Outperformance versus Representational Tradeoffs

The outperformance of women in the legislative arena has been linked to two distinct factors: patterns of sex-based selection in elections and the institutional barriers that women face in office. ^{Footnote 1} Anzia and Berry (Reference Anzia and Berry2011) develop a theory of sex-based selection in elections to explain their finding that female legislators bring more money to their districts than men. The main argument is that only the most ambitious and qualified women run due to perceived or actual gender discrimination in the electoral process (see also Pearson and McGhee Reference Pearson and McGhee2013). ^{Footnote 2} They further test the argument by leveraging variation in discrimination across districts, and they use district ideology as a proxy for the prevalence of sex-based selection in a district. If conservative districts tend to have higher levels of discrimination, the spending advantage of women legislators should be greater than that in liberal districts. While district conservatism is not a perfect indicator of the degree of gender bias in elections, they do find that the positive effect of female representation on spending is larger in conservative districts where public attitudes are likely to be less open to women in politics (Anzia and Berry Reference Anzia and Berry2011). ^{Footnote 3}

Others have instead pointed to institutional reasons for women’s outperformance in the legislative arena. Scholars and officeholders alike have documented the institutional barriers that affect women in elected office (Boxer Reference Boxer1993; Hawkesworth Reference Hawkesworth2003; Schroeder Reference Schroeder1998). In an analysis of over ten thousand speeches, Pearson and Dancey (Reference Pearson and Dancey2011) find that congresswomen speak more on the House floor than their male colleagues on a range of issues. They attribute this pattern in part to the additional incentive that women have to prove their legislative credentials in a predominately male institution. Pearson and Dancey (Reference Pearson and Dancey2011) do not examine differences in floor speech patterns across legislative contexts because their empirical focus is on the U.S. House, but one implication is that sex differences would diminish as the pressure women face to prove their legislative credentials decreases. And although the authors do not extend the argument to constituency service, the pressure on women to demonstrate their expertise likely extends to this aspect of their job as well. It may be that women in more male-dominated institutions or in institutions with fewer women in positions of power have additional incentive to be responsive on casework. ^{Footnote 4}

We would expect either of these factors to lead women legislators to outperform their male counterparts on constituent service. In fact, earlier research on the relationship between gender and service responsiveness suggested that women legislators do spend more time on casework than men. In a four-state survey of state legislators, Richardson and Freeman (Reference Richardson and Freeman1995) find that women receive more requests from constituents and that women believe they put more emphasis on constituency service than other legislators in their state. Thomas (Reference Thomas1992) draws on a survey of city council members and shows that women spend more time on constituent concerns than male councilors. Additionally, several studies in the 1970s and 1980s found that women emphasized their obligations to the community more than male legislators (Antolini Reference Antolini and Flammang1984; Diamond Reference Diamond1977; Flammang Reference Flammang1984; Johnson and Carroll Reference Johnson and Carroll1978). We refer to this expectation as the Outperformance Hypothesis. We include two sub-hypotheses to account for the distinct factors identified here that may shape women’s outperformance in the legislative arena.

• Outperformance Hypothesis: Women legislators are more likely to respond to constituent requests than male legislators.

• Electoral Bias: Women legislators outperform their male counterparts because of perceived or actual gender bias in the electoral process. Outperformance should increase (decrease) in districts where gender bias in elections is more (less) prevalent.

• Institutional Bias: Women legislators outperform their male counterparts because of the pressure to demonstrate their expertise in male-dominated insti-tutions. Outperformance should increase (decrease) in institutions that are more (less) male-dominated.

At the same time, there is also reason to suspect the opposite might be true and that women underperform their male counterparts on constituency service. First, the findings just discussed were based on women legislators’ perceptions of the time they spent on casework, rather than objective measures of responsiveness, and women may have over reported their efforts due to gendered social expectations at that time. More recently, Butler (Reference Butler2014) conducted a field experiment of state legislators and found that men respond to non-women’s issues at higher rates than women, although female legislators are more responsive on women’s issues. ^{Footnote 5} Second and more importantly, all of the previous studies focus on a single dimension of representation; there are no analyses that examine gender differences in service aspects of representation in conjunction with policy aspects of representation. Legislators have limited time and resources to allocate to their various duties, and they face tradeoffs in how to fulfill their responsibilities (Ellickson and Whistler Reference Ellickson and Whistler2001; Freeman and Richardson Reference Freeman and Richardson1996). Harden’s (Reference Harden2016) analysis of the multidimensional nature of representation demonstrates that some legislators devote more time and resources to policy responsiveness while others spend more on constituent service.

Gender scholars have examined the policy dimension of representation, but they have yet to consider representational tradeoffs: whether women underperform their male counterparts on constituency service as a result of their increased legislative activity. The legislator surveys that uncovered gender differences in constituency service in the 1970s and 1980s were not analyzed alongside other legislative activities, nor have the legislative productivity studies in recent years been discussed in relation to non-policy aspects of representation. It may be that women legislators are policy types but not casework types, devoting more attention to policy than their male counterparts but less to constituent service. Women may be more productive than men in some areas and less so in others, and they may trade off constituency service in order to pursue their policy agenda. We refer to this expectation as the Tradeoffs Hypothesis.

• Tradeoffs Hypothesis: Women legislators are less likely to respond to constituent requests than male legislators as a result of their increased legislative activity.

In sum, there are good theoretical reasons to expect women legislators to outperform or underperform on constituency service relative to their male counterparts. It may be that either sex-based selection in elections or increased pressure for women to prove their legislative credentials in male-dominated institutions leads female legislators to outperform their male counterparts not only in the legislative realm but also in non-policy aspects of representation. On the other hand, the resource and time constraints that all legislators face may mean that female legislators’ higher policy productivity leads to lower service productivity than their male legislators. We test these hypotheses in the sections that follow.

Audit Study of State Legislators

We conducted an audit study of state legislators to examine whether female legislators are more or less responsive to constituent requests than their male counterparts. ^{Footnote 6} Our design is similar to Butler and Broockman’s (Reference Butler and Broockman2011) and Butler’s (Reference Butler2014). We used a 2x2 design and we randomized the gender of the constituent and the gendered content of the e-mail. We focus mainly on gender differences in legislator responsiveness, but we briefly discuss the treatment in the last section of the article. We used a total of ten aliases, five male and five female. ^{Footnote 7} The text of the e-mail that was sent to state legislators is provided in figure 1. All legislators received a request for voter registration information. We chose voter registration for a few reasons. We wanted our requests to require little effort due to ethical concerns. ^{Footnote 8} We also sought to use an issue that has minimal partisan divides because we did not want to signal the ideology of the individual. ^{Footnote 9} In addition, voter registration has already been used in several studies, so we could compare our results with those for additional validation.

Figure 1

E-mail sent to legislators

Our sample includes more than 6,000 state senators and representatives. We obtained their e-mail addresses in March 2016 through state legislature websites and online searches. Like previous studies, we treat responses from either the legislator or a staff member as equivalent, with the level of analysis being the legislator’s office (Butler Reference Butler2014; Butler and Broockman Reference Butler and Broockman2011; Carnes and Holbein Reference Carnes and Holbein2015). ^{Footnote 10} Many decisions that affect representation are made at the office level, and elected officials rely on staff to help with various aspects of their work. Unlike most previous studies, we also examine whether the reply was helpful because we are interested in the quality of the response as well. Replies were coded as helpful if they included any of the following information: an e-mail link to online voter registration; an e-mail link to information on how to register; an address, e-mail address, or phone number of a government office; a name, e-mail address, or phone number of an individual at a government office; or an offer to personally deliver a voter registration card. ^{Footnote 11} The majority of helpful replies consisted of e-mail links to or contact information of the county or state board of elections.

We merged the audit study data with bill sponsorship data from LegiScan, a website that tracks legislation at the state level. ^{Footnote 12} This allows us to examine legislator responsiveness while accounting for variation in policy activity. One implication of the Tradeoffs Hypothesis is that gender differences in responsiveness may diminish when legislative activity is included in the models. LegiScan provides data on all bills that were sponsored in each state legislature, as well as the various sponsors and cosponsors in the legislative session in which we sent out the e-mail request. The LegiScan data offer a host of new opportunities for scholars to explore gender differences in legislative activity at the state level (see also Holman and Mahoney Reference Holman and Mahoney2018). While previous analyses of male and female state legislators have examined various subsets of states, we use these data to measure sponsorship patterns across our entire sample. Of our full sample of state legislators, we were able to match 98.3% of them to the bill sponsorship data. Bill sponsorship is measured as the logged value of the total number of bills the legislator sponsored.

We include a variety of individual and district-level control variables in the models, including party, chamber, and minority party status. We include party leader, committee chair or vice-chair, and years in state legislative office to capture seniority and experience. The party leader and committee chair data are from the National Conference of State Legislatures and the state Yellow Books, respectively. We collected the year in which legislators were first elected to the legislature from Project Vote Smart and legislator websites and use their total number of years in office. We draw on two variables to capture the degree of electoral competition faced by the legislator, both of which are from Ballotpedia. We use the legislator’s previous vote share to account for the possibility that women win by smaller margins than men and thus may be more responsive to constituents. We also control for whether the legislator was up for reelection in the year of our study. To measure district demographics, we use Tausanovitch and Warshaw’s (Reference Tausanovitch and Warshaw2013) ideology estimates, and we draw on census data for district population and data from the National Historical Geographic Information System for the median income of the district. State fixed effects are included in all of the models to account for state-level factors that influence responsiveness such as legislative professionalism and variation in electoral competition.

Legislator Gender and Constituency Service

We use OLS regression to examine whether women are less responsive to constituent requests than men in light of their additional policy efforts. ^{Footnote 13} The results of the audit study are provided in table 1. ^{Footnote 14} We can see that female legislators are more, not less, responsive to constituent requests than male legislators. In addition, not only are women more likely to respond to voter registration requests (column 1), they are also more likely to provide information that will help constituents navigate the registration process (column 2). The predicted probability of responding is four percentage points higher for female legislators than for men (59 and 55%, respectively) and the probability of providing a helpful response is six percentage points higher for women than for men (40 and 34%, respectively), controlling for legislator and district characteristics. ^{Footnote 15} The magnitude of the effect is similar to that in Butler and Broockman’s (Reference Butler and Broockman2011) study as well. ^{Footnote 16}

Table 1

Women are more likely to reply, and reply helpfully, to constituent requests

Note: OLS regression coefficients with robust standard errors in parentheses. All models include state fixed effects. **p<0.01, *p<0.05, †p<0.10.

We are also interested in whether this relationship holds when we take into account the increased policy activity of female legislators. The results in columns 3 and 4 suggest that it does. In fact, the coefficients remain virtually unchanged with the inclusion of bill sponsorship data. Even when accounting for bill sponsorship activity, women legislators are more likely to respond, and to respond helpfully, to constituent requests than their male counterparts. Bill sponsorship is actually positively associated with legislative responsiveness to our e-mail request. Thus, in our sample of state legislators, there is limited evidence of a tradeoff between policy activity and constituency service; rather, those who sponsor more bills are also more responsive to constituent requests. The findings suggest that legislators who are more active in one aspect of representation do not necessarily underperform their colleagues in other areas. Here, it is not the case that women legislators devote less time to constituent service because of the extra attention they devote to policy work. ^{Footnote 17}

With respect to the control variables, state senators and party leaders are more likely to respond to constituent requests and more likely to respond helpfully. Legislators from conservative districts are more likely to reply as well, but this relationship is driven mostly by women, which will be addressed in more detail later. Minority party legislators and those who have served more years in office are less responsive to constituent requests. Those who represent more populous districts are less likely to reply, perhaps because they receive a greater number of constituent requests. Lastly, legislators who represent wealthier districts are not more likely to reply but they are more likely to provide a helpful response.

Electoral and Institutional Factors

The previous section demonstrated that women legislators are more likely to respond to constituent requests than their male counterparts, even when accounting for their increased policy activity. Scholars have identified two different reasons for why women outperform their male counterparts in the legislative realm, delineated earlier as the Electoral Bias and Institutional Bias hypotheses. The Electoral Bias hypothesis suggests that that only the most ambitious and qualified women run for office due to perceived or actual discrimination in elections (Anzia and Berry Reference Anzia and Berry2011; Pearson and McGhee Reference Pearson and McGhee2013). The Institutional Bias hypothesis instead highlights the barriers that women officeholders face in male-dominated legislatures (Hawkesworth Reference Hawkesworth2003; Pearson and Dancey Reference Pearson and Dancey2011). In this section, we leverage district and state variation to gain further insight into whether outperformance varies across electoral and institutional contexts in the way these hypotheses suggest.

Our particular concern is whether outperformance increases in districts where gender bias in elections is more prevalent and whether outperformance decreases in legislatures with more women in office. We measure gender bias in elections in two ways. First, we follow the empirical approach in Anzia and Berry (Reference Anzia and Berry2011) and use district conservatism, and we draw on Tausanovitch and Warshaw’s (Reference Tausanovitch and Warshaw2013) state legislative district ideology estimates. Second, we use Pyeatt and Yanus’s (Reference Pyeatt and Yanus2016) measure of “women-friendly” state legislative districts where it is available (WFD Index). The measure is an extension of Palmer and Simon’s (Reference Palmer and Simon2008) seminal research showing that wealthier, more educated, and more urban districts tend to elect more women to congressional office. Of course, we do not know that women face higher hurdles in more conservative or less women-friendly districts, but these measures attempt to capture variation in the electoral environment that may affect legislators’ behavior in office. Higher values correspond to increasing conservatism and more women-friendly districts, respectively. We include an interaction between each measure and legislator sex to test the Electoral Bias hypothesis. We would expect the coefficient on the district ideology interaction term to be positive and the women-friendly district interaction term to be negative, as outperformance should increase in districts that are more conservative and decrease in districts that are more open to women in politics.

To measure gender bias in legislative institutions, we use data from the Center for American Women and Politics (CAWP) (2016) and Klarner (Reference Klarner2013) to calculate the percentage of female legislators in the lower or upper chamber. While there is also evidence that women receive less support from their male colleagues as their ranks increase (Kanthak and Krause Reference Kanthak and Krause2012), women as a group nevertheless face unique barriers due to their marginalized status in male-dominated legislative institutions. The additional incentive women have to prove their credentials should diminish as women advance in legislative institutions. To test this hypothesis, we first include an interaction between the percentage of women in the state legislative chamber and legislator sex. Second, because the marginalization of women in legislatures extends beyond numerical representation, we also include an interaction between legislator sex and the percentage of female committee chairs in the chamber to measure women’s influence and power in the chamber. ^{Footnote 18} If the Institutional Bias argument were right, we would expect the coefficient on both interactions to be negative, as women should face less pressure to demonstrate their expertise when they are more numerous in public office and when they comprise a larger proportion of committee chairs. To be sure, it is difficult to wholly disentangle electoral and institutional factors, but we think our variables capture either more of the electoral or institutional context in a way that maps onto the arguments we seek to test.

The results are presented in table 2. Our findings provide strong evidence that the outperformance of women in conservative and less women friendly districts extends beyond legislative activity and into constituent service as well (columns 1 through 4). The interaction between legislator sex and gender bias in elections is as hypothesized. The effect of female representation on legislator responsiveness increases in more conservative districts (columns 1 and 2) and decreases in districts that are more open to the election of women (columns 3 and 4). The magnitudes of the relationships are substantial as well.

Table 2

Sex-based selection vs. male-dominated institutions

Note: OLS regression coefficients with robust standard errors in parentheses. Models 1-4 include state fixed effects. **p<0.01, *p<0.05, †p<0.10.

Figure 2 presents the predicted probability of responding and responding helpfully for male and female state legislators across values of district conservatism. The probability that female legislators respond is 58% for an average district in California, compared to 67% for an average district in Alabama. (Average district conservatism is -0.32 in California and 0.26 in Alabama.) Similarly, the probability that a female legislator responds helpfully is 39% in California and 44% in Alabama. In addition, the 9 percentage point difference in the probability of responding between the average male and female legislator in Alabama (58 and 67 percent, respectively) is more than twice as large as the effect of gender in the full sample in table 1 (4 percentage points). Interestingly, the likelihood of responding does not differ between male and female legislators in very liberal districts, but there are fewer observations as well.

Figure 2

Predicted probability of responding and responding helpfully by district conservatism and state legislator sex

Note: Predicted probabilities are calculated from the models in columns 1 and 2 of table 2.

The results are similar with the women-friendly district measures (columns 3 and 4 in table 2). For example, a shift from the average level of women friendliness in Connecticut state legislative districts to the average level in South Dakota state legislative districts leads to a 9 percentage point increase in the likelihood that women legislators respond to constituent requests (59 and 68%, respectively). Taken together, the patterns are consistent with Anzia and Berry’s (Reference Anzia and Berry2011) finding that women legislators bring more money to their districts and that women in conservative districts bring home even more. Our findings uncover yet another way in which conservative female legislators behave differently in office than liberal female legislators. We are unable to address whether the electoral hurdles for conservative women are actually higher than they are for liberal women, but we do see differences in how they behave in office. ^{Footnote 19}

However, we find little evidence that the outperformance of female legislators decreases as the number of women in office or in positions of power increases (columns 5 through 8 in table 2). ^{Footnote 20} The interactions are negative but not statistically significant, and the probability of responding overlaps for male and female legislators across levels of women’s representation in the chamber (columns 5 and 6) and values of female committee chairs (columns 7 and 8). ^{Footnote 21} However, one crucial point to note is that women are underrepresented across state legislative institutions and perhaps there are too few legislatures at the very high end of this range for us to adequately test this argument. For example, there are only 103 male legislators and 80 female legislators in the sample who were in legislative chambers with more than 40% women, or a mere 3% of the sample of state legislators. The average percentage of women in the legislative chamber is 24%, and the lack of cases at the upper end of the distribution limits how fully we can test this hypothesis. Nevertheless, in the sample of state legislators here, there is little evidence of a relationship between the presence or influence of women in office and responsiveness across female legislators.

Alternative Explanations

The findings provide additional support that female legislators outperform their male counterparts across a range of representational activities. This section examines several alternative explanations that would cast doubt on the idea that quality differences between male and female legislators are the reason that women are more responsive to constituent requests than men. First, we look at a couple of ways in which differences in legislative staff might matter for our results. We consider whether female legislators hire more competent staff than male legislators, and whether staff responses, not legislator responses, might account for this relationship. To address this possibility, we limited our analysis to states with no legislative staff because, in these states, responsiveness cannot be attributed to having more competent staff (columns 1 and 2 of table 3). ^{Footnote 22} In addition, we coded whether each reply came from the legislator or a member of her staff. We excluded staff replies to see whether the same pattern occurred among the subset of replies that came from legislators (columns 3 and 4 of table 3). ^{Footnote 23} We can see in columns 1 and 2 that, even in states where there are no legislative staff, female legislators are more responsive to constituent requests than men. The magnitude of the relationship remains the same as in the analyses in the previous section. Similarly, in columns 3 and 4, women are more likely to respond when staff replies are excluded from the analysis. Thus, there is little indication that this gender difference is due to staff responses.

Table 3

Legislator responsiveness, considering staff and ideology

Note: OLS regression coefficients with robust standard errors in parentheses. All models include state fixed effects. **p<0.01, *p<0.05, †p<0.10.

Another possibility is that women legislators are more likely to hire female staff and that female staff are more competent than male staff. We cannot measure this directly, but one implication is that staff replies from the offices of female legislators should be more likely to come from women than staff replies from the offices of male legislators. We coded whether the staff reply came from a male or female name, and the responses from the offices of female legislators are as likely to come from women as those from the offices of male legislators. ^{Footnote 24} These results are not presented here but are further discussed in online appendix J.

Second, female state legislators might be more responsive because they are more liberal, on average, than their male counterparts (Carroll and Sanbonmatsu Reference Carroll and Sanbonmatsu2013; Thomsen Reference Thomsen2015, Reference Thomsen2017). Several studies have shown that liberal legislators spend more time and energy on constituency service than conservative legislators who believe the role of government should be limited (Cain, Ferejohn, and Fiorina Reference Cain, Ferejohn and Fiorina1987; Ellickson and Whistler Reference Ellickson and Whistler2001; Freeman and Richardson Reference Freeman and Richardson1996). We include Bonica’s (Reference Bonica2014) estimates of state legislator ideology, with higher values corresponding to ideological liberalism. The results are presented in columns 5 and 6 of table 3. Consistent with previous findings, liberal legislators are more likely to respond to constituent requests, but women are still more responsive than their male counterparts.

Finally, it may be that women are more likely to respond to female constituents than male legislators. ^{Footnote 25} In addition, a host of studies have shown that women legislators devote more attention to women’s issues, so perhaps women are also more responsive to gender-related appeals (see also Butler Reference Butler2014). As noted earlier, we included an experimental aspect in our study to delve into these possibilities further. In half of the e-mails, we added a gender appeal to examine whether legislators respond differently to requests that reference gendered concerns. We invoked gender by noting that the individual was a mother or father of two and concerned about the rising costs of childcare. ^{Footnote 26} We interacted legislator sex with both the gender of the constituent and the gendered nature of the appeal. We also interacted constituent gender with the gender appeal to see if women legislators were still more likely to respond. These results are provided in table 4. None of the interactions reach conventional levels of significance. Women respond at higher rates to all requests, but they are not additionally more likely to respond to female constituents or gender-based appeals. ^{Footnote 27} Our results differ from Butler’s (Reference Butler2014) finding that women are more responsive on women’s issues, but it may be that our gender-related treatment was less overt since it was in the context of voter registration. The sample in Butler’s analysis is also smaller than ours, and our goal was to examine responsiveness across a more generalizable set of legislators. Again, the results here are consistent across a host of specifications and analyses: women legislators are more likely to respond to constituent requests than their male counterparts, even after accounting for their increased policy activity.

Table 4

Legislator responsiveness and gender-based representation

Note: OLS regression coefficients with robust standard errors in parentheses. All models include state fixed effects. **p<0.01, *p<0.05, †p<0.10.