Rafael de Moraes Baldrighi
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
1-24
The determinants of inbound
international tertiary students
in the developing world:
the global south dimension
Os determinantes da entrada de
estudantes internacionais de ensino
superior no mundo em desenvolvimento:
a dimensão do sul global
Los determinantes del ingreso de
estudiantes internacionales de educación
superior en el mundo en desarrollo:
la dimensión del sur global
DOI: 10.21530/ci.v19n2.2024.1517
Rafael de Moraes Baldrighi
1
Abstract
Over recent decades, the internationalization of universities has
become a global norm. Tertiary student mobility literature identifies
push-pull factors driving cross-border flows, but does it explain
why developed countries dominate as destinations while emerging
ones, the Global South, export students? Using Ordinary Least
Squares multiple regression, we analyzed pull factors influencing
inbound students across emerging countries. Our findings highlight
the significant role of geopolitical factors, with language, academic
excellence, migrant networks, and hosting capacity also positively
affecting student inflows.
Keywords: Student Mobility; Academic Internationalization, Global
South; Higher Education.
1 Master’s Degree in International Relations at the Institute of International Relations
(IRI) – University of São Paulo (USP). Rafael is a public servant and researcher for
the Brazilian Government at the Institute for Applied Economic Research (Ipea).
(rafaelbaldrighi@gmail.com). ORCID: https://orcid.org/0000-0002-7527-3404.
Artigo submetido em 16/08/2024 e aprovado em 02/12/2024.
ASSOCIAÇÃO BRASILEIRA DE
RELAÇÕES INTERNACIONAIS
Copyright:
This is an open-access
article distributed under
the terms of a Creative
Commons Attribution
License, which permits
unrestricted use,
distribution, and
reproduction in any
medium, provided that
the original author and
source are credited.
Este é um artigo
publicado em acesso aberto
e distribuído sob os termos
da Licença de Atribuição
Creative Commons,
que permite uso irrestrito,
distribuição e reprodução
em qualquer meio, desde
que o autor e a fonte
originais sejam creditados.
ISSN 2526-9038
The determinants of inbound international tertiary students in the developing world [...]
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
2-24
Resumo
Nas últimas décadas, a internacionalização das universidades tornou-se uma norma global.
A literatura sobre mobilidade estudantil no ensino superior identifica fatores de atração e
expulsão que impulsionam os fluxos transfronteiriços, mas será que isso explica por que os
países desenvolvidos predominam como destinos, enquanto os emergentes, do Sul Global,
exportam estudantes? Utilizando regressão múltipla por mínimos quadrados ordinários,
analisamos os fatores de atração que influenciam os estudantes internacionais em países
emergentes. Nossos resultados destacam o papel significativo de fatores geopolíticos, além
do impacto positivo da língua, excelência acadêmica, redes migratórias e capacidade de
acolhimento.
Palavras-chave: Mobilidade Estudantil; Internacionalização Acadêmica; Sul Global; Ensino
Superior.
Resumen
En las últimas décadas, la internacionalización de las universidades se ha consolidado
como una norma global. La literatura sobre movilidad estudiantil terciaria analiza factores
de atracción y expulsión que impulsan los flujos transfronterizos, pero ¿explica por qué los
países desarrollados son destinos principales mientras que los emergentes, el Sur Global,
exportan estudiantes? Usando regresión por mínimos cuadrados ordinarios, analizamos los
factores de atracción de estudiantes en países emergentes. Nuestros hallazgos destacan la
relevancia de factores geopolíticos, así como el impacto positivo del idioma, la excelencia
académica, las redes migratorias y la capacidad de acogida.
Palabras clabe: Movilidad Estudiantil; Internacionalización Académica; Sur Global; Educación
Superior.
Introduction
For the last three decades, the internationalization of Higher Education
Institutions (HEIs) has become a consolidated, and almost obligatory, process
for well-ranked universities. Mostly inspired by the European experience, HEIs
all around the world are developing internationalization strategic plans and
policies, with the phenomenon of tertiary mobility, both inbound and outbound,
as one of their main components. According to the Organization for Economic
Rafael de Moraes Baldrighi
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
3-24
Co-operation and Development (OECD), between 1998 and 2018, the total number
of international tertiary students has grown by 4.8% on average, yearly, reaching
almost 6 million in 2018 (OECD 2019).
However, those values are significantly imbalanced. In 2018, students from
emerging countries represented the bulk of the outgoing flows migrating toward
developed countries, with Asia accounting for 57% of all mobile students in
OECD states, and China and India alone representing more than 30% of that
number. The United States (US), on the other hand, received 18% of the world’s
flow, followed by Australia and the United Kingdom (8% each). In that year, less
than 30% of the world’s mobile students were enrolled in non-OECD countries.
The vast literature on tertiary student mobility highlights push-pull factors
that can help us better understand this imbalance in which developed countries
are the main destinations and emerging countries – here, the Global South
(Mignolo 2002; Docquier and Rapoport 2012; Van Bouwel and Veugelers 2013;
Beck and Pidgeon 2020), that is, countries where emancipation from Western
established economic and political discourse used to legitimize cultural control
is unfolding – are the exporters of students. Among the push factors are the
value of holding a foreign degree, national isolation (geographical or cultural),
and political/economic issues (academic freedom, censorship, employment
opportunities, and income expectations). Among the pull factors are the existence
of well-ranked institutions, a dynamic job market, hosting capacity, geographical
proximity, and a plethora of other features that a country can display that are
‘attractive’ to foreign students (Beine, Romain and Ragot 2014; Caruso and De
Wit 2014; Didisse, Nguyen-Huu and Tran 2018).
Here, we analyze the tertiary mobility phenomenon with a focus on the
Global South, given the disparity in the number of inbound international students,
dealing with issues such as unequal globalization, brain-drain/brain-gain, former
colonization ties, and the predominance of OECD HEIs in university rankings.
We do so, since it is important to know which policies developing countries can
design to attract students, and since tertiary mobility has become increasingly
relevant over the last decades, both for the growing number of mobile students
(OECD 2019) and the increase in monetary terms of education services (Caruso
and De Wit 2014).
Therefore, this paper aims to study the pull factors, or the attractiveness,
of inbound mobile students in a cross-section comprising emerging countries
in the year of 2017 using Ordinary Least Squares (OLS) multiple regression
The determinants of inbound international tertiary students in the developing world [...]
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
4-24
models. The dependent variable is the proportion of higher education incoming
students per country relative to the total number of students studying abroad
– regardless of its nature: mobility, research, whole course, undergraduate,
graduate, Ph.D., post-doc, and so on. We limit the analysis to 2017 since this is
the year with the highest availability of data for mobile students in emerging
countries in the Unesco Institute for Statistics (UIS) database. Also, it predates
the 2020-2022 global COVID pandemic that severely impacted mobility numbers.
The explanatory variables are: spoken language, academic excellence, migrant
network, geographical distance, income expectations, hosting capacity, and cost
of living.
We opted for a cross-section since we deal with a significant lack of data on
mobility for non-OECD countries (Beine, Romain and Ragot 2014; Rumbley 2012),
hindering the construction of a panel. Also, the dependent variable and some
explanatory factors are (quasi) constant over time, such as distance, language,
and institutional features (Kahanec and Králiková 2011; Didisse, Nguyen-Huu
and Tran 2018). Our innovation is the use of non-OECD countries as the sample
– which is not abundant in the literature and official databases on the subject,
as the next section shows – and the creation of three dummy variables to assess
which dimension better represents the Global South – to try to capture a set of
country-specific aspects that differentiate developing nations from developed
ones (Caruso and De Wit 2014). Most studies on the subject, qualitative or
quantitative, focus on European and OECD countries only.
Our findings, after running ten models, are that out of the three dummies
used, the one that best encapsulates the Global South facet was the existence of
a former colonial regime in a given country. Also, spoken language, academic
excellence, a network of migrants, and hosting capacity are significant and
positively related to our dependent variable. This paper is structured as follows:
in the next section, we highlight the pull factors of tertiary mobility and present
the Global South perspective. In “Descriptive Analysis and Data Presentation”,
we display the variables and the data collection and treatment procedures. In
“Main Findings”, we present the empirical application of the models and the
outputs of the regressions. A final section summarizes our findings.
Rafael de Moraes Baldrighi
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
5-24
Literature Review: Pull Factors of Student Mobility
and the Global South
Pull Factors of International Mobility
The literature on international mobility of tertiary students highlights some
factors that can either attract or refrain students from a specific destination.
Those are commonly divided into push-pull factors (Caruso and De Wit 2014).
The former is related to national features of the country of origin that propel
students to move abroad; and the latter regards the attractiveness of a state to
lure international students. Although this dynamic is in decline in migration
studies, it remains relevant in works on academic mobility, which is a singular
phenomenon regarding migrants. Here, we do not focus on the factors ‘at home’
that push individuals to seek education abroad, but on how a country can attract
foreign students.
But why is it relevant to woo international students? Besides from shielding
the minds that would be lost due to student migration, the act of hosting highly-
skilled migrants has direct and indirect gains in improvements in education,
productivity, and research. Also, the externalities of receiving foreign students range
from a more diverse and multicultural society in the micro level, to international
cooperation and regime strengthening in the macro level – Docquier and Rapoport
(2012) and Van Bouwel and Veugelers (2013). Sure, internationalization processes
encompass international publications, the promotion of English-taught courses,
bilateral agreements, and a plethora of other features that a university can display
to advance in international cooperation. Its main perceptible component, though,
is the mobility, incoming and outgoing, of students.
The first pull factor is language proximity. Beine, Romain and Ragot (2014)
find that a common official language between origin and destination countries
positively impacts the flow of students. This is also noted by Junqueira and
Baldrighi (2020) when analyzing if widely spoken languages are more attractive
to students by comparing Spanish and Portuguese. Didisse, Nguyen-Huu and Tran
(2018) use a four-tier measure for language proximity (from a common official
language to the similarity of different languages) and find a positive relationship
for all measures. Equivalent results can be found in Kahanec and Králiková (2011)
and OECD (2011) that highlight the national language of instruction as one of
the most important factors to attract foreign students, with English, French,
German, Spanish, and Russian being the most inviting ones.
The determinants of inbound international tertiary students in the developing world [...]
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
6-24
Academic excellence, what Van Bouwel and Veugelers (2013) call ‘the quality
dimension’, is also a pull factor that positively affects the number of mobile
international students in a country. Measures of the academic impact of a
country’s scientific publications, expenditure per student, or the number of national
universities in HEI rankings are the general proxies used to estimate this factor.
Using the top 200 universities in the Shanghai Ranking – Academic Ranking of
World Universities (ARWU) –, Beine, Romain and Ragot 2014 find that quality
of education is a significant but moderate attractor when compared to other
pull factors. Van Bouwel and Veugelers (2013) find, regardless of the indicator
used (Shanghai Ranking, the Times Higher Education – THE – Ranking, and the
relative impact of a country’s scientific publications), a positive relationship for
the flow of inbound students. Similar results can be found in Caruso and De Wit
(2014), who detect a positive and significant relationship between the number
of incoming students and public expenditure per student. In those rankings,
there is a strong bias towards educational institutions from the Global North.
A network of similar social actors to which a student can rely on when
abroad is another factor that is positively related to incoming students in the
literature. The proportion of foreigners living in a country (Junqueira and Baldrighi
2020), the number of high-skilled migrants (Beine, Romain and Ragot 2014), and
even internet users (Didisse, Nguyen-Huu and Tran 2018) are instrumentalized
to measure such network effect. The higher the number of students in the
destination country, the higher the flow of students from the same origin since
the “presence of country nationals at destination tends to act as a magnet for
international students” (Beine, Romain and Ragot 2014: 51). According to these
authors, the existence of a migration network in the destination country can also
reduce migration costs. Furthermore, surveys can also capture this phenomenon,
as Mazzarol and Soutar (2000) and Pedro and Franco (2015) demonstrate, after
surveying mobile students in Australia and Portugal, respectively.
Commonly, the geographical distance between countries negatively impacts
the proportion of students attracted. In all six models presented by Beine, Romain
and Ragot (2014), distance has a negative coefficient and is statistically significant
at the 1% level. The same is true for Van Bouwel and Veugelers (2013) who use
five simple gravity models. Caruso and De Wit (2014) also observe this negative
impact and feature geographical distance as an economic pull factor since smaller
distances generally translate into smaller transportation costs. From this, we can
also infer that the existence of bordering countries, analyzed dyadically, facilitate
Rafael de Moraes Baldrighi
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
7-24
travel procedures and have more flexible visa policies, which is accounted for
by Van Bouwel and Veugelers (2013).
Another common pull factor is the expected future income in the destination
country. Caruso and De Wit (2014) find a positive, significant at 1%, relationship
for the gross domestic product (GDP) per capita and incoming students. The
same is true for Beine, Romain and Ragot (2014). Using gross annual wage for
workers with tertiary education level, these authors also find a positive and
significant at the 1% level impact on the dependent variable. Didisse, Nguyen-
Huu and Tran (2018) apply economic and socio-demographic factors as proxies,
such as youth unemployment and average enrolments in tertiary education, and
find, respectively, a negative and a positive relationship.
Hosting capacity is another variable. A vast number of national universities,
a big and dynamic job market, and several housing or funding opportunities
can act as a magnet to attract and increase the number of foreign students. The
most common proxy used is population. Beine, Romain and Ragot (2014) show
that students are sensitive to this factor. However, the measures used differ:
the authors use the total population (logged) as a proxy and compare it to the
total number of students enrolled at the university of destination during a given
academic year for Italy and the United Kingdom (UK). Similarly, and dyadically,
Van Bouwel and Veugelers (2013) use the student population in the host and
the sender to instrumentalize this positive relationship in basic gravity models.
Cost of living can also impact the number of inbound tertiary students and
the literature points to a negative relationship. Caruso and De Wit (2014) proxy
this factor by the current inflation change, whereas Beine, Romain and Ragot
(2014) employ Numbeo’s Consumers Price Index. Also, tuition fees are important
when analyzing the costs of student migration. However, this data is, overall, not
available for a good number of countries (Rumbley 2012; Didisse, Nguyen-Huu
and Tran 2018) and Beine, Romain and Ragot (2014) find that, although living
costs have a negative strong impact on incoming students, fees are insignificant.
This happens since mobile students often benefit from stipends or fellowships
to cover them. This is highly debatable, though, as Caruso and De Wit (2014)
find that cost of living alone does not discourage the inflow of foreign students:
it only does when combined with tuition fees. Collinearity issues in regression
models can also arise when measuring the cost of living since, depending on the
proxy, it is often strongly correlated with income per capita, public expenditure,
or fees (Caruso and De Wit 2014).
The determinants of inbound international tertiary students in the developing world [...]
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
8-24
Other factors, which receive less attention from the literature, can also
be found in quantitative models and surveys, such as violence – which often
include homicides, committed crimes, and other phenomena that are hard to
grasp, such as political repression, xenophobia, and racism (Mazzarol and Soutar
2000; Caruso and De Wit 2014; Junqueira and Baldrighi 2020); public policies
designed to attract students – which include scholarship policies, the promotion
of English-taught courses, visa restrictions, HEIs’ agreements, and migration
opportunities, which, again, suffer from lack of data (Kahanec and Králiková
2011; Rumbley 2012); and other cultural and religious dyadic factors that may
impact the bilateral flow of students.
The Global South Perspective
Besides explanatory factors, scholars often come from different perspectives
and theoretical backgrounds to explain why students migrate, such as human
capital theory – the mobility as an investment to grab job opportunities or to
increase future income (Rosenweig 2008); a consumption choice – the search for
a better education than at home (Van Bouwel and Veugelers 2013; Beine, Romain
and Ragot2014); social capital theory – students being attracted to countries where
they can find a similar social network in a cumulative causation process (Van
Bouwel and Veugelers 2013; Pedro and Franco 2015); and, lastly, from a critical
point of view on globalization and human migration (Mignolo 2002), which
the literature identifies as ‘the Global South perspective’. From this standpoint,
migration flows originate in the so-called periphery with the so-called core, or
the Global North, as the destination. The latter ensure their dominant position
by retaining minds and talents from peripheral countries in a brain-drain/brain-
gain cycle, mainly from former colonies to former colonial powers.
However, we do not use this perspective naively, blaming globalization
only and what some authors call the Geopolitics of Knowledge (Mignolo 2002;
Beck and Pidgeon 2020). Even though this dependent and imbalanced relation
does exist, Docquier and Rapoport (2012) and Van Bouwel and Veugelers (2013)
show that student mobility from the periphery to the core can create positive
externalities at home – on technological, educational, and political issues – and
not necessarily lead to more unequal or neocolonial (Buckner and Stein 2019)
relations. Here, we list five factors and conduct two hypothesis tests to argue for
the development of a model specifically designed for inbound mobile students
in the Global South.
Rafael de Moraes Baldrighi
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
9-24
The first factor is the imbalanced flow. Besides the numbers in the introduction
brought by OECD (2019), we can compare the proportion of international students
per region of the world. With UIS’ National Monitoring data for inbound mobility
rate,2 we find that, in the year 2017 (the reference year for data collection in
our study), the world’s average for this index is at 2.4%. In the Global South:
Sub-Saharan Africa, 1.7%, the Arab States, 3.06%, Asia-Central, 2.16%, Asia-
Southern, 0.16%, Asia-Eastern, 0.85%, Asia-South-Eastern, 1.07%, and Latin
America and the Caribbean, 0.73%. In developed countries, the inbound rate is
at 7.33% in North America and Western Europe, 3.43% in Central and Eastern
Europe, 4.27% in Japan, and 21.27% in Oceania (Australia/New Zealand). That
is, except for the Arab States – possibly due to the high rate of international
students in the Gulf states and the Syrian and Palestinian diasporas –, all the
Global South regions are below the world’s average and significantly below
developed countries’ average.
Furthermore, as presented before, the majority of incoming students in the
world are from emerging countries, creating a ‘natural’ flow that consolidates
mobility as a movement of millions of minds leaving the Global South toward
developed countries. UIS’ net flow of internationally mobile students
3
gives us a
hint that, in general, emerging countries present a deficit when calculating the
difference between incoming and outgoing students, whereas OECD countries
tend to register a surplus.
Another factor is academic excellence. As presented, university rankings are
used as proxies to measure the quality dimension (Van Bouwel and Veugelers
2013). Among those, are the ARWU/Shanghai Ranking, THE, and Quacquarelli
Symonds (QS) World University Ranking. Although these can be helpful and
adequate measures of academic excellence, there is a strong predominance of North
American and European universities, excluding a huge deal of universities from
emerging states, making it difficult to compare Global South countries’ academic
excellence. Sure, ARWU has a strong bias toward China. Also, Brazil, Chile, India,
and South Africa have a good number of HEIs in these rankings. However, to
analyze academic excellence in the Global South, we need an inclusive criterion,
contrasting with Van Bouwel and Veugelers (2013), which limits it to the top 200
in the ARWU ranking, and Didisse, Nguyen-Huu and Tran (2018), who limits it
2 According to UIS’ glossary, it is the “number of students from abroad studying in a given country, expressed
as a percentage of total tertiary enrolment in that country”.
3 That is the difference between the number of students hosted and the number of students sent abroad.
The determinants of inbound international tertiary students in the developing world [...]
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
10-24
just to ARWU’s top 100. Restricting the number of HEIs analyzed to just a couple
hundred would severely hinder a thorough analysis of Global South academic
excellence, since few emerging countries would make it to the podiums.
The third feature that marks the mobility of students leaving emerging
countries is a former colonial tie. This phenomenon is captured by Beine, Romain
and Ragot (2014) and Didisse, Nguyen-Huu and Tran (2018) who show a positive
and significant relationship between former colonial links and incoming students.
For example, Brazilians are the largest group of international students in Portugal,
Indians are the second-largest in the UK, and Moroccans and Algerians are,
respectively, the first and the second major groups of inbound students in France.
Therefore, a sample comprising emerging countries only (since states4 in the
Global South were, mainly, former colonies), would help us better capture the
determinants of inbound mobility toward them.
The fourth point is the late evolution of internationalization in emerging
countries. Not only the development of universities originated in Europe, but, in
that continent, scholars have been mobile for centuries (Van Bouwel and Veugelers
2013). Furthermore, in the last three decades, European countries have engaged
in highly successful projects to strengthen student mobility in the continent via
stipend and fellowship programs supported by legal international obligations,
standardization of credits, and multilateral commitments embodied by the Magna
Charta Universitatum (1988) and the Bologna Declaration (1999). Institutions
such as Campus France and the Deutscher Akademischer Austauschdienst (DAAD)
are well known all over the world. In contrast, efforts in the Global South have
been much more modest and came much after European initiatives.
Also, due to the high quality and tradition of American universities, the
US has dominated the scientific and academic scenario since the mid-1900s,
attracting a huge number of scholars and skilled students over the last decades.
Chen and Barnett (2000) classify countries into three categories they created: a
core destination of students (Western Europe, Oceania, US, and Canada); what
they call the semi-periphery, such as Eastern Europe (mainly Russia);5 and the
4 In the Global South sample, we include countries such as Russia and Turkey. Arguably they can be considered
as former colonial powers, but due to the inclusion criteria explained in the next section, they made it to the
emerging dataset.
5 Russia (and the USSR) is a very important destination country for mobile students and can be included as
a traditional destination. However, due to the sampling criterion used in this paper – presented in the next
section –, we considered the Russian Federation as an observation in the emerging dataset. The col_pow45
dummy tries to deal with this issue, though.
Rafael de Moraes Baldrighi
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
11-24
periphery, which is not attractive to foreign students, such as Latin America, Africa,
and Asia. Thus, path dependence and late development of internationalization
policies corroborate our ‘emerging’ model since Global South countries are not
yet ready to play in the big leagues on an equal footing.
Lastly, the fifth feature regarding Global South inbound mobile students is
the lack of quantitative studies in the literature about them. Almost all of the
empirical studies cited here use European and OECD countries as sample. Sure,
it is undeniably true that the availability of data for those countries is higher,
counting not only on UIS’, but also on OECD’s and Eurostat’s databases – the
UOE data sets (Rumbley 2012). Most of the works on the subject in non-European
and emerging countries use qualitative methodologies (Junqueira and Baldrighi
2020), or assume a critical and theoretical point of view, without diving into
statistical analyses (Mignolo 2002; Buckner and Stein 2019). There are studies
that address some attractive variables to include developing countries, but there
is no quantitative model, to date, that attempts to better understand the flow
of international students in those nations. Thus, our model including emerging
countries tries to remedy this lack in the literature.
To check on the properness of these five assumptions on why we should
model emerging countries, we conduct a hypothesis test comparing UIS’ inbound
mobility rate for North America and Western Europe (NAWE) with the mean
proportion of our sample for the Global South. For descriptive statistics, NAWE’s
rate is 7.33%. Although we could rely on UIS’ average proportion for the several
Global South regions in its database, we chose to calculate the mean presented
in our sample due to the lack of data for emerging countries in UIS’ database,
which affects regions differently.
Loading our dataset into STATA v.14, we can calculate the mean proportion of
the inbound mobility rate of the Global South countries in our sample. Using the
1.5 Interquartile Range (IQR) technique to remove outliers, we excluded seven
6
observations from a total of 77 countries. This left us with a mean statistic of
2.42% (similar to the world’s 2.4%). We also found a sample standard deviation
of approximately 2.28%. We then affirm that the mean proportion of Global South
countries is less than NAWE’s half. Even if we are dealing with proportions,
since we calculated their average (descriptive statistics), we use the hypothesis
test for the mean (µ) and not the proportion’s π.
6 Except for Jordan, all the other six countries excluded are small Gulf or Caribbean states.
The determinants of inbound international tertiary students in the developing world [...]
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
12-24
H0: µ=3.67%
H1: µ<3.67%
After calculating this one-tailed test at a 0.001 significance level,
7
the Z-score
we got was approximately -4.14, way beyond the critical value to the left (-3.08).
Therefore, we can reject the null-hypothesis and confirm that there is evidence
to support the alternative one.
And just to strengthen this argument, we calculated another hypothesis test
regarding the difference in the sample means. We contrasted the “Dataset_Emerging”
mean with the one calculated from “Dataset_World” (more on those in the next
section). The last one is comprised of 36 developed countries and, after applying
the 1.5 Interquartile Range (IQR) technique, we were left with 34 observations,
a sample mean of 8.73 and a sample standard deviation of approximately 5.03.8
H0: µDataset_Emerging = µDataset_World
H1: µDataset_EmergingµDataset_World
We obtained a Z-score of -6.975, which indicates that it lies in the rejection zone
to the left, way beyond any critical values at the 0.05, or 0.01 significance levels.
Descriptive Analysis and Data Presentation
We created two datasets for this study: “Dataset_Emerging” and “Dataset_
World”. The first comprises 77 observations from the Global South and the second
includes those 77 plus 36 developed countries, totaling 113. In trying to identify
the proportion of inbound students enrolled in a country’s education system, we
first used the descriptive statistics of the absolute share (that is, the number of
international tertiary students divided by the total number of tertiary students
in a given country), i.e., UIS’ inbound mobility rate. However, we found that,
due to the small number of tertiary students in some emerging countries, this
would add bias toward less populated states.
For example, China, hosting almost 200 thousand international students, has
a smaller rate (0.36%) than DR Congo (0.44%), which hosts only 2038 mobile
7 We used the formula to calculate the Z-score since, even though we do not know the population
standard deviation, we are dealing with a large sample (n=70).
8 The formula used was .
Rafael de Moraes Baldrighi
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
13-24
students in the year analyzed. We ran one model using this rate as the dependent
variable (Table 2), but the best way to capture a country’s inbound mobility was
to calculate the proportion of international students hosted regarding the total
number of the sample’s mobile students. Using this method, China gets its share
of more than 10%, while DR Congo is at no more than 0.13%. We then assess
the relevance of a country in the Global South scenario of international mobility,
not its national proportion of foreign students. This technique also shields us
against multicollinearity issues between the number of inbound students, the
number of migrants, and the total population.
The explanatory variables for “Dataset_Emerging” are: i. spoken language;
ii. relevance in university rankings; iii. migration network, iv. average distance;
v. expected income; vi. hosting capacity; and vii. cost of living. Table 1 presents
further details on them. For the first explanatory variable, we used a dummy
in which ‘1’ corresponded to a country that de jure/de facto speaks an official
United Nations language (i.e. Arab, Chinese, English, French, Russian, or Spanish)
or German (OECD 2011).9 Secondly, academic rankings are common proxies
for assessing higher education excellence in studies about inbound mobility.
For this independent variable, we use ARWU/Shanghai ranking, but inclusively:
the top 500 universities in 2017.
The third explanatory variable is proxied by the total number of international
migrants in a country. Again, using the same logic applied for the dependent
variable, we could take the national proportion regarding a country’s population
only, but it would create a strong bias toward less populated countries. We then
divide the number of migrants in a given state by the total sum of migrants in
all countries of our sample for the year 2015.10 Average distance, the fourth
explanatory variable, is proxied by the mean distance between a country’s main
economic center
11
(or its capital, which, for the majority of our sample, coincide)
9 As mobilities are generally performed in urban and academic environments, we applied an inclusive criterion.
Some countries of our sample, such as former African colonies or Soviet Republics in Central Asia, speak
a variety of languages. In situations in which the former colonial power language predominates in urban
administrative, bureaucratic, and trade issues, we considered a positive result for the ‘UN languages plus
German’ criterion. For example, India, Kazakhstan, Cote D’Ivoire, and Malaysia are countries in such a situation.
All the ambiguous situations were analyzed using the CIA World Factbook and the CEPII (2011) dataset.
10 We do not use the values for 2017 since the UN has quinquennial publications and the year 2015 is the closest
one with such data.
11 For example, we use São Paulo and not Brasília for Brazil, and Istanbul and not Ankara for Turkey. However,
for the majority of the countries in the sample, the capital city and the main economic center coincide: Buenos
Aires for Argentina, Santiago for Chile, Mexico City for Mexico, and so on.
The determinants of inbound international tertiary students in the developing world [...]
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
14-24
and its ten closest destinations. As the literature points out, this relationship is
negative and we selected the closest national entities and calculated the mean
distance to check how far a country is from its neighbors. Expected income, the
fifth variable, is measured by a country’s GDP per capita in US Dollars in 2017.
Hosting capacity, as the literature suggests, is proxied by the total population
of a country in 2017. Lastly, the cost of living is proxied by World Bank’s Price
level ratio of PPP conversion factor (GDP) to market exchange rate, a continuous
index number that considers the US equal to ‘1’,12 in 2017. “Dataset_World” has
all those variables plus three dummies named i. emerging, ii. col_pow45, and
iii. oecd. They were designed to capture country-specific factors that are present
in Global South countries, based on a dummy created by Caruso and De Wit
(2014) to differentiate Western and Eastern Europe.
We must highlight that we deal with pull aspects only and not dyadic factors
(or, how country-A is attractive to country-B students given their distance, if
they speak the same language, if A’s expected income is higher, and so on), i.e.,
push-pull determinants. Therefore, we focus only on a country’s attractiveness
in general, not relative to some other country. We study these monadic factors
mainly due to the lack of international mobility dyadic data for emerging states
(for example, Mexico and Russia do not have the data on the nationality of the
students hosted), which hinders dyadic analyses.
Furthermore, we chose the year 2017 for data collection since it is the most
recent one (that predates the global COVID pandemic) with the highest availability
of data for emerging countries in UIS’ database. The indicators on international
mobility are part of the National Monitoring series, which means that we are relying
on a country’s good-will to annually inform its numbers to Unesco. It not only
hinders the creation of a panel or a time-series for the observations but forces us to
include data from 2015 or 2016 as proxies for 2017 values for important countries
in our sample (such as Israel and Egypt) that did not provide the 2017 indicators.
For the developed countries, all the data on mobility is from 2017.
Lastly, regarding the inclusion criterion, “Dataset_Emerging” contains countries
from Latin America, Africa, Asia (including the Middle East, the Caucasus, and
Turkey), and European countries that are not members of the European Higher
Education Area (EHEA), or even though are its members, were not part of the
European Union (EU), the European Economic Area (EEA), or the European Free
Trade Agreement (EFTA) in 2017 – that is, Belarus, Bosnia and Herzegovina, Moldova,
12 We did not use Numbeo’s indexes due to their 'user input' methodology.
Rafael de Moraes Baldrighi
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
15-24
North Macedonia, Serbia, Russia, and Ukraine. “Dataset_World” comprises all the
countries from “Dataset_Emerging” plus all the remaining countries from Europe,
the United States, Canada, Japan, Australia, and New Zealand. All the observations
were included if their inbound mobility data were available in UIS’ database.
Table 1 summarizes and further explains the variables used in our models. The
variables for distance (dist10), expected income (income), and hosting capacity
(population) also have their data presented in natural logarithms (ln_dist10,
ln_income, and ln_population). We performed this logarithmic transformation to
obtain a more normalized dataset since these three variables are highly skewed.
Table 1 – Data Presentation
Variable Source Definition
inbound_rate UIS (2020) The proportion of foreign inbound tertiary students in a country
divided by the total number of students in that country in 2017
inbound_sample UIS (2020) The number of foreign inbound tertiary students per country divided
by the total sum of inbound students in our sample
Lang CEPII (2011),
CIA (2023)
Spoken language. '1' if a country's language is one of the official
languages of the United Nations plus German, '0' if it is not
arwu500 ARWU (2017) Number of universities in the top-500 ARWU/Shanghai ranking in 2017
mig_rate UN DESA
(2019)
International migrant stock (percentage of the total population)
in 2015
mig_sample UN DESA
(2019)
Number of migrants per country divided by their total sum of our
sample
dist10 CEPII (2011) Mean distance between a country and the 10 closest national entities.
This variable has a version with logarithmic transformation “ln_dist10
income UN (2017) United Nations Stats 'GDP, Per Capita GDP – US Dollars' in 2017. This
variable has a version with logarithmic transformation “ln_income
population UN (2017) The total population of a country in 2017. This variable has a version
with logarithmic transformation “ln_population
liv_cost World Bank
(2020)
Price level ratio of PPP conversion factor (GDP) to market exchange
rate in 2017. Index number (United States = 1)
emerging De Wit and
Caruso (2014)
The dummy takes the value of unity for countries in
“Dataset_Emerging”
col_pow45 CEPII (2011)
The dummy takes the value of unity if a country had colonies or
possessed territories that became independent in similar colonial
relationship after 1945. It is based on the “col45” variable CEPII (2011)
oecd OECD (2024) The dummy takes the value of unity for OECD members in 2017
Source: Elaborated by the author (2024).
The determinants of inbound international tertiary students in the developing world [...]
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
16-24
Main Findings
After data collection from several sources, the values were tabulated into
two Microsoft Office Package 2019 Excel spreadsheets, each containing one
dataset (“Dataset_Emerging” and “Dataset_World”). To run the OLS models,
we used STATA v.14 software, and the commands and codes used are presented
in a complementary script file submitted to the journal. The equation with
the highest adjusted R-squared value – model (2) below – for “Dataset_
Emerging” can be described as follows. The variables are indexed by country “i”
(i = 1, […], 77).
yinbound_samplei = β0 + β1langi + β2arwu500i + β3mig_samplei + β4ln_dist10i +
β5ln_populationi + β6ln_incomei + β7liv_costi + εi
Table 2 – Inbound Internationally Mobile Students in the Global South
Dependent Variable
(1) (2) (3) (4)
inbound_rate inbound_sample
lang 2.867524
(2.573091)
0.0026857
(0.0036851)
0.0021498
(0.0036834)
0.0019062
(0.0036369)
arwu500 0.1369286
(0.2524916)
0.0020745***
(0.0003656)
0.0019236***
(0.0003496)
0.0021348***
(0.0003631)
mig_rate 0.2309748**
(0.094137)
mig_sample 0.6558668***
(0.087333)
0.6717908***
(0.0869927)
0.7077024***
(0.0758839)
ln_dist10 -7.840511***
(2.56518)
-0.0007607
(0.0038138)
-0.000545
(0.0038315)
0.0000506
(0.0037632)
ln_income 0.3865359
(1.417735)
0.0024032
(0.0020221)
0.0009839
(0.001729)
ln_population -2.100384***
(0.7822091)
0.0013246
(0.0013134)
0.0015992
(0.0013044)
0.0015992
(0.0012403)
liv_cost 16.05242
(10.86713)
-0.02115
(0.015859)
-0.0112308
(0.0135247)
constant 79.97882***
(20.78983)
-0.0271534
(0.031483)
-0.0302044
(0.0315739)
-0.008717
(0.0274763)
Observations 77 77 77 77
Adjusted R-squared 0.4531 0.6973 0.6939 0.6955
F-Value 9.99 26.01 29.72 29.93
Note. Standard errors are presented in parenthesis. *significant at 10%. **significant at 5%. ***significant at 1%.
Rafael de Moraes Baldrighi
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
17-24
By analyzing the regressions’ outputs, we can confirm that the model
significantly improves when using inbound_sample as the dependent variable
instead of inbound_rate. It is also interesting to note that hosting capacity (ln_
population) has a negative coefficient in regression (1), which strengthens our
argument that using the inbound rate proportion would create a bias toward less
populated countries. Another problem with using inbound_rate as the dependent
variable is found in liv_cost: it yields a positive coefficient, contrasting with
previous works, probably due to the high inbound rate of the Gulf and Caribbean
states, all of which present a high cost of living. Furthermore, from Table 1, we
can infer that all statistically significant results in regressions (2), (3), and (4)
are following the literature, that is, arwu500 and mig_sample. In regression (1),
ln_dist10 and mig_rate present expected results, while ln_population, for the
reasons stated, did not.
Moreover, to avoid common collinearity issues when measuring the cost
of living, we conducted regressions (3) and (4) with either ln_income or liv_
cost. The linear correlation between those variables is positive and moderately
strong (Pearson’s R = 0,61) and regression (4) yielded slightly higher adjusted
R-squared and F-values than regression (3), but with a change in the coefficient
for ln_dist10, contrasting with the literature. However, overall, even if presenting
a lower F-value when compared to models (3) and (4), regression (2) produced
the highest adjusted R-squared and all the (significant) coefficients in accordance
with the literature.
The Global South Dimension
To improve our model and to reinforce our argument that emerging countries
significantly differ from developed ones regarding their attractiveness, we increase
the number of observations in the sample and add three new explanatory dummy
variables. We now turn to the “Dataset_World”. The new explanatory dummy
variable named emerging takes the value of ‘1’ when a country is also part of
the “Dataset_Emerging”. The col_pow45 dummy takes the value of unity if a
country had colonies or possessed territories that became independent after a
similar colonial relationship since 1945, based on the “col45” variable present on
CEPII (2011). Lastly, the third dummy, oecd, takes the value of unity if a country
was an OECD member in the year 2017.
The determinants of inbound international tertiary students in the developing world [...]
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
18-24
We designed these dummies based on Caruso and De Wit (2014) to try to
capture a set of latent features that differentiate developed countries from emerging
ones. For example, when choosing a Global South country as a destination instead
of, say, Australia, several hard to measure variables may play a significant role,
such as xenophobia, fear of violence, poor infrastructure, and even unrest about
higher education excellence due to a feeble performance in rankings. This dummy
tries to capture a socio-economic relationship in the Global South dimension.
Also, even if most of the emerging countries were colonies, a specific metric
to evaluate the impact of recent/lasting colonial powers is interesting and has
widespread application in the literature. However, based on CEPII (2011), we
included not only the traditional and lasting colonial empires, such as the British
and the French, but countries that lost territories that became independent
countries since then (such as South Africa and Australia), the heirs of republic
unions (such as the Russian Federation and Serbia), and countries that hold
a dominion over some other territorial entity to this day (such is the case for
Morocco). This dummy tries to capture a geopolitical relationship in the Global
South dimension.
Lastly, due to the high availability of data and the institutional background
linking Higher Education (more specifically international mobility) and OECD
countries, distinguishing between this organization’s members and other states
seems a fruitful effort in testing for a Global South dummy. Sure, not all countries
in the ‘club of the rich’ are developed, such are the cases of Chile, Mexico, and
Turkey. However, due to a more ‘institutional’ approach fomented by the minds
at the OECD and the path dependence of the organization, we opted to also
include the emerging countries in OECD as positive cases for this dummy.
Table 3 displays the outputs of the regression models for “Dataset_World”
with the inclusion of the three dummies. The equation with the highest adjusted
R-squared and F-value – model (9) – for “Dataset_World” is described as follows.
All the variables are indexed by “i” (i = 1, […], 113).
yinbound_samplei = β0 + β1col_pow45i + β2langi + β3arwu500i + β4mig_
samplei + β5ln_dist10i + β6ln_incomei + β7 ln_populationi + εi
Rafael de Moraes Baldrighi
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
19-24
Table 3 – Inbound Internationally Mobile Students in the World
Dependent
Variable
(5) (6) (7) (8) (9) (10)
inbound_sample
emerging -0.0046366**
(0.0022677)
col_pow45 0.0071923***
(0.00195)
0.0071977***
(0.001924)
0.0073857***
(0.0019401)
oecd 0.0030197
(0.0024209)
lang 0.0018903
(0.0014507)
0.0026073*
(0.0014716)
0.0024168*
(0.0013782)
0.0022028
(0.0014684)
0.0024192*
(0.0013668)
0.0021949
(0.0013598)
arwu500 0.0007822***
(0.0001081)
0.0007548***
(0.0001073)
0.0007877***
(0.0001021)
0.0007779***
(0.0001079)
0.0007885***
(0.0000937)
0.0007701***
(0.0001006)
mig_sample 0.3881636***
(0.0643785)
0.3998374***
(0.0636816)
0.3662754***
(0.0611196)
0.3925821***
(0.0643063)
0.3658788***
(0.0578457)
0.3810935***
(0.0592626)
ln_dist10 0.0000755
(0.0013482)
0.0010699
(0.0014145)
0.0001037
(0.0012739)
0.0004147
(0.0013719)
0.0001064
(0.0012614)
-0.0000609
(0.001263)
ln_income 0.0010745
(0.0008314)
0.0005992
(0.0008515)
0.0007834
(0.0007896)
0.0008164
(0.0008547)
0.0007957
(0.000529)
ln_population 0.0009951**
(0.0004592)
0.0008778*
(0.000456)
0.0005365
(0.0004513)
0.0007952
(0.0004852)
0.000535
(0.0004437)
0.0004838
(0.0004482)
liv_cost 0.0023953
(0.0049826)
-0.0010845
(0.0051954)
0.0000996
(0.0047489)
-0.0008528
(0.0056103)
0.0035837
(0.003197)
constant -0.0267969**
(0.0125074)
-0.0227745*
(0.0124782)
-0.0170375
(0.0121107)
-0.0229148*
(0.0128568)
-0.0170894
(0.0117991)
-0.0098933
(0.009737)
Observations 11 3 11 3 11 3 11 3 11 3 11 3
Adjusted
R-squared 0.9024 0.9052 0.9128 0.9029 0.9136 0.9128
F-Value 148.88 134.74 147.61 131.45 170.31 168.58
Note. Standard errors are presented in parenthesis. *significant at 10%. **significant at 5%. ***significant at 1%.
The increase in the number of observations alone (5) improves the explanatory
power of our model significantly. As expected, since all the independent variables
were selected after a literature review of papers analyzing OECD countries.
Furthermore, the inclusion of the three dummies also helped in increasing the
models’ adjusted R-squared (models 6,7,8,9, and 10) and F-values (models 9 and
10). Also, the emerging dummy has a negative impact on the dependent variable,
while col_pow45 and oecd are both positively related to inbound_sample.
The determinants of inbound international tertiary students in the developing world [...]
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
20-24
Once again, we conducted two separate models (with either ln_income
or liv_cost) to avoid collinearity issues. In this dataset, the linear correlation
between the two variables is stronger (Pearson’s R is equal to approximately
0.8). This time, the regressions with only one of the two variables (models 9
and 10) yielded higher adjusted R-squared and F-values than when the two were
included (models 5, 6, 7, and 8), possibly due to this stronger linear relation. For
the significant results, all of them produced coefficients similar to the literature
(lang, arwu500, mig_sample, and ln_population).
An interesting finding is derived from ln_dist10 outputs, even if not significant:
the literature frequently presents a negative relationship, which makes sense.
However, not using push-pull dyadic factors and just analyzing the attractiveness
of a destination as a monadic factor may explain its positive coefficient in models
(5,6,7,8, and 9): important inbound destinations are located in the ‘edges’ of the
world, such as the US, Australia, Canada, Japan, Russia, Argentina, and China.
As the literature often employs dyadic factors using OECD countries (which are
mostly from Europe, a continent in which destinations are relatively closer when
compared to the rest of the world), our result seems plausible when expanding
the sample and focusing on pull factors only.
Also, observing our variables in Tables 2 and 3, one could speculate about
the multicollinearity between the number of incoming students, the number of
migrants, and the population of a country. Although the number of inbound
international students may indeed contribute to an increase in the total number
of migrants, which, in its turn, may aggregate in the total population, this
does not happen in our sample. We dealt with this issue by using the variables
inbound_sample and mig_sample, which do not compare the number of incoming
students to the total number of students of a country, nor the proportion of
migrants in a country’s population, but their share in our sample. This eliminates
expected multicollinearity effects when the number of inbound students affects
the number of migrants and, successively, the population. Furthermore, we only
used natural logarithm variables for distance, hosting capacity, and income in
the models for the sake of standardization, and since the regressions, overall,
yielded better outputs when doing so.
Lastly, out of the three dummies, the one that produced the best results was
col_pow45. This is why we excluded income and cost of living (9 and 10) using
this dummy. Not only it brought higher adjusted R-squared and F-values, it was
significant at the 1% level – emerging was at the 5% level and oecd was not
Rafael de Moraes Baldrighi
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
21-24
significant in any of the measures. This strengthens the argument presented in
the literature that a colonial link, or a recent domination relationship, boosts the
flow of international students and that, perhaps, the geopolitical dummy is more
suitable to distinguish the countries than the socio-economic ones (emerging and
oecd). The results, however, were very similar and yielded comparable outputs.
Conclusion
Our conclusions can be summarized as follows:
i. The use of a larger dataset comprising the whole world presents better
results than using emerging countries only. The inclusion of three
dummies, the Global South dimension, also captures this.
ii. Most of the significant results in the ten models followed the literature.
It is worth mentioning the importance of academic excellence (arwu500)
and migration network (mig _sample) in all models – Caruso and De Wit
(2014). Even though yielding a smaller number of significant results in
the regressions, language (lang) and hosting capacity (ln_population)
are also relevant to better understand inbound flows – Beine, Romain
and Ragot (2014).
iii. The use of a given country’s proportion of international students
among its total number of enrollments (inbound _rate) proved to yield
biased results when compared to the proportion of inbound students
per country in the total number of mobile students in our sample
(inbound _sample).
iv. Some variables, even though not significant, produced results that
contrast with the current literature – Rumbley 2012; Didisse, Nguyen-
Huu and Tran 2018. That is the case for ln_dist10 and liv_cost (Table 3).
The first case can be explained by the monadic approach – instead of a
dyadic one – and the inclusion of non-European countries. The second
one is harder to grasp, but due to its high collinearity with ln_income,
when excluding the cost-of-living variable from “Dataset_World” –
model (9) –, the output displayed better results.
This article tried to address a lack in the literature and on databases on
emerging countries’ inbound mobility. Using inbound_sample to measure the
The determinants of inbound international tertiary students in the developing world [...]
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
22-24
dependent variable, adopting more inclusive criteria when estimating the pull
factors, and creating the emerging, col_pow45, and oecd dummies proved to be
effective techniques not only to recognize the need for a different approach to
Global South countries but also to design a quantitative model including them.
The lack of data regarding international mobility in emerging countries is
blatant and we tried to remedy this by choosing the year in UIS’ database in
which data for the Global South is more abundant, which forced us to create a
cross-section and use OLS techniques. Sure, this can make us wonder if our results
were due to the sheer luck of choosing an atypical year for the international
mobility scenario.
However, three reasons can advocate for the relevance of the 2017 results
when compared to other studies: i. the (quasi-)constant nature of the share of
international students per country, their languages, population, and distance
(Didisse, Nguyen-Huu and Tran 2018); ii. the reduced number of observations
(77 and 113) in our sample – other studies with such a limitation, such as
Kahanec and Králiková (2011) and Caruso and De Wit (2014), use OLS as well;
and iii. our results were in accordance with the literature that uses panels
and/or time-series for modeling OECD/European countries, who benefit from
more abundant data. However, if more (dyadic) data for emerging countries
become available in the next years, the creation of a panel or a time-series may
help us better understand the evolution of the pull factors presented here. Also,
future research that overcomes the push-pull dynamics, which are declining in
migration studies, are welcome.
Overall, our models yielded a satisfactory number of statistically significant
explanatory variables and good adjusted R-squared and F-values and the
dummy that seems to better capture the Global South dimension is col_pow45.
These results can and should help policymakers when designing national
policies to attract inbound students and boost HEI internationalization in
emerging countries.
Rafael de Moraes Baldrighi
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
23-24
References
Beck, Kumari; Michelle Pidgeon. 2020. Across the Divide: Critical Conversations on
Decolonization, Indigenization, and Internationalization. In International Education
as Public Policy in Canada, 384-406. Edited by Merli Tamtik, Roopa Desai Trilokekar,
and Glen A. Jones. Québec, Canada: McGill Queen’s University Press.
Beine, Michel; Romain, Noël; Ragot, Lionel. 2014. Determinants of the international
mobility of students. Economics of Education Review, no. 41: 40–54.
Buckner, Elizabeth; Sharon Stein. 2019. What Counts as Internationalization?
Deconstructing the Internationalization Imperative. Journal of Studies in International
Education 24, no. 2: 1-16.
Caruso, Raul; De Wit, Hans. 2014. Determinants of Mobility of Students in Europe:
Empirical Evidence for the Period 1998-2009. Journal of Studies in International
Education 19, no. 3: 1-18.
CEPII. 2011. CEPII’s distances measures: the GeoDist Database CEPII. Paris, France.
Chen, Tse–Mei; Barnett, George A. 2000. Research on International Student Flows
from a Macro Perspective: A Network Analysis of 1985, 1989 and 1995. Higher
Education 39, no. 4: 435-453.
CIA. 2023. The World Factbook, Field Listing: Languages. https://www.cia.gov/library/
publications/the-world-factbook/fields/402.html Retrieved October 28, 2020.
Didisse, Jonas; Nguyen-Huu, Thanh T.; Tran; Thi A-D. 2018. The Long Walk to Knowledge:
On the Determinants of Higher Education Mobility to Europe. The Journal of
Development Studies 55, no. 6: 1-21.
Docquier, Frédéric; Rapoport, Hilel. 2012. Globalization, brain drain and development.
Journal of Economic Literature 50, no. 3: 681–730.
Junqueira, Cairo G. B.; Baldrighi, Rafael de M. 2020. A internacionalização do ensino
superior no Brasil analisando comparativamente a mobilidade internacional de
estudantes face à realidade latino-americana. Carta Internacional 15, no. 3: 29-54.
Kahanec, Martin; Králiková, Renatá. 2011. Pulls of international student mobility. IZA
Discussion Paper No. 6233, Bonn, Germany: IZA.
Mazzarol, Tim; Soutar, Geoffrey N. 2000. Push-pull Factors Influencing International
Students Destination Choice. The International Journal of Educational Management
16, no. 2: 82-90.
Mignolo, Walter D. 2002. Geopolitics of Knowledge and the Colonial Difference. The
South Atlantic Quarterly 101, no. 1 57-96.
OECD. 2011. Education at a glance 2011, Paris, France.
OECD. 2019. Education at a glance 2019, Paris, France.
The determinants of inbound international tertiary students in the developing world [...]
Rev. Carta Inter., Belo Horizonte, v. 19, n. 2, e1517, 2024
24-24
Pedro, Eugénia; Franco, Mário. 2015. The importance of networks in the transnational
mobility of higher education students: Attraction and satisfaction of foreign mobility
students at a public university. Studies in Higher Education 41, no. 9: 1627–1655.
Rosenzweig, Mark R. 2008. Higher education and international migration in Asia: Brain
circulation. Annual World Bank conference on development economics, 59–100.
Rumbley, Laura. 2012. So many data, so little clarity. In Internationalisation Revisited:
New dimensions in the Internationalisation of Higher Education, 125-134. Edited
by J. Beelen and H. De Wit. Amsterdam, The Netherlands: CAREM.
Un Desa. 2019. United Nations Population Division. https://www.un.org/en/development/
desa/population/migration/data/estimates2/countryprofiles.asp Retrieved October
16, 2020.
Unesco. 2020. Unesco Institute For Statistics. http://data.uis.unesco.org/ Retrieved
October 20, 2020.
Van Bouwel, Linda; Veugelers, Reinhilde. 2013. The determinants of student mobility
in Europe: The quality dimension. European Journal of Higher Education 3, no.
2: 172–190.
World Bank. 2020. Price level ratio of PPP conversion factor (GDP) to market exchange
rate. https://data.worldbank.org/indicator/PA.NUS.PPPC.RF?end=2019&start=2013
Retrieved November 5, 2020.