Structural Inequality in Political Science

Opportunity Structure for Discrimination

Petersen and Saporta’s (Reference Petersen and Ishak2004) theoretical framework of the opportunity structure for discrimination (OSFD) provides a critical lens through which we can better understand the intricate relationship between resource distribution, reward allocation, and the pervasive presence of structural inequality in a social system. The central arguments of OSFD posit that occupational inequality is mainly manifested through three discriminatory mechanisms: allocative, within-job wage, and valuative mechanisms (Petersen and Saporta Reference Petersen and Ishak2004). The allocative mechanism involves discrimination that has placed historically excluded people into lower-ranking jobs. The within-job wage mechanism, also known as reward gap, subjects historically excluded people to lower wages when performing the same job as their majority counterparts. The valuative mechanism devalues occupations where historically excluded people are most represented (Barnett, Baron, and Stuart Reference Barnett, Baron and Stuart2000; Castilla Reference Castilla2005). In spite of its origins in the occupational realm, OSFD can be seamlessly integrated into the academic setting because of the numerous parallels between academic and occupational systems.

Resembling the roles played by job and wage in the occupational realm, publication and citation constitute two primary pillars of the academic system. The desire to make one’s scholarly contribution known to others (through publication) and recognized by others (through inward citation) is a constant theme underlying the routine work of academics (Small Reference Small2004). As pointed out by Merton (Reference Merton1988), a distinct characteristic of science is that it establishes the “private property” of knowledge by “having its substance freely given to others who might want to make use of it.” Publication is a critical means for individual scientists to claim knowledge ownership and build symbolic capital. Nevertheless, publications are not weighted equally. Some publications receive more recognition than others, contributing to an uneven distribution of academic impact (Bourdieu Reference Bourdieu1988) and an expanding divide between the symbolically rich and poor (Kristensen Reference Kristensen2018), as illustrated by the “Matthew Effect” (Merton Reference Merton1957, Reference Merton1968, Reference Merton1988).

Situated within the context of academic publishing, this study will focus on the research article as its key unit of analysis, examine citations as the primary reward obtained by articles, and identify journals as the primary venue for acquiring citation rewards. This structure parallels the occupational context depicted in the OSFD, where the individual employee is the unit of analysis, wage is the primary reward, and job is the primary venue for earning wage rewards. Table 1 provides additional clarity on this analogy.

Table 1

Comparison of OSFD’s constituents within two contexts

Given their functional equivalence, we utilize the concept of allocative mechanism to explore the publication gap and adapt the concept of within-job wage mechanism to elucidate the within-journal citation gap. Finally, as a discipline evolves, multiple lines of inquiry emerge, and scholars cluster into different topic areas to form their own niches. Men and women, as well as racial/ethnic majorities and minorities, often specialize in different subfields (Grant and Ward Reference Grant and Ward1991; Kozlowski et al. Reference Kozlowski, Lariviere, Sugimoto and Monroe-White2022; Maliniak, Powers, and Walter Reference Maliniak, Powers and Walter2013; Teele and Thelen Reference Teele and Thelen2017). The affinity between social categories and topic selection compounds the issue of inequality, giving rise to the evaluation gap, which is similar to valuative mechanism in OSFD.Footnote ¹

In sum, this study will examine 1) publication gap: the degree to which articles authored by historically excluded scholars are disproportionately sorted into lower-tier journals; 2) within-journal citation gap: the disparity in citation counts between historically excluded authors and their majority counterparts when their works are published in the same journals, and 3) evaluation gap: the degree to which minority-concentrated specialties/topics receive fewer citations than majority-concentrated specialties/topics.

Data and Measures

We retrieved all research articlesFootnote ³ published in SSCI-indexed political science journals between 1970 and 2020 from the Web of Science (WoS) database. A total of 211,425 articles from 286 journalsFootnote ⁴ were archived on November 21, 2021.

A quick inspection of the data revealed that a few journals were devoted to publishing short commentaries; moreover, the commentaries published in these commentary-oriented journals were mostly classified as “articles” rather than “letters” or “discussions.” To remove these anomalous cases, we calculated the average page length and the average number of references cited per article for all journals based on the assumption that commentaries are shorter and cite fewer references than original research. We identified 21 commentary-oriented journals, including The Nation, The New Republic, and Economic and Political Weekly, whose articles on average have only 4.05 pages and cite only 2.53 references. For details, please refer to online appendix note S1. After removing these outliers, 265 journals and 163,621 articles were retained. Unless stated otherwise, the findings discussed here are derived from the sample without these outliers.

Author Identity

To capture the demographic characteristics of an article’s authorship, we formulate two types of measures: first author-based (FA-based) and team-based measures. FA-based measures only consider the identity of the first author of a given paper, whereas team-based measures account for the identities of all authors listed in the byline. In computing team-based measures, we assume that the authorship of an article is evenly distributed among contributors, with each author receiving equal fractional credit. This approach is instrumental in avoiding distortion caused by articles with extremely long author lists.

Gender:Footnote ⁵ 1) FA-based Gender_i is equal to 1 if the first author of article i is female and equal to 0 if the first author is male; 2) Team-based Gender_i measures the percentages of female authors named in the byline of article i. In line with previous research (AlShebli, Rahwan, and Woon Reference AlShebli, Rahwan and Woon2018; Dion, Sumner, and Mitchell Reference Dion, Sumner and Mitchell2018; Hofstra et al. Reference Hofstra, Kulkarni, Sebastian, He, Jurafsky and McFarland2020; Teele and Thelen Reference Teele and Thelen2017), we employ Genderize.io and Forebears.io APIs to estimate the gender of a given author based on the forename. Genderize.io and Forebears.io are two genealogical archives that have recorded the gender and racial/ethnic characteristics of more than 15 million and 28 million names across countries, respectively. According to AlShebli, Rahwan, and Woon (Reference AlShebli, Rahwan and Woon2018), Hofstra et al. (Reference Hofstra, Kulkarni, Sebastian, He, Jurafsky and McFarland2020), and Teele and Thelen (Reference Teele and Thelen2017), Genderize.io has a low error rate of 5%–7% when predicting authors’ genders across all fields, and a further reduced error rate of 2% in political science. Given that forename analysis is less sensitive in detecting Asians’ gender and Genderize.io tends to over-guess women (Teele and Thelen Reference Teele and Thelen2017), we manually check the top 300 most productive Asian scholars who were predicted to be female by the two archives (online appendix note S6) and rectify misclassified gender labels. Since WoS did not consistently document authors’ first names before 2007, the gender-related analyses are confined to data from 2007 onward.

Race/Ethnicity: 1) FA-based Race/Ethnicity_i refers to the racial/ethnic identity of the first author of article i; 2) Team-based Race/Ethnicity_i measures the shares of article i’s authors identified with a certain racial/ethnic group. Consulting the World Value Survey’s Cultural Map (Inglehart and Welzel Reference Inglehart and Welzel2005) and Huntington’s (Reference Huntington1993) taxonomy of civilizations, we construct a classification scheme with eight distinct racial/ethnic groups: Anglo-Saxon, European, Slavic, Asian, African/African American, Hispanic/Latino, Jewish, and Others (refer to online appendix notes S2 and S6).

We combine surnames and geolocations to predict the race/ethnicity of scholars (refer to online appendix note S3). Surnames have long been recognized as an effective indicator of race/ethnicity and an enduring symbol of ancestry (Hanks 2003). A wide array of studies published in esteemed academic journals, such as AlShebli, Rahwan, and Woon (Reference AlShebli, Rahwan and Woon2018), Kozlowski et al. (Reference Kozlowski, Lariviere, Sugimoto and Monroe-White2022), and Hofstra et al. (Reference Hofstra, Kulkarni, Sebastian, He, Jurafsky and McFarland2020), have opted to infer race/ethnicity based on surnames. However, previous studies have provided evidence that surname analysis can encounter notable challenges when attempting to distinguish individuals of African descent from those of Anglo descent (Lauderdale and Kestenbaum Reference Lauderdale and Kestenbaum2000; Mateos Reference Mateos, Webber and Longley2007). The presence of African/African American individuals who have Anglo-sounding surnames might be undesirably underestimated.

To address this limitation, we join a nascent body of studies that enhance surname analysis by incorporating geolocation information using the Bayes theorem (Elliott et al. Reference Elliott, Fremont, Morrison, Pantoja and Lurie2008; Fiscella and Fremont Reference Fiscella and Fremont2006; Haas et al. Reference Haas, Elliott, Jacob W., Adams, Wilson-Frederick, Mallett, Gaillot, Haffer and Haviland2019; Jackman Reference Jackman2004; Imai and Khanna Reference Imai and Khanna2016). Under Bayesian models, the probability that a person falls into a certain racial/ethnic category is constrained by the surnames that they bear as well as the neighborhood where they live. If an author resides in an ethnically heterogeneous neighborhood, the model’s uncertainty will increase, resulting in a more cautious and less confident estimation of race/ethnicity. To validate Bayesian estimates, we trained two human coders to annotate races/ethnicities of a random sample of 1,000 authors (refer to online appendix note S6). The validation results confirm that the Bayesian model outperforms the surname-only model by a large margin, achieving accuracy rates of over 90% in all racial/ethnic groups (Gill and Lee Reference Gill and Walker2005). Nevertheless, when breaking down the validation metrics by racial/ethnic groups, we observe a tendency to overestimate the representation of African/African American individuals. Considering this, we take an additional step to manually examine the races/ethnicities of the top 300 system-identified African/African American scholars who have published the largest number of articles. This manual verification process allows us to rectify misclassified cases and to ensure greater accuracy.

Region_. 1) FA-based Region_i equals 1 if the first author of article i is affiliated with an institution in the Global SouthFootnote ⁶; 2) Team-based Region_i denotes the percentage of authors affiliated with institutions in the Global South.

Team Diversity

As suggested by Dion, Sumner, and Mitchell (Reference Dion, Sumner and Mitchell2018), diverse scholarship might contribute to decreasing structural inequality in academic publishing. Between-group gaps might be narrowed if an increasing number of minority scholars could participate in cross-identity collaboration and could get involved in demographically diverse teams. Therefore, we are interested in testing the effectiveness of team diversity in alleviating alleged publication and citation gaps. To this end, we formulate a series of diversity-related factors that might affect publication and citation gaps.

• Mixed-Gender_i is a dummy variable that equals 1 if the research team comprises both men and women and 0 if the team comprises entirely women or men.

• Cross-Racial/Ethnic_i is also a dummy variable that equals 1 if the research team comprises scholars of different races/ethnicities and 0 if the team solely comprises scholars of the same race/ethnicity.

• Cross-National_i denotes whether article i is produced by an international research team. It takes a value of 1 when the team includes scholars from different countries. It is worth noting that the “country” here refers to the country of affiliation rather than the country of origin.

• Cross-Regional_i equals 1 if the research team comprises scholars from both Global North and South institutions and 0 if the research only involves scholars from either Global North or South institutions.

• Since team collaboration is a prerequisite for team diversity, it is impossible for single-authored papers where teamwork is absent to demonstrate team diversity. Therefore, for single-authored articles, Mixed-Gender_i, Cross-Racial/Ethnic_i, Cross-Regional_i, and Cross-National_i consistently have a value of 0.

• %Outgroup References_i is a proxy for reference diversity. To gauge its magnitude, we consider references cited by article i and measure the percentage of outgroup references.Footnote ⁷ We also develop two types of measures: 1) FA-based %Outgroup References_i measures the percentage of prorated references written by racial/ethnic outgroups for the first author. For coauthored references, authorship is equally distributed among cited authors. 2) Team-based %Outgroup References_i measures the percentage of prorated references written by racial/ethnic outgroups for all authors.

Publication Gap

To gain insight into publication gaps, we examine the gender, race/ethnicity, and regional distribution of authors in SSCI-indexed political science journals. As indicated by figure 1, the largest publication gap surfaces between scholars from the Global South and Global North scholars. Only 9% of articles are authored by Global South researchers. This gap is even wider among the top-quartile (Q1) journals, where Global South authors account for 4% of publications.

Figure 1

Prorated team-based authorship by demographic groups

Notes: The upper panel demonstrates the publication gaps between groups in SSCI-indexed political science journals, excluding commentary-oriented outliers (#Articles=163,374). The middle panel demonstrates the publication gaps in Q1 journals (#Articles= 28,204). The lower panel demonstrates the difference between the gender/ethnic composition of SSCI journal authors and APSA members during the year 2020. To align the racial/ethnic typologies, we combine the categories of “Anglo-Saxon,” “European,” and “Slavic” to approximate the APSA’s “Non-Hispanic White and European American” category while grouping the APSA’s “South Asian or Indian American” and “East Asian or Asian American” categories to approximate our “Asian” category. Due to space limitations, the “Others” category is not displayed. The white bars demonstrate groups that are overrepresented in academic journals compared to their representation within the APSA while the black bars demonstrate underrepresented groups.

While gender and racial/ethnic disparities also exist, we find that they are constant across all quartiles of journals and are generally reflective of the demographic composition of professional society. Female scholars contribute 31% of articles, while male scholars contribute 69%. In terms of the racial/ethnic gap, Anglo-Saxons and Europeans collectively author over two-thirds of publications, surpassing the third-place Slavic-origin scholars by a sizeable margin of 28%.Footnote ⁸ Compared to other disciplines (Freeman and Huang Reference Freeman and Huang2014, Reference Freeman and Huang2015; Hopkins et al. Reference Hopkins, Jawitz, Mccarty, Goldman and Basu2013; Kozlowski et al. Reference Kozlowski, Lariviere, Sugimoto and Monroe-White2022), political science is characterized by a more significant presence of Europeans and a considerably weaker presence of Asians. Nevertheless, there is no significant widening of gender and racial/ethnic gaps in the upper echelon of journal rankings. Minority scholars are not disproportionately sorted into lower-ranked journals.

Moreover, by comparing the gender and racial/ethnic composition of SSCI authors and American Political Science Association (APSA) members (figure 1, lower panel), we find that the representation of minority scholars in academic journals is generally in line with their representation in the APSA member pool. If APSA member composition is considered a baseline measure for the demographic makeup of the discipline, this result implies that the publication gap can be attributed to the imbalanced population structure in the field. With that said, the most significant disparity is observed along gender lines. Female scholars’ publication in academic journals is 5% less than their presence in the APSA. This disparity highlights the distinct challenges encountered by different groups. While the primary challenge for racial/ethnic minority groups is to increase the entry of quality candidates into the academic world, female researchers face more obstacles in maintaining their presence during the publication process and avoiding attrition as they progress through career paths.

Figure 2 further visualizes the trends of group representations. According to subplot A, as women incrementally expanded their presence in political science journals, the disparity between top-tier and lower-tier journals diminished. Both Q1 and non-Q1 journals have dedicated an equal share of 34% of publications to women’s works in recent years. The visibility of Global South scholars also increased. However, their representation in Q1 journals did not improve proportionately. They have remained consistently low-profile in top-tier journals in the past two decades.

Figure 2

Time trends of prorated team-based authorship by demographic groups

Notes: Subplots on the same row have shared y-axes. The filled areas between trendlines are indicative of between-journal publication gaps. Racial/ethnic group data before 1976, gender data before 2007, and regional data before 1998 are not displayed since they are too sparse to indicate reliable and consistent temporal patterns.

In terms of the racial/ethnic gap, the representation of Anglo-Saxons dwindled drastically from 46% in 1981 to 27% in 2020, while the presence of Europeans increased moderately from 33% to 38%. When these two groups are combined, their total representation declines by 14%, implying a general trend toward de-Westernization and greater racial/ethnic inclusiveness in the political science discipline. Against this backdrop, scholars with racial/ethnic minority background generally improved their representation over the past five decades, with the exception of Jewish scholars. Scholars of African descent demonstrated a sharp increase in the late 1990s, in tandem with the rapid democratic transitions across sub-Saharan Africa. The Journal of Democracy (JoD) contributed the most to this abrupt shift, which devoted 24 articles to African politics between 1997 and 1999, half of which were authored by African/African American individuals.

Furthermore, the red and blue trendlines are closely intertwined for most racial/ethnic groups, indicating that a group’s representation in Q1 journals is on par with its overall presence in the field. Notwithstanding, the Slavic group stands out as an exception, significantly underrepresented in top-tier journals despite growing visibility (subplot E). A closer look at the data reveals that 54.3% of articles authored by Slavic-origin scholars were published in lower-tier journals specializing in Eastern European studies, such as Politicka Ekonomie and Osteuropa, which had a narrower audience and received fewer citations than generalist journals.Footnote ⁹ This homophily between identity and topic/outlet selection probably penalizes Slavic researchers, relegating Eastern European studies to a less rewarding subfield.

Works by Hispanic/Latino scholars are also seen more frequently in lower-tier journals, yet the disparity is not as strong as it is for Slavic scholars. As shown in subplot G, the difference between the red and blue lines, which was relatively minor during 1997–2006, increased from 2007 to 2012. However, starting in 2013, the gap began to narrow until the two lines intersected in 2020. Among articles written by Hispanic/Latino scholars, 25.9% are published in specialized journals that focus on Latin American studies. While the ratio is only half that of Slavic scholars and has been decreasing recently, it still suggests that Hispanic/Latino scholars might face a certain level of marginalization.

Scholarly Collaboration and Team Diversity

The growing diversity in journal publications is closely linked to the rise of a collaborative culture. As depicted by subplot A of figure 3, collaborative papers have continuously increased in the past five decades. The uptrend is even more prominent among Q1 journals. The red line is always higher than the blue line, and the gap between them becomes wider over time. This finding suggests that top-tier journals have led the process of collaborative knowledge building by publishing more multiauthored papers than other journals.

Figure 3

Time trends of scholarly collaboration in political science

Notes: Subplot A displays the percentages of collaborative articles in political science journals in general. Subplots B-E illustrate the percentages of cross-identity collaborative articles out of all collaborative articles. Collaborative articles refer to articles authored by at least two scholars, while cross-identity collaboration refers to research work that involves scholars from at least two distinct demographic groups.

Team diversity is also increasing. Subplots B, C, and D show that a growing percentage of scholars participate in cross-identity collaboration. Mixed-gender collaboration grew by 12%, accounting for approximately half of the collaborative papers in 2020 in both top-tier and lower-tier journals. However, in top-tier journals, articles by same-gender teams are still more prevalent. This disparity has been reduced recently but has not yet been reversed.

In terms of race/ethnicity, cross-racial/ethnic collaboration also demonstrates an upward trend. As indicated by the red-filled area in subplot C, cross-racial/ethnic teams enjoy a publication premium in top-tier journals. International collaboration is also becoming increasingly pervasive and conducive to publication in top-tier journals.

However, North–South collaboration does not exhibit a similar rising trend. Throughout the studied period, North–South collaboration only accounts for 21–30% of collaborative works in top-tier journals, overshadowed by North-only or South-only collaboration. This finding underscores the difficulty in facilitating effective collaboration between the Global North and South.

Between-Journal Variation

The visibility of collaborative works varies across journals. As indicated by figure 4, political science journals on average have dedicated 24.4% of articles to research conducted by cross-racial/ethnic teams (including Anglo-European teams) and 18.8% to mixed-gender teams. Most (78.6%) journals are located above the diagonal line, implying that cross-racial/ethnic collaboration is more prevalent than cross-gender collaboration. Dot colors represent the co-disciplines. All journals co-listed in Communication and Public Administration categories appear in the top-right quadrant, being more acceptive of cross-identity collaboration.

Figure 4

Scatter plot of SSCI-indexed political science journals (2007–2020)

Notes: Journals are positioned according to the percentage of articles written by mixed gender teams (horizontal axis, log scale) and cross-ethnic teams (vertical axis, log scale). Dot size is determined by journals’ 5-Year Impact Factors in 2020. The top 10 journals with the highest mean values of 5-Year Impact Factors over time are highlighted in red. Journals that are no longer indexed by SSCI are shown in the minimum size. Colors are determined by WoS discipline categories. The horizontal and vertical dashed lines denote the average proportions of articles contributed by cross-racial/ethnic and mixed-gender teams, respectively, while the slant dashed line denotes the diagonal where cross-ethnic collaboration rate is equal to mixed-gender collaboration rate.

Furthermore, almost all top journals have published higher-than-average percentages of cross-identity collaborative works. This includes leading general journals such as the American Journal of Political Science, Journal of Politics, British Journal of Political Science, and the American Political Science Review. In addition, top-tier specialized journals accommodate larger proportions of cross-identity collaboration than lower-tier specialized journals. For instance, Political Analysis has dedicated 54% of multi-authored publications to papers written by mixed-ethnic teams, even though the most prominent cross-racial/ethnic collaboration type in PA is Anglo-European collaboration followed by Anglo-Asian collaboration. This tendency is also observed in Political Science Research and Methods. It is worth noting that JoD is the most inclusive journal for authors with racial/ethnic minority backgrounds. However, its single-authored tradition results in 71.1% of minority-led articles having only one author, placing it in the lower-left quadrant.

Evaluation Gap

The evaluation gap concerns the degree to which the citation impact of a specialty/topic is associated with the share of contributions from minority populations. To elucidate this issue, we unpack the homophily between author identity and research specialty at the discipline and topic levels. At the discipline level, we group journals by their cross-listed WoS disciplines, calculate the relative representation of minority scholars in each co-discipline, and visualize the results using a heatmap.

As shown in subplot A of figure 5, female scholars are highly concentrated in Women’s Studies, while racial/ethnic minority scholars and Global South scholars are more concentrated in Area Studies and Social Issues. Racial/ethnic minorities are also overrepresented in History. These minority-concentrated fields are all located in the lower half of the graph, implying that they are less cited than average. High-impact co-disciplines, such as Communication and Geography, are in various shades of blue. This finding further indicates that women and other minority scholars are less present in more rewarding specialties.

Figure 5

Citation impact and relative representation of women and minority scholars by co-disciplines and topic areas

Notes: Color indicates the degree of over- and underrepresentation of groups relative to their average proportions in all fields and topic areas. Details of topic composition can be found in online appendix table S6.

At the topic level, we also find a clear affinity between gender/ethnic identity and topic selection.Footnote ¹⁰ While Anglo-Saxon and European scholars tend to study elections (Topic 47), racial/ethnic minorities are more committed to studying the foreign policies of developing countries (Topic 1), national security (Topic 18), and authoritarianism (Topic 12). Women devote a larger proportion of their work to gender politics (Topic 4) and social movements (Topic 19), while men are relatively more focused on political economy (Topic 10). Global South scholars intersect with women and racial/ethnic minorities, who focus their research on the foreign policies of developing countries (Topic 1), social movements (Topic 19), and authoritarianism (Topic 12). In line with the findings regarding co-disciplines, minority-concentrated topics receive fewer citations than average. A further investigation (online appendix note S4) reveals that identity-topic affinity is more prevalent among women and racial/ethnic minority authors, while men and racial/ethnic majority authors are more ubiquitous across various topics and less likely to be trapped in the evaluation gap.

Citation Gap

To cast more systematic light on the issue, we implement ordinary least squares regression models to examine the extent to which the citations that an article obtains are associated with author identity and team diversity. As mentioned earlier, author identity can be quantified in FA-based and team-based ways. Models 1 and 2 in table 2 represent the FA-based and team-based effects, respectively. Most coefficients are consistent across these two models in terms of relative effect sizes, implying that our results are robust to different specifications. For the sake of clarity, we mainly report model 1’s results. Additionally, author identity and team diversity might have heterogenous effects depending on authors’ affiliations—historically excluded authors experience a reduced disadvantage once they make their way to top universities. To address this heterogeneity, we employ six additional models that focus on a subset of articles led by authors from the top 10, 20, 30, 40, 50, and 100 universities. University rankings are established based on team-based Affiliation Prestige, as described in the Data and Measures section. Due to space constraints, we present only two of these six models in the main text, while providing details of the remaining models in online appendix table S7. The racial/ethnic composition of the top 15 universities is displayed in online appendix figure S3.

Table 2.

Regression models predicting article citations

Notes: Dependent Variable: Log₁₀(#Citations_i+1). As this variable was positively skewed, we applied the logarithm transformation to conform it to normality. To ensure that all entries can undergo log transformation, we add 1 to each observation, thereby avoiding negative and zero values. Since the models include gender-related variables, they are largely confined to data from 2007 onward, given the large volume of missing values in WoS’s author first name column before 2007.

First, the models corroborate that author identity is an important source of variation in citation impact. Articles authored by women tend to receive more citations than those authored by men when journal prestige is included as a control, and when the intersection of gender, race, and affiliation is disaggregated. In other words, when published in the same journal on the same topic, women’s work may accrue more citations than men’s work.

Unlike female authors, racial/ethnic minorities and Global South authors are in a clearly disadvantaged position in garnering citations. The consistently negative coefficients indicate that articles authored by racial/ethnic minorities and Global South scholars tend to receive fewer citations than those authored by Anglo-Saxon and Global North scholars, even when published in the same outlet and focusing on the same topic. The largest citation gaps emerge among Hispanic/Latino (B_{Hispanic/Latino} = -.075, t = -10.8) and Slavic authors (B_Slavic = -.073, t = -9.9). Articles led by Hispanic/Latino or Slavic authors receive 17.2% and 15.9% fewer citations than those led by Anglo-Saxon authors.

The model also affirms that cross-identity collaboration is a potent cure for the citation gap. Cross-national and cross-racial/ethnic collaboration can inspire impactful research (B_{cross-national} =. 088, t = 14.8, B_{cross-racial/ethnic}=. 016, t = 3.2). All else being equal, cross-national teams attract 22.5% more citations than domestic teams, and cross-racial/ethnic teams attract 3.8% more citations than racially homogeneous teams. Collaborating with outgroup members is an effective strategy for historically excluded scholars to break through the citation gap and expand their research impact to a wider audience. However, if the cross-national team comprises scholars from the Global South and North, the positive effect of internationality will be undercut. South–North teams, which are cross-national by nature, receive only 6.4% more citations than domestic teams, lower than the citation premium (22.5%) for cross-national teams within the Global North or South.

Additionally, although we have seen a growing trend of mixed-gender collaboration, the regression results suggest that the citation disparity between mixed-gender and same-gender works is not statistically significant. This result is congruent with Maliniak, Powers, and Walter (Reference Maliniak, Powers and Walter2013), who find that mixed-gender teams do not outperform same-gender teams in terms of citations received in studies of international relations.

By examining the cited references, we find that both the quantity and the diversity of references positively impact citations. If reference diversity, i.e., %Outgroup References, increases from 0% (fully ingroup) to 100% (fully outgroup), an article’s citation count will increase by 15.9%. Relatedly, reference quantity, i.e., #References, is also conducive to citation impact (B_#references =.003, t = 49.1). Citing more references and citing broadly beyond the group boundary both can attract wider attention and more recognition.

Finally, models 3 and 4 reveal that author identity and team diversity can have varying effects depending on the authors’ affiliation. In model 3, the impacts of author identity (both gender and race/ethnicity) decrease to insignificant levels and all indicators of team diversity diminish in statistical significance within the top 10 high-status universities. Nonetheless, when we expand our investigation to include the top 50 universities, we observe that most variables related to author identity and team diversity regain statistical significance, with the exception of articles by European scholars. This means that even African/African American, Jewish, Slavic, Asian, and Hispanic/Latino scholars within the circle of top 50 universities face a significant citation gap with Anglo-Saxon scholars. Model S4 in online appendix table S7 further substantiates that European-origin scholars are the least vulnerable to citation gaps and their citation disparity with Anglo-Saxon scholars is statistically insignificant in the top-100 circle. In contrast, scholars of African descent appear to be the most vulnerable to citation gaps, who face a statistically substantial citation gap, even when they secure affiliations with the top 20 universities.