they were in 1992 (the year of Justice Thomas confirmation 3

5 hearing), female candidates can succeed at the polls by running on topics where they are advantaged, like poverty, sexual harassment, and health care (Palmer and Simon 2005). But when masculine issues dominate the national agenda, female candidates must formulate strategies to weaken the stereotypes and establish perceptions of issue competency on these traditional male issues (Palmer and Simon 2005, 45). One such strategy may be to develop a legislative portfolio that demonstrates expertise in a masculine policy area. Doing so would allow women to establish their credibility on an issue that their gender does not typically allow them to own. For instance, Michelle Swers (2007) shows that after the 9/11 attacks female senators used bill sponsorship to strengthen their reputations on national defense policy. On average, Democratic women sponsored more homeland security related legislation in the 108th Congress than did co-partisan men or Republicans of either sex. Republican women were more active sponsors of soft defense bills (those extending benefits to veterans and military personnel) in both the 107th and the 108th Congress. Swers (2007) concludes that stereotypes about women s policy expertise creates an additional hurdle for women senators who seek to gain credibility on defense issues. The women recognize this vulnerability and devote extra effort to building their reputations [on these issues] (581). Swers (2007b) findings suggest that women in Congress seek to balance their legislative portfolios by attending both to stereotypically feminine and masculine policy topics. These findings are echoed by Osborn and Mendez (2010), who find that female Senators are more likely than men to address women s issues, and defense and foreign affairs in their floor speeches. These authors also find that women are no less likely than men to address a wide range of non-women s issues. Whether women in the House of Representatives use the type of balancing strategy identified by Osborn and Mendez (2010) has not been explored in the literature. 4

6 Further, the degree to which women have employed this strategy before and after the 9/11 attacks is unknown. However, studies find that voters have long preferred candidates with competence on masculine topics and candidates who possess masculine character traits (e.g., Huddy and Terkildsen 1993; Kahn 1996; Druckman and Ostermeir 2004). This means that even before 9/11, female members of Congress had an incentive to craft policy portfolios that included introductions on masculine issues. This would allow women to demonstrate expertise in these areas on the campaign trial a strategy that comports with studies finding that female candidates prefer to run on their credentials (as opposed to their personal characteristics) (Fox 1997). Hypotheses We expect to find that women in the House balance their legislative agendas by incorporating some masculine issues into their legislative portfolios. This should allow female legislators to demonstrate expertise in these areas on the campaign trail and should help them ward off potential challengers. This may be especially true for electorally vulnerable women, who were recently reelected by a small margin. These women will have the greatest incentive to shore up their credentials by focusing on salient, masculine issues before the next election cycle. Because the gender stereotypes this balancing strategy is designed to combat predate the 9/11 attacks, we expect to find it in use throughout the time period studied. The use of this strategy should also result in female legislative portfolios that are larger and more diverse than are those of men. This is because female MCs are known to be more active sponsors of legislation related to women s issues. If this heightened activity on issues like workplace equality and women s health is coupled with heightened attention to issues like defense and international affairs, it means 5

7 the agendas women pursue in office are likely larger and broader than are those of their male colleagues. Safe women, in particular, may sponsor a large number of bills across many topic areas, as prior research finds that legislators who are electorally secure sponsor more legislation than do other members of Congress (Garand and Burke 2006). This is perhaps because safe members are able to focus on crafting good public policy, whereas vulnerable members must devote more time to campaign activities like fundraisers and events in their districts. Further, vulnerable members will want to focus their energies narrowly on issues that will benefit them at the polls. We, therefore, expect safe women to have the largest, most diverse legislative portfolios, followed by marginal women, and then men. Data and Methods To test our theoretical expectations, we rely on data from the Congressional Bills Project (Alder and Wilkerson ). The dataset includes a record of every bill introduced during our period of study (1963 to 2010 or the 88th through the 110th Congress). For each bill, the dataset includes a number of variables that are useful for our purposes, including the name, sex, and Congressional district of the sponsor. It also includes the committee(s) to which the bill was referred, whether the sponsor served on that committee, and whether the sponsor served as the chair or ranking member of the committee. Finally, the Congressional Bills Project (CBP) data also includes the primary topic area of each bill. These topic codes correspond to the coding scheme developed by the Policy Agendas Project and include 19 major topic categories. These broad categories include topics such as health, education, defense, transportation, and agriculture. 1 Some of these categories clearly align with topics considered traditional women s 6

8 issues that is, topics related to care and compassion. Those categories are health, education, social welfare, and community development and housing. 2 Additionally, the topic of civil rights and liberties contains legislation designed to combat inequality against women and minorities, as well as legislation related to reproductive rights. Finally, the topics of defense, foreign affairs, and macroeconomics clearly encompass stereotypically masculine issue areas. Therefore, while the topic codes are broad, they allow us to examine the sponsorship of feminine and masculine issue bills. The coding scheme also identifies a number of gender-neutral categories, such as transportation, government operations, and public land and water management. 3 The CBP data do not contain information about district characteristics or margin of electoral victory. 4 For that information, we relied on data from the Census Bureau and Polidata Presidential Election Results. 5 These three datasets were merged together based on the congressional district number. After merging the datasets, we construct aggregate measures of the legislative agendas of each Representative per session of Congress. Our dependent variables are, therefore, indicators of sponsorship behavior for each legislator ij, where i is the individual member and j indicates the session of Congress. To test our hypotheses, we construct three dependent variables with these data: a count of the number of bills introduced by each MC in each session, a measure of the degree of concentration in each member s agenda during each session, and the propensity of each member to sponsor legislation in each of the 19 different topic areas. The construction of the first dependent variable is self-explanatory as this is simply a count of the number of bills sponsored. Table 1 displays the descriptive statistics for this variable, first overall and then separated by the sex of the legislator. Notice that, overall, the minimum number of bills sponsored by both men and women in a single session is zero and the maximum is 432. The overall mean is approximately 7

9 22.4. Men, on average, have a higher level of bill sponsorship than do women with an average of about 23 and a standard deviation of 27. Women meanwhile, have a lower average with 18.4 per session, although the standard deviation is much lower with a value of [Table 1 about here.] Assessing the number of issue areas with sponsorship could provide a simple measure of the diversity of each Representative s legislative agenda. However, this type of measure would only indicate the breadth of a Representative s agenda, not the depth of it. To thoroughly test our theory, we need an indicator that can measure both breadth and depth because we believe women may not be able to establish issue competency by sponsoring a single bill in an area like national defense while focusing primarily on a topic her sex allows her to own. To better capture the diversity of each Representative s portfolio, we construct a Herfindahl-Hirschman Index (HHI) for each legislator, by Congress. 6 Herfindahl-Hirschman Indices are utilized by economist to test market concentration and competition. When used for this purpose, low scores indicate a high degree of competition within a market and high scores indicate the presence of a monopoly. The measure has also been used in political science to consider the level of competition in elections (Stigler 1972), the effective number of parties in multiparty systems (Laakso and Taagepera 1979), and the diversity of interest organization populations (Gray and Lowery 2000) (Boydstun, Bevan, and Thomas III 2014, 181). Here, we use the measure to determine whether a single issue monopolizes the agenda of a given legislator, or whether the legislator s attention is distributed equally across many topics. The equation below shows how the HHI is calculated. The measure is created by squaring the percentage of the agenda comprised of each issue and then summing 8

10 those squared values. 7 The index ranges from zero indicating diffusion of attention across all topics to 10,000 indicating a concentration of attention on a single topic. In the next section, we use Time Series Ordinary Least Squares Regression with the HHI as the dependent variable to estimate agenda concentration. HHI = N i=1 S 2 i (1) Our final dependent variable moves beyond the count and concentration of agenda diversity. We seek to examine the types of issues Representatives will focus on, or move away from, as their electoral security changes. To do so, we calculate the proportion of each member s legislative agenda that is devoted to each of the 19 issue areas. We then use this dependent variable to test the effects of electoral security and gender on issue prioritization. Because legislators can offer bills pertaining to a number of different issue areas during the same session of Congress, we must control for the potential of simultaneous and un-modeled correlation in the behavior of the individual members. We therefore utilize a time series seemingly unrelated regression technique (SUR). This technique allows for both the estimation of multiple equations and for the error terms of each equation to be contemporaneously correlated with one another. 8 Constructing Independent Variables In all models, our primary explanatory variable is the sex of the legislator. This variable is provided in the CBP data. The electoral security of individual members of 9

11 Congress is also a key explanatory variable. Utilizing data on Congressional election returns from 1962 to 2008, we construct a categorical variable to indicate the electoral security of each member. We use a categorical rather than continuous variable because the lack of competition in House elections over this time frame makes the use of the margin of victory an imperfect measure of electoral vulnerability. 9 This is especially problematic in the single party south, but the incumbency advantage makes the raw count difficult to implement in statistical analysis for House Members nationwide. We, therefore, simplify electoral competition into categories of marginal or safe electoral seats. Borrowing from Mayhew (1974b), we classify marginal seats as those with levels of competition under 20 percentage points, and safe districts as those with electoral spreads above 20 percentage points. 10 Table 2 reports the summary statistics for our key independent variables, broken down by sex. We also include a lagged dependent variable in each of our regressions. 11 [Table 2 about here.] The models also include variables measuring a number of other factors known to influence the number and type of bills sponsored by Representatives. Party dynamics, institutional position, committee assignments, and tenure in office all affect legislator behavior. Party leaders, committee chairs, and senior members of Congress typically have more resources at their disposal than do rank-and-file members. These resources include additional staff, greater control over the agenda, and greater access to specialized information. Such members are, therefore, more likely to sponsor legislation (Schiller 1995), more likely to see the bills they sponsor debated in committee (Krutz 2005), and more likely to participate in subcommittee markup (Hall 1996). Freshman legislators and members of the minority party, on the other hand, are less likely to sponsor legislation than are other members (e.g. Garand and Burke 2006; Schiller 10

12 1995). For these reasons, we control for majority party status, the number of sessions the member has served in Congress, status as a committee chair, membership in the House leadership, 12 and whether or not the member began their career by replacing a deceased spouse. We also include a variable that indicates whether the individual is a member of a racial minority group and the member s partisanship. District factors also shape the priorities legislators pursue. For instance, legislators from urban areas are unlikely to focus on farm subsidies just as legislators from rural areas are unlikely to focus on urban renewal. Instead, members of Congress craft legislative portfolios that will appeal to their constituents. We, therefore, include in the model information about the districts each member serves. We collected information from the Census Bureau that identifies urban districts, gives the percentage of the district that is African American, and lists the median income for the district in 2012 dollars. 13 We also include the two-party Democratic vote share in the district for the previous presidential election. Because these indicators are all highly correlated with one another, we use principal components factor analysis to construct district-specific factor scores for these demographic and political characteristics. We include this factor score in each of our models. Finally, we include Congress dummy variables to account for fixed effects and within Congress specific variation. Thus, our models test the effects of our predictor variables within sessions of Congress. This is an important choice because there are many differences between sessions that might affect sponsorship decisions. For example, reforms to House rules in the early 1970s were aimed at curbing the influence of seniority in the exercise of legislative power. These reforms may have allowed members to openly sponsor more legislation without fear of retribution at the committee level or from party leadership. Changes at this time also allowed for the co-sponsorship of bills. Institutional changes like these shape legislator behavior 11

13 by altering the rules that govern the body. Other changes from session to session, like changes in the ideological make up of the body, levels of polarization, and the number of women in office, might also influence the sponsorship behavior of MCs. Our modeling strategy accounts for these differences because all between-congress variance is explained by the fixed effects. 14 Results We begin our analysis with an examination of the differences in the overall sponsorship behavior across both sex and electoral security. Table 3 reports the coefficient estimates and standard errors of our first Time Series Negative Bi-Nomial Model predicting the total number of bills sponsored by each member. Overall, the results in Table 3 comport with our expectations that women sponsor more legislation than men. We also find mixed support for our expectation that electoral security effects the behavior of men and women differently. Both safe and marginal women sponsor more legislation than do safe men, the omitted category. Committee chairs also introduce significantly more legislation than do others. The substantive significance of coefficients from negative binomial models is difficult to interpret. For this reason, we plot the predicted level of bills sponsored with 95% confidence intervals. As evident in Figure 1, we have mixed support for our theoretical claims. Safe women have a statistically significantly higher level of bill sponsorship than both categories of male MCs, with over 26 bills sponsored per session. Likewise, marginal female members sponsor a greater number of bills than both categories of male legislators with roughly 25 bills per session. As expected, both marginal and safe men have the lowest level of bill sponsorship with both groups sponsoring roughly 21 bills in a given session. We see no statistical distinction, how- 12

14 ever, within these groups marginal and safe members of the same gender sponsor statistically indistinguishable numbers of bills. [Table 3 about here.] [Figure 1 about here.] Moving beyond the simple count of bills sponsored, we now turn our attention to the analysis of the Herfindahl-Hirschman Indices. As we have previously outlined, a lower value for the HHI indicates a more diverse and evenly spread out legislative agenda, while higher values indicate a concentration of attention on a smaller set of issues. Figure 2 shows the trends of the HHI s over time for safe and vulnerable men and women (respectively). The simple visual provides evidence that marginal women generally have the most diverse agendas, although there is a good deal of variation over time. Moreover, safe women have the second most diverse agendas over time. This evidence lends support to our argument that marginal women and safe women will craft agendas that are more diverse than are those of men. [Figure 2 about here.] We next use the HHI as our dependent variable in a Time Series OLS model. The coefficients and Z-scores are reported in Table 4. The results show that safe women have the most diverse legislative agendas and that when compared to safe men, this difference is statistically significant. The coefficient for marginal women is also large and negative, however, it is not statistically significant. In Figure 3, we plot the estimated HHI s of each group. Here again we find that safe women have the most diverse legislative agendas to a degree that is statistically significant when compared to both safe and marginal males. Marginal women again 13

15 have the second most diverse agendas, however, the large confidence bounds make this estimate indistinguishable from those of other groups. [Table 4 about here.] [Figure 3 about here.] Where do women focus their attention? While we have focused our previous analyses on the number of bills sponsored and the legislator s agenda diversity, we now examine the specific issue areas that receive more or less attention from MCs. Table 5 reports the results of our seemingly unrelated regression estimations. The dependent variable reported in the left hand column is the percentage of each MCs legislative agenda devoted to that issue area. The columns report the regression coefficients for these individual groups, with safe males serving as the baseline. 15 [Table 5 about here.] Our results strongly support our theoretical claims. We see consistent evidence of strategic behavior in the bill sponsorship patterns of women, both safe and marginal, when compared with their male counterparts. Our estimates indicate that both groups of women balance their legislative agendas by sponsoring legislation in areas traditionally associated with women s interest and in areas traditionally stereotyped as masculine issue areas. Both safe and marginal women devote more attention to the issues of civil rights, civil liberties, and minority issues, and health than do their male colleagues. Additionally, safe women focus more on labor, employment, and immigration, law, 14

16 crime, and family and international affairs (a masculine topic) than do other legislators. Marginal women devote a larger share of their attention to education than do members of the other three groups and safe women devote more attention to social welfare. Despite this focus on feminine topics, however, marginal women also devote a larger share of their agendas to defense than do members of the other three groups. This finding offers particularly strong support for our theory. As expected, we find that electorally vulnerable women prioritize this salient, masculine issue area. This supports the idea that women use sponsorship as a way to establish credibility on topics that are important to voters, but that their gender does not allow them to own. Finally, we find no differences in the propensity of men and women to sponsor bills related to the environment, transportation, community development and housing, banking and finance, and space, science and technology. In addition to sponsoring more legislation on a range of feminine and masculine topics, female MCs keep pace with their male colleagues on a broad range of additional issues. Safe and marginal women are less likely than safe men to focus on macroeconomics (a masculine topic), and public lands, and water management (a gender neutral topic). These findings demonstrate that women do not focus on the issues their gender allows them to own to the exclusion of other topics. Instead, women craft diverse legislative agendas that encompass stereotypically feminine and masculine topics, as well as gender-neutral ones. As a result of this balancing strategy, women introduce more legislation on average than do male legislators, and this legislation is typically spread across a larger number of topics. 15

17 Conclusion We have considered the strategic behavior of women in the U.S. House of Representatives as they attempt to represent the broad interests of women while simultaneously combating gender stereotypes and threats from potential challengers. By examining sponsorship behavior across the full range of issues legislated in Congress, we have demonstrated that women build more diverse legislative agendas than do similarly situated males. Men are able to develop more narrow legislative agendas that demonstrate expertise in a few areas presumably those of particular interest to their constituents and to the representatives themselves. Women on the other hand, must or at least believe they must demonstrate expertise on a broader range of issues to be considered competent. They fortify their credentials on masculine topics like defense and international affairs by sponsoring legislation in these areas while also attending to district interests and issues that are of special significance to women. Safe women in particular are the most active members of Congress as compared with safe males. They sponsor the most bills, on average, and those bills span a wide range of topics. A similar pattern is also true for marginal women, however, the dearth of women overall in Congress limits our statistical power in fully examining this relationship. Overall, our findings provide a more complete portrait of the influence women have had in Congress over the past half-century. We now know with certainty that women are not single-issue representatives who focus their attention narrowly on women s issues. Rather, the legislative portfolios crafted by women are the most diverse in the chamber. These findings strongly suggest that female representatives balance constituency needs and the promotion of women s interests with their own strategic electoral needs. 16

18 Of course, these data do not reveal the degree to which women are successful in these efforts. The question of whether these bills find their way out of committee, onto the floor, and are ultimately enacted is beyond the scope of this project. How often women mention their legislative agendas on the campaign trail is also a topic for future studies. Nevertheless, our findings clearly show that female MCs pursue a number of legislative priorities in addition to the substantive representation of women. As women continue to increase their seat shares in legislative bodies at the state and national levels, scholars should strive to examine the full range of issue priorities women pursue. By focusing narrowly on their efforts to advance women s issues, we run the risk of developing a skewed perception of what female legislators do. Moreover, the influence of female legislators on policy topics traditionally associated with men would seem a promising area for future research. Women bring distinct perspectives to policy issues like crime (Kathlene 1995) and weapons testing (Clark and Clark 1986) that are not considered women s issues. The substantive importance of female representation in Congress, therefore, extends beyond levels of activity on topics like childcare, gender discrimination, reproductive rights, and so on. Researchers might investigate the qualitative nature of the legislation women introduce on masculine topics looking to see if they offer distinct perspectives or policy solutions. 17

19 Notes 1 The 19 major topic areas include macroeconomics, civil rights, minority issues, and civil liberties, health, agriculture, labor, employment, and immigration, education, environment, energy, space, science, and technology, law, crime, and family issues, social welfare, community development and housing, defense, transportation, foreign trade, international affairs and foreign aid, government operations, and public lands and water management. In 2014, the coding scheme was updated to create a standalone immigration category. Our analysis does not reflect this change. 2 Community development and housing includes bills related to low and middle income housing programs, veteran housing assistance, elderly housing, and aid for the homeless. 3 These are coding decisions made by the Congressional Bill Project team. We do not collapse the variables into masculine or feminine issue areas. 4 Additionally, if a member does not sponsor any bills during a given session, he or she does not appear in the CBP data during that session. To resolve this, we compared complete member rosters for each session to the bill sponsorship data and added members who did not put forth bills. A total of 109 members did not introduce legislation in at least one session over the time period of our analysis. We include these members in the analysis only for bill sponsorship where zero is a meaningful number. Including a zero in the analysis of spread or issue areas is not logically sound. These models test competition within an individual s legislative agenda and would not be appropriate to compare zeros across all categories. 5 These data are available by request from Princeton University s Data and Statistical Services. 6 In each session, this value is constructed only for members who have at least one bill introduction. If a member does not have a legislative portfolio, we cannot assess the diversity of that portfolio. 7 The index can alternatively be calculated using proportions rather than percentages, in which case, it ranges from zero to one. 8 If these equations were unrelated, the coefficients and standard errors would be identical to running 19 distinct Time Series OLS models. We estimated these individual equations and concluded that the error structure is related, and thus a SUR modeling strategy is more appropriate. 9 In the appendix we estimate models using a continuous variable and an interaction term. The results indicate the same relationship and levels of statistical significance as using the indicator variables. 18

20 10 Like Mayhew, we estimated models with both the 10% and 20% cut-point and found substantively similar results. 11 Cumby-Huizinga tests for autocorrelation reveal serial correlation in our data. This is not surprising given the nature of MCs behavior, as their previous term would be a good indication of what their current term will feature. We also test for stationary in our series and conclude our dependent variables are stationary. Given these two conditions, we are confident modeling our processes with a lagged dependent variable is appropriate (Keele and Kelly 2006). 12 We classify party leadership as Speaker, Majority and Minority Leader, and Majority and Minority Whip. 13 This data was collected in the census year value and converted to 2012 dollars using the Bureau of Labor Statistics Consumer Price Index Inflation calculator. 14 In alternative specifications of the model, these indicators are included. Using both of these measures is problematic as they are both highly correlated and trend over time. We have accounted for this in alternative model specifications by differencing each measure, but concluded the fixed effects were the more appropriate manner to capture these effects. 15 In the appendix, we report the same models split by party. There are some issues that may have a partisan slant on sponsorship behavior. To account for this, we make cross-gender comparisons within the same party. 19

21 References Boydstun, Amber E., Shaun Bevan, and HF Thomas III The Importance of Attention Diversity and How to Measure it. Policy Studies Journal 42: Clark, Cal, and Janet Clark The Old vs. the New Gender Gap: A Case Study of Deploying the MX in Wyoming.. Druckman, James N, Lawrence R. Jacobs, and Eric Ostermeir Candidate Strategies to Prime Issues and Image. American Journal of Political Science 66: Fox, Richard Gender Dynamics in Congressional Elections. Thousand Oaks, CA: Sage. Garand, James C., and Kelly M. Burke Legislative Activity and the 1994 Republican Takeover: Exploring Changing Patterns of Sponsorship and Cosponsorship in the U.S. House. American Politicts Research 34: Gray, Virginia, and David Lowery The Population Ecology of Interest Representation: Lobbying Com- munities in the American States. Ann Arbor, MI: =University of Michigan Press. Hernson, Paul S., J. Celeste Lay, and Atiya Stokes Women Running as Women: Candidates Gender, Campaign Issues, and Voter Targeting Strategies. Journal of Politics 65: Huddy, Leonie, and Nayda Terkildsen The Consequences of Gender Stereotypes at Different Levels and Types of Office. Political Research Quarterly 46:

22 Kahn, Kim F The Political Consequences of Being a Woman. New York: Columbia University Press. Kathlene, Lyn Alternative Views on Crime: Legislative Policymaking in Gendered Terms. Journal of Politics 57: Keele, Luke, and Nathan J. Kelly Dynamic Models for Dynamic Theories: The Ins and Outs of Lagged Dependent Variables. Political Analysis 14: Laakso, Markku, and Rein Taagepera Effective Number of Parties: A Measure with Application to West Europe. Comparative Political Studies 12: Mayhew, David R. 1974a. Congress: The Electoral Connection. New Haven: Yale University Press. Mayhew, David R. 1974b. Congressional Elections: The Case of the Vanishing Marginals. Polity 6: Osborn, Tracy, and Jeanette Morehouse Mendez Speaking as Wome: Women and Floor Speeches in the Senate. Journal of Women, Politics & Policy 31: Palmer, Barbara, and Dennis Simon When Women Run Against Women: The Hidden Influence of Female Incumbents in Elections to the U.S. House of Representatives, Politics and Gender 1: Sellers, Patrick Strategy and Background in Congressional Campaigns. American Political Science Review 92: Stigler, George J Economic Competition and Political Competition. Public Choice 13:

23 Sulkin, Tracy Campaign Appeals and Legislative Action. Journal of Politics 71: Swers, Michele Building a Reputation on National Security: the Impact of Stereotypes Releated to Gender and Military Experience. Legislative Studies Quarterly 32: Windett, Jason Harold Gendered Campaign Strategies in US Elections. American Politics Research 42(4):

24 A Appendix one [Table A.1 about here.] [Table A.2 about here.] [Table A.3 about here.] [Table A.4 about here.] [Table A.5 about here.] [Table A.6 about here.] 23

25 Figure 1: Predicted Count of Bills Sponsored by Gender and Legislative Competition Predicted Number of Bills Sponsored Safe Male Marginal Male Safe Female Marginal Female % Confidence Interval 24

26 Figure 2: Mean Herfindal-Hirschman Index 25

27 Figure 3: Predicted Herfindal-Hirschman Index by Gender and Electoral Security 95% Confidence Interval Safe Male Marginal Male Safe Female Marginal Female Predicted Legislative Spread

28 Table 1: Summary Statistics for Dependent Variables Dependent Variables Min Max Mean SD Overall Total Bills Sponsored HerfindahlHirschman Index Female Legislators Total Bills Sponsored Herfindahl-Hirschman Index Male Legislators Total Bills Sponsored Herfindahl-Hirschman Index

29 Table 2: Summary Statistics for Key Independent Variables Variable Frequency Percent Safe Females Marginal Females Safe Males 6, Marginal Males 2, Democratic Women Republican Women Democratic Men 5, Republican Men 4,

30 Table 3: Time Series Negative Binomial Predicting the Number of Bills Sponsored Variable Coefficient Bills Sponsored t * (0.00) Marginal Male (0.02) Safe Female 0.219* (0.05) Marginal Female 0.175* (0.07) Widow (0.18) Service Time * (0.00) Racial Minority * (0.05) Committee Chair 0.238* (0.03) Party Leader * (0.07) Party (0.00) District Effects Factor * (0.02) Constant 2.732* (0.06) Wald Chi N 7612 * p<0.05 Cell entries report coefficients from a Time Series Negative Binomial regression. Congress dummy variables are not included in the presentation of the models. Standard errors are in parentheses. 29

31 Table 4: Time Series OLS Model Estimating Agenda Concentration Variable Coefficient HHI t * (39.48) Marginal Males (44.02) Safe Females * (84.06) Marginal Females (158.43) Widow (265.89) Service Time * (4.89) Racial Minority (79.07) Committee Chair (75.69) Party Leader (164.72) Party (.42) District Effects Factor (35.68) R N 7493 * p<0.05 Cell entries report OLS Coefficients with standard errors in parentheses. Congress dummy variables are not included in the presentation of the models. 30

32 Table 5: Seemingly Unrelated Regressions Predicting Legislative Issue Concentration Dependent Variable Marginal Males Safe Females Marginal Females Macroeconomics * * (.21) (.42) (.65) Civil Rights, Liberties, and Minority Issues *.767* (.123) (.262) (.40) Health * 4.484* (.28) (.56) (.87) Agriculture * (.22) (.41) (.67) Labor, Employment, and Immigration -.704* 1.277* (.24) (.49) (.76) Education * (.20) (.40) (.62) Environment (.22) (.44) (.67) Energy * (.19) (.39) (.60) Transportation (.23) (.48) (.75) Law, Crime and Family *.722 (.25) (.50) (.77) Social Welfare * (.18) (.37) (.57) Community Development and Housing (.15) (.30) (.45) Banking, Finance, and Domestic Commerce (.24) (.49) (.74) Defense * (.28) (.58) (.89) Space, Science, Technology (.13) (.26) (.40) Foreign Trade -.604* (.28) (.58) (.89) International Affairs * (.16) (.33) (.51) Government Operations * (.35) (.72) (1.12) Public Lands and Water Management.798* * (.34) (.70) (1.07) + p<0.10, * p<0.05 Cell entries are regression coefficients for the Marginal Males, Safe Females, and Marginal Females variables. In the left hand column are the dependent variables from the 20 regressions estimated in the Seemingly Unrelated Regressions equation. For space considerations, only the variables of interest are reported. Standard errors 31 are reported in parentheses.

33 Table A.1: Time Series Negative Binomial Predicting the Number of Bills Sponsored by Party Variable Democrats Republicans Bills Sponsored t * 0.010* (0.00) (0.00) Marginal Males 0.058* (0.01) (0.30) Safe Females 0.240* 0.201* (0.00) (0.01) Marginal Females * (0.28) (0.01) Widow (0.82) (0.28) Service Time * (0.01) (0.49) Racial Minority (0.07) (0.19) Committee Chair 0.187* 0.301* (0.00) (0.00) Party Leader * * (0.00) (0.01) District Effects Factor * (0.00) (0.81) Constant 2.588* 2.989* (0.00) (0.00) Wald Chi N * p<0.05, Cell entries report coefficients from a Time Series Negative Binomial regression. Congress dummy variables are not included in the presentation of the models. Z- scores are in parentheses. 32

34 Table A.2: Seemingly Unrelated Regressions Predicting Legislative Issue Concentration for Republicans Dependent Variable Marginal Males Safe Females Marginal Females Macroeconomics * * Civil Rights, Liberties, and Minority Issues Health * Agriculture.633* Labor, Employment, and Immigration Education Environmental Energy Transportation Law, Crime, and Family *.417 Social Welfare Community Development and Housing Banking, Finance, and Domestic Commerce Defense Space, Science, Technology Foreign Trade * International Affairs * Government Operations * Public Lands and Water Management 1.386* * Other p<0.10, * p<0.05 Cell entries are regression coefficients for the Marginal Males, Safe Females, and Marginal Females variables. In the left hand column are the dependent variables from the 20 regressions estimated in the Seemingly Unrelated Regressions equation. For space considerations, only the variables of interest are reported. 33

35 Table A.3: Seemingly Unrelated Regressions Predicting Legislative Issue Concentration for Democrats Dependent Variable Marginal Males Safe Females Marginal Females Macroeconomics Civil Rights, Liberties, and Minority Issues * 1.465* Health -.764* 3.71* 6.337* Agriculture -.608* Labor, Employment, and Immigration -.857* 1.48* Education.684* * Environmental Energy * Transportation -.747* Law, Crime, and Family Social Welfare.662* 1.565*.051 Community Development and Housing Banking, Finance, and Domestic Commerce Defense Space, Science, Technology Foreign Trade * International Affairs Government Operations * Public Lands and Water Management * * Other p<0.10, * p<0.05 Cell entries are regression coefficients for the Marginal Males, Safe Females, and Marginal Females variables. In the left hand column are the dependent variables from the 20 regressions estimated in the Seemingly Unrelated Regressions equation. For space considerations, only the variables of interest are reported. 34

36 Table A.4: Time Series Negative Binomial Predicting the Number of Bills Sponsored: Alternative Competition Interaction Variable Coefficient Bills Sponsored t * (0.00) Gender 0.152* (0.06) Previous Electoral Margin * (0.00) Gender*Margin (0.00) Widow (0.17) Service Time * (0.00) Racial Minority * (0.05) Committee Chair 0.238* (0.03) Party Leader * (0.07) Party (0.00) District Effects Factor (0.02) Constant 2.755* (0.06) Wald Chi N 7555 * p<0.05, Cell entries report coefficients from a Time Series Negative Binomial regression. Congress dummy variables are not included in the presentation of the models. Z- scores are in parentheses. 35

37 Table A.5: Time Series OLS Model Estimating Legislative Spread: Alternative Competition Interaction Variable Coefficient HHI 0.402* (0.01) Gender (135.42) Previous Electoral Margin (0.66) Gender*Margin (2.65) Widow (266.01) Service Time * (4.89) Racial Minority (79.27) Committee Chair (75.73) Party Leader (164.74) Party (0.42) District Effects Factor * (36.40) R N p<0.10, * p<0.05 Cell entries report OLS Coefficients with p-values in parentheses. Congress dummy variables are not included in the presentation of the models. Standard Errors are in parentheses. 36

38 Table A.6: Time Series Negative Binomial Predicting the Number of Bills Sponsored: Including Party Control Variable Variable Coefficient Bills Sponsored t * (0.00) Marginal Male (0.02) Safe Female 0.218* (0.05) Marginal Female 0.174* (0.07) Widow (0.18) Service Time * (0.00) Racial Minority * (0.05) Committee Chair 0.239* (0.03) Party Leader * (0.07) Party (0.00) District Effects Factor * (0.02) Majority Party Member 0.731* (0.23) Constant 2.731* (0.06) N Wald Chi * p<0.05, Cell entries report coefficients from a Time Series Negative Binomial regression. Congress dummy variables are not included in the presentation of the models. Standard Errors are in parentheses. 37