2.1 Bourdieu’s Field Theory

Bourdieu’s Field Theory is a sociological framework for understanding human behavior. It divides society into two objective structures: “primary” and “secondary.” Primary objective structure refers to the means and methods of social resource distribution and utilization, while secondary objective structure refers to the concrete thought patterns and behavioral methods of practical activities [27]. Field Theory specifically comprises three dimensions: field, capital, and habitus. “Field” refers to an “objective network” among various positions [28,29]; “Capital” denotes accumulated foundations within social fields [30]; “Habitus” shares a symbiotic relationship with fields—it originates from fields and manifests as dispositions and behavioral orientations [30,31].

In the field of higher education, Bourdieu’s Field Theory has been widely applied to uncover the power structures and capital competition both within universities and between universities and their broader social environments. Universities are not only spaces for knowledge production and dissemination but also complex fields constituted by the interaction of multiple forms of capital—academic, social, cultural, and economic [32,33]. Scholars employing Field Theory have analyzed academic hierarchies, disciplinary divisions, resource allocation, and evaluation systems within higher education, emphasizing that academic status often stems from unequal capital accumulation [34–36]. Such inequalities are continuously reproduced through habitus. For instance, the academic habitus cultivated by faculty members through research activities influences their scholarly orientations and teaching practices, thereby shaping the institutional academic style and the overall pattern of knowledge production [37].

Moreover, Field Theory has been instrumental in examining the tensions between globalization and localization in higher education [38,39]. As international competition intensifies, universities face increasing pressure to compete for prestige and resources in the global academic field. This process reshapes institutional habitus and capital logic. Some studies suggest that in the pursuit of global rankings and research performance, universities have developed a “logic of academic capital dominance,” which may compromise educational equity, academic autonomy, and the broader mission of talent development [40]. From a field perspective, researchers have also explored how education policies are formulated and implemented differently across institutional contexts: the positions and capital structures of policymakers, institutions, and academic actors determine their interpretations and practices of policy objectives [41,42].

Building on its extensive application in higher education research, Field Theory also provides a valuable analytical framework for understanding the mechanisms of talent cultivation [43]. Within this framework, key actors in the STT cultivation field include universities (core actors), governments, enterprises, and research institutes. Capital elements encompass human resources (faculty) and educational/R&D capital investments. Habitus refers to concrete measures implemented for cultivating STT, including achievement transformation programs, international academic conferences, and collaborative research. While these significantly influence regional STT cultivation, specific impact pathways remain unclear. Therefore, using Field Theory as the theoretical framework allows us to better capture the relational and multi-actor dynamics underlying STT cultivation, particularly the ways in which multiple institutional conditions interact rather than operate independently.

2.2 Research perspectives on provincial innovation

Against the backdrop of rapidly evolving global innovation landscapes, regions have become increasingly important functional units within national innovation systems. As a meso-level entity possessing administrative authority, fiscal arrangements, and policy autonomy, the “province” is not only the terminal point of policy implementation but also a frontier for institutional innovation and scientific and technological deployment. PI is widely regarded as an intermediary mechanism connecting national strategies with local resources. Its innovation capability is crucial for enhancing national technological self-reliance and building regional economic resilience [44,45].

Traditional Regional Innovation System (RIS) theory emphasizes synergistic effects among local institutional environments, innovation networks, and knowledge flows [46]. Within this framework, PI systems are seen as outcomes of interactions within the industry-university-government triple helix structure, operating through geographical proximity, trust mechanisms, and institutional embeddedness [47,48]. Recently, Innovation Ecosystem theory further highlights the dynamic evolutionary nature of local innovation, emphasizing the interactive structures and strategic adaptability of multiple actors within complex networks [49,50]. This perspective contends that local innovation capacity depends not only on resource allocation but also on the joint support of institutional evolution, organizational learning, and relational capital [51].

Compared to cities or metropolitan areas, provinces possess not only cross-city resource integration capabilities but also stronger policy tools and governance mobilization capacities, making them crucial spatial units for understanding Chinese-style innovation governance and institutional adaptability. Studying PI helps identify spatial heterogeneity within China’s innovation ecosystems and reveals interactive logics among institutional configurations, technology diffusion, and talent mechanisms.

2.3 Talent cultivation and provincial innovation

Talent is a core driver of regional innovation systems, especially amid growing trends of “deglobalization” and “technological localization,” which heighten local dependence on indigenous STT [52]. As primary cultivators of STT, universities are undergoing profound institutional repositioning within PI systems [53]. They are no longer passive knowledge suppliers but increasingly transforming into embedded innovation actors, assuming key roles in coordinating networks, integrating resources, and supporting local innovation systems [54].

A large body of empirical research has provided a solid foundation for understanding the role of universities within regional innovation systems. Overall, existing studies have evolved through three main stages. Early research primarily focused on the spatial agglomeration effect of talent, revealing that the concentration of STT significantly promotes regional knowledge creation and technological innovation, thereby enhancing the absorptive capacity and economic competitiveness of RI [55,56]. As research deepened, scholars began to emphasize the intermediary and hub role of universities [57,58]. Empirical findings indicate that universities serve as core actors in knowledge diffusion and technology transfer within RI networks. Through activities such as research collaboration, university–industry partnerships, technology commercialization, and innovation-oriented education, universities have effectively driven industrial upgrading and innovation synergy. The accumulation of educational and research capital is regarded as a crucial driver of sustained regional innovation growth [59–62]. More recently, studies have further highlighted the structural impact of talent cultivation mechanisms, suggesting that universities, by training, attracting, and reproducing STT, are actively reshaping the knowledge structure and institutional environment of RI. Factors such as educational resource allocation, disciplinary development, research culture, and international collaboration together constitute the core supporting mechanisms of RI [20,21].

Although extensive research has yielded substantial findings in the fields of RI and talent cultivation, scholars remain divided on the mechanisms through which universities exert their influence within innovation systems. Some studies, adopting an institutional-environment perspective, emphasize the decisive role of local government policies, innovation ecosystem construction, and institutional arrangements in shaping the effectiveness of talent, arguing that universities’ innovative performance largely depends on external institutional incentives and resource allocation efficiency [22,23,63]. Others, from an organizational-mechanism perspective, contend that internal factors, such as research culture, academic habitus, resource distribution structures, and governance models constitute the fundamental determinants of universities’ talent innovation capacity and their contribution to regional collaboration [64,65]. In addition, debates over talent agglomeration versus mobility effects have become another focal point: the traditional view highlights how the spatial concentration of STT promotes knowledge spillovers and technological innovation, whereas more recent studies argue that cross-regional mobility, network collaboration, and open knowledge sharing can, under certain conditions, exert even stronger effects on innovation performance [56,66,67]. Overall, while existing research has revealed statistical correlations between talent and innovation, it still lacks an in-depth explanation of how universities, as institutional actors, systematically drive RI capacity through talent cultivation mechanisms within the multidimensional fields of institutions, organizations, and networks.

The above divergence largely stems from limitations in research perspectives and analytical frameworks. On the one hand, many studies tend to simplify RI performance as being driven by a single or a small number of factors—such as institutional environment, internal university mechanisms, or talent agglomeration and mobility—while overlooking the interdependence and mutual embeddedness of these factors [68,69]. This reductionist tendency makes it difficult to explain why the role of universities in RI appears “highly significant in some cases but limited in others.” On the other hand, mainstream empirical approaches are often built upon linear and additive-effect assumptions, treating institutional variables, organizational characteristics, and talent indicators as independent explanatory factors, with little consideration of potential nonlinear effects, threshold effects, or multiple equivalent pathways (equifinality) that may arise from different configurations [25,70]. Under such frameworks, research can identify which factors matter but rarely explain under what contextual and configurational conditions these factors actually interact to produce innovation outcomes [71]. As a result, scholarly understanding of how universities function within RI has remained fragmented and static.

3.1 Research framework

PI capability is a critical link for driving new-old kinetic energy conversion and achieving high-quality local economic development. STT, as core resources for knowledge production and technological application, play a driving role in regional innovation systems [53]. Talent agglomeration not only stimulates regional innovation vitality but also helps build localized innovation networks and institutional foundations. Therefore, enhancing university talent cultivation quality and establishing institutionalized talent support mechanisms have become practical pathways for localities to boost innovation capacity.

According to endogenous growth theory, universities serve as vital nodes for knowledge accumulation and technology diffusion, undertaking dual functions of STT education and research innovation. Bourdieu’s Field Theory further provides an analytical framework, emphasizing that actors within specific fields are dually influenced by capital structures and habitus systems, constructing behavioral boundaries through capital operations and cognitive actions [27]. In the regional innovation context, university talent cultivation mechanisms represent crucial processes of capital conversion and habitus perpetuation within the “education field.”

Additionally, the innovation ecosystem perspective highlights multi-actor interactions in local innovation, typically analyzing the “industry-university-government” triple helix structure [47], emphasizing synergistic relationships among universities, governments, and industries. Accordingly, this paper constructs a theoretical model of how STT cultivation influences PI through the “field-capital-habitus” three-dimensional structure.

Specifically in field level: indicators, Industry-University-Research Collaboration (IUCR) and Government Attention to Scientific Talent (GAT), are selected, measuring institutional linkages between universities and external organizations; Capital level: Human Capital (HC) and Educational and Research Capital (ERC) are selected to reflect universities’ resource allocation capabilities in talent cultivation; Habitus level: International Scientific and Technological Exchange (ISE) and Transformation of Scientific and Technological Achievement (TSA) are selected to measure internalization and output capacities of university research activities toward local knowledge systems.

These six variables constitute the configuration of conditions for the analysis, allowing for an exploration of the equifinal mechanisms through which university talent cultivation influences PI performance, as shown in Fig 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Model of configurational effects.

https://doi.org/10.1371/journal.pone.0344374.g001

3.2 Research method

This study adopts the fuzzy-set Qualitative Comparative Analysis (fsQCA) as its primary research tool. Proposed by sociologist [43], the fsQCA method integrates the case-oriented sensitivity of qualitative research with the systematic variable handling of quantitative approaches. This approach effectively identifies the mechanisms through which configurations of multiple conditions drive the outcome, making it particularly suited for meso-level analyses that account for the interdependent effects of various institutional and social factors.

The core idea of QCA lies in treating social phenomena as configurations composed of multiple conditions. By applying set theory and Boolean algebra, QCA identifies the “sufficiency” and “necessity” of various condition combinations in relation to the outcome variable. Currently, QCA has evolved into three types: crisp-set QCA (csQCA), multi-value QCA (mvQCA), and fuzzy-set QCA (fsQCA). Among these, fsQCA allows variables to have membership scores between 0 and 1, thus more accurately reflecting the “partial membership” nature of social variables. This feature gives fsQCA a significant advantage in dealing with gradual variables such as regional institutions, educational investment, and collaborative relationships.

The influence of university-based STT cultivation on PI capacity is characterized by high complexity and diverse pathways. Nonlinear coupling and complementary mechanisms often exist among different conditions. The fsQCA method is well suited to address such a structure, where “multiple conjunctural causation” leads to the same outcome. It can identify various configurational paths leading to high levels of innovation capacity. This study follows the standard fsQCA procedure, including: (1) direct and indirect calibration of variables; (2) necessity analysis (typically based on a consistency threshold of 0.9); (3) generation of the truth table and extraction of complex, parsimonious, and intermediate solutions; and (4) robustness testing and interpretation. This method helps overcome the limitations of traditional linear models and more accurately reveals the multiple pathways through which university talent mechanisms empower PI under different institutional configurations.

3.3 Data sources

To ensure both broad representativeness and structural comparability in the fsQCA analysis, the QCA approach emphasizes sample coverage and heterogeneity during case selection (Ragin, 2008). Although the cultivation of STT in universities exerts a generally shared influence on PI capacity, it is also shaped by diverse historical, political, social, and economic factors across regions, resulting in significant regional disparities. Based on this, the present study selects China’s 31 provincial-level administrative units (including autonomous regions and municipalities directly under the central government) as cases for configurational comparative analysis at the provincial level (see supporting documents).

The core data used in this study are derived from authoritative statistical sources, including the China Regional Innovation Capacity Evaluation Report, the Compilation of Science and Technology Statistics of Higher Education Institutions, and provincial Government Work Reports. To ensure the timeliness of data and account for potential time lags in the effect of different indicators, this study adopts the three-year average (2021–2023) of each variable as the final measurement value, following common practices in relevant literature. This approach enhances the stability and temporal validity of the variables.

Specifically, the outcome variable “PI” is sourced from the China Regional Innovation Capacity Evaluation Report. This composite indicator integrates five dimensions: knowledge creation, knowledge acquisition, enterprise innovation, innovation environment, and innovation performance, providing a comprehensive assessment of each province’s innovation capability. Regarding condition variables, IUCR and ERC are also taken from the same report. The former reflects the degree of collaborative innovation networks at the local level, while th e latter measures the capital investment capacity of universities in the talent cultivation process.

The variable GAT, indicating the prioritization of science and technology talent in local policy agendas, is quantified through textual analysis of provincial Government Work Reports from 2021 to 2023. Specifically, the frequency of keywords related to “S&T talent” is used to extract the degree of policy attention. This text-based approach has been widely used to quantify governmental policy attention in official reports [72]. As it captures agenda priority rather than implementation outcomes, any random measurement noise is more likely to weaken the observed associations (i.e., yield conservative estimates) than to generate spurious patterns, therefore is unlikely to alter our main conclusions [73]. Meanwhile, data related to university-based talent—such as HC, ISE, and TSA—are sourced from the Compilation of Science and Technology Statistics of Higher Education Institutions. These indicators cover the number of research personnel, the frequency of international collaboration, and the effectiveness of research commercialization, offering a systematic representation of the resource structure and external connectivity of university talent cultivation mechanisms. It should be acknowledged that the indicators used to measure STT are based on aggregate university-level statistics, which include both locally cultivated and externally recruited personnel.

4.1 Measurement and calibration

Drawing upon Bourdieu’s field theory, the cultivation of STT in universities is not merely a singular educational act but a systemic mechanism shaped by the collaboration of multiple actors—industry, academia, government, and research sectors—alongside capital input and academic cultural production. To identify the causal configurations through which university-based talent cultivation mechanisms influence PI, this study takes PI capacity as the outcome variable and selects six condition variables grounded in Bourdieu’s field theory. Drawing on Seppo & Lilles (2012) and Meng et al. (2024), this study selects IURC and GAT as field-level indicators to capture the structural configuration of the innovation field at the provincial level—respectively reflecting inter-organizational relational patterns and policy orientation [72,74]. Drawing on Okoye et al. (2022), this study selects HC and ERC as capital-level indicators to reflect the availability and mobilization of human and financial resources for talent cultivation [75]. Given Bourdieu’s habitus primarily denotes individual-level internalized dispositions reproduced through practice, it is difficult to measure directly in a province-level study using secondary data and may raise concerns about concept–measurement inconsistency. To address this issue, we draw on educational research that extends the notion of habitus to an institutional/organizational level (institutional habitus), conceptualizing it as stabilized, observable practice patterns embedded in universities’ routines and organizational arrangements [76,77]. Therefore, drawing on Byrd (2018), we operationalize habitus using two indicators that reflect enduring practice orientations: ISE and TSA [78].

Calibration in fsQCA follows the direct method proposed by Ragin (2008), which requires the specification of three qualitative anchors to distinguish full membership, the crossover point, and full non-membership in a set. In the absence of theoretically grounded or substantively agreed-upon absolute thresholds for provincial-level indicators, prior methodological guidelines and applications of configurational comparative methods recommend the use of percentile-based anchors as a pragmatic and theoretically coherent solution [79–81]. Specifically, the 95th percentile of each variable is set as the threshold for full membership (1.0), the 50th percentile as the crossover point (0.5), and the 5th percentile as the threshold for full non-membership (0.0). This approach transforms continuous variables into fuzzy-set membership values.

To avoid ambiguity caused by cases at the 0.5 crossover point, a small constant (0.01) is uniformly added to all cases with a membership score of 0.5. This adjustment enhances the discriminative power of the model. The study covers 31 provincial-level units in China, ensuring both theoretical comprehensiveness and compliance with fsQCA’s requirement for case heterogeneity. The original values, quantile thresholds, and calibration settings of each variable are presented in Table 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Variable description and calibration analysis table.

https://doi.org/10.1371/journal.pone.0344374.t001

4.2Necessity analysis

A configurational analysis was conducted on the conditions of university-based STT cultivation and their influence on PI. As a first step, it is essential to examine whether any of the individual conditions constitute necessary conditions. The consistency threshold for identifying necessary conditions ranges from [0, 1], with a commonly accepted benchmark of 0.900. A condition is considered necessary if its consistency score exceeds this threshold and the corresponding scatter plot shows a concentration on the right-hand side or if more than one-third of the data points lie on the horizontal line at 1.0.

As shown in Table 2 and illustrated by the scatter plots in Fig 2, HC, ERC, and ISE emerge as necessary conditions for achieving high levels of PI. In contrast, under conditions of non-high PI, although the absence of ERC exceeds the 0.900 reference value, its scatter plot does not exhibit the required pattern of right-side concentration or alignment along the horizontal line for more than one-third of cases. Therefore, it does not qualify as a necessary condition.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Necessity analysis table for individual condition variable.

https://doi.org/10.1371/journal.pone.0344374.t002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. XY plot.

https://doi.org/10.1371/journal.pone.0344374.g002

In light of these findings, further analysis is required to explore whether there are additional conditions that, while not individually necessary, may jointly contribute to high innovation outcomes. It is also important to examine the synergistic effects of various factors in the context of non-high innovation performance across provinces.

4.3Configurational analysis

Configurational analysis reveals the multiple causal pathways formed by combinations of conditions. Following existing studies [82], and given the small-N nature of this study, the frequency threshold was set to 1, the consistency threshold to 0.800, and the Proportional Reduction in Inconsistency (PRI) threshold to 0.700. Based on the output format suggested by [46], as shown in Table 3, the analysis yielded two configurations associated with high-level PI (Table 4), and four configurations associated with non-high-level PI. Each configuration represents a possible causal path.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Grouping for the realization of high level of PI/non-high level of PI.

https://doi.org/10.1371/journal.pone.0344374.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Driving paths and development strategies for high-level PI.

https://doi.org/10.1371/journal.pone.0344374.t004

The solution consistency for high-level PI is 0.907, indicating that the two configurations account for 90.7% of the cases demonstrating high innovation performance. The solution coverage is 0.843, meaning that these two configurations explain 84.3% of all high-innovation cases. Through the configurational analysis, the study identifies the diverse roles of different conditions in achieving high levels of PI and demonstrates strong explanatory power.

4.4Analysis of high-level provincial innovation pathways

1. S1: Technology Commercialization-Oriented Talent Cultivation Pathway.

In this configuration, TSA and ISE serve as the core habitual elements, forming the main innovation-driven pathway alongside ERC. Although there is a lack of effective platforms for IUCR and insufficient policy support, universities leverage abundant resource input and an endogenous research orientation to embed TSA mechanisms throughout the entire talent cultivation process. This enables a closed-loop educational model of “from classroom to laboratory to market.” Crucially, the interaction among TSA, ISE, and ERC forms a mutually reinforcing mechanism: TSA relies on ERC to transform research outputs into practical innovations, while ISE expands the cognitive and cooperative boundaries that enhance the utilization efficiency of ERC and stimulate further transformation activities. Through this feedback cycle, resources, international exchange, and technological achievement transformation continuously strengthen one another, forming a self-propelling system that drives high-level provincial innovation. This pathway highlights the synergistic mechanism between capital input and internationalization, whose essence lies in institutional self-propulsion realized through habitual reinforcement.

Beijing and Hubei represent this pathway. Beijing, in particular, has established a systematic framework for technology commercialization within its universities, effectively institutionalizing the local incubation of research outcomes. It also promotes cross-border mobility of talent and research resources through “Belt and Road” science and technology cooperation. Hubei, despite limited resource coordination, accelerates the flow of research resources into regional industrial systems through policy guidance and laboratory system construction, thereby embedding a results-oriented habitus into its policies. This configuration shows that even in the absence of favorable structural conditions, universities with strong habitual capacity and resource accumulation can still drive a leap in PI capability.

1. S2: Government-Led Talent Cultivation Pathway.

This configuration centers on the GAT, using policy design and resource allocation to compensate for structural deficiencies in IUCR. In contexts where habitual elements are present but collaborative mechanisms are lacking; the government becomes the most dynamic institutional actor in the field. Through specialized talent programs and the provision of financial and institutional resources, GAT interacts synergistically with ERC to enhance universities’ internal capabilities. The government’s intervention not only mobilizes universities’ latent talent supply capacity but also embeds policy incentives into their organizational routines, gradually transforming external policy pressure into internalized institutional habitus. In this way, a policy–resource–habitus interaction mechanism emerges, through which state-led initiatives bridge the structural gap between the educational and regional innovation fields and collectively generate provincial innovation outcomes.

Sichuan and Anhui are typical cases. Facing challenges such as a shortage of high-end talent and mismatched talent structures, Sichuan introduced policies like the “Cuiqing Project” to form a closed-loop system of talent cultivation through institutional selection and targeted funding. Anhui, guided by its “Strengthening the Province through Talent” strategy, restructures university talent training systems around industrial chains and grants universities greater autonomy in talent evaluation and achievement transformation. This configuration illustrates that in regions with immature structural fields, institutional reconstruction driven by the government can become a key external force in upgrading university-based technological talent systems.

Overall, the fsQCA results reveal that multiple configurations can lead to high provincial innovation performance. Although the dominant conditions vary, with some emphasizing technology achievement transformation and others relying on government-led coordination, they converge toward the same outcome. This reflects the principle of equifinality in configurational analysis, demonstrating that distinct combinations of field, capital, and habitus can operate as alternative but equally effective routes to innovation success. The findings thus highlight that both endogenous, university-driven mechanisms and exogenous, policy-driven mechanisms can serve as complementary pathways through which universities embed themselves within PI.

4.5 Analysis of innovation pathways in non-high-level provinces

1. NS1: “Capital–Habitus Deficiency” Pathway

This configuration reflects a typical dilemma of dual deficiencies in both resources and habitus. In the absence of sufficient ERC, HC, and support for scientific activities, even a certain degree of structural connectivity within the field (e.g., IUCR or GAT) is insufficient to drive high-level PI. This pathway indicates that universities lack the institutional energy to output effectively, talent systems are weak, and the formation of academic habitus is constrained, making it difficult to sustain innovation networks.

Guangxi exemplifies this pathway. Constrained by its economic development level and geographical location, local universities have limited capacity for research investment and scientific output. The system for cultivating STT remains underdeveloped. Although local governments have attempted to introduce resources through reforms such as “joint construction by province and ministry,” the lack of academic habitus and capital foundations has rendered these policies largely ineffective in fostering genuine innovation.

1. NS2: Capital Deficiency Pathway

In this configuration, while there are strong GAT and some willingness to collaborate, the severe lack of ERC and HC input means that the university training system lacks the necessary “energy input” to complete the knowledge production and transformation cycle. Policy focus alone cannot compensate for the material void, and university-industry-research collaboration fails to operate effectively. As a result, the talent mechanisms in universities cannot support PI.

Gansu is a representative case. With a weak economic foundation and limited resource accumulation in universities, the localization of talent and transformation of scientific achievements face multiple obstacles. Although the government has issued measures aimed at optimizing the allocation of scientific resources, the endogenous capability of the research system remains low, and institutional drivers initiated by government intervention have struggled to take root effectively.

1. NS3: Habitus Deficiency Pathway

This pathway presents a scenario of “policy and investment in place, but lacking culture.” Although there is a certain level of ICUR and GAT for talent, universities are weak in TSA and ISE. The immature habitus system prevents research activities from being integrated into local innovation ecosystems. The absence of an educational habitus undermines the application and transformation of knowledge, resulting in a fragmented innovation chain.

Hebei serves as a typical example. Although it has introduced programs to cultivate high-skilled talent and emphasized the central role of universities, a longstanding focus on traditional manufacturing has limited motivation for research transformation. Local universities struggle to establish a habitual pattern of international scientific cooperation, leading to knowledge isolation and talent outflow, which significantly constrains the release of regional innovation capacity.

1. NS4: “Field–Habitus Insufficiency” Pathway

This configuration shows that even if universities have capital support and a foundation for external collaboration, the absence of GAT and the weakness of achievement transformation mechanisms hinder the activation of key institutional dynamics. Without clear goals or incentives, universities as talent cultivation fields struggle to drive breakthroughs in PI.

Heilongjiang is a representative province. Despite a reasonable foundation in research investment and collaboration, the lack of a systemic governmental strategy for STT means that universities cannot effectively embed their research outcomes into the local economy. The transformation mechanisms remain underdeveloped, and university-industry-government collaboration is limited to superficial interactions. As a result, high capital investment fails to translate into high innovation output.

4.6 Robustness test

As a set-theoretic approach, fsQCA emphasizes configurational causation, and its results are generally expected to be stable to reasonable variations in key analytical choices [83,84]. Following common practice, we assess robustness by adjusting two aspects that may affect solution identification: the PRI consistency threshold and the calibration anchors. Specifically, we first increase the PRI threshold to 0.75 to test whether the identified configurations depend on a more permissive rule for handling contradictory configurations. Under this stricter PRI setting, the resulting configurations remain exactly the same as in the baseline analysis, suggesting that the main solutions are not driven by borderline contradictory cases.

We then examine robustness to alternative calibration stringency by changing the three qualitative anchors used in the direct method. In addition to the baseline anchors (Calibration B: 95th/50th/5th percentiles), we apply a less stringent set of anchors (Calibration A: 75th/50th/25th percentiles). As shown in Table 5, the two dominant configurations for high PI reappear under both calibration schemes, with only modest changes in consistency and coverage values. The solution consistency remains high (above 0.90), and overall coverage remains comparable across specifications. A new configuration emerges under the less stringent calibration; however, its coverage is small, indicating a peripheral pattern that does not materially affect the study’s main conclusions. Overall, the configurational relationships remain largely stable across these parameter perturbations, supporting the robustness of our findings.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Robustness check under alternative calibration anchors.

https://doi.org/10.1371/journal.pone.0344374.t005

5.1 Research conclusions

This study is based on Bourdieu’s field theory, and from the three-dimensional structure of “field–capital–habitus,” it constructs a theoretical model of how the mechanism for cultivating STT in universities affects PI capacity. Using fsQCA, configuration identification was conducted on relevant data from 31 provinces in China. The study finds that China’s PI capacity generally presents a pattern of both spatial agglomeration and regional disparity. Innovation highlands are mainly distributed in coastal regions such as Beijing-Tianjin-Hebei, the Yangtze River Delta, and the Pearl River Delta, forming an initial “belt-like” innovation distribution pattern. However, there is still a significant imbalance in the allocation of innovation resources, and coordinated governance mechanisms urgently need to be strengthened.

At the level of necessity analysis, HC, ERC, and ISE are identified as fundamental conditions for high-level PI. This indicates that high-quality cultivation of STT depends on stable resource input and sound knowledge connectivity mechanisms. Furthermore, through fsQCA, two typical high-level innovation paths are identified: one is the “achievement-driven” path led by technology commercialization and international cooperation, which emphasizes a results-oriented habitus in the internal research and education mechanisms of universities; the other is the “policy-guided” path with government attention to STT at its core, highlighting the stimulating effect of the policy field on the innovation capacity of universities. These two paths respectively demonstrate how the cultivation of STT can be embedded into the PI system under conditions of habitus advantage or institutional traction.

Meanwhile, the study also identifies four typical non-high-level PI paths, including the “capital–habitus” dual-deficiency type, capital-deficient type, habitus-deficient type, and “field–habitus” insufficient type. These paths all reflect that the absence of a single institutional element cannot be completely substituted by other conditions, indicating that high-level innovation relies on the coordinated configuration of multiple institutional elements. Notably, across all configurations leading to high-level innovation, at least one element of the technical acquisition set—GAT or TSA—emerges as a consistently present condition, alongside the identified necessary factors. This shows that the empowering effect of the talent mechanism in universities is not realized automatically but requires institutional embedding and alignment with cultural habitus to promote the organic linkage of knowledge, resources, and institutions.

In the last, this study not only expands the theoretical understanding of the relationship between talent mechanisms and regional innovation, but also, through configuration analysis, reveals multiple paths at the provincial level, providing strong support for formulating localized talent policies and institutional design.

5.2 Research implications

5.3 Research limitations

This study still has certain limitations. First, the data used in this study cover a three-year period and include both internally cultivated and externally recruited personnel. As such, the indicators reflect the overall talent capacity of universities rather than the specific effects of internal cultivation. However, this relatively short time span may still limit the ability to capture the long-term dynamics of talent cultivation. Future research could incorporate longitudinal or micro-level data to better distinguish between cultivation and recruitment mechanisms and to further explore their differentiated impacts on provincial innovation performance. Second, some key constructs are measured with proxy indicators from secondary data, which may raise construct-validity concerns. For example, text-based measures from official reports capture issue salience in policy discourse rather than policy implementation or outcomes, and may be affected by rhetorical emphasis. Future research could incorporate more direct measures of policy implementation, such as funding allocation records or program-level participation data, to complement text-based indicators of policy attention. Finally, the contextual specificity of this study may limit the generalizability of its findings. The results are most applicable to countries or regions with governance systems similar to China’s—where universities are closely integrated into government-led innovation frameworks. In more market-driven or decentralized contexts, the mechanisms identified here may operate differently. Future research could therefore conduct comparative analyses to test the adaptability of these pathways across diverse institutional settings.