2.1. Diffusion of innovation theory

The Diffusion of Innovation Theory, developed by Everett Rogers in 1962, is a key framework that explains how innovations—including new ideas, products, or practices—spread within a social system over time. According to this theory, the diffusion of innovations is a social process involving four main elements: innovators, communication channels, time, and social systems. Innovators are the individuals who first adopt an innovation; they typically have higher risk tolerance and socio-economic status. Communication channels refer to how information about the innovation is transmitted, which can include mass media or interpersonal communication. The time dimension refers to the entire process from the first introduction of the innovation to its widespread adoption, typically passing through five stages: knowledge, persuasion, decision, implementation, and confirmation. Finally, the structure, norms, and values of the social system significantly influence the acceptance and spread of innovations. The Diffusion of Innovation Theory has been widely applied across various fields, including public health, education, and marketing, to understand how innovations are adopted and spread [11–13].

In the sports sector, the effectiveness of policies within the sports industry, such as health promotion initiatives or programs aimed at increasing access to sports facilities, is closely tied to public acceptance and the scale of government promotional efforts. These factors collectively determine the success of policy implementation and the broader adoption of innovations within the industry [14]. In China, local governments have effectively stimulated the growth of the sports industry through innovative policies, including tax incentives and targeted funding support. However, the diffusion rate and effectiveness of these policy innovations vary considerably across different regions. These differences are influenced by factors such as the degree of government support, levels of socio-economic development, and cultural context [15].

2.2. Policy innovation diffusion

The exploration of how policy innovations spread began in the United States during the 1960s, drawing from fields like communication studies, sociology, and geography. In recent decades, this area has undergone significant transformation, becoming an essential domain within policy sciences as both a phenomenon to observe and a theoretical framework for analyzing policy processes. Innovation diffusion can be categorized into three distinct phases: the initial phase (1969–1990) concentrated on theories based on single factors; the second phase (1990–2000) saw the emergence of integrated theories; while the third phase (2000-present) is marked by advancements in comprehensive models [16]. The current body of literature can be distilled into three primary themes:

2.3. Special funds for the development of the sports industry

While the term Special Funds for the Development of the Sports Industry is not widely recognized internationally, there is considerable research on how governments financially support the sports sector. The academic literature on this topic can be broadly categorized into two main areas: the use of tax policies to foster the development of the sports industry and the provision of direct financial subsidies.

3.1. Data collection

The data for this study were drawn from authoritative and reliable sources to ensure analytical rigor and precision in examining the diffusion of policy innovations related to the Special Funds for the Development of China’s Sports Industry. Primary data were sourced from the China Statistical Yearbooks, which provide comprehensive statistics on economic, social, and sector-specific variables across provinces, including public expenditures, demographic trends, GDP, and investments in the sports industry. Provincial government reports were analyzed to trace the allocation and utilization of the special funds, offering insights into regional policy adoption and implementation. Socioeconomic indicators from the National Bureau of Statistics of China ensured cross-regional consistency and comparability, while government announcements and policy documents provided valuable information on the timing, structure, and progression of the policy’s implementation.

The analysis covers 31 provincial-level administrative regions in mainland China, excluding Hong Kong, Macao, and Taiwan. The observation period extends from 2006 to 2022. The year 2006 provides a pre-adoption baseline, while the first recorded provincial adoption occurred in Beijing in 2007. The province–year data were organized as a discrete-time risk set: each province contributed annual observations until the year of its first recorded adoption or until the end of the observation period if no adoption occurred. After its first adoption, a province exited the risk set and contributed no subsequent observations. To preserve temporal ordering and reduce potential simultaneity, selected time-varying covariates were entered using their one-year-lagged values.

3.2. Model specification

This study employs an Event History Analysis (EHA) model to investigate the factors influencing the diffusion of China’s special funds policy for sports industry development, following Berry and Berry’s (1990) pioneering application to policy innovation, such as state lottery adoption. EHA, a standard econometric tool for dynamic processes, examines transitions from non-adoption to adoption of the policy at the provincial level, as seen in prior Chinese studies on fiscal reforms and residence permit systems. The event is defined as a province’s first recorded adoption of the policy. Province–years prior to adoption remain in the risk set. Given the absence of precise adoption dates, we adopt a discrete-time hazard model with year as the observation unit, estimated via binary logistic regression in Stata 17.0. The hazard function, denoted as λ(t), represents the probability of policy adoption at time t. The logit-transformed model is:

(1)

To capture time-varying covariates in a province-year panel dataset, Equation 1 integrates economic and institutional drivers while addressing potential reverse causality through one-period lags for select variables (t-1). The independent variables are: Real per capita GDP (inflation-adjusted); : Fiscal self-sufficiency ratio; : Sports lottery sales; : Number of sports enterprises; : Public demand for sports; : Competitive pressure; : Learning pressure; : Administrative pressure. The left-hand side represents the log-odds of adoption probability, while serves as the time-varying error term to account for unobserved heterogeneity, ensuring a robust analysis of policy diffusion dynamics akin to frameworks in Walker [16] and Gray [37].

3.3. Research and variable measurement

3.3.2. Independent variables.

The independent variables are chosen to reflect the key factors hypothesized to influence the adoption of the special funds policy for the development of the sports industry. These variables include economic, financial, social, and external pressures. All variables are measured, as shown in Table 1.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Main Variables Measurement and Data Sources.

https://doi.org/10.1371/journal.pone.0354242.t001

Real Per Capita GDP, which is inflation-adjusted gross domestic product per person to account for price changes and provide a more accurate measure of economic output by avoiding inflation distortions, serves as a primary indicator of provincial economic capacity [38] and directly determines fiscal resources for policy innovation. Higher real per capita GDP fundamentally enables provincial adoption of sports funding policies by transforming general economic capacity into specific policy-implementation resources: elevated economic output directly expands fiscal reserves through taxation and fiscal autonomy—as demonstrated in Portugal where municipal sports financing relied strictly on local fiscal capabilities—thereby alleviating the core constraint of policy experimentation [39]. Concurrently, enhanced economic resources strengthen institutional capacity to execute complex funding mechanisms, evidenced by China’s finding that economic scale dictates sports policy diffusion through administrative mobilization efficiency [40]. Crucially, this economic-policy nexus operates via targeted resource-channeling wherein developed provinces prioritize sports investments as empirically validated in specialized initiatives such as demonstration cooperatives where policy efficacy was contingent on economic-priority alignment [41]. Building on these theoretical and empirical foundations, Per Capita GDP emerges as a critical factor influencing the adoption of innovative funding policies in the sports industry.

1. Hypothesis 1: Provinces with higher levels of Real Per Capita GDP are more likely to adopt special funds policies for sports industry development.

The fiscal self-sufficiency ratio, operationalized as the proportion of local government revenue relative to total expenditures, serves as a critical indicator of subnational financial autonomy. This metric captures the extent to which jurisdictions can mobilize internal resources without relying on intergovernmental transfers, thereby granting them enhanced discretion in resource allocation and policy experimentation [42]. In the context of the sports industry, by reducing dependency on centralized funding structures, fiscally autonomous jurisdictions can test novel instruments—including performance-based subsidy systems and public-private partnership models—without being constrained by homogenized budgetary guidelines from upper-level governments. For instance, local governments with robust fiscal capacities have established specialized funds to provide directed subsidies to sports enterprises undergoing service-oriented transformation, particularly in emerging sectors like sports tourism and health services [43]. Concurrently, fiscally autonomous jurisdictions leverage their discretion to implement innovation-driven policy frameworks that integrate technological advancements with cultural preservation, such as digital infrastructure upgrades for sports facilities combined with localized heritage initiatives to stimulate public participation [44]. Based on these insights, the following hypothesis is proposed:

1. Hypothesis 2: Provinces with higher fiscal self-sufficiency ratios are more inclined to adopt innovative policies for special funds supporting the development of the sports industry.

Sports lottery sales generate self-reinforcing fiscal cycles that directly enable innovative policy adoption. First, by funding tax incentive frameworks, specifically cost-plus-deduction policies for sports enterprises, that lower innovation barriers [45]. Second, they enable targeted investments in grassroots initiatives, including mass sports promotion [46], thereby ensuring direct alignment between fiscal resources and core policy objectives [47]. Third, lottery-driven revenue growth necessitates regulatory adaptation to new market dynamics—as seen when China’s blockchain-integrated ticket systems compelled real-time policy updates for fraud prevention and consumer protection. Crucially, this process transforms lottery revenue from a passive fiscal reservoir into an active lever for policy experimentation and implementation.

1. Hypothesis 3: Provinces with higher sports lottery sales are more likely to implement innovative policies for special funds supporting the development of the sports industry.

The number of sports enterprises drives the diffusion of sports industry special funds through demand-side pull, institutional capacity, and network effects [48]. High-density regions such as Zhejiang and Fujian foster collaborative ecosystems where enterprises collectively advocate for tailored fiscal support, including tax incentives for equipment manufacturers, while institutional feedback mechanisms like Fujian’s Sports Industry Development Council align funding allocation with sector needs. [49]. Entrepreneurial clusters including Shenzhen’s sports e-commerce ecosystem amplify policy diffusion via peer learning and benchmarking, enabling successful models like PPP stadium funding to spread regionally through inter-provincial cooperation platforms [50]. This creates a policy laboratory effect, where dense networks reduce transaction costs and accelerate the replication of specialized funding practices across jurisdictions. We therefore proposed:

1. Hypothesis 4: Provinces with a higher number of sports enterprises are more likely to adopt innovative policies for special funds supporting the development of the sports industry.

Public demand for sports refers to the preferences and potential needs of residents for sports participation, fitness services, sporting events, and related goods and facilities [51]. Growing public interest in health, leisure, and physical activity may place pressure on provincial governments to expand sports services and introduce financial instruments that support the development of the sports industry. Previous research suggests that stronger demand for accessible facilities and participation opportunities can encourage governments to adjust infrastructure investment and funding allocation [52–53]. Because directly comparable annual data on sports participation and household sports expenditure were not consistently available for all 31 provincial-level regions over the observation period, this study uses provincial population density as an indirect proxy for the concentration of potential public demand. Population density does not directly measure residents’ preferences or actual participation; rather, it captures the concentration of potential users and the corresponding pressure on sports facilities and services. On this basis, the following hypothesis is proposed:

1. Hypothesis 5: Provinces with higher public demand for sports are more likely to adopt innovative policies for special funds supporting the development of the sports industry.

Competitive pressure among provinces often manifests as yardstick competition, where local governments benchmark their performance against neighboring regions in areas like economic development and public services, including the sports industry [54]. This dynamic encourages officials to monitor and respond to peers’ achievements to avoid lagging in regional development and comparative performance assessments, and to strive for advantages in evaluations by higher authorities and career advancement. For instance, innovative funding mechanisms for sports venues in one province pressure others to adopt similar policies, like establishing special funds or tax incentives. This mirrors U.S. municipal competition for professional teams through publicly subsidized stadiums [55]. In China, rivalry to host events like the 2021 National Games drove Shaanxi Province to invest heavily in venues and infrastructure across multiple cities, promising economic growth despite debates over fiscal sustainability. Such inter-jurisdictional competition thus spurs facility investment and policy innovation to gain a competitive edge. Consequently, the study hypothesizes:

1. Hypothesis 6: Provinces under higher competitive pressure are more likely to adopt innovative policies for special funds supporting the development of the sports industry.

Learning pressure arises when provincial governments voluntarily observe and emulate successful policy approaches from peer regions, often to refine their own strategies in response to shared economic or social challenges [56]. Unlike administrative pressure, learning pressure operates horizontally, driven by self-initiated benchmarking and adaptation rather than hierarchical directives [57]. For instance, officials might scrutinize how a neighboring province has effectively allocated special funds to boost sports tourism, then adapt similar mechanisms to mitigate risks and improve outcomes in their own context. This process of evidence-based adaptation—monitoring proven practices in comparable areas—helps reduce uncertainty, sharpen decision-making, and promote policy efficacy [4]. We therefore proposed:

1. Hypothesis 7: Provinces experiencing higher levels of learning pressure are more likely to adopt special funding policies for sports industry development.

Administrative pressure, defined as the influence exerted by higher-level authorities on local governments through formal directives, is a key driver of policy adoption in China’s governance system [58]. In this study, administrative pressure is measured by the annual number of central directives and policy documents relevant to sports industry development, which serves as a proxy for the intensity of vertical policy signals. Although this system efficiently disseminates policy priorities—aligning provinces with national objectives like sports-led growth—it frequently obscures the boundary between mandating compliance and sparking innovation, as local officials might treat directives as inflexible commands rather than prompts for originality. Central orders for sports funding, for example, enforce standardization but may erode local autonomy by imposing uniform frameworks that disregard regional disparities in capabilities or demands [59]. Administrative pressure mitigates uncertainty via accountability mechanisms, permitting measured innovation exemplified by adaptive funds for quota management. It can also reduce flexibility, forcing rote conformity that overlooks context-specific needs and breeds inefficiency [4]. Navigating this interplay—central guidance versus local leeway—aims to uphold national uniformity in sports policy while accommodating tailored adjustments. Accordingly, the study hypothesizes:

1. Hypothesis 8: Provinces facing higher administrative pressure are more likely to adopt innovative policies for special funds supporting the development of the sports industry.

4.1. Descriptive statistics and correlation analysis

Table 2 provides a detailed summary of the descriptive statistics for all variables, including means, standard deviations, minimums, and maximums. These statistics reveal substantial heterogeneity across regions, particularly in terms of economic and policy-related factors such as real per capita GDP, fiscal self-sufficiency, and sports lottery sales. The variability observed in these indicators highlights disparities in regional economic capacity, which likely influence the ability to adopt and implement policy innovations. Similarly, the range of values for variables related to the sports industry, such as the number of sports enterprises and public sports demand, underscores differences in industry scale and population concentration and potential demand for sports services, offering important context for subsequent analyses.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Descriptive Statistics of Variables.

https://doi.org/10.1371/journal.pone.0354242.t002

Table 3 presents the correlation analysis results for the adoption of sports industry development special fund policies, revealing varied relationships between policy adoption and the explanatory variables. Policy adoption is significantly and positively correlated with sports lottery sales, administrative pressure, and real per capita GDP. Fiscal self-sufficiency shows a moderate positive correlation at the 0.05 significance level, while the number of sports enterprises exhibits a weak positive correlation at the 0.10 significance level. However, policy adoption is not significantly correlated with public demand for sports or competitive pressure, providing no preliminary bivariate support for H5 and H6. These bivariate results indicate that neither the population-density proxy for public demand nor competitive pressure has a clear unadjusted association with policy adoption; their relationships are examined further in the multivariate models. The high correlation between sports lottery sales and the number of sports enterprises (r = 0.888) also warrants further assessment of multicollinearity, as reported in Section 4.2. Overall, the mixed correlation patterns provide a foundation for the subsequent multivariate analysis, particularly for distinguishing institutional and financial determinants from market-oriented factors in the diffusion of sports industry fund policies. The results also highlight the need for caution when applying general policy diffusion theories to specialized policy contexts such as sports industry funding programs.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Correlation Matrix of Variables.

https://doi.org/10.1371/journal.pone.0354242.t003

4.2. Multicollinearity tests

To ensure the accuracy of regression results, independent variables must remain relatively independent. Multicollinearity diagnostics are essential, commonly using the Variance Inflation Factor (VIF). Per Gujarati and Porter (2009), VIF > 10 indicates severe issues that could distort estimates. In this study, all VIF values are below 10 (mean VIF = 4.45), suggesting no severe multicollinearity, as detailed in Table 4.

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Results of Multicollinearity Tests.

https://doi.org/10.1371/journal.pone.0354242.t004

Despite acceptable VIFs, the correlation matrix reveals notable pairwise correlations. Higher VIFs indicate moderate multicollinearity, potentially inflating standard errors. As a supplementary exploratory diagnostic, principal component analysis (PCA) was used to examine the underlying correlation structure among the explanatory variables, with results in Table 5. PCA extracted two main components: the first captures socio-economic development and pressure factors (high loadings on Learning Pressure, Per Capita GDP, Competitive Pressure, Sports Lottery Sales, and Number of Sports Enterprises), while the second reflects fiscal and market demand factors (high loadings on Fiscal Self-Sufficiency and Public Demand for Sports). These components explain most variance without the original inter-correlations, as shown in Table 6.

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. PCA Results – Eigenvalues, Variance Explained.

https://doi.org/10.1371/journal.pone.0354242.t005

Download:

• PNG
  larger image
• TIFF
  original image

Table 6. Rotated Component Loadings for the First Two Principal Components.

https://doi.org/10.1371/journal.pone.0354242.t006

4.3. Regression analysis

Hypotheses were tested in Stata 17.0 using binary logistic regression within a discrete-time event history framework. Each province contributed annual observations while it remained at risk of adoption and exited the risk set after its first recorded adoption. The regression results are reported in Table 7. Models 1–3 separately examine policy-subject, policy-target, and policy-environment factors. Model 4 combines policy-subject and policy-target factors, Model 5 combines policy-subject and policy-environment factors, and Model 6 combines policy-target and policy-environment factors. Model 7 is the fully specified model incorporating all eight explanatory variables across the three groups of factors. Hypothesis evaluation is based primarily on Model 7, while Models 1–6 show how the estimated relationships change across different combinations of explanatory variables. Table 7 reports model-based standard errors, whereas Section 4.4 examines the sensitivity of the results to province-clustered standard errors and other alternative specifications.

Download:

• PNG
  larger image
• TIFF
  original image

Table 7. Results of the Model Regression Analysis.

https://doi.org/10.1371/journal.pone.0354242.t007

The policy-subject factors exhibit differentiated relationships with policy adoption. In Model 7, fiscal self-sufficiency is positively and significantly associated with adoption (β = 7.3619, SE = 3.1214, p < 0.05), supporting H2. Sports lottery sales are also positively and significantly associated with adoption (β = 1.4840, SE = 0.7025, p < 0.05), supporting H3. These findings indicate that provincial fiscal discretion and policy-specific financial resources are important enabling conditions for the establishment of sports industry development funds. By contrast, real per capita GDP is not statistically significant in the fully specified model (β = −0.3348, SE = 1.2406), and H1 is therefore not supported. This result suggests that general economic development does not independently explain differences in provincial adoption once more directly deployable fiscal resources and the remaining explanatory factors are considered.

The policy-target factors reveal a more complex pattern. Although the number of sports enterprises is positively associated with adoption in the reduced Model 2, its coefficient becomes negative and statistically significant in Model 7 (β = −1.4305, SE = 0.5940, p < 0.05). Because H4 predicts a positive relationship, this result does not support H4. The negative coefficient nevertheless raises the possibility of a compensatory-adoption pattern in which provinces with a comparatively weaker enterprise base have stronger incentives to establish dedicated funds to stimulate industrial development. This interpretation is post hoc and should be treated cautiously because the proposed compensatory mechanism is not directly tested by the regression model. Public demand for sports is negative but statistically non-significant in Model 7 (β = −13.1657, SE = 8.6067); H5 is therefore not supported in the fully specified model.

The policy-environment factors likewise provide limited evidence of independent relationships in the fully specified model. Competitive pressure is negative and statistically non-significant (β = −1.1382, SE = 1.2232), while learning pressure is positive but statistically non-significant (β = 5.1068, SE = 3.3532). H6 and H7 are therefore not supported. Administrative pressure is positively and significantly associated with adoption in several reduced specifications, but the coefficient becomes statistically non-significant after all explanatory variables are included in Model 7 (β = 0.2859, SE = 0.2018). H8 consequently does not receive support in the fully specified model. The change across specifications suggests that administrative pressure may operate as a contextual or conditional influence rather than as an independent determinant of provincial adoption timing. Overall, the results indicate that policy-specific fiscal resources are more consistently associated with adoption than general economic capacity, market conditions, or intergovernmental pressures.

4.4. Sensitivity analyses

To assess whether the findings from the fully specified model were sensitive to alternative methodological choices, three additional analyses were conducted within the discrete-time event history framework. First, Model 7 was re-estimated using standard errors clustered at the provincial level to account for within-province dependence across annual observations. Second, a centered linear duration term was added to Model 7 to account for temporal variation in the baseline hazard of policy adoption. Third, real per capita GDP, fiscal self-sufficiency, sports lottery sales, and public demand for sports were lagged by two years rather than one year. All sensitivity specifications retained the explanatory variables included in Model 7.

The province-clustered and linear-duration specifications used the same 376 province-year observations as Model 7. Because constructing the two-year-lagged variables required excluding the first available observation for each province, the effective sample size of the two-year-lag specification was reduced to 345 province-year observations. All three specifications retained the same 18 policy-adoption events. As shown in Table 8, the coefficient for the number of sports enterprises remained negative across all three specifications. It was statistically significant in the province-clustered specification and marginally significant in both the linear-duration and two-year-lag specifications. These results indicate that the negative direction of the relationship is relatively stable, but they do not support the positive relationship predicted by H4. Administrative pressure remained positively signed but statistically non-significant in all three specifications. Therefore, H8 was not supported by the sensitivity analyses.

Download:

• PNG
  larger image
• TIFF
  original image

Table 8. Sensitivity Analyses Compared with Model 7.

https://doi.org/10.1371/journal.pone.0354242.t008

Fiscal self-sufficiency remained positive and statistically significant in the province-clustered and two-year-lag specifications but became statistically non-significant after the linear duration term was introduced. Sports lottery sales also remained positive, although its statistical significance varied across specifications. Overall, the sensitivity analyses provide stronger evidence regarding the direction of several relationships than regarding the stability of their statistical significance.

4.5. Micro-mechanisms of policy diffusion: Qualitative evidence from stakeholder interviews

The qualitative follow-up focuses primarily on four predictors that were statistically non-significant in the fully specified model: real per capita GDP (H1), public demand for sports (H5), competitive pressure (H6), and learning pressure (H7). Learning pressure nevertheless exhibited a weak positive bivariate correlation at the 0.10 significance level, but this relationship was not sustained after the remaining covariates were included. The results for H4 and H8 require a different analytical interpretation. H4 produced a statistically significant relationship in the direction opposite to that hypothesized, whereas H8 was positive in several reduced models but statistically non-significant in the fully specified model and the sensitivity analyses. Accordingly, the purpose of this section is not to provide qualitative confirmation for every unsupported hypothesis, but to examine possible mechanisms for which the interview materials provided directly relevant evidence.

This section draws on the thematic analysis of semi-structured interviews with 30 stakeholders, including policymakers, implementers, and beneficiaries. Consistent with the procedures described in Section 3.3.3, meaning-bearing transcript segments served as the coding units, while complete interview transcripts served as the units for comparing themes across participants. The qualitative evidence is used to illustrate possible mechanisms rather than to estimate the prevalence of particular views among participants or provinces. The principal themes and illustrative evidence are summarized in Table 9. All quotations were translated from Mandarin Chinese into English by the authors and checked against the original transcripts.

Download:

• PNG
  larger image
• TIFF
  original image

Table 9. Thematic Coding of Strategic Local Governance Behaviors.

https://doi.org/10.1371/journal.pone.0354242.t009

Although the theoretical framework predicted a positive relationship between economic development and policy adoption, real per capita GDP was not statistically significant across the regression specifications. The interview accounts provide one possible explanation: greater economic capacity does not necessarily translate into a stronger commitment to sports industry fund policies. Participants described how some provincial governments prioritized highly visible and commercially oriented projects, such as e-sports arenas and sports tourism complexes, because these projects were perceived to generate more immediate political or economic returns. Such prioritization may divert attention and resources from less visible community fitness programs, thereby weakening the expected relationship between economic development and policy adoption. As one interviewee explained, “Economic growth provides resources, but we prioritize projects that deliver quick political returns rather than addressing grassroots needs” (Interviewee #7, 2023). This account illustrates a possible mechanism of strategic resource allocation and policy decoupling; it does not establish the prevalence of this behavior across provinces.

Public demand for sports was hypothesized to be positively associated with policy adoption. However, its coefficient in the full model was negative but statistically non-significant, while the sensitivity specifications produced consistently negative estimates that were statistically significant or marginally significant. Because these estimates were opposite to the positive relationship predicted by H5, they did not support the hypothesis. The interview accounts suggest that expressions of public demand may sometimes be translated into highly visible projects, such as sports tourism developments, rather than less visible community-based fitness services. Such responses may not correspond to the forms of sports provision that residents demand in their daily lives, potentially weakening the expected relationship between public demand and policy adoption. As one interviewee explained, “When public demand is high, we build large complexes to demonstrate action, but actual fitness sites are neglected” (Interviewee #12, 2023). This account illustrates how a visible policy response may symbolically acknowledge public demand without necessarily strengthening the substantive connection between demand and policy adoption. It therefore offers a possible interpretation of the statistically non-significant and model-dependent result rather than evidence of a systematic negative effect.

Interprovincial competition was also expected to promote policy adoption. However, competitive pressure was not statistically significant in either the full model or the sensitivity analyses reported in Section 4.4. The interview accounts suggest that symbolic compliance may provide one possible explanation. Some participants described instances in which existing expenditures were relabeled or reclassified to demonstrate compliance with higher-level assessments without necessarily creating substantively new policy commitments. As one interviewee explained, “Competition leads us to relabel old expenditures to meet superior assessments” (Interviewee #18, 2023). Such relabeling may create formal evidence of policy responsiveness without a corresponding increase in substantive policy innovation or resource commitment. These accounts suggest that when administrative evaluations emphasize visible responses, competitive pressure may encourage formal compliance rather than substantive innovation. This mechanism offers one possible interpretation of why competitive pressure was not significantly associated with policy adoption in the quantitative models.

Learning pressure showed a weak positive bivariate correlation at the 0.10 significance level. However, the coefficient was no longer statistically significant after administrative pressure and the other covariates were included in the full model. The interview accounts suggest that selective policy adaptation may provide one possible explanation for this model-dependent result. Some participants described how recognizable elements of policies adopted elsewhere were retained while their implementation was adjusted to local political and economic priorities. As one interviewee explained, “We retained the ‘digital payment’ branding but redirected funds to align with leadership priorities” (Interviewee #5, 2023). A related interview account described how commercially oriented projects could be reframed to align with both industrial-development and public-health objectives (Interviewee #21, 2023). These accounts suggest that policy learning may involve selective reinterpretation rather than direct replication, with greater attention given to policy elements that can be aligned with visible industrial or administrative priorities. This mechanism may help explain why the weak bivariate association between learning pressure and policy adoption was not sustained in the full model. The interview evidence is interpreted as illustrating a possible mechanism of selective adaptation rather than establishing how frequently this behavior occurred across provinces.

Taken together, the qualitative findings suggest that the lack of support for the hypothesized positive relationships should not be interpreted as evidence that the underlying policy mechanisms are irrelevant. Instead, economic resources, public demand, interprovincial competition, and policy learning may be strategically reinterpreted during provincial implementation. Resource allocation may favor highly visible projects, expressions of public demand may be translated into symbolic policy responses, competitive pressure may encourage formal compliance, and policy learning may involve selective adaptation rather than direct replication.

The H4 and H8 results further reinforce this interpretation of conditional and strategically mediated diffusion. The negative coefficient for the number of sports enterprises is inconsistent with the positive relationship predicted by H4 but may indicate a compensatory pattern in which provinces with a comparatively weak enterprise base have stronger incentives to introduce dedicated policy support. Because the interviews were not designed to estimate the prevalence of this mechanism, this interpretation remains tentative. Similarly, the attenuation of the administrative-pressure coefficient in the fully specified model suggests that national directives may provide a common institutional background without independently determining differences in provincial adoption timing. These findings identify boundary conditions of conventional policy-diffusion explanations rather than invalidating the broader theoretical framework.

5.1. Research conclusions and discussion

Conventional policy-diffusion explanations generally assume that economically stronger jurisdictions, more mature target markets, and greater intergovernmental pressure are more likely to adopt policy innovations. The findings of this study qualify that expectation. In the diffusion of provincial sports industry development funds, adoption was not uniformly associated with greater economic development, a larger sports-enterprise base, stronger public demand, or greater horizontal and vertical pressure. Instead, the results reveal three more differentiated dynamics: the importance of deployable policy-specific resources, the possibility of compensatory policy intervention, and the strategic local translation of institutional signals.

First, the results distinguish general economic capacity from resources that can be directly mobilized for a specialized policy instrument. Real per capita GDP did not demonstrate an independent relationship with adoption, whereas fiscal self-sufficiency and sports lottery sales were positively and significantly associated with adoption in the fully specified model, supporting H2 and H3. Both variables retained positive coefficients across the sensitivity specifications. These findings suggest that aggregate wealth and deployable policy capacity should not be treated as equivalent: a province may possess substantial economic resources without allocating them to a sports industry fund, whereas fiscal discretion and sports-related revenue can be mobilized more directly. This interpretation connects policy-diffusion research on socioeconomic capacity [61] with fiscal federalism research emphasizing subnational discretion. The interview accounts provide a complementary explanation for the non-significant GDP result by indicating that general economic resources may sometimes be redirected toward projects offering greater political visibility or more immediate commercial returns rather than converted into dedicated funding commitments.

Second, the policy-target results challenge the assumption that a larger enterprise base necessarily precedes government intervention. The number of sports enterprises was negative and statistically significant in Model 7 and remained negative across the sensitivity specifications. H4 was therefore not supported in its hypothesized positive direction. One possible post hoc interpretation is a compensatory-adoption logic: provinces with a comparatively weak enterprise base may have stronger incentives to establish dedicated funds to stimulate industrial development, whereas provinces with more mature markets may rely more heavily on existing commercial and financial networks. This interpretation is relevant to research on the dynamic transformation of sports industry services, but it remains tentative because the analysis did not directly test the proposed compensatory mechanism and the enterprise measure overlaps empirically with other indicators of provincial resource capacity. Public demand for sports likewise did not demonstrate the positive independent relationship predicted by H5. Interview participants described how expressions of demand could be translated into highly visible construction or promotional projects rather than community-based services, suggesting that demand is filtered through local definitions of political visibility and administrative feasibility. This interpretation is consistent with research showing that the effects of government sports expenditure depend on how resources are configured and allocated.

Third, the policy-environment findings suggest that institutional pressures operate as contextual and strategically mediated influences rather than automatic adoption triggers. Competitive and learning pressures did not exhibit independent relationships in the fully specified model. Administrative pressure was positive and significant in several reduced models but became statistically non-significant after all explanatory factors were included in Model 7 and remained non-significant in the sensitivity analyses. H8 therefore did not receive support as an independent determinant of adoption. Nevertheless, national directives may still place sports industry development on provincial agendas and define legitimate policy instruments without independently explaining why some provinces adopt earlier than others. The qualitative findings illuminate this distinction. Accounts of expenditure relabeling suggest that competition may encourage formal compliance without producing substantively new commitments, while accounts of selective adaptation indicate that provinces may retain recognizable policy language but redirect implementation toward local political and commercial priorities.

Complementing the regression analysis, the thematic analysis of 30 stakeholder interviews identifies strategic resource allocation, symbolic responses to public demand, formal compliance, and selective policy adaptation as possible mechanisms connecting institutional conditions to provincial responses. These themes are not used as independent tests of the hypotheses or as estimates of how frequently particular practices occurred among participants or provinces. Rather, they provide an actor-centered account of how economic resources, social demand, competition, and policy learning may be interpreted and reframed during local implementation. The qualitative and quantitative components therefore perform complementary functions: the regression analysis evaluates the hypothesized relationships, while the interview evidence helps illuminate the strategic processes through which policy conditions are translated into provincial responses.

Taken together, the findings suggest a resource–compensation–translation interpretation of policy diffusion in this specialized field. Adoption depends more clearly on resources that provincial governments can discretionarily mobilize than on general economic development; industrial policy may respond to perceived market weakness as well as market maturity; and external policy signals are filtered through strategic local adaptation rather than transmitted automatically into substantive adoption. The contribution of this study therefore lies not in maximizing the number of statistically supported hypotheses, but in identifying the boundary conditions of a uniform, additive model of policy diffusion. By showing how deployable fiscal resources, perceived industrial needs, and local interpretation interact, the study offers a more context-sensitive explanation of policy innovation diffusion within China’s decentralized yet hierarchically coordinated governance system.

5.2. Policy recommendations

To address the identified challenges in policy diffusion, this study proposes targeted and feasible measures that build on existing governance structures in China’s sports sector.

First, to counter fiscal incentives that skew resources toward commercial projects, the central government could establish a national sports industry development fund administered by the General Administration of Sport. This fund would prioritize resource allocation to underdeveloped regions, such as those in the western and northeastern provinces, through needs-based grants allocated according to regional fiscal capacity and community access needs. Such an initiative is politically practical, as it can be piloted through central-local partnerships to ease fiscal pressures on lagging areas while promoting greater local autonomy and reducing long-term dependency on direct subsidies, thereby fostering balanced regional development as supported by recent studies on economic disparities.

Second, to mitigate selective mimicry and policy homogenization, local authorities should introduce digital monitoring systems to track policy adaptation and implementation, using standardized indicators that compare stated policy commitments with documented fund allocations, actual disbursements, beneficiaries, and implementation outcomes. Complementing this, inter-regional talent exchanges and training programs could encourage context-sensitive innovations, with rollout leveraging established e-governance platforms for cost efficiency and incentives rewarding effective implementations to ensure broad stakeholder engagement.

Finally, to bridge enforcement gaps and resolve vertical-horizontal conflicts, intergovernmental coordination could be strengthened via collaborative forums backed by clear legal delineations of responsibilities and integrated information-sharing networks, drawing on international best practices for enhanced accountability. Anchoring these in national policy directives would facilitate rapid adoption, minimizing redundancies and supporting adaptable, sustainable diffusion aligned with broader goals of equitable sports governance.