Green Innovation or Expedient Compliance: Carbon Emission Reduction by Heavily Polluting Enterprises Under Green Finance Reform and Innovation Pilot Zone

Abstract

The effective design of green financial policies is crucial for balancing the operational pressures of heavily polluting enterprises with the goal of sustained carbon emission reduction. This study investigates the impact of the Green Finance Reform and Innovation Pilot Zone (GFRIPZ) policy by employing a multi-period difference-in-differences (DID) model based on firm-level panel data from 2012 to 2021, covering A-share listed enterprises in Shanghai and Shenzhen. The results show that GFRIPZs significantly reduced carbon emissions in pilot regions, with heterogeneous effects observed across enterprise types—particularly among large enterprises, state-owned enterprises, and those located in financially developed areas. To uncover the underlying mechanisms, we compare two behavioral responses: green innovation, marked by long-term investment in green technologies, and expedient compliance, involving short-term, strategic compliance behaviors. Our findings indicate that GFRIPZs did not effectively promote green innovation. Instead, it has encouraged a shift from productive capital investment toward un-productive, symbolic actions aimed at fulfilling policy requirements. These responses risk undermining the long-term objective of green transformation and may contribute to a broader shift from real economic activity toward speculative or less productive investments, raising concerns about the quality and sustainability of the low-carbon transition.

1. Introduction

In 2020, China proposed the goals of peak carbon emissions and carbon neutrality to the international community, signifying that its economy and society are undergoing a wide-ranging and profound transformation. Heavily polluting enterprises, as key sources of carbon emissions, play a pivotal role in advancing the green economic transition, which is vital for realizing carbon peaking and neutrality targets and promoting sustainable socio-economic development [1]. To strengthen the financial frameworks supporting green development and assess the effectiveness of financial tools in driving industrial green transformation, China launched the Green Finance Reform and Innovation Pilot Zone (GFRIPZ) in 2017, selecting eight regions within five provinces as pilot areas. This initiative marks an important transition from blueprint to practice in China’s efforts to advance green financing.

The GFRIPZ policy primarily seeks to enhance the efficiency of financial resource distribution in designated areas, mobilize financial instruments in support of environmentally sustainable projects, and limit capital access for heavily polluting enterprises, thereby encouraging their shift toward green development [2,3]. By creating a more restrictive financial environment and enforcing higher emission reduction obligations for heavily polluting enterprises, the policy delivers a dual message of encouraging ecological transition and deterring environmental degradation [4]. In this context, heavily polluting enterprises may pursue green technological innovation as a means to shift toward environmentally friendly production methods and products, thereby achieving carbon emission reductions [5,6]. This represents a trend-driven innovative solution. However, given that empirical outcomes often deviate from the intent underlying policy, GFRIPZs may incentivize heavily polluting enterprises to adopt strategic responses. Specifically, to escape the constraints of environmental compliance, these enterprises may temporarily adjust their investment-related behavior by reducing their production to reduce their carbon emissions [7]. Such an approach constitutes a short-term expedient response that mitigates surface-level manifestations of the problem while leaving its fundamental drivers unaddressed.

Therefore, this study seeks to address the following research questions. First, can GFRIPZs curb carbon emissions from heavily polluting enterprises within the jurisdiction in which they have been implemented? Second, do the mechanisms through which these policies impact reduction in carbon emissions by corporations align with the original aim of establishing the pilot zones? Specifically, when facing green financial constraints, do heavily polluting enterprises respond by actively investing in green technological innovation? Or do they instead adopt short-term, expedient compliance strategies involving minimal adjustments? Finally, how can GFRIPZs be improved to encourage heavily polluting enterprises to achieve a green transformation? Exploring these questions contributes not only to a deeper understanding of the micro-level implications of the GFRIPZ initiative but also provides theoretical guidance for policymakers in crafting economic and financial strategies that facilitate finance-driven green transformation in pursuit of dual-carbon targets.

A substantial body of literature has focused on evaluating the influence of the GFRIPZ policy on carbon emission reductions at broader macro or sectoral scales. The findings of these scholars suggest that the enforcement of the policy has the potential to lower carbon emissions in targeted areas [8,9,10]. From the perspective of enterprises, existing research has predominantly focused on mechanisms for improving environmental performance and controlling pollution. It has been argued that the GFRIPZ initiative fosters these outcomes by stimulating enterprises to invest in green innovation [3,11,12]. Another body of scholars contends that the GFRIPZ initiative reshapes the scale of external financing available to enterprises by optimizing the distribution of financial resources. As a result, enterprises are driven to reorient their investment behavior, limiting capital flows to high-emission production and ultimately contributing to emission reductions [13,14,15]. Considering the significant role that heavily polluting enterprises play in carbon mitigation, it is imperative to deepen the analysis of how varying implementation approaches influence the durability of corporate emission reductions and the effective advancement of green transformation. Current research largely overlooks a comparative evaluation of the two fundamental pathways—green innovation and strategic investment—through which heavily polluting enterprises react to green finance policy interventions. Building on theoretical analysis, this study utilizes panel data on heavily polluting enterprises listed on the A-share markets in Shanghai and Shenzhen from 2012 to 2021 to investigate the mechanisms through which the GFRIPZ policy influences carbon emission reductions among high-emission enterprises in China’s pilot regions.

This research advances the literature through several marginal contributions. First, whereas previous studies have largely centered on the influence of GFRIPZs on environmentally friendly or general enterprises, our analysis presents novel micro-level insights into its effects on carbon emission mitigation within heavily polluting enterprises. Second, our work captures the heterogeneous responses of heavily polluting enterprises in comparison with general enterprises when facing environmental regulations and analyzes whether these enterprises actively promote green technological innovation or merely engage in strategic investments. Grasping these mechanisms is essential for formulating green financial policies that effectively reconcile operational demands with the objectives of green transition. Finally, we explore the reasons why heavily polluting enterprises opt for the strategies of trend-driven innovation and expediency, and examine whether such behavior hinders a sustainable reduction in carbon emissions and green transformation. We further examine the potential shift from the real economy to the virtual economy, induced by the policy. These findings offer valuable insights for the continued refinement of the GFRIPZ initiative.

2. Theoretical Analysis and Research Hypotheses

2.1. GFRIPZ and Carbon Emissions

As a market-oriented regulatory instrument, green finance policy seeks to address environmental problems by integrating environmental costs into financial decisions, thereby internalizing the adverse externalities associated with corporate environmental degradation. This, in turn, exerts a significant influence on the behavioral choices of actors in financial markets [16,17,18]. China operates a bank-dominated financial system in which financial resource allocation predominantly occurs via indirect financing channels facilitated by commercial banks [19,20,21,22,23,24,25]. For an extended period, financial capital has predominantly been directed toward asset-intensive sectors like steel, petroleum, and chemicals, which are generally characterized by lower credit risk [26]. Characterized by high pollution levels, the rapid expansion of these industries has significantly aggravated environmental challenges and heightened the urgency of carbon emission mitigation in China [27].

Drawing on institutional theory, enterprises’ behavior is not solely determined by economic efficiency but also by the necessity to conform to external institutional pressures in order to gain legitimacy and secure access to critical resources [28,29]. In response to such pressures, enterprises tend to align their strategies and actions with prevailing policy expectations to ensure organizational survival and maintain legitimacy within a dynamically evolving institutional environment. This theoretical perspective helps to explain the behavioral responses of heavily polluting enterprises under the GFRIPZ policy. As a state-led institutional arrangement, GFRIPZs exert coercive institutional pressure by embedding environmental objectives into financial regulations. Following the introduction of the GFRIPZ policy, the pilot zones have actively pursued carbon peaking and neutrality targets through innovations in carbon finance, carbon accounting, and carbon sink initiatives. By imposing stringent restrictions on lending to projects characterized by high energy consumption and significant emissions, authorities have effectively constrained the financing channels available to heavily polluting enterprises. This includes the imposition of punitive high-interest loans on heavily polluting enterprises in the credit markets within these jurisdictions, and the adoption a “one-vote veto” system to refuse loans [30]. GFRIPZs have exerted notable influence on the financial sector, as evidenced by research highlighting its role in tightening financing restrictions and channeling financial resources toward green enterprises and sustainable projects [31]. These pressures have compelled heavily polluting enterprises to adopt measures that satisfy environmental protection requirements and alleviate financing constraints [32]. In doing so, enterprises are driven to adjust their behavior in line with policy expectations, ultimately leading to reductions in carbon emissions.

As heavily polluting enterprises face greater pressure to reduce carbon emissions, they may exhibit distinct strategic adjustments and environmental outcomes. However, existing research on the relationship between GFRIPZs and corporate carbon emissions has primarily focused on environmentally friendly or average enterprises, neglecting the heterogeneous behavioral responses of heavily polluting enterprises [32,33]. This study takes heavily polluting enterprises as the focal research subject and proposes the following hypothesis:

H1: 

The implementation of GFRIPZs contributes to a reduction in carbon emissions from heavily polluting enterprises.

2.2. GFRIPZs and Green Technological Innovation

Research on green technological innovation originated in the 1980s and 1990s. The exploration of its conceptual connotations has gone through four stages: response orientation, governance orientation, goal orientation, and content orientation [34,35]. Green technological innovation is widely recognized in academic discourse as the incorporation of new products, processes, services, and managerial approaches aimed at reducing environmental risks, lowering pollution levels, and minimizing the negative impacts of resource use across their entire lifecycle [36]. Because the gains from green technological innovation predominantly contribute to public welfare and cannot be fully internalized by enterprises, market-driven incentives are typically inadequate to motivate enterprises to invest in these innovations. Therefore, environmental policies must be implemented to provide additional support [37]. The Porter Hypothesis suggests that well-designed policies for environmental regulation can promote technological innovation by enterprises [38]. Organizational legitimacy theory provides an important lens to understand enterprises’ behavioral responses to environmental policies. According to Suchman (1995) [39], organizations seek legitimacy to ensure social acceptance, build trust with stakeholders, and maintain access to essential resources. In this context, adopting visible green practices—such as investing in green technologies—can serve as a means to obtain moral and cognitive legitimacy, particularly under increasing external scrutiny from governments, investors, and the public. For heavily polluting enterprises, aligning with the expectations set by the GFRIPZ policy may help reinforce their legitimacy within the institutional field, thereby securing their organizational survival. Existing studies have explored how GFRIPZs encourage corporate green technological innovation, primarily from the dual perspectives of constraints and incentives. The results suggest that these policies not only impose more stringent environmental standards and finance-related pressure on enterprises to penalize polluting behaviors but also reward their green technological innovation. This strategy safeguards the unique competitive edge of enterprises’ green technologies and products, thus motivating them to pursue green technological innovation [12,40,41,42].

Nonetheless, some researchers have expressed skepticism regarding the consistency of the Porter Hypothesis with firm-level behavior, contending that the influence of environmental regulation on green innovation within enterprises is not definitively established [43,44]. On one hand, both the ambiguous impact of environmental regulation on green technological innovation and the varying intensity of GFRIPZs contribute to a growing disconnect between theoretical assumptions and practical outcomes. Current research remains divided on how the intensity of environmental regulation affects corporate green technological innovation. The extant literature has thus far produced equivocal results, with some studies suggesting a positive linear relationship between the two phenomena, while others have proposed a “U-shaped” or “inverted U-shaped” relationship instead [45,46]. The strength of environmental regulation within the GFRIPZ framework remains indeterminate. While this policy constrains the access of heavily polluting enterprises to credit financing, it does not extend its influence to alternative external or internal funding sources, nor does it involve the imposition of mandatory administrative penalties. Conversely, given the high-risk profile, significant transition costs, and external benefits linked to innovation activities, enterprises’ decisions to invest in such endeavors are highly contingent upon the availability of sufficient financial support [47,48,49]. Research has shown that obtaining external financing can significantly increase corporate investment in R&D and enhance the level of technological innovation [50]. However, the GFRIPZ initiative has tightened financing constraints for heavily polluting enterprises, further complicating their efforts to pursue green technological innovation, which inherently demands considerable capital and long-term commitment.

The uncertainty around the direction of the relevant policy after the conclusion of the five-year pilot period has also led to a sense of complacency among enterprises. If the cost of complying with environmental regulations is perceived to be lower than the economic benefits derived from extensive development, profit-maximizing managers are generally disincentivized to initiate green technological innovation proactively [51,52]. While the Porter Hypothesis provides a compelling rationale for regulation-induced innovation, this optimistic view assumes consistent enforcement and sufficient incentives. In practice, however, the outcomes of the GFRIPZs are mixed. The tightening of financial constraints and ambiguity in enforcement may not encourage genuine green innovation but instead incentivize symbolic compliance and strategic adaptation. That is, enterprises may signal environmental commitment through low cost or superficial actions without undertaking substantive technological transformation. This behavior, aimed at maintaining organizational legitimacy, aligns with organizational legitimacy theory yet deviates from the innovation centric expectations of the Porter Hypothesis. Furthermore, the risk of greenwashing, misrepresenting environmental performance to gain reputational advantage, becomes more pronounced in a weak enforcement environment. This shows that whether the Green Finance Innovation Policy can promote green technological innovation by enterprises and thereby reduce their carbon emissions still requires verification. Therefore, we propose the following:

H2: 

GFRIPZs fail to achieve a reduction in carbon emissions by promoting green technological innovation among heavily polluting enterprises.

2.3. GFRIPZ and Strategic Investment

Resource dependence theory posits that organizational behavior is shaped not only by internal goals but also by the constraints and opportunities imposed by the external environment. According to Reitz and Pfeffer (1979) [53], when enterprises encounter resource scarcity resulting from environmental shocks or policy interventions, they adopt adaptive strategies to maintain organizational autonomy and ensure continued access to critical resources. These responses often involve adjustments to internal structures, a realignment of resource allocation, and modifications in investment behavior. In the context of green finance reform, the GFRIPZ policy imposes substantial constraints on access to financial capital for heavily polluting enterprises. By tightening the financing environment, the policy increases external dependence while reducing enterprises’ control over critical resources, especially credit. According to resource dependence theory, such pressure compels enterprises to take strategic actions to buffer against environmental uncertainty and maintain operational stability. Enterprises in heavily polluting industries are generally larger in scale and possess a higher proportion of fixed assets, which lends them better convenience in financing, compared with other enterprises. This provides them with stronger motivation for expanding their industrial investments. Large-scale productive investment is a direct contributor to an increase in carbon emissions by corporations [54,55,56,57,58,59]. If a company’s investment activities aim to expand its scale of production, this will directly increase its energy consumption and generate more carbon emissions. The more stringent environmental standards imposed under GFRIPZs have prompted heavily polluting enterprises to proactively slow down their core business investments and avoid disorderly expansion, thereby resulting in a reduction in their carbon emissions.

Moreover, as enterprises will not abandon their pursuit of profit maximization, they may shift to non-productive investments when the returns from their productive investments decline owing to green regulation. Financial investment can serve as a useful indicator of the propensity of heavily polluting enterprises to engage in non-productive investments. Research suggests that allocating financial assets can help mitigate the decline in effective demand caused by falling profits in the real economy, guard against risks of liquidity, and broaden financing channels while improving the efficiency of financing [60,61]. When the GFRIPZ policy increases the costs and difficulty of financing for heavily polluting enterprises, and reduces their main business revenue, financial investments that yield short-term gains become a rational choice for these enterprises based on a trade-off between risk and return [62,63]. Their preference for non-productive investments has a crowding-out effect on productive investments, where this further reduces their carbon emissions [64,65,66]. Based on this, we propose the following:

H3: 

GFRIPZ policy reduces emissions by discouraging productive investment and encouraging strategic adjustments.

3. Research Design

3.1. Sample Choice and Data Sources

This study employs panel data from all A-share-listed heavily polluting enterprises spanning the period from 2012 to 2021. To refine the sample, we excluded enterprises in the financial industry, as well as those designated as ST, *ST, or PT. Companies that went public in or after 2012, along with those exhibiting significant data deficiencies, were also removed to ensure data integrity. Macro-level variables such as energy consumption and GDP were collected from various editions of the China Statistical Yearbook, while provincial deposit and loan data were sourced from the China Financial Yearbook. Policy-related information was obtained from publicly issued regulations by the central government. Patent data were retrieved from the National Intellectual Property Administration, and firm-level financial and operational data were drawn from the CSMAR database. The final dataset consisted of 4990 firm-year observations. To minimize the influence of extreme values, continuous variables were winsorized at the 5th and 95th percentiles within each industry group.

3.2. Model Design

Since the pilot zones were established in batches and were progressively expanded, we adopted a multi-period difference-in-differences (DID) model to examine the impact of GFRIPZs on carbon emissions from heavily polluting enterprises. The use of this model is justified by the quasi-natural experimental nature of the policy implementation. The designation of pilot zones was driven by top-down administrative decisions based on regional-level factors such as green development foundations, environmental governance capacity, and financial infrastructure, rather than firm-level emission characteristics. Enterprises have no influence over whether their region is selected for the pilot, and thus the policy shock can be regarded as exogenous to individual enterprises. Furthermore, the clear temporal and spatial boundaries of the GFRIPZs provide a credible setting for causal identification, minimizing concerns regarding reverse causality. The model is as follows:

$${CO}_{2it}={\beta }_0+{\beta }_1\times {Event}_{it}+{\mu }_i{X}_{it}+{\lambda }_i+{\nu }_t+{\epsilon }_{it},$$

(1)

where $i$ and $t$ represent the enterprise and year, respectively, ${CO}_2$ denotes carbon emissions by the enterprise, and $Event$ represents the interaction between the dummy variables for time and group, indicating whether enterprise $i$ was subject to the green finance reform and was within the innovation pilot zone in year t. $X$ represents a series of control variables, ${\beta }_0$ is a constant term, ${\beta }_1$ represents the net effect of the GFRIPZ, $\lambda$ and $\nu$ denote individual-fixed and time-fixed effects, respectively, and $\epsilon$ is a random disturbance term.

To further investigate the mechanisms through which Green Finance Reform and Innovation Pilot Zones (GFRIPZs) influence corporate behavior from the perspectives of green technological innovation and strategic investment, and to assess whether heavily polluting enterprises respond with proactive innovation or adopt a stopgap strategy, we constructed the following models to examine the impact of the independent variable on potential mediating mechanisms, drawing on the methodological approach proposed by Jiang (2022) [67]:

$${Mediator}_{2it}={\beta }_0+{\beta }_1\times {Event}_{it}+{\mu }_i{X}_{it}+{\lambda }_i+{\nu }_t+{\epsilon }_{it},$$

(2)

where $Mediator$ represents the mechanism encompassing both “green technological innovation” and “strategic investment”. The definitions of the remaining variables are consistent with those in the model represented by Equation (1).

3.3. Variable Definitions and Descriptive Analysis

3.3.1. Dependent Variable

In this study, corporate carbon emissions (${CO}_2$) serve as the dependent variable. Due to the limited availability of firm-level carbon disclosure data and the difficulties associated with direct measurement, the existing literature often adopts the emission factor approach for estimation. This study follows that convention, with the estimation process detailed as follows:

Step 1: To ensure consistency with the characteristics of China’s carbon emissions, this study calculated emission coefficients for various fossil fuels by integrating their lower heating values, carbon content per unit of calorific value, and carbon oxidation rates. These parameters were obtained from the China Greenhouse Gas Inventory Research and the Updated Guidelines for the Preparation of Greenhouse Gas Inventories (Trial), as summarized in

Table 1. Carbon emission factors.

Fuel	Low Calorific Value (PJ/104 t, 108 m3)	Carbon Content Per Unit Calorific Value (t/TJ)	Oxidation Efficiency
Coal	0.210	26.320	0.940
Coke	0.280	31.380	0.930
Crude Oil	0.430	20.080	0.980
Gasoline	0.440	18.900	0.980
Kerosene	0.440	19.600	0.980
Diesel	0.430	20.200	0.980
Fuel Oil	0.430	21.100	0.980
Natural Gas	3.890	15.320	0.990
Electricity	0.581 (tCO2/MWh)

.

Step 2: Carbon emissions from fuels were calculated based on the methods provided by Liu et al. [68] and Shan et al. [69], as follows:

$${CE}_{ij}={AD}_{ij}\times {NCV}_i\times {CC}_i\times O_i$$

(3)

where ${CE}_{ij}$ refers to carbon dioxide emissions from fossil fuel $i$, ${AD}_{ij}$ is the consumption of fossil fuel $i$, and ${NCV}_i$ is the net calorific value of $i$ denoting carbon dioxide emissions per unit of net calorific value for fossil fuel $i$ and is the efficiency of oxidation of $i$.

We used the method proposed by Chapple et al. [70] to approximate carbon emissions by corporations based on industrial energy consumption:

$${CO}_2^{enterprise}={\displaystyle \frac{C_{enterprise}}{C_{industry}}}\times \sum _i\left(E_i^{industry}\times {EF}_i\right)$$

(4)

where ${CO}_2^{enterprise}$ refers to the estimated carbon dioxide emissions of the enterprise, $C_{enterprise}$ is the enterprise’s main business operating cost, $C_{industry}$ is the total main business operating cost of the industry, $E_i^{industry}$ is the consumption of fossil fuel $i$ at the industry level, and ${EF}_i$ is the carbon dioxide emission factor of fossil fuel $i$. A direct emission factor was used for electricity, while the carbon emission factor was calculated for fossil fuels based on the low calorific value, carbon content per unit calorific value, and rate of carbon oxidation of each kind of fuel.

Despite its practicality, this proxy approach also has certain limitations. First, it assumes a uniform relationship between operating costs and energy consumption across enterprises within the same industry, which may overlook firm-level differences in efficiency. Second, the accuracy of the results largely depends on the quality and representativeness of the emission factors used. To address this, we manually compiled data from Chinese sources to better reflect the actual emission conditions of Chinese enterprises. Third, for enterprises whose cost structures deviate significantly from industry norms, this method may underestimate or overestimate their emissions. Although these limitations may introduce measurement errors, the use of a consistent and standardized method across all enterprises helps mitigate systematic bias and ensures that the relative comparisons underlying the DID estimation remain valid.

3.3.2. Independent Variable

The principal explanatory factor in our analysis is a treatment–period interaction dummy. Its estimated coefficient captures the net influence of the GFRIPZs on the carbon emissions of heavily polluting enterprises. We construct this variable as the product of two dummies: a policy indicator, equal to 1 for enterprises situated in pilot zones and 0 otherwise, and a time indicator, equal to 1 from the year of policy activation onward and 0 beforehand.

With State Council approval, the first wave of Green Finance Reform and Innovation Pilot Zones (GFRIPZs) were launched in June 2017 across five provinces and eight specific areas—Zhejiang, Jiangxi, Guangdong, Guizhou and Xinjiang. A second wave added Lanzhou New District in Gansu Province in November 2019. Chongqing joined in August 2022, but because this date falls outside our sample window, Chongqing was not classified as a pilot region in this study. Accordingly, we assign 2017 as the policy-start year for the first-batch regions and 2019 for the second-batch region.

3.3.3. Mechanism Variables

To further explore the underlying mechanisms through which the GFRIPZ policy influences carbon emissions, this study examines whether the observed emission reductions stem from enterprises’ genuine green technological innovation or from strategic behavioral adjustments. Specifically, we constructed two groups of mechanism variables to capture these distinct pathways: one reflecting green innovation activities, and the other indicating shifts in enterprises’ investment behavior. The detailed variable definitions are as follows.

To evaluate whether green innovation serves as a key channel through which the GFRIPZ policy affects enterprise behavior, we draw on the methodology proposed by Ley et al. [71]. Overall, green innovation ($EnvrPat$) is proxied by the logarithm of total green patent applications, combining green utility model and green invention patents. Utility model innovation ($EnvrUtyPat$) is measured as the log of green utility model patents, whereas invention-oriented innovation ($EnvrInvPat$) is captured by the log of green invention patents. Together, these indicators reflect both the scale and quality of enterprises’ green R&D activities.

In terms of strategic investment responses, heavily polluting enterprises may adjust their investment behaviors in reaction to the financing environment shaped by the GFRIPZs. Because inventory can be adjusted quickly within production, heavily polluting enterprises often alter inventory levels in response to changes in financing conditions. The introduction of the GFRIPZs is therefore expected to suppress inventory investment and, in turn, reduce carbon emissions [72]. We gauge productive investment using inventory investment ($Inventory$), calculated as inventory cash value divided by beginning-of-year total assets. Non-productive investment is proxied by financial investment ($FI$), defined as the sum of trading and derivative financial assets plus net investment property, scaled by total assets and expressed as a percentage.

3.3.4. Control Variables

We selected a series of characteristic variables at the financial and governmental levels that may influence carbon emissions from heavily polluting enterprises. They included the firm size ($Size$), its asset-to-liability ratio ($Lev$), return on assets ($ROA$), nature of ownership ($SOE$), size of the controlling shareholder ($Top$), duality of roles ($Dual$), board size ($Board$), age ($Age$), proportion of fixed assets ($FIXED$), intensity of capital ($Capital$), ratio of cash flow ($Cashflow$), and balance of equity ($Balance$).

Table 2. Variable summary and descriptive statistics.

Var	Definition	Mean	SD	Min	Max
CO2	Corporate carbon emissions (104 t)	160.388	980.765	0.271	19,574.780
Event	0/1 dummy variable	0.117	0.322	0	1
EnvrPat	The natural logarithm of the sum of green utility model patents and green invention patents	0.858	0.322	0	5.645
EnvrInvPat	Natural logarithm of green invention patents	0.512	0.858	0	5.209
EnvrUtyPat	Natural logarithm of green utility patents	0.608	0.899	0	5.476
Inventory	Inventory/Total assets at the beginning of the year	10.516	8.101	0.038	70.305
FI	(Trading financial assets + Derivative financial assets + Net investment property) × 100/Total assets	1.152	2.887	0	26.865
Size	Natural logarithm of total assets at the beginning of the year	22.670	1.351	19.952	28.543
Lev	Total liabilities at the end of the year/Total assets at the end of the year	0.470	0.196	0.091	0.852
ROA	Net profit/Average total assets balance	0.033	0.048	−0.113	0.231
SOE	State-controlled enterprises are coded as 1, and others as 0	0.544	0.498	0	1
Top	Number of shares held by the largest shareholder/Total number of shares	0.350	0.157	0.003	0.900
Dual	Chairman and general manager are the same person, coded as 1; otherwise, coded as 0	0.179	0.383	0	1
Board	Natural logarithm of the number of board members	2.168	0.190	1.609	2.708
Age	Natural logarithm of (current year—year of company establishment + 1)	2.958	0.306	1.386	3.829
FIXED	Net fixed assets/Total assets	0.347	0.166	0.003	0.796
Capital	Total assets/Operating revenue	2.148	1.553	0.318	13.075
Cashflow	Net cash flow from operating activities/Total asset	0.059	0.060	−0.099	0.224
Balance	Sum of the shareholding proportions of the second to fifth largest shareholders/Shareholding proportion of the largest shareholder	0.626	0.543	0.014	3.074

presents a summary of the variables and descriptive statistics used in this study. The mean carbon emissions (${CO}_2$) by heavily polluting enterprises was 1.604 million tons, with a standard deviation of 9.808, a minimum value of 0.003 million tons, and a maximum value of 195.748 million tons. This indicates significant variation in carbon emissions among heavily polluting enterprises. The mean value of green technological innovation ($EnvrPat$) by them was 0.858, with a standard deviation of 0.322; the mean value of their inventory ($Inventory$) was 10.516, with a standard deviation of 8.101; and the mean value of their financial investment ($FI$) was 1.152, with a standard deviation of 2.887. This suggests substantial differences among heavily polluting enterprises in terms of the levels of green technological innovation, investment in inventory, and financial investment. The control variables all fell within reasonable ranges of distribution.

4. Empirical Results and Analysis

4.1. Results of Baseline Regression

We used a multi-period DID model to estimate the impact of the GFRIPZ on carbon emissions from heavily polluting enterprises. The results of regression after controlling for individual- and time-fixed effects are presented in

Table 3. Baseline regression results.

	CO2
	(1)	(2)	(3)	(4)	(5)	(6)
Event	−22.099 ** (9.838)	−22.952 ** (9.788)	−23.986 ** (9.960)	−20.998 *** (8.098)	−21.549 *** (8.012)	−22.523 *** (8.183)
Controls	NO	YES	NO	NO	YES	NO
Controls2	NO	NO	YES	NO	NO	YES
Time effect	YES	YES	YES	YES	YES	YES
Firm effect	YES	YES	YES	YES	YES	YES
Constant	162.979 *** (1.153)	−649.767 *** (222.852)	−207.960 * (126.397)	156.851 *** (0.941)	−692.114 *** (225.298)	−232.833 (146.015)
N	4990	4990	4990	3992	3992	3992
Adj.R2	0.961	0.961	0.961	0.965	0.965	0.965

*, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

. Column (1) includes only the core explanatory variable Event. Column (2) reports the baseline regression results, which include the core explanatory variable and the first-order terms of all control variables. Column (3) further includes the second-order terms of all control variables to account for possible non-linear effects of these enterprise characteristics on carbon emissions. The results show that the coefficients of regression of the GFRIPZs on carbon emissions from heavily polluting enterprises were consistently and significantly negative. This indicates that after the implementation of the policy, the reduction in emissions from heavily polluting enterprises in the pilot regions was significantly greater than that of those in non-pilot regions. To further address potential concerns regarding sample period selection bias, we re-estimated the regressions using a restricted sample from 2013 to 2020. As shown in the last three columns of Table 3, the results remain consistent with those obtained from the full sample, with the coefficients on the GFRIPZ policy variable remaining significantly negative across all specifications. This robustness check confirms that the observed emission reductions are not driven by specific time window choices. Therefore, the GFRIPZs reduced CO₂ emissions from heavily polluting enterprises, thus confirming Hypothesis H1.

4.2. Parallel Trend Test

Since the validity of the multi-period DID model hinges on the parallel trend assumption, we employed an event study approach to estimate year-specific policy effects. This allowed us to test whether the treatment and control groups followed similar emission trajectories prior to the policy’s implementation. As shown in Figure 1, the estimated coefficients for the years prior to the policy intervention from year minus four to year minus one are statistically insignificant and centered around zero. The confidence intervals for these coefficients all include zero, indicating no significant differences in carbon emissions between pilot and non-pilot regions before the policy took effect. After the implementation of the policy, the estimated coefficients became significantly negative, and their confidence intervals did not include zero. This suggests that the GFRIPZ policy effectively reduced carbon emissions from heavily polluting enterprises in the treated regions relative to the control group, supporting the presence of a robust policy effect.

In the two years following the policy rollout, carbon emissions from heavily polluting enterprises in pilot regions declined significantly compared to those in non-pilot areas. In contrast, during the final two years of the sample—when the GFRIPZ policy overlapped with the COVID-19 pandemic and broader industrial disruptions—the differences in emissions narrowed, likely due to pandemic-related production constraints. This outcome is considered reasonable under the circumstances and does not violate the parallel trend condition. To strengthen the robustness of this inference, we conducted additional tests to isolate and account for the potential confounding impact of the pandemic.

4.3. Robustness Test

4.3.1. Excluding the Influence of the COVID-19 Pandemic

We considered data from 2012 to 2021 in this study. The period from 2020 to 2021 was marked by the outbreak of the COVID-19 pandemic, which had a severe negative impact on China’s industrial production and subsequently affected corporate carbon emissions. To eliminate interference by the pandemic on the experimental results, we adjusted the sample period to 2012–2019, and the results of the corresponding regression are presented in

Table 4. Regression results after excluding the impact of the COVID-19.

	CO2
	(1)	(2)
Event	−17.897 ** (7.241)	−18.789 ** (7.400)
Controls	YES	NO
Controls2	NO	YES
Time effect	YES	YES
Firm effect	YES	YES
Constant	−604.985 *** (238.737)	−183.319 (130.867)
N	4990	4990
Adj.R2	0.966	0.965

**, and *** indicate statistical significance at the 5%, and 1% levels, respectively.

. The findings indicate that regardless of whether the regression included first-order or second-order terms of the control variables, the coefficients of regression remained significantly negative at the 5% level of significance. This supports the conclusion that the GFRIPZs reduced carbon emissions from heavily polluting enterprises in the pilot zones. Moreover, the absolute values of the coefficients were smaller than those in the baseline regression, suggesting that the COVID-19 pandemic negatively influenced carbon emissions from heavily polluting enterprises. This further validates the results of the parallel trend test, which showed no significant difference in carbon emissions between pilot and non-pilot regions in the third and fourth pre-policy periods.

4.3.2. Placebo Test

We artificially advanced the years of implementation of the GFRIPZs by two and three years, respectively. Specifically, the year 2017 was adjusted to 2015 and 2014, while 2019 was adjusted to 2017 and 2016, resulting in a fabricated core explanatory variable. The period of policy evaluation was maintained from 2012 to 2021, and the regression was conducted once again. As the time of implementation of the policy was not the actual time of its promulgation, the coefficients of regression should theoretically have been insignificant. The results of regression are shown in

Table 5. Placebo test results.

	CO2
	(1) Two Years in Advance	(2) Three Years in Advance
Event	−4.908 (10.380)	−5.758 (11.087)
Controls	YES	YES
Time effect	YES	YES
Firm effect	YES	YES
Constant	−656.833 *** (213.631)	−656.354 *** (214.162)
N	4990	4990
Adj.R2	0.961	0.961

*** indicate statistical significance at the 1% levels, respectively.

. Columns (1) and (2) present the results of regression when advancing the policy by three and four years, respectively. In both cases, the coefficients of regression were insignificant, which to some extent verifies that the results of the baseline regression in this study were not obtained owing to unobserved random factors.

4.3.3. Excluding Interference by Major Environmental Policies

The implementation of the GFRIPZ was initiated in 2017, while the sample period of this study spanned from 2012 to 2021. In this decade, the Air Pollution Prevention and Control Action Plan and the Interim Measures for Environmental Protection Ministry Interviews may have influenced carbon emissions from heavily polluting enterprises. The State Council implemented the Air Pollution Prevention and Control Action Plan in September 2013, and it required “strengthening the comprehensive management of air pollution in industrial enterprises and reducing pollutant emissions.” In May 2014, the Ministry of Environmental Protection issued the Interim Measures for Environmental Protection Ministry Interviews, which stipulated that “exceeding the national total emission control targets for key pollutants or failing to achieve the national targets for improving environmental quality, controlling the intensity of carbon emissions, and using measures to ensure soil safety” would be among the conditions for interviews. To enhance the credibility of the results, we needed to determine whether the core conclusions remained significant after excluding interference owing to these major environmental policies. To this end, we first removed the sample enterprises located in the 57 high-target cities under the Air Pollution Prevention and Control Action Plan and re-ran the regression. The results are shown in

Table 6. Results of excluding the impact of major environmental policies.

	CO2
	(1) Excluding the Air Pollution Prevention and Control Action Plan	(2) Excluding the Interviews by the Ministry of Environmental Protection
Event	−19.027 * (9.728)	−27.533 ** (10.918)
Controls	YES	YES
Time effect	YES	YES
Firm effect	YES	YES
Constant	−656.833 *** (213.631)	−656.354 *** (214.162)
N	4990	4990
Adj.R2	0.922	0.962

*, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

(1). We subsequently excluded the 73 provinces and cities that had been interviewed between 2012 and 2021 and re-ran the regression, with the results presented in Table 6 (2). They show that the coefficients of the core explanatory variables remained significantly negative, indicating that the Air Pollution Prevention and Control Action Plan and the Interim Measures for Environmental Protection Ministry Interviews interfered minimally with the conclusions of this study.

4.3.4. Adjusting the Proportion of Winsorization

Two-sided 5% winsorization was applied to the baseline regression. To eliminate the possibility of coincidental results due to the choice of the proportion of winsorization, we adjusted it to two-sided 1% and conducted the regression again. The results are shown in

Table 7. Results of adjusting the proportion of winsorization.

	CO2
	(1)	(2)
Event	−36.638 *** (11.976)	−37.469 *** (12.049)
Controls	YES	NO
Controls2	NO	YES
Time effect	YES	YES
Firm effect	YES	YES
Constant	−525.546 ** (238.737)	−154.174 (153.361)
N	4990	4990
Adj.R2	0.958	0.958

**, and *** indicate statistical significance at the 5%, and 1% levels, respectively.

. They indicate that after changing the winsorization proportion, the coefficients of regression of the core explanatory variables remained negative at the 1% level of significance, regardless of whether the first- or second-order terms of the control variables were regressed. This is consistent with the original conclusion and suggests that the choice of winsorization proportion did not affect our findings.

4.4. Heterogeneity Analysis

4.4.1. Heterogeneity in Firm Size

Given that firm size often shapes an enterprise’s resource endowment, compliance capacity, and exposure to regulatory scrutiny, it is a key factor likely to influence how enterprises respond to green finance policies. To explore whether the GFRIPZs has had heterogeneous effects on carbon emissions from enterprises of different sizes, we divided the sample into larger and smaller enterprises based on the “Standards for Classification of Small and Medium Enterprises,” and the “Statistical Classification of Large, Medium, Small, and Micro-enterprises.” The results of regression, as shown in

Table 8. Results of heterogeneity analysis.

	Firm Size		Nature of Property Rights		Regional Development Level
	(1)	(2)	(3)	(4)	(5)	(6)
Event	−27.721 ** (12.144)	−5.248 (5.920)	−36.405 * (18.541)	0.738 (5.695)	−47.171 ** (23.276)	−0.619 (28.826)
Controls	YES	YES	YES	YES	YES	YES
Time effect	YES	YES	YES	YES	YES	YES
Firm effect	YES	YES	YES	YES	YES	YES
Constant	−706.596 ** (298.254)	−600.693 *** (200.345)	−970.798 *** (374.389)	−615.110 *** (121.461)	−680.171 * (400.528)	−654.127 *** (182.274)
N	4060	930	2711	2267	2750	2240
Adj.R2	0.961	0.635	0.961	0.811	0.962	0.906

*, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

(1) and (2), indicate that the GFRIPZ had a negative impact on carbon emissions from large-scale heavily polluting enterprises at a 5% level of significance, with the policy’s inhibitory effect being stronger than the average. However, its impact on smaller heavily polluting enterprises was not significant.

This divergence may stem from differences in compliance capacity and resource availability. Larger enterprises generally have more formalized management structures, better access to financial and technological resources, and greater reputational exposure, making them more sensitive to policy signals and reputational risks [73]. Moreover, they are more likely to be under direct scrutiny by regulators and financial institutions, which enhances their responsiveness to green finance constraints. In contrast, small enterprises often lack the technical expertise and financing channels needed for green up-grading, and may also operate under weaker regulatory oversight [74,75]. These structural disadvantages hinder their ability to respond effectively to green policy incentives. Therefore, without targeted capacity-building support, green finance policies may exacerbate environmental performance disparities across firm sizes.

4.4.2. Heterogeneity in Property Rights

Ownership structure constitutes a critical institutional factor that may condition how enterprises respond to environmental policies. State-owned enterprises (SOEs) and non-state-owned enterprises (non-SOEs) differ significantly in terms of their regulatory obligations, access to resources, and alignment with government priorities, all of which may influence their behavior under green finance constraints [76]. To investigate whether the GFRIPZs had heterogeneous effects on carbon emissions from heavily polluting enterprises with different property rights, we divided the samples into state-owned enterprises (SOEs) and non-state-owned enterprises (non-SOEs). The results of regression, shown in Table 8 (3) and (4), reveal that the GFRIPZs significantly inhibited carbon emissions by SOEs, exceeding the average, while their inhibitory effect on non-SOEs was not significant.

SOEs are not only economically significant but also politically accountable to government mandates. Their dual role as market participants and agents of public policy makes them more responsive to state-led green initiatives. They face stronger pressure to fulfill social and environmental responsibilities, and their financing is often tied to policy compliance, reinforcing their alignment with national sustainability goals [77]. Non-SOEs, on the other hand, are primarily driven by profit motives and may prioritize short-term financial viability over long-term environmental performance. Without sufficient regulatory pressure or incentive alignment, these enterprises may view green finance policies as burdensome rather than transformative.

4.4.3. Heterogeneity in Levels of Regional Financial Development

Differences in levels of financial development across regions may influence the reduction in carbon emissions due to the GFRIPZs. We used the ratio of deposits and loans of financial institutions to the GDP of each province as an indicator of regional financial development, and divided the sample regions into “high” and “low” categories for regression analysis. The results, shown in Table 8 (5) and (6), indicate that the GFRIPZs had a significant negative impact on carbon emissions from heavily polluting enterprises in regions with high levels of financial development, such that they exceeded the average. By contrast, regions with lower levels of financial development did not exhibit a significant reduction in emissions.

Heavily polluting enterprises in regions with high financial development were accustomed to convenient financing before the implementation of the policy and relied on external funds for production and operation. When the policy imposed constraints on financing, these enterprises could not obtain sufficient funds, which forced them to halt excessive productive investments and reduce the scale of production, which in turn reduced their carbon emissions. Enterprises in regions with poor financial development were already accustomed to relying on internal funds due to difficulties in obtaining external financing. This led them to become more cautious in their investment-related decisions and to large-scale, unplanned production. As a result, the policy’s impact on them was less severe and their carbon emissions did not decrease significantly.

5. Mechanism Analysis

The key to stimulating the enthusiasm of market entities for green transformation through green finance lies in balancing the relationship between environmental protection and corporate economic benefits. The above results consistently support the notion that the GFRIPZs helped reduce carbon emissions from heavily polluting enterprises. In this section, we examine the mechanisms underlying this phenomenon from the perspectives of “green technological innovation” and “strategic investment,” and explore the strategies adopted by heavily polluting enterprises. The analysis aims to verify whether these enterprises prioritize the green, low-carbon transformation and upgrade or the pursuit of short-term economic benefits.

5.1. Test of Mechanism of Green Technological Innovation

Green technological innovation is a crucial pathway for heavily polluting enterprises to achieve green and low-carbon development. It can enhance the capabilities of green manufacturing by such enterprises through technological advancement and the effects of learning and technological accumulation. This enables energy conservation and a reduction in emissions, improves the efficiency of resource use, and ultimately reduces corporate carbon emissions [78]. To more intuitively demonstrate differences in the levels of green technological innovation between heavily polluting enterprises in pilot and non-pilot regions before and after the implementation of the policy, we provide a graph of the average trend in green technological innovation among heavily polluting enterprises. Figure 2 shows that the average level of green technological innovation in both pilot and non-pilot regions exhibited an upward trend with the implementation of the policy. However, the average green technological innovation in non-pilot regions was significantly higher than that in pilot regions. Notably, the average green technological innovation among heavily polluting enterprises in the pilot regions increased significantly after the implementation of the policy. From the perspective of the trend alone, heavily polluting enterprises in the pilot regions exhibited a tendency toward green technological innovation after the implementation of the policy. Moreover, Figure 2 reveals that both in pilot and non-pilot regions, and both before and after the implementation of the policy, the average level and rate of growth of innovations in the green utility model were significantly higher than those of innovations in green inventions. This indicates that compared with substantive green technological innovation, heavily polluting enterprises had a stronger inclination toward strategic green technological innovation. Consequently, the impact of the GFRIPZs on the quality of green technological innovation remains difficult to assess.

We further examined the impact of the GFRIPZs on green technological innovation among heavily polluting enterprises within the relevant jurisdictions by using the model given in Equation (2), focusing on green technological innovation, innovations in the green utility model, and innovations in green invention.

Table 9. Green technology innovation mechanism test results.

	(1) EnvrPat	(2) EnvrInvPat	(3) EnvrUtyPat
Event	0.060 (0.073)	−0.003 (0.054)	0.069 (0.064)
Controls	YES	YES	YES
Time effect	YES	YES	YES
Firm effect	YES	YES	YES
Constant	−5.839 *** (1.289)	−3.755 *** (1.019)	−4.656 *** (1.101)
N	4990	4990	4990
Adj.R2	0.629	0.627	0.556

*** indicate statistical significance at the 1% levels, respectively.

reports the results of the mechanistic test based on green technological innovation. The coefficients of the core explanatory variables were not significant, indicating that the implementation of the GFRIPZs did not have a significant positive impact on either the overall level or the quality of green technological innovation by heavily polluting enterprises. Therefore, it did not yield a reduction in carbon emissions by promoting green technological innovation among heavily polluting enterprises. This confirms Hypothesis H2b.

In July 2024, the Central Committee of the Communist Party of China and the State Council issued the Opinions on Accelerating the Comprehensive Green Transformation of Economic and Social Development, which proposed requirements for promoting green and low-carbon transformation and upgrade of traditional industries. These include driving green and low-carbon transformation of heavily polluting industries such as steel, non-ferrous metals, and petrochemicals, promoting energy-conserving, low-carbon, and clean production technologies and equipment, advancing process upgrades, continually updating mandatory standards for carbon emissions, and curbing the reckless initiation of low-level projects that consume a large amount of energy and generate high emissions. Green technological innovation and application serve as the technological foundation for achieving green production and consumption and present a critical pathway for realizing green development. Promoting continual improvements in the industrial green total factor productivity through technological innovation is fundamental to the green transformation of China’s industries. The reduction in carbon emissions from heavily polluting enterprises through green technological innovation aligns with the requirements of green transformation. However, the GFRIPZs did not significantly positively impact the level or quality of green technological innovation among heavily polluting enterprises in the pilot regions. Nevertheless, the inclination and potential for green technological innovation by these enterprises still shows some positive signs. The policy requires further adjustments and optimization to stimulate these companies’ intrinsic motivation for innovation, with the aim of ultimately achieving green industrial transformation and sustainable development.

5.2. Mechanism Test of Strategic Investment

To visually illustrate the changes in productive and non-productive investments by heavily polluting enterprises, we used data on their investment in inventory and financial investment to generate Figure 3. The graph shows that the level of investment in inventory by heavily polluting enterprises exhibited an overall downward trend, while their financial investment showed a general upward trend. Before the establishment of the first batch of pilot zones, the investment in inventory by heavily polluting enterprises in pilot zones was higher than that by enterprises in non-pilot zones, while financial investment by the two groups remained largely consistent. After the establishment of the first batch of pilot zones, the heavily polluting enterprises in pilot zones’ investment in their inventory significantly declined below that of companies in non-pilot zones, whereas their financial investment surpassed those of heavily polluting enterprises in non-pilot zones. This suggests that heavily polluting enterprises are generally shifting toward reducing their productive investments and increasing their non-productive investments, and policies for green financial innovation are likely contributing to this phenomenon.

We replaced the variables in the model in Equation (2) with investment by heavily polluting enterprises in inventory and their financial investment to examine the impact of the GFRIPZs on their investment strategies within the pilot regions.

Table 10. Strategic investment mechanism test results.

	(1) Inventory	(2) FI
Event	−0.874 ** (0.412)	0.479 ** (0.224)
Controls	YES	YES
Time effect	YES	YES
Firm effect	YES	YES
Constant	16.294 * (9.593)	−2.187 (5.616)
N	4990	4990
Adj.R2	0.824	0.572

*, ** indicate statistical significance at the 10%, 5% levels, respectively.

reports the results of the mechanistic test based on corporate investment strategies. They show that the GFRIPZs had a negative impact on companies’ investment in their inventory and a positive impact on their financial investment, with a 5% level of significance. This suggests that these enterprises adopted behaviors of “strategic investment” in response to the GFRIPZs, thereby validating Hypothesis H3a. Constraints on financing and targets for reductions in carbon emissions limited the ability of the enterprises to obtain external financing, thus reducing their scale of production. Although this reduced their carbon dioxide emissions in the short term, the current targets for carbon emissions have not yet met the requirements for achieving peak emissions and carbon neutrality. Therefore, relying solely on reducing the scale of production to meet the goals of reducing carbon emissions is not sustainable. In the event of economic downturn, heavily polluting enterprises may face broken capital chains, which will increase their risk of bankruptcy and hinder sustainable industrial development. Moreover, the increase in financial investment by heavily polluting enterprises in the pilot regions is noteworthy. Research has shown that excessive financialization can increase corporate financial risks, hinder innovation, reduce total factor productivity, and risk economic development shifting from the real economy to the virtual economy [79,80,81].

The contrasting responses of heavily polluting enterprises to the GFRIPZ policy, as evidenced by their limited engagement in green technological innovation and notable shifts in financial investment behaviors, reveal differentiated behavioral logics shaped by institutional constraints and firm-specific characteristics. Although green innovation represents a forward-looking and sustainable path to decarbonization, many heavily polluting enterprises find it difficult to pursue this route due to high innovation costs, extended payback periods, and uncertainty surrounding future returns [82]. In the absence of a well-developed green finance system and supportive institutional arrangements, these enterprises often lack the absorptive capacity and risk tolerance required for long-term technological upgrading [83]. Consequently, instead of undertaking substantive innovation, many firms adopt short-term and flexible financial strategies, such as reducing inventories or reallocating capital toward financial assets, in order to demonstrate compliance and relieve regulatory or financing pressures. These adaptive strategies reflect a form of “window dressing” behavior, in which enterprises meet surface-level policy expectations without achieving genuine environmental transformation [84]. This divergence underscores a fundamental institutional dilemma within green finance reform in developing economies. When policy instruments emphasize external constraints without providing adequate incentives and capacity-building mechanisms, they may inadvertently promote symbolic adaptation rather than substantive structural change.

6. Conclusions and Policy Recommendations

6.1. Conclusions

This study draws on panel data of A-share-listed enterprises in Shanghai and Shenzhen from 2012 to 2021, employing a multi-period difference-in-differences (DID) framework to assess the impact of the Green Finance Reform and Innovation Pilot Zone (GFRIPZ) policy on carbon emissions among heavily polluting enterprises. The analysis particularly focuses on identifying whether emission reductions stem from the stimulation of green technological innovation or from shifts in strategic investment behavior. The empirical findings offer novel insights into the transmission mechanisms of green financial policy. First, the GFRIPZ policy demonstrably lowered carbon emissions among targeted polluters within the pilot zones. Second, heterogeneity tests reveal that this effect was more pronounced among large-scale enterprises, state-owned enterprises (SOEs), and those located in regions with advanced financial systems. Conversely, the policy showed limited or no significant impact on smaller enterprises, non-SOEs, or enterprises in financially underdeveloped areas. Third, the evidence suggests that the policy has not effectively incentivized genuine green innovation. Rather than increasing long-term investment in green technologies, enterprises responded by adjusting their capital allocation strategies—reducing investments in productive assets while expanding non-productive expenditures—to meet carbon reduction requirements. This adaptive behavior reflects a superficial compliance with environmental goals, potentially weakening the real economy and undermining the pursuit of long-term sustainability. This study reveals the heterogeneous carbon reduction responses of heavily polluting enterprises under the GFRIPZ policy, enriching the theoretical understanding of enterprise behavior under policy interventions and offering important insights for policymakers to tailor supportive mechanisms and adopt differentiated approaches in the promotion of green finance.

6.2. Policy Recommendations

First, given the demonstrated effectiveness of the GFRIPZs in reducing carbon emissions, the scope of the pilot program should be further expanded. Emphasis should be placed on leveraging market-oriented instruments to unlock the emission-reduction potential of large-scale, state-owned, and financially mature heavily polluting enterprises. Doing so can create positive spillover and demonstration effects that encourage small- and medium-sized enterprises to follow suit. To amplify this effect, policymakers could consider integrating green credit scoring systems that differentiate between enterprises based on their environmental performance, thereby guiding capital flows toward those making substantive progress in decarbonization.

Second, the study finds that the GFRIPZs have not yet successfully incentivized green technological innovation among heavily polluting enterprises. Instead, many enterprises have pursued expedient investment strategies that prioritize short-term financial gains over long-term sustainability. To address this, financial institutions in pilot zones should adopt more targeted credit appraisal mechanisms that assess enterprises’ environmental commitment and technological readiness, rather than relying solely on historical emissions profiles. Furthermore, the introduction of conditional interest rate subsidies for enterprises that demonstrate measurable progress in green innovation could provide stronger incentives for long-term investment. Policymakers might also consider establishing public–private co-investment platforms to support high-risk, high-potential green technology projects, thereby lowering the financial barriers to innovation.

Third, enterprise managers must recognize that relying solely on short-term strategic investments is not a viable path for sustainable development. This behavior may provide temporary relief but risks long-term competitiveness. In the current policy environment, enterprise leadership must adopt a forward-looking perspective, benchmarking against industry leaders in green transformation and actively seeking to upgrade internal innovation capacities. Regulatory bodies may also consider supporting capacity-building initiatives and environmental management training programs to enhance managerial awareness and readiness for green transition.

Fourth, the findings of this study have broader implications for other developing countries undergoing similar green finance reforms. The reliance on financial engineering rather than technological upgrading underscores a key challenge for green policy effectiveness in emerging markets. In regions where financial systems are less developed, enterprises may lack access to the long-term capital needed for genuine transformation. Therefore, when replicating the GFRIPZ framework, it is crucial for other countries to tailor the policy according to local institutional and market conditions. This includes ensuring that green financing instruments are accompanied by regulatory enforcement mechanisms, innovation incentives, and accessible credit channels for small- and medium-sized enterprises.

6.3. Limitations and Future Directions

While this study provides important insights into the effects of the GFRIPZ policy on carbon emissions among heavily polluting enterprises, several limitations should be acknowledged. First, the estimation of firm-level carbon emissions was based on proxy calculations due to the unavailability of direct emission data, which may have introduced measurement errors despite standardization efforts. Second, the analysis focused solely on listed enterprises in China’s A-share market, which may limit the generalizability of the findings to smaller, unlisted, or non-public enterprises. Third, although the study examines the mechanisms of green technological innovation and strategic investment, it does not fully capture other potential behavioral responses by enterprises. These may include adjustments in supply chain management, changes in operational strategies, or evolutions in corporate governance and organizational culture in response to green finance policies.

Future research could address these limitations through several avenues. First, as environmental disclosure regulations continue to improve, the use of firm-level carbon emission data reported directly by enterprises would enhance measurement accuracy and enable more precise policy evaluation. Second, expanding the sample to include non-listed enterprises or constructing matched datasets covering both public and private enterprises could improve the external validity of the findings. Third, further exploration of additional mediating or moderating factors—such as managerial characteristics, environmental disclosure practices, and regional enforcement intensity—may offer deeper insights into the behavioral responses of enterprises. Finally, comparative studies across different developing countries undergoing green financial transitions could help identify the institutional and regulatory conditions under which such policies are most effective, thereby informing the broader applicability of green finance reforms.