Keywords

1 Introduction

International student mobility is not a new phenomenon (Altbach 2005; Guruz 2008). However, there has been a significant growth in scale in the last few decades. The Organization for Economic Cooperation and Development (OECD) estimates that the number of international students across the world has increased from 0.6 million in 1975 to 2.4 million in 2004 (OECD 2006) and a staggering 4.6 million in 2015 (OECD 2017). In Europe, international student mobility is regarded as one of the most important drivers for synchronizing the continent’s disparate higher education systems to the same heartbeat.

In this context, the Erasmus program is considered the “flagship of European co-operation” (Barblan 2002) in higher education. Approximately 4.4 million higher education students have participated in the Erasmus+ program in the three decades since the program was set up in 1987, and the program continues to steadily increase in popularity (European Commission 2017). This year, the European Commission has pledged “to triple the Erasmus+ budget (going beyond the Junker Commission’s proposal to almost double the envelope)” (Rubio 2019, (1) for the 2021–2027 programming period. The policy decisions at the European level are mirrored by national and institutional trends of students’ participation in Erasmus outgoing mobilities (see Fig. 1):

Fig. 1
figure 1

Participation in Erasmus outgoing mobilities: national trends for Romania and institutional trends for the West University of Timisoara Source: Compiled by authors from data from The National Agency for Community Programmes in the Field of Education and Vocational Training (ANPCDEFP)—Erasmus programme data, The National Council for Financing Higher Education (CNFIS)—student cohorts country level data, The West University of Timisoara—student cohorts institutional data

Full size image

Notwithstanding the increasing popularity of and investment in international student mobility, the actual labour market benefits for individuals are still widely debated (Di Pietro 2019). One of the major claims has been that individuals who study abroad enjoy better labour market outcomes than their non-mobile peers (Wiers-Jenssen and Try 2005, Wiers-Jenssen 2008, 2011). Specifically, both credit and degree mobility are said to lead to a better insertion into the labour market (and thus to decreased unemployment), above-average earnings, a more prestigious occupation, and a higher likelihood that graduates will work outside their country of permanent residence/citizenship.

However, the actual impact of student mobility on labour market outcomes is less clear, as research is scarce, the evidence used is often “qualitative and anecdotal” or is prone to bias (Di Pietro 2019). Traditionally, the effect of mobility on employability has been measured using interview or survey data collected from (1) students who participated in mobility programs, (2) university administrators in charge of study abroad programs, and (3) employers (Di Pietro 2015, 2019). While these types of studies provide important insights about the benefits associated with international student mobility they can be affected by social desirability bias (Bowman and Hill 2011) and they rarely account for causality.Footnote 1 Thus, linking mobility and employability in a causal relationship is challenging due to selection and self-selection effects, i.e. “[s]tudents who study abroad may differ from students who do not in unobserved characteristics that are likely to affect labor market outcomes” (Di Pietro 2019), and omitted variable bias which can lead to over-estimating or under-estimating the impact of international mobility on employability.

Various econometric techniques can be used to mitigate the selection problems inherent to observational data, including the use of fixed effects, instrumental variables, various propensity score matching techniques, and regression discontinuity designs (Schneider et al. 2007). While such techniques are invaluable in reducing bias that results from omitted variables and various forms of selection, no analytic technique can provide valid estimates if the data on which the analysis is performed is of questionable quality. Obtaining high-quality data by means of surveys is highly resource-intensive. In this chapter, we take an alternative approach, that of using register data to answer questions about the benefits of international student credit mobility.

According to Andersson and Nilsson (2016, 4), in national (or institutional) contexts in which “there is access to national registers that cover the entire population”, register data on income, occupation, unemployment (Nilsson 2017, 79) enable more “penetrating” (Andersson and Nilsson 2016, 4) and more cost-efficient analyses. Using register data offers the possibility to use already existing population-level data and compare the actual employment outcomes of mobile and non-mobile students. In this chapter, we present an analysis based on register data from university records and national employment and baccalaureate exam records of 8 cohorts of graduates between 2007 and 2014 from the West University of Timisoara (UVT), a leading comprehensive university in Romania. By demonstrating the utility of pre-existing data sources in answering policy-relevant research questions through the case of this single institution, we want to send a broader message to ministers of education and higher education leaders: to release existing register data to the research community. In this way, the linkages between education and labour market outcomes can be rigorously and efficiently tested, descriptions of population parameters from which samples are drawn can be more robust, and policy-makers and institutional leaders can have access to the evidence needed to make informed decisions.

In order to illustrate the utility of combining data from pre-existing registers, the chapter analyses the impact of credit mobility on labour market outcomes using institutional- and national-level data. In our analyses, we sought to answer the following question: Does credit mobility have a positive impact on graduate employability? To answer this overarching question, we establish the predictors of international student mobility and test whether credit mobility is significantly positively associated with labour market insertion, income levels and occupational prestige. Specifically, we address the following research questions: (1) What are the predictors of participation in Erasmus mobility among the specific population of graduates we analyse? (2) Does participation in the Erasmus student mobility program predict insertion on the labour market within that population? Among those graduates who had an active work contract, is participation in the Erasmus student mobility program predictive of (3) an above-average salary or (4) having a more prestigious occupational category?

To answer these questions, the chapter proceeds as follows. First, an overview of existing studies exploring the relationship between international student mobility and employment outcomes is provided. Second, the methodology employed and the analytic sample are described. Third, the results of the data analysis are presented and discussed. Finally, the concluding section highlights avenues for further research and makes an argument for the benefits of using existing register data to test theoretical claims in higher education research.

2 Literature Review

There is a plethora of theoretical arguments linking international student mobility with individual benefits that are expected to translate into better employment outcomes for graduates, but relatively few empirical studies have tested these causal claims (Di Pietro 2015, 2019; Wiers-Jenssen and Try 2005). This section explores the theoretical expectations and empirical evidence brought forward by previous research on the relationship between study abroad and labour market outcomes.

The theoretical expectation linking mobility and employability is that individuals who study abroad will accrue non-economic benefits (i.e., skills, mobility capital) that will ultimately transform into economic benefits (i.e., favourable labour market outcomes such as domestic or international employment, higher wages, higher occupational prestige) (Crăciun and Orosz 2018). This is expected on the premise that the skills acquired by individuals through studying abroad are marketable (i.e., valued by employers) (Di Pietro 2015) and because mobility widens the job search area of graduates beyond the domestic labour market (i.e., more and better employment opportunities) (Di Pietro 2019).

International student mobility is perceived and expected by participants, university administrators, and employers to have a positive effect on all aspects of a worker’s skill set. First, mobility has been shown to have a positive impact on individuals’ cognitive skills, particularly foreign language proficiency (Canto et al. 2013; Llanes et al. 2016), problem-solving, and decision-making skills (Bikson et al. 2003). Second, studying abroad exposes students to foreign cultures which is expected by employers to have a positive impact on their non-cognitive skills, especially inter-personal and inter-cultural skills, confidence and self-reliance (Bikson et al. 2003; Bracht et al. 2006; King et al. 2010; Matherly 2005). Third, mobility can improve job-specific skills for jobs that have an international component or give students the skills and experience to pursue an international career (Di Pietro 2019).

Table 1 The relationship between international student mobility and labour market outcomes
Full size table

Cognitive, non-cognitive and job-specific skills are all valued in the workplace, so the expectation is that they will be rewarded by employers through hiring, higher wages and more prestigious jobs. Table 1 provides an overview of research findings from existing studies on the relationship between mobility and these employability indicators.

Several studies have shown that there is a positive relationship between mobility and various labour market outcomes (European Commission 2014, 2016; Di Pietro 2019; Schnepf and D’Hombres 2018) “though they provide mixed results about the magnitude of this effect” (Di Pietro 2013, 6). Others find no difference between mobile and non-mobile students in terms of employment outcomes (Koda and Yuki 2013; Messer and Wolter 2007;Wiers-Jenssen 2011).

3 Data and Methodology

In order to test whether there is a statistically significant relationship between international credit mobility and employment outcomes, the chapter relies on an original dataset constructed from institutional and national register data. We linked three sources of register data to create the analytic data set: (1) university register data of individuals who completed a bachelor’s and/or master’s degree at UVT, and national register data on (2) baccalaureate exam scores and (3) labour market outcomes. The data comes from an institutional, ICT intensive, tracer study (Proteasa et al. 2018).

The initial dataset was based on university register data of individuals who completed a bachelor’s and/or master’s degree at UVT. This data set contains, among other things, information on UVT graduates’ age, gender, the start year of their bachelor’s and/or master’s degree, the field of study, whether they received social and/or merit-based scholarships during their studies, and whether they participated in Erasmus mobility during their time at UVT. This dataset was then matched with publicly available data on the graduates’ baccalaureate exam scores launched in 2004 (Ministerul Educatiei Nationale 2019) and with information about UVT graduates’ labour market outcomes. Information about UVT graduates’ labour market outcomes was requested and received from ReviSal, a mandatory national register for all employees in Romania that was launched in 2011 (Guvernul României 2011).

Matching ReviSal data with university records allowed us to capture information on whether UVT graduates had an active working contract with an employer operating in Romania, as well as information on the highest salary and highest occupational category associated with each individual’s working contract(s) during the period 2011–2018. This meant that we were able to test whether mobile students actually enjoyed better insertion into the labour market, above-average earnings and/or a more prestigious occupation as compared to non-mobile students.

The raw dataset included information on individuals who completed at least a bachelor’s degree at UVT (n \(=\) 20,707). From this dataset, a number of observations were excluded for various theoretical and practical reasons that are discussed next. First, UVT graduates who could not be matched with ReviSal data were excluded from the analytic dataset, as a lack of match to ReviSal records meant that no conclusions could be drawn about their labour market outcomes. Second, UVT graduates who started their bachelor’s degree program in or before 2007 were also excluded from analysis, as the UVT Erasmus mobility records we had access to only start from the academic year 2007/2008. Third, UVT graduates who started their bachelor’s degree program in or after 2015 were also excluded from analysis, because they could not have completed their studies by January 2018, the date when the ReviSal export was received. Finally, UVT graduates whose baccalaureate exam score was missing were also excluded from the analysis, as we used their performance on this national exam as a proxy for academic ability. As a result of delimitating the raw dataset in this way, we ended up with an analytic dataset of n \(=\) 16,565, which includes information on the labour market outcomes of both mobile (n \(=\) 719) and non-mobile (n \(=\) 15,846) UVT graduates.

Table 2 Descriptive statistics of UVT graduates in the analytic sample
Full size table

Table 2 provides summary statistics for the variables used in our data analysis. The majority of the individuals (88.5)Footnote 2 in our analytic sample were born between 1988 and 1995. They started their bachelor’s degree program at UVT between 2007 and 2014. The average baccalaureate score among UVT graduates who participated in Erasmus credit mobility was 8.95, which is statistically significantly higher (t \(= -\)8.7507; p \(=\) 0.000) than the average baccalaureate score of UVT graduates who did not participate in Erasmus credit mobility (8.63). Moreover, the average age at which UVT graduates completed their bachelor’s degree was 22.7, which is statistically significantly lower (t \(=\) 4.4199, p \(=\) 0.000) than the average age at which UVT graduates who did not participate in Erasmus credit mobility completed their bachelor’s degree (23.4).

To test what predicts labour market outcomes, the log odds of (1) having an active contract in Romania in the period 2011–2017 (insertion), (2) having an active contract that is associated with an above-average salary (earnings), and (3) having an active contract that is associated with a managerial or professional job (occupational category) were modelled as a function of UVT graduates’ gender, field of BA study, year in which they started their BA, their age at graduation from BA (22 or below versus above 22), their performance in the baccalaureate exam (in quartiles), their receipt of a merit-based scholarship or social scholarship during their BA, whether they enrolled in or completed an MA degree at UVT, and whether they participated in Erasmus mobility during their studies at UVT.Footnote 3

For more robust results, we also estimated the relationship between Erasmus credit mobility, labour market insertion, earnings and occupational category with the help of propensity score matchingFootnote 4 models, using the same set of co-variates as the logistic regression models discussed above. Detailed results from the logistic and propensity-score matched models are discussed in the next section, and the regression tables are available from the authors upon request.

4 Data Analysis and Results

(1) What Are the Predictors of Participation in Erasmus Mobility Among UVT Graduates?

Comparing mobile and non-mobile students,Footnote 5 the profile of Erasmus participants becomes apparent. Even though 78% of the mobile students are women, all other things being equal, gender is not predictive of participation in Erasmus mobility among the UVT graduates in our analytic sample. Over the years, higher mobility rates in the Erasmus program have been observed for women, at around 60% (Brooks and Waters 2011; Souto-Otero 2008; Teichler et al. 2011). Previous research has shown that for Romania, the gender gap is even bigger, with females representing 70% of mobile students at the national level (Souto-Otero and McCoshan 2006, 4). However, “[t]he feminisation of higher education is apparent at all levels of study” in national student populations (Orr et al. 2011, 59). Therefore, the tilted balance towards higher female participation rates can be in part accounted for by the general structure of national student populations.

Notwithstanding, other factors are predictive of participation in mobility. First, the year in which students began their BA studies is a positive predictor of mobility: with each year, the likelihood of participating in the Erasmus program was 12% higher. This finding is consistent with the growth in popularity, accessibility and funding of the Erasmus program over the years (European Commission 2017, 2018b), and reflected in the national and institutional trends—see Fig. 1 in the introductory section.

Second, age at the time of BA graduation is a significant negative predictor of credit mobility: each additional year in age is associated with a 28% lower likelihood of participating in the Erasmus program. This finding is in line with the predictions of empirical studies on migration which “overwhelmingly conclude that the relation with age is negative, i.e., that the likelihood of migration decreases with age” (Zaiceva 2014, 4).

Third, academic ability is a predictor of Erasmus mobility. Students’ performance at the baccalaureate exam (our proxy for academic ability) is a significant positive predictor of mobility. A one-unit increase in the baccalaureate exam score is associated with a 47% higher likelihood of participating in the Erasmus program. The receipt of a merit-based scholarship is also significantly positively related to mobility. Those students who received a merit-based scholarship were 3.2 times more likely to participate in Erasmus than those who did not receive such a scholarship. The fact that Erasmus grants are awarded on academic merit and that “Erasmus appears to be much more selective in Eastern Europe (where 20% of applicants are rejected)”Footnote 6 (European Commission 2016) could explain the magnitude of the relationship. There is no evidence of a link between the receipt of a social (i.e., need-based) scholarship and credit mobility in the analytic sample.

Fourth, degree level is significantly positively associated with Erasmus mobility. Compared to those who never enrolled in a master degree program at UVT, those who enrolled but did not complete were 91% more likely to participate in mobility, while those who completed a master degree were 97% more likely to participate in mobility. This makes intuitive sense, as those who remain affiliated with a higher education institution longer have more opportunities to apply for an Erasmus scholarship. All in all, the typical Erasmus mobility participant at UVT is young, academically able and more likely to pursue graduate education.

(2) Does participation in Erasmus mobility predict insertion into the labour market?

Participation in Erasmus mobility is significantly negatively associated with insertion in the domestic (i.e., Romanian) labour market in our analytic sample.Footnote 7 Graduates who were mobile during their studies at UVT (either during bachelor or master degree programs) were 40% less likely to have an active work contract with an employer operating in Romania, compared to non-mobile graduates. Rather than implying that mobile graduates are less likely to be employed, this finding is consistent with the body of evidence that suggests that mobile students are more likely to work abroad after graduating (Parey and Waldinger 2011; Di Pietro 2012; Rodrigues 2013; Teichler and Janson 2007; Varghese 2008) and, thus, less likely to appear in the national employment database with an active contract. As migration research has shown, people with a migratory experience have an increased propensity for re-taking this step. “Once someone has migrated, therefore he or she is very likely to migrate again, and the odds of taking an additional trip rise with the number of trips already taken” (Massey et al. 1993, 453). Through the experience of mobility, students acquire ‘mobility capital’ and are likely to look for and take up jobs outside the domestic labour market (Rodrigues 2013; Wiers-Jenssen 2008).

All other things being equal, gender, baccalaureate exam results, the year when the bachelor degree program started, and the receipt of social scholarship (our proxy for socio-economic status) were not predictive of labour market insertion in our analytic sample of UVT graduates. As previous studies have also shown, field of study is predictive of employment status in our analytic sample. Compared to social scientists, humanities and arts graduates and physical education and sports graduates are significantly less likely to have an active work contract, while graduates from natural sciences, mathematics, biology and biomedicine are significantly more likely to have an active work contract. Also, having a master degree is significantly positively associated with labour market insertion. Compared to those who never enrolled in a master degree program, master graduates are twice as likely to have an active work contract in our analytic sample. This finding is in line with human capital theory predictions.

(3) Among those who have an active work contract, does participation in Erasmus mobility predict an above-average salary?

Erasmus mobility is significantly positively associated with earnings. All other things being equal, those who participated in Erasmus mobility during their studies at UVT were 75% more likely to have a higher-than-average monthly salaryFootnote 8 associated with their active work contract. Results from the propensity score matching model (with the same specification as the logistic regression model discussed above) also suggest that participation in Erasmus mobility is positively associated with an above-average monthly salary among those UVT graduates who had an active contract.

Comparing mobile UVT graduates, the duration of study abroad is not predictive of differences in earnings. In other words, there was no statistically significant difference in the likelihood of having an above-average salary, regardless of whether the mobile student experienced a short-term mobility period (operationalized as 5 months or less), or a long-term Erasmus mobility period (operationalized as more than 5 months). The year in which students participated in Erasmus credit mobility is predictive of earning differences. The year of mobility is negatively associated with earnings, that is, among mobile UVT graduates, the likelihood of having an above-average salary decreases with every academic year. This finding is consistent with the expectation that, over time, as Erasmus mobility became more and more common, it became less valuable in accessing higher-paid positions. Alternatively, the negative relationship could be explained by the fact that students who went on Erasmus earlier have had more time on the labour market, and their longer work experience explains their higher likelihood of having higher-paid positions.

(4) Among those who have an active working contract, does participation in Erasmus mobility predict having a managerial or professional occupation?

Prior research found that “five to ten years after graduation, significantly more Erasmus alumni (64%) than non-mobile alumni (55%) hold a management position. The difference is especially large in Eastern Europe (70% compared to 41%)” (European Commission 2016). Contrary to the expectation of higher occupational prestige, in our analytic sample, Erasmus mobility is not predictive of a more prestigious occupational category. All other things being equal, those who participated in Erasmus mobility during their studies at UVT were not more likely to have an active work contract with a managerial or professional occupation. The results from the propensity score matching model (with the same specification as the logistic regression model discussed above) also suggest that participation in Erasmus mobility is not associated with managerial or professional occupations among UVT graduates with an active work contract. Our result might be explained by the fact that higher education attainment in Romania is the lowest in the European Union and, as such, having a tertiary degree per se is highly valued by employers when they are looking to fill managerial or professional positions. Nevertheless, the mechanisms of obtaining a higher status job may be underspecified in our model. Further research should look into the micro-causalities at play on the local labour market as they might provide an explanation for these findings.

Within the sub-group of mobile UVT graduates, the duration of study abroad is not predictive of occupational category either. In other words, both short- and long-term Erasmus mobility periods are associated with a similar likelihood of holding a managerial or professional job. The year of mobility is not associated with the occupational category either: the likelihood to have a managerial or professional position is the same irrespective of the academic year in which the Erasmus mobility took place.

5 Conclusion: Benefits and Limitations of Register Data

In this chapter, we analysed a unique dataset from Romania to illustrate how register data can be used to answer questions about the benefits of Erasmus credit mobility. We found that the typical UVT graduate who participated in Erasmus mobility and then got employed in Romania after graduation is a woman who has a bachelor degree in social sciences or in humanities, who did not receive a social scholarship but received a merit-based scholarship, who enrolled at one point in time in a master degree program at UVT, and who has a managerial or professional occupation.

In terms of labour market outcomes, our analyses indicate that, all else being equal, participation in Erasmus mobility is (1) significantly negatively associated with insertion in the domestic labour market; (2) significantly positively associated with above-average income among those who do work for an employer operating in Romania, and (3) not predictive of a managerial or professional occupational category.

Working with register data is resource-efficient, but it has its own limitations. Findings presented in this chapter are limited to graduates of a single university in Romania, as a spin-off from an institutional, ITC intensive, tracer study (Proteasa et al. 2018). The limitations of the Romanian employee register are reflected into our findings: as the national registry includes only labour contracts, the results may be less relevant for fields in which self-employment is common, such as law, psychology and even computer sciences. A further limitation of using data from a single national register is that we had no employment information about graduates who were employed outside of Romania. The tracer study used interval measures of graduates’ salaries (Proteasa et al. 2018), which makes the measurement of our earnings data less precise. And while working with data from ReviSal allowed us to observe the labour market outcomes of more UVT graduates than would have been possible with the use of an alumni survey, our analytic dataset still misses information about theoretically important characteristics of the UVT graduates, such as their marital status, number of children, and the educational background of their parents.

Nevertheless, register data shows promising avenues for research and encourages the efficient use of resources by using data that is already collected for administrative purposes both at the national and institutional level. The availability of register data enables researchers to test linkages between higher education and a wide range of individual institutional and societal outcomes. At the very least, register data can provide researchers with good descriptive population parameters from which samples can be drawn for further research. Ministries and higher education institutions should consider these benefits when evaluating requests for data release for research purposes.