Talent hat, cross-border mobility, and career development in China

https://www.webofscience.com/api/gateway/wos/peer-review/10.1162/qss_a_00312

Scientific career movement is fundamental for scientific advances, as it not only accelerates the circulation rate of knowledge across institutions and national borders but also enriches the scientific curriculum for researchers (Deville, Wang et al., 2014; Petersen, 2018; Trippl, 2013; Verginer & Riccaboni, 2021). Studies show that nations with scientific openness to international mobility and collaboration are linked to stronger research (Sugimoto, Robinson-García et al., 2017; Wagner & Jonkers, 2017; Wagner & Leydesdorff, 2005). In recent decades, many countries have taken action to increase cross-border talent recruitment. For instance, the Young Thousand Talents program in China (Yang & Marini, 2019) cultivates and recruits potential future rising star scientists overseas, and it has attracted more than 3,000 young talents worldwide. The elected scholars (with a “talent hat”) will get a big bonus, startup funding from the government, and an extra bonus, funding, and PhD student quota from the employer university. Besides China, many countries have developed specific visas for attracting foreign talent, for instance, the “EB-1” in the United States and the “Researcher Visa” in Germany.

However, scientists often experience dilemmas in career movement, especially cross-border mobility (Petersen, 2018; Xu, Braun Střelcová et al., 2022): On one hand, career mobility will bring new research opportunities in new environments, collaborators, and maybe a startup funding for rebuilding the lab, which could benefit scientists greatly; on the other hand, mobility may cause short-term research discontinuity and the loss of resources from the previous institution, collaborators, and funding agencies. In China, the organizational structure often involves a “big team,” (i.e., one super PI leads a team of multiple investigators). This creates a unique dynamic for younger researchers, particularly recent PhD graduates or postdocs, who must transition to independence and relocate to establish their scientific careers. Consequently, they face the intricate challenge of balancing family, research, and relocation, which remains a formidable task (Azoulay, Ganguli, & Zivin, 2017; Liénard, Achakulvisut et al., 2018; Ma, Mukherjee, & Uzzi, 2020; Malmgren, Ottino, & Nunes Amaral, 2010).

This prompts us to ask two questions: How do scientists benefit from international talent recruitment, and how do they maintain a stable and continuable scientific career after cross-border mobility? In recent decades, there have been increasing discussions about talent, impact, employment, and mobility, including the statistical modeling of mobility patterns, career development, and policy implications (Cao, Baas et al., 2020; Clauset, Arbesman, & Larremore, 2015; Deville et al., 2014; Netz, Hampel, & Aman, 2020; Petersen, 2018; Vásárhelyi, Zakhlebin et al., 2021; Way, Morgan et al., 2017; Zhang, Deng et al., 2019; Zweig, Siqin, & Huiyao, 2020), etc. Shi, Liu, and Wang (2023) found that the Young Thousand Talents Program in China (talent with cross-border movement) is successful in attracting people of high academic caliber but not the best compared with those who get the talent hats without returning to China. Zhao, Wei, and Li (2023) and Cao et al. (2020) found that the returnees with cross-border mobility did not show advantages on scientific productivity (Zhao et al., 2023) but were more likely to publish higher impact works and more internationally active than domestic counterparts (Cao et al., 2020). Zweig et al. (2020) found that domestic policies are crucial to attracting abroad scholars. Further, the career development of moved scientists is controversial. Zweig, Changgui, and Rosen (2004) found that scholar with overseas PhDs benefits more in terms of people’s perceptions and technology transfer and produce more disruptive work (Zhao, Li et al., 2019), while Li and Tang (2019) found that Chinese international returnees are pessimistic about career promotion.

To quantitatively clarify the effects of cross-border movement and talent hats on future scientific performance, we manually collected 10 years of the scholars who were enrolled in the Young Thousand Talents program in China and manually matched their publication, citation, and collaboration records via a large-scale scientific corpus. The challenge to conduct this research is how to differentiate the indigenous factors that influence the future status of researchers other than mobility. For example, prior scientific impact will predict future performance for scientists. It has been an intractable problem for a long time to obtain persuasive results and unbiased estimation in this comparison because the treated group is incomparable with the control group with regard to group size and prior academic attributes. In recent decades, matching designs have been used in observational data to reduce the confounding influence in science of science and bibliometrics, including topics in gender disparities, citation prediction, mobility, prizewinning, team performance, field growth, collaborations, etc. (Azoulay et al., 2017; Huang, Tian, & Ma, 2023; Jin, Ma, & Uzzi, 2021; Reschke, Azoulay, & Stuart, 2018; Shi et al., 2023; Zhu, Jin et al., 2023) and network tools have been developed to study scientific careers and science (Fortunato, Bergstrom et al., 2018; Wang, Song, & Barabási, 2013; Way, Morgan et al., 2019; Yang, Chawla, & Uzzi, 2019; Zeng, Shen et al., 2017). In general, the matching techniques are used to find observational “twins” before the treatment and estimate the treatment effects by comparing the posterior difference between them.

In this work, we investigated about 2.6 million scholars with more than 10 publication records and their 65.1 million papers from 2000 to 2021 using OpenAlex (Priem, Piwowar, & Orr, 2022) and manually curated 1,563 Young Thousand Talents and their publication records, which enable us to track the future scientific performance (Petersen, Riccaboni et al., 2012; Sinatra, Wang et al., 2016) for each scholar. We constructed a comparison on three groups of scholars:

1. Returnees-Enrolled: Cross-border scientists with a talent hat (i.e., scientists who moved to China and enrolled in the Young Thousand Talents program who received large pay, funding, and other resources).

2. Returnees-Nonenrolled: Cross-border scientists without a talent hat (i.e., scientists who moved to China who did not enroll in the talent program (see Figure S3 in the Supplementary material).

3. Nonreturnees: Noncross-border scientists (i.e., scientists who have not had cross-border movements (see Figure S4 in the Supplementary material).

Our work innovatively introduces multidimensional scientific performance measures to analyze the academic trajectories of scientists affected by cross-border mobility or talent enrollments. In addition to traditional metrics, such as article numbers and citation count, we have incorporated evaluative measures from journal-centric and collaborator-centric perspectives. These encompass the annual count of papers published by authors in Journal Citation Reports (JCR) Q1 journals and the number of collaborators per year. For awardees of the Youth Thousand Talents Program, we utilized the most comprehensive data set, comprising 1,563 valid pieces of information across seven cohorts of awardees.

We found the advantages of cross-border talents in future scientific career success by disciplines and years. However, not everyone benefits the same from the cross-border talent program. Indeed, some of them thrived quickly and developed into top scientists among their peers, yet some did not. We further explored the potential environmental changes of the talents and the matched contenders and found that scientists who reassemble their collaboration network with new collaborators in new institutions after cross-border movement significantly improve their academic performance, while changing research directions after mobility may have risk their citation gain, which raises strong policy implications on cultivating cross-border talents.

In 2011, China’s central government announced the Young Thousand Talents program to attract younger rising star scientists overseas to China. The chosen scholars (conferred with a talent hat/title) will be given a substantial bonus and startup funding from the government, and the extra bonus, funding, and Ph.D. student quota from the employer university, which is a significantly higher compensation package than the for scholars who are not chosen. We manually compiled data on 1,563 talents who successfully enrolled in the program and returned to China across seven cohorts from 2011 to 2013 and 2015 to 2018 (refer to Section 1 in the Supplementary material for more details).

To quantify the cross-border mobility and the talent hat effects, we compared the Returnees-Enrolled with scientists who moved to China without a talent hat (Returnees-Nonenrolled) and scientists who did not conduct cross-border movements (Nonreturnees). Specifically, as shown in Figure 1A, we performed a two-step matching procedure to identify undistinguishable control groups from the millions of scientists in the database. At first, in the “exact” step, we conducted a 1: N match (i.e., for each cross-border talent scientist, we matched at most 300 unmoved scientists and 200 moved (to China) scientists who have the same discipline, close research career starting year, and comparable total number of publications and citations with the talent scientist).

Next, we integrated four metrics—yearly publications, citations, Q1-publications, and collaborators—to represent scientists’ academic performance. In the “refining” step, we further aligned these metrics for identical scientists to ensure similar career development trajectories before the talent recruitment year (refer to Figure S5 and Tables S1 and S2 in the Supplementary material). To accomplish this, we utilized established matching techniques, particularly the synthetic control method (SCM) detailed in Section 5, which assigns weights to each contender from the first step to better align the academic performance curves of talent scientists while accounting for time-dependent confounding effects.

This process yielded 556, 581, 575, and 282 matched pairs on publications, citations, Q1-publications, and collaborators, respectively, for the Returnees-Enrolled and the Returnees-Nonenrolled. Additionally, we obtained 1,240, 1,435, 1,361, and 476 matched pairs of scientists on these four metrics, respectively, for the Returnees-Enrolled and the Nonreturnees. The fewer matching pairs for the Returnees-Enrolled and the Returnees-Nonenrolled suggest that talents selected by the program are top-tier scientists, posing challenges in finding matches among scientists who missed the program. We also conducted validation using the coarsened exact matching and dynamic optimal matching methods (refer to Section S2.2 in the Supplementary material for detailed procedures).

Figures 1B–1E present the annual counts of publications, citations, Q1-publications, and collaborators for both Returnees-Enrolled and Returnees-Nonenrolled scientists, respectively. Using the SCM method, these groups exhibit nearly identical trends before their movement, as indicated in the grey areas. However, after moving, the Returnees-Nonenrolled show significantly higher numbers of publications and Q1-publications compared to those without talent hats. Additionally, Returnees-Enrolled exhibit slightly higher citation impact and more collaborators than Returnees-Nonenrolled. Similarly, Figures 1F–1I compare Returnees-Enrolled with Nonreturnees, showing that Returnees-Enrolled exhibit significantly higher performance in these four metrics following mobility.

For quantitative analysis, we employed DID regression models to delve deeper into the impacts of mobility and the talent program, controlling for individual variations and time cohorts. As outlined in Table 1, Returnees-Enrolled scientists, on average, published approximately one more paper per year, including one more Q1 paper after moving, and received about 18% more citations annually compared to scientists who moved to China without the talent hat (Returnees-Nonenrolled). The coefficients of the cross-term were positive but insignificant between the Returnees-Enrolled and the Returnees-Nonenrolled, indicating that the talent program brings more but limited collaborators for the talents compared to other returnees not enrolled in the talent program. Furthermore, Returnees-Enrolled outperformed the Nonreturnees, showcasing an average of two more papers, 1.5 more Q1 papers per year, and about 33% more citations. The result also shows a notable increase in collaborators for scientists moving to China with talent hats compared to those who did not undertake mobility. This aligns with the fact that relocation typically brings more collaborative opportunities.

Studies show that incentives in the early career stage will improve the scientist’s future performance (Bol, de Vaan, & van de Rijt, 2018; Zhu et al., 2023). Analogously, based on the DID models, we evaluated the yearly differences between the talented scientists and their contenders, as shown in Figure 2. The talented scientists displayed a growing trend in scientific performance after relocating, compared with their contender groups. Following their relocations, scientists might focus on settling down and restructuring their research teams, such as recruiting students, research assistants, or procuring equipment; thus, their scientific performance initially remained comparable. However, the discrepancy grew progressively within the first 5–6 years post-receiving the talent hat. On average, Returnees-Enrolled scientists published approximately three more papers than Returnees-Nonenrolled scientists in the 6th year after arriving in China. Furthermore, mobility to China appears to offer scientists more resources, with the talent program providing adequate funding for recruiting research assistants and a conducive academic environment, which also serves as a nurturing ground for scientists to produce high-quality work. Indeed, the talent program significantly stimulated scientists’ ability to produce high-quality work, measured by Q1-publications, as demonstrated in Figures 2C and 2G. The rising coefficients of Q1-publications in the comparative trials aligned with the trends observed in publications (Figures 2A and 2E). Similarly, we observed that the Returnees-Enrolled scientists benefit from limited advantage from the talent program in the number of collaborators compared to the Returnees-Nonenrolled (Figure 2D), but gain a greater benefit from international mobility compared to the Nonreturnees (Figure 2H).

For younger scientists, startup funding is important to build their labs, including recruitment of PhD students, purchase of practical equipment and materials, technical and environmental maintenance, etc. The talent program gives sufficient resource support, and the chosen talents will be given sufficient startup funding from the government and extra funding and student quota from the employer university, which is important for experimental or resource-intensive research, such as in materials science and computer science, which need substantial funding and labor support. Indeed, in Figure 3A–D, when we compared the talents with their contenders by disciplines, we found that talent in the resource-intensive disciplines benefited more from the program. In general, in materials and computer science, the talents benefit more in four aspects of academic performance than both their Nonreturnees and Returnees-Nonenrolled, and in materials science, environmental science, computer science, chemistry, and biology, the talents benefited significantly more than their Nonreturnees. Conversely, the gaps in academic performance in four aspects are smaller in mathematics (Figure S6 in the Supplementary material).

When we checked the talent benefits by different year cohorts, in Figure 3E–H, we saw a decrease in the talent benefits compared to the two groups of contenders. Two potential reasons account for this result. On the one hand, in recent years, with the number of returnees increasing, many Chinese universities have also launched local parallel talent programs and offered enticing compensation packages to attract younger scholars studying abroad who do not enroll in the talent program. Therefore, the privileges of talents are diluting over time. On the other hand, the benefit gaps between the talents and the contenders are increasing with time based on Figure 2, so we do not have enough observational years for scientists in recent cohorts; for instance, for scientists in the 2018 cohorts, we have only witnessed less than 4 years (as of 2021 for our database) after they moved to China (Figure S7 in the Supplementary material).

Investigating the performance of scientists following mobility experiences requires a nuanced understanding of the underlying mechanisms at play. Mobility introduces scientists to new scientific environments, characterized by diverse collaborators, institutions, and research directions. This exposure triggers a process of adaptation and exploration, where scientists may leverage their newfound connections to pursue novel research avenues. The dynamics of collaboration, institutional affiliations, and research focus undergo significant shifts during this period, reflecting the multifaceted nature of scientific mobility. By examining how these factors interact and influence scholarly output through logistic regression modeling, we aim to unravel the intricate mechanisms driving scientists’ postmobility performance.

In our initial analysis, we utilized the classical measure of variety, entropy, to characterize the diversity of collaborative authors and institutions among different groups of scientists. Additionally, we employed entropy to assess the breadth of their research interests. Without loss of generality, we denote the collaborators of a scientist before mobility as X; then the variety of their collaborators is given as: H(X) = −$∑i=1n$P(x_i) · log₂(P(x_i)), where n is the number of possible outcomes of X and P(x_i) is the probability of the ith outcome of X. Higher H(X) indicates greater randomness in the group of collaborators, signifying a broader range of coauthors, while lower entropy values suggest more consistency or repetition among coauthors.

In this study, we analyze the variety of collaborators and collaborative institutions among different groups of scientists. Our findings reveal that Nonreturnees demonstrate relatively stable patterns in these two aspects, while Returnees-Nonenrolled and Returnees-Enrolled exhibit a broader range of collaborators, particularly following their mobility experiences (Figure 4A). Scientists who experience new scientific environments often seek connections with various peers within their research field and maintain collaborations with former colleagues prior to their mobility. Compared to Returnees-Nonenrolled, Returnees-Enrolled, who are enrolled in talent programs targeting outstanding young scientists, have established more connections with scientists before their mobility. This observation aligns with the focus of talent programs on fostering high-quality collaborations. Additionally, the broader spectrum of collaborators corresponds to a wider array of collaborative institutions (refer to Figure 4B).

An analysis of the diversity of research directions among the three groups of scientists reveals distinct patterns. Mobility appears to stimulate scientists’ interest in exploring various research directions, especially among returnees who are not enrolled in talent programs (Figure 4C). This indicates that experiences of mobility significantly influence researchers’ inclination to expand their scientific interests and pursue diverse lines of inquiry. The absence of enrollment in talent programs may enhance the impact of mobility on researchers’ openness to exploring new research directions, underscoring the importance of such programs in cultivating focused research trajectories and advancing scientific knowledge. Conversely, scientists tend to focus more on the same research direction when they do not experience mobility. Moreover, Returnees-Enrolled exhibit greater stability in their diversity of research directions, likely due to their established reputation in their current field from previous research endeavors, leading them to delve deeper into their specialized area of study.

In light of the observed trends in scientific movement, we now turn our attention to the primary challenge faced by researchers: navigating the changes in the scientific environment. The movers will get new affiliations with new colleagues, potential extra funding, and students while being faced. We identify a series of factors demonstrably linked to academic success, which includes the change rates of (1) collaborators D_A, (2) range of collaborative institutions D_I, and (3) topic directions D_C, the details of which are described in Section 5. Additionally, we factor in differences in team size (ΔTeam Size). Controls, including career inception year, academic discipline, year of movement, and scientist grouping, are also applied (refer to Section 5 and Section S3.3 in the Supplementary material for an in-depth explanation).

We employ logistic regression models to examine the correlations between these variables and scholarly outputs after movement. The models contrast Returnees-Enrolled with their Nonreturnee and Returnees-Nonenrolled peers in the first step of our two-step matching process. Scholarly success is encoded as a set of binary dependent variables, reflecting whether a scientist’s output surpasses the median in publication and citation counts, Q1-publications, and collaborator quantity among their contemporaries.

Table 2 highlights that broadening one’s collaboration network substantially enhances prospects for future scientific performance. Nevertheless, transitioning to a different research focus adversely impacts the four aspects of academic performance, as significantly (p-value < 0.001) indicated by the coefficients for D_C (−0.2345 for publications, −1.6650 for citations, −0.2585 for Q1-publications, and −0.5114 for collaborators). This decline is attributable to established scientists losing underlying advantages, such as their leadership and reputation in initial fields of study. Recognition in new areas necessitates a period of adjustment during which their contributions might go unnoticed. Further analysis in Figure 5 explores how potential success in key academic metrics fluctuates with different scientific factors, estimated from the models in Table 2. The findings confirm that scientists with over 80% new collaborators postmigration exhibit at least a 50% likelihood of thriving in publications, citations, Q1-publications, and establishing a successful network of collaborators. Shifts in research focus slightly dampen productivity, high-quality output, and network expansion, while significantly curtailing citation rates. Moreover, an increase in average coauthors per publication, indicative of larger team sizes, inversely affects publication success rates. The variability in team size is directly associated with the breadth of one’s collaborative network.

Our results highlight the role of startup funding for early career scientists, especially in experimental disciplines that require more resources. However, because the Young Thousand Talents program only covered a small fraction of young scientists (about 4,000), we suggest that more support should be given to other young scientists who did not receive this kind of funding. This would help to foster a more inclusive and diverse scientific community in China. We also show that the talent program has less impact on some theoretical disciplines, implying the need for discipline-specific career development policies. However, it is important to acknowledge that our analysis is constrained by the absence of comprehensive data on other forms of research funding. We face constraints in quantifying resources, and future research endeavors that encompass a broader range of funding sources would enable a more comprehensive understanding of the intricate interplay between funding mechanisms and scientific productivity among early career scientists in different disciplines.

Our work raises an open question for cultivating cross-border scientists and science development. The younger PIs, who have just become independent from their collaborators and mentors, inevitably need to change their research directions, especially for cross-border scientists whose research environments undergo significant shifts when they relocate. Such transitions align with the general pattern observed in science, often referred to as the “pivot penalty,” where researchers may face challenges and productivity dips when they switch their research focus. Despite the initial challenges, these shifts offer opportunities for growth, interdisciplinary collaboration, and innovative contributions to science (Hill, Yin et al., 2021). Exploring new research directions is necessary for disruptive science and should be encouraged; however, we found that changing research directions after moving was associated with a decrease in both publications and citations, suggesting a risk for young scientists who are newly independent of their collaborators and mentors. We recommend that the host institutions provide more guidance and flexibility for young talents to establish their research agendas in a new environment, for instance, by extending the tenure track probationary period.

Moreover, we observed that young scientists who downsized their teams were more likely to achieve higher levels of productivity (Figure 5M reflects this trend), with smaller team sizes after the cross-border move associated with increased productivity. Previous studies showed that smaller teams tend to introduce more disruptive ideas (Wu, Wang, & Evans, 2019). This implies that downsizing teams may allow young scientists to challenge established paradigms and achieve greater success in their research careers, which, in return, can also serve as an internal motivator for increasing their productivity.

During the past decade, the Young Thousand Talents program has undoubtedly made significant strides in attracting international scientific talent to China. Yet, our findings precipitate a number of questions and implications concerning the future trajectory of such initiatives. As for policy implications, our study indicates that more inclusive and differentiated policies could be implemented to bolster the program’s breadth and depth, considering the diverse needs across scientific disciplines. The efficacy of the Young Thousand Talents program, while evident, is not without its drawbacks, such as a potential overemphasis on certain research areas and reliability decreasing with time. Specifically, our analysis indicates that while this program enhances the academic performance of scientists in terms of both the quality and quantity of scientific publications, there are variations across disciplines. For instance, individuals specializing in materials science tend to experience more significant improvements compared to those in mathematics. Moreover, the validity period of this program typically spans 3–5 years. Beyond this timeframe, the differences in four aspects of academic performance between participants who were enrolled in the program and those who were not begin to narrow. While the program has been beneficial in providing opportunities for relocated scientists to restart their careers, it has not been as successful in cultivating long-term active and highly productive scientists. As such, there is a need for the program to reassess its objectives and make modifications to nurture prolific scientists. The question of whether adjustments are necessary to sustain and enhance the program’s objectives remains open, encouraging a dialog on its continued evolution.

In terms of methodology, the use of matching design has demonstrated value in assessing participant outcomes. Nevertheless, its limitations are clear: challenges in capturing the full gamut of variables that define a scientist’s potential. Moreover, we opted for OpenAlex due to its comprehensive publication records and the rich details it provides on interactions among publications, authors, and institutions. However, this choice may compromise some data specificity. This data set may lack the curation depth of commercial databases such as Web of Science (WoS) and Scopus, which are known for their selective high-quality content and provide advanced bibliometric tools and author disambiguation services. To mitigate this, future research will incorporate additional data sources such as ORCID. This integration aims to enhance our ability to accurately track and analyze scientists’ mobility patterns, thereby enriching our study’s depth and precision.

Assessing academic performance is a complex task, especially when considering factors beyond productivity and citations. Recognizing the multifaceted nature of scholarly work, we aimed to broaden our analysis by incorporating various metrics. These metrics, spanning four dimensions, including the number of publications, citations, publications in JCR Q1 journals, and the extent of collaboration, provide insight into researchers’ ability to produce high-quality work and foster connections within the scholarly community. While we acknowledge the importance of additional measurements, such as annual research funding acquisition, their inclusion is impeded by the lack of comprehensive data concerning funding in countries all over the world. Nevertheless, our examination, rooted in the four dimensions of academic performance derived from the OpenAlex data set, makes a significant contribution to the ongoing scholarly conversation on this topic.

Our study offers a nuanced understanding of the dynamics of recruiting and retaining international talent in higher education. It underscores the imperative for refined, data-driven policymaking, adaptable to the rapidly changing global science and technology landscape. A critical takeaway for other higher education systems is the importance of targeted policy design. Our results indicate that strategies tailored to the specific needs of international talent are more effective. Additionally, providing sufficient support and resources is essential for retaining these talents. The study highlights the necessity of ongoing policy evaluation and adaptation. After nearly a decade, the Young Thousand Talents program clearly requires updating to meet new challenges and shifts in the global academic context, particularly in encouraging scientists to pursue innovative research directions upon settling in China. This can be achieved by introducing new initiatives for scientists venturing into novel research areas and expanding the program to encompass a broader range of research fields and a more diverse talent pool.

We thank the anonymous reviewers for their insightful and constructive feedback.

Yurui Huang: Data curation, Investigation, Visualization, Writing—original draft. Xuesen Cheng: Formal analysis, Writing—review & editing. Chaolin Tian: Validation, Writing—review & editing. Xunyi Jiang: Data curation, Validation, Writing—review & editing. Langtian Ma: Data curation, Validation, Writing—review & editing. Yifang Ma: Conceptualization, Project administration, Supervision, Writing—review & editing.

The authors have no competing interests.

This work was supported by the National Natural Science Foundation of China (grants No. NSFC62006109 and NSFC12031005), the 13th 5-year plan for Education Science Funding of Guangdong Province No. 2020GXJK457, the Stable Support Plan Program of Shenzhen Natural Science Fund No. 20220814165010001, and partly sponsored by SUSTech Research Series No. SUSTECH2020C007. The computation in this study was supported by the Center for Computational Science and Engineering of SUSTech.

The OpenAlex data is publicly available. The code and data for reproducing the main results in this work are available in GitHub: https://github.com/YuruiHuang/QSS2024Proj/.