Listening and responding to citizens are central to the function of representative democracy. Candidates run on platforms that appeal to the issues and needs of citizens, and many expect elected officials to represent and respond to citizens’ priorities and needs. Various theories of democracy require that public officials and public policy respond to constituents’ concerns and perspectives (for example, see Pitkin 1967; Dahl 1971; Esaiasson and Wlezien 2017). However, in actual democratic governing, elected officials’ responsiveness to their constituents is often neither equitable nor random. This can have dramatic effects on who is represented, who informs policy, who feels inclined to reach out to elected officials, and even who chooses to run for office. These consequences, if they exist, are especially important at lower levels of government, where citizens are arguably in the best position to influence and correspond with elected officials and where the actions of citizens and local interest groups can powerfully shape agendas and institutions (e.g., Dynes, Hartney, and Hayes 2021; Anzia 2022).

Based on existing studies and theories about intersectionality, we examine elected officials’ responsiveness and propose that the combination of the identity of the constituent, the identity of the elected official, and the substance of the constituents’ requests strongly influences responsiveness. Using a large-scale (N = 23,738) audit study of state, county, and local elected officials, we collect data on elected officials’ responsiveness to constituent requests along two behavioral measures - if officials open and reply to constituents’ emails. We confirm many of the same basic inequalities in responsiveness along the lines of race and gender that have been observed by others, and going beyond existing studies, we find that Black men are systematically ignored by elected officials– even more than Black or White women. This happens irrespective of the kinds of messages that Black men send to elected officials. This emphasizes the importance of constituents’ identities when understanding responsiveness from elected officials. In contrast, we observe inequities for Black women and White women conditional on if they bring up race or gender directly. Our findings may appear to contrast with some of the core expectations of original articulations of intersectionality that would predict compounded disadvantages for Black women; we discuss how recent trends in the political involvement and representation of Black men and women contextualize this contrast in ways supported by different intersectional theories of politics. The consistent ignoring of Black men and conditional responsiveness for women indicate the importance of an intersectional approach and that inequities in responsiveness from government are often a function of combinations of identities and messages. Black men’s voices are consistently ignored more than other constituents. The implications for current politics are manifold: it deepens racial inequalities in representation, can run roughshod over ideas that Black men bring to the fore, and may have cascading effects on political participation and engagement among Black men.

Testing Intersectionality in Responsiveness

Scholars concerned about responsiveness have documented that elected officials’ understanding of and responsiveness to their constituents is not evenly distributed. In the United States, for example, elected officials and their staff are more likely to respond to and meet with donors than other constituents (Kalla and Broockman 2016) and more likely to discount the views of constituents who disagree with them (Butler and Dynes 2016). In addition, elites respond unevenly depending on constituents’ identities, such as their racial, ethnic, and gender groups (Butler and Broockman 2011; Costa 2017; Magni and Ponce de Leon 2020; Kalla, Rosenbluth and Teele 2018).

Existing research answers some questions about listening to constituents while leaving others unresolved, often isolating only one identity. One of the initial experiments in this area, for example, varied the race of two masculine constituents (Butler and Broockman 2011). This suggests that researchers understand how responsiveness varies by race among men but not responsiveness by race among women or across genders. Studies of gender, on the other hand, do the opposite (e.g., Magni and Ponce de Leon 2020). While one notable study does include variation in race and ethnicity while studying gender, the focus of this research was on gender and the variation of ethnicity was limited and not a large component of the study (Kalla, Rosenbluth, and Teele 2018).Footnote 1

There are other limits to these studies as well. Some work considers how different types of requests promote more or less engagement from political officials, generally focusing on service and policy-focused requests with some variation across studies about which messages prompt more responsiveness (Butler, Karpowitz, and Pope 2012; Costa 2017). Few studies, if any, evaluate the identity of the elected official or the alignment of that identity with constituents’ identities.Footnote 2 Yet to date, no study of responsiveness has brought together the messenger (constituent), the message (the constituents’ request or inquiry), and the receiver (the elected official).

To build on this literature, we apply an intersectional approach on the three different levels mentioned in the previous paragraph. These are the level of the constituent, the content of constituents’ messages, and the constituent-official dyad. We note that this is only one way to think about intersectionality in this context - there are many other different ways to take an intersectional lens to this topic that would also yield important insights.

Intersectional Constituent Identity

Intersectional approaches propose that individuals’ marginalization depends on the intersection of their identity groups. Crucially, even groups that seek to address social inequalities often emphasize the goals of specific subgroups rather than the larger group as a whole (Crenshaw 1989, 1991; Strolovitch 2007). One common example is how some feminist movements prioritize the interests, needs, and perspectives of White women, often to the exclusion of women of color (Mcintosh and Barrett 1985; Crenshaw 2013), and how antiracist movements emphasize the interests of men of color to the detriment of women of color (Crenshaw 1989, 1991). Much of the explicit motivation of these early articulations of intersectionality specifically explored the position of Black women and the compounded and complex disadvantages they faced legally, politically, and socially due to their specific status as women who are Black (Crenshaw 1989, 1991).

At the same time, social and political context determine the specific identity intersection that is most disadvantaged and marginalized. Considering which group faces the most marginalization therefore requires an understanding of how identities combine and interact within a given situation (e.g., Hancock 2007; Dhamoon 2011). Some have summarized this perspective by stating that as an approach, intersectionality is “interested in how the differential situatedness of different social agents affects the ways they affect and are affected by different social, economic and political projects” (Yuval-Davis 2011, p. 4) and positionalities “tend to be different in different historical contexts and are often fluid and contested“(Yuval-Davis 2006, p. 199). Put more succinctly, “There is no neat hierarchy of advantage or disadvantage” (Cassese 2019, p. 11). As a concrete example, in the domain of representation, some explore the degree to which elected officials who are women of color face compound forms of marginalization and can at the same time use their combined racial and gender identities to act as effective coalition builders and advocates for a range of marginalized and vulnerable groups (Dawson 2001; Philpot and Walton 2007, Fraga et al. 2008; Brown and Banks 2014, Shah, Scott, and Juenke 2019, Reingold, Widner, and Harmon 2020). This does not suggest that officials of color who are women face no significant disadvantages, just that those disadvantages can, in some circumstances, be accompanied by different advantages. Abandoning a single-axis approach to marginalization (Crenshaw 1989) thus opens up many ways to thoroughly consider how the combination of different identity groups and political environments shape the form that inequities and power take in society. In this way, intersectionality shows that the manifestations of marginalization and privilege fluctuate based upon combinations of identity and social contexts. This calls for deeper, careful analysis of the functions of identity and context together, which we apply to representation next.

When elected officials interact with constituents, their behavior is undoubtedly determined by identity across multiple dimensions, rather than one type in isolation (see Kalla, Rosenbluth, and Teele 2018). We focus on the intersection of the two critical forms of identity in American politics, race and gender as they relate to responsiveness and representation from elected officials. Because of their chronic salience to American politics, they are likely some of the most consequential forms of identity for elites in responding to their constituents.Footnote 3

Intersectional theory points to places where existing research findings may overlook important patterns in listening and responsiveness. We first, then, outline our expectations based on previous work and then how intersectional theories shift these predictions. Published studies find a consistent pattern of lower responsiveness for racial and ethnic minorities (e.g., Butler and Broockman 2011) and that these patterns seem to be produced by racial bias rather than strategic, electoral considerations (Butler 2014; Costa 2017). From this research, we would expect to find lower levels of responsiveness towards racial minority groups (H1).

In a similar way, work on gender finds evidence of a mild, pro-woman effect where elected officials respond more to women than to men (e.g., Kalla, Rosenbluth, and Teele 2018; Rhinehart 2020; Magni and Ponce de Leon 2020). Building on this work, and ignoring combinations of identity, we expect to find higher levels of responsiveness towards women (H2). On its face, this expectation may appear counter to women’s position of marginality in American society in a number of areas– for example, we would expect women to receive less responsiveness in the context of job markets (e.g., Mishel 2016). However, in the specific context of politics and engagement with elected officials, women appear to have a mild advantage in the form of higher levels of responsiveness.Footnote 4

We use one intersectional perspective to weave together H1 and H2 to make predictions about the combined gender and racial identities of constituents. In so doing, we consider both the structural, social, and historical marginalization groups face and the way those disadvantages connect to the specific context of responsiveness. As noted earlier, this is only one way to apply intersectionality and other ways of intersectional thinking might lead to different valuable expectations.

Our first intersectional prediction is that we expect that the lowest levels of responsiveness should be for Black men (H3). Initially, this may seem to contradict some of the foundational articulations of intersectional theory, which suggest that Black women should typically experience more marginalization than Black men, White women, and White men (Crenshaw 1989, 1991). However, we suggest that in the contemporary political context and in the domain of elected officials’ responsiveness, distinct political and social stereotypes of Black men should lead to the lowest levels of responsiveness from elected officials to that group. This comports with prominent intersectional theories that require an exploration of identities, their combinations, and the specific social and political circumstances in the analysis at hand. This way of thinking about intersectionality - as linked to specific circumstances and environments - has become increasingly important as applications of intersectionality have expanded beyond the case of the United States and the case of Black women used by Crenshaw and others (Carbado, Crenshaw, Mays, and Tomlinson 2013). This application of intersectionality also implies that our expectations about the groups most marginalized can shift dramatically with the context. In a different domain - such as in workplace discrimination - or with a different element of politics - such as in candidate evaluations - we would have very different expectations about the relative marginalization faced by Black men, Black women, and other identity groups. In that sense, we present this as an intersectional hypothesis for these circumstances, not as the only way to apply or understand intersectionality in this area.

We expect the lowest responsiveness to Black men because across combinations of race and gender, Black men are the group often viewed as most interpersonally threatening and least engaged in politics of the four racial, gendered groups we consider (Trawalter, Baird, and Richeson 2008; Plant, Golpen, and Kunstman 2011). This should therefore prompt more distrust and discrimination of this group (Mendes et al. 2002; Richeson and Bean 2011) and corresponding disregard by elected officials.

Data from the U.S. Census suggests that Black men have, historically, been the least likely to vote compared to White men, White women, and Black women (e.g., Census Bureau 2022).Footnote 5 Recent and established work finds Black women are engaged in various forms of political participation and engagement more than Black men (Darcy and Hadley 1988; Farris and Holman 2014; Ezie 2021), and contemporary conversations in the media and public mirror these findings (Herdon 2020). Political representation of Black women and Black men in government bodies also reinforces the voting and participatory differences between these groups, with much of the recent gains in descriptive representation for people of color coming from increased electoral success of women of color and Black women in particular (Hardy-Fanta, Lein, Pinderhughes, and Sierra 2016, Reingold 2019). To summarize these recent patterns, the gender gap between officials of color is actually smaller than for White officials - in the 118 th US Congress, for example, 42% of officials of color are women compared to only 23% of White officials (Schaeffer 2023; Congressional Research Service 2024). This is not to state that Black women and White women do not experience barriers to participation or representation in politics– there is clear evidence that they do. Crucially however, our expectation focuses specifically on the way that Black men are or are not listened to by elected officials.

In this area, there are a number of other stereotypes and narratives about Black men specifically that encourage low levels of responsiveness. Black men are often stereotyped as as criminal, violent, and immoral more than other men and Black women (Hurwitz and Peffley 1997; Kelly 2007; Skorinko and Spellman 2013; Avery, Oh, and Cooper 2021), which could also motivate a lack of engagement by elected officials with Black men in politics.Footnote 6 Black men are linked, statistically and in shared social stereotypes, to disproportionately high rates of incarceration and subsequent disenfranchisement (Pettit 2012; Uggen, Larson, Shannon, and Pulido-Nava 2020). This in turn, promotes lower levels of political participation for Black men (Burch 2013; King and Erickson 2016, White 2022). This goes above and beyond the disproportionate rates of incarceration among Black women - since 2000, for example, rates of incarceration among Black women have fallen at about twice the pace as that of Black men and the rate for Black men continues to dwarf that of Black women and other racial and gender groups (Norwood 2023).Footnote 7 To be clear, this does not imply that Black women do not face their own distinct forms of marginalization when it comes to representation and responsiveness. They undoubtedly do. In our cases here, we expect responsiveness to Black women to be lower than that of White men and White women as explained below. Our claim is also not that the sum total of marginalization experiences faced by Black women are smaller than those faced by Black men (in this, we agree with similar concerns by intersectional researchers - see Carbado, Crenshaw, Mays, and Tomlinson 2013). Our prediction is contextually-oriented by the specific case of how a set of raced and gendered perceptions can determine levels of responsiveness.

With this context in mind, we expect the highest level of responsiveness to be towards White men (H4). This is for many reasons. White men are the best represented (of the four groups we consider here) among elected officials in general (Butler and Broockman 2011; Schaffner, Rhodes, and La Raja 2020; Schaeffer 2023), receiving more descriptive representation. In addition, although a gender gap exists with respect to voting, this gap is significantly smaller for Whites than other racial groups - only 1–3% for Whites compared to 8–10% for Blacks, for example (See Igielnik 2020; Center for American Women and Politics 2023). This combined with the continued connections between politics and masculinity (in political stereotypes and norms of politics; see Schneider and Bos 2019; Winfrey and Schnoebelen 2019) makes White men a continually important political constituency for most elected officials.

The contrary patterns in the previous two hypotheses - about White and Black men - could lead to H2, where women experience more responsiveness than men. This would create an example of Simpson’s paradox (Wagner 1982; Kievit et al. 2013). Specifically, because existing research has not considered the intersection of different identities, the overall patterns observed in the extant literature may obscure important and conflicting patterns among men of different racial groups. The penalty for Black men in responsiveness may generate what appears to be an overall penalty for men, masking the advantage of White men.

We also suggest something similar regarding White women and Black women. Based on the political participation of Whites and women and the absence of negative racial stereotypes of White women, we expect that responsiveness will be higher for White women than Black men or Black women but lower than White men (H5). For Black women, however, we expect that responsiveness will be higher than that of Black men while still lower than White men and women given the discussion earlier (H6). We expect this to be produced by a number of factors, including patterns of political participation by race, the share of the population held by Black vs. White women, and negative stereotypes of Blacks as a racial group. As with the discussion of responsiveness to Black and White men, failing to recognize the different racial dynamics (e.g., the position of Black women compared to White women and Black women to Black men) could produce the overall findings in the literature summarized in H1 and H2. We summarize these intersectional predictions in Table 1.

Table 1 Constituent identity, intersectional predictions
Full size table

Intersectional Message Content

The next layer of our application of intersectional theory involves messages about identity. Some research finds elected officials respond more to service than policy contacts (Butler, Karpowitz, and Pope 2012). Meta-analyses of this same work find that responsiveness does not differ between policy and service requests (Costa 2017). So far, these studies have not explored how elite responses change with qualitatively different topics.

We propose that the issue constituents inquire about can further inequities in responsiveness because of how it relates to constituents’ identities. Issues that evoke identity will shape how the constituent is perceived, specifically those that connect to the constituents’ identity. Research shows support for this possibility, finding that many Whites want to avoid discussing race when having interracial experiences (Appiah et al. 2021), that Whites distance themselves more from Blacks when race is made salient (Goff, Steele, and Davies 2008), and that Whites think differently when anticipating an interracial (as opposed to an intraracial) conversation about race (Sommers, Warp, Mahoney 2008). Other research on different topics supports this general pattern of messenger-message interaction: for example, men who discuss gender-focused topics influence women differently than other women (Moser and Branscombe 2021).

Drawing from these perspectives, we expect that elected officials will, on average, be less likely to listen to constituents from marginalized gender and racial groups when they reference topics about their identity groups (H7). We expect this to be produced by a hesitancy among many elites to avoid identity-focused topics with marginalized constituents and a lack of such hesitancy when engaging with non-marginalized constituents. In the context of intersectionality, this means that marginalized groups face an additional penalty when advocating for topics directly relevant to their groups. There are, potentially, a number of interesting interactive hypotheses between this prediction and the predictions discussed in the prior section - for example, do Black men face an especially large penalty when talking about race? What about Black women discussing gender or race? As we lack much established research about how identity combines with message content, we evaluate these possibilities in a more exploratory way (rather than testing a string of additional hypotheses).

Intersectional Official Identity

The final element of our connection between responsiveness and intersectionality emphasizes the relational role of intersectionality and responsiveness. It involves two individuals. Both have identities that influence power dynamics, marginalization, and levels of responsiveness (e.g., Crenshaw 1989; Strolovitch 2007). Despite this, most existing studies in the area of elite responsiveness do not consider the identity of the official; those that do often reach conflicting findings and consider only one type of identity at once (see Butler and Broockman 2011; Broockman 2013; Kalla, Rosenbluth, and Teele 2018).

Building on intersectional theory, we anticipate that elected officials from advantaged groups will be least responsive to marginalized groups (e.g., women, constituents of color, women of color) who send group-relevant messages (H8). This is because these constituents are less powerful politically and because these group-relevant messages are the most challenging to officials from advantaged groups. We focus specifically on elected officials who are both White and men, as they come from the politically more advantaged combination of racial and gender groups (White men) as discussed by intersectional and other theoretical approaches to political science.

As in the previous section, there are a number of ways that the official identity can intersect with constituent identity and the message content. How might elected officials who are White respond to constituents who are Black men and talk about race? How might that compare to the responses of elected officials who are people of color and male? This leads to an array of different combinations - with 48 distinct groupings based on elected official identity, constituent identity, and message topic.Footnote 8 Our intersectional theory suggests that these different combinations are important, and we expect differences in responses to constituents and message topics based on the intersectional identities of the officials. At the same time, we lack clear theoretical and empirical guidance about the precise way each of these groupings should compare to other groupings. We therefore consider specific differences in a more exploratory way rather than providing an expectation for each distinct official-constituent-message combination.

Design and Procedure

We test these hypotheses through a large audit study on local elected officials conducted early in 2021. Prior to data collection, our team gathered every publicly available email for elected officials at the state, county, and local level to evaluate patterns of responsiveness in a wide and generalizable way. However, we excluded state legislators from our sample due to the large number of experimental studies that have already been conducted with those elected officials. This information was gathered through Google’s Civic Information API as well as supplemented by our team of research assistants (see Online Appendix for more details).

This process created a dataset with 23,738 elected officials, making it one of the largest audits to date in political science. Sample sizes in other audit studies of elected officials generally range from about 1,500 to 11,000 (e.g., Butler and Broockman 2011; Butler and Crabtree 2017; Kalla, Rosenbluth and Teele 2018). We sampled a much larger group of elected officials due to the nature of our predictions, which consider the crossing of racial and gender categories, add the additional layer of message content, and propose expectations about different kinds of elected officials. This design creates 12 different treatment conditions and subgroup analyses for elected officials who are White men, men of color, White women, and women of color. As a result, we create estimates for a minimum of 60 different groups (12 conditions overall and 48 conditions for each of the four elected official subgroups). The complexity of this study and number of comparisons lead us to select such a large sample. With these elected officials, we estimated the race and gender of each elected official based on their name (using the gender and wru R packages) and verified this automatic coding with human searches done by research assistants.Footnote 9

The officials described in Table 2 vary greatly in the offices they hold. This group includes city council members, county attorneys, county council members, sheriffs, and more. As noted by Table 2, the sample includes a large proportion of county-level officials, some of whom may not regularly receive many constituent emails.Footnote 10 Regardless of the office they hold, however, we suggest that our standard of representation - listening to and acknowledging the concerns of constituents - should apply. This kind of responsiveness can operate anywhere from a complete, robust answer to an inquiry from a constituent to the elected official forwarding the inquiry to someone who they see to be more qualified in providing a response. Both of these ends of the spectrum count, in our view, as elite responsiveness.

Table 2 Sample composition
Full size table

We estimate that our sample contains more men than women (63.5% men) and Whites than people of color (60.2% White). With respect to gender, the composition of our sample is similar to other studies of local and state elected officials (Lee, Adair, and Mendes 2020; Cagaanan, Dean, and Lee 2022; Center for American Women and Politics 2022). In the area of race, we document a lower percentage of Whites than other studies– some, for example, contain samples that are more than 70 or 80% White (Butler and Broockman 2011; Schaffner, Rhodes, and La Raja 2020). We note, however, that our sample is not directly comparable to these others as our sample contains officials at three different levels (state, county, and local) and from a wide range of offices (city councils, executives, sheriffs, county attorneys, etc.). Because of our treatment randomization, the specific composition of our sample does not interfere with the causal inferences we draw from our study.

We created four different email addresses that correspond to different identity combinations. The names used in these email addresses are shown in Table 3. These names are based on Census data,Footnote 11 other studies (Druckman and Shafranek 2020), research on first names (Tzioumis 2018), and work on how the combination of first and last names influences racial perceptions (Gaddis 2017).Footnote 12

Table 3 Constituent names
Full size table

In line with our expectations, we also varied the nature of the constituent’s message. By this, we mean that we drafted emails that, although about the same broad policy area, expressed different racial and gender concerns that intersect differently with the senders’ and receivers’ identities. We sent elected officials messages about crime that: explicitly discussed racial groups (using terms like Black directly), implicitly mentioned racial groups (through the use of coded language like “inner-city”), or brought up gender. Our focus on these two kinds of racial messaging is motivated by studies that find different reactions to explicit and implicit racialized statements generally (e.g., Mendelberg 2001; Banks and Hicks 2019, Thompson and Busby 2023; cf. Valentino, Neuner, and Vandenbroek 2018).

Table A.1 of the Online Appendix provides the exact text of these emails and their subject lines. The messages and subject lines all raise concerns about the topic of crime and reference a claim a constituent heard about that topic. The constituents indicate that they live in the same area as the official, have heard some information about crime, and are looking for additional information on the topic. The gender message brings up concerns about women in crime, the implicit racial message asks questions about the amount of crime committed by those in inner-city areas, and the explicit racial message asks about crimes committed by “Black people”. We vary the racial content in the explicit and implicit racial conditions this way to follow studies of racial messaging, which emphasize the distinction between indirectly and explicitly mentioning specific racial groups (e.g., Mendelberg 2001; Slocum 2001; Hurwitz and Peffley 2005, Valentino, Neuner, and Vandenbroek 2018, Thompson and Busby 2023).Footnote 13 The subject line also contained the gendered, implicitly racialized, and explicitly racialized message to ensure that the treatment is provided even to subjects who do not open the emails. As discussed in the following paragraphs, the email addresses used to send the emails contained the racialized names presented in Table 3. These two pieces of information - the sender email and the subject line - present the treatments to all respondents prior to opening the emails. The combination of the identities in Table 3 and the messages allow us to consider both constituent identity and message content in an intersectional way.Footnote 14 We also include typos and grammatical errors within the body of all of the email in order to add realism.Footnote 15

We identify 13,139 distinct government-official combinations in our sample. These are areas that have multiple officials in our sample of the same local government organization (for example, multiple city councilors in the same city). We assign all members of these groups to the same cluster and randomly assigned each of these 13,139 official-office clusters to a treatment condition and gave all elected officials in those clusters the same treatment assignment. We use this randomization to avoid treatment spillover across respondents, especially those who may regularly interact with each other.

We created a series of email addresses with the names from Table 3 and a string of numbers to administer this experiment (e.g., [email protected]). We assigned one condition to each combination of identities and messages which resulted in twelve different experimental conditions. The emails were sent directly to each elected official and mentioned that the writer of the email had reached out to various elected officials to avoid causing suspicion when elected officials within the same cluster got the same email. We sent many hundreds of emails each day from each condition,Footnote 16 and we sent the emails between February 16, 2021 to March 15, 2021. Emails were sent only on weekdays and during normal business hours. We waited two weeks after sending the last set of emails to record the elected officials’ responses as either a response or non-response.Footnote 17 Officials who responded after the data collection period were classified as unresponsive.

We acknowledge, as others working in this area have, that elected officials’ staff may be reading and responding to these messages. With our focus on county and city-level officials, most are unlikely to have many, if any, staff members. Further, staff members operate at the instructions of elected officials and are subject to their approval; as such, public officials are ultimately responsible for the things their staff members do. While we acknowledge that the unit of analysis is most accurately the email address of elected officials, the findings we observe still have important implications for responsiveness in American democracy. As a result, we use the term “elected official” as shorthand for the office of an elected official.

Using email tracking softwareFootnote 18 and research assistants, we recorded two pieces of information for each elected official - if that person (1) opened and (2) replied to the email message. Both variables are important and related to responsiveness in different ways. Replying is the more direct measure of responsiveness and matches closely what most other studies of elected officials in this area have used (see Table A.26 of the Online Appendix for more discussion of the measures in other audit studies). On the other hand, opening an email, in and of itself, does not similarly qualify as responsiveness. It does not indicate more than the barest amount of attention and engagement and involves no direct interaction with a constituent. At the same time, opening is a prerequisite to further forms of engagement with constituents, indicates at least a passive reception of the constituent contact, and can represent a significant breakdown in the process of responsiveness - that of failing to listen to or ignoring a constituent entirely. We therefore consider both opening and replying as different aspects of elected officials’ behavior that connect theoretically and practically to responsiveness. Examining both also helps to illustrate where breakdowns in larger responsiveness occur - in a failure to listen at all or a hesitation to interact directly with constituents.

We recorded any reply from elected officials, whether or not the official actually addressed the question posited in the original email. Emails that were simply forwarded to another official without an initial response were classified as opening the email but not responding. Emails that generated automatic replies from the server or the organization were not counted as responsive unless the official followed up with a direct response. Furthermore, we removed emails that were bounced back from the server because the official did not have the opportunity to read or respond to them. The bounceback rates averaged at about 11% for each condition. Additionally, because we gathered information from Google’s API, some emails were sent to officials who no longer held office. These officials were removed from the dataset after we received a response notifying us of their status.

Prior to conducting our research, we worked with our institutional review board to ensure that we correctly balanced our research goals with all of the ethical considerations inherent to these kinds of studies. These concerns typically focus on not obtaining informed consent from elected officials and protecting constituent-official trust and relationships (see Online Appendix for more discussion of these ethical issues).Footnote 19

Data and Results

Figure 1 presents bar graphs showing the average rate of replying and opening in each of the twelve treatment conditions. These graphs indicate that reply rates were much lower than open rates and that there are significant differences across the treatments. Replies were lowest in the explicit racial conditions, with reply rates around four to five(percent). The one exception to this is the implicit racial condition for constituents with Black, masculine names, which is slightly lower than the corresponding response for messages explicitly referencing racial groups.

Fig. 1
figure 1

Opening and replying by condition

Full size image

Figure 1, however, does not account for either the structure in our data or that we randomized our treatments at the level of the cluster. In our more formal analyses of both opening and replying, we use linear regression with clustered standard errors and random effects for each cluster to test our hypotheses. We use this specific method in keeping with best practices using this kind of data (see Green and Vavreck 2008; Arceneaux and Nickerson 2009; and Golima 2020). Standard errors are clustered on the level of the official-office clusters we created for randomization. We consider alternative methods of analyses that operate at the level of the cluster (rather than the elected official) and observe the same patterns as shown in the figures that follow (see Online Appendix figures A.1 through A.3 for more details).

We begin by testing H1 and H2, our predictions about replicating patterns in previous work with respect to separately considering constituents’ race and gender. Recall again that this information, which is connected to the name of the sender, was visible to all officials prior to deciding to open or reply to the emails. Figures 2 and 3 present the treatment effects between the gender and racial conditions, accounting for the structure of the experiment and randomization.(see Online Appendix for detailed models).

Fig. 2
figure 2

Opening by race and gender. Note: The baseline for the feminine treatment is the masculine-sounding name treatment and for the Black treatment is the White-sounding name treatment. This figure shows the results of these comparisons for opening constituents’ emails, along with corresponding 95% confidence intervals

Full size image
Fig. 3
figure 3

Replying by race and gender. Note: The baseline for the feminine treatment is the masculine-sounding name treatment and for the Black treatment is the White-sounding name treatment. This figure shows the results of these comparisons for replying to constituents’ emails, along with corresponding 95% confidence interval

Full size image

Our results provide consistent support for H1, as elected officials open and reply less to messages from Black than White constituents. We confirm H2 only for replies, as elected officials open emails from men and women at about the same rate but reply more often to women. This affirms the importance of constituent identity in responsiveness, along the lines of what prior studies have previously found.

Next, we examine H3, H4, H5, and H6, our expectations about constituent identity combinations. At this stage, we collapse across the different messages and focus on constituent identity. We find broad support for the idea that identities and the combination of identities are drivers of inequities in responsiveness. We confirm H3 and H4, shown in Fig. 4. No group receives higher listening and responsiveness than White men. The messages from Black men, in contrast, experience the least responsiveness from elected officials, both in whether the message is opened and replied to. Compared to White men, officials are 9.9 percentage points less likely to open emails from Black men (p < 0.001) and about 2.9 percentage points less likely to reply (p < 0.001). Responsiveness to Black men is also lower than to Black women or White women. Messages from Black and White women are opened less frequently than those from White men (3.9 percentage points, p = 0.01; and 5.7 percentage points, p = 0.01, respectively), but do not garner fewer replies than those sent by White men. This is partial support for H5 and H6. These findings demonstrate the importance of an intersectional lens in understanding responsiveness - the disadvantages faced by Black constituents, for example, depend on gender. Further, other groups (both Black and White women) face barriers to representation in being initially listened to but not in replies.

Fig. 4
figure 4

Responsiveness by constituent identity. Note: The baseline in each of these graphs is responsiveness to emails with a White, masculine sounding name. This figure shows the results of these comparisons for both opening and replying to constituents’ emails, along with corresponding 95% confidence intervals

Full size image

We now consider the content of constituents’ requests (H7), the idea that marginalized constituents who bring up identity-relevant topics will face lower levels of responsiveness (the exact text of these messages can be found in Table A.1 of the Online Appendix). As noted before, these topics were mentioned in the subject line so that they are presented to all respondents, even those who do not open the emails. Importantly, although all of the messages were about crime, the specific attribute we experimentally manipulated was how the message referenced gendered and racial identities. Patterns above in Fig. 1 indicate that the message content matters generally - we observe a decreasing pattern of responsiveness as constituent messages talk increasingly about race. Direct estimates of message effects, averaged across constituent identities, confirm these patterns (see Online Appendix).Footnote 20

When we disaggregate these conditions and look at the message along with the sender, we continue to find a striking pattern of lower responsiveness toward Black men (see Figs. 5 and 6). Black men’s messages are opened 10 percentage points less than White men in all three message types. With respect to replies, Black men again face a lack of responsiveness in the implicit racial condition and the gender condition when compared to White men (3.1 and 4.5 percentage points respectively). Often, this gap is also present when comparing to White women and Black women. This pattern of results suggests that the lack of responsiveness to Black men is not contingent on what they talk about. As a result, we fail to find support for H7 with respect to Black men - the penalties they face seem to consistently apply across all three topics. This is a stronger than expected confirmation of H3.

Fig. 5
figure 5

Opening rates by constituent identity and message. Note: The baseline in each of these graphs is responsiveness to emails with a White masculine sounding name. This figure shows the results of these comparisons along with corresponding 95% confidence intervals

Full size image
Fig. 6
figure 6

Reply rates by constituent identity and message. Note: The baseline in each of these graphs is responsiveness to emails with a White, masculine sounding name. This figure shows the results of these comparisons along with corresponding 95% confidence intervals

Full size image

We do, however, observe results consistent with H7 for different kinds of marginalized constituents. White women, for example, prompt lower open rates than White men when they discuss race implicitly or when they discuss gender. In terms of reply rates, White women prompt a drop in responsiveness in comparison to White men only when they discuss gender (and not either racialized topic). This supports H7 as White women face the most consistent penalties in responsiveness when they discuss the messages linked to gender. This may be due to a number of factors, such as stereotyping of women discussing gender as less important to politics or more general dismissiveness of women’s concerns about gender.

Black women, on the other hand, are a more complex case. As noted by larger research on intersectionality, they have marginalized identities with respect to gender and race. As such, H7 implies the possibility that these kinds of constituents may receive lower levels of elite responsiveness for messages about race and gender. This is not, however, what we observe. Figures 5 and 6 indicate that Black women face a penalty with respect to opening messages relative to White men when they discuss race (either implicitly or explicitly) but not when they mention gender. The same pattern does not emerge for replies. This suggests that responsiveness breaks down at different places for Black women than Black men, a point that would not be clear without our focus on both opens and replies. This partial confirmation of H7 only with respect to race also indicates the value of an intersectional lens in considering responsiveness - Black women face different challenges than Black men and White women when interacting with their elected officials. These difficulties intersect with different political topics in ways that are unique to Black women. Our overall conclusion about H7, then, is one of mixed support.

We next consider H8 and how gaps in responsiveness vary based on the identity of the elected official. To do this, we divide our sample of elected officials into four groups - elected officials who are White and men, elected officials who are White and women, elected officials of color who are men, and elected officials of color who are women. Ideally, we would consider specific non-White racial groups, but our sample size and the precision of our coding system for race did not allow for those kinds of comparisons.Footnote 21 As there is some additional uncertainty around the automated coding of race and gender that we employ here, we also urge some caution in interpreting these results.

As noted earlier, when we examine the identity of an elected official, we are considering how the identity of the official influences the responsiveness of their office and staff. Our research therefore considers if the identity of the elected official has an impact on the behavior of their office. Other research facing the same constraints still finds important patterns by the racial identity of elected officials (e.g., Broockman 2013).

We find very little variation by the racial and gender identity of the elected official. Among White officials who are men, we again observe the lowest response and engagement with Black men (figures A.4 and A.5 in the Online Appendix). We see the same overall pattern of effects for elected officials who are White and women and those who are people of color and men (see figures A.6 through A.8 of the Online Appendix). For elected officials of color who are women, in contrast, we do not observe differences in responsiveness based on constituent identity, which supports for H8 (see figures A.10 and A.11 of the Online Appendix). As the sample size of this group is the smallest and given the uncertainty in the coding for race and gender based solely on names, we also approach these results with some caution.

Discussion

We show that the intersectional approach to responsiveness we implement identifies specific inequalities in representation. We find the strongest and most consistent effects at the level of constituents’ identities; this aligns with our expectations where our hypotheses were the most specific and grounded in prior work. In that sense, we observe distinct patterns of inequities in responsiveness related to who constituents are and how their racial and gender identities intersect. At this level, we test the intersection of race and gender within the context of representation and responsiveness. Ultimately, we find that both matter in determining if elites choose to listen and respond to their constituents. At a general level, this kind of uneven responsiveness poses pressing problems for democracy - one-on-one communications with elected officials can be politically important and shape elites’ sense of what their constituents need and want. If elected officials engage less with one group of constituents, the perspectives of this group will likely be left out of discussions of policy, campaigning, and other activities in politics. On an applied level, this is exactly what we find to be the case for Black men. Lack of attention or listening by elected officials can lead to resentment toward the system of politics. This gap in listening could help to explain the recent turn, among a small, albeit noteworthy portion of young Black men becoming more expressively supportive of the anti-elitist and resentment focused political approach of Donald Trump (Brown and Sanders 2024; Rascoe et al. 2024) due to feelings of being excluded by Democratic representatives.

Looking across these elements of our design, this research suggests that constituent identities, message topics, and official identity can all play a part in responsiveness. At the same time, constituent identities seem like the largest drivers in our results. In light of this, we suggest that it is only through designs like the ones we have used here that researchers can identify the relative importance of these different features of official-citizen communications and various levels of intersectionality. Without this kind of design, the way these different elements of responsiveness compare to each other would be uncertain.

Ultimately, we show that Black men are listened to and interacted with the least. Although we explain this in terms of the social and political stereotypes of Black men, we acknowledge this pattern of results may be produced by a combination of mechanisms. In resolving this inequity, we encourage researchers to consider the processes that lead to the disparities we observe. Regardless of the origin of these gaps, our data indicate that the voices of Black men are not heard in the same ways as others, and they are consistently undervalued among elected officials in our study. We observe this across the different kinds of messages we use, suggesting that this pattern is not tied to one particular topic or policy area. This diminished listening and responsiveness may compound and reinforce institutional inequalities, especially in areas that directly influence and interest this group of Americans. Such inequity has troubling implications for representation, political incorporation, and a host of other crucial components of democracy in the United States.

This does not suggest that other groups - be they Black women, White women, or other groups we did not include in this study - do not face other important representational disadvantages. For example, our study considers only narrowly how officials respond to constituent communications - this leaves unexplored the degree to which policies that elected officials promote correspond to the interests and views of various groups. We also do not consider the way different formal institutions do or do not encourage political engagement or responsiveness to these various social and political groups. Other work interested in intersectionality and representation should take up these points and evaluate how those contexts influence the expectations and findings we demonstrate here.

This research also indicates that what constituents talk about can also influence responsiveness. In our study, elected officials are least likely to reply to messages explicitly about race, regardless of the identity of the constituent. There seem to be some topics, and some kinds of misinformation, that elected officials hesitate to respond to. The fact that this pattern regarding directly racial messages emerges largely for replying to constituent messages (and not for opening the emails) suggests that elected officials choose not to take on these topics rather than ignoring them completely.

There are problematic consequences of the tendency of public officials to respond more to more innocuous messages and to tread carefully around controversial topics. In terms of public relations, this calculus makes sense and relates to the electoral pressures elected officials face. When it comes to governing, however, this pattern raises a number of troubling possibilities. Politics are, often by their nature, contentious and the most important issues facing a society frequently involve the strongest levels of disagreement; yet, what we observe suggests that elected officials often choose to avoid these kinds of important, difficult topics. In the end, this evasiveness may either function as tacit approval of problematic views or result in a political discourse that only allows for the most palatable topics. Either way, this kind of avoidance has serious social and political consequences.

Overall, we find only limited evidence for differences in these patterns based on the combined racial and gender identity of elected officials. These patterns are not as stark as the effects of constituents’ identities. There are some differences in how White elected officials who are women react to other White women, and elected officials of color who are women display different patterns of responsiveness than other kinds of elites. However, we generally observe a high degree of correspondence between how elected officials who are men, regardless of their race, react to these statements from constituents. The patterns of responsiveness we observe also appear similar for White officials who are men and women. Theoretically, we maintain the importance of evaluating intersectionality among the identity of elected officials, and our approach is necessary to document these kinds of patterns of similarity and difference between groups of elites.

Responsiveness is a crucial element to democracy and involves the direct channels between elites and the masses. We show that large inequalities in responsiveness exist, especially as it relates to who is reaching out to an elected official and what they are saying. Many may intuitively guess that elites exercise discretion in considering and responding to constituents’ concerns– but our project demonstrates that this discretion involves systematic patterns of bias involving the identity and concerns of constituents. Insofar as representative democracy would be improved by more equitably involving citizens from all walks of life and with all kinds of concerns, these findings indicate significant barriers in realizing these benefits.

.