Reproducing Hierarchies at the APSA Annual Meeting: Patterns of Panel Attendance by Gender, Race, and Ethnicity

ABSTRACT Research on the political science profession has shown that homophilous research networks—that is, those organized along the lines of gender and race/ethnicity—reproduce hierarchies. Research networks composed of white men experience the most prestige and lead to the most opportunities. This study documents homophilous networks in a setting where they likely are nurtured: academic conferences. Drawing data from the 2019 Annual Meeting of the American Political Science Association, we examine the correspondence between the gender and the racial/ethnic composition of section members, panelists, and audience members for four research sections: Political Methodology; Political Psychology; Race, Ethnicity, and Politics; and Women and Politics. We find that attendees’ and panelists’ gender and racial/ethnic identity largely mirror the dominant gender and racial/ethnic group in their section. These findings indicate that homophily manifests at academic conferences and that efforts to diversify research networks should consider who listens to whom in these settings.

W omen, people of color, and women of color are underrepresented in the professoriate. Political science-like other disciplinesremains dominated by men and white individuals. For 2018-2019, the American Polit-ical Science Association (APSA) Project on Women and Political Minorities reported that among the largest PhD-and MA-granting institutions, women and nonwhite faculty members comprised 27.8% and 20.2% of tenured professors, respectively (American Political Science Association 2019). Women, people of color, and women of color are underrepresented in other metrics of professional achievement as well. For instance, women publish in top political science journals less often than men and in proportions considerably less than their presence at research universities (Atchison 2017;Maliniak, Powers, and Walter 2013;Mitchell, Lange, and Brus 2013).
One explanation for the persistent underrepresentation of women and people of color is the homophilous nature of professional networks. Prestigious networks are dominated by those who are male and white, and these patterns reproduce over time. The basic mechanism behind these patterns is that "like gravitates to like": people cultivate relationships with people who look like them. Yet, not all networks function equally. In the academy, (white) men's homophilous networks confer social capital on junior men and help them achieve insider status, whereas the homophilous networks of women and/or scholars of color provide critical support but generally do little to improve their outsider status in the profession (see Atchison 2018 for more discussion). Homophily thus shapes access to professional opportunities along lines that preserve traditional gendered and raced hierarchies, and not only in the academy-homophily also shapes access to top managerial positions in business (Holgersson 2013) and recruitment within political parties (Bjarnegård and Kenny 2015).
This article documents homophilous research networks in political science and builds on other studies that have noted "like gravitating to like" in the field. For instance, collaborator teams typically are single-sex groups (Atchison 2017(Atchison , 2018Teele and Thelen 2017), especially single-sex groups of men (Brown and Samuels 2018). We focus on homophilous networks at a key place where they likely are nurtured: academic conferences. Our data come from the 2019 APSA Annual Meeting, a flagship conference in the discipline.
The project was inspired by anecdotal observations that gender appeared to influence patterns of panel attendance in the Women and Politics research section, in which women comprise more than 90% of members (Mealy 2018). In this section, homophilous research networks are dominated by women, inverting the typical pattern in political science. However, we noticed that when men presented on gender and politics panels, more men seemed to attend. If men attend to hear other men but not to hear women-even when women conduct most of the subfield's research-then panel attendance demonstrates the presence of homophilous networks. Thus, our central hypothesis is that panel attendance both reflects and reinforces homophilous networks in the discipline, along the lines of both gender and race/ethnicity.
To investigate our expectations, we collected population data at the session level for the following four research sections at the 2019 APSA Annual Meeting: Political Methodology (a section with predominantly men members); Women and Politics (a section with predominantly women members); Race, Ethnicity, and Politics (a section with predominantly nonwhite members); and Political Psychology (a section with a gender and racial/ethnic composition that resembles the discipline as a whole). 1 Our analyses show that (1) the gender and racial/ethnic identity of attendees and panelists largely mirrors the dominant gender and racial/ethnic group in their section; and (2) variation in the composition of the panels by gender and by race/ethnicity correlates with variation in the composition of the audience by gender and by race/ethnicity. These findings indicate that conferences maintain homophily and that efforts to diversify academic networks should consider who listens to whom at disciplinary meetings.

HOMOPHILOUS RESEARCH NETWORKS
This study builds on research about professional networks in political science (Brown and Samuels 2018). Women graduate students in top political science programs continue to describe their department as a "boys' club" (Almasri, Read, and Vandeweerdt 2021). Women political scientists receive fewer invitations to subfield conferences, workshops, and seminars, settings in which scholars find new coauthors and build name recognition (Barnes and Beaulieu 2017). Women's work is promoted less ; women publish and are cited less frequently than men, especially in the top political science journals (Atchison 2017;Dion, Sumner, and Mitchell 2018;Mitchell, Lange, and Brus 2013); and they are underrepresented as editors of top journals (Palmer, van Assendelft, and Stegmaier 2020). Their work also is underrepresented in political science graduate syllabi and reading lists (Colgan 2017;Diament, Howat, and Lacombe 2018;Hardt et al. 2019;Smith et al. 2020). Similar gaps in leadership, publication, and opportunities affect political scientists of color (Lemi, Osorio, and Rush 2020). Women of color experience forms of exclusion based on both gender and race/ethnicity: they comprise only 11% of all political science faculty members (Michelson and Monforti 2021), and they face sharply unequal access to and opportunities for professional advancement (Behl 2017;Sinclair-Chapman 2019).
In summary, research indicates that homophilous networks dominated by white men systematically limit the upward mobility of individuals traditionally underrepresented in political science. Political scientists inside these networks accumulate opportunities and are regarded as the experts, whereas those outside face compounding disadvantages.
Nonetheless, there is no research (to our knowledge) that documents and analyzes homophilous research networks at academic conferences. Conferences are venues where attendees cultivate new relationships and renew existing ones, and conference programs provide information about "who's who" in specific areas of study. Conference presentations and interactions shape scholars' future opportunities-such as who is invited to present at seminars and join coauthorship teams and whose work is cited. We hypothesize that patterns of panel attendance at the 2019 APSA Annual Meeting are shaped by gender homophily and by race/ethnicity homophily, as follows: • Hypothesis 1: Men conference attendees turn out at higher rates to hear men panelists (and women attendees turn out at higher rates to hear women panelists). • Hypothesis 2: White conference attendees turn out at higher rates to hear white panelists (and nonwhite attendees turn out at higher rates to hear nonwhite panelists).
Our central hypothesis is that panel attendance both reflects and reinforces homophilous networks in the discipline, along the lines of both gender and race/ethnicity.
T h e P r o f e s s i o n : R e p r o d u c i n g H i e r a r c h i e s a t t h e A P S A A n n u a l M e e t i n g

DOCUMENTING PATTERNS OF PANEL ATTENDANCE
In 2018, the year we designed this study, 63.1% of APSA members were men. White members (non-Hispanic white, European, and American) comprised 61.5% of the association membership and nonwhite members (Black, Asian, Latinx, Native American, and Middle Eastern) comprised 17.3%, with no data reported for the remaining 21.2% (Mealy 2018).
Our study focuses on four APSA organized sections, each with a different membership profile. This variation permits structured comparisons among section membership, panel presenters, and panel attendees. We examine the Women and Politics section, which has predominantly women members (92%); Political Methodology, a section with predominantly men members (80% men); Race, Ethnicity, and Politics, a section with even numbers of men and women (50% women and 50% men); and Political Psychology, a section that mirrors APSA membership (37% women members) (Hidalgo et al. 2018). This case selection also allows us to compare a predominantly nonwhite section-Race, Ethnicity, and Politics, with only 45% white members in 2018-to three sections that are predominantly white. In 2018, Women and Politics, Political Psychology, and Political Methodology were 74%, 75%, and 77% white, respectively (American Political Science Association 2018). 2 Unfortunately, APSA did not report 2018 data intersectionally, so we lack data by gender and by race/ethnicity for these sections.
We collected population data on all sessions sponsored or cosponsored by each of these four sections at the 2019 APSA Annual Meeting, excluding Sunday sessions due to well-known low attendance on that day. By "sessions," we mean traditional paper panels, roundtables, and author-meets-critics panels. Our final sample includes 104 sessions over Thursday, Friday, and Saturday. Political Methodology and Political Psychology each sponsored or co-sponsored 22 sessions; Race, Ethnicity, and Politics offered 26 sessions; and Women and Politics offered 34 sessions. The dataset includes more observations for Women and Politics relative to the other sessions.
To discern homophilous networks, we needed to answer two questions for each session: Who is on the panel and who is in the audience? Answering these questions about whether "like follows like" in a conference setting required collecting data on both audience members and panelists, including demographic details (gender and race/ethnicity), institution type, professional status, and other reasons that people attend panels (personal ties to a panelist and interest in networking). These other reasons could also be correlated with homophily. Combined, this information allows us to assess the relative importance of homophilous research networks formed along various dimensions, such as being affiliated with an R1 university.
We began with the audience. Research assistants fielded a paper survey at the beginning of all 104 sessions (see online appendix A for the complete survey questionnaire). This survey needed to be distributed, completed, and collected before the session began, so we prioritized efficiency over precision. We designed a brief survey, asking respondents only six questions about themselves and their reasons for attending the session. For instance, because we anticipated having too few responses for comparisons between specific racial and ethnic groups, we asked respondents to identify as white or nonwhite-recognizing that this question would not fully capture the diversity of attendees. Similarly, our question about institution type followed the Carnegie classifications, but the response option of "PhD-granting institution" may have introduced noise. Faculty members in a non-PhD-granting department but at a PhD-granting university may have been confused about whether to mark themselves as at a "PhD-granting institution" (see online appendix B for more details about the survey methodology).
Data on panelists come from the 2019 APSA Annual Meeting program. A "panelist" is anyone listed on the program with a role in the session, including discussants, chairs, authors, coauthors, and roundtable or authors-meet-critics presenters. Further, we do not distinguish between presenting and non-presenting coauthors: because these distinctions were not listed in the 2019 program, attendees would not know ex ante which authors would be present, and all authors could potentially draw attendees based on homophily.
The final dataset is organized by session, so panelists and audience members are included at every session they participated in or attended. For instance, if the same woman graduate student attended three sessions in Political Methodology, she is counted at all three sessions. We therefore have a session-level dataset with sessions nested in four research sections. Additional information about ethics, data availability, and replication is in online appendix C, and replication materials are posted in the Harvard Dataverse (see Piscopo et al. 2022).
For each session, we calculate the proportion of attendees and panelists who are men and women, white and nonwhite, and from an R1 university. We also calculate the proportion of attendees who are graduate students or tenure-line faculty members (as opposed to non-tenure-line faculty, undergraduates, or researchers at nonacademic institutions).

WHO PARTICIPATES IN THE ANNUAL MEETING?
Descriptive statistics provide an initial snapshot of how participation in the 2019 APSA Annual Meeting is both gendered and raced.

Whose Voices Are Heard?
Based on the conference program, 912 panelists participated in sessions sponsored or co-sponsored by Political Psychology; Political Methodology; Race, Ethnicity, and Politics; and Women and Politics. An important caveat is that panelists are not grouped randomly. Some fully formed sessions are submitted, and other sessions are created by section organizers. Section chairs may aim to avoid all-male panels or all-white panels and may use personal knowledge or names to make inferences (however problematic) about potential participants' gender and race/ethnicity. Unfortunately, we cannot observe whether organizers' own networks, commitments to diversity, or inferences about panelists' identities influenced their decision making. We can only observe their final decisions as reflected in the final program.
With these caveats in mind, we find that 89% of panelists across the four sections were from PhD-granting institutions, with a high of 93% in Political Methodology and a low of 84% in Women and Politics. Women comprised 51% of all panelists, but their participation varies considerably by section: women were 21% of Political Methodology panelists; 39% of Political Psychology panelists; 57% of Race, Ethnicity, and Politics panelists; and 76% of Women and Politics panelists. Therefore, women panelists were overrepresented relative to their section membership in Race, Ethnicity, and Politics by about eight percentage points; underrepresented in Women and Politics by about 15 percentage points; and represented in proportion to their membership in Political Methodology and Political Psychology (see figure 2).
For race/ethnicity, we had data for 775 panelists. The proportion of white panelists was 77% in Political Psychology; 69% in Political Methodology; 68% in Women and Politics; and 43% in Race, Ethnicity, and Politics. Relative to their section membership, white panelists are underrepresented in all sections except Political Psychology. (Note: When comparing Annual Meeting data to APSA section data for race/ethnicity, we use the proportion of white individuals rather than individuals of color because APSA reported section membership as percentage white.) Figure 1 displays the data intersectionally. The top panel reports each section's panelists by gender and race/ethnicity, arranged in descending order by the proportion of white men panelists. Given that we lack intersectional data for 15% of panelists, these proportions are only suggestive. That said, white men predominate as panelists in both Political Methodology and Political Psychology, whereas white women (followed by women of color) predominate in Women and Politics, and women of color (followed by white women) predominate in Race, Ethnicity, and Politics.

Who Listens?
Our survey respondents at the 2019 APSA Annual Meeting included 904 audience members. Not all audience members participated in the survey because they may have arrived mid-session or declined to participate (see online appendix B for response rates relative to those present when the session began). The reported proportions therefore refer to audience-member survey respondents, whom we describe as "attendees," for clarity.
Seventy-seven percent of attendees are graduate students or tenure-line faculty members. The remaining attendees are  We had high response rates to questions measuring our primary variables of interest, gender and race/ethnicity: 876 attendees disclosed their gender (97%), 838 disclosed their race/ethnicity (93%), and 821 disclosed both (91%). The proportions of women attendees roughly correspond to their proportions among section members, except in Women and Politics, where significantly more men attend panels than are section members (31% men attendees versus less than 10% men members). Recall that Women and Politics sessions also overrepresented men panelists relative to section membership. The proportion of white attendees is highest in Political Psychology (higher than the proportion of white members) and lowest for Race, Ethnicity, and Politics.

Proportion of Panelists and Attendees by Gender and by Race/Ethnicity
As for respondents who provided both their gender and race/ ethnicity, figure 1 (bottom panel) shows that white men predominate as attendees in both Political Methodology and Political Psychology, as they did among panelists. White women predominate among attendees in Women and Politics (followed by women of color), also reflecting the same pattern among panelists. White women and women of color attend Race, Ethnicity, and Politics panels in similar proportions, although women of color are better represented among Race, Ethnicity, and Politics attendees.

Like Follows Like
To summarize, we find evidence for homophilous research networks by gender and by race/ethnicity, as shown in figure 2. Where like does not follow like, our findings nonetheless suggest important patterns about who speaks and who listens. For example, slightly more women attend versus present in Political Methodology, meaning that men's voices predominate as experts, even as more women are listening. In Women and Politics, a different pattern appears: disproportionately more men present and attend than are section members, meaning that men scholars are gaining increasing presence and voice, but not joining the section. Insofar as section membership signals commitment and service to the subfield (e.g., becoming a member means paying dues), men are not supporting the Women and Politics section at the same rate that they are taking advantage of its sessions. Similarly, white members are attending Race, Ethnicity, and Politics sessions without joining the section.

ADDITIONAL EXPLANATIONS FOR PANEL ATTENDANCE
The survey also asked attendees to indicate all of their reasons for attending a session. Pooling attendees across sections, 86% indicated that they had an interest in the section, 30% indicated interest in a paper, and 20% indicated a personal connection with one of the panelists. 3 Surprisingly-and perhaps counter to our intuition that attendees view panels as important moments for networking-only 19% of respondents indicated an interest in networking.
This trend notwithstanding, we did find statistically significant variation for interest in networking by attendee rank. Perhaps unsurprisingly, more graduate students (20%) affirmed their interest in networking compared to tenure-line faculty members (14%) (p = 0.025). We also find patterns by section type and by attendees' gender (but not by race/ethnicity). Interest in networking was lowest among attendees at Political Methodology panels (9% of attendees) and highest among attendees at Race, Ethnicity, and Politics panels and Women and Politics panels (25% of attendees for each section). Of those interested in networking, women were less keen than men in all sections other than Women and Politics, where women comprised 66% of interested networkers. These findings, although suggestive, are consistent with the presence of homophilous networks. In men-dominated sections, men are more interested in networking, but in the one women-dominated section (Women and Politics), women are more interested in networking. In other words, respondents express more interest in networking in settings where most others look like them.

HOMOPHILOUS RESEARCH NETWORKS AND PANEL ATTENDANCE
The descriptive statistics indicate that panelists and attendees largely mirror the gender and racial/ethnic composition of sections, which is not surprising. After all, section membership captures the demographics of those conducting research in that area, and these trends likely shift slowly over time. To confirm the patterns that emerge in the descriptive statistics, we estimate ordinary least squares (OLS) regression models with two dependent variables: the proportion of men attendees and the proportion of white attendees. Our main independent variables are the proportions of men panelists and white panelists, respectively, and we control for section type (with Political Methodology as the reference category) and attendees' interest in networking. We do not control for institution type or position due to low variation: as discussed, most APSA attendees and panelists are graduate students or tenure-line faculty members at R1 universities.
White men predominate as panelists in both Political Methodology and Political Psychology, whereas white women (followed by women of color) predominate in Women and Politics, and women of color (followed by white women) predominate in Race, Ethnicity, and Politics.
The regression results reported in table 1 emphasize the existence of homophily. The more men panelists, the more men audience members (models 1 and 2), and the more white panelists, the more white audience members (models 3 and 4). Relative to Political Methodology, the proportion of male attendees decreases for Women and Politics, whereas the proportion of white attendees decreases for Race, Ethnicity, and Politics. Audience members' interest in networking does not change these results. The adjusted R 2 values indicate that about 31% of the variation in audience composition by gender can be explained by panelist composition by gender, and about 23% of the variation in audience composition by race/ethnicity can be explained by panelist composition by race/ethnicity.

CONCLUSION
Professional networks organized along the lines of gender and race/ethnicity can contribute to in-group and out-group effects, shaping scholars' access to prestige and resources. Our study of panel attendance at the 2019 APSA Annual Meeting reveals important patterns: when panelists are predominantly white and male, audience members are also predominantly white and male. When these patterns diverge, we see more diversity, but in ways that reinforce gendered hierarchies along traditional lines. In Political Methodology, for example, women attend panels more than they present on them, meaning that expertise remains masculinized even as women evince interest in the section. In Women and Politics, men attend and present in roughly

White members White panelists White attendees
T h e P r o f e s s i o n : R e p r o d u c i n g H i e r a r c h i e s a t t h e A P S A A n n u a l M e e t i n g similar proportions, but women still carry the service load (section membership). Finally, most APSA Annual Meeting presenters and attendees are from R1 universities, whereas less conventionally prestigious sections (Race, Ethnicity, and Politics and Women and Politics) have greater shares of attendees from non-PhD-granting institutions. Our data thus provide evidence that homophilous research networks are nurtured and maintained at professional conferences. These analyses cannot account for the qualitative experiences of marginalized scholars who seek access to majority-men and majority-white networks, but they underscore how the mechanism of "like gravitating to like" creates and reinforces patterns of exclusion. Our study provides an important reminder to those who are focused on diversity: academic conferences are sites where homophilous research networks develop but also where professional networks can become more diverse.

ACKNOWLEDGMENTS
The authors thank Betsy Super and her team at the American Political Science Association for their support of the project and for sharing data. We also thank the 16 research assistants who carried out data collection at the 2019 APSA Annual Meeting and Cathy Jenkins for technical support. We gratefully acknowledge funding from the Centennial Center for Political Science and Public Affairs, as well as the assistance of Bucknell University's Office of Sponsored Projects. We also thank Dawn Teele and Denise Walsh for their support of this project.

DATA AVAILABILITY STATEMENT
Research documentation and data that support the findings of this study are openly available at the PS: Political Science and Politics Harvard Dataverse at https://doi.org/10.7910/ DVN/8AMHQW.

SUPPLEMENTARY MATERIALS
To view supplementary material for this article, please visit http:// doi.org/10.1017/S1049096522001032.

CONFLICTS OF INTEREST
The authors declare that there are no ethical issues or conflicts of interest in this research. ▪ N O T E S 1. As of 2020, the Women and Politics research section was renamed the Women, Gender, and Politics section. We use the section name at the time of our study. 2. Contemporaneous estimates place the white membership of the Political Methodology section even higher, at 81% (Hidalgo et al. 2018, 4).  Notes: *=10%, **=5%, ***=1%. †Two sessions were dropped from the dataset due to zero survey responses from attendees.
In Political Methodology, for example, women attend panels more than they present on them, meaning that expertise remains masculinized even as women evince interest in the section. In Women and Politics, men attend and present in approximately similar proportions, but women still carry the service load (section membership).