Has the Environmental Tax Reform Encouraged Companies to Invest Across Regions?—Evidence from Chinese Listed Companies

Abstract

Against the dual backdrop of the global green transition and China’s construction of a unified national market, the impact of environmental policies on the spatial allocation of capital has emerged as a core concern for academic researchers and policymakers alike. China’s implementation of the Environmental Protection Tax Law in 2018 completed the historic transition from the long-standing pollution discharge fee system to a formal environmental tax regime. While differentiated interprovincial tax rates under the new regime have the potential to reshape cross-regional capital flow patterns, existing literature lacks systematic micro-level empirical evidence regarding how environmental tax reform affects firms’ cross-regional investment and the underlying mechanisms driving such effects. This study constructs a sample of Chinese A-share-listed firms covering the period 2012 to 2024 and applies the difference-in-differences (DID) method to systematically examine the causal impact, transmission channels, and heterogeneous boundary conditions of environmental tax reform on firms’ cross-regional investment. The baseline estimation results confirm that environmental tax reform significantly stimulates cross-regional investment activities among firms located in provinces that raised environmental tax rates after the reform. This core conclusion remains robust across a series of validity checks, including parallel trend assumption tests, placebo tests, propensity score matching combined with difference-in-differences (PSM-DID) estimation, and the exclusion of confounding effects from contemporaneous policy interventions. Mechanism analysis identifies two core transmission channels through which the reform exerts its effects: rising pollution abatement costs and alleviation of corporate financing constraints. Further heterogeneity tests reveal that the promoting effect of the reform on cross-regional investment is more prominent for capital-intensive firms and firms located in regions with lower fiscal pressure and firms operating in regions with stricter environmental law enforcement. This study provides new micro-level empirical evidence supporting the applicability of both the pollution haven hypothesis and the Porter hypothesis in the Chinese institutional context and offers actionable policy insights for optimizing the design of the environmental tax system, guiding corporate green transformation, and facilitating coordinated regional development.

1. Introduction

Against the dual background of global green transformation and China’s efforts to construct a new development pattern, the dynamic interactive relationship between environmental policy design and capital allocation efficiency has emerged as a core topic of widespread academic concern. In 2018, the Environmental Protection Tax Law of the People’s Republic of China was officially implemented, marking the formal completion of the transition from the long-standing pollution discharge fee system to a standardized environmental protection tax system. Distinct from administrative fee collection mechanisms, the environmental protection tax internalizes the negative externalities of corporate pollution activities through targeted tax leverage; its economic impact is not limited to constraining enterprises’ polluting production behaviors but also exerts a far-reaching influence on enterprises’ long-term investment decision-making and cross-regional location selection strategies. It is noteworthy that this reform endows provincial-level governments with discretionary authority to set specific applicable tax rates within the unified range specified by the national legal framework. The resulting inter-provincial differences in actual environmental tax burdens may trigger cross-regional flows of capital from regions with higher tax rates to regions with relatively lower tax rates, inducing a potential “pollution haven” effect within the country [1]. This effect creates a potential practical tension with the established policy objectives of building a unified national market and promoting the free and efficient flow of production factors across regions [2,3].

Early research on cross-regional capital flows was deeply rooted in the theory of eclectic paradigm of international production, which explained enterprises’ foreign investment behavior from three core dimensions: ownership advantages, internalization advantages, and locational advantages. Among these dimensions, locational advantages are usually interpreted as consisting of traditional economic factors such as regional market size, factor prices level, and infrastructure construction quality. In recent years, however, a growing body of researchers has confirmed that the institutional environment, as a non-traditional locational factor, is also a key force shaping enterprises’ cross-regional investment decisions [4]. The impact of environmental regulations on capital flows has long been the focus of theoretical debate: the pollution haven hypothesis holds that strict environmental regulatory standards will increase enterprises’ environmental compliance costs, thus motivating polluting enterprises to relocate to regions with relatively loose regulatory requirements. By contrast, the Porter hypothesis argues that appropriately designed environmental regulation can effectively stimulate enterprises’ innovative activities, and the resulting “innovation dividend” is sufficient to offset even exceed the additional compliance costs brought by regulation. Nevertheless, the existing theoretical and empirical debates on these two hypotheses have largely focused on the context of cross-border foreign direct investment (FDI), while there are still very few studies exploring the impact of environmental tax reform on the spatial allocation of capital among Chinese provinces, especially a notable lack of causal identification studies based on firm-level microdata [5,6].

This paper treats the 2018 Environmental Protection Tax reform as a quasi-natural experiment, leveraging the fact that provincial governments were authorized to set local tax rates above the national minimum, employs the difference-in-differences (DID) method, and uses data on Chinese A-share listed companies from 2012 to 2024 to examine the impact of the tax reform on enterprises’ cross-regional investment behavior and its underlying transmission mechanisms [7]. This study identifies the causal effect of environmental policy adjustment on domestic inter-provincial capital allocation, integrates the pollution cost mechanism and financing constraint mechanism into a unified analytical framework to reveal the complete logic of how the tax reform drives enterprises’ cross-regional investment decisions, and expands the explanatory power of the pollution haven hypothesis and Porter hypothesis in the specific institutional context of China as a large developing economy [8,9].

The empirical results show that the implementation of the tax reform significantly promotes enterprises’ cross-regional investment activities, and this effect is transmitted through two parallel channels: the increase in regional pollution compliance costs and the alleviation of green financing constraints for enterprises [10]. The impact of the tax reform also shows significant heterogeneity characteristics along dimensions such as enterprise capital intensity, regional fiscal situation, and local environmental regulation enforcement intensity, which provides micro-level empirical evidence for the applicability of the two classical hypotheses in China’s institutional context [11,12].

The rest of this paper is structured as follows: Section 2 reviews the relevant literature, constructs the theoretical analysis framework, and proposes research hypotheses, Section 3 introduces the data sources, variable measurement, and empirical model design, Section 4 reports the baseline regression results, robustness test results, heterogeneity analysis findings, and mechanism test results, and Section 5 discusses the research conclusions, policy implications, and the limitations of this study in terms of sample composition, variable measurement accuracy, and the complete causal chain of the financing constraint channel.

2. Theoretical Background and Research Hypotheses

2.1. Literature Review

2.1.1. Environmental Taxation and Capital Location Choice: Theoretical Divergences and Evolution

The association between environmental regulatory stringency and capital location selection has constituted a core research topic in environmental economics and international business scholarship since the initial formulation of the pollution haven hypothesis by Walter and Ugelow in 1979. The foundational logic of this hypothesis centers on the following causal chain: stringent environmental regulations impose additional compliance costs on production entities, and profit-maximizing pollution-intensive enterprises are consequently incentivized to relocate operational facilities to jurisdictions with comparatively lenient regulatory frameworks. This theoretical proposition is, however, contested by the Porter hypothesis, which was advanced by Porter and van der Linde in 1995. The Porter hypothesis posits that appropriately designed environmental regulatory instruments can stimulate firm-level technological innovation, and the efficiency improvements generated by such innovation can offset or even surpass the incremental compliance costs, thereby suppressing rather than inducing cross-regional relocation of polluting activities.

In recent years, a growing body of scholarship has recognized that the effect of environmental regulations on capital flows does not follow a unidirectional linear pattern but is instead contingent on the interaction of multiple contextual factors, including institutional design, market structure, and firm-level heterogeneity [13,14]. Current empirical evidence is predominantly focused on cross-border capital flows in the form of foreign direct investment, while investigations into the spatial allocation of capital across Chinese provincial jurisdictions remain comparatively limited [15]. Given the substantial volume of interprovincial capital flows within China, the impact of regional heterogeneity in environmental policy on domestic capital allocation efficiency warrants systematic examination, particularly against the policy backdrop of ongoing efforts to construct a unified national market [16,17].

A burgeoning body of empirical research has further expanded the scholarly discourse on this topic. Studies focused on the Chinese context provide robust empirical evidence that environmental regulations exert a statistically significant influence on the location decisions of polluting industries, underscoring the pivotal role of regional regulatory heterogeneity in shaping domestic capital flows [18]. Parallel research examining the European Union’s Carbon Border Adjustment Mechanism (CBAM) and its impact on foreign direct investment (FDI) demonstrates that climate policy instruments can elicit complex industrial relocation responses, even within advanced economy contexts [19,20]. Collectively, this line of inquiry points to a growing scholarly consensus: environmental policy no longer operates solely as a constraint on firm behavior, but functions as an active force reshaping the geography of capital [21]. Notably [22], within China’s institutional framework—where provincial-level governments retain substantial autonomy in implementing environmental governance policies—interprovincial disparities in environmental tax rates constitute a particularly well-suited empirical setting for testing the domestic applicability of the pollution haven hypothesis [23].

2.1.2. Corporate Cross-Regional Investment: From Traditional Factors to Institutional Drivers

Theoretical perspectives on the determinants of corporate cross-regional investment have undergone gradual evolution in recent decades. Dunning’s OLI paradigm, first proposed, long served as the dominant analytical framework in this field [24]. In its initial formulation, the paradigm conceptualized locational advantages primarily in terms of traditional economic attributes, including market size, factor prices, and infrastructure endowment. Dunning later revised and expanded the framework, integrating institutional quality into the definition of locational advantages and emphasizing the role of both formal and informal institutions in shaping corporate location decision-making.

In the Chinese context, institutional factors play an especially pronounced role in investment decision-making: environmental regulations, fiscal decentralization arrangements, and local government behavioral incentives interact to form the complex institutional environment that constrains and enables corporate investment activities. Existing research on corporate cross-regional investment in China has predominantly focused on traditional determinants such as transportation infrastructure, inter-jurisdictional tax competition, and regional financial development, with relatively few studies positioning environmental regulations as a core explanatory variable. For example, Duan et al. (2021) demonstrate that the opening of high-speed rail lines significantly facilitates cross-regional investment by reducing geographical transaction friction [24], while Dong et al. (2023) show that local government fiscal pressure distorts the efficiency of bank credit resource allocation, indirectly affecting firms’ interprovincial investment capacity [25,26]. While these studies provide valuable empirical insights, they largely neglect the role of environmental policy as a distinct institutional driver of spatial capital reallocation.

2.1.3. Environmental Tax Reform and Corporate Investment Behavior: Research Progress and Limitations

Empirical research on the economic consequences of environmental tax reform emerged comparatively recently, but the volume of relevant scholarship has expanded rapidly following the implementation of China’s Environmental Protection Tax Law in 2018. Existing studies identify two primary pathways through which environmental tax reform influences corporate behavior. The first is the pollution cost channel: the environmental tax internalizes previously externalized environmental costs into corporate decision-making frameworks, directly increasing the tax burden and pollution governance expenditures of high-polluting enterprises. The second is the financing constraint channel: environmental tax reform establishes a standardized mechanism for tax-related information sharing while sending clear policy signals encouraging green development. This institutional arrangement alleviates information asymmetry between enterprises and financial institutions to a certain extent, thereby improving financing conditions for eligible firms [27].

While existing studies have laid an important analytical foundation for understanding the economic implications of environmental tax reform, critical research gaps remain [28]. Most existing scholarship focuses on the direct impact of environmental protection taxes on corporate green innovation or financial performance, with limited attention paid to the spatial dimension of corporate investment strategies. Additionally, existing studies typically examine the pollution cost and financing constraint pathways separately, with few attempts to integrate both mechanisms into a unified analytical framework. This fragmentation limits the ability to fully capture the underlying logic through which environmental protection tax reform drives inter-regional corporate investment [29,30,31].

Recent literature has begun to explore the multifaceted impacts of the 2018 Environmental Protection Tax Law. Liu et al. (2022) investigate the moderating role of CEO green experience in shaping corporate responses to the environmental protection tax, finding that executive environmental background significantly influences firms’ disclosure strategies and policy response boundaries [10]. He et al. (2023) examine the effect of environmental taxation on enterprises, providing empirical evidence that the tax reform has catalyzed sustainable capital allocation among Chinese listed firms [11]. With respect to financing outcomes, Cui (2024) employ a quasi-natural experimental design to show that the “fee-to-tax” reform affects corporate debt financing costs, with the direction of the effect contingent on firms’ pre-reform environmental performance [12]. While these emerging studies make valuable contributions to the literature, they primarily focus on single-dimensional outcomes—either green innovation, financial performance, or financing costs—without systematically addressing the spatial dimension of corporate investment responses. The present study addresses this research gap by examining how the reform simultaneously reshapes firms’ investment location choices through the dual channels of cost pressure and financing alleviation [32,33].

2.2. Theoretical Framework and Hypotheses

This section synthesizes insights from the pollution haven hypothesis, cost-compliance theory, information asymmetry theory, and institutional theory to develop a theoretical framework explaining the causal link between environmental tax reform and firms’ cross-regional investment decisions and proposes testable research hypotheses based on this framework.

2.2.1. The Pollution Haven Hypothesis and Main Effects

The core proposition of the pollution haven hypothesis holds that regional disparities in environmental regulatory intensity alter the relative compliance costs faced by enterprises; rational market actors have incentives to shift production activities from highly regulated regions to less regulated jurisdictions to maintain market competitiveness. The theoretical mechanism operates through three interconnected stages. First, environmental regulations impose direct compliance costs on firms, including pollution control expenditures, emission-related tax payments, and production process adaptation investments. Second, when these costs vary systematically across administrative jurisdictions, they create spatial gradients in the overall cost of conducting business operations. Third, rational firms respond to these cost gradients by reallocating productive capacity to lower-cost locations, provided the expected cost savings exceed the fixed costs associated with cross-regional relocation.

China’s 2018 environmental tax reform provides a particularly appropriate institutional context for testing the domestic validity of this hypothesis. Under the unified national tax law framework, provincial-level governments were granted authority to set specific tax rates within statutory limits, resulting in measurable interprovincial variation in actual environmental tax burdens. This institutional design creates a quasi-experimental setting: firms operating in provinces with higher tax rates face heavier environmental compliance burdens, while those in low-tax provinces benefit from relative cost advantages. The resulting tax rate differentials generate precisely the type of spatial cost gradient that the pollution haven hypothesis predicts will trigger capital reallocation across jurisdictions.

Firms facing increased environmental tax burdens may adopt two primary adaptation strategies: on-site technological transformation or cross-regional operational relocation. For capital-intensive enterprises, the latter strategy is often more economically attractive given the high sunk costs associated with technological upgrading and their substantial baseline tax liabilities. We assume that the assignment of provincial tax rate changes is plausibly exogenous to individual firms’ investment decisions. While provincial governments strategically set tax rates based on local environmental carrying capacity, pollution status, and socio-economic development goals [34], individual firms are unlikely to influence these provincial-level policy decisions. Following the standard difference-in-differences framework, our identification relies primarily on the parallel trends assumption, which requires that, absent the tax rate adjustment, treated and control provinces would have followed similar trends in firms’ cross-regional investment. We empirically validate this assumption through event-study specifications in Section 4.2.1.

Based on the preceding theoretical analysis, interprovincial tax rate disparities created by the environmental tax reform are expected to generate a systematic “push” effect, incentivizing firms located in high-tax regions to expand their cross-regional investment activities. Accordingly, the following research hypothesis is proposed:

H1. 

(Main Effect): Environmental tax reform exerts a statistically significant positive effect on firms’ cross-regional investment behavior.

2.2.2. Cost-Compliance Theory and the Mediating Effect of Pollution Costs

Cost-compliance theory is grounded in the premise that environmental regulations require enterprises to allocate additional resources to meet specified environmental standards, which increases their pollution control expenditures and overall tax burdens. The theoretical logic linking environmental tax reform to cross-regional investment through the pollution cost channel requires explicit articulation, as this mechanism operates through a multi-stage transmission process [35].

The environmental protection tax reform affects corporate pollution costs through two primary avenues. The transition from the relatively flexible pollution discharge fee system to the statutory tax system fundamentally increases the rigidity of environmental cost imposition, while provincial autonomy in rate-setting creates substantial interprovincial variation in actual tax burdens. For firms located in high-tax provinces, the reform represents not merely a change in the legal form of environmental charges but a material increase in operational compliance costs.

A critical analytical question is why increased pollution costs would translate into cross-regional investment rather than on-site pollution abatement. The answer lies in the relative cost structure facing enterprises: for capital-intensive firms, the sunk costs of technological upgrading (including equipment retrofitting, research and development expenditures, and production disruption losses) often exceed the costs of establishing new operational facilities in lower-tax jurisdictions, making relocation a more flexible and less costly adaptation strategy.

The pollution cost channel operates through two complementary firm-level mechanisms. The first is the cost comparison effect: when an environmental tax is imposed in a firm’s home province, firms systematically compare these costs against those in alternative jurisdictions and rationally choose to relocate production capacity to locations where cost differentials are favorable. The second is the cost pass-through constraint effect: in competitive product markets where firms have limited pricing power, the ability to pass environmental tax costs to consumers through price increases is constrained. When cost pass-through is incomplete, profit margins are compressed, creating strong incentives for firms to pursue cost reductions through geographic reallocation of production activities [36].

Pollution cost pressure is proxied using a binary indicator based on whether a firm’s operating costs exceed the sample median, a measure that captures overall cost intensity rather than direct environmental expenditures (detailed measurement specifications are provided in Section 3.1.2, and limitations of this measure are discussed in Section 5.3).

Based on this theoretical reasoning, environmental tax reform is expected to increase pollution costs for affected firms, and this cost increase serves as a key transmission mechanism driving cross-regional investment. Accordingly, the following hypothesis is proposed:

H2. 

(Mediating Effect): Pollution costs play a mediating role in the impact of environmental tax reform on firms’ cross-regional investment; specifically, environmental tax reform promotes cross-regional investment by increasing firms’ pollution-related operational costs.

2.2.3. Information Asymmetry Theory and the Mediation Effect of Financing Constraints

Cross-regional investment requires substantial upfront capital expenditure that typically exceeds the internal financial resources of most firms; therefore, alleviation of financing constraints can enable firms to translate relocation incentives into actual investment behavior [37].

Environmental tax reform may alleviate financing constraints through an information transmission channel. The reform establishes formal information-sharing platforms that standardize the transmission of corporate environmental performance data to financial institutions, reducing banks’ information acquisition costs associated with assessing firms’ environmental risk profiles. This improved information transparency generates quantifiable environmental performance indicators, draws market attention to corporate environmental practices, and allows firms with strong environmental records to signal their creditworthiness to green investors. When financial institutions effectively process this standardized environmental information—as increasingly encouraged by banking regulatory authorities—the environmental risk premium for eligible firms may decline, and overall credit availability may improve [38].

It is critical to emphasize that the financing constraint mechanism does not produce uniformly effective across all firms [39]. For enterprises with strong environmental performance that qualify for tax incentives or exemptions, the financing benefits associated with increased information transparency may be substantial: their demonstrated environmental commitment serves as a positive signal that attracts preferential green credit allocation. Conversely, for high-polluting enterprises with poor environmental performance that face increased tax burdens, the establishment of standardized information channels may produce ambiguous effects: while information transparency improves, the revealed environmental risks may cause financial institutions to tighten rather than relax credit conditions. Therefore, the operation of the financing constraint pathway is significantly contingent on a firm’s environmental performance and its associated risk profile as assessed by financial institutions.

Financing constraints are measured using a binary median-split WW index, following the empirical specification employed; alternative specifications using the continuous WW index and the SA index are reported in the robustness analysis (detailed measurement specifications are provided in Section 3.1.3).

Based on the preceding theoretical reasoning, the following third hypothesis is proposed:

H3. 

(Mediating Effect): Financing constraints mediate the impact of environmental tax reform on firms’ cross-regional investment; specifically, environmental tax reform promotes cross-regional investment by alleviating firms’ external financing constraints.

3. Research Design

All statistical analyses in this study are performed using Stata 18.0, with standard errors clustered at the provincial level to address potential heteroskedasticity and intra-group serial correlation.

3.1. Model Specification

3.1.1. Baseline Regression Model

This study takes the implementation of China’s Environmental Protection Tax Law as a quasi-natural experiment and adopts the difference-in-differences (DID) approach to quantitatively evaluate the causal effect of this policy reform on firms’ cross-regional investment decisions. The baseline econometric specification is constructed as follows:

$${\mathit{Invest}}_{i, t} ={ \beta }_0 +{ \beta }_1\mathit{Treat} \times \mathit{Post} +{ \beta }_2{\mathit{Controls}}_{i, t} + {\mu }_i + {\lambda }_t +{ \epsilon }_{i, t}$$

(1)

where i and t represent year and firm, respectively; the core explanatory variable is the interaction term Treat × Post; the dependent variable is ${\mathrm{Invest}}_{\mathrm{i}, \mathrm{t}}$; ${\mathrm{Controls}}_{\mathrm{i}, \mathrm{t}}$ is the set of control variables; ${\mathsf{\beta }}_1$ are the coefficients of each term; ${\mathsf{\mu }}_{\mathrm{i}}$ and ${\mathsf{\lambda }}_{\mathrm{t}}$ represent firm and year fixed effects, respectively; and ${\mathsf{\epsilon }}_{\mathrm{i}, \mathrm{t}}$ is the random disturbance term. Standard errors are clustered by province to account for heteroskedasticity and within-province serial correlation.

3.1.2. Model of the Mediation Effect of Pollution Costs

To identify the underlying transmission channel through which environmental tax reform affects corporate cross-regional investment, this study first tests the mediating role of pollution cost pressure. Drawing on the analytical frameworks of prior studies, we construct a three-way difference-in-differences (DDD) model to examine whether the policy effect varies systematically with the pollution cost burden faced by firms. The specific model specification is

$${Cost}_{i,t}={\alpha }_0+{\alpha }_{1}Treat\times Post\times Cost+{\alpha }_2{Controls}_{i,t}+{\mu }_i+{\lambda }_t+{\epsilon }_{i,t}$$

(2)

${Cost}_{i,t}$ represents a firm’s pollution cost pressure, defined as the composite cost burden arising from environmental regulatory compliance, including direct tax expenditures, pollution treatment costs, and associated adaptation investments. Under the environmental tax reform, regional variations in tax rates create heterogeneous compliance cost pressures: firms in high-tax provinces face steeper cost increases, while those in low-tax provinces experience milder pressure. When the cost differential between the home province and potential destination provinces becomes sufficiently large, firms have an incentive to relocate investments to regions with lower environmental compliance costs.

In terms of variable measurement, our main indicator for $\mathrm{cost}$ is a binary dummy variable, which takes the value of 1 if a firm’s operating cost is above the annual industry median, and 0 otherwise. This measurement is based on the rationale that the environmental tax reform imposes heavier compliance burdens on pollution-intensive enterprises, which will be reflected in their relatively higher operating costs compared to industry peers. This median-split approach effectively distinguishes between firms facing high and low pollution cost pressure and covers the full sample without relying on specialized environmental disclosure data. Considering that operating costs also include non-environmental components such as labor, raw materials, and logistics expenses, which may introduce measurement bias, we adopt two alternative proxy variables for robustness checks. The first alternative is the ratio of a firm’s actual environmental protection investment to its total assets, with data sourced from the environmental investment disclosure module of the CSMAR database. This indicator directly captures the resources allocated by firms to environmental compliance and is more conceptually aligned with the definition of pollution cost pressure. The second alternative is emission intensity, measured as the ratio of SO₂ emissions to operating revenue, with data obtained from the CNRDS China Environmental Database. This output-based indicator reflects the inherent pollution intensity that drives the difference in compliance costs across firms. We expect consistent estimation results across all three measurement specifications if the pollution cost mechanism is valid. It should be noted that limited by the availability of public corporate environmental data in China, none of the above indicators can perfectly capture the theoretical construct of pollution cost pressure caused by the environmental tax reform; therefore, we report results for all specifications to ensure the robustness of mechanism inference.

3.1.3. Model of the Mediation Effect of Financing Constraints

To further test the second transmission channel, we construct another DDD model to examine whether the policy effect is heterogeneous across firms with different levels of financing constraints [40,41].

$${WW}_{i,t}={\gamma }_0+{\gamma }_{1}Treat\times Post\times WW+{\gamma }_2{Controls}_{i,t}+{\mu }_i+{\lambda }_t+{\epsilon }_{i,t}$$

(3)

This model aims to identify how financing constraints moderate the impact of environmental tax reform on cross-regional investment. The core variable ${\mathrm{WW}}_{\mathrm{i},\mathrm{t}}$ measures the degree of financing constraints faced by firms in year, which reflects the difficulties firms encounter when accessing external funds, including high borrowing costs, limited credit channels, and insufficient loan availability. The theoretical basis of this mechanism is that the environmental tax reform has promoted the establishment of cross-departmental environmental information-sharing platforms and standardized corporate environmental disclosure requirements, which may reduce the information asymmetry between banks and enterprises, lower the environmental risk premium for high-quality firms, and expand credit access for enterprises with better environmental performance [42].

Our primary measure for FC is a binary dummy variable based on the WW index, which takes the value of 1 if a firm’s WW index is above the sample median, and 0 otherwise [43,44]. The WW index is a widely used composite measure of financing constraints that incorporates multiple firm characteristics, including size, profitability, dividend policy, leverage, and industry growth, capturing the multidimensional nature of firms’ external capital access. The binary conversion is adopted for two reasons: first, it facilitates clear identification of heterogeneous treatment effects between firms with high and low financing constraints in the DDD framework, where the coefficient of the triple interaction term Treat × Post × FC reflects whether firms facing more severe financing constraints exhibit a larger investment response to the reform; second, this approach is consistent with the operational practice in recent related studies. To address the potential information loss caused by converting a continuous index into a binary indicator, we employ two alternative measurement specifications in robustness checks. The first alternative retains the continuous WW index in the triple interaction model, which preserves the full information of the original index and interprets the interaction coefficient as the change in policy effect per unit increase in financing constraint level. The second alternative uses the SA index constructed as SA = −0.737 × Size + 0.043 × Size² − 0.04 × Age, where Size is the natural logarithm of total assets, and Age is the number of years since the firm’s establishment. A key advantage of the SA index is that it is calculated solely using two relatively exogenous firm characteristics (size and age), which are unlikely to be affected by the environmental tax reform, thus avoiding potential endogeneity issues associated with the WW index, which incorporates outcome variables such as profitability and leverage. We expect consistent directional results across all three specifications if the financing constraint mechanism holds, despite possible variations in coefficient magnitude and statistical significance [45,46,47].

3.2. Variable Selection

3.2.1. Cross-Regional Investment

The dependent variable in this study is firm cross-regional investment (Invest). Following the approach in existing literature, this variable is measured by counting the number of domestic subsidiaries established by listed companies outside their province of registration.

The dependent variable is measured as the number of domestic subsidiaries established outside the province of registration. This measure captures the breadth of firms’ geographic expansion and aligns precisely with the provincial-level policy variation. We acknowledge that the subsidiary count reflects geographic breadth rather than investment depth; our robustness analysis therefore employs three alternative specifications: LN_Invest (log of the subsidiary count plus one), Invest_ratio (share of out-of-province subsidiaries), and Invest_capital (log of the total registered capital of out-of-province subsidiaries).

3.2.2. Environmental Protection Tax Reform

The primary explanatory variable is the environmental tax reform interaction (Treat × Post). After the Environmental Protection Tax Law came into effect in 2018, provincial governments were authorized to set specific tax rates within the statutory range, leading to heterogeneous rate adjustments across regions. Specifically, 12 provincial administrative regions maintained the original pollution discharge fee standards unchanged under the new environmental tax framework, while 19 provincial administrative regions raised tax rates above the original fee levels. Based on this policy variation and following the grouping method of prior studies, we construct the Treat dummy variable, which takes the value of 1 if a firm is registered in a province that raised environmental tax rates (treatment group), and 0 if it is registered in a province that kept rates unchanged (control group). The Post dummy variable is set to 1 for observations from 2018 onwards (the post-reform period) and 0 for observations before 2018 (the pre-reform period).

This variable design leverages the strong exogeneity of provincial tax rate decisions, the clear policy implementation time point, and the substantial cross-regional variation to identify the causal effect of the reform. The validity of our DID identification strategy relies on two key assumptions. The first is the parallel trends assumption, which requires that cross-regional investment trends of treatment and control group firms would have been parallel in the absence of the reform. We formally test this assumption using event-study methodology in Section 4.2.1, and find no evidence of pre-treatment divergence between the two groups. The second is the conditional independence assumption, which requires that treatment assignment (operating in a province with raised tax rates) is independent of potential outcomes after controlling for covariates and fixed effects. This assumption is plausibly satisfied for three reasons: first, provincial tax rate decisions were made by provincial people’s congresses in December 2017 based on macro-level factors, including regional environmental carrying capacity, pollution status, and socio-economic development goals, rather than the investment decisions of individual firms, ensuring the exogeneity of treatment assignment; second, the policy was implemented nationwide on 1 January 2018 simultaneously, eliminating potential biases caused by staggered adoption timing correlated with local economic conditions; third, tax rate setting was a top-down political process involving multiple provincial departments, leaving limited room for individual firms to influence the final outcome. We further address potential threats to identification through a series of robustness tests, including PSM-DID matching, placebo tests, exclusion of anticipation effects, and alternative sample compositions, all of which confirm the stability of baseline results [48].

3.2.3. Control Variables

To mitigate the impact of confounding firm characteristics on cross-regional investment decisions, we include a set of control variables covering firms’ basic attributes, financial conditions, and governance structures. Specifically, the control variables include firm age (Age), firm size (Size), debt-to-asset ratio (Lev), return on assets (Roa), shareholding ratio of the largest shareholder (Top1), operating revenue growth rate (Growth), board size (Board), current ratio (Liq), return on equity (ROE), CEO-chairman duality (Plu), proportion of independent directors (Indratio), net operating cash flow (Cash), management expense ratio (Mfee), separation of ownership and control rights (Separa), and proportion of institutional investors (Inst_ratio). As noted in the model specification section, all regressions control for firm fixed effects and year fixed effects to absorb unobserved time-invariant firm heterogeneity and common time shocks. Definitions of all key variables are summarized in

Table 1. Table of Key Variable Definitions.

Type	Variable Name	Symbol	Variable Description
Dependent Variable	Company Cross-Regional Investment	Invest	The number of domestic subsidiaries established by a listed company in provinces other than the one where its parent company is located
Explanatory Variables	Dummy variable for environmental tax reform	Trea × Post	Treat is a dummy variable. It is set to 1 when the tax rate in the parent company’s province increases, and 0 otherwise. Post is a dummy variable for the implementation of the Environmental Protection Tax Law. It is set to 1 when the sample year is in or after the policy implementation year, and 0 otherwise.
Intermediary variables	Financing constraints	WW	Companies with a WW index greater than the median are assigned a value of 1, otherwise 0
Control variables	Company Age	Age	Current year − Year of establishment + 1
	Company Size	Size	Logarithm of total assets at year-end
	Debt-to-Equity Ratio	Lev	Total Liabilities at Year-End/Total Assets at Year-End
	Profitability	ROA	Net Income/Average Total Assets
	Company Growth	Growth	(Current-period revenue − Prior-year revenue)/(Prior-year revenue)
	Board Size	Board	Natural logarithm of the total number of board members
	Proportion of Independent Directors	Indratio	Proportion of Independent Directors among Board Members
	Shareholding Concentration	Top 1	Shareholding Ratio of the Largest Shareholder
	Company Liquidity	Liq	Total Current Assets/Total Assets
	Return on Equity	ROE	Net Income/Average Shareholders’ Equity
	Dual Role	Plu	Are the Chairman and the General Manager the same person?
	Cash Flow	Cash	Measured as the ratio of cash flow from operating activities to total assets
	Management Expense Ratio	mfee	Ratio of administrative expenses to operating revenue
	Separation ratio	Separate	Difference between control and ownership
	Proportion of institutional investors	Insti_ratio	Total number of shares held by institutional investors as a percentage of total shares of the listed company
Fixed effects	Company	Id	Control company fixed effect
	Year	Year	Control for Year Fixed Effects

.

3.3. Data Sources and Sample Selection

This study takes all A-share listed companies in China from 2012 to 2024 as the initial research sample. The data used in this study are collected from multiple sources: first, the geographic distribution data of subsidiaries are manually compiled from the notes of “long-term equity investments” in the annual reports of listed parent companies, following the method of prior studies, and a subsidiary is defined as a cross-regional investment if its registration province is different from that of the parent company; second, firm-level financial and governance data are obtained from the CSMAR database; third, provincial environmental tax rate data are manually collected from official policy documents and public announcements of provincial development and reform commissions and ecological environment departments; fourth, provincial-level macroeconomic data are sourced from the official website of the National Bureau of Statistics of China and provincial statistical yearbooks.

To ensure data reliability and the validity of empirical results, we filter the initial sample according to the following criteria: (1) exclude firms that were under special treatment (ST or *ST) during the sample period; (2) exclude observations with missing values for key variables; (3) exclude listed companies in the financial and insurance industries due to their unique accounting standards; (4) exclude firms whose registration place or main business location changed across provinces during the sample period to avoid the interference of relocation decisions; (5) winsorize all continuous variables at the 1st and 99th percentiles to eliminate the impact of extreme values. After the above screening process, we obtain a final sample of 40,965 firm-year observations for empirical analysis.

Table 2. Descriptive Statistics of Key Variables.

VarName	Obs	Mean	SD	Min	Median	Max
Invest	40,965	9.3773	16.333	0.00	4.00	107.00
Treat	40,965	0.7373	0.440	0.00	1.00	1.00
Post	40,965	0.6229	0.485	0.00	1.00	1.00
Age	40,965	20.7280	6.246	8.00	20.00	38.00
Size	40,965	22.2680	1.309	19.90	22.06	26.34
Lev	40,965	0.4186	0.206	0.05	0.41	0.91
ROA	40,965	0.0336	0.066	−0.27	0.04	0.20
Growth	40,965	0.1344	0.360	−0.57	0.08	2.06
Board	40,965	2.1107	0.197	1.61	2.20	2.64
Indratio	40,965	0.3780	0.054	0.33	0.36	0.57
Top 1	40,965	0.3376	0.149	0.08	0.31	0.75
Liq	40,965	0.5747	0.203	0.10	0.59	0.95
ROE	40,965	0.0549	0.146	−0.75	0.07	0.36
Plu	40,965	0.2993	0.458	0.00	0.00	1.00
Cash	40,965	0.0477	0.068	−0.16	0.05	0.24
mfee	40,965	0.0861	0.070	0.01	0.07	0.43
Separa	40,965	5.3538	7.622	0.00	0.45	29.36
Insti_ratio	40,965	42.6782	24.729	0.37	43.80	91.65

reports the descriptive statistics of key variables. The dependent variable, Invest, has a mean value of 9.3773, indicating that sample firms established an average of approximately 9 out-of-province subsidiaries during the observation period. Its standard deviation is 16.333 with a coefficient of variation of 1.74, reflecting substantial heterogeneity in cross-regional investment scales across firms. The minimum value of 0 and the maximum value of 107 further confirm that some firms do not conduct any cross-regional investment, while some large enterprises have realized a nationwide strategic layout through extensive subsidiary establishment. The treatment group dummy Treat has a mean of 0.7373, meaning that 73.73% of sample firms are located in provinces that raised environmental tax rates, providing a sufficiently large treatment group for policy effect identification. The policy period dummy Post has a mean of 0.6229, indicating that 62.29% of observations are from the post-reform period, with relatively balanced distribution of observations before and after the policy implementation to support dynamic effect analysis. The statistics of all control variables are within reasonable ranges: firm age covers different development stages, the average debt-to-asset ratio is 0.4186 with a maximum of 0.91 within the normal range for listed companies, and the average return on assets is 3.36%, consistent with the overall profitability profile of Chinese listed firms. No abnormal distributions are observed for other governance and financial indicators, ensuring the reliability of subsequent regression analyses.

4. Empirical Results and Analysis

4.1. Baseline Regression

Table 3. Baseline Regression Results.

	(1)	(2)	(3)
	Invest	Invest	Invest
Treat × Post	3.315 ***	0.781 ***	0.822 ***
	(20.58)	(3.62)	(3.93)
Age			−0.150
			(−0.60)
Size			6.535 ***
			(33.47)
Lev			−1.030 **
			(−1.99)
Roa			−0.692
			(−0.35)
Growth			−0.561 ***
			(−4.12)
Board			0.113
			(0.19)
Indratio			−1.663
			(−0.90)
Top1			−3.687 ***
			(−3.62)
Liq			−2.093 ***
			(−3.95)
ROE			−2.423 ***
			(−2.80)
Plu			0.095
			(0.65)
Cash			1.805 **
			(2.25)
mfee			9.927 ***
			(8.57)
Separa			−0.015
			(−1.18)
Insti_ratio			0.009
			(1.25)
_cons	7.854 ***	9.037 ***	−131.238 ***
	(71.92)	(83.78)	(−18.43)
Firm FE	No	Yes	Yes
Year FE	No	Yes	Yes
N	40,965	40,965	40,965
Adj. R2	0.01	0.75	0.77

Note: All variables are defined in Table 1. **, and *** denote significance levels of 5%, and 1%, respectively; the numbers in parentheses are t-values, and standard errors are clustered at the provincial level.

reports difference-in-differences (DID) estimates capturing the causal effect of the environmental protection tax reform on firms’ cross-regional investment activities. Column (1) presents unadjusted estimates without control variables; column (2) incorporates firm- and year-fixed effects to account for time-invariant firm-specific heterogeneity and common year-level shocks; column (3) adds the full set of firm-level and provincial-level covariates specified in the empirical model. Across all three specifications, the coefficient of the core interaction term Treat × Post remains positive and statistically significant at the 1% level, providing direct empirical support for Hypothesis 1 that the reform exerts a positive effect on firms’ cross-regional investment.

Beyond statistical significance, the economic magnitude of the estimate is critical for evaluating the reform’s practical impact. The coefficient of 0.822 in the fully specified model (column 3) indicates that, on average, firms located in provinces that raised environmental tax rates after the reform established 0.82 additional out-of-province subsidiaries relative to firms in provinces that maintained pre-reform tax rate levels. To contextualize this effect: the sample mean of the dependent variable, Invest, is 9.38 subsidiaries, so the 0.82 increase corresponds to an 8.7% rise in cross-regional investment activity. Relative to the sample median of 4 subsidiaries, the effect translates to a 20.6% increase, demonstrating that the reform’s impact is particularly pronounced for firms at the median level of geographic expansion. This 8.7% estimated effect falls within the 5–10 percentage point range of increased relocation probability for polluting enterprises due to interprovincial environmental tax differentials reported in prior work, confirming the external validity of the baseline findings. Combined, these figures indicate a moderate-to-large economic effect, confirming that the environmental protection tax reform has materially altered the spatial allocation of corporate investment across Chinese provinces.

4.2. Robustness Tests

4.2.1. Parallel Trends Test

The validity of the DID identification strategy rests on the parallel trends assumption, which requires that cross-regional investment trajectories of treated and control firms follow statistically indistinguishable patterns prior to policy implementation. This study employs an event study framework to test this assumption, constructing interaction terms between treatment group status and year dummies to capture temporal differences in outcome variables between the two groups. To avoid multicollinearity with the full set of year fixed effects, the period immediately preceding the reform (2017) is set as the baseline reference period and omitted from the regression, so all estimated coefficients represent effects relative to this pre-policy benchmark.

Figure 1 presents the event-study estimates of the dynamic treatment effects. Estimation results confirm that all interaction coefficients for periods prior to 2018 are statistically insignificant at conventional levels, with 95% confidence intervals consistently containing zero. This finding verifies that no systematic pre-trend differences exist between the treatment and control groups, satisfying the parallel trends assumption and validating the causal interpretation of the baseline DID estimates.

4.2.2. PSM-DID

To address potential concerns that systematic observable differences between treated and untreated firms prior to the reform may bias the baseline estimates, this study adopts the propensity score matching (PSM) DID (PSM-DID) framework as a robustness check. Given heterogeneous treatment effects in the multi-period DID setting, the panel is converted to cross-sectional observations for matching, with all baseline covariates included in the propensity score estimation. Three matching algorithms—1:1 nearest neighbor matching, radius matching, and kernel matching—are applied to construct matched samples, with regressions re-estimated on each matched subsample.

As reported in

Table 4. PSM-DID Regression Results.

	1:1 Nearest Neighbor Matching		Radius Matching		Kernel Matching
	(1)	(2)	(3)	(4)	(5)	(6)
	Invest	Invest	Invest	Invest	Invest	Invest
Treat × Post	4.304 ***	2.026 ***	5.298 ***	2.504 ***	5.392 ***	2.482 ***
	(12.91)	(6.21)	(21.53)	(10.52)	(21.90)	(10.29)
Control	NO	YES	NO	YES	NO	YES
_cons	8.71 ***	−119.34 ***	9.25 ***	−118.86 ***	9.18 ***	−120.57 ***
	(51.20)	(−30.71)	(73.40)	(−43.54)	(72.77)	(−43.81)
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
N	12,533	12,533	24,503	24,503	24,723	24,723
Adj. R2	0.01	0.17	0.02	0.20	0.02	0.20

Note: All variables are defined in Table 1. *** denote significance levels of 1%, respectively; the numbers in parentheses are t-values, and standard errors are clustered at the provincial level.

, the Treat × Post coefficient remains positive and statistically significant at the 1% level across all three matching specifications [49]. This consistency confirms that the baseline results are not driven by observable pre-treatment differences between the treatment and control groups.

4.2.3. Placebo Test

A placebo test with random treatment assignment is conducted to rule out the possibility that the baseline results are driven by unobserved random shocks or omitted variables. The test procedure holds the official reform implementation timeline constant but randomly assigns hypothetical environmental tax rate adjustments to provinces to generate 500 pseudo-treatment groups. Figure 2 plots the distribution of the 500 placebo coefficients.

The resulting distribution of 500 placebo coefficients approximates a standard normal distribution with a mean value close to zero, and the actual baseline estimate of 0.822 lies in the far right tail of this distribution. This result confirms that the observed effect of the reform is unlikely to be a product of random chance, further validating the robustness of the core conclusions.

4.2.4. Controlling for Concurrent Policy Interference

Two potential policy shocks overlapping with the environmental protection tax reform are addressed to isolate the net policy effect. First, the 2014 Notice on Adjusting Pollution Discharge Fees mandated all provinces to raise the sulfur dioxide discharge fee standard to no less than 1.26 yuan per pollution equivalent by June 2015, with 15 provinces completing the adjustment ahead of schedule in early 2015 and the remainder implementing the change by the June 2015 deadline. The staggered implementation of this policy may confound estimates of the 2018 reform effect, so all 2015 observations are excluded from the sample, and the model is re-estimated. Results reported in columns (1) and (2) of

Table 5. Excluding Concurrent Policy Interference.

	“Notice on Adjusting Pollution Discharge Fees”		“Water Pollution Prevention and Control Action Plan”
	(1)	(2)	(3)	(4)
	Invest	Invest	Invest	Invest
Treat × Post	0.803 ***	0.811 ***	1.445 ***	1.844 ***
	(3.46)	(3.62)	(2.96)	(5.00)
Control	No	Yes	No	Yes
_cons	9.124 ***	−134.274 ***	10.681 ***	−131.093 ***
	(74.41)	(−17.79)	(34.32)	(−12.99)
Firm FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
N	36,941	36,941	30,163	30,163
Adj. R2	0.76	0.78	0.76	0.76

Note: All variables are defined in Table 1. *** denote significance levels of 1%, respectively; the numbers in parentheses are t-values, and standard errors are clustered at the provincial level.

show the Treat × Post coefficient remains 0.811 and significant at the 1% level, indicating no material change to the core conclusion after accounting for this prior policy shock.

Second, the 2015 Action Plan for Water Pollution Prevention and Control (the “Water Ten Measures”) introduced overlapping environmental governance requirements that may affect firms’ investment decisions, with effects potentially extending into the reform implementation period. To isolate the reform’s net effect from this confounding policy, all observations from 2015 and earlier are excluded, restricting the sample to the 2016–2020 period. Re-estimation results reported in columns (3) and (4) of Table 5 show the Treat × Post coefficient is 1.844 and remains significant at the 1% level, confirming the positive effect of the reform persists after adjusting for this overlapping policy shock.

4.2.5. Other Robustness Tests

Two additional robustness checks are conducted to address sample composition and measurement sensitivity. First, the sample scope is adjusted in two separate specifications.

Table 6. Adjusted Sample Scope.

	Excluding Policy Provinces		Excluding Municipalities
	(1)	(2)	(3)	(4)
	Invest	Invest	Invest	Invest
Treat × Post	0.927 ***	0.892 ***	0.646 ***	0.650 ***
	(4.12)	(4.09)	(2.98)	(3.12)
Control	No	Yes	No	Yes
_cons	9.053 ***	−134.307 ***	8.336 ***	−137.443 ***
	(84.27)	(−17.50)	(82.44)	(−21.65)
Firm FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
N	33,355	33,355	31,656	31,656
Adj. R2	0.75	0.78	0.74	0.77

Note: All variables are defined in Table 1. *** denote significance levels of 1%, respectively; the numbers in parentheses are t-values, and standard errors are clustered at the provincial level.

reports the results of the sample adjustment tests. In the first adjustment, observations from Hebei, Shanghai, and Shandong provinces are excluded. These three provinces raised pollution discharge fee standards in 2017, one year prior to the formal implementation of the environmental protection tax law, potentially creating policy anticipation effects that could bias treatment group assignment. Re-estimation on the remaining sample yields a Treat × Post coefficient of 0.892, significant at the 1% level, consistent with baseline results. In the second adjustment, observations from the four centrally administered municipalities (Beijing, Tianjin, Shanghai, and Chongqing) are excluded, as these jurisdictions exhibit systematic differences from other provinces in policy enforcement, economic structure, and regulatory intensity. The re-estimated coefficient is 0.650, remaining positive and significant at the 1% level, confirming that the results are not driven by the unique characteristics of these municipalities.

Second,

Table 7. Alternative Measures of the Dependent Variable.

	(1)	(2)	(3)	(4)	(5)	(6)
	LN_Invest	LN_Invest	Inv_ratio	Inv_ratio	Inv_cap	Inv_cap
Treat × Post	0.058 ***	0.054 ***	0.033 ***	0.022 ***	0.224 *	0.269 *
	(3.87)	(3.80)	(7.65)	(5.28)	(1.69)	(1.84)
Control	No	Yes	No	Yes	No	Yes
_cons		0.001 ***		−0.000 ***		0.014 ***
		(2.81)		(−2.93)		(3.44)
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
N	38,859	38,859	38,859	38,859	38,859	38,859
Adj. R2	0.79	0.82	0.00	0.06	0.13	0.72

Note: All variables are defined in Table 1. *, and *** denote significance levels of 10%, and 1%, respectively; the numbers in parentheses are t-values, and standard errors are clustered at the provincial level.

presents the results using alternative measures of the dependent variable. The first proxy is the natural logarithm of the number of out-of-province subsidiaries plus 1 (LN_Invest), capturing the absolute scale of cross-regional investment. The second proxy is the share of out-of-province subsidiaries in a firm’s total subsidiary count (Invest_ratio), measuring the relative importance of cross-regional investment in the firm’s overall portfolio. The third proxy is the natural logarithm of the total registered capital of out-of-province subsidiaries plus 1 (Invest_capital), capturing the financial scale of cross-regional investment activities. Regressions using these three alternative dependent variables all yield positive and statistically significant coefficients on Treat × Post: 0.054 (significant at 1%) for LN_Invest, 0.022 (significant at 1%) for Invest_ratio, and 0.269 (significant at 10%) for Invest_capital. These consistent results across multiple measurement approaches confirm the core conclusion is not sensitive to the specific operationalization of cross-regional investment.

4.3. Mechanism Test

Table 8. Results of Mechanism Tests.

	Increased Pollution Costs		Easing of Financing Constraints
	(1)	(2)	(3)	(4)
	Invest	Invest	Invest	Invest
Treat × Post × Cost	4.395 ***	0.645 ***
	(33.78)	(4.22)
Treat × Post × WW			−2.019 ***	−0.537 ***
			(−16.55)	(−4.43)
Control	No	Yes	No	Yes
_cons	8.398 ***	−131.401 ***	9.935 ***	−128.265 ***
	(160.13)	(−18.49)	(207.54)	(−17.60)
Firm FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
N	40,965	40,965	40,965	40,965
Adj. R2	0.73	0.77	0.75	0.77

Note: All variables are defined in Table 1. *** denote significance levels of 1%, respectively; the numbers in parentheses are t-values, and standard errors are clustered at the provincial level.

reports the results of the mechanism tests using triple interaction specifications. The coefficient on the triple interaction term Treat × Post × Cost in column (2) is 0.645, statistically significant at the 1% level. This result indicates that firms facing higher pollution cost pressure exhibit a stronger cross-regional investment response to the reform, supporting Hypothesis 2. In economic terms, this coefficient implies that the reform’s effect on cross-regional investment is 0.645 subsidiaries larger for firms with above-median pollution costs than for firms with below-median pollution costs, corresponding to a 6.9% additional increase relative to the sample mean of Invest. This magnitude confirms that pollution cost pressure is not only statistically significant but also economically meaningful as a transmission channel for the reform’s effect.

The coefficient on the triple interaction term Treat × Post × WW in column (4) is −0.537, statistically significant at the 1% level. Since higher WW index values indicate more severe financing constraints, this negative coefficient implies that firms with looser pre-reform financing constraints exhibit a stronger cross-regional investment response to the reform, supporting Hypothesis 3. Economically, this differential effect corresponds to a 5.7% additional increase in cross-regional investment for firms with below-median financing constraints relative to their more constrained counterparts, reflecting the critical role of financing capacity in enabling firms to act on cross-regional relocation incentives.

It is important to note that the financing constraint mechanism operates conditionally. The binary WW index specification captures the average effect across heterogeneous firm subgroups but cannot fully disentangle countervailing effects: firms with strong environmental performance may gain improved access to green credit following the reform, while high-polluting firms may face tightened credit conditions despite enhanced information transparency.

4.4. Heterogeneity Test

This section analyzes heterogeneous effects of the reform across three dimensions identified in the theoretical framework: firm-level capital intensity, local fiscal pressure, and local environmental regulatory intensity. All heterogeneity tests use median splits of the relevant moderating variables for subsample regression.

4.4.1. Firm Level

Following prior work, capital intensity is measured as the ratio of fixed assets to the number of employees, with firms split into above-median (high) and below-median (low) capital intensity subsamples. Regression results reported in

Table 9. Heterogeneity in Capital Intensity.

	High Capital Intensity		Low Capital Intensity
	(1)	(2)	(3)	(4)
	Invest	Invest	Invest	Invest
Treat × Post	2.954 ***	1.897 ***	−0.286	−0.252
	(9.49)	(6.67)	(−0.54)	(−0.54)
Control	No	Yes	No	Yes
_cons	4.595 ***	−121.137 ***	10.228 ***	−126.289 ***
	(8.29)	(−42.11)	(18.65)	(−42.20)
Firm FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Difference	(1)–(3) = 1.391 **		(2)–(4) = 0.779 *
	p = 0.003		p = 0.064
N	19,789	19,789	19,793	19,793
Adj. R2	0.02	0.21	0.03	0.26

Note: All variables are defined in Table 1. *, **, and *** denote significance levels of 10%, 5%, and 1%, respectively; the numbers in parentheses are t-values, and standard errors are clustered at the provincial level.

show that the Treat × Post coefficient is 1.897 and significant at the 1% level for the high capital intensity subsample, but statistically indistinguishable from zero for the low capital intensity subsample. A between-group coefficient difference test confirms this heterogeneity is statistically significant at the 10% level. This pattern arises because capital-intensive firms have higher pollution emission intensities and face larger increases in tax burdens under the reform, while the high sunk costs of on-site emission reduction technology upgrades make cross-regional relocation a more cost-effective adaptation strategy. In contrast, low-capital-intensive firms face smaller tax shocks and greater operational flexibility, allowing them to adjust to the reform through internal optimization rather than geographic relocation. The 1.897 coefficient for high-capital-intensity firms corresponds to a 20.2% increase in cross-regional investment relative to the sample mean, highlighting that capital-intensive firms are the primary drivers of the spatial capital reallocation induced by the reform.

4.4.2. Government Level

Following existing research, local fiscal pressure is measured as the ratio of (fiscal expenditure minus fiscal revenue) to fiscal revenue, with provinces split into above-median (high) and below-median (low) fiscal pressure subsamples. Results reported in

Table 10. Heterogeneity in Local Fiscal Pressure.

	High Local Fiscal Pressure		Low Local Fiscal Pressure
	(1)	(2)	(3)	(4)
	Invest	Invest	Invest	Invest
Treat × Post	0.333	0.373	3.498 ***	2.932 ***
	(0.55)	(0.62)	(7.82)	(6.45)
Control	No	Yes	No	Yes
_cons	44.851 ***	43.021 ***	27.574 ***	−17.713 ***
	(21.38)	(7.94)	(30.67)	(−5.70)
Firm FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Difference	(1)–(3) = −2.968 ***		(2)–(4) = −1.81 **
	p = 0.000		p = 0.025
N	7849	7849	28,606	28,606
Adj. R2	0.07	0.13	0.11	0.13

Note: All variables are defined in Table 1. **, and *** denote significance levels of 5%, and 1%, respectively; the numbers in parentheses are t-values, and standard errors are clustered at the provincial level.

show that the Treat × Post coefficient is 2.932 and significant at the 1% level for the low fiscal pressure subsample, but 0.373 and statistically insignificant for the high fiscal pressure subsample. The between-group difference is statistically significant at the 5% level. This pattern reflects that local governments with low fiscal pressure have greater administrative capacity and fiscal flexibility to enforce the environmental protection tax strictly, leading firms to face the full statutory tax burden and stronger relocation incentives. In contrast, governments facing high fiscal pressure may adopt lenient enforcement practices to protect local tax bases and employment, effectively neutralizing the reform’s cost pressure on local firms. The 2.932 coefficient for low fiscal pressure regions corresponds to a 31.2% increase in cross-regional investment relative to the sample mean, nearly four times the size of the baseline average effect.

4.4.3. Regulatory Aspects

Following prior methodology, local environmental regulatory intensity is measured as the proportion of environment-related terms in annual provincial government work reports, with provinces split into above-median (high) and below-median (low) regulatory intensity subsamples. Results reported in

Table 11. Heterogeneity in Environmental Regulatory Intensity.

	High Environmental Regulatory Intensity		Low Environmental Regulatory Intensity
	(1)	(2)	(3)	(4)
	Invest	Invest	Invest	Invest
Treat × Post	3.594 ***	2.012 ***	1.349	0.704
	(6.65)	(3.86)	(0.72)	(0.39)
Control	No	Yes	No	Yes
_cons		1.353		4.838
		(0.42)		(1.41)
Firm fixed effects	Yes	Yes	Yes	Yes
Year fixed effects	Yes	Yes	Yes	Yes
Difference	(1)–(3) = 1.761 **		(2)–(4) = 1.889 **
	p = 0.036		p = 0.028
N	15,369	15,369	14,799	14,799
Adj. R2	0.00	0.13	0.13	0.23

Note: All variables are defined in Table 1. **, and *** denote significance levels of 5%, and 1%, respectively; the numbers in parentheses are t-values, and standard errors are clustered at the provincial level.

show that the Treat × Post coefficient is 2.012 and significant at the 1% level for the high regulatory intensity subsample, but 0.704 and statistically insignificant for the low regulatory intensity subsample. The between-group difference is statistically significant at the 5% level. This pattern demonstrates that the reform’s effect is amplified in regions with strict complementary environmental enforcement, where the tax operates as part of a comprehensive regulatory regime that creates strong compliance pressure. In regions with weak enforcement, firms can absorb or circumvent the tax burden without relocating. The 2.012 coefficient for high regulatory intensity regions corresponds to a 21.4% increase in cross-regional investment relative to the sample mean, approximately 2.4 times the baseline average effect.

5. Conclusions and Implications

5.1. Main Conclusions

Taking the implementation of the 2018 Environmental Protection Tax Law as an exogenous policy shock, this study empirically investigates the influence of this regulatory reform on firms’ cross-regional investment decisions and the underlying transmission mechanisms, based on 40,965 firm-year observations of China’s A-share-listed enterprises spanning the period from 2012 to 2024. The core findings of the research provide systematic answers to the three core questions raised at the outset of the study, as outlined below.

First, the environmental protection tax reform has exerted a significant positive effect on the cross-regional investment scale of firms located in provinces where applicable tax rates were raised following the reform. Quantitatively, the average treatment effect corresponds to an 8.7% increase in cross-regional investment relative to the sample mean. This core conclusion remains robust across a series of validity tests, including event-study analysis to verify the parallel trend assumption, placebo tests with randomly assigned treatment groups, propensity score matching combined with difference-in-differences (PSM-DID) estimation to address selection bias, control for the confounding effects of concurrent policy initiatives implemented during the sample period, adjustments to the sample inclusion criteria to exclude potentially anomalous observations, and substitution of the dependent variable with alternative measures of cross-regional investment.

Second, mechanism analysis identifies two distinct and parallel transmission pathways through which the reform affects cross-regional investment. The first is the pollution cost channel: the rise in environmental tax rates directly increases firms’ environmental compliance costs in their original locations, creating strong incentives for firms to relocate part of their production capacity to jurisdictions with lower environmental tax burdens to reduce operating expenditures. The second is the financing constraint channel: the standardized environmental tax collection and information disclosure system introduced by the reform improves the information transparency of firms’ environmental performance, effectively reducing information asymmetry between firms and financial institutions and expanding access to green credit resources for eligible enterprises, thus providing sufficient financial support for their cross-regional expansion activities. Notably, the operation of this financing pathway is not universal but conditional on firms’ actual environmental performance and their corresponding risk assessment results in the credit evaluation system of financial institutions.

Third, the impact of the environmental protection tax reform on cross-regional investment exhibits significant heterogeneity across different firm characteristics and regional institutional contexts. Specifically, capital-intensive firms show a far stronger investment response to the tax reform, with their cross-regional investment scale increasing by 20.2% relative to the sample mean; this is because the high sunk costs associated with end-of-pipe emission reduction upgrades in existing production sites make cross-regional relocation a comparatively more cost-effective strategy for such firms. In addition, firms located in regions with lower fiscal pressure and stricter environmental law enforcement also exhibit larger increases in cross-regional investment, highlighting that the actual effectiveness of environmental tax policy is critically dependent on the local government’s institutional capacity and long-term commitment to consistent regulatory enforcement [50].

5.2. Policy Recommendations

The empirical findings outlined above yield targeted policy implications that are directly aligned with the observed patterns of cross-regional investment responses to the environmental tax reform, without extending to institutional design issues beyond the scope of the study’s evidence.

Our results confirm that interprovincial tax rate differentials drive capital flows, creating a tension between local environmental goals and the national objective of building a unified market. While provincial autonomy in rate-setting allows tailored environmental policies, the resulting capital mobility may undermine environmental effectiveness if firms relocate rather than reduce pollution. Policymakers should consider establishing interregional coordination mechanisms—such as minimum effective tax floors or interprovincial revenue-sharing arrangements—to prevent “race-to-the-bottom” competition while preserving local flexibility.

The finding that fiscal pressure attenuates policy effectiveness highlights the need for central-level enforcement standards and oversight. Relying solely on local governments’ voluntary compliance may be insufficient, particularly in fiscally constrained regions where enforcement forbearance is tempting. The central government could consider establishing dedicated funding for environmental tax administration in high-need regions and implementing routine compliance audits.

The financing constraint mechanism operates heterogeneously across firms with different environmental performance. While the environmental tax information-sharing platform improves credit access for environmentally responsible firms, it may tighten financing for heavy polluters. Policymakers should design differentiated green credit policies that provide transitional financing support to firms actively undertaking emission reduction upgrades, avoiding a “one-size-fits-all” approach that could force viable firms into premature exit while merely shifting pollution across provincial borders.

Capital-intensive firms exhibit the strongest relocation responses due to their high sunk costs of on-site upgrading. Policymakers should provide targeted transition support—including technology upgrading subsidies, workforce retraining programs, and reasonable phase-in periods—to guide these firms toward “relocation as upgrading” rather than simple pollution transfer.

5.3. Limitations and Directions for Future Research

This study has several limitations that warrant acknowledgment. The sample covers all A-share listed companies, and a large proportion of non-polluting or low-emission enterprises naturally exhibit lower sensitivity to the environmental protection tax, which may dilute the estimated policy impact. While limiting the sample to heavily polluting industries could enhance homogeneity, it would also diminish the study’s value for assessing the policy’s general applicability. Future research could explore weighted regression approaches that assign greater weight to pollution-intensive industries.

This study has several limitations. First, the sample includes many non-polluting firms that are naturally less sensitive to the reform, potentially diluting the estimated effects. Second, our mechanism variables rely on proxy measures subject to measurement error; future research with direct environmental expenditure data could enhance precision. Third, the causal chain of the financing constraint channel—from information-sharing platforms to improved bank credit—relies on statistical correlations rather than direct institutional evidence. Finally, our analysis captures short-to-medium-term effects (2018–2024); long-term consequences for relocated subsidiaries and host-region environments await further research.