Gendered patterns of global scholarly migration: Evidence on mobile researchers from bibliometric data

Highly skilled people are the most mobile population group worldwide in the current era of knowledge-based economies. The links between scholarly mobility and globalization are of special interest as these facilitate the exchange of information. Such knowledge production greatly benefits national and international science (Netz et al., 2020; Robinson-Garcia et al., 2019). However, we assume that gender disparity, the so-called glass-ceiling, whereby female scholars advance less quickly than their male peers, exists in the diaspora of the highly skilled, as elsewhere in academia (Huang et al., 2019; Larivière et al., 2013). The exploration of the intersection of gender and international scholarly mobility is a critically important step for pursuing a more balanced gender representation in academia and ensuring the sustainability of academic careers for researchers, regardless of gender. There is very little research in this area to assess this and, if necessary, propose redress. Hence, this study analyses the interplay between the globalization of scientific knowledge, internationalization of academia, and gender inequalities.
This study is unique in creating a robust computational methodology to offer a nuanced view of gender inequalities and dynamics in scholarly mobility across different countries over time. It relies on large-scale bibliometric data of over 33 million Scopus article and review publications from an exhaustive set of 10 million published researchers worldwide between 1996 and 2020. The pre-processed bibliometric data allows us to analyze the global scholarly mobility over time, identify the internationally mobile researchers and differentiate their gender using a systematic process composed of name-gender dictionaries and deep learning methods. This novel methodology allows an investigation of researchers’ respective mobility trajectories over four time periods between 1998 and 2017. For a greater depth to the understanding of global scholarly migration, this study analyzes the evolution of gendered patterns of international mobile researchers over time. We map changes in direction, intensity, and diversity of scholarly migration worldwide. The emigration, immigration and net migration rates of both female and male researchers across different countries are further quantified to provide more solid evidence for the observed patterns, with important administrative and policy implications.
Data and methods This study relies on large-scale bibliometric data of over 33 million Scopus article and review publications from an exhaustive set of 10 million published researchers worldwide between 1996 and 2020. The cleaned and pre-processed bibliometric data allows us to analyze the global scholarly mobility over time. The residence country of each year is assigned to each
author according to their most frequent (mode) country(ies) during the publishing year and that of the most recent year of publishing. Accordingly, based on their migration events or lack thereof, we can identify the migrant researchers as well as their migration trajectory. To reduce the effect of yearly fluctuations, we use four time periods (1998-2002, 2003-2007, 2008-2012, and 2013-2017) to group the migration events according to the migration year, to further investigate the temporal trends of the international gendered migration in academia over 20 years. The only metadata that can be used to indicate an author’s gender from Scopus is the author’s name. We elaborated a three-step process to infer the genders of researchers by combining 1) a large dictionary of names and genders i.e., worldwide gender-name dictionary (WGND) which includes 6.2 million names from 182 different countries, 2) a gender detection tool Dempgraphicx based on a deep learning BERT embedding model with sub-word tokenization, and 3) an application programming interface (API) called genderize.io. By applying our name-based process in name-gender dictionaries from different countries, we validate that the name-based process we design works quite well for almost all tested countries, covering western countries and Asian countries in our evaluation. The precision rates and recall rates are over 90% in these countries. China is an exception, having a large number of names with uncertain genders in the original test dataset. Finally, we obtained the genders for 89.27% of author profiles in our dataset.
The emigration spread (ES) and immigration spread (IS) in terms of the variety of origin and destination countries respectively measure the extent to which the mobile researchers are dispersed across destination countries, and the extent to which they come from a diverse array of origin countries (Czaika & de Haas, 2014). Specifically, ???????????????? (Eq. (1)) and ???????????????? (Eq. (2)) measure the extent to which bilateral emigration flows (????????????????????) among total emigration flows (????????????????) from country ???? and bilateral immigration flows (????????,????????????) among total imigration flows (????????????????) to country ???? are diverse given a period of time ???? on a country level (???????? is the number of countries in each period).
????????????????=1−Σ(????????????????????????????????????)2????????????=1 (1)
????????????????=1−Σ(????????????????????????????????????)2????????????=1 (2)
On the global level, ???????????????????????????????????? (Eq. (3)) and ???????????????????????????????????? (Eq. (4)) are calculated in a similar way to measure the extent to which the total global migration flows of researchers are dispersed across destination and origin countries, respectively among all the migration flows ????????during the period ????.
????????????????????????????????????=1−Σ(????????????????????????)2????????????=1 (3)
????????????????????????????????????=1−Σ(????????????????????????)2????????????=1 (4)
Preliminary findings
The number of female and male published researchers worldwide witnessed considerable growth during the past decades. The result shows that the figure of female mobile researchers rose from 29,161 (4.3% of all published female researchers) during 1998-2002 to 79,370 (4.6%) during 2013-2017. The figure of male mobile researchers rose from 91,684 (6% of all published male researchers) to 167,182 (5.6%) over the same time periods. The gender difference in the representation of transnational scholarly migration has gradually shrunk, although there is still a long way off to achieve the gender equality. At the country level (Fig. 1), the scholarly migration from and to the US still dominates the globalization of the high-skilled for both genders. However, the growth of the male emigration flows noticeably outpaced that of the male immigration flows. Most migratory countries in the figure indicate a similar increase pattern for female and male mobile researchers, in either direction over the 20 years. That contributed to a relatively stable gender migration of researchers across national borders for a long time. The only exception is Japan, where the influx of male researchers even slightly scaled back in the latest period compared to the earlier period.
Fig.1. Immigration and emigration of researchers by gender and country between periods 1998-2003 and 2013-2017 (in thousands)
The migration spreads of mobile researchers measure the extent to which the mobile researchers are dispersed across destination countries (emigration spreads), and the extent to which they come from a diverse array of origin countries (immigration spreads). Overall, mobile researchers of both genders tended to move from a relatively smaller number of countries (solid lines in Fig. 2) to a more diverse array of destinations (dotted lines) during all time periods. And the gap between emigration and immigration spreads have become narrower with time. In addition, the female migration pattern presented a relatively larger difference between the diversity levels of outflows and inflows, indicating that female researchers from a smaller number of countries took more opportunities to go abroad, and the increased mobility was more dispersed in their destinations.
Fig.2 Global scholarly migration spread by gender with time
Considering the variation in migration spread of different countries over time, Fig. 3 illustrates the gendered migration spreads of researchers at the country level. It dividing the 18 major migration countries into three groups: (a) Constant high migration spread, (b) Fast-growing and average migration spread, and (c) Low migration spread. Albeit the rapid increase in scholarly migration spread across South Korea, its spread values are still below those of most other countries, making it partitioned into the category of Low migration spread. In group (b), Canada and Japan experienced a fast development in their migration spread of researchers while the other three countries maintained a relatively moderate migration spread without remarkable fluctuations.
(a) High migration spread
(b) Fast-growing and average migration spread
(c) Low migration spread
Fig.5 Country-level scholarly migration spread by gender
In most of the countries with high level migration spreads, the emigration or immigration spreads of female and male researchers always followed similar trends with close values of spread measurements. In another word, when the migration of male researchers from these countries becomes more spread across destination countries, the female emigrant researchers also go to an increasingly diverse array of destinations. This is especially remarkable in the recent two periods. However, for some countries in the other two groups including India, Brazil and Israel, the migration spreads presented an opposite trend among female researchers and male researchers during some periods.