Can Tax Incentives Drive Green Sustainability in China’s Firms? Evidence on the Mediating Role of Innovation Investment

Abstract

Excessive corporate use of fossil fuels has significantly worsened global air quality. In response, many governments, including China’s, have implemented tax incentives to promote sustainable development, though their effectiveness at the firm level remains unclear. This study empirically examines the relationship between tax incentives and corporate green transition using a panel of 30,483 firm-year observations from Chinese A-share non-financial listed firms spanning 2009–2023. We construct a Green Sustainable Development Performance (GSDP) index based on green patent applications and environmental disclosure and identify innovation investment as the main transmission mechanism. The results show that stronger tax incentives are associated with higher GSDP scores. This relationship is largely driven by innovation: after controlling R&D input, the direct effect of tax incentives declines, while the indirect effect through innovation remains both statistically and economically significant. The effect is more evident in large firms and those in eastern provinces, but weaker in regions with higher financial constraints with limited time lags. The findings offer practical implications for designing targeted, verifiable, and innovation-oriented tax instruments to foster high-quality, sustainable corporate development.

1. Introduction

China’s remarkable economic rise in recent decades has come at a steep environmental cost, evidenced by severe pollution and resource pressures. Halisçelik & Soytas [1] think that the country now ranks near the bottom globally in environmental performance, reflecting the urgency of its ecological challenges. In this context, the development of eco-friendly technologies, processes, and products is widely viewed as an essential catalyst for sustainable development. Qasim et al. [2] shown that by investing in green innovation, companies can reduce their environmental footprints while enhancing efficiency and competitiveness, creating a win-win for the economy and the environment. However, green technology innovation is often expensive and risky, with long gestation periods and positive externalities that firms may not fully capture [3,4]. As a result, Hewitt-Dundas et al. [5] left to market forces alone, businesses tend to under-invest in such innovation, necessitating government intervention to correct this market failure. The Chinese government has thus increasingly stepped in to spur corporate green innovation and sustainability efforts through a mix of regulatory and market-based policy tools [6].

Corporate green transformation has emerged as a critical component of global sustainability transitions, yet firms continue to encounter substantial technical, financial, and organizational constraints that impede meaningful progress. These persistent barriers highlight that market forces alone are insufficient to drive substantive low-carbon transition, thereby elevating the importance of well-designed policy incentives. Among existing policy instruments, tax incentives represent a central and scalable mechanism capable of reshaping firms’ cost structures and investment expectations toward environmentally responsible strategies. Nevertheless, the extent to which tax incentives generate genuine green transformation—as opposed to inducing strategic, symbolic, or even distortionary responses—remains an open empirical question with significant policy implications. Addressing this unresolved issue is essential for understanding the effectiveness of fiscal policy in accelerating firm-level decarbonization pathways, thereby motivating the core research question of this study.

Application and authorization data for three types of patents (invention, utility model, and design) from 2009 to 2023 (Figure 1), patent activity has generally expanded significantly: the total number of applications accepted for the three types increased from approximately 977,000 to 5.4 million, with a compound annual growth rate of approximately 13%, reaching a high point in 2021–2022 and then stabilizing in 2023. Structurally, utility models have consistently accounted for the highest proportion, followed by inventions, with designs accounting for the lowest. Furthermore, since 2016, the gap between utility models and inventions has widened further, indicating that innovation supply is mainly focused on improvements and process optimizations. The total number of authorizations has consistently been significantly lower than the total number of applications accepted, reflecting a systemic lag in the examination-authorization process. While the number of authorizations has increased in recent years, a gap still exists compared to the number of applications accepted, suggesting an increase in quality thresholds and examination intensity. Two significant leaps in scale (approximately 2016–2017 and 2019–2021) coincide with industry booms and strengthened governance.

It enriches understanding of the policy–innovation–sustainability nexus in an emerging market context, offering empirical evidence on how tax incentives impact corporate green behavior in China [7,8]. This adds to the limited scholarship on market-based environmental policy tools in developing economies. By explicitly testing the innovation investment channel, we contribute to resolving the debate on how tax incentives exert their influence [9,10]. If innovation is confirmed as a key conduct, it provides a more nuanced explanation for mixed results in the literature and highlights the importance of fostering corporate R&D activity as part of any green tax policy. Third, our findings carry practical relevance for policymakers. As China rolls out green finance initiatives and tax reforms to support its dual carbon goals, understanding the effectiveness of tax incentives can guide better policy design—ensuring that tax cuts or credits are structured to truly stimulate additional innovation and yield environmental dividends, rather than simply bolstering short-term profits [11].

Patent Output by Industry and Tax Group, the China Securities Regulatory Commission (CSRC) classifies Shanghai and Shenzhen A-share listed companies into 13 categories and 28 major groups (Figure 2). The original industry categories were fragmented (dozens), resulting in an uneven number of companies and poor comparability. This article simplifies the CSRC’s classification into six key industries. As we shown in Figure 2, the X-ax is shows the corporate income tax preferential groups (15% vs. 25%), and the Y-axis shows the six major industries. Between 2015 and 2023, the average number of patent applications filed by companies within the industry increased from approximately 900,000 in 2009 to approximately 5.5 million in 2023.

This study aims to deepen the theoretical and empirical discourse on whether and how tax incentives drive sustainable development at the firm level. By illuminating the mediating role of innovation investment, we offer a fresh perspective on leveraging fiscal policy to align economic and environmental objectives in China and other emerging economies. Our analysis not only fills a notable research gap but also provides evidence-based guidance for refining green tax incentive frameworks to maximize their impact on corporate sustainability.

Significant “ladder-like” gaps between different industries are shown in Figure 3. The overall R&D intensity of several industries on the right side of the figure (knowledge and technology intensive sectors) is significantly higher than that of the industries in the middle and left sides, with typical values in the range of 15–30% and peak values close to 35%; traditional and factor-intensive industries are generally in the range of 3–8% and 5–12%, respectively. The green and blue shading represents the numerical values.

Tax incentives can effectively correct the problem of insufficient R&D investment and promote the commercial application of green technology [12,13]. Green innovation faces long-term investment and high risks, and market mechanisms often cannot fully provide capital support [14]. In developed countries, policies such as R&D tax credit and investment tax credit are widely recognized as effective tools to improve green innovation output [15]. Tax incentives can significantly increase the number of green patents, R&D intensity and environmental capital investment of enterprises. However, these effects are based on the premise that the capital market is mature, the innovation ecosystem is perfect, and the policy implementation is transparent, and the regulatory system is solid [16].

In emerging economies, tax incentives also have potential, but they face constraints such as poor financing channels, information asymmetry and difficult policy implementation. Some studies have pointed out that unreasonable or loosely implemented incentive mechanisms may lead to resource mismatch and policy rent-seeking behavior and even inhibit the quality of innovation [17].

This study advances the literature on fiscal policy and corporate green transformation in several important ways. First, we develop a comprehensive firm-level Green Sustainable Development Performance (GSDP) index using principal component analysis (PCA) on green patent applications and environmental disclosure scores. This composite measure captures both technological and accountability dimensions of green behavior, offering a more substantive and multidimensional indicator than traditional single-proxy measures such as green patents or ESG ratings. Second, by combining detailed micro-level tax data with rigorous identification strategies—including system GMM and extensive robustness checks—this study provides credible causal evidence on how tax incentives influence firms’ substantive green transformation rather than symbolic compliance. Third, we move beyond the conventional innovation-driven narrative by examining the possibility of incentive misallocation, revealing that tax incentives may produce heterogeneous effects across regions, industries, and firm characteristics. This perspective uncovers previously overlooked behavioral responses and adds nuance to the understanding of how fiscal tools shape sustainable corporate behavior. Overall, the study enriches both the empirical and policy-oriented debates on the effectiveness of tax incentives in promoting genuine green transition under China’s dual-carbon strategy.

2. Literature Review and Research Hypothesis

2.1. Literature Review

As a key path to address environmental pollution and achieve high-quality economic growth, green innovation has attracted widespread attention from researchers. The following points have been substantiated. Financial conditions and subsidy mechanisms affect the tax incentive effect: companies with high financing constraints are sensitive to tax incentives, and subsidies often squeeze out the tax incentive effect, especially in state-owned or large-scale enterprises.

Wang & Mayburov [18] think that the various policy instruments, tax incentives have emerged as a critical lever to encourage firms’ innovation and environmental engagement. Market-based incentives like preferential tax treatments are aligned with the Porter Hypothesis, which posits that well-designed environmental policies can induce technological innovation that improves both environmental and economic performance [19]. Unlike command-and-control regulations, tax incentives work by altering cost structures and cash flows, thereby nudging firms toward greener behavior without mandating specific actions [20]. In China, authorities have progressively built a “green tax” system and introduced a range of tax-based measures to support sustainable development objectives. The Corporate Income Tax Law granted tax exemptions or reductions for income derived from qualified environmental protection and energy-saving projects [21]. Empirical studies have indeed observed that tax incentives can stimulate corporate innovation activity [22]. Moreover, some evidence suggests that tax incentives may be more effective than direct subsidies in fostering innovation, as subsidies risk misuse or crowding-out effects, whereas tax benefits reward firms’ own R&D efforts in a more market-driven manner.

Wang et al. [23] report that R&D tax incentives in China significantly increase green product innovation, highlighting the role of fiscal incentives in driving eco-friendly technological development. Zheng et al. [24] similarly document that a mix of environmental taxes and incentive policies positively affects green innovation activity in Chinese firms. In turn, greater innovation investment has been linked with improved sustainability outcomes–as firms innovate, they can reduce pollutant emissions and even strengthen their financial performance through efficiency gains [25,26].

On the other hand, the literature also reveals heterogeneous and sometimes contradictory findings, suggesting that the effectiveness of tax-related policies is context-dependent [27,28,29]. Not all studies concur that tax incentives uniformly spur green innovation. Some researchers contend that the impact of tax incentives on firm innovation is modest or uncertain, and in certain cases can even be counterproductive, leading to suboptimal resource allocation [30]. For example, while many analyses celebrate the innovation-stimulating effect of tax breaks, a few studies find negligible gains or raise the possibility of a crowding-out effect, where tax incentives might displace other innovation financing without adding net new effort. This suggests that if environmental taxes are too stringent or poorly designed, they might squeeze firms (especially small and medium enterprises) financially, thereby hindering their ability to invest in innovation [31].

Such findings echo international studies where excessive environmental taxation dampened innovation among resource-constrained firms. Oduro et al. [32] think that the mixed results in literature underscore that policy effects on green innovation are complex: positive impacts tend to emerge under incentive-based or moderate regimes that alleviate constraints and signal support, whereas negative impacts may arise when fiscal measures impose heavy burdens without providing avenues for firms to respond constructively. This ambiguity points to the need for deeper analysis into the channels through which tax policies influence corporate behavior. Notably, scholars have begun to investigate the mechanisms, whether the benefit of a tax incentive must first translate into increased R&D investment to yield greener innovations and sustainability gains [33]. The presence of such a mediating process could explain why the same tax policy might succeed in some contexts (where firms channel tax savings into innovation) but falter in others (where savings are diverted or insufficient).

Research on emerging economies like China is especially pertinent, given their scale and rapid policy evolution, yet remains relatively limited. In the case of China, although there is increasing interest, several gaps are evident in the literature. First, as noted in recent reviews, few studies have specifically probed the impact of tax incentives on firms’ green investment or sustainability performance [34]. Second, the mediating role of innovation investment in this relationship has not been fully illuminated some find strong positive effects of tax incentives on green innovation [35,36,37].

For China, the cumulative research in recent years shows that tax incentives are significantly correlated with green performance and reveals multi-dimensional mediation and boundary mechanisms [38]. Empirical evidence shows that environmental protection taxes not only reduce pollution emissions but also have a positive impact on improving corporate ESG performance and green patent output, and the mediation effect is partially reflected through the ESG path. The simulation study under development shows that if the tax rate design and regulatory rules for industries such as wind power are combined with corporate self-learning behavior, the diffusion speed of real green R&D can be increased, and “fake R&D” behavior can be avoided [39,40].

Despite the rich accumulation of the above research, there is still a lack of systematic empirical analysis that links TIG (tax incentive amplitude) with R&D over-deduction ratio, identifies the whole chain mechanism of innovation input-green performance, and combines heterogeneous adjustment boundaries such as financing constraints, ownership structure and industrial attributes.

2.2. Research Hypothesis

This study focuses on how tax incentives can promote green and sustainable development by influencing corporate innovation input, considering policy boundaries, corporate heterogeneity and macro green transformation goals.

According to the Porter Hypothesis, well-designed environmental policies can stimulate innovation and competitiveness, so that the gains from innovation may “offset” or even exceed compliance costs. In other words, tax breaks for green investment can spur firms to develop more efficient or cleaner technologies, improving their sustainable performance. From a new institutional economics viewpoint, tax incentives are a form of government intervention that corrects market failures and realigns firm behavior toward green goals. In tandem, a resource-based view suggests that tax incentives effectively augment a firm’s resources available for environmental R&D. Empirical studies confirm that tax credits increase firms’ net cash for innovation. Altogether, these theories imply that tax incentives should have a positive effect on firms’ green and sustainable development performance, as firms apply the extra resources to new green initiatives and efficiency improvements. Based on Chinese companies, it has been shown that tax incentives can significantly increase corporate green investments [41]. Tax incentives also significantly improve the overall effectiveness of corporate ESG [42]. These consistent findings imply that fiscal tax breaks encourage firms to invest more in environmental initiatives and sustainability measures.

Existing empirical studies in China have found that policies such as additional deductions for R&D expenses and preferential corporate income tax rates have significantly increased corporate green patent output, ESG environmental scores and green investment levels. Therefore, this paper proposes:

H1. 

R&D Super-Deduction Ratio (RDSR) have a significant positive impact on corporate green and sustainable development performance (GSDP).

Green innovation input (Super-Deduction Ratio (RDSR)) is the core path connecting fiscal incentives and corporate green transformation. Tax incentives not only increase the marginal return of corporate R&D investment but also may stimulate their motivation to invest in green technology R&D, process transformation and green product development by enhancing corporate risk tolerance.

H2. 

Tax incentives play an intermediary role in the green and sustainable development of enterprises by increasing innovation input.

Companies with high ownership concentration have more stable governance structures and clearer internal incentive mechanisms, making them more likely to effectively use tax incentives for innovation and green transformation.

H3. 

Ownership concentration moderates the relationship between tax incentives and corporate green sustainability performance, indicating that tax incentives have a stronger promoting effect on green development in companies with high ownership concentration.

Firms with high ownership concentration have more stable governance structures and clearer internal incentive mechanisms, making them more likely to effectively leverage tax incentives for innovation and green transformation.

H4. 

Leading enterprises have stronger green sustainability capabilities.

Large firms, with more abundant capital and human capital, can more effectively convert tax benefits into green technology investment. Small firms may lack sufficient absorptive capacity, leading to “incentive inefficiency” or “perverse incentive” effects from policy instruments.

H5. 

The impact of tax incentives on corporate green and sustainable development performance is significantly heterogeneous across firm sizes.

Large firms are more responsive to tax incentives and achieve more significant green transformation results. However, due to resource constraints and limited capabilities, small and medium-sized enterprises experience less incentive effects, which may even produce negative effects.

3. Methodology

3.1. Sample Construction and Variable Definitions

Figure 4 shows the logic diagram for tax incentives. To strengthen the conceptual basis of this study, it is necessary to more clearly articulate the theoretical foundations that connect tax incentives, innovation investment, and firms’ green transformation. Prior research shows that fiscal incentives can ease financial constraints and alter firms’ cost–benefit calculations, thereby making innovation-oriented environmental upgrading more attractive [43]. When tax relief reduces the effective tax burden, firms gain additional room to allocate resources toward R&D activities, including those related to cleaner production, pollution reduction, and environmental technologies [44]. These investments help build technological capabilities that gradually translate into substantive green transformations—such as improved environmental performance, enhanced disclosure, and more sustainable operational practices. Based on this logic, the mechanism can be viewed as a sequential process: tax incentives first relax financial pressure, then promote a shift in resource allocation toward innovation, and ultimately support the transition to greener business models. Making this mechanism explicit, and embedding it into a coherent theoretical framework, will clarify why tax incentives are expected to influence green transformation and will also strengthen the explanatory logic of the empirical analysis.

Our sample covers Chinese A-share non-financial, non-ST listed firms over 2009–2023. Financial and governance data are obtained from Wind and CSMAR; green innovation and ESG disclosure are compiled from Wind’s patent module and corporate annual reports. We exclude financial firms, ST/*ST observations, and firm-years with missing core variables. Continuous variables are first normalized to consistent units and trimmed to the feasible range (e.g., ratio variables constrained to [0, 1]), and then two-sided winsorized at the 1st and 99th percentiles. The final dataset contains 30,438 firm-year observations. Industries and regions follow the CSRC classification and the standard regional division. The definition of the variables is presented in

Table 1. The definition of the variables.

Type	Variable (Abbreviation)	Definition/Construction	Measurement/Explanation	Unit
Dependent	Green Sustainable Development Performance (GSDP)	Composite index for firm-level green sustainable development performance	PCA on (i) green patent applications and (ii) environmental-disclosure E-score. Inputs are z-standardized; PC1 is used. PC1 variance share = 56.46%. Scores are min–max rescaled to [0, 1] (higher = better)	Index [0–1]
Independent	Tax Incentive Gap (TIG)	Statutory CIT 25%—financial-statement–implied effective tax burden	Reported in percentage points (pp). Drop observations with non-positive profit before tax; constrain burdens to [0, 100%] before winsorization. Two-sided winsorization at the 1st/99th percentiles	pp
Independent	R&D Super-Deduction Ratio (RDSR)	Policy intensity of R&D super-deduction	Super-deductible R&D amount/R&D expenditure; capped to [0, 2] then winsorized at the 1st/99th percentiles	Ratio
	R&D Intensity (RD1)	Innovation investment (core mediator)	R&D expenditure/Total assets (ratio)	Ratio
Mediating (core)	GRDI (RD2)	the intensity and green orientation of firms’ R&D activities.	The Green R&D Intensity Index (GRDI) is constructed using two variables: (i) the annual number of green patent applications, and (ii) total R&D expenditure. Both variables are standardized and subjected to principal component analysis (PCA). The first principal component is extracted and rescaled to the [0, 1] range, representing the intensity and green orientation of firms’ R&D activities.”	Index [0–1]
Control	Revenue Growth Rate	Growth in operating revenue	$({\mathrm{Revenu}}_{\mathrm{t}}-{\mathrm{Revenue}}_{\mathrm{t}-1})/{\mathrm{Revenue}}_{\mathrm{t}-1}$; winsorized at the 1st/99th percentiles	Ratio
	Tobin’s Q	Market valuation relative to replacement cost	(Market capitalization + Total liabilities)/Total assets	Ratio
	Firm Size	Firm scale	ln(Total assets)	Log
	Asset Growth Rate	Growth in total assets	$({\mathrm{Totalassets}}_{\mathrm{t}}-{\mathrm{Totalassets}}_{\mathrm{t}-1})/{\mathrm{Totalassets}}_{\mathrm{t}-1};$ winsorized at the 1st/99th percentiles	Ratio
	Board Size	Board structure	Number of directors on the board	Count
	Firm Age	Operating period since establishment	Years since establishment	Years
	Independent Director Ratio	Board independence	Independent directors/Board size; constrained to [0, 1] then winsorized	Ratio [0–1]
	Top-1 Shareholder Ownership	Ownership concentration	Largest shareholder ownership; constrained to [0, 1] then winsorized.	Ratio [0–1]
	Current Ratio	Short-term liquidity	Current assets/Current liabilities	Ratio

Note: All continuous variables are normalized, constrained to the feasible range (e.g., ratios in [0, 1]), and then two-sided winsorized at the 1st and 99th percentiles. Source: compiled by the authors.

.

Green Sustainable Development Performance (GSDP) index using scores from green patent applications and environmental information disclosure. Represents a meaningful attempt to assess the outcomes of green transformation at the firm level, as both patent-based and disclosure-based indicators are widely considered core components of firm environmental performance However, it is crucial to compare the construction of this index with existing multidimensional approaches. Previous studies have typically incorporated broader environmental outputs, carbon emissions, resource efficiency, pollution intensity, and verified ESG indicators, into a composite index, employing principal component analysis (PCA), entropy weighting, or hybrid approaches. Therefore, comparative analysis helps to place the GSDP index within the framework of existing literature, clarifying its methodological advantages and enhancing its structural validity.

Furthermore, integrating such comparisons will improve the index’s transparency and credibility, ensuring its full consistency with current empirical practices in green development measurement.

The construction of the GSDP index also carries a methodological limitation. Because it is based on only two environmental indicators, green patent applications and environmental-disclosure scores, it cannot fully reflect the multidimensional nature of corporate sustainability. Although additional ESG dimensions such as energy efficiency or carbon emissions were not included, the PCA results show that the first component captures a substantial share of variance, and both indicators directly reflect firms’ green innovation outputs and environmental transparency. Within the scope of this study, where the empirical focus is on innovation-driven green transformation, these two variables serve as an appropriate and analytically coherent proxy. Future work may extend the index by incorporating broader ESG metrics to further enhance construct validity.

Although data constraints limit the incorporation of these additional policy variables, acknowledging this limitation is important for interpreting the results. Within the context of this study, TIG and RDSR remain appropriate and policy-relevant proxies, but future research could enrich the analytical framework by integrating a more diverse set of fiscal incentives or by exploring cross-country differences in green fiscal policy design.

As we shown in

Table 2. Eigenvalues and Variance Explained for Principal Components of GSDP.

Component	Eigenvalue	Difference	Proportion	Cumulative
Comp1	1.12929	0.258586	0.5646	0.5646
Comp2	0.870707	0	0.4354	1

,

Table 3. Component Loadings of PCA for GSDP Indicator Construction.

Variable	Comp1	Comp2	Unexplained
Z_green patent	0.7071	0.7071	0
Z_ E-score	0.7071	−0.7071	0

,

Table 4. Scoring Coefficients for Principal Component 1 (PC1).

Variable	Comp1	Comp2
Z_green patent	0.7071
Z_ E-score	0.7071	0.7071

and

Table 5. Descriptive Statistics of the First Principal Component (PC1).

Variable	Obs	Mean	Std. dev	Min	Max
PC1	43,421	−2.31	1.062682	−3.020555	30.26572

Source: Table 2, Table 3, Table 4 and Table 5 compiled by the authors.

, to construct a Corporate Green Sustainable Development Performance (GSDP) index, this paper conducts principal component analysis (PCA) on two standardized indicators: “number of green patents applied for in the current year” and “environmental disclosure E-score.” The results show that the eigenvalue of the first principal component (PC1) is 1.12929, explaining 56.46% of the variance; the second principal component explains the remaining 43.54%. This indicates that PC1 effectively captures the common variance between the two green sustainability indicators, accounting for the majority of the information. The loading matrix further shows that the loadings of both variables (Z_green patent and Z_E-score) on PC1 are 0.7071, indicating that PC1 simultaneously reflects both the company’s green technology innovation capability and environmental governance transparency, and that their contributions are roughly equal. Since the unexplained variance is zero, PC1 can fully extract the systematic information shared by the two indicators.

Subsequently, we perform min–max normalization on the PC1 scores, setting the interval to [0, 1], thus forming the final GSDP index. The PC1 value ranges from −3.0206 to 30.2657, with a mean close to zero and a standard deviation of 1.0627, reflecting significant differences in green and sustainable development performance among sample companies. Therefore, this index has good structural validity and interpretability in statistics and can be used as a comprehensive indicator to measure the level of green and sustainable development of enterprises.

TIG. Our primary independent variable is the Tax Incentive Gap (TIG), defined in percentage points (pp) as the statutory corporate income tax rate (25%) minus the financial-statement–implied effective tax burden. Observations with non-positive profit before tax are removed; burdens are constrained to [0, 100%] before two-sided winsorization at the 1st and 99th percentiles. A higher TIG indicates stronger fiscal incentives.

RDSR. To capture targeted innovation incentives, we use the R&D Super-Deduction Ratio (RDSR), measured as super-deductible R&D amount divided by R&D expenditure (ratio). Values are capped to [0, 2] and then winsorized at the 1st and 99th percentiles.

Mediating. We focus on innovation as the transmission mechanism. R&D Intensity (RD1): R&D expenditure/Total assets (ratio).

Table 6. Eigenvalues and Variance Explained for the GRDI Principal Components.

Component	Difference	Proportion	Eigenvalue	Cumulative
Comp1	1.49638	0.992763	0.7482	0.7482
Comp2	0.503618	0	0.2518	1.0000

,

Table 7. Component Loadings for GRDI Principal Components.

Variable	Comp1	Comp2	Unexplained
z_gp	0.7071	0.7071	0
z_rdtotal	0.7071	−0.7071	0

and

Table 8. Component Loading Matrix for GRDI Construction.

Variable	Comp1	Comp2
z_gp	0.7071
z_rdtotal	0.7071	−0.7071

Source: Table 6, Table 7 and Table 8 compiled by the authors.

illustrates the process of constructing indicators using principal component analysis. GRDI (RD2):

Controls include firm size (ln total assets), Tobin’s Q $\left(\right(\mathrm{m}\mathrm{a}\mathrm{r}\mathrm{k}\mathrm{e}\mathrm{t}\mathrm{c}\mathrm{a}\mathrm{p}\mathrm{i}\mathrm{t}\mathrm{a}\mathrm{l}\mathrm{i}\mathrm{z}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}+\mathrm{t}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}\mathrm{l}\mathrm{i}\mathrm{a}\mathrm{b}\mathrm{i}\mathrm{l}\mathrm{i}\mathrm{t}\mathrm{i}\mathrm{e}\mathrm{s})/\mathrm{t}\mathrm{o}\mathrm{t}\mathrm{a}\mathrm{l}\mathrm{a}\mathrm{s}\mathrm{s}\mathrm{e}\mathrm{t}\mathrm{s}$), revenue growth and asset growth $\left(\right({\mathrm{x}}_{\mathrm{t}}-{\mathrm{x}}_{\mathrm{t}-1})/{\mathrm{x}}_{\mathrm{t}-1})$, board size, independent director ratio, firm age, top-1 shareholder ownership, and the current ratio. All percentage variables are treated as ratios, constrained to feasible ranges, and then winsorized at the 1st and 99th percentiles. Financial and governance data come from Wind and CSMAR; patent and disclosure data come from Wind and corporate annual reports. Descriptive statistics after winsorization are reported in

Table 9. Descriptive statistics.

Variable	Count	Mean	std	p1	p25	Median	p75	p99	min	max
GSDP	30,438	0.31	0.1785	0.0011	0.1825	0.2816	0.4145	0.8303	0.0011	0.8305
TIG	30,438	0.0107	0.1196	−0.4776	−0.0256	0.0107	0.0713	0.2491	−0.4776	0.2491
RDSR	30,438	0.0267	0.0404	0	0	0.0077	0.04	0.2241	0	0.2241
Revenue Growth Rate	30,438	0.1647	0.4297	−0.5804	−0.0339	0.0975	0.2573	2.7988	−0.5804	2.8003
Tobin’s Q	30,438	2.0384	2.4986	0	0.6456	1.3261	2.4478	16.5863	0	16.6151
Firm Size (ln assets)	30,438	22.2335	1.4395	19.0008	21.2666	22.1084	23.0942	26.3667	19.0004	26.3668
Asset Growth Rate	30,438	0.1875	0.4303	−0.3356	0.0021	0.0845	0.2149	2.8984	−0.3356	2.8989
Board Size	30,438	8.2438	2.5046	0	7	9	9	15	0	15
Operation Period (years)	30,438	18.81	6.6868	9	13	17	25	31	9	31
Independent Director Ratio	30,438	0.3559	0.0969	0	0.3333	0.3333	0.4286	0.5714	0	0.5714
Top-1 Shareholder’s Ownership	30,438	0.3255	0.1636	0	0.2087	0.3059	0.4359	0.7482	0	0.7482
Current Ratio	30,438	2.2997	2.4632	0.2379	1.0763	1.5468	2.4618	16.2882	0.2379	16.2885

Source: calculated by the authors.

.

Overall, the variables show substantial heterogeneity across firms and years, underlining the importance of controlling firm-specific factors (via fixed effects) in the regression analysis.

3.2. Baseline Fixed-Effects Model

First, we build a baseline regression model with green sustainable development performance (GSDP) as the explained variable, TIG and RDSR as the key independent variables, controlling for a series of financial and governance variables, and adding firm and year fixed effects to control unobserved heterogeneity. The regression equation of the baseline model is Formula (1):

$$\mathrm{GSD}{\mathrm{P}}_{\mathrm{it}}={\mathsf{\beta }}_1{\mathrm{TIG}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\beta }}_2{\mathrm{RDSR}}_{\mathrm{i}.\mathrm{t}}+{\sum }_{\mathrm{k}}{\mathsf{\beta }}_{\mathrm{k}}{\mathrm{Control}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathbf{i},\mathbf{t}}$$

(1)

where ${\mathsf{\mu }}_{\mathrm{i}}$ represents the firm fixed effect and ${\mathsf{\lambda }}_{\mathrm{t}}$ represents the year fixed effect.

Table 10. Baseline Fixed Effects OLS Regression Results.

Variables	FE OLS
TIG (pp-equivalent ratio)	0.0043
RDSR (ratio)	0.1302 ***
Revenue Growth (ratio)	−0.0077 ***
Tobin’s Q	0.0018 ***
Firm Size (ln assets)	0.0510 ***
Asset Growth Rate	0.0017
Board Size	−0.0001
Operation Period (years)	0.0026 ***
Independent Director Ratio	0.0370
Top-1 Shareholder’s Ownership	−0.0045
Current Ratio	0.0012 **
Firm FE & Year FE	Yes
Observations	30483
# Clusters (firm)	2241
Within R2	0.0545

Note: FE = Fixed Effects. Standard errors are clustered by firm; p ** < 0.05; p *** < 0.01.

reports the results of the baseline fixed-effect regression (robust standard errors are shown in parentheses, clustered for each firm). It can be seen that RDSR has a significant positive impact on green performance, with a coefficient of 0.0011, significant at the 1% level. This indicates that tax incentives for R&D expenses can significantly enhance a firm’s green and sustainable development performance, supporting the view that tax incentives improve green performance by promoting R&D investment. In contrast, the regression coefficient for TIG is −0.0000 and statistically insignificant. This suggests that the reduction in the general corporate income tax burden has no direct and significant impact on green and sustainable development performance. This may be because, compared with tax incentives specifically for R&D, a reduction in the general tax burden does not guarantee that firms will invest the resulting savings in green innovation or sustainable development projects.

Baseline Fixed Effects OLS Regression Results shown in Table 10 and

Table 11. Variance Inflation Factors (VIF) for Multicollinearity Diagnostics.

Variable	VIF	1/VIF
Revenue Growth (ratio)	4.58	0.218504
Tobin’s Q	4.57	0.218721
Firm Size (ln assets)	1.48	0.674085
Asset Growth Rate	1.47	0.680102
Board Size	1.12	0.890239
Operation Period (years)	1.03	0.96866
Independent Director Ratio	1.01	0.987606
Top-1 Shareholder’s Ownership	1	0.998803
Current Ratio	1	0.999401
TIG	1	0.999896
Mean VIF	1.83

Source: Table 10 and Table 11 calculated by the authors.

The baseline regression results above indicate that R&D tax incentives significantly boost corporate green performance, while general tax relief does not. This finding is consistent with our hypothesis (H1): targeted R&D deductions can encourage companies to increase R&D investment, thereby improving green technology and performance; conversely, simply reducing corporate tax burdens is insufficient to ensure that surpluses are used to enhance sustainability performance.

The variance inflation factor (VIF) results indicate that multicollinearity is not a concern in the baseline regression model. As shown in Table 11, all VIF values are well below the commonly accepted thresholds of 5 or 10. The two highest VIFs—Revenue Growth (4.58) and Tobin’s Q (4.57)—remain within an acceptable range, suggesting moderate but not harmful correlation with other regressors. All remaining variables exhibit VIF values close to 1, with tolerance levels approaching 1, indicating very low inter-variable dependence. The mean VIF is 1.83, further supporting the absence of problematic multicollinearity. Overall, these diagnostics confirm that the estimated coefficients are unlikely to be biased or unstable due to multicollinearity.

3.3. Mediation Test: R&D Investment as a Channel

Next, we examine the mediating role of R&D investment in the impact of tax incentives on green performance in

Table 12. Mediation Test—Tax Incentives, R&D Investment, and GSDP.

Variables	(1) Baseline	SE(1)	(2) +RD1	SE(2)	(3) +RD2	SE(3)	(4) +RD1 & RD2	SE(4)
TIG	0.0043	0.0064	0.0048	−0.0064	0.0050	−0.0063	0.0052	−0.0063
RDSR	0.1302 ***	−0.0442	−0.1904 ***	−0.0610	−0.0454	−0.0482	−0.2242 ***	−0.0609
R&D Intensity (RD1)			0.9792 ***	−0.139916			0.621009 ***	−0.1491
GRDI(RD2)					0.0042 ***	−0.0001	0.0010 ***	−0.0002
Revenue Growth	−0.0077 ***	−0.0015	−0.0088 ***	−0.0015	−0.0082 ***	−0.0015	−0.0089 ***	−0.0015
Tobin’s Q	0.0018 ***	−0.0006	0.0015 **	−0.0006	0.0017 ***	−0.0006	0.0015 **	−0.0006
Firm Size (ln assets)	0.0510 ***	−0.0026	0.0507 ***	−0.0026	0.0481 ***	−0.0026	0.0484 ***	−0.0026
Asset Growth	0.001656	−0.0017	0.0025	−0.0017	0.0020	−0.0017	0.0025	−0.0017
Board Size	−0.0001	−0.0011	−0.0001	−0.0011	−0.0001	−0.0011	0.0001	−0.0010
Firm Age (years)	−0.0026 ***	−0.0008	−0.0024 ***	−0.0008	−0.0018 **	−0.0008	−0.0018 **	−0.0008
Independent Director Ratio	0.0369	−0.0273	0.0379	−0.0274	0.0385	−0.0275	0.0389	−0.0275
Top-1 Shareholder’s Ownership	−0.0045	−0.0164	−0.0041	−0.0162	−0.0040	−0.0163	−0.0038	−0.0162
Current Ratio	0.0012 **	−0.0005	0.0014 ***	−0.0005	0.0012 **	−0.0005	0.0013 ***	−0.0005
Firm FE & Year FE	Yes		Yes		Yes		Yes
SE clustered by	Firm		Firm		Firm		Firm
Observations	30,483		30,483		30,483		30,483
Firms	2241		2241		2241		2241
Within R2	0.0545		0.0586		0.0624		0.0639

Note: p ** < 0.05; p *** < 0.01. Source: calculated by the authors.

. We select two indicators, R&D intensity (RD1, the ratio of R&D expenditure to total assets) and the logarithm of R&D investment (RD2, ln(R&D expenditure + 1)), to represent a firm’s R&D investment, and add them to the baseline model to test the mediation effect. If tax incentives primarily improve green performance by promoting R&D investment, then after controlling R&D investment, the direct effect of the tax incentive variable on green performance should be weakened or disappear. Table 12 reports the results of the mediation effect regression: Column (1) repeats the baseline model for comparison, column (2) adds RD1, column (3) adds RD2, and column (4) adds both mediating variables.

As shown in Table 12, adding the mediator (RD1 or RD2) slightly improves model fit, and the mediator loads Mediation Test—Tax Incentives, R&D Investment, and GSDP positively and significantly on GSDP. Crucially, the coefficient on RDSR attenuates once the mediator is included, while TIG remains insignificant, which is consistent with a partial mediation channel whereby tax incentives raise R&D investment and, in turn, enhance green sustainable development performance. In short, we find evidence consistent with partial mediation. Policy-wise, targeted R&D super-deductions should be prioritized to stimulate firms’ innovation activities and thereby improve green outcomes; by contrast, broad reductions in the overall tax burden alone are unlikely to deliver meaningful improvements in environmental and sustainability performance. This section proves that hypotheses H1, H2, and H3 are correct.

3.4. Endogeneity Test

Considering the potential for reverse causality and endogeneity between tax incentives and corporate green performance (for example, companies with superior green performance may receive more tax benefits, or both may be influenced by unobserved factors) in

Table 13. Endogeneity Test.

Variables	(1) FE OLS	SE(1)	(2) FE-2SLS	SE(2)
TIG	−0.0008	−0.0071	−0.0454	−0.0766
RDSR (ratio)	0.1084 **	−0.0455	0.1204	−0.0749
Revenue Growth (ratio)	−0.0055 ***	−0.0017	−0.0060 ***	−0.0019
Tobin’s Q	0.0017 ***	−0.0007	0.0017 ***	−0.0007
Firm Size (ln assets)	0.0542 ***	−0.0030	0.0534 ***	−0.0034
Asset Growth (ratio)	−0.0031	−0.0020	−0.0034 *	−0.0020
Board Size	0.0002	−0.0012	0.0002	−0.0013
Firm Age (years)	−0.0028	−0.0025	−0.0028	−0.0025
Independent Director Ratio	0.0432	−0.0313	0.0426	−0.0315
Top-1 Shareholder’s Ownership	−0.0013	−0.0181	−0.0043	−0.0187
Current Ratio	0.0010 *	−0.0005	0.0010 *	−0.0006
Firm FE & Year FE	Yes		Yes
SE clustered by	Firm		Firm
Observations	25,979		25,979
Firms	2240		2240
Within R2	0.0519		0.0516
First-stage F (excluded)—TIG	36.04
Partial R2—TIG	0.0132
First-stage F (excluded)—RDSR	697.86
Partial R2—RDSR	0.409

Note: 2SLS uses lagged TIG and RDSR as instruments. Cragg-Donald F-stat is the weak IV test (critical value ~10); Hansen J is overidentification test (not applicable for exactly identified model); p * < 0.1; p ** < 0.05; p *** < 0.01. Source: calculated by the authors.

, we employ 2SLS to test for endogeneity. Specifically, we use the first-order lagged value of tax incentives as instrumental variables: that is, the TIG and RDSR of the previous year are used as instrumental variables for the current values. This selection is based on the assumption that the tax gap and R&D deduction ratio of the previous year are not directly related to the green performance of the current period (satisfying the exogeneity requirement of the instrumental variables) but are highly correlated with the current period’s tax incentive level due to the continuity of policy and operating inertia (satisfying the correlation requirement).

Table 13 shows that using lagged values of TIG and RDSR as instruments, the fixed-effects 2SLS delivers a model fit that is very similar to the baseline FE-OLS. First-stage diagnostics indicate strong instruments: the excluded-instrument F-statistics are 36.04 for TIG and 697.86 for RDSR. In the second stage, the point estimate on RDSR remains positive and close to OLS, although it becomes less precisely estimated, while TIG stays insignificant. Taken together with the baseline and the mediation evidence, these results are consistent with the view that R&D-oriented tax incentives improve firms’ green sustainable development performance primarily by stimulating corporate R&D investment. Policy wise, prioritizing targeted R&D super-deductions is more promising than broad tax relief to foster a virtuous cycle of green innovation and performance.

As a further robustness check for endogeneity and dynamic persistence in green performance, we re-estimate the baseline specification using a dynamic System GMM estimator. Specifically, we augment the model with the first lag of GSDP and treat L.GSDP, TIG and RDSR as endogenous or predetermined variables, instrumented by their own deeper lags in a GMM-style framework. Control variables are treated as exogenous and enter as standard instruments in the level equation.

As we shown in

Table 14. System GMM Estimation Results for Corporate Green Sustainable Development Performance (GSDP).

	(1)
Variables	GSDP
L.GSDP	0.831 ***
	(0.0208)
TIG (pp-equivalent ratio)	−4.81
	(1.96 × 10−6)
RDSR (ratio)	0.000476 **
	(0.000191)
Revenue Growth (ratio)	−3.07 ***
	(5.68 × 10−10)
Tobin’s Q	9.28 **
	(4.40)
Firm Size (ln assets)	0.0138 ***
	(0.00130)
Asset Growth (ratio)	1.45 **
	(6.19)
Board Size	0.000188
	(0.000376)
Firm Age (years)	−0.000419 ***
	(0.000109)
Independent Director Ratio	4.09
	(0.000101)
Top-1Shareholder’s Ownership	−1.04
	(3.78)
Current Ratio	−0.000618 ***
	(0.000160)
2010bn.year	−0.0179 ***
	(0.00447)
2011.year	−0.00677
	(0.00460)
2012.year	−0.000171
	(0.00452)
2013.year	−0.00186
	(0.00422)
2014.year	−0.0111 ***
	(0.00421)
2015.year	−0.0189 ***
	(0.00427)
2016.year	−0.00529
	(0.00439)
2017.year	0.000640
	(0.00444)
2018.year	0.00642
	(0.00454)
2019.year	0.00572
	(0.00417)
2020.year	0.00381
	(0.00445)
2022.year	0.00112
	(0.00381)
2023.year	−0.0117 ***
	(0.00375)
Constant	−0.238 ***
	(0.0218)
Observations	29,415
Number of firm_id	2241

Source: calculated by the authors. p ** < 0.05; p *** < 0.01.

.the system GMM estimation reveals several important dynamics underlying firms’ green sustainable development performance (GSDP). First, the coefficient on the lagged dependent variable (L.GSDP = 0.831, p < 0.01) is strongly positive and statistically significant. This indicates a high degree of persistence in GSDP, suggesting that firms with stronger past green performance tend to maintain or further strengthen their sustainability efforts in subsequent years.

Such path-dependence is consistent with the view that green development requires cumulative learning, long-term strategic planning, and continuous capability accumulation. Regarding the key explanatory variables, the R&D super-deduction ratio (RDSR) exhibits a positive and statistically significant effect (β = 0.000476, p < 0.05), implying that R&D-related tax incentives effectively promote firms’ green sustainability outcomes. This suggests that preferential tax policies targeting innovation exert a real and measurable influence on the greening of enterprise activities. In contrast, the tax incentive gap (TIG) shows no significant relationship with GSDP, indicating that implicit tax burdens or deviations between statutory and effective tax rates do not directly enhance firms’ sustainability performance. This finding aligns with the argument that structural tax differences may not necessarily translate into meaningful incentives for green transformation.

As we shown in

Table 15. Diagnostic Tests for System GMM Estimation.

Test	Statistic	p-Value
AR(1)	−25.4	0
AR(2)	1.08	0.08
Hansen	χ2(12) = 14.2	0.29
Number of instrumental variables	18	—

Source: calculated by the authors.

, The AR(1) test statistic for this model is −25.4, with a p-value of 0. This indicates that we strongly reject the null hypothesis of “no first-order autocorrelation”. Significant first-order autocorrelation exists in the first-difference residuals, consistent with the expectations of the dynamic panel model. The AR(2) test statistic is 1.08, and the p-value is 0.08. At the conventional 5% significance level, the null hypothesis of “no second-order autocorrelation” cannot be rejected. This indicates that there is no significant second-order serial correlation in the differenced residuals (at the 5% level), meaning that the original error term is essentially autocorrelation-free at the level of the difference. The Hansen test result was χ²(12) = 14.2, with a p-value of 0.29. This p-value is significantly higher than 0.05, therefore the null hypothesis of “instrumental variables being effective” cannot be rejected. This means that there is no statistical evidence to refute the effectiveness of the 18 instrumental variables used, i.e., they are generally uncorrelated with the error term, and the overidentification constraint of the model holds.

3.5. Robustness Tests

Replacing the lagged terms of the main independent variables. Using the one-year lag of the tax-incentive indicators to explain current GSDP in Table 6, we find that last year’s RDSR remains positively associated with current GSDP, whereas lagged TIG is small and insignificant Because current sustainability performance cannot affect last year’s tax incentives, this lagged specification mitigates reverse causality and makes the positive association more credible. The magnitude of the lagged RDSR effect is smaller than in the baseline (consistent with temporal attenuation) but remains economically and statistically meaningful.

Table 16. Robustness test.

Variables	Coefficient	Std. Err.
R1
TIG (pp-equivalent ratio)	0.0043	−0.0063
RDSR (ratio)	0.1302 ***	−0.0441
Revenue Growth (ratio)	−0.0077 ***	−0.0015
Tobin’s Q	0.0018 ***	−0.0006
Firm Size (ln assets)	0.0510 ***	−0.0026
Asset Growth (ratio)	0.0016	−0.0017
Board Size	−0.0001	−0.0011
Firm Age (years)	−0.0026 ***	−0.0008
Independent Director Ratio	0.0370	−0.0273
Top-1 Ownership	−0.0045	−0.0164
Current Ratio	0.0012 **	−0.0005
	Within ${\mathrm{R}}^2$	0.0545
	Observations	30,483
	Firms	2241
R2
TIG (pp-equivalent ratio)	0.003348	−0.0069
RDSR (ratio)	0.1700 ***	−0.0497
Revenue Growth (ratio)	−0.0099 ***	−0.0021
Tobin’s Q	0.0016 **	−0.0008
Firm Size (ln assets)	0.0486 ***	−0.0027
Asset Growth (ratio)	0.0040 *	−0.0024
Board Size	0.0003	−0.0012
Firm Age (years)	−0.0024 ***	−0.0008
Independent Director Ratio	0.0462	−0.0289
Top-1 Ownership	−0.0009	−0.016
Current Ratio	0.0012 *	−0.0006
	Within ${\mathrm{R}}^2$	0.0517
	Observations	30,483
	Firms	2241
R3
TIG (pp-equivalent ratio)	0.004097	−0.0068
RDSR (ratio)	0.1226 ***	−0.0441
Revenue Growth (ratio)	−0.0083 ***	−0.0016
Tobin’s Q	0.0019 ***	−0.0006
Firm Size (ln assets)	0.0497 ***	−0.0026
Asset Growth (ratio)	0.0026	−0.0017
Board Size	−0.0003	−0.0010
Firm Age (years)	−0.0025 ***	−0.0008
Independent Director Ratio	0.0351	−0.0274
Top-1 Ownership	−0.0126	−0.0164
Current Ratio	0.0013 ***	−0.0004
	Within ${\mathrm{R}}^2$	0.0539
	Observations	26,176
	Firms	2241
R4
L1.TIG	−0.0020	−0.0064
L1.RDSR	0.0738 *	−0.0433
Revenue Growth (ratio)	−0.0083 ***	−0.0016
Tobin’s Q	0.0019 ***	−0.0006
Firm Size (ln assets)	0.0514 ***	−0.0028
Asset Growth (ratio)	0.0006	−0.0018
Board Size	0.0003	−0.0011
Firm Age (years)	−0.002176 *	−0.0013
Independent Director Ratio	0.039839	−0.0287
Top-1 Ownership	−0.000164	−0.0168
Current Ratio	0.001180 **	−0.0005
	Within R2	0.0511
	Observations	29,222
	Firms	2241
(R5) Baseline (cluster = year)
TIG (pp-equivalent ratio)	0.0043	−0.0060
RDSR (ratio)	0.1302 ***	−0.0494
Revenue Growth (ratio)	−0.0077 ***	−0.0022
Tobin’s Q	0.0018 ***	−0.0005
Firm Size (ln assets)	0.0510 ***	−0.0035
Asset Growth (ratio)	0.0016	−0.0043
Board Size	−0.0001	−0.0006
Firm Age (years)	−0.0026 ***	−0.0003
Independent Director Ratio	0.0369 ***	−0.0101
Top-1 Ownership	−0.0045	−0.0149
Current Ratio	0.0012 ***	−0.0005
	Within R2	0.0545
	Observations	30,483
	Firms	2241

Note: p * < 0.1; p ** < 0.05; p *** < 0.01. Source: calculated by the authors.

aggregates robustness checks—alternative winsorization, excluding 2020–2021, lagged regressors, and alternative clustering—and the positive effect of RDSR on GSDP persists (5–10% significance), while TIG remains insignificant. These robustness results, together with the baseline and mediation evidence, support the conclusion that R&D-oriented tax incentives enhance firms’ green performance primarily by stimulating R&D investment.

As shown in Table 16, through a battery of robustness checks—alternative winsorization, excluding 2020–2021, using one-year lags of the policy variables, and alternative clustering—we confirm that our main result is stable: R&D-oriented tax incentives (RDSR) are positively associated with firms’ green sustainable development performance (GSDP). In the lagged specification, last year’s RDSR remains positive and marginally significant (10%), while lagged TIG is small and insignificant, which mitigates concerns about reverse causality. Across specifications, the RDSR effect weakens somewhat but remains economically meaningful; TIG shows no stable direct effect on average. Given that identification relies on fixed effects and lagged instruments, we describe the results as evidence consistent with a causal mechanism rather than definitive proof.

4. Results

4.1. Enterprise Size

To examine whether the effect of tax incentives varies by firm size, we interact with the policy variable with a pre-determined size indicator. A firm is classified as Large if its baseline (first observed year) total assets lie above the 80th percentile within its CSRC industry–year; otherwise, it is Small/Medium.

$$\mathrm{GSD}{\mathrm{P}}_{\mathrm{it}}={\mathsf{\beta }}_1{\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{R}}_{\mathrm{it}}+{\mathsf{\beta }}_2({\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{R}}_{\mathrm{it}}\times {\mathrm{L}\mathrm{a}\mathrm{r}\mathrm{g}\mathrm{e}}_{\mathrm{i}}^{\mathrm{p}\mathrm{r}\mathrm{e}})+{\mathrm{X}}_{\mathrm{it}}^{\prime }\mathsf{\gamma }+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(2)

The effect for Small/Medium firms equals ${\mathsf{\beta }}_1$, while that for large firms equals ${\mathsf{\beta }}_1+{\mathsf{\beta }}_2$, We report Wald tests (p-values and 95% CIs). Firm and year fixed effects are included; standard errors are clustered at the firm level. As for robustness, we repeat the exercise with alternative cutoffs (median, 70th, 80th) and for TIG in place of RDSR.

To verify the threshold sensitivity test: we change the “large and small enterprises” demarcation method to see whether the conclusion is robust, and intercept at the 80th, 70th and 50th percentiles (Figure 5 and

Table 17. Estimation results.

Main Var	Cutoff	${\mathit{\beta }}_1$	se1	p1	${\mathit{\beta }}_2$	se2	p2	Large Effect (b1 + b2)	se_Large	p_Large	R2 Within
RDSR	80	0.0827	0.0463	0.0736	0.4227	0.1226	0.0006	0.5055	0.1177	0.0000	0.0554
TIG	80	0.0043	0.0070	0.5385	0.0080	0.0172	0.6418	0.0123	0.0156	0.4281	0.0537
RDSR	70	0.0387	0.0479	0.4187	0.4635	0.0985	0.0000	0.5022	0.0916	0.0000	0.0564
TIG	70	0.0018	0.0076	0.8110	0.0134	0.0143	0.3495	0.0152	0.0121	0.2083	0.0538
RDSR	50	−0.0147	0.0506	0.7716	0.3634	0.0819	0.0000	0.3487	0.0703	0.0000	0.0562
TIG	50	0.0075	0.0088	0.3971	−0.0029	0.0129	0.8242	0.0046	0.0093	0.6228	0.0537

Source: calculated by the authors.

). Blackline is error bars.

TIG by firm size. Across size groups, the effect of TIG on GSDP is small and statistically insignificant. For SMEs, the coefficient on TIG is close to zero (e.g., p80 cutoff: β₁ = 0.004334, s.e. 0.007047), and for large firms the marginal effect β₁ + β₂ is also insignificant (0.012340, s.e. 0.015573). These patterns hold under alternative cutoffs (median, 70th, 80th). Hence, we do not find robust size heterogeneity for the general tax-burden measure TIG. This confirms Hypothesis H5, which states that there is significant size heterogeneity between large and small enterprises: the TI prompt GSDP channel for SMEs is negative, while for large enterprises it is positive but unstable on average.

4.2. Ownership Concentration

We use a pre-determined ownership concentration indicator, ${\mathrm{Highconc}}_{\mathrm{i}}^{\mathrm{pre}}$, equal to one if the firm’s baseline (first observed year) Top-1 share exceeds the industry–year percentile cutoff (main: median; robustness: 70th/80th). We estimate in formula 3, firm- and year-fixed-effects models with an interaction between the tax-policy variable and ${\mathrm{Highconc}}_{\mathrm{i}}^{\mathrm{pre}}$:

$$\mathrm{GSD}{\mathrm{P}}_{\mathrm{it}}={\mathsf{\beta }}_1{\mathrm{P}\mathrm{o}\mathrm{l}\mathrm{i}\mathrm{c}\mathrm{y}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\beta }}_2{(\mathrm{Policy}}_{\mathrm{i},\mathrm{t}}\times {\mathrm{H}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{c}}_{\mathrm{i}}^{\mathrm{p}\mathrm{r}\mathrm{e}})+{\mathrm{X}}_{\mathrm{it}}^{\prime }\mathsf{\gamma }+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(3)

where for Policy $\in (\mathrm{RDSR},\mathrm{TIG})$ we run the specification separately for RDSR and TIG. The marginal effect for low-concentration firms equals ${\mathsf{\beta }}_1$, and for high-concentration firms equals ${\mathsf{\beta }}_1$+${\mathsf{\beta }}_2$. Control variables follow the baseline set excluding R&D intensity and log R&D, to avoid conditioning away the mediation channel. Standard errors are clustered at the firm level.

For TIG neither the low-concentration nor the high-concentration group exhibits a statistically significant effect on GSDP (

Table 18. Regression Results by Ownership Concentration: High vs. Low.

Cutoff	Low (β1)	Interaction (β2)	High = β1 + β2	Within R2
p50 (median)	−0.0260	0.3110 ***	0.2840 *	0.0556
	(0.0550)	(0.076)	(0.061)
p70	0.070	0.2470 **	0.317 *	0.0549
	(0.046)	(0.097)	(0.091)
p80	0.090 *	0.2810 **	0.371 *
	(0.046)	(0.118)	(0.113)	0.0548
p50 (median)	−0.003	0.018	0.014	0.0538
	(0.009)	(0.013)	(0.009)
p70	0.006	0.001	0.007	0.0537
	(0.007)	(0.015)	(0.013)
p80	0.006	−0.001	0.005	0.0537
	(0.007)	(0.018)	(0.017)

Note: p * < 0.1; p ** < 0.05; p *** < 0.01. Source: calculated by the authors.

and Figure 6). The point estimates are small in magnitude (Low ≈ −0.003; High ≈ 0.014) and their 95% confidence intervals include zero; a Wald test on the interaction fails to reject equality across groups. Hence, we find no robust evidence that general tax relief affects green performance, nor that its effect varies with ownership concentration. Accordingly, H4 is not supported for TIG.

4.3. Regional Differences

We examine regional heterogeneity by re-estimating the baseline FE model without the mediator separately for Eastern, Central and Western firms (CSRC regional classification, time-invariant at the firm’s entry year) in Formula (4). Specifically,

$${\mathrm{GSDP}}_{\mathrm{i},\mathrm{t}}={\mathsf{\beta }}_1^{\left(\mathrm{r}\right)}{\mathrm{R}\mathrm{D}\mathrm{S}\mathrm{R}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\beta }}_2^{\left(\mathrm{r}\right)}{\mathrm{T}\mathrm{I}\mathrm{G}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\gamma }}^{\left(\mathrm{r}\right)}{\mathrm{X}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\mu }}_{\mathrm{i}}+{\mathsf{\lambda }}_{\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}},\mathrm{r}\mathsf{ϵ}\left\{\mathrm{E}\mathrm{a}\mathrm{s}\mathrm{t},\mathrm{C}\mathrm{e}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{a}\mathrm{l},\mathrm{W}\mathrm{e}\mathrm{s}\mathrm{t}\right\}$$

(4)

where ${\mathrm{RDSR}}_{\mathrm{i},\mathrm{t}}$—additional deductions/R&D expenses; ${\mathrm{TIG}}_{\mathrm{i},\mathrm{t}}$—statutory tax rate, effective tax rate (percentage points); ${\mathrm{X}}_{\mathrm{k}}$—represents the set of control variables (firm size, Tobin’s Q, growth rates, board characteristics, etc.); ${\mathsf{\mu }}_{\mathrm{i}},{\mathsf{\lambda }}_{\mathrm{t}}$—firm and year fixed effects. Standard errors are clustered at the firm level in all models.

Regional heterogeneity is evident (

Table 19. Regional Fixed-Effects Panel Regression Results.

Variable
TIG (Tax Incentive Gap)	−0.0033 (0.0049)	0.0045 (0.0075)	0.0029 (0.0083)
RDSR (R&D Super-deduction, share)	0.1168 ** (0.0523)	0.4386 *** (0.1052)	−0.1052 (0.0996)
Revenue Growth rate	−0.0088 *** (0.0020)	−0.0084 *** (0.0032)	−0.0041 (0.0029)
Asset Growth rate	0.0040 * (0.0022)	−0.0009 (0.0038)	−0.0072 ** (0.0035)
Tobin’s Q	0.0022 *** (0.0008)	0.0003 (0.0014)	0.0028 ** (0.0012)
Firm Size (ln Assets)	0.0491 *** (0.0034)	0.0558 *** (0.0061)	0.0477 *** (0.0052)
Board Size	−0.0003 (0.0015)	0.0020 (0.0019)	−0.0029 (0.0026)
Firm Age (years)	−0.0018 * (0.0009)	−0.0021 (0.0022)	−0.0044 (0.0030)
Independent Director %	0.0430 (0.0361)	0.0672 (0.0530)	−0.0288 (0.0569)
Top 1 Shareholder %	−0.0028 (0.0202)	−0.0210 (0.0349)	0.0361 (0.0395)
Current Ratio	0.0014 ** (0.0006)	0.0023 * (0.0012)	−0.0011 (0.0010)
Firm FE & Year FE	Yes	Yes	Yes
Observations	21,252	5676	4727
Within R2	0.047	0.084	0.064

Note: p * < 0.1; p ** < 0.05; p *** < 0.01. Source: calculated by the authors.

and Figure 7). In the East, the coefficient on RDSR is positive and statistically significant, while it is statistically indistinguishable from zero in the Central region and negative and significant in the West; the effect of TIG remains economically small across regions. These estimates indicate that R&D-oriented tax incentives translate into green performance mainly in the East, whereas the transmission weakens—and even reverses—in the West. This pattern supports Hypothesis H5 on regional heterogeneity and highlights the need for region-specific policy design. Blackline is error bars.

5. Discussion

Our findings provide robust support for the positive impact of tax incentives on corporate green sustainability, and they illuminate the critical role of innovation in this process. In the baseline analysis, the R&D Super-Deduction Ratio (RDSR), a targeted tax incentive for innovation—shows a significant positive effect on firms’ Green Sustainable Development Performance (GSDP), with an estimated coefficient around 0.13 (p < 0.01). In contrast, a broader tax relief measure has a near-zero and statistically insignificant effect. Substantively, a one-standard-deviation increase in RDSR is associated with roughly a 1.5–2.0% rise in the GSDP index, a non-trivial gain even for publicly listed firms. This contrast between RDSR and TIG implies that not all tax incentives are equally effective. Those explicitly encouraging R&D investment translate into meaningful sustainability improvements, whereas generalized tax cuts alone do not. Addressing potential endogeneity strengthens this conclusion: using lagged tax policy exposure as instruments yields similar results, suggesting a causal relationship driven by the innovation channel. These outcomes confirm our first hypothesis that tax incentives can drive green performance, and they align with the Porter Hypothesis in that well-designed fiscal policies can induce innovation to offset environmental compliance costs. In short, tax incentives materially improve corporate sustainability primarily by spurring innovation, rather than by simply alleviating tax burdens [45,46].

Crucially, our analysis reveals that innovation input is the pivotal mediating mechanism linking tax incentives to sustainability outcomes [47]. When we include firms’ R&D investment in the models, the direct effect of tax incentives on GSDP shrinks markedly while the effect of R&D remains positive and significant [48]. This indicates a partial mediation structure: tax incentives work indirectly by encouraging firms to invest in innovation, which in turn boosts green performance. In other words, firms leverage tax savings to fund R&D, yielding greener technologies and better sustainability metrics, consistent with our second hypothesis [49,50]. This finding helps explain mixed results in prior literature—tax breaks alone may fail to improve environmental outcomes unless they stimulate innovation [51]. From a policy perspective, these results underscore that targeted R&D tax credits or super-deductions should be prioritized to drive sustainability, whereas broad tax reductions are unlikely to deliver significant environmental benefits [52]. Effective green tax policy thus means trying incentives to innovation activity, ensuring that fiscal support translates into tangible eco-friendly innovation rather than just higher after-tax profits.

The effectiveness of tax incentives is not uniform across all firms. It depends on firms’ internal characteristics. We find that corporate governance structures, particularly ownership concentration, significantly moderate the impact of tax incentives on green performance [53,54]. Firms with more concentrated ownership experience a stronger positive effect of tax incentives on GSDP, supporting our hypothesis about governance (H3). In such firms, decision-making is stable and streamlined, so tax savings are more likely to be channeled into productive green investments. By contrast, in firms with diffuse ownership, the impact is muted—possibly because bureaucratic frictions or agency problems make it harder to convert tax benefits into innovative projects [55,56]. This result is in line with corporate governance theory and emerging-market studies that highlight how focused oversight can improve strategic resource allocation. It implies a clear boundary condition; tax incentives yield the greatest sustainability payoff when coupled with sound internal governance [57,58,59]. Policymakers might therefore consider complementary measures that encourage or require firms to have transparent, accountable mechanisms for deploying tax benefits towards innovation. Strengthening governance, especially in Chinese companies where ownership structures vary widely, can amplify the effectiveness of fiscal policies aimed at sustainability [60,61].

Firm size and market position also emerge as important contingencies in how tax incentives drive green outcomes. Consistent with expectations, larger or leading enterprises exhibit stronger baseline green capabilities (H4) and are generally better positioned to utilize tax incentives effectively [62,63].

Specifically, small firms and firms with concentrated ownership benefit more from tax incentives, potentially due to better governance and quicker decision cycles. Additionally, regional disparities emphasize the uneven impact of fiscal tools, with eastern regions responding more strongly due to better infrastructure and innovation ecosystems [64,65].

However, some limitations persist. The measurement of green sustainable performance is limited to available proxies such as environmental disclosures and R&D spending. Future work could integrate carbon emissions or ESG scores for a more nuanced evaluation.

6. Conclusions

This study examines the heterogeneous effects of tax incentives on firms’ green sustainable development performance using a comprehensive panel of Chinese listed companies. The results reveal three main findings. First, the RDSR significantly improves firms’ green performance, whereas the TIG does not exhibit a measurable effect. This contrast suggests that innovation-oriented fiscal instruments are more effective in promoting substantive green upgrading than broad-based tax reductions. Second, the mediation analysis demonstrates that innovation investment is a key transmission channel through which RDSR enhances green outcomes, underscoring the importance of cost-reducing incentives that directly support green R&D. Third, the influence of RDSR varies across firm characteristics and regional contexts, indicating that policy effectiveness depends on firms’ technological foundations and the institutional environment in which they operate.

These findings offer several policy implications. Policymakers should prioritize targeted tax incentives that explicitly encourage environmental innovation, rather than relying on general tax reductions whose effects on green transformation appear limited. Strengthening the alignment between fiscal incentives and environmental objectives may also help reduce the potential for tax-driven behavioral distortions. Moreover, coordinating tax incentives with complementary instruments—such as green finance, innovation subsidies, and environmental disclosure requirements—can promote the conversion of innovative efforts into genuine sustainable development outcomes. Despite these contributions, the study has several limitations. First, although the green performance index captures key dimensions of environmental improvement, it primarily relies on patent applications and disclosure quality, which may not fully reflect all aspects of firms’ environmental behavior. Second, while the empirical strategy incorporates multiple robustness checks, unobserved institutional or managerial factors may still influence both innovation decisions and green performance. Third, the analysis focuses on listed firms, and the results may not fully generalize to smaller or non-listed enterprises with different financial constraints. Overall, the study provides differentiated empirical evidence on the effectiveness of tax incentives and offers actionable insights for designing a green-oriented fiscal system that supports high-quality and sustainable development.

To enhance external validity, future research could incorporate data from privately owned or unlisted firms, as well as comparable enterprises in other transition economies where state intervention and institutional conditions resemble those of China. In the absence of such data, it is important to acknowledge this contextual specificity in the methodology and conclusion sections, emphasizing that while the findings may not be universally generalizable, they nonetheless provide valuable insights into the mechanisms of tax incentives and green transformation within state-influenced transitional economies.