The effect of public funding on long-term relocation of experienced researchers: An analysis of the Marie Curie Career Integration Grant Open Access

https://www.webofscience.com/api/gateway/wos/peer-review/10.1162/qss_a_00350

Scientific mobility tends to positively affect individual and institutional scientific productivity and impact (e.g., Fernández-Zubieta, Geuna, & Lawson, 2016; Franzoni, Scellato, & Stephan, 2014; Horta, Veloso, & Grediaga, 2010; Netz, Hampel, & Aman, 2020; Tavares, Sin, & Lança, 2019), by providing access to specific training and research projects (Barjak & Robinson, 2008; Laudel & Bielick, 2019), contributing to knowledge recombination (Scellato, Franzoni, & Stephan, 2017; Sugimoto, Robinson-García et al., 2017), the development of scientific networks, and the transmission of tacit knowledge (e.g., Collins, 2001; Fontes, Videira, & Calapez, 2013; Gibson & McKenzie, 2014; Laudel, 2005; Petersen, 2018; Zubieta, 2009).

For these reasons, policymakers and funding programs have been promoting scientific mobility, especially of early career researchers (e.g., Council of the European Union, 2015; European Commission, 2012a; UNESCO, 2018). More than 20 years ago the European Commission introduced the fundamentals of a European Research Area (ERA) in its political agenda (European Commission, 2000), with the goal to put in place a single, borderless space for research and innovation, and promote the free movement of researchers, knowledge, and innovation within Europe. Indeed, the mobility of scientists in Europe has increased substantially in recent decades, especially at junior level, and nowadays 20–50% of PhD candidates and postdocs are of foreign nationality (Eurostat, 2020). However, mobility predominantly occurs from less attractive and weak systems to more attractive and stronger systems (Ackers, 2005), raising concerns of a so-called “brain drain” (Doria Arrieta, Pammolli, & Petersen, 2017). Some scholars observed that mobility does not necessarily imply a single move from one place to another, but often a “brain circulation” (Johnson & Regets, 1998) whereby researchers move across different destinations, improve their competences, and contribute to network development and knowledge diffusion along the way, including the home country (Coe & Bunnell, 2003). Therefore, specific funding instruments may be required to favor relocation within a certain geographical domain.

The goal of this article is to explore the effects of the Career Integration Grant (CIG), a funding scheme of the Marie Curie Actions under the European Union’s Seventh Framework Programme (FP7), which aimed to reinforce the European Research Area and counter the European brain drain to third countries by supporting the long-term relocation of experienced researchers in EU member states or associated countries. It is important to highlight that “standard” mobility grants target junior scientists and support temporary mobility (e.g., the MSCA individual fellowship, Fulbright programs, and the international mobility grant program of the Swiss National Science Foundation (SNSF)). Hence, the CIG had a sort of monopoly, being the only funding instrument dedicated to supporting the long-term relocation of experienced researchers in the ERA.

We consider the CIG funding scheme, active from 2011 to 2013, and explore its efficacy in increasing the chances of long-term relocation of experienced researchers in the desired destination, namely a specific “host” institution and “host” country, which was the key objective of the grant. We estimate the effect of obtaining a grant on the probabilities of being affiliated to the host institution and host country, 7–8 years after the call, namely 3–4 years after the end of the grant period. We estimate the effect of the grant and conduct second-level analysis to explore whether the grant had a differential effect for specific subgroups of researchers, defined by their traits and the characteristic of the destination. In addition, we investigate whether receiving the grant increased the chances of nontenured academics to obtain a tenured position—7–8 years after the call—and the effect of the grant on scientific productivity—8 years after the call.

We combine three types of data regarding applicants: (a) demographic and proposals characteristics; (b) CV-related information about their career affiliations and positions from 2011 to 2021; and (c) scientific production. To provide evidence of causal relationships we developed a sharp Regression Discontinuity Design (RDD), which is based on the intuition that applicants whose proposal’s grade is in close proximity to the evaluation threshold (just above or just below the funding cutoff) are presumed to be comparable in all significant aspects, except for the funding subsidy.

The results of the analysis reveal that the grant increased the chances of successful relocation to the host institution (+9.4%) and country (+8.2%), and that it was especially effective for subcategories of applicants for which it is commonly more difficult to access alternative sources of funding, namely (a) nontenured, (b) scientists in the soft sciences, (c) non “returnees,” and (d) to nonhigh-ranked institutions.

The article makes three main contributions.

First, it contributes to the literature exploring the efficacy of research grants, and research grants promoting academic mobility more specifically. Previous studies generally found a positive effect on career advancement and on obtaining more grants later (e.g., Bloch, Graversen, & Pedersen, 2014; Bol, de Vaan, & van de Rijt, 2018; Jacob & Lefgren, 2011). Evidence about the impact on scientific productivity is mixed. By employing information on both funded and nonfunded applicants, three studies on the United States did not find any substantial impact on publications nor citations (National Institutes of Health (NIH) grants: Jacob & Lefgren, 2011; Wang, Jones, & Wang, 2019; and the National Science Foundation (NSF): Arora & Gambardella, 2005). The results have been mixed in the European context. Langfeldt, Bloch, and Sivertsen (2015) examined applicants to Danish and Norwegian grant schemes and found a positive effect only on productivity in Norway. Heyard and Hottenrott (2021) and Carayol and Lanoë (2017) found a positive effect on publications and citations of grants awarded respectively by the SNSF and the Agence Nationale de la Recherche (ANR) in France. A recent study employed an RDD to examine the impact of the European Research Council (ERC) scheme and found no effect on research impact and productivity for researchers with a ranking position close to the threshold (Ghirelli, Havari et al., 2023). Moreover, in a review study, Aagaard, Kladakis, and Nielsen (2020) found stronger arguments in favor of increased dispersal of grant funding and against concentration, also because of stagnant or diminishing returns to scale of grant size on research performance.

Very few articles have explored the efficacy and effects of mobility grants. Ackers (2005) analyzed MSCA fellowships funded under the 4th and 5th Framework Programmes (FP4 and FP5). The MSCA programme recognizes and seeks to reconcile the possible tension between the goal of promoting excellence and equality of opportunity at an individual level with balanced growth at regional level. Hence, Ackers examined the implications of MSCA fellowship from two (equality) perspectives: in terms of individual equity and equality of opportunity for researchers as European citizens and in terms of regional equality and balanced growth. From the perspective of individual equity and justice, the scheme played an important role by providing access and opportunity based on individual merit. In terms of regional equality, she found that the distribution and nationality of fellows was “imbalanced,” (i.e., in favor of more attractive countries). However, the imbalance did not necessarily imply a brain drain. In fact, most of the fellows (56%) returned to their home country after 4 years—thus representing an investment from less research-intense regions with little direct cost. Nonreturn may also depend on the lack of opportunities to work in science in the home country, re-entry problems, and struggle to reintegrate, in which the term “brain drain” would be oversimplistic, as the failure to transfer that knowledge cannot be attributed to the migration within the ERA. In addition, MSCA often prevented a “brain drain” to the United States and favored the mobility of fellows to less attractive systems by compensating for comparatively much lower salaries. Baruffaldi, Marino, and Visentin (2020) examined the international mobility grant program of the SNSF, which supported research periods of early career researchers up to a maximum of 36 months at a foreign research institution. They found that the grants increased by 47 percentage points the probability to be abroad in the first year and by 24 percentage points 5 years later, tending to zero in the seventh year. They observed a positive effect on the average impact factor of the published articles, but not on the number of publications nor on obtaining a professorship. Non, van Honk et al. (2022) investigated the effects of an international mobility grant of the Dutch Research Council NWO aimed at early career researchers. They found no significant effect on the probability of leaving academia, the quantity of scientific publications, the citation score, or the number of coauthors. Within the limited literature on the efficacy of mobility grants, which examined grants that promote temporary mobility for professional development of junior researchers (Baruffaldi et al., 2020; Non et al., 2022), our article addresses a specific gap, namely, the effect of a grant supporting long-term relocation of experienced researchers.

Second, the article contributes to the literature on scientific mobility. There are several theoretical perspectives on the drivers of emigration (Cassarino, 2004), such as real income differentials (neoclassical economics: Thorn & Holm-Nielsen, 2006), accumulation of savings and human capital (new economics of labor migration: Dustmann & Weiss, 2007; Stark, 1996), social and institutional factors (structural approach: Cerase, 1974), national identity (transnationalism approach: Portes, 2001), and patterns of interpersonal relationships (social network perspective: Cassarino, 2004). Empirical studies of scientists’ mobility revealed a complex interaction between individual and contextual factors, such as lack of political freedom, economic and social problems in the home country (Thorn & Holm-Nielsen, 2006), institutional location and research intensity (Yan, Zhu, & He, 2020), individual research performance (Abramo, D’Angelo, & Di Costa, 2022; Cattaneo, Malighetti, & Paleari, 2019), demographic traits (El-Ouahi, Robinson-García, & Costas, 2021; Yuret, 2023), and early career mobility experience (Reale, Morettini, & Zinilli, 2019). Scientists are attracted to prestigious universities with a strong educational and research environment (Bratti & Verzillo, 2019; Stephan, Scellato, & Franzoni, 2015), although country characteristics—in term of wealth and strength of the research system—are even more important (Lepori, Seeber, & Bonaccorsi, 2015). However, several studies found that the career progression of mobile foreign researchers tends to be more challenging compared to local and nonmobile ones, and that they are heavily underrepresented among tenured staff compared to junior staff (e.g., Corley & Sabharwal, 2007; Cruz-Castro & Sanz-Menéndez, 2010; Godechot & Louvet, 2008; Lundgren, Claesson Pipping, & Åmossa, 2018; Seeber & Mampaey, 2021; Seeber, Debacker et al., 2023). Scientists who move abroad can also face challenges when trying to come back and make a career in their country of origin (Lawson & Shibayama, 2015; Seeber et al. 2023) and longer stays abroad reduce the likelihood of return (Andújar, Cañibano, & Fernández-Zubieta, 2015; Baruffaldi & Landoni, 2012; Van de Sande, Ackers, & Gill, 2005) as researchers may become “locked in” to the host system, due to stronger personal and professional ties (Casey, Mahroum et al., 2001). Researcher may also get “locked out” if they do not manage to establish scientific prestige or if they lose contacts within the research community in the home country (Casey et al., 2001). The capability to preserve linkages with the home country in fact increases the probability and capability of return (Baruffaldi & Landoni, 2012), as well as public financial support facilitating their return (Andújar et al., 2015).

CIG differs from national initiatives promoting return mobility, such as the Spanish Ramón y Cajal (RyC) program or the Rita Levi Montalcini program in Italy, as it could be used both for return mobility and for relocation in another country in the ERA, and targeted researchers of any nationality. We find evidence that experienced researchers predominantly intended to relocate to their country of origin, but that the grant appears to be mostly effective for the chances of relocation of nonreturnees in a specific institution, compensating for part of the disadvantage experienced by scientists aiming to relocate to a third country.

Finally, the article has policy implications, particularly in terms of the design of grant funding instruments. Considering the large size of resources invested via grant funding, the assessment of their efficacy and the exploration of how to improve their functioning should be a priority. The result of the analysis provides useful insights into what categories benefitted from a similar grant, which can be used for its recalibration or the design of future instruments. Namely, the grant has increased the chances of relocation for subcategories of applicants for which it is commonly more difficult to access alternative sources of funding. This suggests the opportunity to either fund primarily subcategories of applicants that are more in need of financial support, such as nontenured researchers, or increase the sum of the grant for the other categories’ applicants to provide a meaningful contribution.

In the following section we present the data, variables, and methods, and in Section 3 we present the analysis. In Section 4 we discuss the results and their implications, and directions for future research.

The overall objective of the Marie Skłodowska-Curie Actions (MSCA) under the Seventh Framework Programme (FP7) was to make Europe more attractive to researchers, at all stages of their career, by enhancing the training, skills, mobility, and career development of researchers within the ERA and its international dimension. They aimed at strengthening the research human potential and capital at European level by stimulating people to pursue research careers and encouraging top researchers to stay in Europe or attracting those willing to come to Europe. Geographical mobility is a key feature of the MSCA programme, and so are international cooperation, intersectoral collaboration, and the transfer of knowledge between the academic and nonacademic sectors. To that purpose, the MSCA had a set of funding schemes (actions) addressing different priorities. One of those actions was the Career Integration Grant (European Commission, 2011, 2012b, 2013 ¹).

Scientific mobility schemes commonly target early career researchers and temporary mobility (e.g., Horta, Birolini et al., 2021). We explore the efficacy of a European mobility scheme focused on experienced researchers and supporting long-term relocation.

Under the EU’s FP7², the MSCA programme had in place a dedicated funding scheme addressing the return and reintegration of researchers in Europe to counteract a potential loss of talent (so-called “European brain drain”). This scheme targeted experienced researchers (usually at postdoctoral level) willing to relocate in a European country. The funding scheme consisted initially of two subschemes, one called European Reintegration Grants, reserved exclusively to recipients of Marie Curie fellowships, and another one called International Reintegration Grants, which was open to any experienced researcher working outside Europe and willing to relocate in an EU member state or associated country.

In 2011, these two funding schemes were merged into the Career Integration Grants (CIG) for the remaining 3 years of FP7. CIG aimed to improve the prospects for long-term integration of researchers to EU member states and associated countries after a mobility period. They were directed at experienced researchers³ of any nationality, from any country and research field. To be eligible, applicant researchers could not have resided or carried out their main activity in the country where they intended to relocate (host country) for more than 12 months in the 3 years immediately prior to the reference deadline⁴.

The research proposal was jointly submitted by the applicant researcher and the host organization where the project should be executed in case of being awarded the funding. The applicant researchers were free to choose their research topics, with no predefined priority in terms of field or domain of research. For evaluation purposes, though, applicants were requested to choose one among the eight MSCA scientific panels (see Section 2.3). The host organization had to provide evidence supporting the applicant researcher and to ensure their effective and lasting professional integration for a period of at least the same duration of the project, namely an employment contract, temporary or tenured (European Commission, 2012a). Successful applicants received a grant of €25,000 per year for a maximum of 4 years, which consisted of a flat-rate contribution to support the research costs of the researcher at the host organization (e.g., staff employed for the project, travel costs, overheads, management costs).

As for other Marie Curie Actions, the evaluation of proposals consisted of two distinct and consecutive steps. The first one consisted of the evaluation of each proposal by three independent expert reviewers. In this phase, experts produced an Individual Evaluation Report (IER), with comments and scores with respect to their independent assessments. Proposals were evaluated against four criteria: scientific and technological (S&T) quality of the proposal; the researcher quality (i.e., the Curriculum Vitae, CV); the quality of the implementation; and the impact of the proposal. Each criterion was scored from zero to five, five being the maximum score. The total score of the evaluation consisted of a weighted score, converted to a scale from 0 to 100. In a second step, the three expert reviewers who evaluated a given proposal met over a consensus meeting, where they discussed the proposal and reached a consensual outcome, resulting in a Consensus Report (CR), which was not necessarily the average, in terms of scores and comments, of the three reports previously produced by the reviewers. The Consensus Report was converted into an Evaluation Summary Report (ESR) to be sent to applicants with the results of the evaluation. Depending on the funding available each year, proposals were ranked into a main list (those offered the funding), a reserve list, and a list of rejected proposals due to lack of funding or for being scored below the thresholds of quality.

We combine information from different sources. The initial data set for CIG included 4,325 distinct applicants, considering solely the most recent application in cases of multiple submissions. This data set offers comprehensive information, including details such as the applicant’s nationality, desired affiliation, residence at the time of application, scientific panel, evaluation outcome, proposal score, and grant start date. In the second step of data collection and consolidation, we refined the data set to 2,939 researchers, ensuring complete information on their research performance, affiliations, and job positions. For research performance, we extracted data from Scopus⁵, encompassing the following information (as of September 2022, the date of extraction):

1. date of first publication and date of last publication;

2. number of publications⁶ until the year of the call;

3. number of publications until the year of the call, normalized by number of coauthors; and

4. number of publications with coauthors from the host country until the year of the call⁷.

We describe the dependent, independent and control variables.

As a proxy of long-term relocation, we considered whether the researcher was affiliated to the institution of desired relocation (so-called “host institution”) 7–8 years after the year of the call to which they applied. In other words, we controlled whether in year 7 and/or year 8 after the year of the call the researcher was affiliated to the host institution. For those who were funded, this meant 3–4 years after the conclusion of the grant⁹.

We also considered whether the applicant was affiliated to any institution of the desired country of relocation 7–8 years after the year of the call, namely the host institution or another institution in the same country.

We tested whether obtaining the grant increased the chances of nontenured applicants (junior) achieving a tenured position 7–8 years after the call.

Grant funded: yes or no (binary).

We measure research productivity at the year of the call (normalized by the number of coauthors and weighted by Journal SJR) as a factor that can affect the chances of long-term relocation, career progression and research productivity at year 8. In academia, research productivity is in fact a major factor affecting careers progression and achievement of stable positions (e.g., Ginther & Hayes, 2003; Hesli, Lee, & Mitchell, 2012; Lutter & Schröder, 2016; Tien, 2007; Weisshaar, 2017).

Scientific connections, networks, and personal ties contribute to securing academic positions (e.g., Baruffaldi & Landoni, 2012; Jungbauer-Gans & Gross, 2013; Lutter & Schröder, 2016) and support for those intending to return (Ackers, 2008; Morano-Foadi, 2005). Recognizing the importance of scientific collaborations, we considered the standardized number of publications with at least one coauthor affiliated to an institution in the host country, until the year of the call.

This is a binary variable that measures whether the applicant’s nationality was the same as the host country or not. Several arguments suggest that researchers are more likely to relocate successfully in their country of nationality: Nationals are more likely to have personal ties with their country of origin and they are more likely to know the local language, which is beneficial for academic duties and particularly teaching (e.g., Pudelko & Tenzer, 2019; Śliwa & Johansson, 2014)¹⁰.

Researchers’ nationality of a country outside the European Research Area.

The quality of the cultural, social, and economic environment in which an institution is located influences the desirability to move and to stay in such location. We used data from the OECD better life index (www.oecdbetterlifeindex.org, year 2019–22), which includes 10 indicators of well-being at regional level (mostly Nomenclature of Territorial Units for Statistics (NUTS) level 2). A factor analysis identified two factors. As a proxy of the quality of life in the region of the host institution we employ the first factor, which is highly correlated with the indicators measuring quality of education, job opportunities; income, environment, accessibility to service, housing, and community.

The country of destination may affect the chances of long-term relocation. We regrouped countries of destination into two groups. One group identifies highly attractive countries in central and northern Europe that are net receivers of mobility flows and one group corresponds to countries in southern and eastern Europe that are net senders of researchers (see, e.g., Franzoni, Scellato, & Stephan, 2015; Lepori et al., 2015; Macháček, Srholec et al., 2022; Robinson-Garcia, Sugimoto et al., 2019). We constructed a dummy variable that equals 1 for the first group.

The added value of the grant may be greater in low ranked institutions, for which access to alternative sources is typically more difficult. As a proxy of the prestige of the hiring institution we employed data from the SCImago Institutions Rankings (SIR) (year 2018)¹¹. We created a dummy variable by dividing the sample of institutions below the median (1 = low ranked) and above the median (0 = high ranked).

We included dummy variables for the eight disciplinary panels to which applicants could submit their proposals. Moreover, to control whether the grant was more effective in given scientific fields, we regrouped the eight disciplinary panels to which they could submit their proposal in two main groups: Hard Sciences: CHE—Chemistry, ENV—Environmental and Geosciences, LIFE—Life Sciences, MAT—Mathematics, and PHY—Physics; and Soft Sciences: ECO—Economic Sciences, ENG—Information Science and Engineering, and SOC—Social Sciences and Humanities.

Age of the applicant at the year of the call, standardized¹².

Gender of the applicant. Male, Female.

Unobserved variables, like the level of selectivity and the quality in the pool of applicants, may have changed from year to year, so it is important to control for that.

To examine the efficacy of the CIG on the primary policy goal, which is the long-term relocation of applicants at a country and institutional level, as well as on two secondary outcomes, attainment of professorship and research productivity, we implemented a sharp Regression Discontinuity Design (RDD). RDD is often seen as the sharpest tool of causal inference as it approximates very closely the ideal setting of randomized control experiments (see Lee & Lemieux, 2010, p. 282).

Based on the consideration that the simple comparison of the sample of granted and not-granted applicants might lead to biased estimates of the grant efficacy, as the two groups are rather different in terms of baseline characteristics, we leverage the cutoff-based nature of the policy and create a quasi-experimental setup where individuals are assigned to treatment in a way that resembles random assignment (Lee, 2008). In detail, we compare the outcomes of applicants that receive scores close to the threshold at the time of the evaluation. The intuition of this is that around the cutoff of the assignment of the Marie Curie grant, applicants are expected to be similar in all relevant aspects apart from the treatment status, thus allowing us to draw causal inference on the effect of the funding scheme without the need to setup a randomized control trial.

To validate our analytical RDD framework, we first investigate the presence of a clear discontinuity in the treatment variable, which is suggestive that applicants have no direct control over the treatment variable and there is not a suspect concentration of applicants just below the cutoff (Imbens & Lemieux, 2008). Figure 1 illustrates the distribution of the scores of proposals. Second, we investigate whether each covariate is significantly affected by an apparent discontinuity around the cutoff, which could invalidate the assessment due to dissimilarities among applicants (Calonico, Cattaneo et al., 2019; Lee, 2008). Table S1 in the Supplementary material reports RDD regression on each covariate, showing that applicants are comparable above and below the cutoff based on observable characteristics.

Our econometric specification is based on standard logit models when the outcome variable is binary, such as relocation at a country and the institutional level or the attainment of a tenure position. Meanwhile, we employ an OLS model for evaluating the long-term research performance of the applicants.

The average age of the applicants was 36 years (standard deviation 5.2). They were mostly male (62%) and with a diverse disciplinary background, namely Life Sciences (36%), Engineering (15%), Social Sciences (13%), Physics (12%), Chemistry (8%), Environmental Sciences (8%), Mathematics (4%), and Economics (4%). The research productivity of the applicants was of a good level, with an average of 10.1 normalized publications. As a matter of fact, the grant did not cover the salary, so it was unlikely that the host institution would endorse a poor candidate. Moreover, the grantees were significantly more productive than the nongrantees: 15.0 vs. 7.9 normalized publications (after controlling for discipline and age: average of +6.653*** normalized publications—see Table S2 in the Supplementary material). A considerable proportion of the applicants aimed to relocate to a low-ranked institution (39.1%), and—interestingly—these latter applicants displayed a significantly higher scientific productivity, on average +3.005*** normalized publications (Table S3 in the Supplementary material).

To understand which researchers aimed to use the CIG and for what purpose, we describe the characteristics of the applicants. In most cases, the applicant was a European or associated country national, aimed to relocate in their country of nationality (58.7%) or in a third country (28.4%), whereas 669 applications (12.8%) were from nationals of non-EU or associated countries¹³. In turn, in the intention of the applicants, the instrument was intended to support the return to their home country. This was especially the case for Israeli (97.7%), Turkish (95.4%), Spanish (77.5%), and French (73.8%) researchers, whereas only a minority of Italians (44.7%) and Germans (41.7%) applied to relocate in their country of nationality.

Most applications were from nationals of countries that are net exporters of researchers, like Spain (12.0%), Italy (10.0%), Israel (9.3%), and Turkey (9.2%), which together accounted for more than 40% of the applications, but also France (6.7%), Germany (6.9%), and the United Kingdom (4.4%). At the same time, a considerable number of applications (12.8%) came from outside the ERA (i.e., non-EU or associated countries).

The United Kingdom and Spain were the most popular host countries, with the first attracting mostly applications from nonnationals. More precisely, the most popular host countries were the United Kingdom (14.4%), Spain (13.5%), Israel (9.7%), Turkey (9.4%), France (8.1%), Italy (5.9%), and Germany (5.8%); 80.6% the applications to the United Kingdom were from nonnationals, compared to 51% for Germany, 38.6% for France, 30.9% for Spain, 23.9 % for Italy, 7.1% for Israel, and 6.3% for Turkey. Hence, relocation intentions to a third country reflect the usual patterns towards attractive countries.

In sum, applicants applied to the CIG to address various needs: primarily, returnees to countries which are large exporter of researchers, but also mobility to a third country as well as attraction of talent from countries outside Europe, with both European and non-European nationality.

Figure 2 illustrates the applicants’ nationality and host country for countries with at least 100 applications.

Prior to presenting the econometric findings, we provide an overview of the observable characteristics of the estimated sample (Table 1 and Figure 3) and conduct a visual analysis of the outcome variables in relation to the grant proposal scores. Figure 4 depicts the mean probability of applicants relocating at both the country and institutional levels. Bandwidth values for the host and institution reallocation are 4.3 and 5.0, respectively. The linear (red lines) and quadratic (green lines) interpolation lines serve as visual indicators of the presence of discontinuities, in both cases. As indicated by the visual representations, the interpolation lines suggest the absence of quadratic relationships¹⁴.

Table 2 reports the econometric results for our outcome variables, presenting the estimated marginal effects. Specifically, we observe a significant and positive effect of the grant on the relocation of applicants at both the country and institutional levels. Namely, receiving the grant increased the probability of being reallocated after 7–8 years in the host country by 8.2% (p-value: 0.003) and in the host institution by +9.4% (p-value: 0.076). As a robustness test, we further elaborate our approach by employing narrower bandwidths, which effectively reduce the number of applicants under consideration for testing the relationship. This approach allows us to specifically examine applicants in closer proximity to the threshold. Specifically, we reduce the selected bandwidth by 15% for both estimates. The effect of CIG funding on long-term host relocation is found to be 7.3% (p-value: 0.001), while the effect on long-term institution relocation stands at 9.4% (p-value: 0.090).

Control variables indicate that, in general, relocation is more common when the applicant’s nationality is from an ERA country; when the nationality is the same as the host country; when they have a tenured position; and with an increasing number of established scientific collaborations with the host country. Based on the premise that we are operating within an RDD framework, applicants’ gender, age at the call, research performance, and the attractiveness of the host country do not influence the chances of long-term relocation.

Average effects can mask heterogeneous effects for specific subgroups. Hence, we examine whether the funding scheme particularly benefits certain categories of applicants. We extend our analysis by dividing the population of researchers into subgroups based on their individual characteristics and type of destination. Evidence about different effects across subgroups can be informative for a better understanding of the mechanisms that underpin the functioning of the instrument as well as for an insightful recalibration of the instrument itself.

First, we divide the population of applicants based on individual traits that significantly affect the chances of long-term relocation, namely, whether they hold a tenured position and belong to hard sciences or soft sciences disciplines. Furthermore, building on the primary focus of this funding instrument, which is international mobility, we delve deeper into its effectiveness concerning applicants’ choices to move towards specific geographical macro-areas; their country of nationality or a third country; and high- or low-ranked institutions.

The findings presented in Table 3(a) and 3(b) underscore that the overall effect of CIG funding varies among different typologies of applicants. Specifically, obtaining the grant significantly increased the chances of nontenured applicants of long-term relocation in a host institution (+7.8%) and country (+9.1%). It proved effective for scientists operating in the soft sciences, increasing the probability of long-term relocation in the host institution by 9.8% and host country by 14.9%. We also distinguished applicants aiming to return to their country of nationality (i.e., “returnees”) from those moving to a third country (i.e., “nonreturnees”). We found that the grant increased the chances of relocation for nonreturnees at the institutional level (+16.9%). In other words, the grant compensates for part of the disadvantage of relocation to a third country. Furthermore, grantees do not have higher chances of relocation than nongrantees, when directed to northern European countries or South-East European countries. Finally, we explored whether the added value of the grant for those moving towards prestigious or nonprestigious institutions. We found that it significantly increased the chances of relocation for applicants directed toward nonhigh-ranked institutions (+14.1%).

In summary, the grant was particularly effective for categories of applicants for which it is commonly more difficult to access alternative sources of funding, namely nontenured soft scientists moving to a third country and to nonhigh-ranked institution.

We also explore whether obtaining the grant influenced two additional outcome variables, beyond long-term relocation. Specifically, we examine whether it increased the likelihood of junior staff obtaining a tenured position, 7–8 years after the call, and if it affected the long-term research performance of the applicant, measured 8 years after the call. Bandwidth values for long-run professorship and research productivity are 4.85 and 4.95, respectively.

Table 4 presents the results of the regressions. The sample also includes scientists that did not relocate, so the variables related to the characteristics of the desired relocation country, region, and collaborations would not be meaningful and have not been included as controls in the regression. In both instances, the results do not reveal any effect of obtaining a grant, and that other factors play a more significant role in influencing the career progression and productivity of researchers. As a matter of fact, personal and family considerations are often very important drivers of return mobility decisions (e.g., Czeranowska, Parutis, & Trąbka, 2023). This may explain why the grant is not related to greater career progression or productivity. In any case, most of the nontenured applicants did reach a tenured position after 7–8 years, namely 54.7% of the grantees compared to 53.9% of the nongrantees.

This is the first quantitative study to explore the efficacy of a mobility program aimed at tackling unbalanced flows from less to more attractive countries, by supporting the long-term relocation of experienced researchers. We analyzed the CIG, a funding scheme from the Marie Curie Actions under FP7. CIG aimed at improving the prospects for long-term integration of experienced researchers to EU member states and associated countries and addressing the European brain drain. We employed data on applicants to the three calls of CIG, in 2011, 2012, and 2013, from a Scopus data set and CV information, to reconstruct their career track, and to explore whether receiving CIG increased the chances of long-term relocation in the desired institution and/or country of destination, and whether the grant had a different effect for different subgroups of researchers. We also explored whether the grant increased the chances for obtaining a tenure position and the effect on scientific productivity.

We employed a regression discontinuity design and found that obtaining a CIG increased the chances of successful relocation to the host institution (+9.4%) and host country (+8.2%). The grant was particularly effective for some categories of researchers, namely for nontenured researchers, social scientists, nonreturnees, and directed to nonhigh-ranked institutions. In turn, it appears that the grant was particularly valuable for categories of scientists that typically have less access to funding, thus implying a comparative higher value of the grant. This evidence can be valuable for the design of similar funding instruments in the future. Namely, either fund primarily subcategories of applicants that are more in need of financial support or to increase funds for the other categories of applicants to provide a meaningful contribution.

Some limitations should be mentioned. First, we observed a significant, but relatively moderate effect of obtaining a CIG on the chances of long-term relocation. However, it is important to remark that the CIG might have provided leverage for scientists to establish contact with a host institution and to develop a proposal together. Such a relationship and commitment from the host organization arguably increased the chances of relocation, even for unsuccessful applicants. Hence, part of CIG’s efficacy may be unobserved, as not solely related to obtaining the grant. Another possible contributing factor to the observed moderate effect is the well-known crowding out effect, namely the possibility that unsuccessful applicants found alternative sources of support. Also, the fact that the grant was not very effective for the relocation and productivity of scientists returning to their country of nationality may be because personal motivations play a major role for return mobility, and funding support and scientific goals are comparatively less important. A final limitation is also that we did not have the possibility to control for important variables that often affect the decision to relocate, such as civil status and number of children.

It is interesting to compare CIG with the other individual mobility scheme of the Marie Curie Actions: the individual fellowships. The CIG objective was to encourage experienced researchers to establish themselves in Europe, whereas the individual fellowships aim to support researchers to acquire new skills and training or intersectoral experiences by allowing them to move between EU member states or associated countries (in the case of the Intra-European Fellowships); bring to Europe top-class researchers from third countries (i.e., outside Europe) (for the International Incoming Fellowships); or enable European researchers to be trained and gain experience in a third country (mostly the United States) before returning to Europe (in the case of the International Outgoing Fellowships). Data available on the Horizon Dashboard and programmatic evaluation reports (European Commission, 2017, 2018) show that more than half of the Individual Fellowship (IF) applicants and fellows were nationals of Spain, Italy, France, and Germany, and more than 50% of them submitted proposals to perform their work in the United Kingdom, France, or Germany. Nearly two-thirds of all the fellows (with a majority coming from India, China, the United States, Russia, Australia, and Canada) were hosted in the United Kingdom, France, Germany, or Switzerland. In contrast, the CIG attracted large numbers of researchers attempting to return to their home countries after an experience abroad. Hence, CIG was partly complementary in purpose and usage to the other funding schemes under the Marie Curie Actions.

Finally, it is worth reflecting on the history of the CIG and the potential for similar programs in the future. Every 7 years, the EU launches a new Framework Programme (FP) for Research and Innovation, with a revised budget and structure. Mobility programs for researchers have been part of these Framework Programmes from the beginning, though the “Marie Curie” branding was only introduced in 1996. With each new FP, subprograms are typically reviewed and updated, leading to new funding schemes. A call for simplification during the FP7 phase led to a reduction in the number of funding schemes within the MSCA under the subsequent FP, Horizon 2020. No official document explains the decision to discontinue the CIG, but two reasons seem plausible. First, smaller programs are often easier to cut, and CIG was relatively smaller than other initiatives. Second, although CIG specifically targeted long-term relocation for experienced researchers, IFs could, in principle, also support the long-term mobility of senior researchers, despite their primary focus on temporary mobility for junior researchers. Our study indicates, however, that CIG applicants had a different profile than those applying for IFs. CIG applicants tended to be more experienced, often holding tenured positions, and were more likely to move to less popular host countries and institutions. This highlights the importance of maintaining distinct mobility schemes if the goal is to diversify the range of researchers benefiting from EU funding.

We are grateful to the reviewers for their insightful comments and to the Research Executive Agency for providing access to the data used in this study.

Marco Seeber: Conceptualization, Data curation, Formal analysis, Investigation, Supervision, Writing—original draft, Writing—review & editing. Mattia Cattaneo: Data curation, Formal analysis, Investigation, Methodology, Visualization, Writing—original draft, Writing—review & editing. Carlo Debernardi: Data curation, Formal analysis, Investigation, Visualization, Writing—review & editing. David G. Pina: Conceptualization, Resources, Writing—review & editing.

DGP is employed at the European Research Executive Agency, the EU funding agency that was responsible for the management of the FP7 CIG programme.

No funding has been received for this work.

The data are strictly confidential. The anonymized data set is available upon request.