Introduction

“Double first-class” is the abbreviation of world-class universities and first-class disciplines. By 2022, a total of 147 institutions have been designated for this prestigious title, with 20 located in the six provinces of central China. The six central provinces include Anhui, Jiangxi, Henan, Hubei, Hunan and Shanxi, which are collectively known as the Central China. Graduates from these esteemed universities play a vital role in China’s talent market. Their employment choices not only align with their personal interests but also significantly impact regional economic development and the creation of innovative areas1. Since 2017, a number of cities have introduced new policies on talent, gradually setting off a climax of the war for talent2. Among them, “double first-class” universities graduates are the main talent targets that governments and enterprises around the world compete for.

In recent years, the number of college graduates in China has continued to be at a high level, and the employment situation faced by graduates has become more and more severe. To ensure a smooth and orderly transition into the workforce for graduates, various levels of government in China have implemented a range of policies and initiatives3,4,5. Research has examined structural and regional disparities in graduate employment, as well as the effectiveness of related policies. For ‘Double First-Class’ universities, graduate employment outcomes and location choices are critical to advancing China’s higher education and regional economies6,7,8.

The employment mobility of college graduates is a critical indicator of regional talent allocation and socioeconomic development. Graduates from ‘Double First-Class’ universities, as China’s elite human capital, play a pivotal role in driving innovation and economic growth. However, central China, despite its strategic geographic position and dense higher education resources, faces challenges in retaining talent compared to the eastern coastal regions. This study focuses on central China’s ‘Double First-Class’ universities to unravel the spatial dynamics and driving forces behind graduate mobility.

The dynamic mechanisms of graduate employment flow are shaped by a complex interplay of economic disparities, policy incentives, and social networks9. For instance, regional innovation capacity and wage differentials act as ‘pull’ factors, while local enrollment policies and alumni networks create ‘push’ effects. Additionally, the COVID-19 pandemic introduced new dynamics, such as reduced interprovincial mobility and increased preference for local employment. These mechanisms collectively drive the hierarchical and path-dependent patterns observed in graduate mobility, underscoring the need for a spatially explicit analysis to inform policy interventions.

Existing research focuses on graduate mobility patterns, drivers, and impacts. For example, Cui Dan et al. developed an urban suitability model to analyze talent agglomeration factors10. They examined the spatio-temporal patterns and influencing mechanisms of innovative talent agglomeration in China. Li Qin et al. used 2017 mobile population data and spatial analysis tools to study young talent distribution and retention factors11. Wang Yifan et al. concluded, through the lens of labor migration theory, that economic factors and urban comfort are the primary drivers of talent mobility12. In terms of research area and scale, Cui et al. focused on college graduates in the Yangtze River Delta region, adopting a “national-regional” perspective to analyze the spatial characteristics and driving factors of talent mobility13. Methodologically, existing studies often employ cartogram maps, spatial autocorrelation, and social network analysis methods, selecting indicators such as sticking rate and agglomeration degree for spatial analysis. Additionally, models like the Geodetector, Spatial Econometric Interaction Model, and Directed Transfer Model are used to explore the driving factors behind talent mobility14,15,16. While existing studies provide foundational insights into talent mobility, they predominantly focus on economic factors and urban comfort, often overlooking institutional constraints such as China’s household registration (hukou) system and university admission quotas. For instance, Li Qin et al. (2020) emphasized economic drivers but did not account for how centralized enrollment policies shape student mobility. This gap underscores the need to integrate institutional frameworks into analyses of graduate migration patterns. Additionally, prior research has largely focused on the Yangtze River Delta and Pearl River Delta regions, leaving central China underrepresented in the literature. This study aims to fill this gap by examining the unique mobility patterns of graduates from ‘Double First-Class’ universities in central China, which are influenced by both regional economic disparities and national policy frameworks.

The mobility of central China’s ‘Double First-Class’ graduates remains understudied17. There is a need to further strengthen or optimize the understanding of their mobility network patterns and driving factors. Inspired by the concept of “proximity” in evolutionary economic geography, Menzel introduced the ideas of “dynamic proximity” and “spatial dynamics brought about by changes in the spatial movement of actors”18,19,20,21. These concepts provide new perspectives to explain the formation of network “bridges” and suggest that talent mobility is a key driver in the creation of inter-city collaborative innovation networks. Therefore, by analyzing mobility patterns of central China’s ‘Double First-Class’ graduates through a GeoDetector framework, this study identifies key location-choice drivers, offering actionable insights for regional talent strategy formulation. It can support the development of employment policies for graduates from “Double First-Class” universities and offer strategies for retaining and attracting talent to foster high-quality economic development in the Central China22,23,24.

Data and methods

Theoretical logical framework construction

Graduate mobility creates a complex network linking origin, study, and employment cities, shaped by differences in urban characteristics. Drawing on social network and migration theories, this study develops a diamond framework model (Fig. 1) to analyze this network.

Fig. 1
figure 1

Theoretical logical diamond framework model.

Full size image

First, the model captures the triangular relationships among origin, study, and employment cities. Graduates, as carriers of advanced knowledge, drive inter-city knowledge flows and innovation through their mobility. They move from hometowns to universities, forming origin-study connections, and later to workplaces, creating study-employment links. These city nodes—representing key stages in graduates’ trajectories—serve as relational hubs for urban development and network evolution. Thus, the triangular framework characterizes the interconnected mobility patterns of graduates from ‘Double First-Class’ universities.

Second, we examine the drivers of graduate mobility. In the source network, students’ movement from hometowns to study cities is shaped by enrollment programs, city location, university reputation, and career opportunities. These factors critically influence their initial choices. In the employment network, graduates’ movement from study cities to workplaces is driven by economic factors (e.g., income, benefits), city location, and local environment. Workplace decisions also reflect personal preferences, hometown ties, and peer networks, often leading to clustering in regions with strong social connections. These employment networks significantly shape urban spatial structures. Focusing on employment drivers, this study uses the GeoDetector model to analyze four categories—economy, environment, infrastructure, and innovation—and 22 specific factors influencing graduate mobility.

Research methods

Graduate mobility network centrality indicators

The attractiveness of each province and city to graduates is quantified as a sticking rate and net sticking rate indicator. The stickiness rate reflects the ability of the place of employment to stick to graduates, and the expression for the stickiness rate \(F\) is:

$$F = \frac{{Q_{i} }}{P}$$
(1)

where: \(Q_{i}\) represents the number of graduates of the “double first-class” universities in the Central China who went to place \(i\) for employment, and \(P\) represents the number of graduates employed in the Central China. This formula applies to all provinces, regardless of their geographic classification. The net sticking rate further refines this measure by accounting for the proportion of local graduates retained in the province, providing a more nuanced understanding of a region’s ability to attract and retain talent. This methodology allows for a comparative analysis of both central and non-central provinces, highlighting the competitive dynamics between regions.

The net stickiness rate reflects the ability of the place of employment to retain local graduates while sticking to graduates from different places of origin, and the net stickiness rate is expressed as:

$$J = F - \frac{{C_{x} }}{S}$$
(2)

where: \(C_{x}\) represents the number of students from birthplace \(x\) who came to the Central China for employment, and \(S\) represents the total enrollment in the Central China.

Spatial autocorrelation

In space, the existence of correlation between a certain spatial unit and other spatial units around it with respect to a certain attribute in the spatial unit is called spatial autocorrelation. In order to explore the spatial characteristics of talent flow in the Central China, this paper will use the provincial location and the number of talents flowing into each province to measure the spatial autocorrelation of talent flow in the Central China by using Global Moran’s \(I\), which is calculated by the formula25:

$${\text{Moran's }}\;I = \frac{{\mathop \sum \nolimits_{i = 1}^{n} \mathop \sum \nolimits_{j = 1}^{n} w_{ij} \left( {x_{i} - \overline{x}} \right)\left( {x_{j} - \overline{x}} \right)}}{{s^{2} \mathop \sum \nolimits_{i = 1}^{n} \mathop \sum \nolimits_{j = 1}^{n} w_{ij} }}$$
(3)

where I is roughly in the interval [− 1, 1]. \({\text{i}}\), \({\text{j}}\) is the polygon number and wij is the spatial connectivity matrix between i, j. n is the number of provinces, municipalities, and autonomous regions, is the mean value of viscosity rate in the study area, and xi, xj are the observed values of viscosity rate in areas i, j.

OD network construction

Graduates, as the social actors studied in the social network of this paper, have a certain coupling relationship between the network system of their place of origin and the network of their place of employment22. This paper uses OD network analysis method combined with spatial analysis method in order to deepen each other. Take the geometric center of each province, municipality directly under the central government, autonomous region (excluding Hong Kong, Macao and Taiwan) as a node, according to the number of enrollment plans of regional “double first-class” universities i in each province x, to construct the flow matrix from the place of origin to the place of study \(S_{x \to i}\), so as to get the location of the university y and the place of origin of the students x flow network \(G_{x \to y}\). Similarly, we can construct the network from the place of study y to the place of employment. The flow network \(P_{y \to z}\) from the place of study \(y\) to the place of employment \(z\), to construct two inter-provincial directed weighted migration networks with the size of 6 × 31 and 31 × 6.

$$G_{x \to y} = \mathop \sum \limits_{i = 1}^{i \in y} S_{x \to i}$$
(4)
$$P_{y \to z} = \mathop \sum \limits_{i = 1}^{i \in y} T_{i \to z}$$
(5)

Geo-detectors

A geodetector model26 is used to identify the influence of different factors on talent mobility. It is assumed that the study area consists of several sub-regions and spatial heterogeneity exists if the sum of the variances of the sub-regions is less than the total variance of the region. Geo-detectors can reveal the driving factors behind spatially stratified heterogeneity, overcoming the limitations of the number of sample units as well as the statistical assumptions of the variables.

The factor detection expression is:

$$q = 1 - \frac{{\mathop \sum \nolimits_{h = 1}^{L} N_{h} \sigma_{h}^{2} }}{{N\sigma^{2} }}$$
(6)

where: N and \(\sigma^{2}\) represent the number of units and sample variance of the study area provinces, municipalities and autonomous regions, respectively, \(N_{h}\) and \(\sigma_{h}^{2}\) are the sample capacity and intra-stratum variance of the sub-stratum of the hth sub-stratum, respectively, and L is the number of strata of the explanatory variable factor on the stickiness rate. The value of q is [0, 1], and the larger the value of q indicates that the influence factor has a stronger explanatory power on the stickiness rate.

TERGM model

The TERGM (Temporal Exponential Random Graph Model) is employed to explore and predict the spatial evolution of graduate flows. TERGM treats the network across multiple periods as a single entity, capturing the structural, temporal, and spatial dependencies in network evolution. It does this through indexes of endogenous structural variables, temporal dependence terms, node attribute variables, and network covariate variables, thus examining both the endogenous and exogenous mechanisms behind the network’s formation and evolution. In this paper, we analyze the endogenous and exogenous mechanisms driving the evolution of the spatial association network of college graduates from 2018 to 2020. If the spatial association network of graduate mobility is considered an observational sample over time, the TERGM is configured as follows:

$$P\left( {G = g^{t} |g^{t - k} } \right), \ldots ,g^{t - 1} ,\theta ) = \frac{{\exp \left( {\theta^{T} h\left( {g^{t} ,g^{t - 1} , \ldots ,g^{t - k} } \right)} \right)}}{{c\left( {\theta ,g^{t - k} \ldots ,g^{t - 1} } \right)}}$$
(7)

where P is the probability that the sample network g is in all possible occurrences of network G; \(\theta\) is the vector of coefficients of the model to be evaluated, and \(\theta^{T}\) is the \(\theta\) transpose of \(\theta\); \(h\left( {g^{t} ,g^{t - 1} , \ldots ,g^{t - k} } \right)\) is the vector of statistics that contains the endogenous and exogenous mechanism variables of the network, where k must completely contain the time dependence of \(g^{t}\); and c is the normalization constant, which serves to ensure that the model computes values on 0 to 1.

Data sources

This study focuses on undergraduate and postgraduate graduates from public ‘Double First-Class’ universities in central China’s six provinces: Hubei, Hunan, Anhui, Jiangxi, Shanxi, and Henan. These institutions, listed in the Ministry of Education’s ‘National List of Ordinary Colleges and Universities,’ are full-time undergraduate schools. Data on graduate mobility were collected from the 2018–2020 Graduate Employment Quality Reports published on university websites (e.g., Wuhan University: https://info.whu.edu.cn/info/2046/15193.htm). Note that some universities discontinued these reports after 2021. The reports detail graduates’ origin and employment distributions, covering 19 ‘Double First-Class’ universities (excluding the University of National Defense Technology). This study examines mobility from hometowns to study cities and from study cities to first-job locations. Employment data include labor contracts, agreements, targeted training, state employment, flexible employment, and local programs, excluding non-employment categories such as further education or overseas destinations.

All data pertain exclusively to graduates from ‘Double First-Class’ universities in central China. Findings should not be generalized to non-‘Double First-Class’ institutions. The focus on ‘Double First-Class’ universities is justified by their unique role in China’s higher education system, as they are recognized for their academic excellence and research output. These institutions attract top students and are often the primary targets of regional talent policies, making their graduates a critical cohort for studying mobility patterns.

Based on the Push-Pull Theory of population migration, this study utilized the economic-environmental-infrastructure-innovation evaluation framework to construct a comprehensive indicator system that assesses the driving factors behind graduate mobility. By selecting 22 determinants across four subsystems—economy, environment, infrastructure, and innovation (Table 1)—the study provides a nuanced analysis of the factors influencing graduate movement. Data for these 22 determinants for each province, city, and autonomous region were primarily sourced from the China Statistical Yearbook and the China Regional Economic Statistics Yearbook from 2018 to 2020. The 22 determinants were selected based on the push and pull theories of talent mobility and validated through a review of previous studies on talent mobility. Indicators such as tertiary value added and R&D investment were prioritized due to their correlation with graduate retention. In addition, the selection process included two stages of validation: first, a preliminary list of indicators was compiled based on theoretical relevance; second, final indicators were selected based on data availability and statistical significance of the pilot analysis. This rigorous selection process ensured that the selected determinants comprehensively reflected the economic, environmental, infrastructural and innovation-related factors affecting graduate mobility.

Table 1 Driving factor index system.
Full size table

Analysis of results

Characteristics of student source and employment networks

Basic statistical characteristics

The number and percentage of students of “double first-class” universities in central China and the employment situation are analyzed in this statistical analysis, and the statistical results are shown in Table 2.

Table 2 Graduate enrollment and employment statistics table.
Full size table

Table 2 highlights variations in enrollment, employment, and location preferences among graduates from central China’s ‘Double First-Class’ universities. Henan, Jiangxi, and Shanxi provinces prioritize local recruitment, with over 50% of students originating locally. Locally employed graduates exceed 30%, led by Henan at 66.10%, reflecting its strong local ties. Within provinces, graduates cluster in provincial capitals and metropolitan areas, such as Wuhan and Xiangyang in Hubei. Conversely, those seeking external employment migrate to economic hubs like Beijing, Shanghai, and Guangzhou, or coastal provinces such as Zhejiang and Jiangsu.

China’s university admission system allocates enrollment quotas provincially, compelling students to prioritize institutional prestige over geographic preferences. For example, Henan Province recruits 74.8% of students locally (Table 2), partly due to its large population and limited quota allocations for out-of-province universities. This quota system creates a path-dependent mobility pattern, where students from provinces with fewer top-tier universities are more likely to attend local institutions. Thus, while path dependence and meritocratic connections influence mobility, institutional constraints like admission policies play a pivotal role in shaping student source networks. Furthermore, the Gaokao score, which determines university admission, often limits students’ ability to choose their preferred study location, as they must select institutions where their scores are competitive. This systemic constraint reinforces the concentration of local students in central China’s ‘Double First-Class’ universities.

Inter-provincial mobility is overwhelmingly driven by disparities in economic opportunities. Guangdong, Shanghai, and Beijing attract 24.09% of graduates (Table 2) due to their robust job markets and higher wages, overshadowing environmental or infrastructural factors. The eastern coastal regions, with their advanced industrial bases and higher income levels, serve as magnets for graduates seeking better career prospects. For instance, Guangdong Province alone absorbed 15.86% of graduates in 2018, despite having fewer local student sources. This trend highlights the significant role of economic pull factors in shaping graduate mobility patterns. Moreover, the concentration of multinational corporations, tech hubs, and financial centers in these regions further enhances their attractiveness to highly educated talent.

Spatial patterns of mobility across provinces

Path dependence and meritocratic connection are crucial perspectives for understanding the spatial distribution of economic activities and talent mobility patterns. Path dependence underscores the influence of historical events, emphasizing how past decisions shape current outcomes. Meritocratic connectivity is an important mechanism for network evolution, emphasizing the Matthew effect, where new nodes will tend to establish connections with dominant nodes. From the perspective of meritocratic connection, talents are more inclined to choose to flow to cities with better economic development, public services, wage level and innovation environment. From the perspective of path dependence, the experience, knowledge and cognition that talents have accumulated about cities help them to build their own preferences for city choices. Both mechanisms emphasize cumulative advantage and self-reinforcement, with meritocratic connections focusing on the attributes of the target city and path dependence emphasizing the preferences of the talent itself, and the combination of the two mechanisms is more useful for understanding the network pattern of talent mobility.

According to the OD directed network matrix depicting student mobility from their hometowns to their study locations and subsequently to their employment destinations, the spatial patterns of employment and student origin from “double first-class” universities in central China are visualized in Figs. 2 and 3. Figures 2a and 3a specifically illustrate the inflow patterns of graduates employed from these universities to 31 provinces across China. Figure 2a reveals that graduates from the central China predominantly migrate to the eastern provinces, including Guangdong, Shanghai, Beijing, Zhejiang, Jiangsu, and Shandong, which collectively absorb 36.05% of these graduates. Notably, Guangdong attracts the largest proportion of graduates for employment, although its share has declined slightly over the past three years: 15.86% in 2018, 14.68% in 2019, and 13.66% in 2020. Shanghai, Beijing, and Zhejiang also emerge as highly attractive destinations, accounting for 4.68%, 4.69%, and 5.18% of total inflows over the three years, respectively, while fewer graduates opt for employment in other provinces. The career choices of college graduates are influenced by their place of origin, as well as characterized by path dependence and meritocratic principles. Path dependence suggests that graduates tend to seek employment locally or return to their hometowns after studying elsewhere for an extended period. Meritocratic connections, on the other hand, drive graduates to cities offering favorable urban environments and higher incomes. In terms of inter-provincial mobility, meritocratic connections exert significant influence, particularly evident in cities like Guangdong, Beijing, and Shanghai, which attract large numbers of graduates despite having fewer student sources locally. Figure 3a depicts the employment flow network of graduates, highlighting how the capital city of the central China serves as a hub, absorbing local graduates while also sending them to economically developed eastern coastal regions such as Beijing, Tianjin, Hebei, the Yangtze River Delta, and the Pearl River Delta.

Fig. 2
figure 2

The proportion of graduates in employment and students.

Full size image
Fig. 3
figure 3

Cross-provincial flow pattern of “double first-class” universities graduates in central China.

Full size image

Figures 2b and 3b show the pattern of student mobility of “double first-class” universities in the central China. Predominantly, students originate from within the region itself, with the six central provinces contributing 64.09% of these students. This is primarily due to these universities’ significant enrollments within their respective provinces, coupled with targeted recruitment efforts in neighboring provinces. For instance, Wuhan University enrolls approximately 4200 students annually from Hubei, accounting for 30% of its total intake that year, and admits over 1200 students from Henan annually. Supported by the Central and Western Region Enrollment Collaboration Program, provinces within the Central China maintain robust inter-provincial student exchanges, ensuring a substantial and stable flow of enrollments within the region. Beyond the central China, provinces such as Hebei, Guangdong, Guangxi, Zhejiang, Shandong, and Jiangsu constitute the primary sources of students for “double first-class” universities in the central China. Over the past 3 years, these external contributions have remained relatively consistent. This regional enrollment strategy results in a predominant reliance on local student sources within the central China, leading to a decreased intensity in student flows from the central provinces to surrounding areas. The intensity of student flow shows a decreasing state from the central China to the surrounding provinces and cities, generally showing a swirling network layout with a dense center and sparse surrounding.

Spatial pattern of intra-provincial mobility

While eastern provinces like Jiangsu and Zhejiang have denser economic networks, central China hosts significant higher education resources. For instance, Hubei Province, home to 7 ‘Double First-Class’ universities, surpasses Guangdong and Fujian in institutional count. However, graduates still gravitate eastward due to better career prospects. This paradox highlights the disconnect between educational resources and economic opportunities in central China. Despite having a robust higher education system, the region struggles to retain talent due to the limited availability of high-paying jobs and advanced industries. This trend underscores the need for central China to align its educational strengths with economic development strategies to enhance its attractiveness to graduates. For instance, the proportions of college graduates opting to stay in their respective provinces and cities for employment in Hubei, Hunan, Henan, Anhui, Shanxi, and Jiangxi were 33.37%, 36.55%, 66.10%, 44.05%, 37.59%, and 41.01% respectively. There is a big gap with the average proportion of college graduates in East China who can be employed in the provincial area of more than 60%.

To illustrate the hierarchical flow of graduates within the Central China’s municipalities, the distribution of graduates from colleges and universities across various municipalities within the provinces was analyzed.

As depicted in Fig. 4a, Wuhan, Changsha, Hefei, Zhengzhou, Kaifeng, and Taiyuan emerge as focal points for intra-provincial mobility of graduates within the central China. Notably, Zhengzhou in Henan stands out with over 60% of graduates opting to stay within the province, particularly concentrated in the cities of Zhengzhou and Kaifeng. This reflects a distinct hierarchical distribution of employment opportunities within the province. Beyond the provincial capitals mentioned, their surrounding cities also present significant employment prospects. In Anhui, for example, with Hefei as the center, the ratio of the number of graduates flowing in from surrounding cities such as Wuhu, Lu’an and Fuyang all exceeded 2.5%. Similarly, in Hunan, Changsha acts as a major hub, with cities such as Zhuzhou, Yueyang, Changde, and Huaihua also drawing considerable numbers of graduates. Henan, however, exhibits a unique pattern: Zhengzhou, as the provincial capital, attracts a substantial 38.91% of graduates, while Kaifeng, its secondary center, also garners significant inflows at 19.79%. Together, these cities create a multi-core spatial pattern within Henan province.

Fig. 4
figure 4

Flow pattern of graduates in the province.

Full size image

Figure 4b illustrates the network pattern of graduate flows from universities in the central China to cities within their respective provinces. Notably, all universities, except in Hunan where Xiangtan and Changsha host four “double first-class” institutions, are situated in provincial capital cities. For example, Xiangtan University has a large flow to Changsha City, while combined with Fig. 4a, staying in Xiangtan for employment accounts for a smaller percentage of the total, which shows that it exports less graduate flow to the surrounding cities, and that the capital city of the province exhibits a stronger polarization effect. Overall, the graduate flow within the Central China forms a spatial pattern radiating outward from provincial capitals. Intra-provincial mobility itself is based on the premise of path dependence compared to inter-provincial mobility, but there is also a meritocratic connection to cities as graduates choose to stay in the Central China for employment.

Path dependence is evident in the concentration of universities in provincial capitals, which historically serve as economic hubs with accumulated development over centuries. These cities offer well-established knowledge, perception, and experiences, making them the preferred employment destinations for local residents. Consequently, provincial capitals generally boast favorable geographical locations and wage levels, making them ideal choices for both local and non-local graduates seeking optimal career opportunities.

Analysis of graduates stickiness rate

Sticky rate analysis

Using the neighboring space weight matrix based on the Rook contiguity principle for the common boundaries of the 31 provinces, and conducting a spatial autocorrelation analysis of the viscosity rate of graduate inflow from the Central China to these areas, we obtained a global Moran’s I index of 0.27 and a Z score of 2.73, with a P value of < 0.01. This indicates a spatial correlation in the employment location stickiness rate of graduates, as confirmed by a one-tailed normal distribution hypothesis test. Table 3 presents the LISA cluster types for the stickiness rate in each province. Anhui, Jiangxi, and Hunan in the central China fall into the “high-high” cluster, with stickiness rates above 3.5%. Conversely, the three northeastern provinces and the northwestern region are in the “low-low” cluster, with stickiness rates below 1.7%. Fujian, on the right side of Central China, belongs to the “low-high” zone and has a lower inflow of graduates compared to other provinces in the region. Table 3 does not display any “high-low” clusters, suggesting a more concentrated and uniform graduate flow in the central China. Although Guangdong, Shanghai, and Beijing have high stickiness rates, their spatial aggregation effects are not significant, and their stickiness rates are spatially isolated.

Table 3 Spatial autocorrelation results.
Full size table

Analysis of net sticking rate

This paper calculates the net sticking rate of employment by province, defined as the difference between the percentage of graduates entering a province for employment and the percentage originating from that province. This rate reflects the attractiveness of a province for university graduates from the central China. A positive net sticking rate indicates that a province not only retains local graduates but also attracts additional talent, while a negative net sticking rate indicates a loss of graduates to other regions. Overall, Guangdong has the highest net sticking rate at 12.78%, followed by Shanghai and Beijing, with net sticking rates of 4.47% and 4.25%, respectively. Other provinces with relatively high net sticking rates include Zhejiang (2.86%), Jiangsu (1.91%), and Hubei (1.40%). Hubei is the only province in the central China with a high stickiness rate, while Hunan has a positive but low net sticking rate (0.26%). The other provinces in the central China are experiencing a brain drain, with Henan (− 9.08%), Shanxi (− 4.14%), Jiangxi (− 2.43%), and Anhui (− 1.50%) all having negative net sticking rates.

Based on the changes in the net sticking rate over the years 2018, 2019, and 2020, provinces within the central China exhibit varying trends. Henan, Hunan, and Anhui show a positive upward trend in their net sticking rates. Conversely, Hubei, Jiangxi, and Shanxi display a downward trend, indicating a weakening in their ability to attract and retain talent. For other provinces, first of all, East China has a high level of attractiveness to graduates from Central China while its net sticking rate also varies, such as Guangdong, Beijing, Shanghai, Tianjin, etc. The net sticking rate has slightly decreased, while the net sticking rate in Zhejiang, Jiangsu, Shandong, etc. has an upward trend. For the western provinces, the implementation of the Western Development Strategy and initiatives encouraging graduates to work in the west have provided some support to their net sticking rates. Notably, the Tibet Autonomous Region has a positive net sticking rate, and other western provinces have also attracted some graduates from the central China, though their net sticking rates remain relatively low. Despite policies aimed at revitalizing the old industrial zones, the three northeastern provinces continue to struggle with low net sticking rates, indicating a significant insufficiency in their attractiveness to graduates from the central China.

Comparing Table 3 with the LISA aggregation type table of stickiness rate reveals that the central China shows “high-high” aggregation in terms of stickiness rate. This is because the central China retains more graduates due to the large number of local students. However, when focusing on the net stickiness rate, the central China does not exhibit the same level of competitiveness despite its quantitative advantage (Fig. 5). The central China’s lower net stickiness rate is attributed to the economic development, extensive job markets, and higher salary levels in Guangdong, Shanghai, and Beijing. These regions attract graduates from across the country and have fewer local students, resulting in higher net stickiness rates. Consequently, according to formula (2), the Central China’s net stickiness rate is low and often negative, indicating a loss of graduates, while the eastern coastal region has positive and high net stickiness rates.

Fig. 5
figure 5

Annual change in net viscosity rate.

Full size image

Analysis of driving factors

Driving factors for graduate mobility

Based on the framework of economic-environmental-infrastructure-innovation evaluation system adopted in this paper, the economic factors are mainly reflected in the economic diversity, employment level, and the role played by the economic system in attracting graduates to employment, which may affect the level of the city society, including the city’s economic structure, the city’s economic efficiency, and the city’s economic innovation capacity. Therefore economic factors can be measured by the city’s economic strength, economic vitality, industrial structure, foreign exchange capacity; environmental factors are the spatial carrier of urban development, mainly including the regional natural environment and urban landscape green space, based on the natural process, which can provide a variety of services to improve the livability of the city becomes an important consideration for graduates in the choice of employment place. Therefore, the environmental factors to consider pollution control27, urban green space and emissions and other capabilities; infrastructure belongs to the built environment of the city, is to ensure the operation of human and urban environmental system is a key factor, is the city of the “meridian skeleton”. The quality of infrastructure determines whether the urban operation process is smooth, whether the social services provided are comprehensive and perfect, and the safety capacity of the city in times of crisis28. Therefore, the city’s supply and drainage capacity and internal and external connectivity are chosen to measure the infrastructure factors; finally, based on the “double first-class” universities graduates as the research object in this paper, the innovation factor is introduced as one of the driving factors for the city to attract graduates. The number of patents is chosen as an indicator of innovation capacity, R&D investment, the number of scientific and educational personnel and expenditure on scientific and educational funds are chosen as indicators of innovation vitality, and the number of institutions of higher education is chosen as an indicator of innovation carriers, which play a key role in the city’s innovation activities.

Detection and identification of driving factors of graduate mobility

The stickiness rates of 31 provinces showed significant positive spatial correlation (p < 0.01), enabling further analysis using the GeoDetector model (Table 4). Among 22 factors tested, 12 significantly influenced stickiness rates (p < 0.05), with five showing strong explanatory power (p < 0.01): tertiary industry value added (economic), wastewater discharge (environmental), broadband access ports (infrastructure), invention patents (innovation), and science-education expenditures (innovation). Where the number of asterisks represents the significance level of the influencing factor. Innovation factors, led by invention patents, tertiary industry value added, and R&D investment, were the most influential, followed by economic factors. Environmental and infrastructure factors, though less significant, still contributed to regional attractiveness.

Table 4 The result of factor detection.
Full size table

Changes in driving factor detection analysis

This paper compares the results of the three-year driving factor detection analysis for 2018, 2019, and 2020 (Fig. 6) and draws the following conclusions: Firstly, the primary driving factor has shifted from economic factors to innovation factors over time, with the influence of innovation gradually increasing. In 2018, the value added by the tertiary industry, under the economic factor category, had the greatest influence with a q-value of 0.6248. By 2020, R&D investment had become the most influential factor, with a q-value of 0.6642, surpassing the value added by the tertiary industry. Innovation factors are gradually becoming the main influences that dominate the employment flow of graduates, with invention patents and R&D investment becoming the tertiary quantitative indicators with the largest q-values in 2019 and 2020, respectively. This indicates that innovation has a significant attraction effect on talent in the Central China. Secondly, the attraction of economic factors to talent in the central China remains stable. Indicators related to economic factors consistently occupy most of the significant indicators, highlighting that economic factors are still key drivers of talent flow. In contrast, the significance of environmental and infrastructure factors has declined over the years, with the number of significant indicators decreasing from five to three. This suggests that the influence of environmental and infrastructure factors on graduates in the Central China is waning.

Fig. 6
figure 6

Annual change in factor probe results.

Full size image

Due to the COVID-19 outbreak at the end of 2019, there may have been an impact on the employment of 2020 graduates. In terms of employment rates, the proportion of employed graduates increased from 57.67% in 2018 to 59.37% in 2019. The 2019 graduates were employed and thus unaffected by the outbreak. However, the proportion of employed 2020 graduates dropped to 55.7%, even lower than in 2018.

In terms of employment mobility, the proportion of intra-regional employment increased compared to 2019 Central 2020, with graduates preferring local or intra-regional employment. Many high school graduates opted to further their studies and delay employment due to the epidemic. This trend is attributed to society being in a semi-closed state, restricting population mobility, and the closed management of schools, which reduced employment opportunities such as job fairs. Additionally, businesses faced operational crises and began laying off employees to cut costs and increase efficiency. In summary, the epidemic created significant obstacles to the employment of graduates, reducing the intensity of employment flows and limiting the employability of graduates.

Result analysis TERGM prediction results

Using significant indicators from factor detection, we applied the TERGM model for prediction. Goodness-of-fit tests (Fig. 7) showed strong model performance, with ROC and PR curves intersecting at the top-left, indicating accurate simulation of graduate mobility networks. Using 2018–2019 net retention rates to predict 2020 trends, we established edge covariates for key indicators (Table 5). Model 1 (baseline) examined independent covariate effects, while Model 2 incorporated time-memory effects. Results revealed that covariates gradually reduce network stability, reflecting natural edge turnover without significant degradation. Economic, environmental, and infrastructure factors are expected to further destabilize the network, while innovation factors—driven by hubs like China’s emerging innovation cities—positively enhance stability and talent attraction.

Fig. 7
figure 7

TERGM goodness-of-fit test for graduate mobility in “double first-class” universities.

Full size image
Table 5 Robustness test of the mobility network of graduates of “double first-class” universities.
Full size table

Conclusion and discussion

Conclusion

This paper examines college graduates as a unique form of human capital and innovation agents. Utilizing data from the Graduate Employment Quality Reports released by 19 “double first-class” universities in central China over three years (2018, 2019, and 2020), it reveals the spatial characteristics of graduate mobility at both national and regional levels. Employing social network analysis and spatial analysis methods, we illustrate the spatial dynamics of graduate mobility from these “double first-class” universities in central China. The relationship between graduate mobility and stickiness rates and city attributes was parsed using GeoDetector and the following conclusions were drawn:

  1. (1)

    Graduates from central China’s “Double First-Class” universities exhibit hierarchical, distance-decay employment patterns shaped by path dependence and merit-based urban advantages. 48.62% are employed within central China, with 62.65% concentrated in provincial capitals. In contrast, 40.31% migrate to eastern regions like Beijing, Shanghai, and Guangdong, drawn by higher salaries and job opportunities.

  2. (2)

    Talent competition reflects a tug-of-war between central and eastern China. Central China retains half of its graduates due to local ties, but net retention rates are low or negative (e.g., Hubei, Shanxi, Jiangxi). Only Hunan shows positive growth. Eastern regions, with stronger economic appeal, dominate talent attraction.

  3. (3)

    GeoDetector analysis highlights innovation (R&D, patents, higher education) and economic factors (retail sales, tertiary industry value, local finance) as key drivers of graduate retention. Over time, innovation has surpassed economic factors in influence, while environmental and infrastructure impacts have declined. Despite COVID-19’s disruption, regional attractiveness remained stable.

This study contributes to the literature in three key ways. First, it pioneers a network-based analysis of graduate mobility in central China, a region underrepresented in existing research. Second, it integrates the GeoDetector and TERGM models to simultaneously assess spatial heterogeneity and temporal evolution, offering a methodological advancement. Third, the findings reveal the rising dominance of innovation factors over traditional economic drivers, providing policymakers with actionable insights to recalibrate talent strategies in the post-pandemic era.

Discussion

China is transitioning from being a country abundant in human resources to one abundant in human capital. Accelerating the development of a modern workforce characterized by excellent quality, ample quantity, an optimized structure, and reasonable distribution is essential for realizing the talent dividend and promoting high-quality regional development. A well-educated and vocationally trained workforce forms the cornerstone of China’s efforts to harness this talent dividend. This paper focuses on the unique human resources of graduates from ‘double first-class’ universities, examining their mobility patterns and the factors influencing these patterns. The goal is to provide insights for the central China to establish an interactive and coordinated talent mobility mechanism with rational distribution and sharing. Such a mechanism would promote optimal talent allocation, thereby enhancing the region’s competitiveness and innovation capabilities.

The study found that the mobility of graduates from “double first-class” universities has formed a complex network among their places of origin, study, and employment. To attract and retain talent in the Central China, the following recommendations are proposed: (1) Attracting Talent: Focus on graduates from outside the province by establishing innovative practice positions and guiding their movement to local enterprises. This approach aims to draw in talent from their places of origin. (2) Retaining Talent: For graduates employed locally, prioritize employment security, housing security, and support for innovation and entrepreneurship, which are key concerns for talent retention. (3) Promoting University Collaboration: Actively encourage exchanges and interactions among universities and colleges within the Central China to enhance regional innovation vitality and attract innovative talents.

To improve the national distribution of talent in the Central China, it is essential to enhance mechanisms for talent flow among various regions. Areas with strong demand for talent and the capacity to accommodate them should increase investment in talent attraction to maximize benefits. Conversely, less developed regions with weaker capacity should focus on their regional division of labor, seeking appropriate models for talent mobility. This includes strategically integrating economic capacity, educational resources, and development policies to gradually enhance their attractiveness and competitiveness in the process of promoting structural change.