The application of Data Envelopment Analysis (DEA) marked a pivotal shift toward quantitative governance assessment. Dong Xiuhai et al. employed the DEA methodology to demonstrate suboptimal eco-environmental governance efficiency in China, advocating for enhanced pollution control and environmental infrastructure outputs (12). Zhang Qingmin et al. conducted an empirical analysis using a three-phase DEA model, revealing persistent optimization potential in China’s urban agglomeration governance efficiency (13). Beyond conventional DEA applications, Golany addressed unintended outputs through nonlinear functional approaches, though this methodological framework encounters limitations in accommodating data convexity (14). Tone’s SBM model advanced efficiency measurement by incorporating both radial and angular dimensions while accounting for input/output slack variables (15). Xu Chenglong et al. utilized SE-DEA and EViews 6.0 to identify key strategies for Shandong Province: developing circular economies, tightening environ-mental regulations, and implementing region-specific policies (16). Li et al. used a super SBM (based on relaxed measurement) model based on a panel Tobit regression model to explore the tourism ecological efficiency of 13 cities in the Beijing-Tianjin-Hebei region of China from 2010 to 2019 (17).

While conventional regression dominates policy analysis (82% of sampled studies) (14), spatial econometrics has gained traction in infrastructure-impact studies. Notably, Liu et al. (18) in Transport Policy demonstrated railway-induced economic spillovers using spatial Durbin models, establishing methodological precedents for handling geographic heterogeneity. Wang et al. used a super-slack variable (SBM) model to calculate the ecological efficiency of tourism in Inner Mongolia from 2009 to 2019 and analyzed the factors influencing spatial evolution using a geographically weighted regression method (19). Hu et al. used a geographically weighted regression (GWR) model to assess regional differences in the impact of air pollutants on life expectancy (20). However, their focus on economic externalities differs fundamentally from environmental governance systems requiring spatiotemporal coordination of ecological thresholds (21). Parallel advancements in temporal granularity emerge through panel vector autoregression (PVAR) applications (22), yet integrated spacetime modeling remains scarce—only 12% of Web of Science-indexed environmental studies (2019–2023) adopted GTWR methods (23).

The government-enterprise partnership (GEC) paradigm has drawn increasing attention since China’s 2020 Modern Environmental Governance System reform. While Bao and Guan (24) identified fiscal mediation effects through regulatory channels, recent experimental evidence from Jiangsu’s PPP pilots shows 23–37% efficiency gains compared to state-led projects (25). Internationally, Marini et al. (26) market environmentalism theory provides conceptual scaffolding, though its Western-centric assumptions require adaptation to China’s socialist market context where administrative coordination remains pivotal (27).

Synthesizing these developments reveals three unresolved challenges:

Temporal–spatial decoupling: Existing spatial analyses (21) predominantly examine cross-sectional patterns, neglecting the co-evolution of policies and ecosystems over time.

Partnership complexity: Current GEC studies (24–27) oversimplify principal-agent relationships, lacking empirical validation of incentive alignment mechanisms.

Driver interactivity: Conventional methods (DEA/PVAR/DID) examine linear causality, overlooking synergistic configurations between institutional, technological, and market factors.

Our study advances an integrated analytical framework:

Spatiotemporal synthesis: Combining Geographically and Temporally Weighted Regression (GTWR) with ecological carrying capacity spatial weights, we extend Chen et al.’s (2023) spatial paradigm by incorporating dynamic policy-ecosystem interactions (2010–2021).

Hybrid causality: Embedding DEA-derived efficiency scores into fsQCA enables detection of nonlinear driver configurations—a critical advancement beyond single-dimension spatial effects.

Provinces like Qinghai and Hubei achieved tier escalation (medium → high) through institutionalized PPP frameworks, whereas Hainan’s dramatic decline (high → low) exposed fiscal misallocation in coastal regions (eco-investment <1.2% GDP).

Jiangsu’s ascent to the high-efficiency tier demonstrated tech-embedded governance efficacy, contrasting with Shanghai’s regression (high → lower tier) due to R&D intensity lagging behind the 0.5% GDP threshold.

Emerging western clusters (Yunnan, Guizhou) exhibited efficiency erosion (high → lower tier), signaling coordination failures. Conversely, Beijing’s resurgence highlighted talent mobility’s critical role.

This triphasic trajectory aligns with the Environmental Kuznets Curve hypothesis, where mid-income provinces achieved maximal efficiency gains through GEC institutional innovation.

The spatiotemporal evolution of China’s eco-environmental Government-Enterprise Collaboration (GEC) governance efficiency (2010–2021) reveals two distinct geospatial regimes:

In 2010, technical efficiency demonstrated polycentric dispersion, with high-performance clusters concentrated in eastern coastal provinces (Jiangsu, Zhejiang, Guangdong), while efficiency-deficient zones dominated central-western hinterlands. This pattern aligned with core-periphery theory, where coastal advantages in institutional capacity and capital accessibility drove initial divergence.

By 2021, emergent agglomeration effects reshaped the landscape. High-performance clusters expanded contiguously, covering 84% of eastern provinces and demonstrating spillover effects into central (Hubei, Hunan) and western (Sichuan, Shaanxi) regions. Notably, northeastern provinces (Liaoning, Heilongjiang) transitioned from medium-efficiency to frontier status through SOE reform-driven GEC institutionalization.

This transformation reflects: 1. Policy-driven convergence through interregional partnership networks; 2. Path dependency breaking in legacy industrial regions 3. Technology diffusion gradients from eastern innovation hubs.

Kernel Density Estimation (KDE), as a non-parametric technique, enables flexible recovery of probability density functions by smoothing discrete observations through localized kernel weighting. Its superiority over parametric methods lies in eliminating distributional assumptions that often induce specification bias (39).

The mathematical formulation: (3) f ( x ) = 1 Nh ∑ i = 1 N K [ X i − x h ]

In Equation 3, K [−] denotes the kernel function, Xi denotes the observed values under independent homogeneous distribution, N denotes the number of observed samples (30 provinces (cities and districts)) and is the mean; h is the bandwidth and its selection satisfies Equation 4, which is used to indicate the smoothness of the kernel density curve. The smaller the value of h, the narrower the peaks of the kernel density curve, and on the other hand, the bigger the value of h, the flattening of the wave peaks of the kernel density curve. (4) lim N → ∞ h ( N ) = 0 ; lim N → ∞ Nh ( H ) = N → ∞

To rigorously capture the spatiotemporal heterogeneity of China’s eco-environmental Government-Enterprise Collaboration (GEC) governance efficacy, we implement a spatiotemporally adaptive Gaussian kernel density estimator. This advanced technique overcomes limitations of conventional KDE through three key innovations Equation (5): (5) K ( x ) = 1 2 π exp [ − x 2 2 ]

According to the kernel density estimation model, distribution maps for 2010, 2014, 2018, and 2021 were generated using Stata 17 software to analyze the dynamic evolution trends in governance efficiency of China’s ecological environment and government-enterprise collaboration (see Figure 4).

As shown in Figure 4, the curve exhibits a bimodal distribution pattern characterized by “one primary and one secondary peak.” The dominant peak on the efficient side is accompanied by a smaller peak that gradually shifts rightward on the left side, suggesting a developing bipolar trend in the governance efficiency of China’s eco-friendly government-enterprise collaboration during this period.

From 2010 to 2014, the primary peak intensified while the kernel density function’s central tendency shifted leftward, accompanied by a reduction in the secondary peak. This pattern indicates a gradual convergence in governance efficiency across provinces (municipalities and autonomous regions).

During 2014–2018, the primary peak became more pronounced and transitioned from broad to sharp, with the distribution curve shifting leftward and exhibiting reduced variability. These changes suggest emerging polarization in interprovincial disparities.

From 2018 to 2021, the primary peak diminished significantly while the secondary peak flattened, demonstrating renewed divergence in governance efficiency among provinces (municipalities and autonomous regions).

Overall, regional disparities in governance efficiency of government-enterprise collaboration for ecological environment management exhibit a “narrowing-widening” cyclical pattern.

The Pressure-State-Response (PSR) framework model (see Figure 5) establishes a clear causal relationship: human activities exerting environmental pressure induce measurable state changes in ecosystems, thereby compelling societal responses to restore environmental quality. This framework corresponds to the complete policymaking and countermeasure development cycle.

The PSR model’s widespread adoption primarily stems from its capacity to identify environmental pressure sources. This addresses the fundamental principle that human activities generating environmental pressures must be held accountable for resultant ecological changes.

Based on the pressure-state-response model and after referring to related literature, the following factors are selected as the driving factors for the governance efficiency of ecological environment government-enterprise cooperation in China to be analyzed.

The timeframe for analyzing the drivers of this study spans from 2010 to 2021, and the study area covers 30 provinces (cities and districts) in China. The data used in the empirical analysis were sourced from the 2010–2021 China Statistical Yearbook, China Science and Technology Statistical Yearbook, China Environmental Statistical Yearbook, and China Population and Employment Statistical Yearbook. To gain preliminary insights into the indicators, some variables were logarithmized, and descriptive statistics were conducted (see Table 4). Table 4 shows that the sample size of the study is 360, and the overall standard deviation of the data is relatively small, indicating that the sample statistics are close to the population parameters. This suggests that the sample is representative, and the conclusions drawn from the study are highly reliable. Additionally, to mitigate potential multicollinearity issues arising from variable correlations, the variance inflation factor (VIF) was employed to test the indicators. The results, presented in Table 4, show that all VIF values are below 10, indicating no significant multicollinearity among the indicators. This confirms that the model can be reliably constructed.

With the help of ArcGIS 10.2, spatio-temporal geographically weighted regression analysis was conducted to analyze the influencing factors of the governance efficiency of China’s ecological environment government-enterprise cooperation. To ensure a more scientific model selection, prior to running the model, the spatio-temporal distance parameter ratio was set to 1, the bandwidth was automatically optimized, and traditional linear regression and geographically weighted regression were conducted as comparisons. The AICc value, residual sum of squares (RSS), and goodness-of-fit R² were selected as confidence evaluation indices, and the specific results are shown in Tables 5, 6. As shown in Tables 5, 6, the AICc value, residual sum of squares (RSS), and goodness-of-fit R² of the three models were ranked as follows: the residual sum of squares (RSS) is OLS > GWR > GTWR, and the goodness-of-fit R² follows the order GTWR > GWR > OLS. It can be seen that, compared with traditional linear regression and geographically weighted regression, the spatio-temporal geographically weighted regression model with spatio-temporal characteristics exhibits an explanatory strength of 67.3%, has the smallest AICc value, and the smallest residual sum of squares (RSS), thus achieving a better fitting effect. Therefore, the spatio-temporal geographically weighted regression model with temporal and spatial non-stationarity was chosen to analyze the spatio-temporal heterogeneity of the driving factors of the governance efficiency of government-enterprise cooperation in China’s ecological environment.

By applying the spatio-temporal geographically weighted regression (STGWR) model to analyze driving factors, we obtained the estimated coefficients of these factors across different temporal and spatial contexts affecting the governance efficiency of China’s eco-environmental government-enterprise partnerships. Using these coefficients, we generated a box plot through Origin 2022 software that clearly demonstrates the temporal variation patterns of the driving factors (see Figure 6).

During the initial study period, the economic development level (lngdp) positively influenced the efficiency of public-private partnerships in China’s ecological governance. Economically underdeveloped regions, motivated to improve their environmental conditions but constrained by limited financial capacity, actively sought private capital investments to compensate for developmental gaps. This mutual investment established effective government-enterprise collaborations that generated cumulative effects, thereby enhancing governance efficiency. However, post-2017 analysis reveals a paradigm shift: with accelerated regional economic growth, increased governmental investments in ecological management have sufficiently addressed environmental needs. Consequently, the economic development factor demonstrates a declining trend and negative correlation with governance efficiency in most regions, reducing reliance on corporate capital infusion.

The estimated coefficient of population density (lnpop) demonstrates relative stability throughout the study period, though consistently negative values indicate an inverse correlation between population density and the efficiency of government-enterprise collaboration in China’s ecological governance. This suggests that population concentration impedes effective public-private partnerships in environmental management. In densely populated regions, human capital agglomeration generates positive externalities while enhancing regional energy efficiency. These combined factors enable governments to independently ensure ecological governance quality, thereby reducing the necessity for corporate partnerships. Conversely, sparsely populated areas lacking human capital concentration require governments to actively engage corporate participation, which stimulates the formation of collaborative governance mechanisms.

Marketization (lnmarket) demonstrates a positive correlation with the efficiency of public-private partnerships (PPPs) in China’s ecological governance during the study period. Effective environmental governance requires market mechanisms and competitive participation, whereby governments typically conduct competitive bidding processes to select optimal corporate partners. This market-driven approach reduces transaction costs and optimizes resource allocation, thereby enhancing PPP efficiency in ecological management. However, this positive relationship exhibits diminishing marginal returns, reaching an optimal threshold beyond which excessive marketization inversely impacts governance efficiency. Notably, the marketization coefficient shows statistically significant decline post-2016 (p < 0.05), suggesting the attainment of a critical marketization level that initiates efficiency reduction in China’s environmental PPPs.

Technical innovation inputs (Tech) demonstrate a progressively strengthening positive correlation with the governance efficiency of China’s public-private environmental partnerships during the study period. This temporal intensification confirms technological advancement as a pivotal catalyst for enhancing collaborative governance effectiveness. Specifically, these inputs facilitate synergistic government-enterprise alliances that integrate green technologies, industrial restructuring, and sustainable investment mechanisms. Such integration enables systematic greening of traditional industries, thereby achieving effective pollution control and generating cumulative improvements in environmental governance outcomes.

Industrialization intensity (Ind) exhibits a progressively intensifying adverse impact on the governance efficiency of China’s public-private environmental partnerships during the study period. Regions with higher industrialization indexes predominantly contain pollution-intensive industries, which during collaborative governance processes tend to escalate energy consumption and compound environmental degradation. This dual burden of legacy and emerging pollution creates path-dependent governance challenges, ultimately diminishing partnership efficacy in ecological management. Furthermore, the industrialization coefficient displays a non-linear trajectory characterized by initial contraction (2010–2013) followed by sustained expansion (2014–2020), statistically significant at p < 0.01 level. This pattern confirms the persistent spillover effects of industrial intensity on environmental governance systems, aligning with the Environmental Kuznets Curve hypothesis.

China’s openness index (Open) demonstrates a statistically significant positive correlation (β = 0.32, p < 0.01) with the governance efficiency of environmental public-private partnerships (PPPs) during the study period. International trade engagement serves dual functions: stimulating economic growth while concurrently enhancing collaborative governance capacity through technology transfer and environmental standard diffusion. Furthermore, the stability of openness coefficients (SD = 0.08) over time reflects sustained capital inflows and institutional learning effects. These transnational resource flows facilitate PPPs by bridging government fiscal constraints with corporate technological advantages, ultimately generating synergistic governance improvements in ecological management.

Technological advancement (lnsci) demonstrates a progressively intensifying positive correlation (β = 0.42, p < 0.01) with the governance efficiency of China’s environmental public-private partnerships (PPPs). Innovation-driven green production enables enterprises to achieve pollution-intensity decoupling through efficiency-enhancing technological solutions. In environmental governance systems, governments strategically incentivize corporate R&D investments via policy instruments like tax credits (averaging 15% R&D expenditure since 2015) and innovation subsidies. This institutional arrangement creates technological spillover effects, as evidenced by the sustained upward trajectory of the technology coefficient (CAGR 6.7% 2010–2020). The expanding coefficient magnitude confirms the growing catalytic role of technological progress in PPP efficiency enhancement, validating government-led innovation promotion strategies.

Environmental carrying capacity (Env) demonstrates a progressively intensifying adverse impact (β = −0.28, p < 0.05) on the governance efficiency of China’s environmental public-private partnerships (PPPs). Elevated environmental thresholds correlate with accelerated accumulation of industrial effluents and municipal solid waste (MSW), averaging 12.7% annual growth in monitored cities. While urban ecosystems exhibit enhanced resilience initially, surpassing critical capacity thresholds (0.82 on standardized index) triggers systemic rebound effects. Post-threshold analysis reveals 38% reduction in PPP formation rates, as municipalities prioritize autonomous pollution control mechanisms over collaborative governance frameworks during transitional phases (2015–2020).

To visualize the spatiotemporal heterogeneity of key determinants, this study identifies significant drivers (p < 0.05) influencing the efficiency of China’s environmental public-private partnerships (PPPs). Utilizing Geographically and Temporally Weighted Regression (GTWR) coefficients derived from ArcGIS 10.2 analyses, we generated comparative spatial distribution mappings for 2010 and 2021 (Figure 7). This methodological framework effectively captures the evolving spatial non-stationarity of PPP efficiency determinants across two critical policy implementation phases.

Level of economic development (LNGDP). In 2010, the level of economic development in most provinces had moderate-to-weak positive effects on the governance efficiency of government-enterprise cooperation in China’s ecological environment. In terms of spatial distribution, the positive influence of economic development level was mainly located in the eastern, central, and select western regions, while the negative influence was predominantly concentrated in western and northeastern areas. This indicates that at this stage, the economic development level of most provinces remained suboptimal, and governments’ capacity to invest in ecological-environmental management was constrained, necessitating corporate capital infusion and collaborative project implementation. As dual actors in green development, energy regulation, and ecological construction, government and enterprise entities achieved resource optimization through strategic planning and economic governance, thereby progressively enhancing the efficiency of China’s ecological environment government-enterprise cooperation. By 2021, provinces demonstrating moderate or weak positive impacts began to decline. Spatially, limited positive-impact zones persisted in eastern, central, and isolated western regions, with most provinces transitioning to negative-impact status relative to 2010. This shift primarily stems from enhanced governmental investment capacity in ecological management accompanying economic growth, reduced corporate collaboration incentives, and the emergence of regional protectionism manifesting through market access restrictions. Consequently, the economic development level’s positive impact on governance efficiency has entered a phase of diminishing marginal returns.

Technology investment (TECH): R&D investment has a positive effect on the governance efficiency of China’s eco-environmental government-enterprise co-operation on the whole, but there are regional differences in the promotion effect. In 2010, the level of technology investment in most regions had a positive effect on the governance efficiency of China’s eco-environmental government-enterprise co-operation, which was mainly due to the low nationwide baseline of eco-environmental governance efficiency during that period. The investment in scientific research promoted the improvement of enterprise pollution control technology, providing enhanced technical support for government-enterprise cooperation and consequently improving governance efficiency. By 2021, most provinces showed negative impacts of technology investment on governance efficiency, with only eastern coastal provinces maintaining weak positive effects. This shift primarily occurred because, despite increased technical investment in some provinces alongside economic development, technological innovation reached stagnation thresholds where enterprise innovations could no longer adapt to rapidly evolving ecological pollution challenges, thereby limiting the effectiveness of technological investments in China’s ecological environment government-enterprise co-operation governance efficiency.

Openness to the outside world (OPEN): In 2010, the openness level in most provinces negatively impacted the governance efficiency of government-enterprise cooperation in China’s ecological environment. In spatial distribution, positive impacts clustered in select central-western regions, while negative effects concentrated in eastern and northeastern zones. This reflects limited provincial openness during this phase, where foreign trade minimally influenced environmental governance, primarily reliant on governmental funding with scarce cross-sector collaboration. By 2021, provinces demonstrating moderate/weak positive impacts increased. Spatially, negative-impact regions narrowed to Inner Mongolia, Shanxi, Guangxi, and Hainan, contrasting with majority provinces showing improved outcomes. This reversal stems from accelerated economic globalization and regional integration expanding provincial openness, frequent interregional capital flows, and consequent efficiency gains in environmental governance partnerships.

Environmental carrying capacity (ENV): In 2010, provincial environmental carrying capacities predominantly inhibited the governance efficiency of government-enterprise cooperation in China’s ecological environment. Spatially, ENV’s inhibitory effect displayed a north–south attenuation pattern. This suggests provincial ecosystems could autonomously assimilate pollution loads during this phase, where enhanced environmental self-purification capacities reduced dependency on cross-sector collaborations—particularly domestic government-enterprise partnerships—for ecological governance. By 2021, provinces demonstrating strong positive ENV impacts increased, concentrated in Northeast China’s industrial heartlands. Most regions retained negative correlations. The Northeast’s industrial pollution intensity exceeds local environmental assimilation capacities, necessitating robust government-enterprise partnerships for effective remediation. Spatially, environmental carrying capacity maintains inhibitory effects on China’s ecological governance collaboration efficiency.

Qualitative Comparative Analysis (QCA), as a small and medium-sized sample case study tool, reveals the inner laws of complex phenomena by constructing causal mapping relationships between conditional variables and outcomes (52). The method is based on the theory of set theory, combined with the principle of Boolean algebra operation, through the structured processing of case information to form a coding system that can be quantitatively compared, and ultimately through the truth table operation to obtain the regular conclusions of the combination of the key conditions (53). Compared with traditional causal analysis, this technique is particularly suitable for compound research topics where multiple factors act synergistically (54).

This study focuses on the synergistic mechanism of the drivers of government-enterprise cooperative governance in China’s ecological environment, a type of composite public management problem with typical nonlinear characteristics, the causes of which often involve the interaction of multidimensional elements rather than being determined by a single variable. In this research context, the QCA method can not only maintain the advantage of in-depth analysis of case studies, but also reveal the common patterns among different cases through group analysis. In particular, the fuzzy set improvement technique (fsQCA) is adopted, which breaks through the traditional binary variable restriction, adopts the affiliation interval value to characterize the variable attributes, realizes the effective unification of qualitative and quantitative analyses, and provides an innovative solution to deal with the influencing factors characterized by continuous changes (55).

At the level of technical implementation, this study constructs a two-stage analysis framework: firstly, key variables are extracted based on the in-depth parsing of typical cases, and then a multifactor role model is established through group operations. Different from the conventional literature induction method, this research path rooted in the field of practice can more accurately capture the dynamic elements in the real governance situation. In terms of the variable selection mechanism, the method constructs a dual-path selection model: one is to establish a basic variable pool through the systematic combing of theoretical literature, and the other is to refine the contextualized variable dimensions by relying on empirical case studies (56). This study adopts the latter to construct an analytical framework to ensure that the research variables are deeply coupled with local governance practices.

In the previous section, the drivers of governance efficiency in China’s ecological environment government-enterprise partnerships have been analyzed independently. The subsequent section will perform interaction effect analysis on these drivers to examine the synergistic effects produced by factor combinations.

Prior to conducting the configuration analysis, necessity analysis of individual condition variables is required. This process primarily serves to determine whether any single variable constitutes a necessary condition for influencing the outcome variable. The necessity analysis results for governance efficiency condition variables, obtained through fsQCA 3.0 software (see Table 7), demonstrate that no condition variable achieved a consistency score of 0.9. This indicates that individual variables cannot be considered necessary conditions for governance efficiency outcomes. Notably, the environmental carrying capacity (ENV) variable showed a consistency value of 0.885, falling within the 0.8–0.9 range that may be considered indicative of quasi-sufficient conditions.

These findings suggest the existence of multiple concurrent condition relationships affecting governance efficiency in China’s ecological environment government-enterprise collaborations. Further sufficiency analyses are required to identify the specific configuration pathways that influence governance effectiveness in these partnerships.

By conducting fsQCA 3.0 analysis, three configurations of solutions (simple, complex, and intermediate) were derived to explain governance efficiency in China’s ecological environment government-enterprise partnerships. Drawing on methodological frameworks from prior studies (57), and considering the outcome variables’ theoretical implications, this study selects the intermediate solution—generated under full conditional presence—as the primary analytical focus (see Table 8).

Table 8 delineates five distinct configurations (C1–C5) affecting governance efficiency. The overall solution demonstrates acceptable consistency (0.852) and moderate coverage (0.513). All configurations exceed the 0.85 consistency threshold, confirming their status as sufficient condition combinations.

Notably: C1 (coverage = 0.293) prioritizes technological advancement and R&D investment levels; C2 (coverage = 0.252) centers on population density and technological advancement; C3–C5 (coverage = 0.169–0.119) share population density, R&D investments, and technological advancement as core conditions.

Collectively, these configurations identify population dynamics, innovation capacity, and resource allocation as critical determinants of governance efficacy. The unique coverage metric further highlights C1 (0.163) and C2 (0.114) as the most effective explanatory models, accounting for 16.3 and 11.4% of observed variance, respectively.