What Drives the Reverse of Overseas Brain Drain? Identifying the Critical Factors by a Hybrid Grey DANP Technique

Abstract

Against the backdrop of intensified global talent competition, the return of overseas talents has become a key engine driving the enhancement of core competitiveness in developing countries. Accurately identifying its critical driving factors is essential for China to address the challenges of talent introduction. This study constructs a hybrid multiple-criteria decision-making framework to systematically explore the influence mechanism of overseas talent return: first, a 15-criterion decision structure covering economic, policy, educational, technological, and social aspects is established via systematic literature review and two-round Delphi expert surveys; second, the grey DEMATEL-ANP technique is adopted to quantify the inter-relationships and relative weights of the criteria and screen and rank the critical driving factors accurately. Empirical results show that the six core driving factors ranked by importance are talent policy support, economic development level, scientific and technological development strength, public service quality, educational resource supply, and attention to science and technology, with significant synergistic interaction effects among these factors. This research provides a scientific decision-making framework and empirical support for developing countries to formulate targeted talent introduction policies and optimize the talent development ecosystem.

1. Introduction

In the knowledge economy era with fierce global talent competition, high-end talents serve as the core engine of knowledge spillover, technological innovation, and economic growth. Their cross-border flow pattern has become a key variable reshaping national and regional competitiveness [1,2]. Since the 1980s, with the rapid economic rise of newly industrialized countries and some developing countries, a new trend of “reverse brain drain” has emerged in global talent mobility [3,4]. Countries that have traditionally suffered from brain drain (primarily developing countries except for newly industrialized countries with rapid economic rise) have gradually transformed into net talent inflow destinations, offering a new research context for the formulation of global talent strategies [5]. China, a typical representative of this trend, has experienced a historic leap from “large-scale brain drain” to “the number of returnees exceeding the number of emigrants”. Its transformation path and practical experience provide valuable references for developing countries worldwide.

According to statistics from China’s Ministry of Education, from 1978 to 2024, the total number of Chinese people of various types studying abroad reached 8.88 million, among whom 7.43 million completed their studies and 6.44 million chose to return to China for development after graduation, providing robust talent support for advancing national strategic development. Notably, since 2012, a total of 5.63 million returnees have served in China, accounting for 87% of the total number of returnees since 1978. In 2024, 495,000 overseas students returned to China, an increase of 79,400 compared with 2023, a year-on-year growth of 19.1%, indicating an accelerated trend in overseas talent return. This sharp growth in 2024 is closely linked to changes in overseas visa policies (e.g., stricter restrictions and shortened validity periods in some major destination countries) and the continuous expansion of high-quality career opportunities in China’s high-tech and knowledge-intensive industries. During the COVID-19 pandemic (2020–2022), international travel restrictions, disruptions to overseas academic and career activities, and global market volatility further enhanced the relative attractiveness of China’s stable development environment and targeted talent policies, becoming an important contextual factor driving this trend. Since 2020, this trend has evolved into a historic shift with returnees outnumbering outgoing students, driven by post-pandemic global talent mobility dynamics. As shown in Figure 1, the number of overseas students and returnees and the proportion of returnees in China have undergone significant changes over the past 25 years, with the continuous rise in the proportion of returnees and the accelerated growth in recent years fully confirming China’s growing appeal to overseas talents.

Meanwhile, the adjustment of the global economic structure has further boosted the talent return rate. As shown in Figure 2, the source countries of Chinese returnees seeking jobs in 2022 were mainly the UK (41.4%), Australia (20.3%), and the US (9%). This covers major developed countries globally, reflecting the global scope of overseas talent return to China.

Currently, China is in a critical period of economic transformation and industrial upgrading. Traditional labour-intensive industries are gradually being transferred or eliminated. Knowledge-intensive economy and high-tech industries have become the main drivers of development [6], creating an urgent demand for overseas high-end talents [7]. China’s talent return phenomenon not only is an important achievement of its “talent-powered country” strategy but also provides a practical model for other countries facing talent mobility dilemmas worldwide. Therefore, analyzing the core driving factors of overseas talent return responds to China’s practical development needs and offers scientific support for global talent mobility and talent strategy formulation, bearing distinct contemporary and international significance.

Existing studies have confirmed that a region’s attractiveness to overseas talents stems from multi-dimensional factors, including economic development level [8,9], technological development potential [10], quality of life [11], policy support [12,13,14], and humanistic and historical environment [15]. Overseas talent return positively empowers the home country’s regional economic development and technological innovation through knowledge transfer, skill transmission, and capital introduction [16,17]. Their cross-border flow is also the core path of regional knowledge spillover [2,4]. Feng et al. [18] pointed out that the spread of ideas and knowledge flow triggers knowledge spillover. Zweig [4] and Wang et al. [2] further confirmed that the cross-border flow of high-tech talents is the main source of regional knowledge spillover.

Despite the theoretical foundation laid by existing studies, obvious gaps remain: Firstly, most studies focus on the independent impact of single factors, ignoring the interdependent and feedback relationships among various factors [19,20], making it difficult to reveal the complex mechanism of talent return decisions. Secondly, there is a lack of systematic indicator system construction for the return of overseas high-end talents, leading to inaccurate identification of key driving factors. Thirdly, research methods are mostly descriptive statistics or simple regression, which are incompetent in handling the “incommensurable and conflicting” complex problems in multi-criteria decision-making (MCDM) [21,22], failing to fully reveal the inherent interaction mechanism among factors.

This study focuses on the core research problem: how to systematically identify and quantify the critical driving factors of overseas talent return and their interdependent interaction mechanisms, so as to provide a scientific decision-making basis for developing countries to formulate targeted talent introduction policies. And this study sets three specific research objectives: (1) integrate multi-dimensional factors including economy, science and technology, policy, living environment and humanistic ties to construct a comprehensive and systematic indicator system of key influencing factors for overseas talent return; (2) apply the grey DEMATEL-based ANP (Grey DANP) hybrid multi-criteria decision-making model to quantify the weight of each factor, reveal its interaction relationship and accurately identify the critical driving factors; (3) conduct an empirical study of China as a typical sample of developing countries and provide a scientific basis and operable reference for China and similar countries around the world to optimize talent policies and enhance talent attractiveness.

This study’s originality and contributions are threefold. First, by integrating a systematic literature review and the Delphi method, this study develops a systematic 5-dimension, 15-criterion decision-making framework for overseas talent return drivers, addressing the research gap of fragmented indicators in existing studies. Second, this study pioneers the application of the Grey DANP model to identify critical drivers of overseas talent return, which uses grey system theory’s strength in handling problems with “clear extension and unclear connotation” to quantify inter-factor interdependence and feedback relationships, overcoming the applicability limitations of traditional questionnaire-based regression models. Third, with China as a typical sample, this study reveals the shift of overseas talent return motivation from emotion-driven to value-driven, verifies that scientific and technological development level is the core driver, and enriches empirical evidence for global talent mobility research.

The remainder of this paper is structured as follows: Section 2 reviews the relevant literature and constructs a prototype decision structure for overseas talent return. Section 3 elaborates on the principles and procedures of the Delphi method and Grey DANP hybrid model. Section 4 conducts an empirical study to measure the weights and causal relationships of influencing factors. Section 5 discusses the core findings and factor interaction effects based on empirical results. Section 6 presents the conclusions and clarifies the research limitations.

2. Establishing the Decision Structure

The driving mechanisms behind the return migration of overseas talents are shaped by a complex array of factors. Through a systematic review, this study identifies five primary aspects in the existing research on the determinants of talent return.

2.1. Economic and Career Development

Economic opportunities and career prospects serve as key pull factors [13]. Li et al. [23] identify financial returns and quality of life as primary drivers of talent mobility. Yu et al. [24] further confirm that economic opportunities exert a stronger influence than living environment for highly educated talents. In cross-border contexts, Bhardwaj and Sharma [25] argue that return intention increases significantly when overseas earnings fall below expectations, return costs are manageable, and the home country offers higher returns for skill translation. Empirical work by Pham [26] shows that high-skilled returnees boost local wage levels through human capital spillovers, creating a positive cycle. In China, economic incentive policies and local support measures have effectively attracted overseas talents to fill gaps in key technological fields [27], underscoring regional economic development as a central attraction. Notably, career development is a multifaceted factor integrating economic, psychological, and social attributes: economically, it is closely linked to income growth and professional advancement; psychologically, it satisfies talents’ needs for self-esteem and self-actualization; and socially, it is shaped by educational resources, cultural expectations, and institutional contexts, reflecting the synergistic effect of multiple factors.

2.2. Science, Technology, and Innovation Environment

The level of scientific development and commitment to R&D is crucial for attracting technical talent. Research from Jiang et al. [28] and Wang [12] highlights the positive impact of regional technological strength and R&D investment. Focusing on participants of the “Thousand Talents Program,” Liu et al. [29] used CV data to show that strong governmental investment in research is a key driver for high-end talent return. Although the programme expanded to include foreign experts in 2010 [30], China still faces a talent trade deficit. This highlights the critical role of the S&T environment: only an internationally competitive innovation ecosystem can sustainably attract high-value talent [31].

2.3. Policy and Institutional Environment

Policy support and institutional safeguards are important for lowering barriers to return. Studies by Xu et al. [31], Jiang [32], and Kurokawa & Kusakabe [33] show that talent policies, political systems, and public governance directly influence return decisions. Drawing on developed countries’ experiences, Yi [9] notes that targeted policy designs (e.g., streamlined residency and start-up research grants) effectively increase return willingness. Jin et al. [34] add that return costs and domestic labour market conditions are significant institutional factors that policies must address. Furthermore, the quality of public services (e.g., healthcare and education), as an extension of the institutional environment, affects post-return adaptation and development [35].

2.4. Social and Family Ties

Traditional studies regarded cultural attachment and family bonds as core emotional drivers of talent return, arguing that the sense of belonging from shared culture and kinship could offset certain economic trade-offs [30,36]. However, with the deepening of global talent mobility and the upgrading of talent demand in developing countries, the role of social and family ties has evolved from a “dominant driver” to a “supplementary synergistic factor”. Contemporary high-end talents prioritize professional value realization and career development platforms, while social and family ties act as a “catalyst” to consolidate return intentions rather than a decisive factor [16,23]. Notably, emotional factors alone rarely drive high-end talents to return without matching professional opportunities or innovation ecosystems [30]. Social and family ties interact with economic development, technological environment, and policy support.

2.5. Quality of Life and Livability

The demand for a high quality of life is increasingly important for high-skilled talents. Zhu [37] argues that modern talents increasingly prioritize livability over economic opportunity alone. This aspect covers multiple facets: physical and mental health security, economic security, environmental sustainability, work–life balance, and access to quality education and recreational facilities [11,38,39,40,41]. These factors are not independent; for example, clean living environments rely on environmental governance policies, while work–life balance is shaped by industrial development models and social values. Their improvement directly reduces post-return adaptation costs, complementing economic and technological pull factors to form a comprehensive attraction.

2.6. Prototype Decision Structure

Based on the systematic review of five core dimensions affecting overseas talent return, this section constructs a prototype decision structure for the driving factors of reverse brain drain. Specifically, we first clarify the inherent interlinkages and logical relationships among the five dimensions to lay a theoretical foundation for the construction of the indicator system, then propose the initial index structure based on literature evidence, and finally put forward the corresponding research hypotheses according to the synergistic mechanism among the dimensions.

2.6.1. Interlinkages Among the Five Aspects

The five aspects form an interdependent and synergistic system rather than isolated factors, with clear causal and supportive relationships:

Foundation-driven linkage: Economic and career development provides material support for all other aspects. The economic development level determines the investment scale of scientific and technological R&D and educational resources, while career development opportunities directly enhance the attractiveness of the science, technology and innovation environment.

Policy-bridging linkage: The policy and institutional environment acts as a “connecting hub”; talent policies optimize the cost–benefit ratio of talent return, and high-quality public services directly improve livability and reduce social integration costs.

Demand-driven linkage: The quality of life and livability aspect and social and family ties aspect respond to the dual needs of talent for “material security” and “emotional belonging”; factors such as clean air, safe living conditions, and work–life balance rely on the improvement of public services and economic strength, while family relationships and national attachment synergize with career development to promote return intention conversion.

Core-leading linkage: The science, technology and innovation environment aspect is a core pull factor in the knowledge economy era; its development level drives the upgrading of economic industries and the refinement of talent policies, forming a positive cycle of “innovation-driven development → talent agglomeration → further innovation”.

2.6.2. Initial Index Structure

Based on the aforementioned literature, the criteria influencing overseas talent to return are selected and integrated. Next, these criteria are classified into different aspects according to their definitions and functions. Following these steps, this study is able to propose a prototype of a decision structure consisting of five perspectives: economic and career development; science, technology, and innovation environment; Policy and institutional environment; social and family ties; and quality of life and livability.

Table 1. Prototype decision structure.

Aspect	Criterion	References
Economic and Career Development	Financial returns	[23,34]
	Quality of life	[23]
	Economic opportunities	[24,25]
	Local wage levels	[26]
	Economic incentive policies	[27]
	Career prospects	[13,16,23,27]
Science, Technology, and Innovation Environment	Level of scientific development	[28]
	Commitment to R&D	[28]
	Governmental investment in research	[29]
	Internationally competitive innovation ecosystem	[31]
Policy and Institutional Environment	Talent policies	[31]
	Political systems	[22]
	Public governance	[33,35]
	Targeted policy designs (educational facilities, streamlined residency, and start-up research grants)	[9]
	Domestic labour market conditions	[34]
Social and Family Ties	Cultural attachment	[30]
	Family bonds	[36]
Quality of Life and Livability	Excellent climate and livable natural environment	[37]
	Physical and mental health security	[11]
	Economic security	[38,39,40,41]
	Convenience and sustainability of living environment	[38,39,40,41]
	Work–life balance	[38,39,40,41]
	Recreational facilities	[38,39,40,41]

shows these perspectives and the criteria used to assess them.

2.6.3. Hypotheses

Based on the above analysis of the interdependent synergistic relationships among the five core aspects and the 15-criteria decision structure, two research hypotheses are proposed to empirically test the driving mechanism of overseas talent return:

H1. 

The driving factors of overseas talent return are not independent; the five core aspects (economic and career development, science and technology and innovation environment, policy and institutional environment, quality of life and livability, and social and family ties) exhibit significant interdependent, feedback, and synergistic interaction effects.

H2. 

In the knowledge economy era, the level of scientific and technological development is the primary driving factor for overseas talent return, and its influence weight is higher than traditional economic factors and social–family emotional factors.

3. The Proposed Grey DANP

Based on the foundational MCDM framework proposed by Tzeng and Huang (2011) [42], this study innovatively develops a Grey DANP hybrid model by integrating grey system theory with the DANP method. Traditional questionnaire-based regression methods are unsuitable for this research, as they can only verify the independent effect of single factors and fail to depict the nonlinear interaction and causal transmission among multi-dimensional driving factors. The core advantage of grey system theory lies in solving uncertain problems with clear extension and unclear connotation, which perfectly fits the inherent fuzziness and information incompleteness in expert evaluations of talent return drivers. Compared with the conventional DANP method, our Grey DANP model effectively reduces the subjective bias of expert evaluations and markedly improves the robustness of evaluation results [43,44,45,46].

Specifically, the Grey DANP model is utilized to rank the influencing factors of overseas talent return. Among them, grey DEMATEL is used to test the causal relationship between influencing factors, and then the relative weight of the criterion is calculated using the ANP method based on the causal relationship obtained from grey DEMATEL.

3.1. Relationships Among Driven Factors

3.1.1. Build an Index Structure for Influencing Factors

The Delphi method is an appropriate method to reach expert consensus and is used to select the aspects and criteria for overseas reverse brain drain. In this study, the quartile deviation (QD), defined as one-half the interquartile range (IQR), which is the difference between the 25th and the 75th percentiles in a frequency distribution [47], is applied to determine consensus. According to the measurement of consensus proposed by Faherty [48], the items that received a QD < 0.6 were considered to have achieved high consensus, and a moderate consensus was defined as a QD > 0.6 and < 1. In practice, an accepted consensus is achieved when the QD is no larger than one [47,48]. If consensus is not reached, we have to continue to repeat the questionnaires until it is. The operation steps of the Delphi method are referred to in Figure 3.

3.1.2. Establish Initial Direct Influencing Matrix

The questionnaire, as the core tool for data collection, is designed and validated based on rigorous academic standards, with detailed information as follows:

Questionnaire type: This study uses a pairwise comparison questionnaire developed based on the established five-dimension, 15-criterion indicator system (Appendix A 

Table A1. Questionnaire design.

Questionnaire
Dear, Firstly, I would like to thank you for your availability of promptly responding to the invitation to be a respondent to this questionnaire. The questionnaire aims to conduct a factor analysis on attracting more international students to serve their home country, establish a framework design for the influencing factors of overseas talent return, and better serve relevant academic research. We highlight that the confidentiality of your personal data is guaranteed. Name of the specialist: Company or institution: Time of experience: Position: This questionnaire is a pairwise comparison, please fill in the table below indicating the influence among the barriers, as follows: 0—no influence|1—very low influence|2—low influence|3—high influence|4—very high influence|
Abbreviation	Factors	A1	A2	A3	B1	B2	B3	B4	C1	C2	C3	C4	D1	D2	E1	E2
-	Economic and Career Development	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
A1	Economic Level	0
A2	Cost of returning to the country of origin		0
A3	Career development			0
-	Policy and Institutional Environment	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
B1	Cultural Assets				0
B2	Public Service					0
B3	Talent Policy						0
B4	Educational Level							0
-	Quality of Life and Livability	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
C1	Policy and Institutional Environment								0
C2	Natural Environments									0
C3	Recreational Facilities										0
C4	Convenience of Life											0
-	Social and Family Ties	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
D1	Family relationships												0
D2	Attachment to the motherland													0
-	Science, Technology, and Innovation Environment	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-
E1	Development Level of Science and Technology														0
E2	Attention to Science and Technology															0

Note: The definitions of the letter-number abbreviations in this table refer to Figure 5; the same below.

). This questionnaire is designed to quantify the mutual influence intensity between each pair of factors, which aligns with the analytical logic of the Grey DANP model for examining inter-factor interdependence.

Questionnaire structure: The questionnaire comprises three sections. The first section collects only the core background information of respondents’ main reasons for overseas stay, to ensure respondent anonymity and reduce filling concerns. The second section details the DEMATEL-specific 5-point influence scale (0 = no influence, 1 = very low influence, 2 = low influence, 3 = high influence, and 4 = very high influence) with supporting examples, to standardize respondents’ understanding of the scoring rules. The third section includes the full pairwise comparison matrix for all criteria, with standardized instructions requiring respondents to score the influence intensity of each row factor on each column factor per the specified scale.

Validation process: (1) Content validity: The initial questionnaire was developed based on a systematic literature review of high-quality studies on overseas talent return. Then, interdisciplinary experts were invited to conduct rounds of Delphi consultation to revise the indicator expression, scale rationality, and matrix design of the questionnaire; ambiguous and inappropriate items were deleted and optimized, and the final formal questionnaire was formed after expert consensus was reached, which fully ensures the content validity of the questionnaire. (2) Structural validity: The causal relationship between factors obtained from the questionnaire data is completely consistent with the theoretical framework constructed in Section 2 of the manuscript, and the total influencing matrix passed the consistency test, which verifies the good structural validity of the questionnaire. (3) Pilot test: A pilot survey was conducted among 20 individuals with overseas study or work experience and three academic researchers in related fields. Feedback on readability and operability was collected to revise the questionnaire, ensuring respondents could accurately understand evaluation requirements. (4) Reliability check: The consistency of expert evaluations was tested using the QD. All QD values of the pairwise comparison results were <1, indicating sufficient consistency of expert opinions and reliable questionnaire results [48]. Please refer to Appendix A Table A1 for the complete questionnaire.

3.1.3. Construct Grey Number Matrix

The elements in the direct impact matrix obtained from the questionnaire survey are all semantic variables of expert comments. Based on the grey number corresponding to the semantic variables (

Table 2. Semantic variables for expert evaluation.

Semantic Variable	Interval Grey Number	Evaluation Value
no influence	[0.00,0.00]	0
very low influence	[0.00,0.25]	1
low influence	[0.25,0.50]	2
high influence	[0.50,0.75]	3
very high influence	[0.75,1.00]	4

), the direct impact matrix is transformed into a grey number matrix. Furthermore, the grey number matrix is clarified using formulas to obtain the initial matrix.

To standardize multi-expert opinion integration and avoid subjective weight assignment in the Delphi method,

Table 3. Semantic variables for expert evaluation weight.

Semantic Variable	Interval Grey Number
not important	[0.0,0.3]
slightly unimportant	[0.3,0.5]
important	[0.4,0.7]
more important	[0.5,0.9]
very important	[0.7,1.0]

defines semantic variables for expert weight evaluation and their corresponding interval grey numbers.

To ensure the rationality and scientificity of expert weight assignment in the Delphi method and the subsequent grey number matrix construction, the semantic variables for expert weight evaluation are defined with corresponding interval grey numbers, as presented in Table 3.

3.1.4. Calculate Direct Impact Matrix

The direct impact matrix obtained from the questionnaire is composed of semantic variables from expert evaluations. Based on the interval grey numbers corresponding to the semantic variables, we first convert the direct impact matrix into a grey number matrix. Then, we perform whitening and standardization on the grey number matrix following the classic grey system theory framework and integrate the evaluation results of all experts to obtain the final direct impact matrix A. The detailed calculation formulas and standardization steps for grey number whitening and expert evaluation integration are presented in Appendix B.1.

3.1.5. Calculate the Comprehensive Impact Matrix

We standardize the direct impact matrix A to obtain the standardized direct impact matrix N and then calculate the comprehensive impact matrix T, which quantifies the total direct and indirect influence between each pair of criteria. The core calculation formula is as follows:

$$T=N{(\mathrm{I}-N)}^{-1}$$

(1)

where I is the identity matrix and the spectral radius of matrix N is less than 1 to ensure the convergence of the matrix calculation. The detailed standardization process of the direct impact matrix is shown in Appendix B.2.

3.1.6. Calculate the Degree of Influence (${\mathrm{d}}_i$) and the Degree of Being Influenced (${\mathrm{r}}_j$)

The degree of influence ${\mathrm{d}}_i$ is the sum of each row in the comprehensive impact matrix T, representing the total direct and indirect impact of the corresponding factor on all other factors. The degree of being influenced ${\mathrm{r}}_j$ is the sum of each column in the comprehensive impact matrix T, representing the total impact of all other factors on the corresponding factor. The detailed calculation formulas for ${\mathrm{d}}_i$ and ${\mathrm{r}}_j$ are provided in Appendix B.3.

3.1.7. Calculate the Centrality and Causal Degree of Each Influencing Factor

Centrality $P_i$ represents the position and degree of influence of a certain influencing factor in the entire system evaluation process, and a larger centrality indicates that the role of the influencing factor is more critical. The degree of causation $E_i$ can be divided into positive and negative factors, indicating that the factor has a significant impact on other factors and $E_i>0$ is a factor that affects other factors, known as the causal factor. The factor $E_i<0$ that is greatly influenced by other factors and is influenced by other factors is called the outcome factor.

$$P_i=\left\{R_i+D_j\left|i=j\right.\right\}$$

(2)

$$E_i=\left\{R_i-D_j\left|i=j\right.\right\}$$

(3)

3.1.8. Analyze Model Calculation Results

Construct a Cartesian coordinate system using causation and centrality as the horizontal and vertical axes and draw a causal relationship diagram of influencing factors. At the same time, mark the positions of each influencing factor, and analyze the causation and centrality of each factor and their corresponding importance. Conduct a comprehensive analysis based on the ranking of influence, influence, centrality, and cause, and ultimately effectively identify the key factors affecting the return of overseas talents.

3.2. Obtaining Weights of Driven Factors

Based on the comprehensive influence matrix obtained from grey DEMATEL, we construct the unweighted super matrix of the ANP method and obtain the weighted super matrix through normalization processing. Finally, the limit super matrix is calculated to get the stable relative weight of each criterion [49,50]. The detailed construction and normalization formulas of the unweighted super matrix and weighted super matrix are presented in Appendix B.4. The core formula for calculating the limit super matrix to obtain the final weight of each criterion is as follows:

$$W^{\ast }=\underset{k\to \infty }{\lim }{W}_w^k$$

(4)

where $W_w$ is the weighted super matrix and $W^{\ast }$ is the convergent limit super matrix, from which the final weight value of each criterion can be extracted.

The flowchart of the proposed hybrid model in this paper is shown in Figure 4.

4. Empirical Study

4.1. Formal Decision Structure

In order to establish an authoritative decision structure, five experts are invited to provide advice on the aspects and criteria. The background of the five invited experts is shown in

Table 4. Professional backgrounds of the invited five experts for the Delphi survey.

Expert	Organization	Position	Duties	Seniority (yr)
A	Human Resources and Social Security Bureau of a certain city in China	Deputy director-general	Organize the implementation of human resources market development plans and human resources mobility policies	23
B	Organization Department of a certain municipal party committee in China	Section chief of talent	Talent attraction and development, as well as the formulation of talent policies	14
C	Science and Technology Bureau of a certain city in China	Deputy director	Responsible for the introduction of intelligence and the formulation of technology policies in the city	18
D	A headhunting company	Senior talent advisor	Assist employers in recruiting overseas returnees	15
E	School of Public Administration at a certain university	Professor	Engaged in long-term research in the field of population mobility and distribution	22

.

First, the expert panel is requested to rate the necessity of items to be included in the formal decision structure. A five-point Likert-type scale is used, and the relationship between rating and necessity is shown in

Table 5. Relationship between rating and necessity.

Rating	1	2	3	4	5
Necessity	Strongly unnecessary	Unnecessary	Average	Necessary	Strongly necessary

.

In the first round of the Delphi survey, the panel members suggest that criteria should be amalgamated. According to the meaning of each criterion, the criteria with the same meaning are deleted. After the first round of the Delphi questionnaire, the prototype decision structure shown in Table 1 is integrated into five aspects and 22 criteria. The integrated decision structure is shown in

Table 6. The integrated decision structure.

Aspect	Criteria
Economic and Career Development	Economic level
	Quality of life
	Cost of returning to the country of origin
	Economic incentive policies
	Career development
Science, Technology, and Innovation Environment	Development level of science and technology
	Attention to science and technology
	Internationally competitive innovation ecosystem
Policy and Institutional Environment	Cultural assets
	Public service
	Social system
	Talent policy
	Educational level
Social and Family Ties	Family relationships
	Attachment to the motherland
Quality of Life and Livability	Climate
	Natural environments
	Physical and mental health security
	Economic security
	Recreational facilities
	Convenience life
	Sustainable environment

.

The second-round questionnaire is composed of all integrated items shown in Table 6. Panel members are asked to rate the need to include items in the formal decision structure. We compute the mean and QD of the panellists’ responses, and the results are shown in

Table 7. The results of the second round of the Delphi survey.

Criteria	Ascending Order					Average Value	QD	Classification
Economic level	4	4	4	4	5	4.2	0.25	High Consensus
Quality of life	1	2	4	4	5	3.2	1.50	No Consensus
Cost of returning to the country of origin	4	4	4	5	5	4.4	0.50	High Consensus
Economic incentive policies	2	2	3	4	5	3.2	1.25	No Consensus
Career development	3	4	4	4	5	4	0.50	High Consensus
Development level of science and technology	4	4	5	5	5	4.6	0.50	High Consensus
Attention to science and technology	4	5	5	5	5	4.8	0.25	High Consensus
Internationally competitive innovation ecosystem	1	1	3	5	5	3	2.00	No Consensus
Cultural assets	3	3	4	5	5	4	1.00	Moderate Consensus
Public service	5	5	5	5	5	5	0.00	High Consensus
Social system	1	2	2	3	3	2.2	0.75	Moderate Consensus
Talent policy	4	5	5	5	5	4.8	0.25	High Consensus
Educational level	3	4	4	4	5	4	0.50	High Consensus
Family relationships	4	4	4	4	5	4.2	0.25	High Consensus
Attachment to the motherland	3	4	4	4	4	3.8	0.25	High Consensus
Climate	4	4	4	5	5	4.4	0.50	High Consensus
Natural environments	4	4	4	4	5	4.2	0.25	High Consensus
Physical and mental health security	1	2	3	3	4	2.6	1.00	Moderate Consensus
Economic security	1	1	4	4	5	3.0	1.75	No Consensus
Recreational facilities	4	4	4	4	5	4.2	0.25	High Consensus
Convenience life	3	4	4	4	4	3.8	0.25	High Consensus
Sustainable environment	1	2	3	3	4	2.6	1.00	Moderate Consensus

.

Based on the results shown in Table 7, the criteria that did not reach consensus are deleted. Furthermore, the criteria with an average value lower than three are removed from the decision structure because they are deemed unnecessary. The formal decision structure is shown in Figure 5.

4.2. Identifying Critical Driven Factors

4.2.1. Data Collection

In this study, the survey respondents are overseas returnees. We define overseas talents based on their education level and occupational categories. Specifically, in the questionnaire, we define talents as the following three categories: (1) highly educated talents refer to individuals who have obtained a bachelor’s degree or higher overseas; (2) high-skilled talents, who are classified as “professional and technical personnel” in overseas occupations; (3) technological talents refer to individuals who are classified as “scientific researchers” or “engineering technicians” in their overseas professions.

A total of 300 questionnaires are distributed exclusively to Chinese nationals with overseas experience, and 287 valid questionnaires are collected. Among the 287 valid samples, the number of respondents in the three preset categories are 110, 86 and 91, respectively.

4.2.2. Obtaining Relevant Weights of Each Attribute Using Grey DANP

Following the Grey DANP procedure, the total influence matrix of all criteria is first computed (

Table 8. The total influence matrix.

	A1	A2	A3	B1	B2	B3	B4	C1	C2	C3	C4	D1	D2	E1	E2	d
A1	0.230	0.232	0.325	0.275	0.205	0.129	0.313	0.041	0.027	0.202	0.257	0.097	0.120	0.323	0.293	3.069
A2	0.116	0.061	0.126	0.117	0.069	0.039	0.120	0.010	0.009	0.058	0.089	0.063	0.086	0.109	0.103	1.175
A3	0.173	0.142	0.110	0.156	0.102	0.051	0.147	0.014	0.013	0.074	0.111	0.088	0.099	0.151	0.143	1.572
B1	0.254	0.193	0.258	0.141	0.151	0.074	0.208	0.035	0.035	0.117	0.162	0.090	0.107	0.241	0.215	2.281
B2	0.241	0.177	0.201	0.200	0.107	0.119	0.209	0.035	0.049	0.147	0.177	0.078	0.124	0.228	0.215	2.306
B3	0.229	0.151	0.220	0.218	0.151	0.079	0.241	0.020	0.019	0.114	0.133	0.157	0.173	0.232	0.206	2.342
B4	0.319	0.197	0.292	0.258	0.197	0.197	0.199	0.027	0.026	0.126	0.194	0.114	0.148	0.319	0.302	2.913
C1	0.105	0.075	0.073	0.082	0.070	0.057	0.070	0.007	0.038	0.081	0.060	0.025	0.047	0.058	0.053	0.901
C2	0.090	0.060	0.072	0.081	0.069	0.057	0.069	0.006	0.006	0.094	0.074	0.040	0.062	0.057	0.052	0.889
C3	0.111	0.078	0.078	0.072	0.074	0.044	0.076	0.008	0.008	0.037	0.093	0.025	0.046	0.092	0.073	0.914
C4	0.190	0.131	0.170	0.144	0.149	0.066	0.136	0.014	0.014	0.123	0.085	0.059	0.086	0.141	0.132	1.640
D1	0.103	0.084	0.115	0.095	0.061	0.064	0.098	0.008	0.008	0.052	0.053	0.033	0.114	0.085	0.079	1.052
D2	0.162	0.129	0.157	0.159	0.090	0.097	0.163	0.013	0.012	0.075	0.138	0.107	0.061	0.139	0.131	1.633
E1	0.339	0.216	0.284	0.287	0.217	0.144	0.312	0.056	0.055	0.162	0.241	0.097	0.120	0.211	0.292	3.032
E2	0.309	0.182	0.269	0.224	0.183	0.131	0.262	0.039	0.038	0.132	0.181	0.086	0.105	0.296	0.170	2.607
r	2.971	2.106	2.751	2.509	1.894	1.347	2.624	0.331	0.356	1.591	2.048	1.160	1.498	2.681	2.457

Note: The definitions of the letter-number abbreviations in this table refer to Figure 5; the same below.

). This matrix quantifies the direct and indirect interactions between factors. Based on the prominence (d + r) and relation (d − r) values (

Table 9. Prominence and relation of each factor.

	d	r	d + r	d − r
A1	3.0688	2.9710	6.0397	0.0978
A2	1.1752	2.1063	3.2815	−0.9311
A3	1.5721	2.7510	4.3230	−1.1789
B1	2.2812	2.5094	4.7905	−0.2282
B2	2.3063	1.8945	4.2007	0.4118
B3	2.3418	1.3474	3.6892	0.9944
B4	2.9126	2.6237	5.5363	0.2889
C1	0.9015	0.3312	1.2326	0.5703
C2	0.8892	0.3559	1.2451	0.5333
C3	0.9141	1.5910	2.5051	−0.6769
C4	1.6399	2.0485	3.6883	−0.4086
D1	1.0516	1.1596	2.2112	−0.1081
D2	1.6330	1.4979	3.1309	0.1351
E1	3.0318	2.6810	5.7127	0.3508
E2	2.6068	2.4574	5.0642	0.1495

), the 15 criteria are divided into two groups:

(1)

Cause group (d − r > 0): Economic level (A₁), public service (B₂), talent policy (B₃), educational level (B₄), climate (C₁), natural environments (C₂), attachment to the motherland (D₂), development level of science and technology (E₁), and attention to science and technology (E₂). These factors exert an active influence on other criteria.

(2)

Effect group (d − r < 0): Cost of returning to the country of origin (A₂), career development (A₃), cultural assets (B₁), recreational facilities (C₃), convenience of life (C₄), and family relationships (D₁). These factors are mainly affected by other criteria.

Subsequently, the total influence matrix is normalized to obtain the weighted super matrix. The limiting super matrix (

Table 10. The limiting super matrix.

	A1	A2	A3	B1	B2	B3	B4	C1	C2	C3	C4	D1	D2	E1	E2	RANK
A1	0.107	0.107	0.107	0.107	0.107	0.107	0.107	0.107	0.107	0.107	0.107	0.107	0.107	0.107	0.107	2
A2	0.041	0.041	0.041	0.041	0.041	0.041	0.041	0.041	0.041	0.041	0.041	0.041	0.041	0.041	0.041	11
A3	0.055	0.055	0.055	0.055	0.055	0.055	0.055	0.055	0.055	0.055	0.055	0.055	0.055	0.055	0.055	10
B1	0.080	0.080	0.080	0.080	0.080	0.080	0.080	0.080	0.080	0.080	0.080	0.080	0.080	0.080	0.080	7
B2	0.083	0.083	0.083	0.083	0.083	0.083	0.083	0.083	0.083	0.083	0.083	0.083	0.083	0.083	0.083	5
B3	0.082	0.082	0.082	0.082	0.082	0.082	0.082	0.082	0.082	0.082	0.082	0.082	0.082	0.082	0.082	6
B4	0.104	0.104	0.104	0.104	0.104	0.104	0.104	0.104	0.104	0.104	0.104	0.104	0.104	0.104	0.104	3
C1	0.033	0.033	0.033	0.033	0.033	0.033	0.033	0.033	0.033	0.033	0.033	0.033	0.033	0.033	0.033	13
C2	0.031	0.031	0.031	0.031	0.031	0.031	0.031	0.031	0.031	0.031	0.031	0.031	0.031	0.031	0.031	15
C3	0.032	0.032	0.032	0.032	0.032	0.032	0.032	0.032	0.032	0.032	0.032	0.032	0.032	0.032	0.032	14
C4	0.057	0.057	0.057	0.057	0.057	0.057	0.057	0.057	0.057	0.057	0.057	0.057	0.057	0.057	0.057	9
D1	0.037	0.037	0.037	0.037	0.037	0.037	0.037	0.037	0.037	0.037	0.037	0.037	0.037	0.037	0.037	12
D2	0.057	0.057	0.057	0.057	0.057	0.057	0.057	0.057	0.057	0.057	0.057	0.057	0.057	0.057	0.057	8
E1	0.109	0.109	0.109	0.109	0.109	0.109	0.109	0.109	0.109	0.109	0.109	0.109	0.109	0.109	0.109	1
E2	0.094	0.094	0.094	0.094	0.094	0.094	0.094	0.094	0.094	0.094	0.094	0.094	0.094	0.094	0.094	4

) is further derived to determine the relative weights of each criterion [51]. As shown in Table 10, the criteria ranking by weight is as follows: E₁ > A₁> B₄ > E₂ > B₂ > B₃ > B₁ > D₂ > C₄ > A₃ > A₂ > D₁ > C₁ > C₃ > C₂.

4.3. Criteria Classification by Impact Level

To simplify the presentation of results and help readers quickly grasp the core driving factors, we classify the 15 criteria into three levels (high/medium/low impact) based on their relative weights (from Table 10) and causal roles (from Table 9). The classification results are shown in

Table 11. Three-level classification of criteria by impact level.

Impact Level	Criteria	Weight Range	Causal Role	Core Characteristics
High	E1	>0.08	All belong to the “cause group” (d − r > 0)	Active driving role; strong influence on other factors; core determinants of talent return
	A1
	B4
	E2
	B2
	B3
Medium	B1	0.04–0.08	Mixed (B1/D2: cause group; A3/A2/C4: effect group)	Intermediate transmission role; affected by high-impact factors while influencing low-impact factors
	D2
	C4
	A3
	A2
Low	D1	<0.04	All belong to the “effect group” (d − r < 0)	Passive response role; mainly affected by other factors; weak direct driving effect on talent return
	C1
	C3
	C2

.

5. Discussion

Based on the empirical results of the Grey DANP model (Section 4.2) and the three-level classification of criteria (Section 4.3), this section focuses on discussing the driving mechanisms of high-impact criteria (the six core factors) and their synergistic effects, while briefly commenting on medium- and low-impact factors to ensure the focus and depth of the discussion. Ranked by weight priority, the high-impact criteria are: development level of science and technology (E₁), economic level (A₁), educational level (B₄), attention to science and technology (E₂), public service (B₂), and talent policy (B₃). All key factors have positive relation values (d − r) (Table 9) and belong to the “cause group”, indicating their active driving role in talent return decisions. Significant interactive linkage effects exist among these factors (Figure 6).

5.1. Core Engine: The Leading Role of Science and Technology Development Level

The development level of science and technology (E₁) has the highest weight (0.109, Table 10) and a prominent degree of d + r = 5.7127 (Table 9), verifying H2 and indicating its core hub position in the talent return driving system. As a key concern of knowledge-intensive talents, it directly determines a region’s capacity to absorb and accommodate high-end talents. Internationally competitive scientific research infrastructure, cutting-edge R&D projects, and a sound industry–university–research collaborative innovation ecosystem provide a core platform for overseas talents to exert their professional capabilities and meet their career development needs. From the perspective of causal interaction, the development level of science and technology (E₁) has a significant positive impact on talent policy (B₃) and public service (B₂) (Table 8). Regional demand for scientific and technological development forces talent policies to upgrade towards precision and high-end orientation. It also encourages public services to optimize in line with the needs of high-end talents. This result confirms that in the knowledge economy era, scientific and technological strength has become the core competitiveness in attracting overseas high-end talents. It also explains the continuous rise in the talent return rate in high-tech industry agglomerations in China in recent years.

5.2. Basic Support: The Enabling Value of Economic Level

Economic level (A₁, weight = 0.107) is the material foundation for the core driving factor E1, with a total influence value of 0.323 on E₁ (Table 8), forming a two-way empowerment cycle. On one hand, the improvement of economic level can directly increase R&D funding, strengthening the hard power of the development level of science and technology (E₁). On the other hand, sufficient financial support can optimize the allocation of educational resources (B₄) and improve the quality of public services (B₂), providing high-quality living and development guarantees for talents [50]. The total influence value of economic level (A₁) on E₁ reaches 0.323 and 0.313 on B₄, confirming the positive cycle of “economic foundation → core resource agglomeration”. Notably, economic level and the development level of science and technology form two-way empowerment: scientific and technological development spawns new industries and new drivers, feeding back economic growth. Together, they constitute the primary attraction engine for talent return.

5.3. Key Support: The Dual Guarantee Function of Educational Level

Ranked third, educational level (B₄) has a positive relation value (d − r = 0.2889), highlighting its active driving role. Its support for talent return is reflected in two aspects. In terms of career development, high-quality higher education resources provide core channels for academic exchanges, industry connections, and R&D cooperation, helping talents enhance professional capabilities and achieve career breakthroughs [17]. In terms of living security, a sound basic education system effectively addresses the worries of talents about their children’s education, reducing family decision-making costs for return. Data in Table 8 shows that the total influence value of educational level (B₄) on both public service (B₂) and talent policy (B₃) is 0.197. This indicates that the improvement of educational level can encourage public services to optimize towards meeting the needs of talent families. It also provides a basis for the precise design of talent policies, acting as a key bridge between career development and living security.

5.4. Institutional Guarantee: The Synergistic Effect of Attention to Science and Technology, Public Services, and Talent Policies

As key driving factors, attention to science and technology (E₂), public service (B₂), and talent policy (B₃) rank fourth to sixth in sequence. Together, they form the institutional and environmental guarantee system for talent return. This result also verifies H1, confirming that the driving factors of overseas talent return form an interdependent synergistic system rather than acting independently. (1) Attention to science and technology (E₂, d − r = 0.1495) conveys the government’s long-term commitment to innovative development. This clear institutional orientation alleviates talents’ uncertainty regarding future development prospects, enhancing their confidence in returning. It also directly influences the design direction of talent policies and the allocation focus of public services. (2) As the basic guarantee for life and work, public service (B₂, d − r = 0.4118) directly affects the adaptation efficiency and living experience of talents after return. The improvement of medical care, transportation, and government services is an important support factor for retaining talents. Its positive interaction with economic level and educational level forms a virtuous cycle of “resource input → service upgrading → talent agglomeration”. (3) With the highest relation value, talent policy (B₃, d − r = 0.9944) serves as an important “transmission hub” among key driving factors. Targeted policies such as salary subsidies, housing support, and research start-up funds can directly reduce the institutional costs of talent return, offsetting the insufficient attractiveness of regional development gaps [52]. Its linkage with the development level of science and technology and economic level ensures policy precision and effectiveness.

Notably, factors related to home country connections (D₁ and D₂) rank 8th and 12th in weight (Table 10), with weak direct driving effects—contrary to traditional views that cultural attachment is the primary emotional driver [30,36]. This reflects the “value-driven” decision-making logic of modern high-end talents, who prioritize professional value realization and innovation ecosystem matching over emotional factors. Emotional ties only play a supplementary synergistic role in reducing post-return adaptation costs.

6. Conclusions and Remarks

6.1. Conclusions

How to effectively attract overseas talents to return and serve their home countries is an urgent issue for developing countries, including China. This study constructs a five-aspect and 15-criterion indicator system for overseas talent return via a systematic literature review and two-round Delphi method and adopts the Grey DANP model to quantify the factor weights and interaction relationships. Empirical results identify the six core driving factors ranked by weight as: E₁ > A₁ > B₄ > E₂ > B₂ > B₃.

The core research findings are as follows: (1) The development level of science and technology is the primary driving factor for overseas talent return. Both its prominence and relation rank among the top. This confirms the key demand of high-end talents for hard power such as research platforms and innovation ecosystems in the knowledge economy era. It differs from the conclusion of some previous studies that “economic factors are the primary driving force”, reflecting the transformation of China’s talent return demand from “survival-oriented” to “development-oriented”. (2) Economic level still plays a fundamental supporting role. As the material guarantee for scientific research and development, educational investment, and public service upgrading, it forms a two-way empowerment cycle with the development level of science and technology. Together, they constitute the core attraction engine for talent return. (3) Educational level, public service, talent policy, and attention to science and technology serve as key supporting factors. They correspond to the core needs of talents in career development, living security, institutional adaptation, and development expectations respectively, jointly forming a comprehensive attraction system of “hard power + soft environment”. (4) Factors related to connections with the home country (family relationships D₁ and attachment to the motherland D₂) and living environment (climate C₁ and natural environments C₂, etc.) have low weights (all ranking after 10th place). Their direct driving effect on talent return is weak, which is consistent with the decision-making logic of high-end talents who pay more attention to development opportunities and professional value realization.

6.2. Theoretical Contribution

Firstly, this study breaks the single-factor analysis paradigm of talent return drivers. Existing studies mostly focus on the independent impact of individual factors such as economy, policy, or culture [8,12,30], ignoring the interdependent and feedback relationships among multi-dimensional factors. This study constructs a five-aspect and 15-criterion indicator system through a systematic literature review and Delphi method and quantifies the synergistic interaction effects among factors using the Grey DANP model. It reveals the causal transmission paths among economic level, scientific and technological development, educational resources, and other factors, thus perfecting the analytical framework of the multi-factor synergy mechanism for talent return.

Secondly, it enriches the methodological system of global talent mobility research. Traditional studies on talent return mainly adopt descriptive statistics or questionnaire-based regression models, which are unsuitable for analyzing such complex multi-criteria decision-making problems and fail to depict the nonlinear interaction and feedback mechanisms among multiple factors. This study is the first to introduce the Grey DANP hybrid model into this field. Leveraging the strength of grey system theory in solving problems with “clear extension and unclear connotation”, it effectively addresses the uncertainty and ambiguity in expert evaluations, providing a new quantitative analytical tool for revealing the interaction mechanism of talent mobility drivers.

Thirdly, it updates the theoretical cognition of talent return motivation in the knowledge economy era. Previous studies often regarded economic factors as the primary driving force for talent return [9,25] or emphasized cultural and emotional attachment as the core emotional driver [30,36]. However, this study confirms that the level of scientific and technological development has become the most critical pull factor for overseas high-end talent return, and the core motivation of talent return has shifted from “emotion-driven” to “value-driven”. This finding reflects the transformation of talent return demand in developing countries from “survival-oriented” to “development-oriented”, enriching the empirical evidence of global reverse brain drain theory and talent competition theory.

6.3. Practical Contributions

(1)

For developing countries: This study identifies the six core driving factors of overseas talent return and their synergistic relationship and provides a scientific decision-making framework for developing countries to formulate targeted talent introduction policies and optimize the talent development ecosystem under the background of global talent competition.

(2)

For China: Combined with the latest talent return data (2001–2024), the study clarifies the core demands of high-end overseas talents for a scientific and technological innovation environment, economic development and educational resources, which can provide a practical reference for China to further refine talent policies, strengthen the integration of the talent chain and industrial chain, and enhance the sustainable attractiveness of overseas talents.

(3)

For regional talent governance: The research reveals the multi-dimensional synergy mechanism of talent return drivers and suggests that regional talent governance should break through the single-policy orientation and construct a comprehensive talent attraction system of “hard power (science and technology, and economy) + soft environment (policy and public service)”, which provides a practical path for regional talent agglomeration in developing countries.

6.4. Research Limitations and Future Study

In future studies, the following limitations and potential research directions need to be further explored and improved. Firstly, this study took China as a typical sample of developing countries to explore talent return drivers, and the conclusions may be affected by China’s unique institutional background and cultural context. There is a lack of cross-country comparative analysis with other developing countries or developed countries, making it difficult to reveal the commonalities and differences in talent return mechanisms under different national contexts. Secondly, the study constructed a five-dimensional and 15-criterion indicator system, but some potential influencing factors were not fully covered. For example, the impact of international geopolitical changes, cross-border talent flow policies of other countries, and the psychological adaptation costs of returnees were not included in the indicator system. Additionally, the measurement of factors such as “scientific and technological development level” and “public service quality” mainly relied on subjective evaluations, lacking the support of objective statistical data for cross-validation. Thirdly, this study also has a limitation in sample information collection. Specifically, to protect respondents’ anonymity, reduce information disclosure concerns and improve the questionnaire response rate during the design phase, we only collected sample information based on the three pre-defined categories of overseas talents and did not collect comprehensive demographic characteristics including age, gender, specific work field and detailed education level. Future research will optimize the questionnaire design, collect comprehensive demographic information on the premise of ensuring respondents’ privacy, and conduct multi-group heterogeneity tests to further verify and improve the universality of the conclusions.