Political Science in Latin America: A Scientometric Analysis

Studies of Latin American political science are centered on analyses of national cases and are based on intellectual narratives about the institutional development and teaching of the discipline. In this article we take a different, more comprehensive and comparative approach. We distinguish and contrast two Latin American political sciences: one native to Latin America and published in journals from five countries in the region, and one foreign and published in journals from five countries outside the region. We use three types of bibliometric measures: 01. reciprocal citations in 23 academic journals indexed in the Scopus database; 02. co-occurrence of terms in titles and abstracts of 5,880 research articles published between 2006 and 2018; and 03. cocitations of authors in the bibliographies of said articles. The network of journals forms a well-defined archipelago with three clusters separated by language (English, Spanish and Portuguese). The community is divided along two main axes: political approaches versus sociological approaches. The study also points out that themes, authors, and methodologies are not significantly different between these two political sciences.

Studies of Latin American political science are centered on analyses of national cases and are based on intellectual narratives about the institutional development and teaching of the discipline. In this article we take a different, more comprehensive and comparative approach. We distinguish and contrast two Latin American political sciences: one native to Latin America and published in journals from five countries in the region, and one foreign and published in journals from five countries outside the region. We use three types of bibliometric measures: 01. reciprocal citations in 23 academic journals indexed in the Scopus database; 02. co-occurrence of terms in titles and abstracts of 5,880 research articles published between 2006 and 2018; and 03. cocitations of authors in the bibliographies of said articles. The network of journals forms a well-defined archipelago with three clusters separated by language (English, Spanish and Portuguese). The community is divided along two main axes: political approaches versus sociological approaches. The study also points out that themes, authors, and methodologies are not significantly different between these two political sciences. Keywords: Science mapping; bibliometric methods; theoretical traditions; political science journals; Latin American Political Science. attests to how the discipline has been adopting an increasingly reflexive attitude (BULCOURF, KRZYWICKA and RAVECCA, 2017). In 2019, Ravecca published 'The Politics of Political Science: Re-Writing Latin American Experiences'. His multimethod research is a good example of this growing capacity for self-analysis.
However, the academic work amassed thus far relies more on national case studies 3 than comparative analyses. Altman (2005) showed that the comparative dimension in collective works in fact resulted from overlapping PS studies in each individual country. Such is the dominant perspective 4 both in the volume compiled by Freidenberg (2017b) and Casas-Casas and Mendéz (2015).
A second characteristic of the works published on this type of PS is the type of analysis employed. In general, the predominant approaches are descriptive and based either on intellectual narratives, or on an analysis of the institutional construction of Latin American PS research and teaching in specific countries (BARRIENTOS DEL MONTE, 2015;BULCOURF and CARDOZO, 2013;HEISS, 2015;MARENCO, 2015).
This article proposes a different approach to Latin American political science -more comprehensive and comparative -based on scientometrics. In summary, "scientometrics is the study of the quantitative aspects of science as a discipline or economic activity. It is part of the sociology of science" (TAGUE-SUTCLIFFE, 1992, p. 01) 5 .
Our goal is to describe the institutional, conceptual, and intellectual structures of this scientific field. We focus on material published at the beginning of the 21st century. We intend to describe the contemporary situation of PS academic production in/about Latin America. We want to answer three questions: 01. how is this 'scientific community' organized (i.e., what are its divisions, connections and hierarchies)?; 02. what is the 'thematic and methodological structure' of Latin American political science?; and 03. who are the 'reference authors' of the discipline in the region? ______________________________________________________________________________________________ 3 For example, for the case of Argentina, see (D' ALESSANDRO, MEDINA and LEIRAS, 2015); for Brazil, (CARDOZO, 2016); for Chile, (GATICA, 2012); for Colombia, (LEYVA and RAMÍREZ, 2015); and for Mexico, (BARRIENTOS DEL MONTE, 2017). 4 The essay by Bulcourf et al. (2014) is one exception as it historically contextualizes the construction of PS in Argentina, Brazil, and Mexico. Barrientos del Monte's book (2014) is another exception to the rule. The author conducted an in-depth historical analysis of the intellectual tensions that shaped the discipline in the region with information from almost 20 countries. 5 For the definition of scientometrics as a science map, see Mingers and Leydesdorff (2015). Scientometric Analysis (2020) 14 (3) e0007 -4/35 We analyzed citation patterns, methodological approaches, research themes, and influential authors in 5,880 articles published between 2006 and 2018 in 23 journals indexed in the Scopus database. We made use of bibliometrics to handle this diversity of journals and the volume of research articles. Bibliometrics, in brief, is the use of quantitative analysis for a given scientific production. Through the analysis of its metadata, we may determine recurring patterns within a knowledge structure.
In addition to 'citation analysis', an important parameter for revealing scientific impact and establishing scales of power and prestige in a community, two other types of bibliometric measures are used in this article: 'co-word analysis' and 'co-citation analysis'. These enable us to build bibliometric networks and map the conformation and connectivity of a given research domain.
Bibliometrics is therefore a powerful analysis tool for mapping the scientific field (i.e., scientometrics) 6 .
The article is divided into five parts. Section two summarizes the characteristics of Latin American political science revealed by the literature. Section three explains the research design, the inclusion and exclusion criteria for journals and articles, and details the meaning and function of the bibliometric measures.
Section four presents five bibliometric networks and, through them, the characteristics of PS scientific production. In the last section, we summarize our findings based on the potential of the scientometric approach to expose and explain the knowledge structure of PS. We emphasize the limitations of the bibliographic coverage of our data source (Scopus), the possible bias resulting from the unit of analysis (research articles) and the problems arising both from the type of research (quantitative) and from how we explored and exposed the data (bibliometric network visualizations). Finally, we address some shortcomings in our study, which suggest further research possibilities.

Bibliographic background
Three major characteristics comprise the academic production of Latin In the absence of comparative policy analysis, the delimitation of its research objects is parochial. Chasquetti (2017) found that only 20% of the articles published in six major Spanish-speaking journals were comparative 7 . Basabe-Serrano and The third characteristic of Latin American PS is its power of self-analysis. One example is a study by Salatino (2017), who developed a wide-ranging analysis of 306 journals published in 14 countries in Latin America. As he sought to explore the discipline's 'scientific communication structure', he concluded that most academic journals in the region were under-institutionalized.
Articles in journals are not the only form of scientific communication in the social sciences, but they may serve as an indirect indicator of the state of a discipline.

Research corpus and bibliometric methods
The aim of this article is to provide an analytical map of the 'underlying structure' and the institutional, conceptual, and intellectual 'interaction dynamics' of PS in the region.
Our study use 'research articles' to analyze Latin American PS standards. Our methodological assumption is that citing and cited articles are in some way related to each other (yet to be determined). In Garfield's classic article (1955), he suggests that this linkage reveals some kind of 'association of ideas' between them. Mingers and Leydesdorff (2015) go a step further and assert that "The act of citing another person's research provides the necessary linkages between people, ideas, journals and institutions to constitute an empirical field or network that can be analyzed quantitatively" (MINGERS and LEYDESDORFF , 2015, p. 01).
We describe the research design, the characteristics of the data source, the selection criteria for the 23 journals and 5,880 documents analyzed, the bibliometric measures, and the software for viewing the citation patterns. We used a distancebased approach to establish networks of journals, terms, and authors. These networks provide, through two-dimensional maps, a succinct representation of the scientific field.

Database
In the selection of Latin American PS journals, we opted to analyze titles indexed in Scopus. Journals in the Scopus database are the most important in their fields of knowledge and obey precise indexing criteria (such as: peer review, diversity in geographical distribution of editors and authors, academic contribution to the field, citedness of journal articles in Scopus, clarity of abstracts, etc). Although Scopus generally has more limited coverage, the database is far more comprehensive for scientific publications in the region than Web of Science.
Nonetheless, there are some important limitations regarding the representativeness of the research universe.
There are more than 25,100 indexed titles (circa 23,500 peer-reviewed journals) in the Scopus database. Regional coverage, however, is concentrated in Western Europe (more than 12,000 titles) and North America (about 6,600 titles).
Distribution between subject areas is also uneven. While health and life sciences are responsible for 46% of the 25,100 titles and physical sciences account for 28%, the entire area of social sciences and arts and humanities comprises only 26% of the database (Scopus, 2020).
In We present below our methodology for selecting the 23 journals in this study.

Inclusion and exclusion criteria
Latin American social and political science journals are not thematically specialized. Most are interdisciplinary and publish papers ranging from social history to economics, or from sociology to public administration.   After selecting the collection of documents published between 2006 and 2018, we chose only the research articles (i.e., we excluded editorials, book reviews and research notes). In order to avoid overtly varied subjects in the articles, or too far afield from PS, we refined the selection of texts in the 'Search within Results' to those that contained the term 'politic*' in their abstracts. The asterisk (*) is a Boolean logical operator that allows for the inclusion of similar words such as

Bibliometric methods
Research articles are the base unit of this study and we observed them from three dimensions: 'journal unit', 'term unit' (words or noun phrases) and 'author unit'. We used four bibliometric indicators: 'citation analysis', 'journal co-citation', 'word co-occurrence', and 'author co-citation'.
Citation analysis is the most common measure in bibliometrics. The unit of information is the number of mentions of an item (TODESCHINI and BACCINI, 2016, p. 49). The relative importance in terms of impact, relevance, influence, and visibility of an author, article, journal, or institution is measured by counting the number of times that one of these items is mentioned in scientific papers. This is a good proxy for the hierarchy of the field of knowledge. Co-word and co-citation are semantic and thematic similarity measures between two documents. They estimate the interdependence between two elements (words, noun phrases, authors, etc). These measures are expressed mathematically through networks of relationships.
We evaluated the reciprocal citations between the 23 journals to discover clusters by thematic, theoretical, linguistic, and geographical affinities and the place   Observe the frequency of direct citations of an item in the corpus of the study to measure its influence.
The number of times an item (journals, terms, authors) is cited is an objective indicator of its relevance or impact in a scientific domain.
Facilitates the ranking by order of importance of a bibliographic element, as defined by practitioners of a discipline or scientific field. journals cocitation cited journals journal titles The relationship between items is determined based on the number of times journals are reciprocally cited (McCAIN, 1991).
Identify the formation of informal communities in the studied area.
The proximity between journals can reveal the institutional structure underlying a scientific field and, by extension, the recurrent practices of a discipline.
co-word recurring words words in the articles' titles and abstracts The relationship between words and their binding strength is determined based on the number of times they occur together in articles (CALLON et al., 1983).
Discover the conceptual framework of a scientific area.
The intensity of the links between the terms and groups of terms in a network is an indicator of their thematic, theoretical, or methodological affinities.
Measures the strength of association between terms and, by extension, between themes, problems, concepts, methods, and research techniques in common as coordinated in a field. author cocitation cited authors bibliographic references The relationship between the items is determined by the number of times documents/authors are cited together in the reference list of articles in the study corpus (White and Griffith, 1981).
Point out the different intellectual traditions that shape the structure of a given study area based on the most relevant authors.
The frequency in which two bibliographic references are cited indicates that they belong to the same school of thought.
Reveals, through the most co-cited references, the intellectual basis and the reference authors in a given research field.

Bibliometric software
When working with a large volume of information it is advisable to limit the analysis of a scientific subject area to a single type of academic publication (research articles) and limit the analysis of articles to their metadata. Networks are a didactic way of presenting relations between this information. Bibliometric mapping allows us "to examine how disciplines, fields, specialties, and individual papers are related to one another. It produces a spatial representation of the findings analogous to geographic maps" (ZUPIC and ČATER, 2015, p. 429).
The software used for bibliometric calculations and the presentation of the networks was VOSviewer version 1.6.12. The name of the software comes from

Results and discussion
We present the results of our bibliometric measurements through two types    Portuguese-speaking community, and Revista de Ciencia Política in the Spanishspeaking community.
Finally, betweenness centrality measures the intermediation capacity of a journal and how much it can function as a bridge or shortest communication path between the other nodes. We may think of this as a measure of strategic position in the network and thus of influence. Revista de Ciencia Política is the major connecting node of the entire graph (0.029) followed by the Journal of Latin American Studies (0.027). In the Brazilian cluster (blue), Dados (0.020) is the connecting node.

Reciprocal affinities: the conceptual map of Latin American Political Science
Co-words analysis is a type of content analysis that uses the co-occurrence of two terms in a corpus of documents. Three measures are important: 01. the presence/absence of certain words; 02. the frequency that these words appear in this corpus; and 03. the proximity between them. There are three assumptions in this type of analysis: 01. that the words in titles and abstracts aptly describe the content of the articles; 02. that a word that occurs simultaneously in two (or more) scientific papers is a valid link between them; and 03. that the occurrence of the same words or the same word pair is very often an indicator of the presence of a research topic, topic of interest, or a form of methodological approach shared by the scientific community.
From this point forward, we present the networks separated by groups of journals from Latin America and about Latin America. Figure 02 and Figure 03 show the strength of association between the most frequent terms in the two political sciences and allow us to recognize concepts or ideas that theoretically and methodologically structure the literature. We can conclude that the large area in the center and towards the left of Figure 02 can be broadly referred to as 'political sociology' and the area on the right as 'mainstream political science' 12 . The conclusion is dependent on the perspective of the articles we selected and filtering for 'politics'. We may dray five conclusions from this bibliometric map: 01. there is a clearly-identifiable research agenda that deals with more traditional problems and ideas of PS (green cluster, to the right of the map), such as the Judiciary (the Courts), the Executive (the Presidency), and the Legislature (Congress) or voting, parties, and elections; 02. methodologically, mainstream political science deals with data, evidence, hypotheses and variables; 03. this research community is more insular as revealed by the proximity between its terms/themes and more distant from the other items in the network; 04. on the other hand, the themes, problems and concepts of the other seven clusters, in the center and on the left in Figure 02, are more varied and much more 'sociological'. There are a few special sociologies: economic sociology (yellow cluster), sociology of violence (blue), sociology of inequality (lilac cluster), sociology of gender (brown), and so on. In a more peripheral position on the network is the orange cluster, which contains topics associated with '21st century socialism' (Bolivia, Venezuela, left, socialism, etc); and 05. the red cluster lists the preferred methods of these sociologies: discourse, paradigm, narrative, and image. Figure 03 is the visual representation of the network of occurrences of terms in 3,692 articles from journals 'from' Latin America. The overlapping of the two figures shows that in both types of journals, intellectual production is structured around the divorce between a mainstream PS (green cluster, right) and political sociology with its typical objects of study (woman, violence, social movements).
Furthermore, the divide of methodological approaches is the same as in Figure 02.
The light blue cluster, on the right and closest to the green cluster of 'pure' PS, comprises terms such as 'data', 'evidence', 'indicator', and 'survey'. On the other hand, the red cluster on the left encompasses terms such as 'theory', 'concept', 'discourse', 'narrative', and 'history', suggesting that there is a 'qualitative' focus along this axis.
______________________________________________________________________________________________ 12 Political sociology, under the definition of Sartori (1969), is an interdisciplinary hybrid that combines the analysis of political facts, variables typical of sociology (social structures and processes), and variables typical of political science (political institutions). Looking at the word networks (Figures 02 and 03) and authors (Figures 04 and 05), we detect the presence of terms/authors more associated with sociology, but whose object is politics. Hence, 'Political Sociology'.
But what does the data tell us about the differences between these two groups? Table 06 lists the most frequently occurring terms in the two networks, which allow us to highlight the diverging research agendas between the two Latin American 'politologies'.

Divergent disciplines: the intellectual map of Latin American Political Science
Maps based on author co-citation analysis (Figures 04 and 05) express the number of times that two authors are cited together in the bibliographic references of the citing articles (WHITE, 1981). In Small's classic definition, "cocitation is the frequency with which two items of earlier literature are cited together by the later literature" (SMALL, 1973, p. 265). Therefore, this bibliometric measure looks at the discipline's past or its base literature. This allows access to schools of thought or intellectual traditions that configure this scientific field (WHITE and GRIFFITH, 1981).

Conclusions, limitations and lines of future research
The main contribution of this article was to present an overview of Latin American Political Science, as a countercurrent to the dominant analyses in the area focused on national case studies and the reconstruction of the discipline's historical institutionalization process. We explored differences and similarities from a comparative analysis between the native PS of Latin America and the foreign PS. One concern was to establish a more recent time frame (2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018) to encompass the events of the past decade. Another novelty was to apply bibliometric measures combined with network analysis to handle a large volume of sources, authors, and references in order to detect the patterns in this literature.
By measuring the distance and proximity between journals, we access the underlying institutional structure of this scientific field. By using academic journals as markers, we single out the existence of informal, feudalized communities and the reasons for their existence. Through the map of co-occurring terms in titles and abstracts, we express the dominant research themes and methodological approaches. Lastly, the map of the cited authors in the 103,334 bibliographic references of the 23 journals from ten different countries indicates the intellectual traditions that structure this field. Table 08 presents the main findings and answers the three questions formulated in the Introduction. This is still an exploratory, descriptive, and comparative study of a large number of items. The type of unit (journals and papers), the investigation technique (bibliometrics), the treatment of information (quantitative) and its exclusive source (the Scopus database) present some limitations. These limitations include the representativeness of the universe and the validity of the measure to select political science articles, in addition to the specific inaccuracies derived from scientometric approaches. Organized in the form of an archipelago: three well-defined journal communities separated by the language of publication (Portuguese, English, and Spanish); these communities have more connections with each other than between any two of them. The Conceptual Framework Attested the existence of two major divisions -thematic and methodologicalthat structure the production of Latin American political science: political science and quantitative methods, on the one hand, and political sociology and qualitative methods, on the other. The Intellectual Structure This division between political sociology and political science is partially confirmed and replicated through the analysis of authors' citations. The difference here is that journals published outside Latin America typically cite more authors from political science; journals published in Latin America cite more sociologists and social philosophers, revealing the importance of the area of political and social theory. Source: Prepared by the authors.
We worked solely with the metadata of articles, excluding, for example, books. Furthermore, due to the nature of the data and the type of treatment, it was not possible to discuss the substantive content of these articles. Thus, we tried to calculate this content indirectly, based on recurring terms in titles and abstracts. But titles are not always accurate or very descriptive of the themes of the articles, and abstracts can be very vague, not detailing, for example, methodologies of analysis. If used in isolation, scientometrics may not be the most appropriate technique for mapping production in the social sciences. Journals both 'from' and 'about' Latin America are publications that include articles on sociology, anthropology, history, social theory, economics, and law studies. Given their multidisciplinary nature, the number of documents selected in the first screening (10,838) and the lack of standardization of the structure of the articles, it was impracticable to use manual methods to identify research articles strictly within the field of political science. We adopted an ad hoc solution to select PS papers. It is thus possible to question the validity of our method: to analyze articles where the term 'politics' and its variants ('political', 'politician') appear in abstracts. The corpus (5,880 articles) may be overestimated. Henceforth, we may consider an agenda encompassing a longer period, more countries, a more representative number of journals, and a type of analysis able to access and explain the content of articles and other bibliographic sources. Adding other academic journals would significantly increase the reliability of our findings and allow us to more confidently express how this field of knowledge is hierarchized, enabling the formulation of explanatory hypotheses. An expansion of the timeframe spanning four decades would allow us to analyze the conceptual dynamics, thematic transformation, and the change (or not) in the reference base of