International applicability of education and migration aspirations

Abstract

1. Introduction

International migration generates major efficiency gains in the global production (Borjas 1995; Clemens 2013). As migrants from most countries are better educated than nonmigrants (Docquier et al. 2009; Grogger and Hanson 2011), the effects of migration are larger than the overall share of migrants in global population (3.6 per cent in 2020) alone would suggest. In addition to reallocating human capital between countries, international migration potentially affects incentives to invest in human capital. Mountford (1997) and Stark et al. (1997) showed that if the mobility of the educated is stochastic, a relatively low probability of being able to migrate from a poor to a rich country may generate brain gain through brain drain. The prospect of being able to migrate to a rich country encourages all potential migrants to invest more in their education, and the human capital of those who are not able to migrate benefits the country of origin. Beine et al. (2001) analyze cross-sectional data for thirty-seven developing countries and conclude that the possibility of a beneficial brain drain “could be more than a theoretical curiosity.” In addition to changing overall investment in human capital, the possibility of migration tends to make internationally applicable education more attractive to students (Poutvaara 2004, 2008). The effects can be substantial. Abarcar and Theoharides (2024) analyze the effects of changes in US visa policies for nurses on educational choices in the Philippines. They find that increased job opportunities in the USA increased both enrollment and supply of nursing programs in the Philippines.

In this article, we analyze perceptions on the international applicability of one’s education in different study fields and in different countries and how these are related to study effort and whether students would like to emigrate. Our empirical analysis is based on student surveys in nineteen universities in seven countries: Czechia, India, Indonesia, Italy, Mexico, the Netherlands, and Spain. In addition to background information such as the study field, students were asked to give an assessment on how applicable their education is in their home country and abroad if they were to move. They were also asked if they would ideally like to permanently emigrate, and whether they currently have plans to emigrate. We refer to the desire to permanently emigrate as migration aspirations, and to current plans to emigrate as migration intentions.

We start by presenting a theoretical framework on decisions by students on how much effort to invest in their studies, whether to plan to emigrate if the realization of a stochastic mobility component allows it, and how these decisions depend on the perceived applicability of one’s education abroad. After that, we study perceptions of the international applicability of one’s education and how these perceptions are correlated with migration aspirations and intentions, and time spent studying. When doing so, we control for potential confounding factors, like willingness to take risks and patience that can be expected to influence both study field choices and desire to migrate. Importantly, we also use between-individual variation in perceived applicability of one’s education to study how this is related to differences in migration aspirations and intentions among students studying in the same field. We include university fixed effects to capture any systematic differences between universities, whether in the quality of education, in the language of teaching, or in the socio-economic background of their student body.

We find substantial differences in the perceived international applicability of one’s education between study fields and countries. Across countries and institutions, students in law are much less likely to find their education to be internationally applicable, both in absolute terms and relative to the applicability of their education in the home country. This pattern prevails in all countries from which we have law students in our sample (Czechia, Mexico, and Spain). Students in STEM degrees and in social sciences find their education most internationally applicable, with three out of four (75 per cent) viewing their education as very applicable or applicable, compared with somewhat more than half of economics and business students (56 per cent) and about two of five law students (41 per cent). In terms of countries, Dutch and Mexican students view their education as most internationally applicable, and Czech students as least applicable. This cannot be explained by different distributions of study fields among respondents from different countries: Czech students in all study fields view their education as less internationally applicable than students studying the same field in each other country, although differences are not always statistically significant.

Our empirical analysis also confirms that the study field and perceived international applicability of one’s education are strongly correlated with migration aspirations and intentions. Importantly, perceived international applicability predicts migration aspirations and intentions also after controlling for study field, university fixed effects, individual demographic background variables, migration networks, as well as risk attitudes and patience, and perceived applicability at home. The association of perceived international applicability with migration aspirations is strong for both genders, while the association with plans to migrate is driven to a large extent by women who would ideally like to work full time.

While we do not claim causality of these effects, we are able to control for a rich set of potentially confounding characteristics. A test following Oster (2019) in Supplementary Appendix A suggests that the selection on unobserved variables would have to be 1–6 times as large as selection on observables to nullify the relationship between perceived applicability and migration aspirations and intentions.

On the theory side, our article is most closely related to Poutvaara (2008) which analyzes public and private educational choices when there are two types of human capital: internationally applicable and country specific. Poutvaara (2008) shows that an improved international applicability increases students’ private incentives to invest in study effort. If students can freely decide on their study field, a higher share chooses internationally applicable education when its applicability abroad improves. The government, instead, has an incentive to reduce the number of study places in internationally applicable education, like economics and engineering, and provide more country-specific education, like law, to keep a larger tax base. We extend Poutvaara (2008) by including multiple study fields, as well as by allowing both returns to and costs of education to differ by gender and country. Other related papers include Andersson and Konrad (2003) and Thum and Uebelmesser (2003) which highlight that the mobility of university graduates helps to alleviate time consistency problems by boosting tax competition. The government’s decision problem has an interesting parallel to the firms’ decisions on providing general and firm-specific training that Becker (1993) analyzed. Just like firms which are unwilling to pay for general training, governments can be expected to be reluctant to pay for internationally applicable education. The main limitation of our model is that we assume that the study field is exogenous. This limitation reflects the corresponding restriction in our survey data. As we interview only university students who have already chosen their study field, our data would not allow testing any predictions related to how the choice of study field depends on willingness to emigrate in the future.

The rest of the article is organized as follows. In Section 2, we present a theoretical framework on how much effort students invest in their education when facing uncertainty about their future mobility, and how this investment depends on the expected returns to human capital at home and abroad given their study field. Section 3 presents the data and the empirical strategy. Section 4 studies students’ perceptions of the international applicability of their education, across study fields and countries. In Section 5, we analyze how perceptions of the international applicability of education are associated with aspirations and intentions to emigrate, both with and without study field and university fixed effects. We also examine how this relationship depends on gender and whether one would like to ideally work full time after graduation. Section 6 concludes.

2. Theory

We analyze decisions by students on how much effort to devote to their studies and whether to plan to emigrate, as well as how these decisions depend on expected returns to their education in the home country and abroad and on migration costs. There are S study fields, denoted by index s, two countries, denoted by A and B, two genders, denoted by index g with values f for female and m for male, and two periods. In the first period, individuals are students and decide how much effort to invest in their studies, given their gender-specific expectations on future labor market returns to such investments at home and abroad and individual migration costs. In the second period, individuals are graduates. At the beginning of the second period, they learn a stochastic component related to their mobility and decide then whether to migrate in case they can migrate, and work the remainder of the second period. Working time is normalized to one. We analyze choices of students in country A after they have entered a university and chosen their study field, to be in line with our empirical analysis that is based on surveys of students who have already chosen their study field. We include gender when considering expected labor market returns to education, given large gender wage gaps that vary across countries (OECD 2024) and study fields (Goldin 2013; Quadlin et al. 2023).

In Equation (1), $αsg$ is a gender-specific basic component for those with study field s in country A and $rsg$ denotes the gender-specific after-tax return to human capital for those with study field s in country A. Modelling log wages as a function of gender-specific returns to human capital is in line with Aksoy and Poutvaara (2021). Whereas Aksoy and Poutvaara (2021) analyzed how the self-selection of female and male refugees and irregular migrants with given stocks of human capital depends on risks related to staying in an unsafe country of origin and risks related to migration, our focus is on the interaction between human capital investments and migration plans.

The utility cost of accruing human capital stock $hj$ is the product of $hj2$ and a gender-specific constant $γg$⁠, $γghj2$⁠. The gender-specific term may reflect gender differences in both psychological effort costs and opportunity costs of time spent studying.

Migration cost has two components. The cost of migration in case of being mobile, $cj$⁠, is known already in the first period, and could be negative for students with a very strong preference for living in B. The stochastic component of the actual mobility status is revealed only at the beginning of the second period, with students being mobile after their graduation with probability p, $0<p<1$⁠. That students’ actual realization of mobility is disclosed only after completing their education may reflect uncertainty related to realized language and cultural skills and family formation, as suggested by Poutvaara (2007), or uncertainty regarding immigration policy in the intended destination country, in case it is not part of a common labor market. The mobile ones can migrate freely with the individual-specific cost while the nonmobile ones cannot migrate as they face infinitely high mobility costs.

Note that the utility with strategy N is deterministic, while the expected utility with strategy M is stochastic. The reason for this difference is that students can always stay, while the realization of the mobility status is stochastic. It is optimal to pursue human capital investment strategy M if and only if $EUjM>UjN$⁠.

To derive empirically testable predictions on how the choice between strategies N and M depends on perceived applicability of one’s education at home and abroad and on risk attitudes, assume next that $ci$ has two components. Both the student and the researcher can observe parameter $σj$⁠, capturing factors like risk aversion. Only the student can observe idiosyncratic parameter $ϵj$ that captures individual-specific taste for migration. It is independently and identically distributed, with density function $ϕ(ϵ)$ and cumulative distribution function $Φ(ϵ)$⁠, obtaining negative values for those students who have an inherent preference for living in the other country. In our empirical analysis, we also test the possibility that family background, measured by parental education, could affect plans to emigrate.

The comparative statics effects with respect to the probability that a student j pursues strategy M are given by:

Part (i) of Proposition 1 implies that expected higher returns to human capital abroad unambiguously increase the probability of pursuing strategy M as does, by part (ii), also the probability of being mobile and by part (iii), the fixed compensation component abroad. The fixed compensation at home and observable migration cost component, instead, reduce the probability of pursuing strategy M (part (iii)). Part (iv) shows that the effect of higher returns to human capital at home may go either way, depending on the relative returns to human capital at home and abroad, and the probability of a graduate becoming mobile. The probability of pursuing strategy M is decreasing in returns to human capital at home if $rsg≥(1−p)(2−p)r^sg$ and increasing in returns to human capital at home if $rsg<(1−p)(2−p)r^sg$⁠.

The intuition for the decreasing probability of pursuing strategy M when $rsg$ increases if $rsg≥(1−p)(2−p)r^sg$ follows the same logic as part (i): higher returns to human capital in A make staying a more attractive strategy. The intuition for the opposite pattern if $rsg<(1−p)(2−p)r^sg$ is that with very low returns to human capital in the home country and a low probability of becoming mobile, students would choose such a low level of human capital that it does not pay off to emigrate even if being mobile. In such a low-investment equilibrium, an increase in returns to human capital in the home country would reduce the cost of strategy M in case of not becoming mobile. This could, in turn, spur more students to pursue strategy M. This finding is related, yet distinct, from previous literature that has shown that the possibility of emigration may encourage investment in human capital in a poor country of origin, resulting in brain gain through brain drain (Mountford 1997; Stark et al. 1997; Beine et al. 2001). We show, instead, that an increase in gender and study field-specific returns to human capital at home may result in an increase in emigration even if the probability of being able to migrate does not change, as a side effect of increased effort investment by students. The overall effect on the human capital stock would depend on the relative magnitude of this induced brain drain effect, compared with the brain gain from higher effort investment by those expecting to stay and those investing more in their human capital hoping to be able to reap higher benefits to their effort investment abroad.

Parts (i), (ii), and (iii) of Proposition 1 give our first two testable predictions:

Equations (3) and (4) imply that investment in education is always increasing in expected returns at home among both those pursuing strategy M and those pursuing strategy N, and also in expected returns abroad among those who would like to emigrate. This gives three further testable predictions:

A challenge in empirical testing of Hypothesis 1 is that perceptions of returns to human capital at home and abroad may be correlated. Therefore, we also analyze a modified version that relates to the difference in the perceived applicability of one’s education abroad relative to perceived applicability at home:

Our model highlights the importance of gender-specific returns to human capital and allows also for gender-specific differences in effort or opportunity costs of investment in study effort. More limited labor market opportunities after graduation would discourage female investment in human capital. On the other hand, more limited job and leisure opportunities for women during their studies would reduce $γf$ relative to $γm$⁠, pushing women’s investment in their human capital in the opposite direction of women investing more in their human capital due to lower opportunity costs. As a result, it could be that women in a country with a high degree of gender discrimination could still invest more in education than men, especially if social norms restrict their leisure opportunities. In the empirical analysis, we control for gender to evaluate empirically which of these mechanisms dominates, in case there are systematic gender differences in educational investments. We also do our main analyses separately for men and women.

3. Data and empirical strategy

3.1. Data

To test our hypotheses, we use data from online surveys carried out in close cooperation with different universities. The participating universities were:

• In Czechia: Masaryk University, University of Ostrava, and University of Economics Prague (VSE),

• In India: IIT Kanpur and Ashoka University,

• In Indonesia: Institut Pertanian Bogor, Universitas Indonesia, Institut Teknologi Bandung, Politeknik Manufaktur Bandung, and Universitas Padjadjaran,

• In Italy: Università Cattolica del Sacro Cuore,

• In Mexico: El Colegio de México (COLMEX), Centro de Investigación y Docencia Económicas (CIDE), Instituto Tecnológico Autónomo de México (ITAM), and Universidad Nacional Autónoma de México (UNAM),

• In the Netherlands: Maastricht University,

• In Spain: Universitat Autonoma de Barcelona (UAB), University of Barcelona, and Carlos III University of Madrid.

We reached respondents by sending an invitation link via university or faculty email lists. Out of the 10,992 students who had opened the link to the questionnaire, 6,084 started to reply and 3,753 finished the whole survey. This corresponds to a completion rate of 34.1 per cent among those who opened the link to the survey and 61.7 per cent among those who started to reply. Four thousand and forty-nine respondents answered the relevant questions for our analysis regarding their study field, the perceived applicability of their education abroad and at home, as well as their migration aspirations and intentions. These are 36.8 per cent of those who opened the link to the survey and 66.6 per cent of those who started to reply. The surveys were conducted between 1 April 2019 and 7 April 2020. We present the target population and number of respondents by universities in Table 1. Data used in this article and replication package of all our analyses are available in the Supplementary data on the journal website.

It should be noted that the universities are not representative of all universities in the country they are located in; we were restricted to carrying out the survey in universities willing to send out the invitation to their students. Additionally, there is self-selection among students in whether they reply, which is a common challenge in surveys. Therefore, estimated distributions of beliefs and migration preferences may differ from the general student population. With our data, we are able to address the degree of self-selection of students in the final analysis sample compared to those students who started the survey. Supplementary Table B1 shows results of an analysis for self-selection in answering questions on migration aspiration and intention, which were asked in the last section of the questionnaire used in this analysis. None of the key variables in our subsequent analysis is statistically significantly related to the probability of being in the final sample. Intuitively, more patient respondents are less likely to drop out of the survey. We cannot check the degree of selection in the initial sample of students who started the survey. However, robustness analyses excluding different countries do not suggest that sample differences across countries drive our results.

Our data contain rich information on respondents from different study fields and universities concerning their studies and career plans as well as their socio-economic background. Table 2 shows summary statistics for key variables in our sample of respondents. Of particular interest for this study are survey questions asking respondents about how they perceived the domestic and international applicability of their education, whether they would ideally like to live permanently in another country and whether they already made plans or started to prepare for a move abroad. Applicability abroad and applicability at home are measured on scales from 1 to 5 with 1 being “Not at all applicable,” 2 being “Not very applicable,” 3 being “Somewhat applicable,” 4 being “Applicable,” and 5 being “Very applicable.” We define migration aspirations as ideally wanting to move permanently abroad (“Ideally move”) based on a binary answer of “Yes” or “No.” We define migration intentions as having concrete plans such as a moving date (“Migration plans”) as opposed to the other categories of “Thoughts, but no plans,” “In principle yes, but no thoughts,” and “Never move.” We perform robustness checks for the functional form of the multinomial variables, that is applicability of one’s education and migration intentions. We also asked respondents about their study effort and about their time and risk preferences, which have been shown to play an important role for migration decisions (Jaeger et al. 2010). We also ask whether they have family or friends abroad. Men and women in our sample differ along several dimensions: men view their education as more applicable abroad and are less likely to want to ideally move. Women are less willing to take risks compared to men. Men are more likely to study STEM majors and women are more likely to study social sciences, humanities, law, and medicine. Women are also more likely to have both friends and family abroad than men.

3.2. Empirical strategy

Our theoretical framework focuses on the relationship between the perceived applicability of one’s education abroad (“Applicability abroad”) and migration aspirations and intentions. In the empirical analysis, we first analyze in Section 4 perceptions of the international applicability of different study fields. We then analyze in Section 5 the role that different study fields play in the context of migration aspirations or intentions. If international applicability varies by study field, we would expect study fields that are generally seen as more internationally applicable, like STEM, to be linked with a higher probability of migration aspirations and intentions. At the same time, individual perceptions of international applicability within study fields might also contribute to individual-level variation in migration aspirations and intentions. We test this in our regressions. In the third set of analyses in Supplementary Appendix C, we link the perceived international applicability of education and migration aspirations with investment in education, measured by how many hours the respondents report to study.

Our final sample focuses on respondents who are citizens or permanent residents in the country they are taking the survey in, thereby excluding temporary students who are already selected on their realized mobility. In addition, we exclude students who are majoring in multiple subjects. While there might be interesting interactions between subjects that inform about perceived applicability and migration aspirations and intentions, we do not have information on which field is the primary study field. Due to our sample size, we cannot estimate a fully nonparametric specification with dummies specific to field combinations. Consequently, we focus on those with a single major (80 per cent of our sample as shown in Table 2).

In one specification, M represents a dummy variable equal to one if the respondent has migration aspirations. In the other specification, it is a dummy variable equal to one if the respondent has migration intentions.

It is important to note that we cannot identify causal relationships in our analysis. An unobserved and idiosyncratic desire to migrate may drive the selection into the study field, as well as perceived applicability abroad and migration aspirations and intentions. We discuss this issue further in Subsection 5.3.

4. Perceptions of international applicability of education

Figure 1a shows how applicable respondents believe their education to be abroad—across study fields and countries in our sample. Respondents assessed the applicability of their education on a scale from 1 (not at all applicable) to 5 (very applicable). The modal answer for all study fields except law is that respondents believe their education to be applicable abroad, whereas for law the modal answer is that education is somewhat applicable abroad. This matches well with our theory presented in Section 2. Knowledge in law is highly country-specific and legal education does not transfer easily to another country so it is reasonable that law students assess the international applicability of their subject lower than students in other fields. Combining the top two categories of applicability, STEM and Social Sciences are viewed as applicable or very applicable abroad by 75 per cent of respondents. Among students of economics and business, 56 per cent of respondents think that their education is applicable or very applicable abroad. For law students, the share of those who report their education to be applicable or very applicable is the lowest (41 per cent). The shares also vary considerably by survey country. Almost 80 per cent of respondents in Mexico (77.5 per cent) and the Netherlands (78.1 per cent) think their education is very applicable or applicable across study fields. Conversely, among respondents in Czechia, only 38.5 per cent selected one of these two categories. Across all countries the fraction of those thinking their education is not at all applicable is very low.

Figure 1.

Perceived international applicability by study field and country. Notes: (A) shows the distribution of perceived applicability among our respondents, first by study field (top panel), then by country in which the survey was taken (bottom panel). We focus on respondents who study a field with at least ten respondents within a given country to minimize noise. (B) shows variation in perceived applicability abroad by country and field. The plot shows the predicted values and 95% confidence intervals from a linear regression of perceived applicability on fixed effects for country-by-field combinations. We restrict the sample to fields with at least ten respondents within a given survey country.

Open in new tabDownload slide

Of course, given the different distributions of study fields across different countries, looking at distributions only across countries or study fields can be misleading. Therefore, Fig. 1b breaks down the responses by study field and country. Law students assess the applicability of their education abroad to be lower than students in other fields. This confirms the idea that law degrees are country specific. Among respondents in Czechia and Spain, STEM students consider their knowledge to be more applicable than respondents in other study fields, while among respondents in Mexico, this is particularly the case among humanities students. In India and Indonesia, we do not observe much difference in perceived applicability across study fields. As there is considerable variation in perceptions of applicability across fields but also across countries, controlling for country or institution fixed effects in our analysis is important to isolate the effects of study field or idiosyncratic variation.

In Table 3, we analyze how different study fields and further observable characteristics are related to how applicable respondents perceive their education to be abroad. We estimate in a linear regression model how individual characteristics explain the perceived applicability of one’s education abroad. We start by including only study fields in column (1). Relative to law, the reference category, other study fields are positively and statistically significantly related to a higher perceived applicability of respondents’ education abroad. The results are robust to the inclusion of university fixed effects in column (2) and further variables in columns (3) and (4). We control for university fixed effects in columns (2), (3), and (4) to capture student selection into university based on unobserved characteristics related to perceived applicability of one’s education abroad. University fixed effects can also account for between-university differences in the language in which one studies, or the international language proficiency of students across countries and even universities. Studying for a doctoral degree is associated with higher perceived international applicability.

Importantly, we are able to proxy for the existence of migrant networks by controlling for whether a respondent has family or friends abroad. Having family abroad is strongly positively correlated with perceived international applicability—potentially capturing the fact that those with family abroad are better informed about opportunities outside of their home country. On the other hand, having friends abroad is correlated to a much lower degree with perceived applicability. Nevertheless, the addition of these important controls does not alter the correlation between study fields and applicability much. Other variables do not seem to matter much; for example, there are no systematic gender differences in perceived international applicability after controlling for university and study-related factors as well as other individual characteristics. In column (4) we additionally include a measure for risk attitudes (on a scale from 1, very risk averse, to 10, very risk prone) and patience (on a scale from 1, not at all patient, to 10, very patient). Results indicate that respondents who report being more patient and less risk-averse perceive their education to be more internationally applicable after controlling for the above-mentioned variables. Higher perceived applicability among patient respondents could be related to their generally higher study effort, which then also boosts the international applicability of their education. Higher perceived applicability among those who are less risk averse could be explained by their willingness to try their luck in new circumstances; it could also be that risk aversion is related to more pessimistic attitudes towards big changes like migration. In Supplementary Table A1, we also control for parental education as an important social factor that might affect students’ opportunities. Controls for this are statistically insignificant and adding those has only minor effect on our estimates.

The perceived applicability variable has five categories, and we have, thus far, been treating it as continuous. We relax this assumption in Supplementary Table A2 by estimating an ordered probit model that treats each of the five categories separately. We run this model controlling for the maximum number of controls (Table 3, column 4). Supplementary Table A2 shows the predicted probabilities from the ordered probit model for each of the categories of perceived applicability for every study field. These results tell the same story as the results from our linear specification—the predicted probability of those studying law to consider their education either “applicable” or “very applicable” (columns 4 and 5) is substantially lower than the same predicted probability for all other study fields. On the flip side, the predicted probability in the other categories that imply lower perceived applicability is higher for law compared to all other study fields. Given the very similar results from the ordered probit model, we continue to treat perceived applicability as a continuous variable in our analysis.

Our theoretical predictions take into consideration that individuals compare returns to education abroad to those at home. Therefore Table 4 examines how the difference in perceived applicability abroad and perceived applicability at home varies between study fields. We find that coefficient sizes for study field dummies are even larger when we consider the difference in applicability than in the previous table. This is suggestive evidence that relative to law, respondents in other study fields perceive their education to be more applicable abroad relative to at home. In Supplementary Table A3, we show that controlling for parental education has only minor effect on our estimates, and the estimated effects of parental education are statistically insignificant.