Green Credit Guideline Influencing Enterprises’ Green Transformation in China

Abstract

To achieve high-quality corporate development, it is essential to undergo green transformation. Unlike previous literature, this paper explores relevant mechanisms affecting firms’ green transformation from a novel perspective of green credit guidelines (GCG). Using A-share listed industrial enterprises in China from 2010 to 2020 as a sample, we combine the generalized moment estimation model (GMM) with the difference-in-difference model (DID) and demonstrate that (1) GCG significantly promotes enterprises’ green transformation measured by the method of super-efficient Slacks-Based Measure and Data Envelopment Analysis (SBM-DEA). (2) GCG tends to improve green technology innovation, which further facilitates firms’ green transformation, and corporate social responsibility (CSR) reinforces the positive relationship between GCG and firms’ green transformation. Correspondingly, we provide policy recommendations for China and other developing countries.

1. Introduction

Green transformation is an inherent requirement for high-quality enterprise development. In June 2022, the Ministry of Industry and Information Technology, the State-owned Assets Supervision and Administration Commission of the State Council, and other departments jointly issued the Action Plan for Industrial Energy Efficiency Promotion, proposing to promote energy efficiency improvement and green upgrading in key areas. At the enterprise level, green transformation is an important driving force for the improvement of the development level of enterprises. Green transformation produces a dynamic gaming effect in the market between buyers and sellers of enterprises and plays an important role in meeting market demand. Through green transformation, enterprises send positive signals to the market, build their reputation and enhance their competitiveness [1].

However, firms’ green transformation has been slow in China’s context. How to promote the transformation becomes a crucial issue. Existing studies have analyzed the factors influencing the green transformation of enterprises from the perspectives of environmental regulation, internal enterprises, etc. [2,3,4]. Although existing literature paves the way for this study, they neglect three aspects: (1) From the green credit guideline side, existing research mainly focuses on the operational mechanism, implementation status quo, and existing problems of green credit guideline, or is based on the perspective of enterprises, analysing the impact of green credit implementation on the business performance of commercial banks from the micro level, and exploring the impact of green credit implementation on enterprise innovation, investment, and efficiency. On the other hand, most of the studies on green transformation of enterprises are carried out from the perspective of environmental regulation. No such work has been performed to promote firms’ green transformation from a novel perspective of green credit policy. There is an urgent need to investigate GCG’s influence on firms’ green transformation at a micro-firm level. (2) More importantly, most of the existing literature uses green technology patents to represent firms’ green transformation, which lacks systematic exploration and requires the construction of comprehensive indicators to measure corporate green transformation. (3) In the economic field, it is crucial to identify cause–effect by solving endogeneity, often neglected in previous studies, and further investigate the influencing mechanism.

The fact is that on the firms’ side, they do not treat green projects as an advantage in competition, instead, as a burden due to positive externality. Meanwhile, most green projects, such as ecological remediation and environmental governance, are costly. As a result, firms lack motivation to engage in green transformation. On the other hand, credit as a basic product of China’s financial system has been used to pursue profit-maximization. There is a new category of “green transition” projects that need lending officers and lending criteria to enable them to be fairly assessed, which is challenging for the credit department. A candidate explanation is as follows: Wicksell’s natural interest rate means that when the supply and demand of capital are equal, that is, when saving and investment are equal, the interest rate is the natural rate of interest, which is equivalent to the expected rate of return on capital. The real market interest rate is the currency interest rate. The equalization of two rates is an important condition for achieving economic equilibrium. Differently, Keynes replaced Wicksell’s natural rate of interest with marginal efficiency of capital, which is a bank rate that makes the sum of the present values of the expected returns within the service life of capital goods equal to the supply price of the capital goods or replacement capital, but the uncertainty of expectations made the investment behavior more complex. In the case of green credit, green projects with an uncertainly expected rate of return may not offset the cost of the green credit (the rate of interest). As a result, firms with green projects face credit constraints, particularly small and medium-sized enterprises, although this paper focuses on relative credit allocation (rather than total credit volume) between green and non-green projects. Another possible reason is that there exists credit rationing induced by adverse selection, moral hazard, and even information asymmetry [5]. That is, the credit department is unwilling to lend credit out to firms with green projects and low profit by taking non-interest rate loan measures to make some green fund demanders withdraw from the credit market to achieve balance. It is estimated that China’s “dual carbon” strategy requires a capital investment of 150–300 trillion yuan before the end of 2060, but the government’s financial support ratio is only 10%. Thus, It is hard for firms having green projects to obtain financial sources.

The Porter hypothesis (PH) states that well-designed environmental regulation stimulates a firm’s innovations, which increase its profits that sufficiently offsets costs induced by environmental restrictions. Similar logic is that green credit policy is inclined to allocate green credit to a firm’s green innovations, provoking a firm to obtain profits that pay burdens borne by green projects.

All of these factors trigger the emergence of green credit. Through a series of reviewing, it is found that China’s green credit policy and carbon credit policy have been enacted roughly at the following times (see Figure 1). Among them, the “Green Credit Guideline” (GCG) issued by Banking Supervision in 2012 institutionalised the binding of China’s green financial policies with corporate environmental performance. Traditional credit usually conducts credit analysis on borrowers based on the principles of character, capacity, capital, collateral, condition, and continuity. Once a lending enterprise encounters an environmental accident, commercial banks will bear the dual losses of funds and reputation. Unlike this, the Guideline focuses on the whole process, such as regulating the overall principles, organizational management, policy systems and capacity building, process management, internal control management and information disclosure, supervision and inspection, and ultimately effectively preventing environmental risks. Therefore, this paper takes the Green Credit Guidelines as the entry point and adopts the DID method to assess the micro-implementation effect of GCG. And the Key Evaluation Indicators for Green Credit Implementation issued by the CBRC in 2014 is a further refinement of the Guidelines, which provides the basis and foundation for subsequent green bank ratings. These motivate us to integrate GCG into firms’ green transformation, and there is an urgent need to facilitate firms’ green transformation from a novel perspective of GCG.

A fundamental question is how we could accurately measure firms’ green transformation. Traditionally, data envelopment analysis (DEA) has certain advantages in analyzing the technical efficiency of multiple inputs and outputs and has a very wide range of applications involving many fields, such as education and agriculture [6,7], and also has a very wide range of applications in the economic field. Therefore, this paper adopts a slacks-based model and data envelopment analysis (SBM-DEA) to measure the green transition evaluation index system of enterprises. In addition, there exists a phenomenon in which firms’ green transformation in urban areas matches green finance development in China. The second question is as follows: can GCG really drive firms’ green transformation? We do see some applications of DID or GMM to green innovation individually, but it is rare to combine three models (DID, GMM, and PSM) to overwhelm the endogeneity problem in the case of green transformation at the micro-firm level of China’s context. Card won Nobel Economics Award in 2021 with the application of DID that can obtain the net utility of policy implementation by two differences. Hansen won Nobel Economics Award in 2013 with the application of GMM that is applicable to the presence of reverse causality. We apply three methods (DID, GMM, and Propensity score matching (PSM)) to solve the endogeneity problem in order to answer why green credit guideline affects a firm’s green transformation. Furthermore, we perform analyses of mediating effects of green innovation and moderating effects of corporate-social-responsibility, which explains how green credit guideline influences a firm’s green transformation in China, while they have been neglected in the literature.

The objective of this paper is to explore the promotion of green transformation of enterprises from a new perspective of green credit guidelines. Using the data of listed enterprises from 2010 to 2020, DEA is used to calculate the green transformation index of enterprises, construct cross terms that can reflect green credit policies, and measure them using a double difference (DID) model. Preliminary findings show that green credit policies have a facilitating effect on the green transformation of enterprises.

This study contributes to the literature in the following ways: theoretically, we apply endogenous growth mechanisms to construct a framework and examine facilitating factors on firms’ green transformation that has been weak in China from a novel viewpoint of green credit policy, which enriches theoretical analysis and has academic value. Practically, by taking advantage of micro-firm level data, this study adopts a super-efficiency SBM-DEA model to construct an index to measure a firm’s green transformation and applies three methods (DID, GMM, and PSM) to solve the endogeneity problem in order to answer why green credit guideline affects firm’s green transformation, which is largely ignored in the literature. More importantly, this paper performs heterogeneity analysis and explores the mediating effects of green innovation and the moderating effect of corporate-social-responsibility, which explains how green credit guideline influences a firm’s green transformation in China. Our findings provide a scientific basis for the Chinese government to formulate national strategies and enterprises to actively participate in the green economy, which is of innovative significance.

This paper is arranged as follows: In Section 2, we perform a literature review. In Section 3, we present the research hypothesis, followed by the study design and data description in Section 4. In Section 5, we discuss empirical analysis. Section 6 was performed by testing the mechanism of green credit guidelines (GCG) influencing firms’ green transformation. In Section 7, we present a heterogeneity analysis of GCG affecting firms’ green transformation. Lastly, in Section 8, we make conclusions and recommend measures for policy-decision makers.

2. Literature Review

2.1. Impact Study of Green Credit Policy

The economic effects of green credit policy are mainly involved in the overall quality of China’s ecological environment through improving firms’ innovation and upgrading industrial structure [8]. Cui et al. [9] suggest that green credit policy mainly affects corporate total factor productivity by promoting technological innovation and enhancing resource allocation efficiency. Chen et al. [10] reveal that green credit policy can effectively promote the rationalisation of industrial structure.

2.2. Research on Factors Affecting Firms’ Green Transformation

As for measuring the transformation, Boussemart et al. [11] reveal that a Malmquist productivity index overestimates productivity changes, while the Luenberger productivity indices, that can simultaneously contract inputs and expand outputs but that can also measure in either input or output orientations. Sun and Zhang [12] apply the entropy weight model to construct a comprehensive index for green transformation from five dimensions: technology, production, emission reduction, environmental protection, and social evaluation at the firm level. Liu and Chen [13] combine the principles of hierarchical analysis and the entropy method to calculate the green transformation of the marine industry from three dimensions: industry development, industrial ecology, and green potential at 11 China’s coastal province level. Wu et al. [14] use the super-SBM and entropy weight models to assess the green transformation index at 78 prefecture-level cities in western China. Unlike the regional or city level, to effectively assess the green transformation, it is required to construct an evaluation framework system based on firms’ unique characteristics at the micro-firm level.

Regarding factors affecting the transformation, Yu et al. [15] reveal that policy change from environmental protection fees to taxes accelerates the green transformation of heavily polluting enterprises. Liu and Chen [13] also demonstrate that environmental regulation has a significant role in promoting the green transformation of the marine industry.

In summary, the existing literature conducts an analysis of the operation mechanism, implementation status, and problems of green credit policy (GCP). It analyzes its influence on commercial banks’ business performance and on firms’ innovation, investment, and efficiency. On the other hand, the literature investigates the factors influencing firms’ green transformation from the perspective of environmental regulation while ignoring the role of green finance, especially GCP. More importantly, most of the existing literature lacks systematic exploration, which requires the construction of comprehensive indicators to measure firms’ green transformation at the micro-firm level.

3. Research Hypothesis

3.1. Green Credit Policy and Firms’ Green Transformation

Generally, green projects have a positive externality, which refers to the activities of individual economic behavior that benefit others or society without cost, resulting in private profit less than social benefits. Firms lack the motivation to provide green projects and face financial constraints. By combining government-directed subsidy with a credit market mechanism, Green Credit Guideline (GCG) is an economically incentivized environmental control policy. GCG supports green innovation with low-interest rates, forces enterprises with backward overcapacity to bolster internal energy conservation and emission reduction measures, and gradually changes industrial structures, thereby promoting firms’ green transformation.

Empirically, Zhang and Lu [16] demonstrate that green credit policy significantly contributes to the innovation performance of restricted enterprises. For a more detailed discussion, see the Section 2. Therefore, the first hypothesis can be proposed:

Hypothesis 1 (H1).

Greencredit guideline (GCG) tends to facilitate firms’ green transformation.

3.2. Mediating Influence of Corporate Green Technology Innovation

Porter’s hypothesis suggests that a win–win situation can be formed between environmental protection and economic benefits, where appropriate and effective environmental regulations can motivate enterprises to innovate and improve productivity, thereby offsetting the additional costs brought by environmental regulations and even bringing more profit opportunities to enterprises. Fundamentally, firms’ green transformation relies on green technology innovation, which requires financial support. Green technology innovation is chosen as a mediating variable in this regard. Green credit cuideline (GCG) provides direct financial support for enterprises’ green technological innovation, which enables enterprises to introduce new technologies and optimize processes on the input side, while on the output side, through the development of new products and processes, it increases market share and profits and facilitates the green transformation of enterprises.

Empirically, Fan and Zhou [17] demonstrate that a higher financing efficiency can promote green innovation. Pan et al. [18] reveal that the positive impact of platform economy on enterprise value is mediated by firms’ technological innovation ability. Pan et al. [19] also demonstrate that all regional innovation systems in the maturity stage belong to the eastern developed regions in China’s case. Thus, the second hypothesis can be proposed:

Hypothesis 2 (H2).

Green credit guideline (GCG) tends to facilitate firms’ green transformation through channels of green technology innovation.

3.3. The Moderating Effect of Corporate Social Responsibility (CSR)

In order to strengthen the positive influence of green credit policy on firms’ green transformation, corporate social responsibility (CSR) is selected as a moderating variable based on the following concern: (1) On one hand, enterprises need to send positive signals to their stakeholders by fulfilling their social responsibilities and demonstrate to all parties in society their determination and actions for green transformation, so as to obtain more credit support. (2) On the other hand, firms with strong CSR are willing to engage in green activities such as green innovation, thereby realizing green transformation and sustainable development.

Empirically, Li et al. [20] reveal that social responsibility in firms tends to be increased by China’s green credit guideline (GCG). On the other hand, Le [21] demonstrates that corporate social responsibility benefits a firm’s sustainable performance. Thus, we proposed the third hypothesis:

Hypothesis 3 (H3).

Corporate social responsibility reinforces the positive influence of the green credit guideline (GCG) on firms’ green transformation.

4. Study Design and Data Description

4.1. The Theoretical Framework

While much attention has been paid to the economic effects of green credit Guideline (GCG), less attention has been paid to GCG’s influencing on firms’ green transformation. Theoretically, the Solow model is viewed as a cornerstone of growth theory in modern macroeconomics; however, it treats technological progress as an exogenous variable [22]. Following the framework of the two-department endogenous growth model from Lucas-Uzawa [23], this study describes a simple theoretical framework for GCG to facilitate firms’ green transformation. In the model, credit or capital can be divided into brown (low-quality) and green (high-quality) capital. Correspondingly, employees are divided into general workers and green innovators. All of them are included in the production function. The simple logic of the endogenous growth mechanism is that in the short term, brow or polluting capital is allocated to general skilled workers, who promote economic growth through brown consumption channels; the initial marginal revenue of polluting capital is greater than that of green capital [24]. However, the marginal pollution control cost increases through environmental channels, which also leads to damage to consumer welfare and the reduction of consumption, making the marginal revenue of polluting capital decrease. If a portion of polluting capital revenue is used to increase research and development investment, learn from practice, and generate knowledge spillovers, it will accelerate the accumulation of green capital. Accordingly, the wage increase of green innovators will accelerate the transformation of general skilled workers to highly skilled workers, leading to an increase in the marginal revenue of green capital. Alternative elucidation is that technological progress relies on relatively expected values of brown and green capitals [25]. Combining the market clearing conditions such as business investment demand for brown and green credits equals the financial supply of brown and green credits; we reveal that on an endogenous growth path, there exists a dynamic equilibrium: marginal return of brown or unsustainable technology equals marginal return of green or sustainable technology. This implies that economic growth quality and improving sustainability seem to be realized simultaneously and endogenously. Policy shocks such as the green credit policy initiated in 2012 may change the economic structure and accelerate firms’ green transformation.

4.2. Enterprise Green Transformation Index Measurement

The Malmquist index concept was originally proposed by Malmquist in 1953, and its essence was the ratio of two distance function values, which was a theoretical concept [26]. In 1978, Charnes, Cooper, and Rhodes proposed the first DEA model to measure technical efficiency through a linear programming approach and then transform it into a distance function value, which was called the CCR model [27]. Eventually, the Malmquist index was transformed from a theoretical concept to an empirical index. Chung and Fare [28] derived the M index with undesired outputs like pollutants and named it the Malmquist–Luenberger (ML) index. Since then, it has been customary to refer to the Malmquist index with non-expected outputs as the ML index. Because the traditional CCR-DEA model is based on input or output orientation, assuming that inputs or outputs move at the same rate leads to input redundancies or output deficiencies and thus affects the accuracy and credibility of efficiency values [29]. In 2001, Tone proposed a new DEA model, the Slacks-Based Model (SBM), by fully taking into account the non-proportional changes of inputs or outputs, and thus better reflecting the efficiency level in the actual production process [30]. Tone further extended the SBM model in 2003 by integrating non-desired outputs into efficiency analysis in a more comprehensive way [31] (

Table 1. Evolution of the SBM-DEA and ML (Malmquist–Luenberger) indices.

Year	Author (s)	Index and Method
1953	Malmquist [26]	First proposed concept of Malmquist (M) Index
1978	Charnes, Cooper, and Rhodes (CCR) [27]	First proposed the DEA model and called CCR model to calculate the M index
1997	Chung and Fare [28]	Further derived the M index with undesired outputs like pollutants and called Malmquist–Luenberger (ML) index
2001	Tone [30]	Proposed a new DEA model by relaxing nonproportional changes of inputs or outputs and called SBM model
2003	Tone [31]	Further extends the SBM model by integrating non-desired outputs into efficiency analysis

).

Currently, the productivity index calculated by DEA models is the Malmquist index and the Luenberger (ML) index, which is an essential Malmquist index by taking into account non-desired outputs. As an exchange, academics treat the productivity Index as a proxy for total factor productivity (TFP) or green TFP. Considering the intrinsic mechanism of firms’ green transformation, this paper adopts the super-efficiency SBM-DEA model based on non-expected outputs and the ML index. The enterprise green transformation index system is shown in

Table 2. Enterprise green transformation index system.

Comprehensive Index	Specific Indicators	Specific Indicators	Unit of Measurement	Property Direction
Corporate Green Transformation (GT)		Net fixed assets	Billion	+
	Inputs	Number of employees	ten thousand people	+
		Environmental Investment	Billion	+
	Desired Outputs	Revenue from main business	Billion	+
	Non-desired outputs	Enterprise sewage charges and environmental protection tax	Billion	−

Note: In the column of “Property Direction”, “+” indicates a positive indicator, and “−” indicates a negative indicator.

. Following Cui and Lin [32], we assume that the Malmquist–Luenberger (ML) index in 2010 is 1, and then multiply it with the ML index of each period to finally obtain green transformation for the period 2010–2020.

Assume that each firm acts as an independent decision unit (DMU, i = 1,2, …,n), which contains three input–output vectors, i.e., inputs p, desired outputs $q_1$, non-desired outputs $q_2$, and sets vectors X = [$x_1,x_2,\dots ,x_n$]$\in R^{pn}$; Y = [$y_1,y_2,\dots ,y_n$]$\in R^{q_{1}n}$; Z = [$z_1,z_2,\dots ,z_n$]$\in R^{q_{2}n}$, and X > 0, Y > 0, Z > 0.

The linear programming form of the slacks-based measure (SBM) model is as follows:

$$\rho =\min \frac{1-\frac{1}{p}\sum _{i=1}^p\frac{s_i^{\_}}{x_{ik}}}{1+\frac{1}{q_1+q_2}(\sum _{j=1}^{q_1}\frac{s_j^{+}}{y_{jk}}+\sum _{r=1}^{q_2}\frac{s_r^z}{z_{rk}})}$$

(1)

$$\mathrm{s}.\mathrm{t}.\left\{\begin{array}{c}{x}_{ik}\ge \sum _{g=1,g\ne k}^n{\lambda }_g{x}_{ig}-s_i^{-},i=\mathrm{1,2},\dots p\\ y_{jk}\le \sum _{g=1,g\ne k}^n{\lambda }_g{y}_{jg}+s_j^{+},j=\mathrm{1,2},\dots q_1\\ z_{ik}\ge \sum _{g=1,g\ne k}^n{\lambda }_g{z}_{rk}+s_r^z,r=\mathrm{1,2},\dots q_2\\ \lambda \ge 0, s^{-}\ge 0, s^{+}\ge 0, s^z\ge 0\end{array}\right.$$

(2)

$s^{-}$, $s^{+}$ and $s^z$ are slack in input factors, desired output, and non-desired output in turn, and the efficiency values can be calculated according to the above formula, and to measure the value of firms’ green transformation, me the ML index is calculated based on Chung and Fare [28]; the following is the formula for the index between period t and period t + 1:

$${ML}^{t+1}={\left\{\frac{\left[1+\overrightarrow{D^t}\left(x^t,y^t,z^t;y^t,{-b}^t,\right)\right]}{\left[1+\overrightarrow{D^t}\left(x^{t+1},y^{t+1},z^{t+1};y^{t+1},{-b}^{t+1},\right)\right]}\times \frac{\left[1+\overrightarrow{D^{t+1}}\left(x^t,y^t,z^t;y^t,{-b}^t,\right)\right]}{\left[1+\overrightarrow{D^{t+1}}\left(x^{t+1},y^{t+1},z^{t+1};y^{t+1},{-b}^{t+1},\right)\right]}\right\}}^{\frac{1}{2}}$$

(3)

4.3. Sample and Data

A sample of Chinese A-share listed enterprises from 2010 to 2020 is manually screened: (1) According to the National Bureau of Statistics mentioned in the FAQ on the statistical system and classification standards, industrial sectors include electricity, heat, gas and water production and supply, mining, and manufacturing. Guidelines on Industry Classification of Listed Companies (revised in 2012) stay in the industrial category in the sample of enterprises. (2) Deleting the ST and *ST samples with abnormal financial status. (3) Excluding the sample with the year of listing after 2010. (4) Remove samples with omitted or negative values of important operational and financial indicators. The data of other enterprise characteristics are obtained from the database of Guotaian (CSMAR). For data that cannot be matched or have contradictions, manual corrections are made by collecting the accurate information disclosed in the annual reports of listed companies and Juchao Consulting.com. The final sample was selected from 504 enterprises with 3491 sample observations. Among them, 316 are in the experimental group (heavily-polluting industries firms), and 188 are in the control group (non-heavy-polluting industrial firms). The environmental protection investment, emission fee, and environmental protection tax of listed companies were obtained from CSR reports, other data at the enterprise level were collected from the Guotaian database, and the marketability index was used from the China Marketability Index compiled by Fan Gang et al. In case of data mismatch or conflict, manual corrections are made by collecting accurate information published in the annual reports of publicly traded companies and information from the Juchao Information Website.

The specific inputs are described as follows: (1) Labor input can be measured by the year-end headcount of employees in a company, while capital input can be calculated using the enterprise’s net fixed assets, which is the residual amount of the fixed assets’ original value after subtracting accumulated depreciation and fixed assets impairment allowance. (2) Referring to Cui and Jiang [33], we use environmental investment as an intermediate input and select two items of construction in progress and overhead for screening key words related to environmental investment, including wastewater, waste residue, waste gas, remediation, and recycling treatment. Output indicators are (1) Expected output, using the enterprise’s main business income to measure, can reflect the core competitiveness of enterprises and better reflect the level of efficiency. (2) For the non-expected output indicators, refer to Sun and Fei [34]; corporate emission fees and environmental protection taxes are used as the non-expected output of enterprises.

4.4. Variable Construction and Interpretation

The explained variable is the green transition of firms. The main explanatory variable of interest is the policy on green credit, and the control variables include firm profitability level, financial leverage, Tobin’s Q, and firm age.

• Explained variable: corporate green transformation (GT) measured by multiple indicators with super-efficient Slacks-Based Measure and Data Envelopment Analysis (SBM-DEA) based on non-expected output and ML index.
• Explanatory variable: green credit guideline (GCG) indicated by experimental dummy variable (ifhp). This paper mainly conducts quasi-natural experiments based on GCG shock, so the experimental group comprises highly polluted firms, i.e., ifhp = 1; non-polluting industrial enterprises are utilized as the control group, i.e., ifhp = 0. A time dummy variable (post): The exogenous shock policy of this paper, the green credit guidelines, was officially released on 29 January 2012. We select 2012 as the policy implementation node; before 2012, the post takes 0; after 2012, the post takes 1.
• Control variables: Profitability (profit), the return on assets of a company = operating profit/total assets. Companies with high profitability levels can have enough capital for equipment renovation and technology upgrades in the short term to improve production efficiency and competitiveness. However, for polluting enterprises, high profitability levels often come at the expense of environmental performance, leading to environmental pollution and waste of resources. Therefore, the influence of profitability levels on the transition behavior of companies, in the long run, is uncertain and may facilitate or hinder enterprises from achieving green development.

Financial leverage (lev), a firm’s gearing ratio = total liabilities/total assets. A high gearing ratio indicates that enterprises have strong financing ability and can use financial leverage to raise sufficient funds for transformation, but it also exposes a larger capital risk, which may affect the long-term running and growth of firms, so the impact of this indicator on the green transformation of firms is as uncertain as the earnings level.

Tobin Q (TobinQ), corporate social wealth creativity is the ability of a firm to create wealth for society in its production and operation activities. Drawing on Zhang [35], TobinQ is used to measure corporate social wealth creativity, which is the ratio of a firm’s market value to its capital replacement cost and represents the wealth created for society per unit of resource wealth. Generally, the stronger the social wealth creativity but, the more innovative the enterprise is, which is conducive to the enterprise’s green transformation. Therefore, we expect the sign to be positive.

Firm age (lnage), expressed as the logarithm of the year in which the firm was established, Margaritis and Psillaki [36] argue that the level of development and capital accumulation of a firm is often related to its time of establishment and that in general, firms with a longer history have more efficient management, which is conducive to improving the TFP of the firm. However, Yu et al. [37] argue that resource-based firms place different emphasis on green innovation at different phases of development, preferring to invest in green technology study and growth in the early stages while focusing more on environmental protection in the later stages so that the maturity of the firm may negatively affect green technology innovation. Therefore, we have no clear expectation on the sign.

Descriptive statistics for all variables are demonstrated in

Table 3. Descriptive statistics.

Symbols	Variables	Observations	Average Value	Standard Deviation	Minimum Value	Maximum Value
$GT$	Corporate Green Total Factor Productivity	3491	1.138	0.781	0.158	2.749
$infp$	Experimental dummy variables	3491	0.671	0.469	0	1
$post$	Time dummy variable	3491	0.812	0.390	0	1
$infp\times post$	Green Credit guideline	3491	0.541	0.498	0	1
$Profit$	Profit Level	3491	0.027	0.155	−7.242	0.775
$Lev$	Financial leverage	3491	0.476	0.237	0.014	4.113

. There is a remarkable difference between the maximum and minimum values of GT of firms, a measure of enterprise green transformation, which implies that the green transformation varies widely between individual firms. The mean value of the dummy variable for the experimental group (Treat) is 0.671, which means that the sample observations of heavy-polluting enterprises account for about 67% of all sample observations. The mean value of the time dummy variable (Time) is 0.812, which suggests that the sample data represents 81.2% of the total sample after GCG publication.

5. Empirical Analysis

5.1. Model Setting and Variable Definition

The empirical research on policy evaluation in the literature generally adopts DID model, which can not only avoid the endogeneity problem of the policy as an explanatory variable but also control the influence between the dependent variable and the independent variable, and make the unobservable time-varying confounding factors excluded by comparing the difference in the impact of the policy changes before and after the policy on the policy pilot area (experimental group) and the non-pilot area (control group), so that the dispositive effect of the policy can be separated from the confounding factors to better analyse and assess the impact of the policy.

To include the lagged one-period value of firms’ green transformation as an explanatory variable in the model so that the model has a better dynamic explanatory power. However, there exists a two-way causal issue between the one-period lag and non-lag of corporate green transformation, which results in an endogeneity problem. Estimation by traditional econometric methods such as ordinary least squares, instrumental variables, and maximum likelihood methods can lead to biased and non-consistent parameter estimates, which can distort the economic meaning of the regression results. Systematic GMM, on the other hand, does not require knowledge of the exact distribution of the random error term, allowing it to be exclusively heteroskedastic and serially correlated, and thus can yield more valid parameter estimates than these traditional estimation methods. The endogeneity problem arising from dynamic panels can be effectively addressed. In our paper, we take advantage of GMM-DID and combine them together to solve endogeneous problems as a benchmark analysis.

$$GT={\beta }_0+{\beta }_{1}ifhp+{\beta }_{2}post+{\beta }_{3}ifhp\times post+{\beta }_{4}profit+{\beta }_{5}lev+{\beta }_{6}TobinQ+{\beta }_{7}Lnage+{\beta }_8{GTFP}_{t-1}+\epsilon$$

(4)

where $GT$ represents firms’ green transformation, and $ifhp$ is the experimental group idiot variable, taking the value of 1 for firms in the heavy pollution industry and 0 for firms in the non-pollution industry as the control group; $post$ is the policy implementation time dummy variable, taking the value of 0 before 2012 and 1 in 2012 and after; ${\beta }_3$ denotes the coefficient of the influence of green credit guidelines on the green transformation of industrial enterprises, and the core variable green credit guideline is denoted by $ifhp\times post$. profit, lev, TobinQ, and Lnage are control variables, representing the profitability level, financial leverage, TobinQ value, and age of enterprises, respectively, and ε is a random disturbance term.

5.2. Parallel Trend Test

This test has to be performed before using a difference-in-difference (DID) model. Scholars usually use two methods to test: 1. Observe whether the trend in the experimental group and the control group prior to the policy shock is consistent, and if it is consistent, the test is passed; 2. The regressions include the interaction term between the dummy and the policy variable at each time point, and if the interaction term coefficient is insignificant before the policy node, the parallel trend test passes. In this paper, method 2 is used to test the parallel trend hypothesis. As Figure 2 shows, The dotted line in the figure indicates the year in which the policy was implemented. The estimated coefficients of interaction terms of dummy and policy variables before the release of GCG are insignificant, which implies that there is no significant change difference between experimental and control groups before the implementation of GCG, which is consistent with the parallel trend test. After the release of GCG, the coefficients of the dummy variables in all periods are greater than zero, which reflects some extent that implementing green credit guidelines has a positive influence on the green transformation of firms.

5.3. Basic Regression Analysis

In the basic regression model, there is a two-way causal issue between the one-period lag of corporate green transformation and corporate green transformation, and the resulting endogeneity problem is inevitable. Estimation by traditional econometric estimation methods such as ordinary least squares, instrumental variables, and maximum likelihood methods can lead to biased and non-consistent parameter estimates, which can distort the economic meaning of the regression results. These methods can only yield reliable estimates if their parameter estimates satisfy certain assumptions. Systematic GMM, on the other hand, does not require knowledge of the exact distribution of the random error term, allowing it to be exclusively heteroskedastic and serially correlated, and thus can yield more valid parameter estimates than other estimation methods. The endogeneity problem arising from dynamic panels can be effectively addressed.

Findings of GCG influencing the green transformation of firms are shown in

Table 4. GMM-DID benchmark regression results.

Variables	Expected Symbols	GMM-DID (1)	GMM-DID (2)	GMM-DID (3)	GMM-DID (4)	GMM-DID (5)
$infp\times post$	+	0.7441 **	0.8593 ***	0.8687 ***	0.8678 ***	0.6949 **
		(0.3694)	(0.3224)	(0.3202)	(0.3342)	(0.2959)
$ifhp$		−0.6746 **	−0.7754 ***	−0.7825 ***	−0.7722 **	−0.6270 **
		(0.3414)	(0.3008)	(0.2957)	(0.3054)	(0.2691)
$post$		−0.6510 **	−0.7188 ***	−0.7551 ***	−0.7416 ***	−0.5876 ***
		(0.2696)	(0.2360)	(0.2323)	(0.2501)	(0.2200)
$profit$	$+/-$		−0.0566 ***	−0.0932 ***	−0.2112	−0.1616
			(0.0104)	(0.0117)	(0.1552)	(0.1491)
$lev$	$+/-$			−0.1334 ***	−0.1719 ***	−0.1416 ***
				(0.0431)	(0.0560)	(0.0547)
$TobinQ$	+				0.0214 **	0.0208 **
					(0.0094)	(0.0081)
$lnage$	$+/-$					−0.0283
						(0.0332)
${GTFP}_{t-1}$		0.9451 ***	0.9261 ***	0.9212 ***	0.8864 ***	0.8590 ***
		(0.0348)	(0.0285)	(0.0292)	(0.0311)	(0.0280)
$AR\left(1\right)>z$		0.000	0.000	0.000	0.000	0.000
$AR\left(2\right)>z$		0.123	0.133	0.138	0.125	0.136
$Hansen Test$ $Prob>chi2$		0.109	0.128	0.150	0.112	0.135
$Constant$		0.6665 **	0.7472 ***	0.8486 ***	0.8526 ***	0.8088 ***
		(0.2594)	(0.2268)	(0.2324)	(0.2354)	(0.2039)
Observations		2977	2977	2977	2802	2802
Number of id		504	504	504	489	489

Note: **, and *** mean that the statistics are significant at the 5%, and 1% levels, respectively, and the values in parentheses are robust standard errors of the regression coefficients.

. In the table, column (1) only tests the impact of green credit guidelines on the green transition of firms, and columns (2) to (5) add listed companies’ profitability level, listed companies’ financial leverage, listed companies’ Tobin value, and listed companies’ corporate age item by item on the basis of the previous column. According to findings in columns (1) to (5) of Table 3, infp×post are significant at the 5% level with positive coefficients, AR (1) is less than 0.1, AR (2) is greater than 0.1, indicating that the regression results are valid by Arellano–Bond test, and the results of Hansen Test is larger than 0.1, indicating that findings are valid by over-identification constraint test. Our results are fully in line with expectations, suggesting that GCG tends to facilitate the green transformation of firms, and hypothesis 1 is verified. A possible explanation is that green credit guideline provides positive incentives for businesses to improve their productivity, thus indicating that GCG has a Porter effect in China.

5.4. Robustness Test

5.4.1. Test for Alternative Method by Propensity Score Matching

Because of the large differences between firms in terms of profitability, age, and stage of development, it is difficult to align the experimental group with the control group on anything other than policy effects. Therefore, before doing double-differencing, the sample needs to be treated so that the control group sample has as similar characteristics as possible to the experimental group sample, thus reducing bias due to non-randomly shuffled samples. Matching the treatment and control groups according to propensity scores can effectively avoid endogeneity problems due to sample self-selection bias. The model was regressed using a combination of propensity score matching and difference-in-difference (PSM-DID) approach with reference to Yang et al. [38]. The results show that after propensity score matching and after adding control variables, the cross-term still has a positive influence on the firm’s green transformation with 1% significance. Detailed results are displayed in

Table 5. Tests for propensity score matching.

Variables	GMM-DID (6)
$infp\times post$	0.9554 ***
	(0.2954)
$ifhp$	−0.8610 ***
	(0.2693)
$post$	−0.7664 ***
	(0.2179)
$profit$	−0.0933
	(0.1553)
$lev$	−0.1495 **
	(0.0588)
$TobinQ$	0.0268 ***
	(0.0080)
$lnage$	−0.0175
	(0.0344)
${GTFP}_{t-1}$	0.8362 ***
	(0.0336)
$Constant$	0.9533 ***
	(0.2208)
$Observations$	2802
$Number of id$	489

Note: **, and *** mean that the statistics are significant at the 5%, and 1% levels, respectively, and the values in parentheses are robust standard errors of the regression coefficients.

.

5.4.2. Randomly Generated Experimental Groups

In this paper, a random sample of heavily polluting companies was selected for placebo testing. Specifically, 316 firms from 504 firms are randomly selected as the treatment group in this paper, assuming that these 316 firms are the heavily polluting firms and the others are the control group. Ensure that the independent variable infp × post constructed by random sampling has no effect on the total factor productivity of the firm. In this paper, 500 random samples were taken from the data, and the mean of the regression estimates after each sample was calculated, as shown in Figure 3. This paper finds that the mean of all the regression coefficients for infp × post is close to zero, indicating no significant effect. Figure 3 also shows the distribution of the 500 regression coefficients and the corresponding p-values. As can be seen in Figure 3, the distributions of the regression coefficients are all closely around zero, with p-values greater than 0.1, further confirming the conclusions of this paper. At the same time, the estimates significantly deviate from the true coefficient of 0.7441, which in turn implies that the conclusions of this paper are more credible.

5.4.3. Shorten the Time Window of the Sample

The research interval of this paper is from 2010–2020 because no other national green credit policies were implemented during this period, and the interference of other policy factors can be excluded. However, the issue of the new Environmental Protection Law in 2018 may have had an impact on green credit. To eliminate this effect, this paper shortens the sample period to before 2018, retains the data from 2010–2018, and re-estimates the empirical model, and the results are displayed in column (8) of

Table 6. Results of shortened sample window time test.

Variables	GMM-DID (7)	GMM-DID (8)
$infp\times post$	0.6949 **	0.6970 ***
	(0.2959)	(0.2624)
$ifhp$	−0.6270 **	−0.5634 **
	(0.2691)	(0.2289)
$post$	−0.5876 ***	−0.6045 ***
	(0.2200)	(0.1949)
$profit$	−0.1616	−0.2339
	(0.1491)	(0.1955)
$lev$	−0.1416 ***	−0.1039
	(0.0547)	(0.0716)
$TobinQ$	0.0208 **	−0.0036
	(0.0081)	(0.0116)
$lnage$	−0.0283	0.0024
	(0.0332)	(0.0384)
${GTFP}_{t-1}$	0.8590 ***	0.9001 ***
	(0.0280)	(0.0491)
$Constant$	0.8088 ***	0.6895 ***
	(0.2039)	(0.2110)
$Observations$	2802	2008
$Number of id$	489	450

Note: **, and *** mean that statistics are significant at 5%, and 1% levels, respectively. Values in parentheses are robust standard errors.

. Column (7) shows the results of the original baseline regression, which is placed in Table 6 for comparison. The regression results in column (8) show that the coefficient on the interaction term is still significant, implying that GCG contributes to firms’ green transformation, which is coherent with the previous benchmark regression results and confirms the reliability of experimental findings.

6. Testing Mechanism of Green Credit Guideline (GCG) Influencing Firms’ Green Transformation

6.1. Test Green Technology Innovation as an Intermediary Channel

According to mechanism analysis, GCG may affect green transformation by improving green technology innovation, and this section introduces green technology innovation-related variables to verify the intermediary influence of green technology innovation in green credit guidelines and green transformation. Thus, a mediating effect model is constructed by referring to Wen et al. [39] and Fang et al. [40]:

$$GT={\theta }_0++{\theta }_{1}ifhp\times post+\mathsf{\Sigma }{\theta }_{\mathrm{i}}\mathrm{Z}+\mathsf{\epsilon }$$

(5)

$$Pat={\eta }_0++{\eta }_{1}ifhp\times post+\mathsf{\Sigma }{\eta }_{\mathrm{i}}\mathrm{Z}+{\epsilon }_1$$

(6)

$$\mathrm{GT}={\delta }_0++{\delta }_{1}ifhp\times post+{\delta }_{2}Pat+\mathsf{\Sigma }{\delta }_{\mathrm{i}}\mathrm{Z}+{\epsilon }_2$$

(7)

In the above-mentioned model, the $GTFP$ means the degree of green transformation of companies, and $ifhp\times post$ means the green credit guideline. $\mathrm{P}\mathrm{a}\mathrm{t}$ represents the green technological innovation of firms, and drawing on the method of Qi [41] to measure green technological innovation, this research uses the number of green patents to estimate green technological innovation.

The effects of corporate green technology innovation on green credit guidelines and corporate green transformation are presented in

Table 7. Analysis of intermediary effects.

Variables	GMM-DID (9)	GMM-DID (10)	GMM-DID (11)
$infp\times post$	0.6949 **	0.3836 **	0.7869 **
	(0.2959)	(0.1750)	(0.4009)
$Pat$			0.1229 **
			(0.0495)
$ifhp$	−0.6270 **	−0.3505 **	−0.7259 **
	(0.2691)	(0.1586)	(0.3657)
$post$	−0.5876 ***	−0.2653 **	−0.6387 **
	(0.2200)	(0.1240)	(0.2884)
$profit$	−0.1616	0.0091	−0.2252
	(0.1491)	(0.0806)	(0.2827)
$lev$	−0.1416 ***	0.1026 **	−0.2515 ***
	(0.0547)	(0.0406)	(0.0768)
$TobinQ$	0.0208 **	0.0025	0.0182 *
	(0.0081)	(0.0076)	(0.0094)
$lnage$	−0.0283	−0.0557 **	−0.0204
	(0.0332)	(0.0246)	(0.0405)
${GTFP}_{t-1}$	0.8590 ***		0.7349 ***
	(0.0280)		(0.0202)
${Pat}_{t-1}$		0.6610 ***
		(0.0680)
$Constant$	0.8088 ***	0.4236 ***	1.0020 ***
	(0.2039)	(0.1292)	(0.2718)
$Observations$	2802	2802	2802
$Number of id$	489	489	489

Note: *, **, and *** mean that statistics are significant at 10%, 5%, and 1% levels, respectively. Values in parentheses are robust standard errors.

. The regression result of the model (5) shows the ifhp×post coefficient ${\theta }_1$ is 0.6949 and significant at the 5% level, which means that green credit guideline positively promotes the green transformation of firms. The result of the model (6) shows that the ifhp×post coefficient ${\eta }_1$ is 0.3836 and significant at the 5% level, indicating that green credit guidelines can promote green technological innovation of companies to some extent. In model (7), the coefficient ifhp × post of GCG seems to be significantly positive (${\delta }_1$ = 0.7869, p < 0.05), and the coefficient ${\delta }_2$ of corporate green technological innovation is also significant (${\delta }_2$ = 0.1229, p < 0.05), which indicates that there is a partial intermediary effect of corporate green technological innovation on the relationship between GCG and corporate green transformation, and hypothesis 2 holds.

6.2. Analysis of the Regulation Mechanism of CSR

To explore a moderating influence of CSR on the correlation between GCG and firms’ green transformation, we add the interaction term of CSR and green credit guideline to model (4) by checking the significance of the interaction term as follows:

$$GTFP={\gamma }_0+{\gamma }_{1}ifhp\times post+{\gamma }_{2}CSR+{\gamma }_{3}ifhp\times post\times CSR+\mathsf{\Sigma }\gamma \mathrm{X}+{\epsilon }_3$$

(8)

In the above-mentioned model, where GT means the degree of green transformation of firms, and ifhp × post indicates green credit guidelines. CSR is corporate social responsibility, and according to the relevant literature, the CSR score released by Hexun.com is chosen as the indicator of corporate social responsibility in this paper. Ifhp × post × CSR is the interaction term between GCG and corporate social responsibility. If the coefficients of CSR and GCG are significant, this means that the extent to which the explanatory variable (GCG) influences explained variable (firms’ green transformation) is moderated by the moderating variable (CSR). ∑γX stands for control variables. ${\epsilon }_3$ denotes a random disturbance term.

The outcome of model 7 suggests that the interaction term of CSR and green credit guideline ($ifhp\times post\times CSR$) seems to be significant at a 5% level. Since CSR is a positive indicator in this paper, it indicates that the higher the CSR score is, the more obvious the facilitating effect of GCG on firms’ green transformation is, i.e., social responsibility makes this promotion effect enhanced, which perhaps because firms with higher social responsibility scores are more willing to invest all green credit in green technology R&D and increase green output after receiving support from green credit policies. While enterprises with lower social responsibility scores may invest the obtained green credit in other purposes and thus have insufficient green output (

Table 8. Analysis of regulation effects.

Variables	GMM-DID(12)	GMM-DID(13)
$infp\times post\times CSR$		0.3037 **
		(0.1467)
$CSR$		−0.0488
		(0.0719)
$infp\times post$	0.6949 **	0.6805 **
	(0.2959)	(0.2682)
$ifhp$	−0.6270 **	−0.6066 **
	(0.2691)	(0.2481)
$post$	−0.5876 ***	−0.6211 ***
	(0.2200)	(0.2000)
$profit$	−0.1616	−0.0049
	(0.1491)	(0.2250)
$lev$	−0.1416 ***	−0.1348 **
	(0.0547)	(0.0579)
$TobinQ$	0.0208 **	0.0144 *
	(0.0081)	(0.0082)
$lnage$	−0.0283	−0.0236
	(0.0332)	(0.0318)
${GTFP}_{t-1}$	0.8590 ***	0.8819 ***
	(0.0280)	(0.0293)
$Constant$	0.8088 ***	0.9780 ***
	(0.2039)	(0.3780)
$Observations$	2802	2802
$Number of id$	489	489

Note: *, **, and *** mean that statistics are significant at 10%, 5%, and 1% levels, respectively. Values in parentheses are robust standard errors.

).

7. Heterogeneity Analysis of GCG Affecting Firms’ Green Transformation

7.1. Enterprises’ Ownership Heterogeneity Analysis

The fact that state-owned enterprises (SOEs) and non-state-owned enterprises (non-SOEs) may be subject to different levels of preference or discrimination in the credit market results in green credit policies having different incentives or constraining effects on them. For example, Nie and Xia [42] found that SOEs and non-SOEs face different treatment in the credit market and that, ceteris paribus, SOEs have less difficulty in accessing bank loan funds and obtaining more loan amounts.

Based on different ownership, we divide the sample into two categories: state-owned and non-state-owned firms aiming to analyze if there is heterogeneity between them in terms of GCG influencing green transformation.

Findings are displayed in

Table 9. Ownership heterogeneity analysis.

Variables	GMM-DID (14) Non-State-Owned Firms	GMM-DID (15) State-Owned Firms
$infp\times post$	0.5143 **	0.3412
	(0.2551)	(0.4085)
$ifhp$	−0.5090 **	−0.2799
	(0.2360)	(0.3714)
$post$	−0.3310 *	−0.3473
	(0.1827)	(0.3120)
$profit$	0.1030	−0.2357
	(0.1854)	(0.1846)
$lev$	0.0050	−0.1940 ***
	(0.0806)	(0.0649)
$TobinQ$	0.0247 **	0.0094
	(0.0110)	(0.0084)
$lnage$	−0.0492	−0.1295 **
	(0.0303)	(0.0596)
${GTFP}_{t-1}$	0.8725 ***	0.7650 ***
	(0.0239)	(0.0348)
$Constant$	0.5596 ***	1.0090 ***
	(0.1782)	(0.3181)
$Observations$	1211	1591
$Number of id$	220	269

Note: *, **, and *** mean that statistics are significant at 10%, 5%, and 1% levels, respectively. Values in parentheses are robust standard errors.

, revealing that interaction term coefficients for non-state and state enterprises are significantly different in the two models. The estimated interaction term coefficient in column (14) is 0.5143, with significance at a 5% level, while the estimated interaction term coefficient in column (15) seems to be insignificant. This indicates that GCG has a stronger positive influence on green transformation for non-state firms, while it seems to have a weaker positive influence on green transformation for state-owned firms. A candidate’s explanation is that it is more difficult for non-SOEs, on the one hand, to reach a long-term and stable partnership and obtain credit support from banks. Therefore, in order to obtain preferential green credit, non-state-owned enterprises must add their investment in green innovation and R&D and actively seek ways to green transformation. The execution of green credit guidelines will form a stronger financing incentive for non-SOEs, thus effectively promoting green transformation for non-SOEs.

7.2. Regional Heterogeneity Analysis

Fan et al. [43] argue that marketization means government deregulation of the economy and the growth of a more liberal economy. According to Wang and Chen [44], the market environment is a key external factor that influences the green innovation of firms. Liu et al. [45] also pointed out that in areas with high regional marketability, there is less government intervention in enterprises and more fluid and transparent market information, which can effectively decrease the information asymmetry problem, thus increasing the financing flexibility of enterprises and reducing financing constraints, which is beneficial to enterprises’ green transformation. On the contrary, in less market-oriented areas, information is not transparent, market information transmission is obstructed, bank credit funds are allocated inefficiently, and enterprises face greater financing constraints, which is not conducive to enterprises’ green transformation. Therefore, in more market-oriented areas, GCG can more effectively promote firms’ green transformation.

To verify whether there are different influences of GCG on enterprises’ green transformation under different degrees of marketization, we refer to Fan et al. [43] and adopt the marketization indicators to estimate the marketization degree of each region. As the average reflects the general situation and average level of a set of data, areas with a marketability index above the average are defined as the high marketability group, and areas below the average are defined as the low marketability group. The provincial regions are matched with the registered locations of the sample enterprises to determine the degree of marketability of the enterprise locations. Based on columns (16) and (17) of

Table 10. Heterogeneity analysis of the degree of regional marketization.

Variables	GMM-DID (16) Highly Marketable Group	GMM-DID (17) Low-Marketable Group
$infp\times post$	0.6150 **	0.4695
	(0.3067)	(0.4458)
$ifhp$	−0.5514 **	−0.4355
	(0.2787)	(0.4064)
$post$	−0.4745 **	−0.3864
	(0.2199)	(0.3354)
$profit$	−0.2635 *	0.4649
	(0.1510)	(0.2846)
$lev$	−0.0962	−0.0859
	(0.0672)	(0.0851)
$TobinQ$	0.0124 *	0.0510 ***
	(0.0072)	(0.0154)
$lnage$	0.0144	−0.2794 ***
	(0.0338)	(0.0607)
${GTFP}_{t-1}$	0.8681 ***	0.6473 ***
	(0.0296)	(0.0473)
$Constant$	0.5671 ***	1.4687 ***
	(0.2097)	(0.3814)
$Observations$	1943	859
$Number of id$	325	164

Note: *, **, and *** mean that statistics are significant at 10%, 5%, and 1% levels, respectively. Values in parentheses are robust standard errors.

, we analyzed the regression outcomes in groups. Findings indicate that the interaction term of treat × post seems to be significantly positive in the high marketization group and insignificant in the low marketization group. This means the influence of the guidelines on promoting firms to achieve green transformation is more significant in regions with higher marketization.

8. Conclusions and Recommendations

Firms’ green transformation is crucial for China to achieve sustainable development and a “dual carbon” target. However, the transformation has been slow. How to promote firms’ green transformation is challenging for various countries, particularly in developing countries like China.

This study attempts to investigate influencing factors on firms’ green transformation from a novel perspective of green credit guidelines (GCG). A-listed firms in Shanghai and Shenzhen from 2010 to 2020 are selected as samples. By using the SBM-DEA method to measure firms’ green transformation and multiple DID models to perform empirical analysis, we demonstrate that (1) GCG notably facilitates firms’ green transformation, and our conclusions still hold after a range of robustness checks. (2) Mediating effect analysis reveals that GCG facilitates green transformation through the channel of improving green technology innovation. (3) Moderating effect analysis suggests that CSR reinforces the positive influence of GCG on firms’ green transformation. (4) Firm ownership heterogeneity means that non-state firms are more significantly affected; regional marketization heterogeneity indicates that firms in regions with a higher marketization are more significantly affected.

Correspondingly, we recommend policy implications as follows. (1) A mechanism should be developed to promote the execution of GCG. First, the process of green credit standardization should be accelerated, and executive branches should develop unified policy implementation standards according to the actual situation to provide effective policy guidance for banks and financial institutions. Secondly, various government departments should strengthen coordination to realize the unified management of green credit standard implementation, make professional assessments of environmental risks of manufacturing enterprises, and issue assessment reports to provide unified standards for green credit issuance and facilitate green credit growth. Again, it is also essential to explore a green credit incentive system for banks and financial institutions. Lastly, according to the difficulty and conditions of implementing green credit policies in different regions, government departments should develop green credit policies that are suitable for their respective levels of economic development. Differentiated measures should also be formulated according to the nature of ownership of firms. For state-owned enterprises, government departments should establish strict assessment and incentive mechanisms to improve their motivation and efficiency of green transformation; for non-state-owned enterprises, government departments should formulate special policies to solve their financing difficulties and expensive financing problems to mitigate the risks of green transformation. (2) A mechanism should be developed to vigorously facilitate green technology innovation. On the one hand, firms can apply for green credit funds with low-interest rates by adopting green technologies and production modes so as to promote their own technological innovation and production transformation. On the other hand, enterprises can obtain green credit funds to increase innovation investment and improve their own innovation ability. In addition, the government should develop a unified green technology standard and green product certification system to meet the market demand. (3) Enhance firms’ consciousness of CSR. Enterprises should integrate the concept of sustainable development into their strategies, realize that environmental costs are part of production costs, and incorporate environmental risks into their risk management systems. Enterprises should first formulate clear green strategies and goals, integrate green concepts into their visions, missions, and values, establish corresponding organizational structures and incentive mechanisms, and promote the participation and implementation of all employees; Secondly, the company should assess its environmental impact and risks, adopt internationally accepted standards and indicators, regularly disclose data and progress in carbon emissions, energy consumption, water use, waste treatment, etc., accept the supervision and evaluation of all social parties, and demonstrate its social responsibility to investors in the market; finally, the company can set up a special organization or hire professionals to regularly evaluate the company’s social responsibility practices and improve the level of social responsibility disclosure. Firms should actively respond to GCG, which is not only an important way to fulfill their social responsibility but also an essential way to enhance their corporate image and credibility. Through green credit, enterprises can gain more recognition and support from the public while enhancing their sense of social responsibility and environmental awareness, promoting their green transformation and upgrading, and improving their competitiveness and sustainable development.

Compared to previous studies, our findings are somewhat consistent with the literature on green credit policy that can promote industrial transformation and upgrading [46], which focuses on the provincial level in China. In contrast, this paper focuses on firm performance at the micro level. This paper examines the impact of GCG on firms’ green transformation, which is different from traditional economic growth. More importantly, this paper uses systematic GMM and PSM-DID to address the endogeneity issues arising from mutual causality and sample selection bias in dynamic panels, which are able to identify causal relationships in a stable manner, which have been largely ignored by the existing literature. Finally, the paper explores the mediating effect of green technology innovation and the moderating effect of corporate social responsibility, which has been overlooked by the previous literature.

This study does have some limitations. (1) This paper adopts the super-efficiency SBM-DEA model, which has the advantage of concerning non-proportional changes in inputs or outputs, non-expected outputs in the efficiency analysis, and no need to set specific production function forms. However, it also has limitations with ignoring the influence of the existence of random errors on the results. (2) Adding carbon footprint/atmospheric pollution data to the system indicators is of great significance to the improvement of the green transformation index, which can be secured from the macro-regional level [47]. However, it is a great pity that the existing research can’t obtain such accurate data from enterprises.

Further research should focus on the following areas. First, the advantage of the Stochastic Frontier Model is that the model considers the influence of the existence of random errors on the results and determines the form of production function in advance, which deserves further research in the future and writing another paper as a comparison. Second, as the collection of data and information from regulators continues to improve, there is an opportunity to progressively enrich research data at the micro-firm level. Future research is needed to include carbon footprint/atmospheric pollution data in the system indicators to accurately measure firms’ green transformation. Third, future research is needed to perform multi-period DID, that is, the current time minus the implementation time of each policy, or to choose different time windows if the assumption on the steady economic structure would be relaxed. Lastly, future research is needed to compare how calculating Malmquist–Lenberger Index is changing in relation to green innovation/investment as opposed to how it is evolving in relation to changes in non-green total factor productivity.

Nevertheless, our research not only provides a reference for China’s sustainable development and high-quality economic construction but also provides a scientific basis for the development of green finance in other developing countries.