Analysis of the Inverted “U” Relationship Between R&D Intensity and Green Innovation Performance: A Study Based on Listed Manufacturing Enterprises in China

Abstract

Environmental innovation represents a pivotal pathway toward achieving energy efficiency improvements, carbon footprint reduction, and ecological sustainability enhancement. The research investigates Chinese manufacturing enterprises listed on domestic stock exchanges throughout 2011–2023. The analytical framework utilizes count-based regression methodologies to explore how R&D investment intensity influences eco-innovation capabilities. Results demonstrate curvilinear associations linking R&D expenditure levels with both substantive and strategic environmental innovation achievements across industrial firms. This outcome successfully passed the turning-point test. Environmental oversight and financial incentives produce divergent moderating influences on innovation trajectories. Regulatory frameworks generate restrictive impacts through narrowing optimal investment ranges and dampening peak innovation outputs, whereas fiscal support mechanisms foster expansive effects via broadening resource availability and amplifying achievement levels. Cross-sectional examination uncovers substantial variations among ownership categories and geographical locations. State-owned enterprises demonstrate significantly lower optimal R&D intensity thresholds. Private firms require substantially elevated thresholds for optimal performance. Inland territories manifest unbalanced innovation dynamics. Coastal areas exhibit symmetric innovation patterns. The research enriches empirical knowledge in eco-innovation studies while offering context-specific strategic insights. The findings establish theoretical foundations and practical guidance for policy architects designing integrated environmental management systems that enhance innovation capabilities.

1. Introduction

Global focus on environmental governance has intensified substantially throughout recent decades. Responding to these circumstances, nations have established significant carbon mitigation pledges while deploying comprehensive response strategies. Within the developing sphere of Earth system governance, establishing robust governance frameworks has emerged as the foremost priority [1]. This holds comparable significance for China, currently undergoing development processes. Realizing resource efficiency, pollution mitigation, and ecological preservation represents a substantial challenge that China will encounter throughout the forthcoming years.

Industrial enterprises face urgent demands for ecological transition due to intensive resource utilization and significant pollutant discharge. Research findings indicate that environmental technology advancement strengthens organizational sustainability practices and elevates ecological performance metrics [2,3]. Green innovation activities generate favorable impacts on enterprise profitability [4,5]. Adopting environmentally oriented strategic approaches allows organizations to reduce pollution outputs while promoting conservation initiatives. Nevertheless, achieving such advantages demands sufficient financial commitment. Organizations also need favorable regulatory environments and robust internal structures.

Environmental innovation has become a central theme within scholarly discourse. Green innovation encompasses strategic activities that balance ecological responsibility with financial viability. These activities require extensive funding commitments, prolonged return periods, and substantial technical uncertainties. Such factors create intricate resource management difficulties for organizations. They also generate complex operational planning challenges. [6]. Research demonstrates that environmental technology investments markedly improve organizational innovation capabilities, although this association varies considerably among different enterprise categories [7]. Technology-intensive sectors, including artificial intelligence, illustrate how research investment intensity stimulates creative capacity development [8]. Such findings indicate that intensified investigative activities can promote comprehensive innovation advancement across industries.

Existing research examines environmental innovation catalysts and their organizational performance impacts. This investigation utilizes diverse conceptual frameworks. These include resource capability perspectives, institutional analysis perspectives, and stakeholder relationship models. Field studies reveal that green innovation activities strengthen enterprise outcomes. These activities operate as protective mechanisms against operational risks. They consequently create varied value opportunities for organizations. Such opportunities benefit enterprises implementing sustainability initiatives [9]. Environmental innovation functions as a transformation driver for ecological advancement among high-pollution sectors. This function broadens scholarly insights into operational dynamics. Such insights enhance understanding of complex innovation mechanisms [10].

The research offers three key contributions to scholarly knowledge. From a methodological perspective, the investigation utilizes count-based regression techniques for examining environmental innovation patent application data. This approach resolves data dispersion challenges that traditional ordinary least squares and panel estimation methods fail to address effectively. Regarding theoretical advancement, the analysis enhances comprehension of curvilinear R&D–green innovation associations through incorporating polynomial R&D intensity variables. Such modeling uncovers distinct response patterns among substantive and strategic green innovation categories while filling analytical voids in current policy mechanism studies. Concerning practical applications, cross-sectional examination demonstrates substantial disparities among geographical areas and enterprise classifications. These findings establish evidence-based foundations for targeted policy formulation and deliver actionable guidance for developing nations. Such guidance supports nations implementing ecological transition approaches that resemble China’s industrial transformation experience.

2. Research Hypotheses

2.1. R&D Investment and Green Innovation Performance

Within China’s “dual carbon” policy framework, industrial enterprises demonstrate growing emphasis on green innovation initiatives. Scientific research and technological development constitute fundamental catalysts for advancement and ecological innovation. Nevertheless, the association between R&D intensity and organizational innovation capabilities reveals substantial intricacy [11]. This complexity stems from various contributing factors.

Contemporary research findings indicate that scientific investment encourages environmental technological advancement while supporting ecological sustainability goals [12,13]. Drawing from such evidence, R&D intensity produces varied impacts on organizational performance across diverse circumstances [14]. Among industrial enterprises, strengthened R&D investment intensity enhances innovation capabilities [15]. Via R&D enhancement activities, organizations can tactically redistribute internal assets toward environmental technology advancement and ecological product development. Nevertheless, such beneficial associations display declining marginal benefits past specific turning points. Intellectual property creation among enterprises correlates significantly with their selected R&D approaches and asset distribution frameworks [16]. Over-investment in scientific activities conversely diminishes innovation effectiveness. These conceptual perspectives indicate that R&D intensity and green innovation performance potentially demonstrate curvilinear associations.

Such adverse associations can be interpreted via three conceptual frameworks. Initially, absorptive capacity threshold effects reveal that organizational knowledge absorption demonstrates accumulative properties. However, surpassing critical limits induces cognitive saturation. This consequently reduces enterprise effectiveness in incorporating external knowledge while hindering innovation endeavors [17]. Subsequently, managerial complexity embedded within dynamic capabilities potentially disrupts innovation mechanisms. Excessive R&D investment amplifies administrative and coordination responsibilities linked to resource integration activities. When R&D investment exceeds organizational management thresholds, negative impacts on innovation processes emerge [18]. Ultimately, diminishing marginal utility principles in asset allocation suggest that R&D investment beyond optimal distribution points creates resource fragmentation effects. These effects conflict with core assumptions regarding asset heterogeneity and uniqueness within Resource-Based View frameworks. This undermines green innovation performance [19].

Recent years have witnessed substantial advancement in green innovation research across theoretical depth and methodological approaches. Some researchers examined urban carbon emission reduction mechanisms through digital economy perspectives [20]. Their findings indicate that digital economies can effectively mitigate urban carbon emissions via green synergy effects of industrial agglomeration. Such insights provide novel theoretical perspectives for comprehending green innovation-driven mechanisms within contemporary economic contexts. Concerning measurement methodology, researchers have explored ecological innovation pathway mechanisms via sustainable development frameworks [21]. Such research highlights the pivotal role of green innovation within economic transformation processes. This analysis reveals that ecological innovation encompasses more than technological advancement expressions. It serves as an essential catalyst for attaining harmonized development between economic expansion and environmental preservation.

Contemporary academic research has advanced differentiation among varied forms of environmental innovation. Specifically, innovation endeavors focused primarily on developing authentic technological competencies within organizations constitute substantive green innovation. Conversely, activities directed toward establishing credibility and competitive positioning represent strategic green innovation [22]. Substantive green innovation performance markedly strengthens enterprise financial outcomes. In contrast, strategic green innovation performance predominantly enhances ecological achievements [23]. Research investigations analyzing both innovation categories reveal that R&D investment produces curvilinear impacts on substantive green innovation performance [24]. Such evidence indicates an inverted U-shaped association between organizational R&D intensity and environmental innovation outputs. Drawing from these observations, the current analysis proposes the following:

H1. 

Manufacturing enterprises exhibit inverted U-shaped associations linking R&D intensity with both substantive green innovation performance and strategic green innovation performance.

2.2. The Moderating Effect of Environmental Regulation

External institutional frameworks shape organizational innovation conduct while producing varied regulatory impacts across diverse settings. Elements including institutional contexts function to moderate how different catalyst forces affect green innovation performance [25,26]. Environmental regulations are fundamentally connected to enterprise green innovation performance activities within existing policy structures. However, scholarly discourse continues concerning the efficacy of environmental regulations in enhancing green innovation performance [10,27].

Environmental policy-innovation associations demonstrate essential curvilinear characteristics that question standardized regulatory approaches. The Porter Hypothesis suggests that well-structured regulations initiate innovation offset mechanisms. However, theoretical advancements highlight crucial regulatory intensity limits beyond which declining benefits occur [28,29]. Modern analytical frameworks acknowledge that environmental regulations function within performance boundaries. Under such conditions, excessive regulatory intensity potentially inhibits rather than encourages innovation outcomes [30].

Conceptual mechanisms governing regulatory impacts function circumstantially via three channels. Initially, innovation compensation processes operate efficiently solely within ideal R&D intensity parameters. Beyond such boundaries, resource misallocation and strategic inflexibility tend to prevail [31]. Subsequently, institutional pressure systems alter organizational conduct through enforcing influences, standardizing influences, and imitative influences. Nevertheless, such effectiveness relies on institutional capabilities and regulatory architecture quality [32]. Finally, stakeholder pressure transmission coordinates regulatory demands with external anticipations. However, mechanism performance fluctuates among different institutional settings [33]. These institutional elements—governance structures, regulatory structures, and market structures—generate intricate, situation-specific effects on innovation behavior. Such complexities necessitate empirical analysis to determine optimal regulatory arrangements [25,34,35]. Consequently, this analysis proposes the following:

H2. 

Environmental regulation moderates both substantive and strategic green innovation performance in manufacturing enterprises.

2.3. The Moderating Effect of Environmental Protection Subsidies

Environmental subsidies function via essentially distinct pathways compared to regulatory approaches. These mechanisms serve as resource enhancement rather than compliance enforcement to coordinate private motivations with ecological goals [36]. Subsidies broaden organizational innovation resource limits while diminishing innovation uncertainties. This potentially facilitates continuous high-intensity R&D investments without regulatory restrictions. Nevertheless, subsidy effectiveness relies fundamentally on receiving organizations’ absorptive capacity and institutional infrastructure. These capabilities determine successful transformation of financial support into innovation capabilities [37].

Conceptual models indicate that subsidies display curvilinear benefits dependent on organizational characteristics and program design features. Although subsidies conceptually strengthen innovation via resource enhancement and uncertainty reduction, threshold phenomena potentially arise when support intensity surpasses enterprises’ effective utilization capacity. This generates declining marginal benefits or asset misdirection [38]. Optimal subsidy effectiveness consequently fluctuates considerably among enterprise categories, geographical contexts, and institutional settings. Such variations necessitate empirical examination of these diverse impacts.

The resource complementarity mechanism explains subsidies’ positive innovation impacts through synergistic rather than substitute effects with private investment [39,40]. Appropriately structured subsidies establish amplifying associations that allow organizations to implement more aggressive R&D strategies beyond inherent financial limitations. This generates beneficial spillover impacts that enhance comprehensive innovation capacity. Such complementarity effects function most efficiently when coordinated with sufficient institutional absorptive capacity. This coordination enables successful transformation of financial support into innovation outcomes.

Subsidies establish beneficial innovation motivations that coordinate private financial interests with ecological goals. This generates reinforcing patterns where improved performance validates ongoing support. Contrasting with regulatory compliance pressures, such resource enhancement mechanisms function via direct provision and uncertainty reduction. These create enduring innovation systems demonstrating varied effectiveness among different institutional contexts. Consequently, the analysis proposes:

H3. 

Environmental subsidies moderate substantive green innovation performance in manufacturing enterprises.

3. Research Design

3.1. Selection of the Sample

The research investigated Chinese A-share industrial enterprises traded between 2011 and 2023 to analyze how R&D investment intensity influences green innovation performance. Information was obtained from multiple databases including China Stock Market & Accounting Research (CSMAR), China Research Data Service Platform (CNRDS), and Wind financial systems. Data quality was ensured through a dual-stage filtering methodology. Initially, organizations with special treatment designation (ST and *ST) were removed due to irregular financial circumstances. Subsequently, enterprises lacking essential variable information were excluded from analysis. Following these filtering procedures, the resulting dataset contained 14,457 enterprise-year observations. Winsorization was implemented at the 1st and 99th percentiles to reduce outlier effects. Statistical computations were conducted using Microsoft Excel and Stata 18.0 software packages.

3.2. Variable Definitions

The research utilized green innovation performance as the outcome variable while treating R&D investment intensity as the primary explanatory factor. The analytical framework additionally included multiple control variables. These encompass capital concentration, debt-to-assets ratio, financial leverage, and return on assets.

Table 1. Variable definition.

Category	Variable Name	Symbol	Measurement	Reference	Data Source
Dependent Variables	Substantive green innovation performance	GI	Number of green invention patent applications	[24,41]	CNRDS
	Strategic green innovation performance	GU	Number of green utility model patent applications		CNRDS
Independent Variable	R&D investment intensity	RD	R&D investment/revenue	[8,10,14]	CSMAR
Regulated variables	Environmental regulation	ERI	Pollution control amount/industrial output value	[42]	CSMAR
	Environmental subsidies	SUB	Natural logarithm of environmental protection subsidy values	[43]	WIND
Control variables	financial leverage	SLEC	Total liabilities of the enterprise/total assets	[24]	CSMAR
	enterprise age	AGE	The number of years since the establishment of the company is taken as the natural logarithm		CSMAR
	Return on assets	ROA	Net income divided by total asset value.		CSMAR
	capital-intensity	CC	Total asset value of the company/operating income		CSMAR
	Cash flow ratio	CASH	Total cash assets/assets		CSMAR
	Enterprise Ownership	SOE	For state-owned, it is 1; otherwise, it is 0		CSMAR
	Equity concentration	SC	Proportion of the top ten shareholders		CSMAR

provides comprehensive details regarding these variables.

3.2.1. Dependent Variable

The outcome variables comprise substantive green innovation performance (GI) and strategic green innovation performance (GU). Strategic green innovation performance (GU) was quantified through the number of green utility model patent applications filed during the observation period. Conversely, substantive green innovation performance (GI) was assessed via the number of green invention patent applications filed by organizations [41].

Certain limitations exist when employing patent categories to differentiate substantive green innovation from strategic green innovation. Some researchers have noted that particular utility model patents may encompass substantive technical improvements. Such classification approaches cannot entirely eliminate conceptual overlap. The same patent may simultaneously possess both substantive and strategic value [44]. Nevertheless, this differentiation method maintains theoretical validity. China’s patent system characteristics cause invention patents and utility model patents to exhibit distinct features. Elevated standards for invention patents increase their likelihood of representing substantive technological breakthroughs. The accessibility of utility model patents renders them more appropriate for rapidly establishing technological barriers. Large-sample investigations allow such systematic differences to effectively capture various dimensions of green innovation.

3.2.2. Independent Variables

The central explanatory variable is R&D investment intensity (RD). This indicator captures the ratio of R&D spending to total operational income. The calculation follows established methodological practices. [45]. R&D intensity functions as the central independent variable within the empirical framework.

Environmental regulation (ERI) and environmental subsidies (SUB) constitute the moderating factors. Environmental regulation intensity (ERI) is constructed as the proportion of annual pollution control investment within each enterprise’s geographical area to regional total industrial output value, consistent with established approaches [42]. This proportion is presented in per-thousand units to enhance interpretive clarity. China’s distinctive institutional context creates challenges for acquiring environmental protection data at the enterprise level. Consequently, environmental regulations measurement is conducted at the geographical level. Environmental subsidies are quantified via the subsequent methodology. Information is sourced from subsidy-related statements within enterprises’ annual documentation. Environment-focused terms, encompassing “green,” “environmental subsidies,” “environmental protection,” and “energy conservation,” are located and compiled to determine aggregate annual environmental subsidies obtained by each enterprise [43]. These figures undergo logarithmic transformation to mitigate distributional asymmetry.

3.2.3. Control Variables

To guarantee reliable empirical outcomes, the investigation includes the subsequent control factors. Financial Leverage (SLEC) is computed as the proportion of total liabilities to total assets. Firm age (AGE) represents the natural logarithm of years since enterprise establishment. This transformation addresses potential heteroskedasticity concerns. Return on assets (ROA) equals the proportion of net income to total assets. Capital Intensity (CC) represents the proportion of total assets to operational revenue. Cash ratio (CASH) denotes cash and cash equivalents divided by total assets. State ownership (SOE) receives binary coding with 1 indicating state-owned enterprises and 0 representing alternative ownership structures. Equity Concentration (SC) captures the percentage of shares possessed by the leading ten shareholders [24].

3.3. Model Section

Preliminary descriptive statistics reveal that the variance of the dependent variables exceeds their respective means, indicating overdispersion in the data. Given this overdispersion, the analysis employs a negative binomial regression model, which is appropriate for count data with variance exceeding the mean [46]. Regarding model selection, this investigation does not employ the zero-inflated negative binomial model. Several reasons justify this decision. The zero-inflated negative binomial model demonstrates instability and susceptibility to multicollinearity issues. Its parameter estimation frequently encounters convergence failures. To guarantee research accuracy, this study consistently utilizes the negative binomial model for all analyses. To address unobserved heterogeneity, industry-fixed effects and year-fixed effects are incorporated across all model specifications. The model is set as follows:

$${GI}_{ij}/{GU}_{ij}={\beta }_0+{\beta }_1{RD}_{ij}{+\beta }_2{{RD}_{ij}}^2+{\beta }_3\sum {controls}_{ij}+{\epsilon }_{it}$$

(1)

To test Hypotheses 2 and 3 regarding the moderating effects of environmental regulation and environmental subsidies, interaction terms between ERI, SUB, and RD are incorporated following established methodology [47]. Models 2 and 3 are specified as follows:

$${\mathrm{G}\mathrm{I}}_{\mathrm{i}\mathrm{j}}/{\mathrm{G}\mathrm{U}}_{\mathrm{i}\mathrm{j}}={\beta }_0+{\beta }_1{\mathrm{R}\mathrm{D}}_{\mathrm{i}\mathrm{j}}+{\beta }_2{{{{\mathrm{R}\mathrm{D}}_{\mathrm{i}\mathrm{j}}}^2+\beta }_3{\mathrm{R}\mathrm{D}}_{\mathrm{i}\mathrm{j}}\ast \mathrm{E}\mathrm{R}\mathrm{I}+{\beta }_4{{\mathrm{R}\mathrm{D}}_{\mathrm{i}\mathrm{j}}}^2\ast \mathrm{E}\mathrm{R}\mathrm{I}+\beta }_3\sum {\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i}\mathrm{j}}+{\epsilon }_{\mathrm{i}\mathrm{t}}$$

(2)

$${\mathrm{G}\mathrm{I}}_{\mathrm{i}\mathrm{j}}/{\mathrm{G}\mathrm{U}}_{\mathrm{i}\mathrm{j}}={\beta }_0+{\beta }_1{\mathrm{R}\mathrm{D}}_{\mathrm{i}\mathrm{j}}+{\beta }_2{{{{\mathrm{R}\mathrm{D}}_{\mathrm{i}\mathrm{j}}}^2+\beta }_3{\mathrm{R}\mathrm{D}}_{\mathrm{i}\mathrm{j}}\ast \mathrm{S}\mathrm{U}\mathrm{B}+{\beta }_4{{\mathrm{R}\mathrm{D}}_{\mathrm{i}\mathrm{j}}}^2\ast \mathrm{S}\mathrm{U}\mathrm{B}+\beta }_3\sum {\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i}\mathrm{j}}+{\epsilon }_{\mathrm{i}\mathrm{t}}$$

(3)

In these models, ERI denotes environmental regulation intensity, while RD * ERI and RD² * ERI capture the moderating effects of environmental regulation on the linear and quadratic terms of R&D intensity, respectively. Similarly, SUB represents environmental subsidies, with interaction terms RD * SUB and RD² * SUB measuring the moderating effects of subsidies on R&D intensity. The coefficients of these interaction terms in Models 2 and 3 capture the magnitude and direction of the moderating effects.

4. Empirical Findings

4.1. Descriptive Statistics

Table 2. Descriptive statistics.

Variables	Count	Mean	Sd	Min	Max
GI	14,457	0.872	2.929	0.000	21.000
GU	14,457	0.768	2.372	0.000	16.000
RD	14,457	4.821	3.783	0.080	22.650
ERI	14,457	0.015	0.015	0.001	0.083
SUB	14,457	4.067	6.181	0.000	16.966
CASH	14,457	0.811	1.203	0.023	7.710
CC	14,457	2.134	1.273	0.475	8.217
SLEC	14,457	0.393	0.189	0.059	0.868
AGE	14,457	1.831	0.956	0.000	3.258
SOE	14,457	0.262	0.440	0.000	1.000
SC	14,457	0.582	0.143	0.246	0.878
ROA	14,457	0.039	0.065	−0.257	0.202

Note: GI denotes the number of green invention patent applications; GU denotes the number of green utility model patent applications.

reports summary statistics for all study variables. The outcome variables exhibit variance measures surpassing their corresponding means, indicating overdispersion characteristics within the dataset. This observation validates the application of negative binomial regression methodology. Such techniques are particularly suited for analyzing overdispersed count observations.

4.2. Correlation Matrix

Table 3. Correlation coefficient table.

Variables	GU	GI	RD	ERI	SUB	CASH	CC	SLEC	AGE	SOE	SC	ROA
GU	1.000
GI	0.682 ***	1.000
RD	0.067 ***	0.086 ***	1.000
ERI	−0.031 ***	−0.025 ***	−0.173 ***	1.000
SUB	0.00400	−0.00300	−0.170 ***	0.168 ***	1.000
CASH	−0.065 ***	−0.058 ***	0.186 ***	−0.024 ***	−0.097 ***	1.000
CC	−0.047 ***	−0.060 ***	0.380 ***	0.016 *	−0.074 ***	0.223 ***	1.000
SLEC	0.139 ***	0.145 ***	−0.216 ***	0.062 ***	0.114 ***	−0.564 ***	−0.161 ***	1.000
AGE	−0.00700	0.021 **	−0.199 ***	0.100 ***	0.137 ***	−0.276 ***	−0.072 ***	0.365 ***	1.000
SOE	0.041 ***	0.079 ***	−0.137 ***	0.158 ***	0.128 ***	−0.122 ***	−0.099 ***	0.259 ***	0.445 ***	1.000
SC	0.016 *	0.0100	−0.00500	−0.055 ***	−0.057 ***	0.161 ***	−0.068 ***	−0.181 ***	−0.461 ***	−0.125 ***	1.000
ROA	0.014 *	0.023 ***	−0.062 ***	−0.035 ***	−0.027 ***	0.213 ***	−0.261 ***	−0.390 ***	−0.193 ***	−0.099 ***	0.245 ***	1.000

Note: The table contains the correlation coefficient, * is used to indicate p < 0.1, ** to indicate p < 0.05, and *** to indicate p < 0.01. GI denotes the number of green invention patent applications; GU denotes the number of green utility model patent applications.

displays the correlation matrix encompassing all study variables. Findings reveal that pairwise correlation coefficients remain below 0.7, while variance inflation factor (VIF) statistics for all variables stay below 10. These diagnostic assessments confirm minimal multicollinearity issues within the analytical models.

4.3. Benchmark Regression

The investigation employs a two-stage methodological framework to explore curvilinear associations [48]. The explanatory factor, R&D intensity, is categorized into low, medium, and high classifications. The turning point of the curvilinear relationship aligns with the medium classification. Low and high classifications occupy the left and right portions, respectively. The medium classification signifies the optimal R&D intensity threshold. This structure investigates how R&D intensity movements from low to optimal thresholds affect green innovation performance. Additionally, it examines transitions from optimal to elevated intensity thresholds. Within this structure, green innovation performance is initially conceptualized as a discrete measure. To strengthen analytical reliability, green innovation performance is subsequently conceptualized as a continuous measure after coefficient derivation. This methodological strategy facilitates the development of curvilinear patterns from regression outputs. Such visualization enhances interpretive clarity of empirical associations.

First, a Hausman test is performed to identify the suitable panel estimation approach. Test outcomes reveal a p-value substantially below 0.01, establishing that fixed-effects specifications are superior to random-effects alternatives. Estimation results are displayed in

Table 4. Benchmark regression.

Variables	(1) GI	(2) GI	(3) GU	(4) GU
RD	0.148 *** (3.68)	0.253 *** (7.22)	0.101 *** (2.88)	0.158 *** (4.58)
RD2	−0.005 *** (−3.03)	−0.007 *** (−4.67)	−0.004 *** (−2.63)	−0.005 *** (−3.16)
CASH		0.051 (1.24)		−0.009 (−0.24)
CC		−0.084 ** (−1.74)		−0.075 * (−1.78)
SLEC		3.811 *** (11.66)		2.744 *** (8.71)
AGE		−0.126 ** (−2.02)		−0.194 *** (−3.13)
SOE		0.497 *** (3.61)		0.317 ** (2.44)
SC		0.071 (0.19)		−0.346 (−0.89)
ROA		5.651 *** (8.46)		4.678 *** (6.80)
Industry-fixed effects	YES	YES	YES	YES
Year-fixed effects	YES	YES	YES	YES
Observed value	14,457	14,457	14,457	14,457
Pseudo R2	0.046	0.069	0.051	0.066
Prob > ${\mathrm{c}\mathrm{h}\mathrm{i}}^2$	0.0000	0.0000	0.0000	0.0000
Log Likelihood	−13,219.28	−12,901.47	−13,030.86	−12,830.97

Note: In the above table, * is used to indicate p < 0.1, ** to indicate p < 0.05, and *** to indicate p < 0.01. GI denotes the number of green invention patent applications; GU denotes the number of green utility model patent applications.

. This table contains four columns of regression estimates derived from Model (1). Columns (1) and (3) analyze the impacts of R&D investment intensity on substantive green innovation performance and strategic green innovation performance, respectively. Meanwhile, columns (2) and (4) display benchmark regression outcomes incorporating control variables.

Columns (1) and (2) indicate that, after controlling for industry-fixed effects and year-fixed effects, the linear R&D intensity coefficient shows positive significance within the substantive green innovation performance framework. Conversely, the quadratic coefficient demonstrates negative significance. Therefore, substantive green innovation performance displays adverse associations with R&D intensity beyond optimal thresholds. The turning point in column (1) reaches 14.2%, whereas the corresponding turning point for column (2) attains 18.0%. Columns (3) and (4) reveal that the linear R&D intensity coefficient maintains positive values while the quadratic component turns negative. The turning point in column (3) reaches 12.0%, while column (4) achieves 16.8%. These estimation outcomes demonstrate that when organizations initially pursue innovation activities and strengthen R&D intensity, green innovation performance shows corresponding enhancements. Nevertheless, as R&D intensity continues escalating beyond optimal turning points, declining marginal benefits emerge. Excessive accumulation of absorptive capacity reduces organizations’ capability to effectively integrate and utilize innovative knowledge. Elevated R&D intensity creates considerable administrative pressures on organizations. Marginal returns to R&D investment deteriorate gradually, causing resource misdirection and ineffective capital deployment.

The fundamental regression examination in Section 4.3 relies upon conceptual frameworks including absorptive capacity thresholds, dynamic capability complexities, and resource allocation principles. Empirical outcomes demonstrate that throughout China’s industrial sector ecological transition, balanced R&D intensity promotes organizational green innovation. Conversely, over-investment in research activities creates obstacles to innovation performance. These findings offer empirical confirmation for Hypothesis H1 developed in Section 2.1.

4.4. Analysis of Regulatory Effect

Following the benchmark regression analysis, this study examines the moderating effects on the relationship between R&D intensity and green innovation performance. The moderating variables are evaluated at their mean values plus and minus one standard deviation. These adjusted values are subsequently incorporated into the regression model to construct two comparative curves. These curves demonstrate the differential effects of R&D intensity on green innovation performance across varying levels of the moderating variables [49]. Since the curves differ only in their moderating variable values while maintaining constant coefficients for all other parameters, they provide robust evidence of the moderating effects on green innovation performance.

Table 5. Regulatory effects.

Variables	(1) GI	(2) GI	(3) GU	(4) GU
RD	0.287 *** (8.08)	0.273 ***	0.170 ***	0.168 ***
		(7.83)	(4.87)	(4.91)
RD2	−0.010 *** (−6.17)	−0.009 ***	−0.006 **	−0.005 ***
		(−5.72)	(−3.74)	(−3.67)
ERI	−6.300 (−1.55)		−2.966 (−0.81)
SUB		0.011 ** (1.96)		0.012 ** (2.10)
ERI × RD	6.921 *** (3.83)		2.991 *	1
			(1.70)
ERI × RD2	−0.512 ***		−0.212 **	1
	(−4.52)		(−2.08)
SUB × RD		0.005 −1.35		0.001
				(0.36)
SUB × RD2		−0.0004 ** (−2.39)		−0.0002
				(−1.04)
controlled variable	YES	YES	YES	YES
Industry-fixed effects	YES	YES	YES	YES
Year-fixed effects	YES	YES	YES	YES
Observed value	14,457	14,457	14,457	14,457
Pseudo R2	0.070	0.069	0.066	0.066
$\mathrm{P}\mathrm{r}\mathrm{o}\mathrm{b}>{\mathrm{c}\mathrm{h}\mathrm{i}}^2$	0.0000	0.0000	0.0000	0.0000
Log Likelihood	−12,882.93	−12,891.83	−12,825.88	−12,824.24

Note: In the above table, * is used to indicate p < 0.1, ** to indicate p < 0.05, and *** to indicate p < 0.01. GI denotes the number of green invention patent applications; GU denotes the number of green utility model patent applications.

reports regression estimates across four specifications. Columns (1) and (3) employ Model (2), while columns (2) and (4) utilize Model (3). Columns (1) and (2) analyze the moderating influences of environmental regulations and environmental subsidies on substantive green innovation performance. These columns present separate analyses for each policy instrument. Columns (3) and (4) explore the equivalent moderating impacts of environmental regulations and environmental subsidies on strategic green innovation performance. These specifications follow parallel analytical frameworks.

Initially, examine the findings displayed in column (1) of Table 5. The descriptive statistics in Table 2 show that the moderating variable “Environmental regulation” has a mean of 0.015 with a standard deviation of 0.015. The upper threshold of 0.029 (mean + one standard deviation) represents high environmental regulation intensity. Conversely, the lower threshold of 0.001 (mean–one standard deviation) corresponds to low environmental regulation intensity. When these threshold values are incorporated into the regression model, two distinct inverted U-shaped curves emerge.

Since the findings in columns (1) to (3) of Table 5 are similar to those in Table 4, detailed mathematical interpretations for individual columns are omitted. Figure 1 derives from the estimation outcomes of column (1). This visualization supplies convincing empirical support for the contingent characteristics of the Porter Hypothesis. The analysis reveals that environmental regulatory intensity functions within specific performance boundaries that substantially modify the R&D intensity-innovation association. Under low ERI circumstances, organizations display a strong inverted U-shaped pattern with a turning point at 31.2% and maximum performance of 1.18. This demonstrates that balanced institutional pressure allows the innovation compensation mechanism to operate efficiently [31]. Such patterns confirm Porter’s central argument that well-structured regulations encourage resource optimization without exceeding organizational capabilities. This enables continuous high-intensity R&D investment within enhanced strategic adaptability limits.

In contrast, under elevated ERI circumstances, the association becomes substantially compressed with an early turning point at 11.0% and markedly diminished peak performance of 0.29. This leftward shift demonstrates how excessive institutional pressure converts the innovation compensation mechanism into a resource limitation mechanism. Under such conditions, compliance enforcement supersedes resource enhancement. The optimal R&D intensity threshold has declined considerably (from 31.2% to 11.0%), reinforcing prior analytical conclusions. These observations advance Porter Hypothesis theory through identifying specific regulatory intensity boundaries beyond which stakeholder pressure transmission mechanisms produce adverse outcomes [33]. Instead of improving innovation effectiveness via regulatory direction, excessive pressure establishes institutional barriers. Such barriers compel premature optimization while constraining innovation investment timeframes.

Column (3) of Table 5 reveals that the Porter Hypothesis displays varied applicability among innovation categories. Strategic green innovation demonstrates increased responsiveness to regulatory intensity boundaries. Figure 2 illustrates these patterns under different ERI circumstances. Under low ERI circumstances, organizations display an inverted U-shaped pattern with a turning point at 22.4% and maximum performance of 0.40. This suggests that the innovation compensation mechanism functions efficiently for strategic innovations under balanced institutional pressure [29]. Under elevated ERI circumstances, the association becomes substantially compressed. The turning point occurs early at 12.0% with diminished peak performance of 0.26. This indicates that institutional pressure mechanisms influence strategic innovation via expedited resource redistribution effects [32].

These observations advance resource allocation theory by illustrating that regulatory pressure generates systematic portfolio redistribution effects. Organizations transfer resources from strategic exploration toward operational compliance under such pressures [33]. The varying impacts on strategic versus substantive innovation supply empirical support for innovation hierarchy theories. Regulatory constraints systematically influence resource allocation toward lower-risk, shorter-term innovation activities under these frameworks. These outcomes offer empirical confirmation for Hypothesis H2 developed in Section 2.2.

Column (2) of Table 5 demonstrates the impact of environmental subsidies on substantive green innovation performance. Descriptive statistics reveal an average of 4.067 with a dispersion measure of 6.181. This produces an upper boundary of 10.248 (mean + one standard deviation) indicating high environmental subsidy levels. Environmental subsidies maintain non-negative characteristics by nature. Therefore, the lower boundary is established at 0, signifying low subsidy intensity. These boundary values are then integrated into the regression framework for comparative examination.

The estimation outcomes derived from column (2) are displayed in Figure 3. This visualization supplies convincing empirical support for resource complementarity theory. Under elevated SUB circumstances, the turning point shows a rightward shift from 13.0% to 19.0%. This offers direct evidence that subsidies operate as resource enhancement mechanisms. These mechanisms broaden organizations’ innovation capacity limits beyond inherent financial constraints. Performance improvements at optimal investment thresholds empirically validate the beneficial spillover effects anticipated by resource complementarity theory. Regulatory mechanisms function through institutional pressure and compliance enforcement. In contrast, subsidies establish positive innovation motivations that coordinate private profit interests with environmental goals via direct resource provision [40].

These observations essentially differ from regulatory impacts. They validate theoretical distinctions between resource enhancement and institutional pressure mechanisms. Environmental regulations compress innovation associations through compliance obligations. Subsidies broaden innovation opportunities through resource liberation. The considerable enhancement in both optimal R&D intensity and peak performance indicates that subsidies function via absorptive capacity improvement. This creates enduring innovation systems that amplify organizational capabilities rather than constraining them. Environmental subsidies demonstrate statistically significant positive moderation of substantive green innovation performance. The rightward turning point shift and performance enhancement supply robust empirical support for hypothesis H3.

4.5. Robustness Test

To strengthen the reliability of these observations, the investigation performs supplementary validation examinations. Alternative indicators are utilized as robustness assessments for the outcome variables. The number of green utility model patents possessed by sample organizations functions as an alternative proxy for strategic green innovation performance. In contrast, the combined total of green invention patents and green utility model patents submitted by organizations represents the alternative indicator for substantive green innovation performance.

Table 6. Dependent variable substitution.

Variables	(1) GI	(2) GI	(3) GU	(4) GU
RD	0.237 *** (7.12)	0.233 ***	0.146 ***	0.148 ***
		(7.10)	(5.61)	(5.70)
RD2	−0.008 *** (−5.50)	−0.008 ***	−0.005 ***	−0.005 ***
		(−5.39)	(−4.71)	(−4.65)
ERI	−5.126 (−1.38)		−0.832 (−0.29)
SUB		0.011 ** (2.09)		0.012 *** (2.95)
ERI × RD	5.248 *** (3.27)		3.132 **
			(2.28)
ERI × RD2	−0.354 ***		−0.167 **
	(−3.74)		(−2.42)
SUB × RD		0.003 (0.79)		0.004
				(1.27)
SUB × RD2		−0.0003 * (−1.93)		−0.0003
				(−1.56)
controlled variable	YES	YES	YES	YES
Industry fixed effects	YES	YES	YES	YES
Time fixed effects	YES	YES	YES	YES
observed value	14,457	14,457	14,457	14,457
Prob > ${\mathrm{c}\mathrm{h}\mathrm{i}}^2$	0.0000	0.0000	0.0000	0.0000
Log Likelihood	−17,955.04	−17,960.59	−22,905.63	−22,900.33

Note: In the above table, * is used to indicate p < 0.1, ** to indicate p < 0.05, and *** to indicate p < 0.01. GI denotes the number of green invention patent applications; GU denotes the number of green utility model patent applications.

displays the corresponding outcomes.

Table 6 displays the outcomes of the initial robustness examination through substituting the dependent variables. The analytical framework and estimation methodology maintain consistency with specifications reported in Table 5.

Columns (1) and (3) present the test results with environmental regulations as the moderating variable. The coefficient of the linear R&D intensity term remains positive and statistically significant, while the quadratic term coefficient continues to demonstrate negative significance. Similarly, the interaction term for the linear component maintains its positive significance, whereas the quadratic interaction term exhibits negative significance, consistent with the benchmark regression findings. The findings demonstrate that environmental regulation retains its moderating influence. This provides robust support for hypothesis H2.

Column (2) demonstrates that under environmental subsidy moderation, the linear R&D intensity coefficient exhibits positive significance. The quadratic coefficient remains negative and significant. The interaction term coefficients align with the benchmark regression findings. This confirms that environmental subsidies maintain their moderating influence on substantive green innovation performance. These results provide additional validation for Hypothesis H3.

Second,

Table 7. Adjusts the sample cycle.

Variables	(1) GI	(2) GI	(3) GU	(4) GU
RD	0.279 *** (5.29)	0.310 ***	0.145 ***	0.168 ***
		(5.98)	(2.93)	(3.53)
RD2	−0.010 *** (−3.94)	−0.011 ***	−0.006 **	−0.007 ***
		(−4.07)	(−2.56)	(−2.98)
ERI	−8.365 ** (−2.04)		−4.982 (−1.22)
SUB		0.019 ** (2.47)		0.018 ** (2.54)
ERI × RD	8.444 *** (3.74)		3.273
			(1.39)
ERI × RD2	−0.635 ***		−0.212
	(−3.88)		(−1.53)
SUB × RD		0.008 (1.47)		−0.004
				(−0.81)
SUB × RD2		−0.001 * (−1.75)		0.0003
				(0.97)
Controlled variable	YES	YES	YES	YES
Industry-fixed effects	YES	YES	YES	YES
Year-fixed effects	YES	YES	YES	YES
Prob > ${\mathrm{c}\mathrm{h}\mathrm{i}}^2$	0	0	0	0
Log Likelihood	−5162.91	−5136.81	−5136.81	−5135.49

Note: In the above table, * is used to indicate p < 0.1, ** to indicate p < 0.05, and *** to indicate p < 0.01. GI denotes the number of green invention patent applications; GU denotes the number of green utility model patent applications.

reports the robustness check results based on a restricted sample period. This study refined the sample period by selecting the 2011–2017 timeframe from the original 2011–2023 range. Table 7 maintains the same model specifications and estimation methodology as Table 5.

Analysis of the restricted sample period reveals that the linear interaction term maintains positive significance. Conversely, the quadratic interaction term retains negative significance. These findings reaffirm the validity of the benchmark regression results. The analysis demonstrates that both environmental regulation and subsidies exert significant moderating influences. This provides robust empirical validation for Hypotheses H2 and H3. Robustness analysis employing the 2011–2017 timeframe validates the stability of most observations. Nevertheless, moderating influences on strategic green innovation performance demonstrate certain sensitivity to more conservative clustering adjustments. This suggests that such relationships may necessitate extended observation periods for consistent detection.

To resolve the endogeneity concern, this investigation employs instrumental variable methodology. The research selects China’s R&D tax incentive policies as instrumental variables. First, regarding the correlation requirement: R&D tax incentive policies directly motivate organizations to enhance R&D investment by reducing actual R&D investment costs. This satisfies the strong correlation requirement between instrumental variables and endogenous explanatory factors. Second, concerning exogenous requirements: Tax incentive policies are consistently developed by the central government according to national innovation strategies. Organizations possess no influence over policy content and implementation timing. Additionally, these policies are input-focused and do not directly target particular innovation outputs. They can only indirectly influence green innovation performance through promoting R&D investment. This satisfies the exogenous and exclusion restrictions of instrumental variables. Therefore, R&D tax incentive policies constitute effective instrumental variables that fulfill these requirements. Based on the tax incentive policies, the instrumental variables (PolicyIntensity) are constructed as follows. Formula (4) contains “I” as the dummy variable for the year. This variable equals 1 when corresponding conditions are satisfied.

PolicyIntensity_t = 0.5 × I (year ≥ 2008) + 0.25 × I (year ≥ 2018) + 0.25 × I (year ≥ 2021)

(4)

Following instrumental variable construction, a first-order lag is applied to eliminate immediate correlation. The green patent count represents count data. Therefore, this investigation adopts the control function approach to address the endogeneity concern [50,51]. Environmental protection subsidies demonstrate no significant influence on strategic green innovation performance. Consequently,

Table 8. Instrumental variable estimation.

	The First Phase of Regression	Second-Stage Regression Results
	R&D intensity (1)	GI (2)	GU (3)
L.Policy Intensity	3.68 *** (9.06)
RD		0.35 *** (2.59)	0.22 (1.36)
RD2		−0.009 *** (−5.13)	−0.006 *** (−3.57)
ERI × RD		7.70 *** (3.95)	5.04 ** (2.51)
ERI × RD2		−0.56 *** (−4.64)	−0.33 *** (−2.66)
controlled variable	YES	YES	YES
fixed effect	YES	YES	YES
R2	0.36	0.07	0.07
F-statistic	82.04

Note: In the above table, ** to indicate p < 0.05, and *** to indicate p < 0.01. GI denotes the number of green invention patent applications; GU denotes the number of green utility model patent applications.

employs environmental regulations as moderating variables. This table displays the control function method regression outcomes. First-stage R&D tax incentive policies exhibit significant positive effects on enterprise R&D investment intensity. Second-stage regression outcomes for each component align with expectations. These results remain entirely consistent with previously stated assumptions.

The first-stage F-statistic equals 82.04. This value substantially exceeds the critical threshold of 10. These results indicate that the instrumental variable successfully passes both the weak instrument test and the non-identifiable tests [52]. The quantity of instrumental variables matches the number of endogenous variables. Consequently, over-identification testing becomes unnecessary.

4.6. Further Analysis

Manufacturing organizations assess substantive green innovation through the number of green patents they possess. These patents significantly enhance energy efficiency and emission reduction capabilities. The investigation utilizes five categorical classifications: state-owned enterprises (SOEs), non-state-owned enterprises (NSEs), and regionally distributed organizations in eastern, central, and western areas. SOEs and NSEs constitute one comparative framework. The eastern, central, and western areas form separate analytical categories. This research investigates how R&D investment intensity particularly influences green innovation performance among different enterprise classifications.

Table 9. Analysis of heterogeneity.

GI	Take Environmental Regulation as the Regulating Variable Panel (1)					The Environmental Subsidy is the Moderating Variable Panel (2)
Variables	State- Owned	Non-State-Owned	Eastern	Mid- Sized	Western	State- Owned	Non-State-Owned	Eastern	Mid- Sized	Western
RD	0.213 *** (3.32)	0.319 *** (7.82)	0.354 *** (8.56)	0.253 *** (3.74)	0.141 ** (2.00)	0.209 *** (3.31)	0.302 *** (7.81)	0.350 *** (8.89)	0.201 *** (2.98)	0.173 ** (2.20)
RD2	−0.009 *** (−2.71)	−0.011 *** (−5.68)	−0.013 *** (−6.28)	−0.010 *** (−3.41)	−0.006 ** (−2.02)	−0.007 ** (−2.25)	−0.010 *** (−5.66)	−0.012 *** (−6.75)	−0.006 *** (−2.05)	−0.010 ** (−2.01)
ERI	−7.012 (−1.36)	−5.922 (−1.04)	−0.037 (−0.01)	−8.274 (−1.05)	−30.703 *** (−3.99)
SUB						−0.001 (−0.10)	0.021 *** (2.97)	0.021 *** (3.30)	−0.007 (−0.65)	−0.031 * (−1.73)
ERI × RD	9.250 *** (3.66)	7.939 *** (2.78)	9.078 *** (3.53)	6.882 ** (2.12)	0.091 (0.03)
ERI × RD2	−0.764 *** (−3.99)	−0.508 *** (−3.32)	−0.613 *** (−4.23)	−0.522 *** (−2.74)	−0.394 * (−1.80)
SUB × RD						0.001 (0.14)	0.001 (0.27)	0.010 * (1.90)	−0.001 (−0.21)	0.004 (0.45)
SUB × RD2						0.0008 (0.26)	−0.0004 (−1.51)	−0.001 *** (−3.13)	0.0001 (0.56)	−0.002 (−1.54)
Control variable	YES					YES
fixed effect	YES					YES
N	3794	10,663	9593	2933	1885	3794	10,663	9593	2933	1885

Note: In the above table,* is used to indicate p < 0.1, ** to indicate p < 0.05, and *** to indicate p < 0.01. GI denotes the number of green invention patent applications; GU denotes the number of green utility model patent applications.

displays the findings of this analysis. Panel (1) demonstrates that environmental regulations exhibit significant moderating effects. Across both SOEs and NSEs, the quadratic R&D intensity term exhibits negative significance, as does the corresponding interaction term. This suggests that the inverted U-shaped relationship remains valid under these conditions, with environmental regulations serving as important moderators.

Figure 4, Figure 5, Figure 6 and Figure 7 display the observations reported in Panel (1). Figure 4 demonstrates that heightened environmental regulatory pressure essentially transforms the R&D intensity-green innovation association within state-owned enterprises. This reveals a remarkable distinction among regulatory contexts. Under low ERI circumstances, SOEs display a distinctive U-shaped pattern with an exceptionally minimal turning point at 0.2%. This suggests swift activation of innovation capabilities through limited R&D investment. High ERI circumstances produce a compressed inverted U-shaped association with a turning point at 8.8%. This indicates that regulatory pressure restricts innovation potential while compelling premature optimization. These outcomes show that environmental regulations establish varying strategic requirements for state-owned enterprises. Stringent regulatory circumstances actually constrain the innovation ceiling rather than improving performance through intensification. Moderate regulatory environments enable SOEs to utilize their institutional advantages for continuous innovation advancement.

Figure 5 demonstrates that environmental regulatory pressure establishes markedly different innovation pathways for non-state-owned enterprises. This illustrates the essential distinction between ownership structures when responding to institutional limitations. Under low ERI circumstances, NSEs display a notable U-shaped association with a turning point at 28.1%. This indicates that private enterprises need considerable resource accumulation to surpass critical mass thresholds. They achieve innovation breakthroughs at 1.09 performance level through this process. High ERI circumstances produce a compressed inverted U-shaped curve with an early turning point at 12.0%. The performance ceiling reduces significantly to 0.39 under these conditions. These observations clarify the resource-intensive characteristics of innovation within market-oriented enterprises. They also reveal the restrictive impacts of regulatory pressure. Turning points vary substantially between elevated and minimal regulatory circumstances. Stringent environmental regulations weaken NSEs innovation potential. They compel premature resource redistribution from exploratory R&D toward compliance activities. Moderate regulatory environments allow NSEs to exploit their market adaptability for continuous innovation performance.

Subsequently, the analysis examines the differential impact of environmental regulations across eastern, central, and western regions. Table 9 reveals that the hypothesized relationships hold for firms in both eastern and central regions. However, western firms exhibit a distinct pattern: the linear interaction term lacks statistical significance, whereas only the quadratic interaction term demonstrates negative significance. This result suggests regional heterogeneity in the moderating influence of environmental regulations, with western firms responding differently to regulatory pressures. Given that central region firms exhibit significant outlier effects that may compromise result reliability, separate graphical analysis is excluded for this sub-sample. Consequently, the study presents moderating effect visualizations exclusively for eastern and western regions. Central region findings are interpreted primarily through regression coefficient analysis.

Figure 6 demonstrates the distinctive configuration of western region organizations. These enterprises display unbalanced responses to environmental regulation. Under low ERI circumstances, western enterprises attain performance levels of 15.5 at 30% R&D intensity. The western region trajectory shows unique features without displaying a notable turning point within standard parameters (0–40%). This configuration differs substantially from alternative regions. High ERI circumstances produce limited influence with flat performance trajectories and early turning point at 6.0%. The substantial regional variation in regulatory effectiveness originates from core institutional and structural differences. Initially, institutional quality fluctuates considerably among regions, establishing varied regulatory transmission pathways. Eastern regions maintain strong legal frameworks and enforcement capacities that facilitate balanced policy execution. Western regions encounter institutional weakness that restricts effective regulatory transmission [53]. Subsequently, regional innovation ecosystems vary in development stages, substantially influencing regulatory impacts. Eastern regions gain from established innovation infrastructure supporting high-intensity R&D investment. Western regions experience deficient ecosystem foundations required for innovation breakthroughs, necessitating considerably elevated resource thresholds [54]. Finally, economic development phases differ among regions, creating essential misalignment between regulatory goals and regional capacities. Western regions depend on resource-intensive development approaches that fundamentally contradict green transformation requirements. Such contradictions demand differentiated policy strategies since uniform regulatory intensity remains insufficient for varied regional circumstances [55].

Figure 7 illustrates regional heterogeneity impacts. Geographic variations essentially transform the environmental regulation-innovation connection. This exposes differences in institutional effectiveness throughout China’s economic landscape. Eastern region organizations display distinctive configurations. Under low ERI circumstances, they exhibit a balanced U-shaped association with a turning point at 28.7% and maximum performance of 0.82. High ERI circumstances substantially compress this association. The turning point moves early to 11.3% while the performance ceiling decreases to 0.28.

Panel (2) demonstrates that environmental protection subsidies exhibit no significant moderating effects on either state-owned or non-state-owned enterprises. This finding contrasts sharply with the pronounced differences observed for environmental regulations in Panel (1). However, this cross-ownership consistency pattern potentially indicates that environmental protection subsidy policies maintain more standardized implementation and distribution mechanisms. This observation aligns with resource enhancement theory. Subsidies function as external capital injections. Their effectiveness depends primarily on enterprise project quality rather than ownership characteristics.

Figure 8 illustrates the findings presented in Panel (2). The Panel demonstrates significant effects for eastern region enterprises. Both linear and quadratic interaction terms achieve statistical significance. These results indicate that environmental subsidies maintain moderating effects. Eastern regions display consistent inverted U-shaped associations under government subsidy moderation. Figure 8 illustrates that elevated environmental subsidy levels produce a rightward movement in the turning point. This increases the R&D intensity threshold for optimal performance. Simultaneously, it raises the curve elevation. Such elevation corresponds to considerably enhanced green innovation performance outputs at optimal performance points. The following mechanisms explain this phenomenon.

Resource enhancement theory indicates that government subsidies operate as external resource infusions that broaden enterprises’ innovation capacity limits. This enables organizations to maintain higher-intensity R&D investments while preserving financial sustainability. The rightward movement from 11.4% to 18.4% in eastern firms illustrates how subsidies permit organizations to expand their innovation investment time frames beyond inherent resource limitations [56,57]. Absorptive capacity theory explains that regional differences in innovation policy effectiveness emerge from varied absorptive capacities among regions. Eastern regions maintain superior absorptive capacity with enhanced capabilities to internalize innovation motivations. Environmental protection subsidies produce intensified innovation motivations that eastern regions convert efficiently. They translate financial support into concrete innovation results through advanced institutional mechanisms and market infrastructure. Institutional environment theory offers further explanations for these regional differences [17,58]. Disparities in resource allocation levels, institutional advancement levels, and economic development levels create considerable variations in how government subsidies affect regional innovation effectiveness. Eastern regions display comprehensive institutional environments alongside elevated marketization degrees. These elements provide superior foundations for subsidy absorption while enabling successful conversion into innovation performance. Eastern regions maintain institutional advancement that optimizes the subsidy-innovation transformation mechanism, enhancing comprehensive policy effectiveness [58,59].

5. Conclusions and Recommendations

This study employs a comprehensive dataset of Chinese manufacturing listed firms covering the period 2011–2023 to examine the impact of R&D intensity on both substantive and strategic green innovation performance. The key results are outlined as follows.

The research provides significant theoretical advancement by revealing the curvilinear association between R&D intensity and green innovation performance. Initially, the investigation establishes that an absorptive capacity threshold exists. This finding challenges conventional assumptions that “more investment yields better outcomes.” The discovery emphasizes that optimizing resource allocation surpasses merely accumulating resources in importance. Subsequently, the contrasting regulatory impacts of environmental regulations and environmental subsidies demonstrate distinct mechanisms. Regulations compress innovation space via constraint mechanisms. Conversely, subsidies broaden organizational capabilities through resource enhancement mechanisms. This offers fresh perspectives for institutional tool classification. Finally, the differential responses of substantive green innovations and strategic green innovations confirm the motivation-based innovation classification framework.

As ecological development theory continues advancing, the research provides theoretical implications for future investigations. Initially, scholars could explore the temporal dynamic evolution of optimal R&D intensity. Such exploration would examine how optimal thresholds change over time. Subsequently, absorptive capacity thresholds may vary across different industries. This variation presents significant research opportunities. Additionally, the applicability of these findings in cross-border contexts requires examination. International validation would strengthen theoretical generalization. Finally, the multi-level green innovation ecosystem theory requires systematic construction. This represents a promising avenue for theoretical development.

Regarding methodological contributions, the investigation introduces several innovations. Negative binomial regression is employed to address overdispersion concerns in patent count data. The turning-point test validates curvilinear relationships. Measurement reliability is enhanced through integrating multiple data sources. The research also exhibits specific limitations. Initially, concerning generalizability issues, the analysis is confined to China’s manufacturing sector. Cross-national and cross-industry validation remains necessary. Subsequently, regarding variable measurement, categorizing different types of green innovation through distinguishing between invention patents and utility model patents cannot entirely eliminate conceptual overlap. Concerning model specification, future research could examine asymmetric models or machine-learning approaches. These methods may reveal more complex nonlinear associations.

According to the results, the following suggestions are put forward:

Business managers should implement targeted R&D investment strategies. The optimal R&D intensity should be established at 16–18% of revenue. State-owned enterprises can accomplish innovation breakthroughs at relatively modest intensity levels (8–12%). Private enterprises require elevated investment thresholds (20–25%). Eastern region organizations can leverage innovation ecosystems to attain high efficiency through moderate investment. Western region enterprises need to substantially enhance their investment to offset infrastructure limitations.

Governments should establish differentiated institutional frameworks. Environmental regulations require careful calibration to avoid excessive measures that inhibit innovation. A progressive compliance mechanism should be implemented instead. Environmental subsidy programs should expand their scope, particularly targeting private enterprises and western regions. Policy coordination should follow a sequential framework of “subsidies first, followed by regulations”.

Other emerging economies should adopt a three-stage policy framework based on China’s experience. The initial phase (1–3 years) should primarily emphasize subsidies to develop innovation capabilities. The intermediate phase (4–7 years) should introduce measured environmental regulations. The final phase (eight years onward) should establish balanced regulatory-subsidy frameworks. Innovative infrastructure development should receive priority attention. Regional differentiation strategies require implementation. Experience exchanges should be enhanced through international policy mechanisms such as South–South cooperation.