From the perspective of policy system construction, a newly introduced policy should seamlessly integrate into the existing policy framework. Such integration is critical for the continuous improvement of the policy system, facilitating a comprehensive structure that aligns with the objectives of talent development. Isolated or singular policies often lack effectiveness, as they fail to generate the spillover effects that arise from the interconnections and complementarities within a cohesive policy framework. The talent policy system is inherently complex and difficult to fully assess, exhibiting the characteristics of a gray system. To address this complexity, this study employs a system coupling model based on gray relational analysis to evaluate the degree of integration and interconnectedness within the talent policy system.

Taking 2014 as the base year for the policy system, and considering that all collected policies remain currently effective, the policy system for a given year is composed of the policies introduced in that year in conjunction with the existing framework. From the perspective of overall system integrity, a lower coupling degree indicates better policy embeddedness. In other words, well-integrated policies enhance the overall cohesion of the system, whereas a higher coupling degree reflects poor embeddedness and a weaker fit within the policy framework.

Given that the original data dimensions differ, we first perform a dimensionless treatment on the data. Let $X_{1}, X_{2}, \ldots X_{i}(i=1,2, \ldots, n)$ be the sub-tool of the talent policy instruments, with its value in year t denoted as $x_{i}(t)(t=1,2, \ldots, T)$. The standardized data is represented as $x_{i}^{\prime}(t)=\frac{x_{i}(t)-\min _{t} x_{i}(t)}{\max _{t} x_{i}(t)-\min _{t} x_{i}(t)}$. Let $Y_{1}, Y_{2}, \ldots Y_{k}(k=1,2, \ldots, p)$ be the primary sub-goal of the talent objectives, where Y_kj denotes the secondary sub-goal j under the primary sub-goal k, and the value of the secondary sub-goal in year t is $y_{k j}(t)(t=1,2, \ldots, T)$, with its standardized data represented as $y_{k j}^{\prime}(t)=\frac{y_{k j}(t)-\min _{t} y_{k j}(t)}{\max _{t} y_{k j}(t)-\min _{t} y_{k j}(t)}$. The correlation coefficient between the policy sub-tool X_i and the talent objective Y_kj in year t is given by:

where ρ∈(0,1) typically takes a value of 0.5. A stronger correlation indicates a higher contribution of the policy tool to the talent objective. The contribution of X_i to Y_kj in year t can be expressed as:

Given the differences in contribution among various policy subsystems to different talent objectives, a smaller variance indicates a better coupling effect. Therefore, the coupling degree of the policy subsystem to the talent sub-goal can be represented as:

where $\lambda_{k}(t)=\sqrt{\left(R_{1}^{k}(t)\right)^{2}+\left(R_{2}^{k}(t)\right)^{2}+\cdots\left(R_{i}^{k}(t)\right)^{2}}$.

The degree of policy compensatory fit indicates the extent to which the “supply” of policies enjoyed by talent aligns with their demand.

where $\cos \theta_{s d}(t)=\frac{\sum_{k=1}^{p} y_{k}^{\prime}(t)}{\sqrt{\left(\sum_{l}^{N} P T_{l}^{\prime}(t)\right)^{2}+\left(\sum_{k=1}^{p} y_{k}^{\prime}(t)\right)^{2}}}$, PT_l′(t) represents the standardized score of policy tools for year t with the standardization method being the same as that for x_i(t). $\sum_{l}^{N} P T_{l}^{\prime}(t)c$ indicates the policy supply situation in year t, while $ \sum_{k=1}^{p} y_{k}^{\prime}(t)$ reflects the talent demand situation in year t.

Let Ψ_sd be the environmental variable. When the policy supply exceeds the talent needs, it is recorded as -1, indicating that θ_sd>45°. Conversely, when the policy support is less than or equal to the talent needs, it is recorded as 1, indicating that θ_sd≤45°.

A higher value of consistency fit and compensatory fit for science and technology talent policies indicates a greater degree of fit, while a lower value of embeddedness fit reflects better integration. Accordingly, the final fit score for science and technology talent policies is calculated by subtracting the embeddedness fit value from 1, and then summing this adjusted value with the consistency fit and compensatory fit values.

The policy texts analyzed in this study were sourced from publicly available science and technology talent policies in China issued between 2014 and 2023. The following selection criteria were applied during the data collection process: First, only policies explicitly addressing talent-related issues were included, while documents with partial or peripheral discussions on talent were excluded. Second, only policies applicable at the national level were collected, excluding those targeting specific regions. For instance, policies such as the “Notice from the Ministry of Finance and the State Taxation Administration on Personal Income Tax Policies for High-end and Scarce Talent in the Hainan Free Trade Port” were not included. Third, the selection was limited to normative administrative documents, work plans, implementation schemes, and similar policy types. Fourth, outdated policies were excluded to ensure that only currently valid policy documents were analyzed, with the most recent versions of policies being used. Based on these criteria, a total of 283 valid policy documents were obtained for analysis.

Using the TF-IDF algorithm to extract keywords from policy documents and analyzing them with VOSviewer software, this study reveals that China’s talent policies from 2014 to 2023 primarily emphasize talent cultivation, training, evaluation, and the graduation and entrepreneurship of university students. The key government departments responsible for issuing these talent policies include the Ministry of Education, the Ministry of Finance, the Ministry of Human Resources and Social Security, the Ministry of Industry and Information Technology, and the Ministry of Science and Technology.

In terms of the evolution of policy fit, the overall fit degree of China’s science and technology talent policies has exhibited an upward trend from 2014 to 2023, increasing from 12.42 in 2014 to 16.66 in 2023. However, the year 2018 marked a trough in policy fit despite the highest number of science and technology talent policies being issued that year. A retrospective analysis of the policy content indicates that, while 2018 witnessed significant attention to the development of science and technology talent—with numerous policies introduced covering areas such as talent selection, funding, rewards, training, the establishment of effective talent management mechanisms, and improvements to talent compensation—these policies faced challenges in quality. Specifically, the prevalence of macro-level guiding opinions and insufficient feasibility for practical implementation led to a lower consistency fit score, ultimately affecting the overall policy fit score. Post-2018, although the number of talent policies issued in China declined, the fit degree displayed a consistent upward trend (Figure 3). This trend is also consistent with China’s actual conditions in different periods. Before 2018, the size of China’s scientific and technological talent pool was not large, so the focus of talent policies was on “expanding the quantity”. The 4th National Survey on the Status of Science and Technology Workers released by the China Association for Science and Technology in 2018 showed that 45.9% of science and technology workers believed that the problem of irrational orientation in scientific and technological evaluation was prominent. In response, multiple government departments began to intensively introduce policies aimed at “breaking the four-onlys” (over-reliance on academic papers, professional titles, academic qualifications, and awards). However, the initial policies were mainly principle-based requirements.

The consistency of policy fit has increased year by year, evolving from universality to targeted breakthroughs; however, there are still gaps in the application of policy tools. From 2014 to 2023, the consistency fit of China’s science and technology talent policies has shown a continuous upward trend, increasing from 11.71 points in 2014 to 15.64 points in 2023 (Figure 4). This reflects a growing alignment between policy tools and the intended goals of science and technology talent policies, demonstrating ongoing improvements in policy quality. However, gaps in the application of certain policy tools remain evident. The use of policy tools has varied across different periods, highlighting the evolving focus of talent development in China. Between 2014 and 2018, the most frequently employed policy tools were those related to talent cultivation and development. In contrast, tools for talent introduction and aggregation, as well as development planning, were less frequently utilized. During this period, China primarily implemented cultivation plans and specific training programs to enhance the overall technical capabilities of its talent pool. From 2019 to 2021, policies related to evaluation and incentives surpassed those focused-on cultivation and development, becoming the primary instruments in China’s science and technology talent strategy. Key measures during this period included providing financial support for talent, improving talent evaluation systems, reforming the professional title system, and selecting outstanding talent. These efforts aimed to address systemic issues in talent evaluation and enhance the selection and motivation of exceptional talent, marking a shift from the relatively universal talent policies of the previous period to a more targeted focus on high-performing individuals.

Post-2022, the application of policy tools became more balanced, with notable increases in management and service policies, as well as policies related to mobility and allocation, compared to the previous two periods. Policies issued during this time were more detailed and specific. Despite these advancements, policies focused on talent introduction, aggregation, and development planning remained underutilized compared to other tools. In terms of policy objectives, little variation is observed across different periods, with a consistent focus on supporting talent and improving talent quality. This reflects the sustained priority of these goals within China’s science and technology talent policy framework.

The fit of policy embedding has shown fluctuations, and the policy system is dynamically improving; however, there are still some areas where goal coupling is suboptimal. As shown in Figure 5, the embedding fit of China’s science and technology talent policies exhibits significant fluctuations across different years. Using 2014 as the baseline year, with its value set at 0, the evolution of embedding fit from 2015 to 2022 reveals a dynamic pattern of both increases and decreases. Notably, between 2017 and 2019, policy embedding showed a relatively poor performance, following an M-shaped trajectory. The decline in embedding fit during this period can be attributed to the continuation of certain policies from 2017 into 2018, which led to stronger integration within the existing framework. For instance, the international selection and training policy for high-end accounting talent and the policy for leading talent in the management of small and medium-sized enterprises were directly inherited from 2017’s framework, resulting in improved policy embedding. In contrast, the introduction of major reforms to talent evaluation systems and mechanisms in 2019 diverged significantly from the existing policy framework, leading to a reduction in embedding fit for that year. From 2021 to 2023, the embedding of policies has shown a notable increase, suggesting that China’s science and technology talent policy system has undergone continuous dynamic improvement between 2014 and 2021. During this period, the system actively explored and refined its policy mechanisms for talent, resulting in fluctuations in embedding fit. After 2021, the policy system reached a more mature and stable state. However, regarding specific objectives, the degree of goal coupling within the talent policy system varies significantly. Overall, coupling is strongest in terms of talent effectiveness, followed by talent support, and then talent mobility. By contrast, the coupling related to talent scale and talent quality remains suboptimal. Specifically, while China’s science and technology talent policies have been effective in driving local economic and social development through talent-related initiatives, their impact on expanding the talent pool and enhancing individual abilities has been relatively limited. Furthermore, an analysis of policy consistency reveals that these policies predominantly aim to enhance talent quality and provide talent support. However, a gap exists between the intended goals of the policies and their actual outcomes, particularly regarding talent scale and quality. Additionally, the degree of goal coupling is influenced by the differential contributions of various policy tools. From the perspective of policy contributions to talent development goals, introduction and aggregation policies and development planning policies make relatively smaller contributions compared to other tools. This underscores the need for the introduction of more robust and impactful policies to address these gaps effectively.

The degree of policy compensatory fit has increased year by year, with supply and demand matching becoming more consistent; however, the policy supply still cannot meet talent demand. From 2014 to 2023, the compensatory fit value of China’s science and technology talent policies has steadily increased, rising from 0.707 in 2014 to 0.999 in 2023 (Table 6). This trend indicates that, after a decade of evolution, the alignment between policy supply and talent demand has generally improved, achieving a relatively high degree of matching. For example, in 2022, eight departments including the Ministry of Science and Technology launched a pilot program for reforming talent evaluation, designating 6 regional entities and 21 universities and research institutions, including Tsinghua University, as pilot sites. Tsinghua University abolished quantitative indicators for academic papers and shifted to evaluating “representative achievements.” This initiative supported Chen Haodong’s team in overcoming the bottleneck in plant gravity sensing research, with the relevant results published in the journal Cell. However, throughout this period, the environmental variable values have consistently remained at 1, signaling that the current supply of talent policies is still insufficient to fully meet talent demand. This finding underscores the need for continued enhancement of policy support for science and technology talent. In terms of the matching degree between policy supply and specific talent demands, there is minimal variation across different talent dimensions. The compensatory fit of China’s science and technology talent policies for talent scale, talent quality, talent effectiveness, talent mobility, and talent support remains relatively stable, fluctuating around 0.75. When analyzed by policy tool type, evaluation and incentive policies and training and development policies exhibit the highest degree of alignment with talent demand. These are followed by management and service policies, while introduction and aggregation policies and development planning policies show lower degrees of fit. These findings highlight areas where greater emphasis on targeted policy adjustments and enhancements is needed to address the evolving demands of science and technology talent.