Abstract
Scholars argue that women’s presence in politics enhances symbolic representation, such as positive evaluations of one’s representative and increased political engagement. However, there is little empirical evidence of these symbolic benefits from descriptive representation. With data from the Cooperative Congressional Election Study panel survey, we examine how a change in the gender of a representative affects individuals’ perceptions of that representative and likelihood to contact them. In general, we find that women express more positive evaluations of female representatives than male representatives, yet they are also less likely to contact female representatives. By contrast, the effect of an elected official’s gender does not significantly affect how men evaluate or engage with that official. However, we also show that partisanship conditions these effects, perhaps due to the fact that gender stereotypes operate differently for Democrats than Republicans. For example, women rate female Republican legislators more positively than they do male Republican legislators, but neither women nor men rate Democratic legislators differently based on their gender. The findings provide strong evidence that gender matters when it comes to representation, but contrary to some conventional wisdom, female elected officials may actually enjoy some advantages in terms of their standing among constituents.
Introduction
A considerable body of research has documented the ways in which women have distinct representational roles in Congress, as well as distinct roles as constituents. Many scholars have claimed that descriptive representation is especially important for women because they historically occupy a marginalized place within politics and society (Hansen 1997; Mansbridge 1999; Phillips 1995; Rosenthal 1995). Women, thus, have distinct policy preferences and interests in relation to men, which draws them together as a group (Carroll 2002; MacDonald and O’Brien 2011; Mendelberg, Karpowitz, and Goedert 2014; Osborn and Mendez 2010). In addition, women are purported to have a distinct style of governing that affects their behavior in office, including engaging with the public (Carroll 2002; Mendelberg, Karpowitz, and Goedert 2014; Reingold 2008; Swers 2002). Women’s presence in political office, therefore, can increase substantive representation of women—the responsiveness and attention to women’s policy interests. Of course, women vary in their interests, preferences, and needs, so the mere presence of women legislators is not necessarily sufficient for substantive representation. But descriptive representation may, nevertheless, influence women’s political behavior in the mass public.
If female elected officials do provide a unique type of representation for women, as the research suggests, then how does this affect the way in which women evaluate and engage with female elected officials? If at least some women are aware that they receive better representation from female elected officials, then they might be prone to evaluate such legislators more favorably. These patterns may also influence the way in which they engage with their representatives. On one hand, women might contact female elected officials more often, owing to the fact that they might feel more efficacious about the prospects of such contact. On the other hand, women may trust female elected officials more than male officials to represent them well, and, thus, feel less of a need to contact them to express their policy concerns.
While previous studies have examined how gender conditions the way in which men and women evaluate elected officials differently, these studies have been limited in their ability to make widely generalizable causal inferences. In this paper, we address these limitations by making use of a unique data source and the exogenous shock caused by a redistricting cycle. Specifically, we draw on data from the 2010–2014 Cooperative Congressional Election Study (CCES) panel survey (Schaffner and Ansolabehere 2015), which included interviews with 9,500 American adults across three election cycles and spanning the 2012 congressional redistricting process. Using an individual fixed-effects approach, we examine the extent to which a change in the gender (but not the party) of an individual’s legislator affects how they evaluate that legislator and the extent to which they initiate and evaluate contact with him or her. Using this unique causal leverage in a real-world setting, we find that women tend to evaluate female legislators more favorably, but that they initiate contact with such legislators less frequently. By contrast, the effect of an elected official’s gender on how men evaluate or engage with that official is contingent on the party of those male constituents. While Democratic men evaluate women legislators more favorably, the opposite is true for Republican men. Overall, our findings suggest that women (and Democratic men) do view female legislators as better representatives, and that these improved evaluations may mean they feel less of a need to contact those legislators to express their views on policy matters they care about. These findings are also conditioned by the party affiliation of the legislator. For example, women rate female Republican legislators more positively than they do male Republican legislators, but neither women nor men rate Democratic legislators differently based on their gender.
Gender and Perceptions of Representatives
While the US-focused research on women and politics generally shows that female elected officials provide a unique type of representation to women in the electorate, there has generally been less agreement on the extent to which female representation influences the attitudes and behaviors of women. Some studies indicate that descriptive representation is associated with symbolic benefits for women in the electorate; female constituents have a greater sense of trust and vertical understanding toward women representatives, possibly leading to a heightened sense of awareness and connection to government (Burns, Schlozman, and Verba 2001; Carroll 2002; Hansen 1997; Verba, Schlozman, and Brady 1995). In addition, sharing descriptive characteristics with a marginalized group often enhances surrogate representation, and affects how constituents and legislators perceive one another (Broockman 2013; Butler and Broockman 2011; Fenno 1978). These attitudinal effects can lead more women to vote and be active in politics (Atkeson and Carrillo 2007; Campbell and Wolbrecht 2006; Dolan 2006; Schramm 1981). However, some research has found little or no empirical evidence of the symbolic benefits of more women in political office. Descriptive representation does not always affect women’s political trust, efficacy, competence, nor engagement (Broockman 2014; Karp and Banducci 2008; Lawless 2004). In fact, Karp and Banducci (2008) find some increased efficacy and activity for individuals represented by women in other countries, but the United States stands out as a case where they find no such effects.
In general, this body of research has focused on the impact of female elected officials on general levels of efficacy and participation among women. However, much less work has considered the effects of descriptive representation on women vis-à-vis their own elected official. That is, many scholars have examined whether women are more likely to trust government or vote when they are represented by a woman, but few have studied whether women are more likely to evaluate their representative favorably or contact their representative if that representative is a woman. Lawless’s (2004) study is one notable exception to the overall lack of attention to these more direct effects of descriptive representation. She finds that women are more likely to approve of their member of Congress when that member is a woman. However, her study goes on to examine whether descriptive representation promotes more general efficacy and participation (such as trust and voting) among women in the electorate, and she finds no such effect.
Our study builds on this body of research in three ways. First, we focus on attitudes and activities that are related to the elected official who is providing the descriptive representation. Specifically, we examine how constituents evaluate female representatives differently from male representatives and the extent to which they initiate contact with those representatives. One’s propensity to contact their representative links symbolic and substantive representation; it indicates a level of political activism and awareness (symbolic), as well as serves as a key component of constituency service (substantive). Descriptive representation is at least partly thought to be important because it increases political efficacy; members of the represented group feel as though they have a voice in government, and feel they can more easily communicate with representatives who resemble them on some key trait. Communication between legislators and constituents, therefore, deserves attention on its own when examining representation. Because letters and e-mails are among the few means of directly communicating with their constituents, representatives take this communication seriously and deal with it daily (Goldschmidt, Cooper, and Fitch 2011). Citizens contact their representative for help accessing governmental services and benefits or to express policy opinions and persuade their legislator to take action on some legislation.
Based on different conceptions of how descriptive representation might influence the relationship between constituents and their legislators, we develop several expectations. On one hand, women may be more likely to contact their legislator when that legislator is a woman. We might expect such an effect because women will feel as though they are more likely to be heard or understood by a female representative than by a male, and, thus, may feel more efficacious about their attempts at communication. That is, they may feel as though their attempts at expressing their views or concerns will be more successful if heard by a female representative than a male representative of the same party. On the other hand, women may feel as though it is less necessary to contact a female representative than a male. After all, if women believe that their views and priorities are more likely to be internalized by a female representative, then they may feel less of an impetus to contact that representative to express those views. Moreover, even if women believe their policy views are shared by male and female representatives alike, they may trust female representatives to carry out those priorities in Congress more than male representatives. This perceived competence may lead women to contact female representatives less frequently than male representatives.
To be sure, we also recognize that the effects of descriptive representation are likely to be conditioned by more powerful forces, such as partisanship. Indeed, recent scholarship on the impact of the gender of politicians has made note of the important ways in which the public’s perceptions of partisanship and gender intersect. For example, Swers (2013) notes that “gendered perceptions of issue competencies are mediated by party reputations”; specifically, women are thought to have a greater level of commitment and expertise on social welfare and feminist issues, but so, too, are members of the Democratic Party in general. This means that issue stereotypes are reinforcing for Democratic women, but in conflict for Republican women. That is, the average citizen would already expect a Democratic member of Congress to be highly attentive to social welfare issues, so there may be little difference in how citizens perceive male versus female Democrats on that dimension. By contrast, because people do not typically expect Republicans to emphasize social welfare issues, they may hold divergent expectations for male versus female Republicans. Specifically, they might anticipate that female Republicans will place more of an emphasis on social welfare issues compared with their male counterparts. Indeed, Schneider and Bos (2016) show that in comparison with Republican men, Republican women are often viewed as closer to Democrats of either gender when it comes to perceived emphasis on issues related to women’s rights, affirmative action, the wage gap, and children’s welfare.
However, it is not just on issue domains where the intersection of gender and partisan stereotypes might be consequential; people also use both gender and party affiliation to stereotype politicians’ personality traits. For example, Hayes (2011) and Winter (2010) find that Democrats tend to be associated with stereotypically feminine traits such as compassion while Republicans tend to be connected to masculine traits such as strength. As a result, a female Democrat may be stereotyped by voters as having more feminine traits both by virtue of their partisanship and their gender, whereas a female Republican may send conflicting signals. Put another way, while Democratic legislators may be viewed as compassionate regardless of their gender, Republican women may be viewed as more honest and compassionate (but weaker leaders) compared with male Republicans. Overall, these extant findings related to the intersection of gender and partisanship lead us to expect that the gender of the representative may be particularly important for constituents who are represented by Republicans. Thus, our second contribution is to investigate whether the effect of gender is conditional on the partisanship of the legislator.
Our third contribution is to provide a stronger empirical test of the impact of representation by female legislators. Indeed, much of the previous research in this area has made use of cross-sectional observational designs, which undermine the extent to which scholars can make strong causal inferences. In recent work, Broockman (2014, 3) notes that “few studies have persuasively identified the causal effect of women’s candidacies or officeholding on women’s mass and elite-level participation in the US.” One of the concerns with observational designs that compare attitudes of individuals who have female representatives with those represented by males is the possibility that women may be more likely to win elections in places where people are less critical of their elected officials or where women participate more in the electoral process. If this is the case, then selection bias would cause us to overestimate the causal effect of legislator gender on how women evaluate their elected officials or participate in the political process.
In this paper, we use a unique large-N panel survey occurring over a redistricting cycle to allow us to make stronger causal inferences about these dynamics than would be possible with a cross-sectional observational analysis. In the following section, we describe our approach to analyzing these data and how they allow us to make inferences about the causal effect of female elected officials on citizen evaluations and activity.
The CCES Panel Survey
Our analysis takes advantage of the redistricting cycle as a natural experiment and a unique panel survey designed specifically to study issues of representation in Congress. The redistricting cycle is a particularly helpful period during which to conduct a study of this nature, because the shifting of congressional seats and boundaries leads to a significant shakeup in terms of representation in the US House. Redistricting affects the representation of individuals in two ways. First, many individuals’ addresses are moved to new districts that are represented by a different legislator. Second, the redistricting and reapportionment process generally leads to increased legislator retirements or defeats, creating an additional shock to the typical stability of congressional representation.
The CCES panel survey includes 9,500 respondents interviewed six times over three election cycles from 2010 to 2014 (Schaffner and Ansolabehere 2015). Of the 9,500 total respondents, 3,956 had a different representative in 2013–2014 than they did in 2011–2012. We begin with this group for our analysis. However, because evaluations and communications with representatives are highly conditioned by partisanship, we attempt to remove this as a factor by limiting our analysis to those instances where an individual had a new representative who affiliated with the same party as the old representative. We also limited the analysis to those individuals who did not change addresses between 2012 and 2014 as the act of moving can cause disruptions to an individual’s propensity to participate in politics (Ansolabehere, Hersh, and Shepsle 2012). There were 2,431 individuals in the sample who met these conditions. Among this group, the most common scenario was to move from one male legislator to another—1,762 individuals fit this condition. Only 81 individuals went from having one female representative to a different female representative of the same party. Among those who experienced a change in the gender of their representative, 280 moved from having a male representative to a female representative of the same party, while 308 changed from a female representative to a male representative.
The panel nature of the survey is particularly valuable for investigating the role of gender on an individual’s evaluation of her representative because it allows us to examine how the same individuals respond when the gender of their representative changes. To estimate the effect of legislator gender, we use cross-sectional time-series regression with individual fixed effects. We use a series of dependent variables (described below), and our key independent variable is the gender of the representative that the individual is evaluating in that particular year. The use of the individual fixed effects means that the model accounts for any time-invariant differences between individuals who experienced a change in their representative’s gender and those who did not. Thus, the estimate that we recover from this approach is essentially the average within-unit effect of having a female (rather than male) representative. This approach is also particularly efficient because it makes use of both types of change that we observe—moving from a male representative to a female representative and vice versa. 1
We are interested in several variables that capture how respondents evaluated their members of the US House in both 2012 and 2014. Each of the questions we detail here was asked in both waves and in relation to the member of Congress who had represented that person during the previous two years. Thus, we are able to examine how a change in the gender of the individual’s House member affected her evaluations on the identical questions. First, we examine the extent to which men and women perceive their representatives to be closer (or farther) from them on the ideological scale. To do this, we draw on questions that ask the individual to place him or herself on a 7-point ideological scale and the representative on the same scale. Our ideological distance measure is simply the absolute value between where the individual places herself and her representative on those scales. The measure ranges from 0 to 6, and has an average of 2.11. If women think that female representatives represent them better on the issues, then we should find that they perceive their female House members to be closer to them ideologically than male House members.
The next two measures we examine ask respondents to rate their House member on a scale from 1 = extremely strong to 7 = extremely weak on the characteristics of competence and integrity. These measures had similar mean values (4.47 and 4.52, respectively) and were recoded so that higher values would be associated with the representative being stronger on each trait. We expect that women will perceive female representatives as having more competence and integrity compared with male representatives.
Our last two measures capture a more general overall evaluation of the House member’s performance in office. The first question asked respondents, “How well do you think [HOUSE REPRESENTATIVE NAME] represents the people in your congressional district?” Respondents could indicate that their member represented the district very well, somewhat well, or that she does not represent the district well. We expect that women are more likely to think that female legislators represent the district well compared with male legislators. The second question asks respondents to indicate whether they approve of the job their House member is doing in office. The response options for this question ranged from strongly approve to strongly disapprove. We expect that women will be more likely to approve of the job that the representative is doing when that representative is female rather than male (Lawless 2004).
To make the magnitude of effects more comparable, each of the above measures was rescaled to range from 0 to 1, with 1 indicating the most favorable evaluation on each question, and 0 indicating the least favorable view. However, for the ideological distance measure, 0 indicates that the individual placed the House member and themselves at the same position on the scale, and 1 indicates that they placed the House member as far away from themselves as is possible.
Finally, we also include an analysis of the respondent’s approval rating of President Barack Obama as a placebo test. A respondent’s evaluations of her representative may be affected by a change in the gender of that representative, but how she evaluates the president should be unaffected by such a change. If we were to find a difference on this measure, then it would be cause for concern about spurious factors that our design is failing to account for.
Results
Because we expect the gender of the legislator to matter differently for male versus female citizens, we estimated separate models for men and women. This means that we estimated twelve fixed-effects panel regression models in total (a model for men and women respondents for each of the five dependent variables described above, as well as the placebo variable). Each model included an indicator for whether the individual was evaluating a male or female respondent in that particular time period (2012 or 2014). The use of individual-level fixed effects means that the coefficient we observe for this indicator variable is an estimate of the effect of having a female (rather than male) representative based on the average within-unit change during this period.
Evaluations
The full results from the fixed-effects models can be found in the online appendix. 2 For ease of presentation, we plot the coefficients and 95 percent confidence intervals from these models in Figure 1. The left-hand plot shows the effects from the six models estimated with female respondents, and the right-hand plot shows the estimates from those models run with male respondents. The x-axis is the size of the treatment effect of having a female representative—that is, the direction and magnitude of change in each dependent variable associated with having a female rather than male representative.
Effects of gender on evaluations of House representatives.
There are a few noteworthy patterns in Figure 1. First, having a female representative does not appear to change how close women think the representative is to them on the ideological scale. This is evidenced from the first coefficient in the left-hand plot, which is just slightly negative but not statistically significant. A similar nonresult is evident for men. In this case, the coefficient is slightly positive, but also lacks statistical significance. Thus, the House member’s gender does not appear to be important for influencing perceptions of ideological proximity for citizens of either gender.
A second noteworthy pattern is the extent to which women’s evaluative perceptions of representatives’ character and ability to do their job is influenced by the gender of the representative. Specifically, when women are represented by a female, they perceive their representative to be significantly more competent and have significantly more integrity than when they are represented by a male. Female respondents also express stronger approval and are more likely to say female House members represent people in their district well when they have a female legislator than male legislator. Each of these effects is statistically significant (p < .05), though the magnitude of the effects is somewhat modest. In general, women increase their ratings of female representatives (compared with male representatives) by a magnitude equivalent to about 1/20th of the range of the scale on each measure.
Men, however, appear unresponsive to the gender of their representative. Having a female House member of Congress did not significantly alter perceptions of that representative’s competence, integrity, representation of the district, or their approval of the representative. In each case, the estimated effect was close to zero.
Finally, note that for both men and women respondents, there is no significant effect for having a female representative on approval ratings of Barack Obama. In fact, for female respondents, the estimated effect is almost precisely zero. This lends additional confidence in our assumption that there are no confounding variables that we are not accounting for in our analysis.
Thus, overall, the gender of a representative does appear to matter for women. Women view female members of Congress as more competent, as having more integrity, and as being better representatives, and, accordingly, they tend to approve more of those female legislators. Recall that this is after explicitly controlling for partisanship in our design by only examining cases where the representative changed, but the party of the representative did not. These patterns also arise despite the fact that we find no difference on our placebo test of respondents’ approval ratings of President Obama. Furthermore, these effects do not appear to be caused by women projecting an expectation of receiving better policy representation from female legislators, as evidenced by the fact that they do not perceive female legislators as closer to them ideologically than male legislators.
Contact
Women rate female legislators more highly than they do male officeholders, but does this affect their propensity to actively reach out to their representative? In the 2012 and 2014 preelection waves of the CCES, respondents were asked whether they had any contact with their House member during the previous two years. Respondents could indicate the type of contact they had with their member of Congress, including that they visited the member’s office, called the office, and e-mailed or sent a letter to the member. They could also report ways in which they had been contacted by the legislator, including that they received an e-mail or letter from the legislator, received a call from the legislator, or met the legislator at a public event.
Table 1 shows the percentage of respondents reporting each type of contact in both 2012 and 2014. Not surprisingly, the most common type of contact that constituents have with their representatives is by mail or e-mail. One-in-five respondents reported sending an e-mail or letter to their representative in 2012, and more than one in three reported receiving mail from their representative. Other types of contact are much rarer. For example, only 1 percent of respondents reported visiting their representative’s office, and only 3 to 4 percent called the representative’s office.
Frequency of Reported Contact with House Representatives, 2012 and 2014.
Percentages calculated using sampling weights. N = 9,500 American adults interviewed in 2012 and 2014.
Because we are interested in the extent to which an officeholder’s gender affects a citizen’s propensity to contact that officeholder, we focus only on the top three types of contact listed in Table 1 as those are all citizen-initiated forms of communication. Thus, our measure of contact is whether the respondent indicated that she visited the member’s office, called the member’s office, or e-mailed or sent a letter to the member’s office. Among the individuals we analyze, 28 percent reported contacting their legislator in 2011–2012 compared with 20 percent who said they did so from 2013 to 2014. Note also that women tend to contact their legislators at a somewhat lower rate than men. However, our interest is in whether women contact female legislators more often than those same women contact male legislators.
We are also interested in the extent to which gender affects whether women are satisfied when they do contact their representatives. The survey asked individuals who reported contact with their representatives whether they were satisfied with that contact. Among the sample that we analyze in this paper, more than 70 percent were satisfied with their contact (with roughly similar levels of satisfaction in the 2012 and 2014 waves of the survey). For those who contacted their members in both years, we examine whether a change in the gender of an individual’s representative affects their reported satisfaction when contacting him or her. 3
In addition, we once again include a placebo test in our analysis. This time, our placebo measure is a question asking respondents whether they attended a local political meeting during the previous year. A change in the gender of a respondent’s legislator may affect whether she sought to contact that legislator, but it should not affect whether she engaged in another unrelated activity such as attending a political meeting.
Figure 2 presents a coefficient plot similar to what we showed above. 4 In this case, we simply plot the effects of representative gender from six models—contact and satisfaction with contact for both male and female respondents, as well as the placebo model for both groups. Positive values on the x-axis would indicate that respondents are more likely to contact a female legislator (compared with a male) or more likely to be satisfied when they contact a female representative compared with a male representative.
Effects of gender on contacting representatives and satisfaction with contact.
The effect at the top of the left-hand plot in Figure 2 indicates a statistically significant and negative effect for representative gender on contacting that member. Specifically, the coefficient here is −.077 (p = .007), indicating that female respondents were 7.7 percentage points less likely to contact a female representative than they were to contact a male representative. Given the overall low rate at which constituents contact their representatives, this amounts to a dramatic decrease in the rate at which women contact their legislators. Indeed, the predicted values from the model indicate that 22.4 percent of women make contact if their representative is a male, compared with just 14.7 percent who do so if the representative is a female. Men, by contrast, are again unaffected by the gender of their representative; having a female member of Congress does not statistically significantly affect their propensity to contact their member.
While the gender of a representative does have a dramatic effect on the rate at which women contact their representatives, it does not appear to influence how satisfied they are when they do make such contact. For both men and women, the coefficient for having a female representative is close to zero and lacks statistical significance.
One explanation for this set of findings might be that women see less of a need to contact women representatives. Most people reach out to their members of Congress about policy issues that are important to them. 5 Therefore, representatives that are seen as more competent, as having more integrity, and as representing their districts well may not need to be contacted as frequently.
Finally, the estimated effects for the placebo test are again close to zero and lack statistical significance. While women are significantly less likely to report contacting their legislator when she is a woman, having a female legislator has no effect on whether they report attending local meetings. Thus, we can be all the more confident that the effect we have uncovered here is not caused by spurious factors.
Effects by Partisanship of the Representative
So far, we have shown that women rate female representatives more favorably than male legislators and they are less likely to contact them, while we found no such differences among men. We now turn to examining whether these results are conditional on the legislator’s party. 6 As mentioned above, our analysis controls for the party of the legislator by limiting our focus to cases where the gender of the representative changed but the party did not. Figure 3 reproduces the analysis from Figure 1, but this time separating the effects based on the party of the legislator and the gender of the respondent. 7 This allows us to test our expectation that the legislator’s gender may matter more when the legislator is a Republican rather than a Democrat.
Effects of gender on evaluations of House representatives, by party of legislator.
The plots on the left-hand side of Figure 3 show the results for the tests of whether the gender of the legislator matters when that legislator is a Democrat. Notably, there appears to be little difference in how either male and female respondents evaluate Democratic men compared with Democratic women. Nearly every effect on the left-hand side of the plot is small (less than 5 points) and not statistically significant. These results are in line with previous research that finds that people expect both male and female Democrats to emphasize social welfare issues and to have high integrity as elected officials (Schneider and Bos 2016).
The right-hand plots in Figure 3 tell a different story. The top right plot, in particular, shows that women evaluate Republican female representatives quite differently than they evaluate Republican male legislators. Specifically, women rate female Republican legislators as being more competent, having more integrity, representing the district better, and they approve more highly of them. For each of these measures, the difference amounts to about one-tenth of the total range of the scale—a substantively meaningful effect. Yet, turning to the bottom-right plot, there is only one significant difference in how male respondents rate female Republicans compared with male Republicans—they see the female Republicans as farther away ideologically (from themselves) than male Republicans. On every other measure, male respondents rate female Republicans roughly the same as they rate male Republicans.
Why does the gender of Republican legislators matter for women, but not for men? Recall that people tend to believe that Democrats care more about feminist and social welfare issues than Republicans (Schneider and Bos 2016). Yet, at the same time, people tend to believe that female legislators care more about these same issues and, indeed, this expectation matches the reality of women’s legislative activity in Congress (Swers 2013). Women tend to care more about feminist and social welfare issues, and they are more likely to expect female legislators (but not male legislators) to share those interests. For many men, this emphasis on feminist and social welfare issues is less important, and, thus, the gender of the Republican legislator would matter less to them. This is precisely the pattern we find in Figure 3.
Figure 4 recreates the results from Figure 2 but again separating the analysis by respondent gender and by the party of the legislator. Consistent with what we find in the pooled analysis, men are no more nor less likely to contact their representative or be satisfied with that contact if the legislator is male or female. Also consistent with the pooled analysis, women are less likely to contact their female representatives than their male representatives, but this gender difference only persists for Democratic legislators. This finding supports our theory that if women trust that their interests are more often prioritized by female legislators, they may see less of a need to contact them. In line with the above discussion, it may be that in the case of Democratic female representatives, the party and gender cues send reinforcing signals, reducing female constituents’ perceived need to contact the legislator. For a Republican female legislator, the signals are conflicting; thus, even though women may rate the female Republican legislator higher than the male Republican legislator, the fact that the party and gender cues conflict may mean they still feel the same need to contact that representative to express their views.
Effects of gender on contacting representatives and satisfaction with contact, by party of legislator.
Effects by Partisanship of the Constituents
We might also expect that such ratings would be related to the distinct ways in which Democrats and Republicans in the public view women in politics. To illustrate the basis for this possibility, Figure 5 presents data from a Pew Research Center survey conducted in 2014. Respondents to the survey were asked whether they thought “men generally make better political leaders than women,” “women generally make better political leaders than men,” or “women and men make equally good political leaders.” Figure 5 shows the percentage of respondents choosing one of the first two statements as a function of their partisanship and gender.
Attitudes on gender and political leadership, by gender and partisanship.
Notably, Figure 5 demonstrates that the evaluation of women versus men as political leaders is a function of both gender and partisanship. For example, Democrats are more likely than Republicans to see women as better political leaders and less likely than Republicans to think that men are superior, even after accounting for gender. Accordingly, we might expect Democratic women to be more favorably inclined toward female elected officials than Republican women, and we might likewise expect Democratic men to evaluate female legislators more favorably than Republican men.
To test this possibility, Figure 6 shows separate estimates for respondents based on both their gender and their partisanship. 8 Notably, the results among women largely hold across party identification. Both Democratic women and Republican women tend to evaluate female legislators more favorably than they evaluate male elected officials. While these effects are not always statistically significant, they are all consistent with the findings presented above for all women. This suggests that women do, in fact, tend to have more favorable views of female legislators regardless of their party.
Effects of gender on evaluations of House representatives, by party of respondent.
Such consistency is not found among men, however. Democratic men evaluated female legislators as having more integrity and as representing the district better, and they also approved more of those female legislators compared with when they were represented by a male. The effects for Republican men are precisely the opposite. Republican men are the one group who provide lower evaluations of female elected officials. It is also worth noting that Republican men are alone in their propensity to view female legislators farther from them ideologically than men. In short, Republican men do not react favorably to being represented by a female legislator, even when the partisanship of their representative does not change.
Finally, Figure 7 reproduces our analysis of the contact variables, but accounting for the partisanship of the respondents. Note that the reduced propensity of women to contact female legislators is largely limited to Democratic women in our sample; the effects are much smaller and not significant for Republican women. Among male respondents, our results are again in opposite directions. Male Republican respondents were 8.6 points more likely to contact female legislators (p = .06) while the treatment effect was negative, but not significant, for Democratic men.
Effects of gender on contacting representatives and satisfaction with contact, by party of respondent.
Conclusion
Scholars, activists, and policy makers alike have advanced arguments for increasing the number of women in office. Descriptive representation is thought to foster many benefits for marginalized groups, including substantive representation on public policy and promoting trust in government. Several different threads of literature suggest that the dyadic relationship between a representative and her constituent should be improved if they share certain characteristics, especially gender (e.g., Carroll 2002; Mansbridge 1999; Swers 2002, 2005). Because communication with constituents is a key element of representation and a daily practice of elected officials, we may expect that shared gender would influence constituents, especially women, to contact their members of Congress. To be sure, some research posits that women’s presence in political office should enhance women’s political participation (see, for example, Atkeson and Carrillo 2007). Moreover, based on previous literature that suggests women feel more positive about government when more women are in office (Burns, Schlozman, and Verba 2001; Hansen 1997; Reingold 2008), women may also express a higher level of satisfaction with communication that is with a female representative.
The findings presented in this paper challenge such expectations. We take advantage of a large-N panel survey situated around the exogenous shock of redistricting to conduct a strong causal test of how a change in the gender of a representative affects evaluations of a representative and communication patterns between constituents and representatives. While women were more likely to evaluate female representatives more favorably than male representatives, they were not more likely to initiate communication with female representatives than when they were represented by men. Of course, many factors could be driving this pattern. Women representatives may have different experiences, qualifications, issue positions, committee assignments, and patterns of outreach to constituents that could affect constituent evaluations and contact. We hope that future research will do more to explore the causal mechanisms that drive our findings.
Our study contributes to the literature in a number of ways. Ultimately, we show that descriptive representation does matter for communication between women and their representatives, however, not necessarily how we might expect. Women tend to view female legislators as more competent and as having more integrity than male legislators and believe they do a better job at representing their district. This may explain why they do not necessarily feel a strong impetus to contact those representatives at a higher rate. Notably, these findings suggest that female elected officials may actually enjoy some (minor) advantages over males in terms of their standing among constituents. After all, male constituents in the aggregate do not penalize female legislators in their evaluations while women express higher evaluations for female legislators.
A second contribution of the paper is methodological. The observational nature of most studies surrounding women’s representation makes it difficult to tease out the causal impact of having a woman as a representative. Our panel approach provides a much better estimate of the causal effect of a representative’s gender by examining how individuals’ evaluations and levels of participation change when the gender of their representative changes for largely exogenous reasons (i.e., the redistricting process). This individual fixed-effects estimate accounts for time-invariant differences across individuals, greatly reducing the number of confounding factors that challenge causal inferences in cross-sectional estimations. While it is still possible that there are time-variant factors that differentiate individuals who experienced a change in the gender of the representative from those who did not, this possibility is less likely. Thus, the effects we identify in this paper—that women evaluate female representatives more favorably and contact them less frequently—are more likely to approximate the true causal effects of having a female (rather than male) officeholder.
Finally, our paper helps to provide important context to understanding how constituents evaluate politicians based on their gender. Specifically, when we examine our effects by partisanship, we find that it is the interaction of gender and partisanship that explains how citizens evaluate female versus male politicians. Indeed, Democratic men joined women from both parties in providing more favorable evaluations of female legislators, but Republican men departed from this pattern and provided less favorable evaluations of representatives who were women. In the wake of an election that saw presidential candidate Hillary Clinton struggle to overcome very negative ratings, particularly from more conservative men, and in which Carly Fiorina struggled to overcome outright sexist attacks during the primary campaign, our findings help to illuminate the segment of the population where female politicians still struggle to gain parity with their male counterparts. It is noteworthy that the number of Democratic women in Congress has continued to climb during the past few decades while the number of Republican women in Congress has not. 9 More research should be done to parse out the effects we find by constituents’ party affiliations and the party of their representatives, but our findings are nevertheless suggestive—Republican men, who make up a majority of Republicans in the electorate, hold less favorable views of female elected officials. Their negative views of female politicians may be holding Republican women back in terms of seeking, winning, and retaining positions in Congress. And this is increasingly leading the Republican Party to become a mostly male caucus.
Footnotes
Acknowledgements
We would like to thank Sara Angevine, Georgia Duerst-Lahti, and Wendy Schiller for feedback on earlier versions of this paper.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Science Foundation, which funded this research (Awards 1154420 and 1430473).
