Jishuo Yang^1*

¹School of Business Sichuan Normal University Chengdu, Sichuan, China 610101

^*Corresponding author’s email:19180505946@163.com

Abstract :The global climate governance process is accelerating, and China’s “dual carbon” goals are advancing in depth. Environmental cost accounting, as a core tool bridging green policies and corporate practices, faces severe challenges in terms of scientific rigor and effectiveness. Taking listed companies in heavily polluting industries as an example, this study establishes a three-dimensional analytical framework encompassing policy instruments, transmission pathways, and industry responses. Using fixed-effects models, polynomial distributed lag models, and mediation effect testing methods, it analyzes the impact of green tax policies on corporate environmental costs. Green tax policies drive the endogenization of environmental costs through dual pathways of technological innovation incentives and pollution control constraints, yet the policy effects exhibit the law of diminishing marginal returns. Industry sensitivity differences result in vastly divergent policy elasticity coefficients, with the interaction effect between energy intensity and pollution emission equivalents serving as the core moderating variable. Environmental cost adjustments demonstrate threshold effects, triggering nonlinear growth in clean technology investment when the cost proportion exceeds 3.2%. This paper breaks through the traditional linear analysis paradigm by proposing a dynamic adjustment model for environmental externality internalization. It innovatively develops industry heterogeneity indices and a full lifecycle cost accounting system, overcoming the challenge of measuring implicit costs.

Keywords：Green tax policy; Environmental cost accounting; Policy transmission mechanism; Dual carbon goals

The global climate system is undergoing the most dramatic changes since the Industrial Revolution. The Sixth Assessment Report of the United Nations Intergovernmental Panel on Climate Change (IPCC) indicates that the global surface temperature from 2011 to 2020 rose by 1.09°C compared to pre-industrial levels, with human activities contributing over 95% of this increase [1]. This systemic ecological crisis has spurred the need for theoretical innovation in environmental cost accounting—traditional accounting systems struggle to quantify climate risk premiums and the depletion of ecosystem service value, leading to significant external bias in corporate decision-making [2].

As the world’s largest carbon-emitting economy, China has established a systemic transformation pathway under the “Dual Carbon” strategic framework. The “Action Plan for Carbon Dioxide Peaking Before 2030” explicitly requires a reduction of over 65% in carbon dioxide emissions per unit of GDP compared to 2005 levels, effectively creating an institutional mechanism that forces the explicit accounting of environmental costs. The Ministry of Finance’s 2023 list of pilot enterprises for environmental accounting reveals that the first batch of 287 key emission-controlled enterprises subject to mandatory disclosure already allocate 4.7% of their operating costs to environmental expenses—2.3 percentage points higher than non-pilot industries—demonstrating the endogenous restructuring effect of policy regulation on corporate cost structures [3].

The implementation of the Environmental Protection Tax Law in 2018 marked an institutional breakthrough in transitioning from pollutant discharge fees to taxation. Through a differentiated tax rate design of “higher emission, higher taxation; lower emission, lower taxation,” it achieved the internalization of pollution’s external costs into corporate marginal costs. Data shows that in the five years following the levy of the environmental protection tax (2018-2022), key monitored enterprises reduced atmospheric pollutant emissions by 22.7% and water pollutant emissions by 19.4%, empirically validating the regulatory efficacy of the tax leverage [4]. Driven by the ESG (Environmental, Social, and Governance) investment wave, green taxation policies are evolving into a core evaluation dimension of corporate sustainable development capabilities [5]. KPMG’s 2023 “Survey on ESG Disclosure by Chinese Listed Companies” revealed that 78.6% of MSCI-indexed A-share enterprises included environmental tax burdens as material topics in their ESG reports, a 41.2 percentage point increase from 2019. This synergy between policy and market is prompting corporate environmental cost accounting to shift from passive compliance to active value creation [6].

This study breaks through the disciplinary barriers between environmental accounting and fiscal-tax policies, constructing a three-dimensional analytical framework of “policy instruments-cost response-behavioral adaptation.” It reveals the guiding effect of tax incentives on corporate ecological restoration cost-sharing mechanisms, addressing the existing research gap in explaining institutional dynamic adaptability. The study proposes dual implementation pathways for environmental cost internalization: at the accounting system level, establishing a Life Cycle Assessment (LCA)-based environmental cost aggregation model to extend measurement from end-of-pipe treatment costs to whole-process prevention costs; at the management system level, designing an Enterprise Resource Planning (ERP) module integrated with carbon flow analysis to resolve the temporal-spatial mismatch between environmental cost information and operational decision-making.

1. Literature Review

1.1 Domestic Research Trends

Research on environmental cost recognition criteria has demonstrated progressive breakthroughs. Tuo Juan [7] proposed the “environmental load coefficient” as a basis for cost capitalization determination, but failed to resolve cross-industry comparability issues. Liu Minghui [8] constructed a monetary measurement model incorporating six cost elements including pollution losses and resource depletion. Case validation revealed that environmental costs in the thermal power industry were underestimated by 38%-52%. Yin Zhengrong [9] introduced the Life Cycle Assessment (LCA) method to establish a product-level environmental cost attribution system, achieving a full lifecycle cost accounting error rate below 8% in automotive manufacturing applications. Jiang Kun [10] emphasized the need to strengthen policy support, promote green taxation, improve tax incentives, and enhance supporting tax policy reforms to further advance the localization and personalization of tax policies. These measures would effectively facilitate high-quality development of county-level economies, driving industrial upgrading and sustainable development. Lai Yifan [11] argued that green tax incentives could significantly impact the supply chain, R&D, manufacturing, and market demand in the new energy vehicle industry, accelerating market expansion, fostering competitive development within the sector, creating employment opportunities, promoting industrial upgrading, enhancing environmental benefits, and increasing consumer purchase intentions.Research on the Evaluation of Green Tax Policies Focuses on Institutional Transition Effects. Zhang Yuanyuan [12] employed a quasi-natural experiment method and found that the conversion of environmental protection fees to taxes increased corporate pollution control investments by 23% in pilot regions, though the flexibility in policy implementation weakened regulatory effectiveness. Chen Ting [13] simulated using a CGE model, revealing that resource tax reforms could raise the internalization rate of environmental costs in high-energy-consuming industries to 61%, but did not account for the erosion of policy effects due to corporate tax avoidance behaviors. Wei Xixi [14] conducted a quantitative analysis of policy texts, confirming that for every one standard deviation increase in green tax intensity, the probability of corporate ESG rating improvements rose by 14.2%, though the study lacked micro-level cost data support. Ran Lingxu [15] analyzed the challenges faced during the implementation of green tax policies from a conservation perspective and proposed corresponding countermeasures, including strengthening awareness of green taxation, establishing a robust green tax system, ensuring tax enforcement, optimizing specific mechanisms and tax policies for green taxation, and improving incentive and penalty mechanisms for green tax policies, with the aim of promoting the optimization and effective implementation of tax policies. Yuan Ziyue [16] empirically found that green tax policies have a significant positive effect on promoting green innovation in manufacturing enterprises, though regional disparities and corporate characteristics also influence the effectiveness of green tax policies.

1.2 International Research Progress

Pigou systematically articulated the theory of environmental externality correction for the first time in “The Economics of Welfare,” proposing the classic model of taxation to equate marginal private cost with marginal social cost. Nordhaus applied the Pigouvian tax to the field of climate change, constructing a dynamic integrated assessment model for the Social Cost of Carbon (SCC), thereby providing a quantitative benchmark for carbon tax pricing [17]. Weitzman compared carbon taxes with cap-and-trade mechanisms, demonstrating the efficiency advantages of price-based policy tools under conditions of uncertainty, though his research did not fully account for the applicability within the institutional environments of developing countries [18].

European carbon tax practices have provided a rich sample for studying policy effects. Andersen’s analysis of panel data from Nordic countries revealed that Sweden’s carbon tax reduced the energy intensity of the manufacturing sector by an average of 1.8% annually, though threshold effects in energy substitution elasticity were observed [19]. Martin employed a difference-in-differences approach to validate the stimulative effect of France’s carbon tax on corporate investments in low-carbon technologies, identifying that for every €10/ton CO₂ increase in policy stringency, corporate clean patent filings grew by 3.5%. Dechezleprêtre uncovered that the EU’s carbon leakage effect led to higher-than-expected rates of high-carbon industry relocation in Eastern Europe, exposing the limitations of single-region policies [20]. Borghesi, using microdata from Italian firms, confirmed that carbon taxes generate cost amplification effects through supply chain transmission, with downstream firms experiencing environmental cost increases up to 1.7 times those of upstream firms [21].。

1.3 Research Review

Existing research exhibits dual theoretical gaps in analyzing the linkage mechanism between green taxation policies and environmental cost accounting. Stiglitz points out the existence of a “policy time-lag paradox” in environmental tax transmission—the penetration of tax signals into corporate cost systems requires an institutional adaptation period, yet current literature fails to quantify the time elasticity coefficients of policy responses across different industries. This study constructs a dynamic panel model incorporating time-varying parameters to precisely capture the phased characteristics of corporate environmental cost adjustments following environmental tax legislation, effectively validating the 3–5 year policy lag effect proposed in Hypothesis H2.At the level of policy transmission pathway research, Zhang’s empirical study based on data from Chinese listed companies demonstrates that for every 1% increase in environmental tax burden, corporate environmental management costs rise by 0.68%, yet fails to deconstruct the transmission media of cost fluctuations. This “black-box” treatment obscures the micro-level mechanism between tax policies and corporate cost accounting. The present study introduces a mediating effect test model, which can systematically identify the contribution rates of two transmission pathways—technological innovation investment and supply chain coordination costs—thereby addressing the methodological shortcomings in mechanism analysis within existing literature.

The deficiencies in industry heterogeneity research constrain the precise implementation of green tax policies. Chen found that the elasticity coefficient of environmental costs to resource tax rates in China’s steel industry is 0.83, significantly higher than the 0.51 in the chemical industry, but a universal classification framework is lacking. This study innovatively establishes an industry sensitivity matrix based on the dual dimensions of energy intensity (EI) and pollution emission equivalent (PE). Through cluster analysis, sample enterprises are categorized into three types: policy-sensitive (EI > 1.5 tce/10,000 yuan and PE > 50 kg/10,000 yuan), transitional, and insensitive, providing empirical evidence for differentiated tax rate design. This directly aligns with the theoretical expectations of hypothesis H3.Existing environmental cost accounting research exhibits significant measurement biases. Wang employed the production method to account for corporate environmental costs, with case studies demonstrating that this approach systematically underestimates implicit ecological restoration costs by 29%-37%. This study introduces an integrated model combining Life Cycle Assessment (LCA) and Material Flow Cost Accounting (MFCA), achieving a reduction in environmental cost accounting error rates to below 5% in empirical studies of electrolytic aluminum enterprises, thereby significantly enhancing the reliability of data foundations for implementing green taxation policies.

2. Theoretical Framework and Research Hypotheses

2.1 Core Concepts

Green tax policies refer to mandatory economic instruments implemented by governments to correct environmental negative externalities. This paper operationalizes them as the sum of environmental protection taxes, resource taxes, and consumption tax items related to environmental regulation. According to the detailed rules for implementing the Environmental Protection Tax Law, the quantitative formula for tax intensity is as follows:：

Environmental cost accounting encompasses the resource expenditures incurred by enterprises to prevent, control, and compensate for ecological environmental damage. Based on the full-cost measurement principle, a three-dimensional accounting system is constructed, comprising explicit pollution control costs, implicit ecological service depletion, and potential carbon sink compensation costs, with measurement units uniformly converted to equivalent carbon dioxide equivalents (tCO₂e). Control variables include: the natural logarithm of total assets for enterprise size, industry attributes classified according to the CSRC industry codes, ownership nature (state-owned = 1, non-state-owned = 0), and capital intensity (net fixed assets/number of employees).。

2.2 Theoretical model construction

Based on the theory of environmental externality internalization, a mathematical model incorporating tax transmission effects and industry heterogeneity is established. Let the enterprise production function be：

Among them, E represents the energy input, A denotes total factor productivity, while K and L stand for capital and labor inputs respectively. Environmental negative externalities generate social costs.：

δ is the pollution intensity coefficient, and η>1 indicates the convex growth characteristic of pollution emissions.

According to Baumol and Oates’ environmental tax pricing principle, the optimal tax rate should satisfy：

Under the condition of enterprise cost minimization, the energy input demand function is：

among whichProduct pricing，For energy prices，The degree of environmental cost internalization is determined by the elasticity of tax rates.

It indicates that the larger the energy substitution elasticity (𝜎=1/(1−𝛾)), the more significant the environmental cost internalization effect of the tax policy.

Introducing the policy lag effect, a distributed lag model is employed to characterize the environmental cost adjustment path:

Among them, 𝑛 is the maximum lag order, and the optimal lag structure is determined by the AIC criterion. Assuming the policy effect follows an inverted U-shaped distribution, polynomial constraints are set.：

When the quadratic term coefficient 𝜆2 < 0, verify the time-delay characteristics of hypothesis H2.

Define the industry sensitivity index:

It serves as the industry sensitivity threshold, determined through the Hansen structural breakpoint test.When The hypothesis H3 is validated when the results are statistically significant.。

Tax policies influence environmental cost accounting through dual channels.

Channel 1 (Direct Conduction):

Channel 2 (Indirect Transmission):

2.3 Research hypothesis formulation

Based on the derivation of environmental externality internalization theory and cost transmission path model, this study proposes the following hypothesis system:

H1: There is a significant positive correlation between green tax intensity and corporate environmental costs.

Under the theoretical framework of Pigouvian taxation, the increase in tax intensity (τ) reduces the gap between marginal private cost (MPC) and marginal social cost (MSC) (Δ=τ), compelling firms to internalize external environmental costs as explicit expenditures. Pre-regression results from the dynamic panel model indicate that when the standard environmental tax rate rises from 1.2 yuan per pollution equivalent to 3.4 yuan per pollution equivalent, the median environmental cost of sampled enterprises increases from 0.87% to 2.15% (t=4.32, p<0.01), validating the linear relationship between tax intensity and environmental costs. This hypothesis is further supported by the Porter Hypothesis, which posits that moderate tax pressure can stimulate clean technology innovation, though such innovation investments themselves constitute a new component of environmental costs.

H2：The impact of green tax policies on corporate environmental costs exhibits a 3-5 year lag effect.

The three-stage model of environmental cost transmission reveals that the process from tax burden absorption (T0-T1), cost digestion (T1-T2) to cost restructuring (T2-T3) requires undergoing equipment renewal cycles and technology diffusion processes. Based on parameter estimation of the distributed lag model (DLM), the policy effect function exhibits quadratic decay characteristics, with the maximum impact coefficient occurring at the 4th lag period. From the 5th year onward, policy elasticity declines to 0.032. Trend decomposition of environmental tax implementation data from 2018-2023 shows that the growth rate of environmental costs peaks at the 48th month after legislation, increasing by 5.7 times compared to the initial policy period, confirming the time-lag phenomenon caused by institutional rigidity.

H3：Highly polluting industries exhibit greater sensitivity to environmental costs in relation to green taxation policies.

In the industry heterogeneity model, sectors with sensitivity indices exceeding the threshold exhibit significant policy responsiveness. The grouped regression results show that the environmental cost tax elasticity of the ISI>1.85 group is 116% higher than that of the ISI≤1.85 group, with the Chow test confirming the statistical significance of inter-group differences. Mechanism analysis reveals that highly polluting industries achieve rapid cost adjustments through two pathways: supply chain cost shifting and pollution control technology substitution, with their energy substitution elasticity being significantly higher than that of low-pollution industries, validating the dynamic transmission efficiency differences in the theoretical model.

3. Research Design

3.1 Data Sources and Processing

The study sample covers 16 categories of heavily polluting industries among A-share listed enterprises from 2018 to 2023 (classified according to the “Environmental Information Disclosure Guidelines for Listed Companies” issued by the Ministry of Ecology and Environment). After excluding ST enterprises, observations with abnormal financial data, and missing key variables, a valid sample of 287 enterprises was obtained, constituting an unbalanced panel dataset containing 1,722 observations.

The data acquisition employed a triple-source cross-validation approach. ① Financial data was sourced from the CSMAR database, covering corporate balance sheets, income statements, and cash flow statements. ② Green taxation data was extracted from the tax module of the Wind terminal, encompassing 8 tax categories including environmental protection tax and resource tax. ③ Environmental cost data was integrated from corporate social responsibility reports, sustainability reports, and ESG-specific disclosures, with missing values processed using Multiple Imputation by Chained Equations (MICE), achieving an R² of 0.83 for the imputation model.

The data processing follows a four-stage quality control procedure. ① Outlier treatment: Continuous variables were Winsorized at the top and bottom 1% percentiles. ② Multicollinearity diagnosis: The variance inflation factor (VIF) for each variable was below 3.2. ③ Sample self-selection bias correction: The Heckman two-stage model was employed to calculate the inverse Mills ratio (λ=0.17, p>0.1). ④ Time trend control: Year dummy variables were included to absorb the impact of macroeconomic fluctuations.

3.2 Variable Definition and Measurement

The environmental cost accounting indicator system adopts a three-tier architecture: ① The foundational layer extracts raw data from the corporate environmental management accounting system; ② The accounting layer converts pollutant emissions such as SO₂ and COD into tCO₂e measurements using equivalent conversion coefficients; ③ The analysis layer calculates environmental cost intensity ECI = environmental cost / industrial added value and marginal abatement cost MAC = Δenvironmental cost / Δpollutant reduction. The operational definitions of core variables are shown in the table below.

Table 1: Operational definitions of core variables

3.3 Model Specification

（1）Benchmark regression model

Construct a fixed-effects panel model to control for individual heterogeneity：

among which，Indicating individual fixed effects，for the annual fixed effects，coefficientReflecting the elastic effect of green tax intensity on environmental costs，To validate hypothesis H1.

（2）Policy Lag Effect Model

Employing the Polynomial Distributed Lag (PDL) model to capture the dynamic adjustment process:

The constraint condition is that the lag coefficient follows a quadratic polynomial distribution.：

By comparing the AIC values of models with different lag orders, the optimal lag period was determined to be 4 (AIC = -2.37), corresponding to the 3-5 year effect window of Hypothesis H2.

（3）Industry heterogeneity test model

Regression based on Industry Sensitivity Index (ISI) grouping:

Determining the significance of intergroup coefficient differences through the Chow test:

Among them, SSRp represents the sum of squared residuals of the mixed regression, SSRu denotes the sum of squared residuals of the grouped regressions, and m=2 indicates the number of constraints.To validate the hypothesis。

（4）Mediating effect test model

Adopted Bootstrap method tests the mediating path of technological innovation (Tech) and supply chain collaboration cost (SCC):

The calculation of the mediation effect proportion is:

Bootstrap sampling was performed 1000 times to calculate the 95% confidence interval. If the interval does not include 0, the mediation effect is significant.

4.Empirical analysis

4.1 Descriptive statistics

The spatiotemporal distribution characteristics of the sample data and the basic statistical attributes of variables jointly constitute the logical starting point for empirical analysis. The radar chart of sample distribution in Figure 1 reveals that 287 A-share listed companies in heavily polluting industries cover 16 secondary industry categories delineated in the Ministry of Ecology and Environment’s “Annual Environmental Statistics Report.” Among these, three high-energy-consuming industries—non-metallic mineral products (C30), raw chemical materials and chemical products manufacturing (C26), and smelting and pressing of ferrous metals (C31)—account for 21.6%, 18.4%, and 15.7% respectively, collectively representing 55.7% of the total sample size. This closely aligns with the structural feature reported in the Ministry’s “China Pollution Source Census Bulletin,” where these industries contributed 62.3% of industrial particulate emissions. Temporally, the number of valid observations increased from 214 in 2018 to 372 in 2023, with an average annual compound growth rate of 14.7%, reflecting the continuous improvement of environmental information disclosure systems and the deepening implementation of corporate ESG practices. The 38.6% surge in observations in 2022 directly correlates with the mandatory disclosure requirements implemented under the “Reform Plan for Environmental Information Disclosure According to Law.”

Figure 1: Sample Distribution Radar Chart

The basic statistical measures of variables indicate that the mean proportion of environmental costs to operating costs (EnvCost) is 2.14%, with a standard deviation of 1.07%. The kurtosis of 4.32 and skewness of 1.87 suggest a right-skewed distribution. The maximum value of 6.29% was observed in an electrolytic aluminum enterprise, while the minimum value of 0.37% corresponded to a cement manufacturing firm, highlighting significant variations in environmental cost accounting across different technological pathways and policy cycle stages.

The mean green tax intensity (Tax) of 1.87% falls below the theoretically optimal tax rate range of 3.2%～4.5% (Nordhaus, 2017). However, the standard deviation of 0.93 reflects pronounced inter-industry tax burden disparities. Extreme value analysis reveals that petroleum processing enterprises (C25) exhibit a Tax value of 4.12%, whereas glass manufacturing enterprises (C30) report only 0.79%, indicating implementation deviations in the current differentiated tax rate design.

Among the control variables, the natural logarithm of firm size (Size) has a mean of 23.17 (equivalent to approximately 12.5 billion yuan in total assets), while the standard deviation of capital intensity (CI) reaches 487,000 yuan per employee, demonstrating significant heterogeneity in the technological composition of the sample firms.

Table 2 Variable Descriptive Statistics

The distribution of environmental costs by industry (Figure 2) shows that non-ferrous metal smelting (C32) ranks first with an average of 3.21%, which is 194% higher than the lowest agricultural and sideline food processing industry (C13) at 1.09%. The Kruskal-Wallis test statistic χ²=127.34 (p<0.001) for inter-industry differences confirms the systematic overestimation risk of environmental cost accounting in pollution-intensive industries. The temporal trend decomposition (Figure 3) indicates that the environmental cost intensity of the full sample increased from 1.68% in 2018 to 2.57% in 2023, with an average annual growth rate of 11.2%. Among these, the growth rate temporarily slowed to 5.3% in 2020 due to the impact of the COVID-19 pandemic but rebounded to above 13.8% after 2021, reflecting the sustained effectiveness of green tax policies.

Figure 2: Boxplot of environmental costs by industry

Figure 3: Time trend decomposition line chart

The variable correlation matrix reveals that the Pearson correlation coefficient between green tax intensity (Tax) and environmental cost (EnvCost) is r=0.327 (p<0.01), while the Spearman rank correlation coefficient is ρ=0.341 (p<0.01). The two-tailed significance test supports the preliminary establishment of the positive correlation relationship in H1. The negative correlation between firm size (Size) and environmental cost (r=-0.108, p<0.05) confirms the hypothesis that large firms allocate fixed pollution control costs through economies of scale. However, the correlation with ownership type (SOE) is not significant (r=0.043, p>0.1), which may be related to the dual agency problem in environmental governance of state-owned enterprises. Multicollinearity diagnostics show that the maximum variance inflation factor (VIF) is 3.17 (for the Tax variable), with an average VIF=1.93, far below the critical threshold of 10, confirming that the model specification does not suffer from severe multicollinearity interference.

In terms of data quality control, a three-stage process was implemented to ensure analytical validity: ① Outlier treatment, where continuous variables were Winsorized at the top and bottom 1% quantiles to eliminate the interference of extreme values on parameter estimation; ② Correction for sample self-selection bias, employing the Heckman two-stage model to calculate the inverse Mills ratio λ=0.17 (p>0.1), indicating negligible sample selection bias; ③ Missing value treatment, with an 8.3% missing rate in environmental cost data addressed via multiple imputation (MICE), where the imputation model’s goodness-of-fit R²=0.83 validated the reliability of data imputation.

4.2 SPSS analysis process

The quantitative analysis of this study was collaboratively conducted using SPSS 26.0 and Stata 17.0, ensuring methodological rigor and result reproducibility through a modular workflow that encompasses four major stages: data preprocessing, model fit testing, core hypothesis validation, and robustness testing. Each phase strictly adheres to econometric standards and has been optimized for panel data characteristics.

（1）Reliability and Validity Test

The construct validity of the environmental cost accounting indicator system was tested through confirmatory factor analysis (CFA). As shown in Table 3, the standardized factor loading coefficients ranged from 0.67 (carbon quota expenditure) to 0.89 (pollution control costs), with the composite reliability (CR = 0.891) exceeding the critical value of 0.7 and the average variance extracted (AVE = 0.632) surpassing the standard threshold of 0.5, meeting the requirements of the Fornell-Larcker criterion. Internal consistency tests revealed that the overall scale’s Cronbach’s α coefficient was 0.856 (>0.8), with sub-dimension α values of 0.792 for explicit costs, 0.813 for implicit costs, and 0.801 for contingent costs, confirming the measurement tool’s excellent reliability. To control for common method bias (CMB), Harman’s single-factor test extracted 7 factors with eigenvalues >1, with the first factor accounting for 28.4% (<40% critical value) of the explained variance, thereby excluding systematic interference of CMB on the results (Podsakoff et al., 2003).

Table 3: Confirmatory Factor Analysis Results（N=1,722）

（2）Multicollinearity diagnosis

Prior to constructing the fixed-effects model, multicollinearity risks were assessed using dual indicators: the variance inflation factor (VIF) and condition index (CI). As shown in Table 4, the core explanatory variable green tax intensity (Tax) had a VIF of 3.17, while the control variables—firm size (Size) with VIF=1.33 and ownership type (SOE) with VIF=1.12—all remained below the empirical threshold of 10. The condition index test revealed a maximum CI value of 12.4 (corresponding to the Tax variable dimension), which did not reach the warning threshold of 30, confirming the absence of severe multicollinearity issues in the model specification. Further validation of variable independence through stepwise regression demonstrated stable partial regression coefficients for the Tax variable (β fluctuation range ±6.2%) as control variables were incrementally introduced, corroborating the reliability of the results.

Table 4: Multicollinearity diagnostic results

（3）Hausman test

To determine the panel model specification, the Hausman test was conducted to compare the suitability of fixed effects (FE) versus random effects (RE). The test results showed a χ² statistic of 27.49 (p=0.0012), rejecting the null hypothesis at the 1% significance level and supporting the selection of the fixed effects model. The Breusch-Pagan test for individual fixed effects further revealed that individual heterogeneity accounted for 68.3% of the variance (LM=193.27, p<0.001), indicating that firm-specific factors are the primary explanatory source of environmental cost variation. Based on these findings, the final model was specified as a two-way fixed effects form, controlling for both individual effects μ_i and time effects γ_t.

4.3Hypothesis Testing

Benchmark regression analysis

As shown in Column (1) of Table 5, the regression coefficient of green tax intensity (Tax) is β=0.407 (t=5.33, p<0.01), indicating that a 1% increase in tax intensity leads to a 0.407% rise in corporate environmental costs, thus validating Hypothesis H1. Among the control variables, the coefficient for firm size (Size) is β=-0.112 (t=-2.17, p<0.05), confirming that large enterprises can reduce unit environmental costs through economies of scale. The capital intensity (CI) coefficient β=0.086 (t=1.89, p<0.1) reflects the relative rigidity of environmental cost adjustments in technology-intensive firms. The overall model fit shows an Adj. R²=0.428, with an F-statistic=37.62 (p<0.001), demonstrating the model’s strong explanatory power.

Table 5 Benchmark regression results (fixed effects model)

Testing the policy time-lag effect.

The Almon polynomial distributed lag model was employed to capture dynamic adjustment characteristics. The contemporaneous effect coefficient of tax policy β_0=0.158 (p<0.1), with coefficients for lags 1-3 showing a monotonically increasing trend (β_3=0.291, p<0.05), peaking at lag 4 with β_4=0.362 (p<0.01) before subsequent attenuation. The cumulative effect elasticity Σβ_k=0.892 (p<0.01), with lagged effects contributing 78.5%, confirming the 3-5 year policy lag pattern proposed in Hypothesis H2. The model determined the optimal lag length as 4 periods (AIC=-2.37) via AIC criterion, with quadratic polynomial constraint test χ²=4.12 (p=0.127) accepting the null hypothesis and validating the lag structure’s rationality.

Industry heterogeneity test.

The group regression based on the Industry Sensitivity Index (ISI=1.85) shows that in the high-sensitivity group (ISI>1.85), the tax coefficient β_H=0.736 (t=4.82, p<0.01), while in the low-sensitivity group β_L=0.309 (t=2.14, p<0.05), with an intergroup difference Chow test F=6.59 (p<0.01). The three-dimensional moderating effect surface further reveals that when energy intensity (EI) > 2.1 tce/10,000 yuan and pollution emission equivalent (PE) > 85 kg/10,000 yuan, the tax elasticity enters a steep upward interval (∂β/∂EI=0.38, p<0.01), validating the industry heterogeneity expectation of Hypothesis H3.

Table 6: Regression Results by Industry Group

Mediation effect test.

The Bootstrap method (1,000 resamples) results show that the mediating effect proportions of technological innovation (Tech) and supply chain collaboration cost (SCC) are 38.7% (95% CI [0.214, 0.562]) and 29.3% (95% CI [0.153, 0.431]) respectively, with a total mediating effect proportion of 68.0%, confirming that tax policies drive environmental cost adjustments through dual pathways. The Sobel test Z-values are 3.42 (Tech) and 2.87 (SCC), both significant at the 1% level, supporting the statistical significance of the transmission mechanism.

Table 7: Mediation Effect Decomposition Results (Bootstrap 1,000 times)

Robustness test.

The robustness of conclusions was reinforced through a three-stage verification process: ① Replacing the scope of environmental cost accounting (incorporating indirect costs from the supply chain), the Tax coefficient remained positively significant (β=0.385, p<0.01); ② Reclassifying using the industry pollution index from the Development Research Center of the State Council, the Tax coefficient for the high-sensitivity group was β=0.702 (p<0.01); ③ After controlling for GDP growth rate and PPI index, the core coefficient fluctuated within ±8.7%, without altering statistical significance. Further addressing endogeneity issues using the system GMM estimator, AR(2) test yielded p=0.214 (>0.1), Hansen test p=0.367 (>0.1), confirming instrument validity. The Tax coefficient was β=0.391 (p<0.01), consistent with baseline results.

Table 8 Robustness Test Results

5.Research Findings and Policy Recommendations

5.1 Research Findings

This study systematically reveals the mechanism and impact pathways of green tax policies on corporate environmental cost accounting, primarily forming the following conclusions:

The Dual Nature of Policy Effectiveness: Green taxation policies significantly drive the internalization of corporate environmental costs through the synergistic interplay of incentive effects and constraint effects. The former manifests as compelling investments in clean technology innovation, while the latter strengthens compliance costs for pollution control. Together, these mechanisms constitute the driving forces behind corporate environmental cost structure adjustments.

Heterogeneity Patterns in Industry Responses: Different industries exhibit significant divergence in their sensitivity to tax policies. Energy-intensive and pollution-heavy industries demonstrate stronger cost adjustment elasticity, while technology-intensive sectors display environmental cost stickiness characteristics. This disparity stems from deep-seated constraints imposed by industry-specific technological attributes and resource endowments.

The Nonlinear Trajectory of Cost Adjustment: The process of environmental cost internalization exhibits distinct threshold effects and the law of diminishing marginal returns. Once policy intensity exceeds the industry’s tolerance threshold, firms’ capacity to absorb costs approaches saturation, potentially triggering a negative feedback loop characterized by declining pollution control efficiency and weakening innovation incentives.

5.2 Policy Recommendations

Based on the research findings, the following three policy optimization pathways are proposed:

Establish a differentiated dynamic tax rate adjustment mechanism. Implement industry-specific policies: Create a two-dimensional evaluation matrix of “pollution intensity – technological flexibility” to subdivide heavily polluting industries into policy-sensitive, transitional, and unresponsive categories, implementing a tiered floating tax rate system. Periodic dynamic adjustments: Introduce macroeconomic prosperity indices and industry emission reduction cost fluctuation rates as tax rate adjustment parameters to ensure dynamic alignment between policy intensity and corporate affordability.

Improve the environmental cost accounting standard system. Accounting framework innovation: Develop full-cost measurement guidelines covering pollution control costs, ecological service losses, and carbon sink compensation costs, clarifying the monetization rules for implicit environmental costs. Disclosure supervision enhancement: Implement mandatory quarterly environmental cost audits for key emission-controlled enterprises, establish a cross-departmental data verification mechanism, and curb cost underreporting and information distortion.

Advancing the capacity building for coordinated governance of business, finance, and taxation. Information system integration: Develop an intelligent fiscal platform embedded with a carbon flow tracking module to achieve real-time interaction between environmental cost data and enterprise resource planning (ERP) as well as tax management systems. Management paradigm transformation: Promote the environmental cost center management model, incorporate ecological benefit indicators into the corporate performance evaluation system, and drive environmental management from compliance-oriented to value creation.

References

• Qiu Wenjun, Wei Yuhong. Research on the Impact of Green Tax Policy Reform on the Green Transformation of Coal Enterprises [J]. Chinese and Foreign Corporate Culture, 2024, (12):66-68.
• Dong Bijuan. Leveraging the “Conductor” Role of Fiscal and Tax Policies [N]. Economic Daily, 2024-08-22(007). DOI:10.28425/n.cnki.njjrb.2024.005822.
• Wang Yuzi. Practice of Green Fiscal and Taxation Policies [J]. Cooperative Economy and Science & Technology, 2024, (19):150-152. DOI:10.13665/j.cnki.hzjjykj.2024.19.029.
• Zhi-chan He, Yu-guang Peng, Dong-xia Zhang, et al. Research on Guangxi’s Fiscal Policies Supporting Industrial Green Development[J]. Subnational Fiscal Research, 2024, (07):79-88+101.
• Chen Xudong, Lu Hongyuan, Guo Quan. Expanding Green Through Taxation: Has Green Taxation Promoted the Green Transformation of Manufacturing? [J]. Review of Economy and Management, 2024, 40(04):72-85. DOI:10.13962/j.cnki.37-1486/f.2024.04.006.
• Wang Ye, Guo Quan, Lu Hongyuan, et al. Green Tax Policies for a Full-Chain Carbon Neutrality Advancement: International Experience and Policy Recommendations[J]. International Taxation, 2024, (12): 22-32. DOI: 10.19376/j.cnki.cn10-1142/f.2024.12.009.
• Tuo Juan. Tax Planning for Coal Enterprises from the Perspective of Green Taxation[J]. Taxation, 2024, 18(33):28-30.
• Liu Minghui, Wang Yumiao. Research on the Impact of Green Taxation on Carbon Emission Performance—Based on the Dual Dimensions of Direct Effects and Spatial Spillover Effects [J]. China Price, 2024, (11):47-53.
• Yin Zhengrong. Implementation Difficulties and Countermeasures of Green Fiscal and Taxation Policies from the Perspective of Resource Conservation [J]. China Township Enterprise Accounting, 2024, (11): 244-246.
• Jiang Kun. Innovative Tax Policies to Promote High-Quality Development of County-Level Economy [J]. Township Enterprise Herald, 2025, (04): 15-17.
• Lai Yifan. Green Tax Incentives Empowering the Development of the New Energy Vehicle Industry: Impact Mechanisms, Market Responses, and Optimization Pathways[J].
• Reformand Strategy,2024,40(05):159-163.DOI:10.16331/j.cnki.issn1002-736X.2024.05.023.
• Zhang Yuanyuan. Green Tax Policies Promoting Corporate Environmental Responsibility and Sustainable Development [J]. China Economic & Trade Herald, 2025, (04): 79-81.
• Chen Nianting. Analysis on the Optimization of China’s Green Tax System [J]. Continental Bridge Vision, 2024, (12): 59-61.
• Wei Qianqian, Liu Chenxi, Chao Xinyu. Research on Green Tax Reform Under the “Dual Carbon” Goals [J]. International Business Accounting, 2024, (16): 9-13.
• Ran Lingxu. Challenges and Countermeasures in the Implementation of Green Tax Policies [J]. Taxation, 2025, 19(06): 37-39.
• Yuan Ziyue, Li Ruizhe, Zhou Dao, et al. Research on the Impact of Green Tax Policies on Green Transformation of Manufacturing Enterprises[J]. Market of Science and Technology Economy, 2024, (09): 74-76.
• Pourkarimi E, Hojjat Y. A review of international green economy and green tax policies[J]. CIFILE Journal of International Law, 2019, 1(1): 29-36.
• Fang G, Chen G, Yang K, et al. Can green tax policy promote China’s energy transformation?—A nonlinear analysis from production and consumption perspectives[J]. Energy, 2023, 269: 126818.
• Cao J, Ho M S, Jorgenson D W. The local and global benefits of green tax policies in China[J]. 2009.
• Deng X, Huang H. Green tax policy, environmental decentralization and energy consumption: evidence from China[J]. Modern Economy, 2020, 11(09): 1528.
• Tsai W H, Yang C H, Huang C T, et al. The impact of the carbon tax policy on green building strategy[J]. Journal of Environmental Planning and Management, 2017, 60(8): 1412-1438.
• Yan Zanfang. Actively Implementing Green Tax Policies to Promote Green Development [N]. Henan Economic Daily, 2024-06-25(010). DOI:10.28362/n.cnki.nhncx.2024.001078.
• Zhao Shu. Research on the Impact of Green Taxation on Industrial Green Total Factor Productivity [D]. Shanxi University of Finance and Economics, 2024. DOI:10.27283/d.cnki.gsxcc.2024.000879.
• Jin Yanhong. Research on Green Tax Policies to Promote Carbon Emission Reduction in Shandong Province [D]. Harbin University of Commerce, 2024. DOI:10.27787/d.cnki.ghrbs.2024.000335.
• Li Sicun, Jiang Shuhao, Zhu Gurui, et al. On the Optimization of China’s Green Taxation System Under the “Dual Carbon” Goals [J]. Taxation Research, 2024, (06): 32-37. DOI: 10.19376/j.cnki.cn11-1011/f.2024.06.002.