The Gendered Division of Labor in Brazilian Political Science Journals

This article analyzes the gendered division of labor in Brazilian political science. We seek to answer two questions: what are the predominant topics in political science that are being published in the discipline’s journals? How are women and men’s authorship distributed in these journals? The methodology involved three stages. First, we built a corpus with 2,363 articles that were classified as ‘political science and international relations’ by the Coordination for the Improvement for Higher Education Personnel (Coordenação de Aperfeiçoamento de Pessoal de Nível Superior – CAPES) and published in the most prominent Brazilian journals between 2005 and 2018. Next, we scraped abstracts and other bibliographic information from publications in the Scientific Electronic Library Online (SciELO) platform and used a topic modeling technique to identify the most recurrent topics. Finally, we associated the identified topics with the authors ’ gender. The data was examined based on two specific types of the gendered division of labor: the ‘horizontal’ and the ‘vertical’. Our results show that women and men as first authors tend to cluster around specific topics (horizontal division), but we did not find a tendency in journals to reject works on the topics in which women are better represented. In other words, differently from what was found by the international literature, the Brazilian journals in our sample do not seem to grant a lower status to these topics (vertical division). It is noteworthy, however, that men are the majority of first authors in all topics, including feminism.

information on the composition of this community (CANDIDO, FERES JUNIOR, and CAMPOS, 2019;MENDES and FIGUEIRA, 2019), it is still unknown whether journals' biases against specific topics could be hindering the inclusion of women in the Brazilian academic community. To address this issue, we examined articles that were published in the most prestigious political science journals. This examination is justified for two reasons. First, the circulation of articles is particularly important to advance the career of Brazilian political scientists, not only because publication rate is an important criterion for research productivity assessments, but also because it is central for disseminating research results (LEITE and CODATO, 2013). Second, the examination of these journals allows us to identify the most prevailing topics in published articles and to assess the female representation within these topics.
Scholars dedicated to the examination of the gendered division of labor faces the double challenge of determining which topics are to be considered 'male' or 'female' and measuring their respective prestige. In that respect, some authors stress that less cautious definitions tend to associate what is masculine with what is most socially valued (GOODWIN, BATES and McKAY, 2020;PANSARDI and VERCESI, 2017). There are numerous ways to categorize academic roles and fields, and that is done with the use of controversial and discretionary criteria. To give one example from within the political science community, it is usual for quantitative research and 'hard' approaches to be associated with the masculine, while qualitative research and 'soft' approaches are commonly associated with the feminine and allegedly undervalued in the field. Even though such an association does exist as a general impression in academia, taking it as an analytical key could reinforce rather than challenge such a stereotype.
The distinction between 'vertical' and 'horizontal' gendered division of labor improves our understanding of the different mechanisms underpinning these inequalities. When we talk about the vertical division of labor, we refer to issues of prestige, power, and status that derive from several aspects of the work relations in a given occupational group, such as salaries, hierarchical positions, or participation in more prestigious subfields. Drawing on this last aspect of the vertical division, we identified the topics in Brazilian political science and looked into their circulation in the form of published articles in academic journals. A topic was defined as prominent when it is more regularly published in the best Brazilian political science journals. By contrast, those topics with lower acceptance will be treated as marginalized in the discipline.
The horizontal division, on the other hand, considers the clustering of men and women in different topics. To address the horizontal division, we considered the gender distribution of the first authors across topics. Except for the artic les that list the authors' names alphabetically, first authors tend to be research coordinators or experienced scholars in specific subfields. First, we consider the hypothesis of a strong horizontal gendered division, with women being concentrated on certain topics and men on others. Second, we consider the hypothesis of a vertical division, with a greater proportion of women working on topics that are less prestigious in the journals. It is noteworthy that both types of the gendered division of labor in authorship were found in the American academic community (KEY and SUMNER, 2019;TEELE and THELEN, 2017;YOUNG, 1995). How they occur in Brazil is still unknown.
This article is divided into four sections. First, we examine the literature on gender inequalities in political science. Second, we discuss our methodology, indicating our parameters for selecting the journals and explaining the technique used in the analysis (topic modeling). Third, we present the results and discuss the different types of the gendered division of labor among political scientists. Finally, we conclude by considering the general contributions of our research to the Brazilian and international debate.

Studies on gender inequalities in political science
The debate on gender inequalities gained particular attention in the discipline of political science at the turn of the 21st century, with the increased regularity of publications on the topic and the dissemination of empirical work across the world.
Such numerous forms of inequalities are noticed by those who suffer from them: women report more discrimination in professional interactions than men (ALLEN and SAVIGNY, 2016;AKHTAR et al., 2005;KANTOLA, 2015KANTOLA, , 2008. issues, concerns about making an impact on the field, lack of information about work practices, and lack of self-confidence. Lack of self-confidence is more often reported by women, who are also particularly affected by the demands of domestic and family life, by time constraints, by gender stereotypes, and by their deviation from the male standard, which is implicit in the best institutional positions. In her influential 'Theory of Gendered Organizations', Joan Acker (1990) postulates that professional institutions are underpinned by both an abstract notion of worker and a presumed sense of gender neutrality, which favor and reflect the male body and sexuality.  (BREUNING and SANDERS, 2007;CARPIUC, 2016;EVANS and MOULDER, 2011;KONIG and ROPERS, 2018;TEELE and THELEN, 2017;YOUNG, 1995;WILLIAMS et al., 2015). Researchers in peripheral countries replicated these studies and found a similar pattern of inequalities, suggesting that these inequalities exist regardless of the scientific communities' position in the geopolitics of knowledge (CAMPOS, FERES, and GUARNIEIRI, 2017;CARPIUC, 2016;CURTIN, 2013;FERNÁNDEZ, 2006;MENDES and FIGUEIRA, 2019).
Studies addressing the issue of gender gap usually focus on the authors' gender not only because academic journals are an important site of communication and prestige within the discipline, but also because their evaluation systems are presumed to be 'blind' to the social and political hierarchies among researchers. Used by the vast majority of the most qualified and renowned journals, the blind peer-review process has become hegemonic. While keeping the authors' total or partial anonymity, this model would in theory guarantee that intrinsic characteristics of scholars, such as gender or race, are not considered when assessing the quality and suitability of their scientific work.
The reality, however, is distant from the ideal. The gender gap in authorship not only reflects how job opportunities are unequally distributed in academia but, in   Kantola (2015), for instance, pointed out that shifts in government policies and university funding in Finland affected gender relations, with negative effects on women. The author, thus, identified a variable that is usually ignored by studies on inequalities in political science: the external interferences of science-related policies.
Fernández (2006), on the other hand, stressed similar conditions experienced by female political scientists in Chile and in other countries where the discipline was institutionalized during the democratic transition.
In Brazil, researching and teaching political science is essentially done at public universities, where candidates must pass a public competition to become a professor and start a career that, in comparison to private universities, has a high level of stability. Women continue to be professors possibly because of these particularities; however, mechanisms to tackle gender inequalities are not used to select professors in Brazilian universities. Women and men experience the same demand for productivity, which is measured mainly by their publication rates in prominent journals. Thus, despite possibly working double or triple shifts and dealing with maternity obligations, women must meet the same goals as men -while men have more time to advance their careers. As the theory of Joan Acker (1990) postulates, entering professional organizations entails adapting to an abstract notion of an ideal worker who, in reality, mirrors the life conditions of men.
Female professors in Brazilian graduate programs in political science are clearly and highly underrepresented, which is not seen in neighboring fields, such as sociology and anthropology. Women represent only 33% of the overall number of professors in political science, in contrast to 47% of the sociologists and 54% of the anthropologists (CANDIDO, FERES JÚNIOR, and CAMPOS, 2019). To promote more diversity in the field, it is crucial to identify its inequalities, which, in turn, require researchers to be in dialogue with similar studies carried out both in Brazil and other countries. Inspired by studies that draw an association between women's lower publication rates and topics in political science (KEY and SUMNER, 2019;TEELE and THELEN, 2017;YOUNG, 1995), we investigated how receptive the Brazilian political science journals were to different topics and whether the patterns we found could be associated with obstacles for women's inclusion -obstacles that have a high impact on their academic career.

Metodology
Our ______________________________________________________________________________________________ 5 We would like to thank Thyago Simas for his assistance in creating the algorithm for scraping data from the SciELO platform and André Félix for his assistance in using the tool to assign a gender to the authors' names. 6 The political science committee considered a list of journals that had published more articles by political scientists; their selected list of journals was given to us by members of this committee. Since most of these journals were included in the SciELO's portfolio in 2005, we selected those issues that were published between 2005 and 2018. We excluded articles without an abstract in English (usually, editorial notes, translations, interviews, etc) and those whose abstracts had less than 300 characters -such exclusion is justified because such small abstracts tend to accompany some type of non-scholarly manuscript. Furthermore, the topic modeling technique requires a certain number of words to function properly. Table 01 shows the distribution of the analyzed articles by year and journal.  We assigned topics to the articles in the corpus with a topic modeling technique, which is used to identify semantic patterns in scientific articles (taken as units of analysis) by forming groups of terms with most co-occurrence. We used the LDA package (Latent Dirichlet Allocation) in R programming language to bring certain documents together and, consequently, keep others apart (SCARPA, 2017).
Topic modeling operates in five stages. First, the set of documents to be analyzed is cleared of words without semantic relevance, keeping only the terms with some meaning. Numerals, pronouns, prepositions, words with less than two letters, etc are eliminated. Secondly, very common and very rare terms are excluded. This is necessary because none of these two classes of words contributes to the definition of minimally regular semantic patterns. The term 'article', for example, is so ubiquitous in the abstracts in our corpus that it would hardly help to characterize any topic; the same goes for a rare term like 'portfolio'. A third step consists of isolating the root words and eliminating the less semantically relevant suffixes and prefixes.
To estimate the number of topics for achieving the best clustering of articles in the corpus, we used the method based on the density of intersections via the LDA package, as proposed by Cao et al. (2009) -once again, in R programming language.
With this method, an algorithm extracts from the corpus a certain number of different topics. This method assumes that the ideal number of topics is that in which the words belonging to more than one topic reach the optimum level. In other words, a certain number of topics are considered to be excessive when most of them are characterized by a large number of identical terms, and, conversely, the number of topics is considered insufficient when most of them have terms so specific that the intersection is too small. Thus, the optimal number of topics is established when there are some intersections -but not many -between the topics. In our case, the optimal number is 15 topics.
We used two outputs of the topic modeling process in our analysis. The first output is a list of topics that were detected in the articles, followed by the ten most recurring terms in each of these topics, as shown in Table 02. From these sets of terms, we can make an interpretative determination of the topics that they represent, as shown in the third column ('assigned label'). The second output is the definition of the predominant topic for each of the articles included in the corpus. In accordance with the topic modeling technique, we chose to work only with the predominant topic, although each article may have its vocabulary associated with more than one topic. In the last stage of data collection and analysis, we did an automated classification of the first authors' gender based on a dataset of the Brazilian Institute of Geography and Statistics (Instituto Brasileiro de Geografia e Estatística -IBGE) that estimates the gender of names with a high degree of certainty. We considered only first authors not just due to technical limitations, but also because including all authors would most likely have produced a false balance, as women generally publish more as second or third authors (CAMPOS, FERES JÚNIOR, and GUARNIERI, 2017).
Because some ambiguous or foreign names were not recognized by the IBGE database, the coding had to be completed manually.

Results
As we can see in Graph 01, articles with a vocabulary associated with 'elections and political parties' predominated, corresponding to 9.6% of the corpus. In We found two surprising topics falling in intermediary positions. First, the number of articles focused on 'gender and feminism' (07%). Despite being a topic in expansion, it was late recognized in legitimizing spaces within the political science community, such as congresses and graduate programs (MATOS, 2016) 7 . A reason that makes this result even more significant is that political scientists who work on gender and want to publish their work have the option of submitting their work to journals that were not included in our sample, journals that have good ratings on CP&RI Qualis and are specifically dedicated to gender issues, such as 'Revista Estudos Feministas' ______________________________________________________________________________________________ (REF) and 'Cadernos Pagu' 8 . The topic 'public opinion and media' was predominant in 6.1% of the articles, contrasting with findings from Antonio Teixeira de Barros and Lucas Silva (2017), who examined a larger set of journals and indicated that political scientists have a low engagement with this topic.
Secondary topics, on the other hand, have rather unclear publication patterns.
'State and nation' is a topic with articles focusing on the formation, consolidation, and operation of national states -a lingering topic that is not significantly addressed by the political science community. Studies on 'collective action and social movements', in turn, have experienced rapid growth in recent years (SZWAKO, DOWBOR, and ARAÚJO, 2020), but they still are a relatively minor topic in political science. It is noteworthy that this topic was 'split' into two separate groups in the topic modeling process, also creating the topic 'political participation'. In the topic list, positioned between 'collective action and social movements' and 'political participation', we have 'quantitative methodologies ', a topic that has been increasingly discussed within the discipline, but still secondary compared to rather established topics.
We will now use the sequence of the assigned topics depicted in Graph 01 to hierarchically classify the topics in political science, that is, 'elections and political parties' is the most prestigious topic, followed by 'international relations and foreign policy', and so forth. After we verified the gender balance within each topic, we will use this hierarchical classification to assess the vertical division of labor.
Of the overall compiled articles, 30.2% had a woman as the first author, while 69.8% had a man in the lead. Over the period considered in the analysis, the proportion of women as first authors increased very slightly.   Review' also has a relatively well-balanced set of topics, although topics such as 'political economy', 'elections and political parties', and 'public policies' were privileged.
Graph 04 shows the horizontal division of labor, which is defined by the gender ratio of authors within each topic. One can see that the gender ratio varies considerably across topics. Articles with topics such as 'state and nation' and 'quantitative methodology' had about 20% of women as first authors, while articles focused on 'gender and feminism' or 'law and justice' are closer to parity, with more than 40% of articles with women as first authors. Still, it is noteworthy that, even in articles focused on gender inequalities and feminism, women are behind men in authorship. It is also true that articles addressing feminist issues may have also been linked to other topics during the topic modeling process (such as 'political theory', 'collective action', or 'political participation'), depending on the vocabulary used in the abstracts.
Equally important is the concentration of women in studies on 'law and justice'. That was a surprising result -it was expected that the representation of women would be higher within the topic of 'gender and feminism', given that women not only played a larger role in the genesis of this topic and in recent assessments on its progress, but they also publish more than men in the journals specially dedicated to gender and feminism (DINIZ and FOLTRAN, 2004;DION and MITCHELL, 2019;KEY and SUMNER, 2019;MATOS, 2016;REID and CURY, 2019;YOUNG, 1995). While the results indicate a clear horizontal division of labor, the same is not true for the vertical division. Graph 04 shows that gendered patterns of publication vary between topics; it also indicates that women are not necessarily overrepresented among scholars in those topics with a lower status in the journals. The topic with the worst publication performance, as shown in Graph 03, is 'state and nation', which also has the lowest proportion of women as first authors. In contrast, works on 'public policies' and 'gender and feminism' -topics with a substantial percentage of women as first authors -are regularly published.
To capture the potential effects of the journals ' different editorial processes and proclivities, we examined the gender ratio of the first authors in the eight journals of our sample (Graph 05). The results show that the gender ratio in each journal does not deviate much from the average (30.2%). These results suggest that the average gender imbalance that was found is unlikely to be the result of discriminatory procedures of one or more journals that were examined. With such findings, it is convenient to ask whether the representation of women in each journal was not being affected by the journal's preference for specific topics 9 . Graph 06 shows, for each journal, the relationship between the percentage of articles authored by women (y-axis) and the percentage of published articles on the five topics in which women publish more: law and justice, gender and feminism, collective action and social movements, public policies, and political participation (xaxis the results suggest that the inclusion of female authors seems to follow the journals' receptivity to the five topics in which women are better represented.
Graph 06. Percentage of articles authored by women by the percentage of published articles on the five subfields with greater female representation in journals Source: Elaborated by the authors.

Conclusions
Political science has not only developed as one of the most male disciplines within the social sciences, but it has also been slow in incorporating gender perspectives in the discipline's studies (ACKER, 1990;BIROLI, 2016;CARPIUC, 2016;FREIDENBERG, 2018;MATOS, 2016 are interdisciplinary and generally unappealing to political scientists, which could explain why established political science journals are increasingly publishing works on gender. American political science, in contrast, has journals such as 'Women and Politics', one of the very few prestigious journals in the discipline with a majority of female authors (YOUNG, 1995).
It is also important to consider the relationship between the relative prestige and recognition associated with each topic in the discipline and the gender balance within it. If we take the average percentage of women as first authors as a threshold (30,2%), we see that except for 'public policies', the topics above this threshold (i.e., those where female authors are better represented) are less established topics, such as 'right and justice', 'gender and feminism', 'collective action and social movements', and 'political participation'. While the topic 'executive-legislative relations' practically meets the threshold value, the rest fall below it. Some of the topics in which male ______________________________________________________________________________________________ overrepresentation is higher are also those that are better established in the discipline, such as 'international relations and foreign policy', 'Brazilian political thought', 'elections and political parties', 'political theory', 'public opinion and media', 'political economy', 'quantitative methodology', and 'state and nation'. Except for the quite specialized topic 'defense and security', all the other topics that fall below the threshold are, so to speak, mainstream, as they have been present in the discipline's curriculums and research agendas for decades, both in Brazil and in other countries.
In conclusion, we identified two different publication patterns associated with the gendered division of labor in Brazilian political science. Although Brazilian political science journals are frequently publishing works from those topics in which women are more engaged ('gender and feminism', for example), the gendered division of labor is still thriving in the Brazilian political science community -with men far in the lead in mainstream topics. Once our results have suggested that Brazilian journals do not give preference to publishing overly male topics of research, the causes of the gender gap in publication rates for men and women and the solutions to tackle it remain to be found.