Introduction

Since 1980, the global trade landscape has gradually evolved from inter-product to intra-product specialization. As the system of international division of labor deepens and value chains expand, a nation’s export competitiveness depends less on quantitative measures like export volume and more on enterprises’ ability to add value to products and their positioning within the global value chains, an increasingly pivotal role as highlighted by Koopman et al. (2012). Benefiting from its comparative advantage in low factor costs, China has witnessed a surge in its export volumes and has emerged as the “world’s factory,” commanding a significant presence in the global value chains. However, despite these “achievements” in export scale, Chinese manufacturing enterprises continue to grapple with challenges such as size without strength and generally low value-added capabilities. For instance, an exported Chinese iPod priced at $150 generates less than $4 in value added (Linden et al. 2009), whereas the value added from an iPhone export represents only about 1.8% of the device’s total value (Kraemer et al. 2011). While Chinese manufacturing has yet to develop a comprehensive export competitive edge, the long-standing production approach characterized by high input, high consumption, and high pollution has also led to a gradual decline in the domestic natural environment, triggering severe pollution issues and social crises, such as the “2005 Songhua River” contamination, the “2009 Hunan Liuyang cadmium pollution,” and the “2011 Yunnan Qujing chromium slag pollution” incidents. This dire situation compels China to hasten the enforcement of environmental management policies. In practice, the fragmented environmental management system spread across varying levels of government hampers the efficiency of delivering crucial environmental public services. Theoretically, it also impedes the enhancement of domestic manufacturing enterprises’ export competitiveness from an institutional arrangement perspective.

Research on environmental decentralization has unfolded along two principal lines: first, the measurement of environmental decentralization, where most studies have used indirect methods, focusing primarily on legal frameworks, institutional evidence, and practical characteristics (Fredriksson and Svensson, 2003; Lutsey and Sperling, 2008). Second, investigations into the impact of environmental decentralization have examined its economic and social effects, including environmental quality, individual behavior, transboundary spillovers, and government governance (Zhuravskaya, 2000; Faguet, 2001). Another body of literature pertinent to this research is the domestic value-added rate in exports (hereafter referred to as export DVAR). This concept largely stems from outsourcing and division of labor theories (Feenstra and Hanson, 1997; Grossman and Rossi-Hansberg, 2008) and covers two main areas: firstly, the measurement of export DVAR has progressed from early industry-level assessments based on input-output tables (Koopman et al. 2012) to more recent micro-level evaluations utilizing resources such as the China Industrial Enterprise Database and China Customs Database (Upward et al. 2013; Kee and Tang, 2016), with the latter method experiencing continuous refinement and widespread application. Secondly, the influence of export DVAR has been examined from various angles, including environmental regulation (Sun et al. 2023), value-added tax implications (Wu et al. 2021), employment effects (Chen et al. 2012), foreign direct investment impacts (Lu et al. 2022), and industrial agglomeration considerations (Liu et al. 2022).

In summary, while research on environmental decentralization and export DVAR has yielded substantial findings within their respective domains, there remains an academic void concerning the interplay between the two. This study aims to bridge this gap by integrating environmental decentralization and export DVAR within a singular analytical framework, endeavoring to augment the extant literature on three fronts: firstly, by assessing the influence of environmental decentralization on corporate export DVAR; secondly, by disaggregating environmental decentralization into administrative, inspection, and monitoring facets, and delving into the effects of such decentralization on export DVAR through the lens of heterogeneous attributes such as trade patterns, ownership structures, and pollution-intensive industries; thirdly, by refining the construction of enterprise DVAR indicators to surmount the limitations identified in the current body of literature.

Theoretical mechanisms and hypotheses

Mechanism analysis

This paper draws on the model of Kee and Tang (2016) to theoretically derive DVAR of the enterprise i’s product k exports.

$$DVA{R}_{ikt}=1-\frac{{\kappa }_{I}/{\varpi }_{ikt}}{1+{({p}_{ikt}^{imp}/{p}_{ikt}^{dom})}^{b-1}}$$
(1)

Where \({\kappa }_{I}\), \({\varpi }_{ikt}\), \({p}_{ikt}^{imp}\), \({p}_{ikt}^{dom}\) and b denote intermediate goods output elasticity, cost markup, imported intermediate goods price, domestic intermediate goods price, and domestic and foreign intermediate goods substitution elasticity, respectively. From Eq. (1), it can be seen that the change in DVAR of a representative enterprise’s product exports is jointly determined by factors such as marginal cost markup \({\varpi }_{ikt}\) and the relative price of imported intermediate goods. The first-order derivation is.

$$\frac{\partial DVA{R}_{ikt}}{\partial {\varpi }_{ikt}}=\frac{{\kappa }_{I}}{{\varpi }_{ikt}^{2}}\frac{1}{1+{({p}_{ikt}^{imp}/{p}_{ikt}^{dom})}^{b-1}}\, >\, 0$$
(2)
$$\frac{{\partial {DVAR}}_{{ikt}}}{\partial ({p}_{{ikt}}^{{imp}}/{p}_{{ikt}}^{{dom}})}=\frac{{\kappa }_{I}(b-1)}{{\varpi }_{{ikt}}}\frac{{({p}_{{ikt}}^{{imp}}/{p}_{{ikt}}^{{dom}})}^{b-2}}{{[1+{({p}_{{ikt}}^{{imp}}/{p}_{{ikt}}^{{dom}})}^{b-1}]}^{2}}\, >\, 0$$
(3)

It is easy to see that the marginal cost markup and the relative price of imported intermediate goods are positively related to the export DVAR of the enterprise’s products, and the economics behind this is evident: the increase in the cost markup of the enterprise’s products, i.e., the expansion of the ratio of the output value of exports to production inputs, will increase the profitability and export DVAR of the enterprise’s exports. The increase in the price of imported intermediate goods will make enterprises reduce the demand for imported intermediate goods to pursue cost minimization, promoting the increase in the export DVAR of enterprises.

Cost markup impact channel

According to industrial organization theory, cost markup is defined as the ratio of a product’s price to its marginal cost, which serves as an indicator of a enterprise’s profitability and market power (De Loecker and Warzynski, 2012; De Loecker et al. 2016). Within the context of China’s environmental decentralization framework, the allocation and adjustment of environmental management authority by the central government dictate the degree of environmental decentralization. An increase in this degree suggests an expansion of local authorities’ powers in environmental governance. Given that the central government often lacks coercive power to enforce environmental policies over local governments, national environmental protection directives may not be fully implemented or enforced due to protectionist tendencies at the local level. Moreover, local environmental protection agencies, which rely on local governments for personnel and funding, often lack the independence and impartiality required for effective environmental management. This decentralization of environmental regulation tends not to favor the intensification of environmental controls. Consequently, this can lead to lower compliance and “market entry threshold” costs for enterprises due to the “compliance cost effect” (Managi and Karemera, 2005). As more businesses enter the market, increased competition may ensue (De Loecker et al. 2016), which in turn can squeeze profit margins and thereby suppress increases in enterprises’ marginal cost markup levels., i.e.,\((\frac{\partial {\varpi }_{{ikt}}}{\partial {{ED}}_{{pt}}})\, <\, 0\).

Based on the previous mathematical model’s derived findings, combined with the analysis in this section, it can be further deduced that environmental decentralization promotes the export DVAR of enterprises’ products through the cost markup influence channel.

$$\frac{{\partial {DVAR}}_{{ikt}}}{{\partial {ED}}_{{ikt}}}=\frac{{\partial \varpi }_{{ikt}}}{{\partial {ED}}_{{ikt}}}\frac{{\partial {DVAR}}_{{ikt}}}{{\partial \varpi }_{{ikt}}}\, <\, 0$$
(4)

R&D innovation impact channel

Numerous studies have proved the role of R&D innovation as a critical driver of enhanced export competitiveness for enterprises (Zhu and Fu, 2013). Intensified environmental protection and heightened regulatory stringency compel businesses to incur additional regulatory costs. These costs are then partially transferred to other enterprises demanding intermediate goods in the form of price premiums on these goods, which can compress profit margins and attenuate competitiveness. As a result, domestic businesses may cut back on the demand for costlier foreign intermediate goods. Conversely, environmental decentralization can have the inverse effect., i.e., \(\left(\partial {{IMP}}_{{ikt}}^{I}/\partial {{ED}}_{{ikt}}\right)\, >\, 0\). Firstly, environmental decentralization can be advantageous for boosting profits and capital accumulation, which in turn enhances the purchasing power for imported intermediate goods, that is, the relative price of imported intermediate goods decreases. An increase in the volume of imported products of a similar nature and heightened domestic competition may “compel” enterprises to escalate their R&D and innovation efforts in order to sustain their market competitiveness (Aghion et al. 1997; Aghion et al. 2001). Secondly, the influx of higher-quality imported intermediate goods also presents domestic enterprises with opportunities for access, learning, and imitation. The resultant “technology spillover” effects may assist these enterprises in elevating their R&D and innovation capabilities (Coe and Helpman, 1995; Cui et al. 2022)., i.e., \((\partial R{D}_{it}/\partial IM{P}_{ikt}^{I})\, >\, 0\). Third, according to the “pecking order” theory of the sources of financing innovation inputs, an increase in own funds will also promote enterprises to engage in R&D and innovation activities, thus increasing export DVAR. Higher relative prices of imported intermediate goods (Kee and Tang, 2016), i.e., \([\partial ({p}_{ikt}^{imp}/{p}_{ikt}^{dom})/\partial R{D}_{it}]\, >\, 0\).

Based on the previous mathematical derivation conclusions\([\partial DVAR{}_{ikt}/\partial ({p}_{ikt}^{imp}/{p}_{ikt}^{dom})]\, >\, 0\), combined with the analysis in this section, it can be further deduced that environmental decentralization promotes the export DVAR of enterprises’ products through the R&D innovation influence channel.

$$\frac{\partial {{DVAR}}_{{ikt}}}{\partial {{ED}}_{{ikt}}}=\frac{\partial {{IMP}}_{{ikt}}^{I}}{\partial {{ED}}_{{ikt}}}\frac{\partial {{RD}}_{{it}}}{\partial {{IMP}}_{{ikt}}^{I}}\frac{\partial ({p}_{{ikt}}^{{imp}}/{p}_{{ikt}}^{{dom}})}{\partial {{RD}}_{{it}}}\frac{\partial {{DVAR}}_{{ikt}}}{\partial ({p}_{{ikt}}^{{imp}}/{p}_{{ikt}}^{{dom}})}\, >\, 0$$
(5)

Based on the above analysis, the following hypotheses are proposed to be examined in this paper.

Hypothesis 1: Environmental decentralization nonlinearly affects the export DVAR of enterprises’ products.

Hypothesis 2: Environmental decentralization has the potential to lower the marginal cost markup, thereby decreasing the export Dynamic Value-At-Risk (DVAR) of enterprises’ products. Simultaneously, it can elevate the export DVAR by fostering R&D innovation. The overall impact of environmental decentralization on a enterprise’s export DVAR will ultimately hinge on the interplay between these two effects.

Empirical study design

Econometric model construction

This paper focuses on the impact of environmental decentralization on export DVAR. The following econometric model is constructed in this paper based on the existing studies.

$$\begin{array}{ll}DVA{R}_{ikt}={\alpha }_{0}+{\beta }_{1}E{D}_{pt}+{\beta }_{2}ED{S}_{pt}+{\gamma }_{1}F{D}_{pt}+{\gamma }_{2}AG{E}_{it}+{\gamma }_{3}MARKE{T}_{pt}\\ \qquad\qquad\quad+\,{\gamma }_{4}HH{I}_{pt}+{\gamma }_{5}SAL{E}_{it}+{\gamma }_{6}F{S}_{it}+{\gamma }_{7}SO{E}_{it}+{\gamma }_{8}SU{B}_{it}+{\phi }_{k}+{\eta }_{t}+{\varepsilon }_{ikt}\end{array}$$
(6)

Where the subscript i denotes the enterprise, k denotes the digit-2 industry, and t denotes the year. The dependent variable, \(DVA{R}_{ikt}\) denotes the enterprise’s export DVAR. The core independent variables, \(E{D}_{pt}\) and \(ED{S}_{pt}\), are the horizontal and quadratic terms of environmental decentralization, respectively, to mitigate the endogeneity effects lagged by one period. The control variables include: \(F{D}_{pt}\), denotes fiscal decentralization, measured as the ratio of central and local fiscal revenues; \(AG{E}_{it}\), denotes enterprise age, measured as subtracting the time of registration from the time of collecting the samples (\(AG{E}_{it}=Tim{e}_{Collecting}-Tim{e}_{Registration}\)); \(MARKE{T}_{pt}\), denotes the level of marketization, measured as the Fan’s index (the marketization index); \(HH{I}_{pt}\), denotes industry concentration, measured as the calculated Herfindahl index; \(SAL{E}_{it}\), denotes sales, expressed as total annual enterprise sales; \(F{S}_{it}\), denotes enterprise’s size, measured as total enterprise assets; \(SO{E}_{it}\), denotes enterprise’s state-owned status, measured as a 0-1 dummy variable; \(SU{B}_{it}\), denotes the intensity of government subsidies, measured as the number of subsidies received by the enterprise. \({\phi }_{k}\), \({\eta }_{t}\), and \({\varepsilon }_{ikt}\) denote industry-fixed effects, year-fixed effects, and stochastic error terms, respectively. All continuous variables in the model are logarithmized to mitigate the possible heteroskedasticity in the regression process.

Indicator measurement

Export DVAR of enterprises

In this paper, drawing upon the methodologies established by Upward et al. (2013), and Kee and Tang (2016), we account for factors such as the indirect imports of intermediate goods and the imports of capital goods when calculating the export DVAR indicator for enterprises. Addressing the issue of indirect imports of intermediate goods is a crucial step in the calculation. This issue is categorized into three distinct types: the first is related to trade intermediaries; the second pertains to transactions of intermediate goods among domestic processing trade companies; and the third involves the international fragmentation of products. The first two types are addressed by building on insights from the existing literature, while for the third type, scholars have proposed various solutions (Koopman et al. 2012; Kee and Tang, 2016). However, since the dataset used in this study spans from 2000 to 2014, it is not feasible to apply values to the variables of interest using these existing studies’ findings.

Furthermore, Wang et al. (2014) observed that the average proportion of domestic value added in China’s export of imported intermediate goods rose to 1.3% in 2007, marking a significant increase from the 0.1% recorded in 1995. Kee and Tang (2016), drawing on extrapolations from Koopman et al. (2014), noted that the share of foreign value added from domestic intermediate goods in China’s exports varied markedly across industries, with averages ranging from 0.4% to 5.7%, all of which exceeded zero. Consequently, the contributions of foreign value added to domestic intermediate goods and of domestic value added to imported intermediate goods are complex and cannot be oversimplified. Diverging from the aforementioned scholars, this paper employs country-sector data from the World Input-Output Database to compute the re-imported value-added rate and the exported foreign value-added rate for each manufacturing sector in China for each year, as per Wang et al. (2017). These rates serve as proxy indicators for the domestic share of value added to imported intermediate goods and the foreign share of value added to domestic intermediate goods, respectively. This approach facilitates a more accurate estimation of the real value of intermediate goods imports, aligning more closely with actual levels.

After calculating the actual imports of capital goods using the method mentioned above, this paper constructs the DVAR indicator for enterprises’ exports.

$$\left\{\begin{array}{ll}DVA{R}_{ikt}^{o}=1-\frac{IM{P}_{ikt}^{o}+IMP{K}_{ikt}^{o}-{D}_{ikt}^{o}}{{Y}_{ikt}}-{\delta }_{id}^{F}+{\delta }_{id}^{D}\\ DVA{R}_{ikt}^{m}=1-\frac{IM{P}_{ikt}^{m}+IMP{K}_{ikt}^{m}-{D}_{ikt}^{m}}{{Y}_{ikt}}-{\delta }_{id}^{F}+{\delta }_{id}^{D}\\ DVA{R}_{ikt}^{x}={\sigma }_{m}\cdot (1-\frac{IM{P}_{ikt}^{o}+IMP{K}_{ikt}^{o}-{D}_{ikt}^{o}}{{Y}_{ikt}}-{\delta }_{id}^{F}+{\delta }_{id}^{D})\\ \qquad\qquad\quad+\,{\sigma }_{o}\cdot (1-\frac{IM{P}_{ikt}^{m}+IMP{K}_{ikt}^{m}-{D}_{ikt}^{m}}{{Y}_{ikt}}-{\delta }_{id}^{F}+{\delta }_{id}^{D})\end{array}\right.$$
(7)

Where the superscript o, m and x denote ordinary trade, processing trade and mixed trade patterns, respectively. \(IM{P}_{ikt}\) and \(IMP{K}_{ikt}\) denote the actual import amount of intermediates and capital goods, \({D}_{ikt}\) denotes the accumulated depreciation of imported capital goods. \({\delta }_{id}^{F}\) and \({\delta }_{id}^{D}\) denote the foreign value-added rate and the return value-added rate of exports, respectively. \({\sigma }_{m}\) and \({\sigma }_{o}\) denote the proportion of processing trade export mode and ordinary trade export pattern, respectively.

Environmental decentralization

While the concept of environmental decentralization is grounded in the theory of environmental federalism and has its theoretical roots in fiscal decentralization (Oates, 1999), the unique historical evolution of China’s environmental management system complicates the task of accurately reflecting the country’s actual level of environmental decentralization through the traditional, narrowly-defined fiscal decentralization paradigm and predetermined legal frameworks. Such approaches can also obscure the structural nuances within environmental governance (Sigman, 2005). Consequently, this paper assesses the degree of environmental decentralization using data related to the staffing distribution across various levels of environmental agencies. The advantages of employing staffing data are manifold: firstly, personnel is the primary conduit through which authorities at different levels deliver public environmental services and execute environmental functions, making the distribution of staff a more objective indicator of the division of environmental responsibilities. Secondly, the fluctuations in institutional size and personnel tend to be more consistent. Thirdly, staffing levels and distribution are key variables for measuring decentralization. Given China’s specific context, environmental decentralization is principally linked to managerial decentralization, and the distribution of staff aligns more accurately with the essence of environmental decentralization. Lastly, the consistency and completeness of the relevant data is superior, rendering it a more practical measure.

Reflecting the nuances of China’s environmental affairs, environmental decentralization is further delineated into the categories of environmental administrative decentralization and environmental enforcement decentralization. The framework for constructing the environmental decentralization index is as follows:

$$E{D}_{pt}=\frac{EP{P}_{pt}/PO{P}_{pt}}{EP{P}_{t}/PO{P}_{t}}\left(1-\frac{GD{P}_{pt}}{GD{P}_{t}}\right)$$
(8)

Where the subscript p and t denote the province and year, respectively. \(E{D}_{pt}\) denotes the environmental decentralization index, \(EP{P}_{pt}\) and \(EP{P}_{t}\) denote the number of administrative staff in provincial environmental protection agencies and local environmental protection agencies, respectively. \(PO{P}_{pt}\) and \(PO{P}_{t}\) denote the number of people in each province and the country, respectively. \(GD{P}_{pt}\) and \(GD{P}_{t}\) denote the economic size of each province and country, respectively. \((1-GD{P}_{pt}/GD{P}_{t})\) denote a weight to deflate the decentralized indicators to mitigate the endogeneity of the indicators. Similarly, the environmental decentralized indicators are further subdivided to construct the environmental administrative decentralization index, \(EA{D}_{pt}\), environmental inspecting decentralization index, \(ES{D}_{pt}\), and environmental monitoring decentralization index, \(EM{D}_{pt}\).

$$EA{D}_{pt}=\frac{EA{P}_{pt}/PO{P}_{pt}}{EA{P}_{t}/PO{P}_{t}}\left(1-\frac{GD{P}_{pt}}{GD{P}_{t}}\right)$$
(9)
$$ES{D}_{pt}=\frac{ES{P}_{pt}/PO{P}_{pt}}{ES{P}_{t}/PO{P}_{t}}\left(1-\frac{GD{P}_{pt}}{GD{P}_{t}}\right)$$
(10)
$$EM{D}_{pt}=\frac{EM{P}_{pt}/PO{P}_{pt}}{EM{P}_{t}/PO{P}_{t}}\left(1-\frac{GD{P}_{pt}}{GD{P}_{t}}\right)$$
(11)

Data sources and processing

The study spans the years 2000 to 2014 and utilizes five distinct data sets:

  1. 1.

    The China Industrial Enterprises Database, sourced from the National Bureau of Statistics of China, encompasses information pertaining to all state-owned and large non-state-owned enterprises. This data set provides details on the age, sales, current assets, and fixed investment of enterprises. To address issues of sample mismatches, missing indicators, and abnormal values, we implement corrective measures following Brandt et al. (2012), which include: (a) excluding enterprises with fewer than eight employees; (b) removing enterprises with negative total assets, gross fixed assets, intermediate inputs, or wages payable; (c) omitting enterprises founded before 1949.

  2. 2.

    The China Customs Database, provided by the General Administration of Customs of China, includes comprehensive import and export data at the enterprise level with 8-digit HS codes. This resource allows for the assessment of export technological sophistication and the amount and quality of imported intermediate goods. Following Ahn et al. (2011), we address the issue of intermediary reliance in China’s enterprise imports and exports by aggregating monthly data into annual figures and excluding abnormal values. This involves filtering out names with “import and export,” “economic and trade, “ and “science and trade “.

  3. 3.

    The World Input-Output Database (WIOD) offers country-sector input-output information. Using the KWW model, the database facilitates the decomposition and calculation of both the domestic value-added rate for each industry in China and the value-added rate for exports abroad.

  4. 4.

    Data on the distribution of personnel across various provincial-level environmental protection agencies is obtained from the China Environmental Yearbook, provided by the National Bureau of Statistics (NBS). In addition, this source provides population figures and economic size data.

  5. 5.

    The China Enterprise Patent Database, from the National Patent Office, contains details on patent applications, including applicant names and application dates. This data is employed to match and calculate the number of enterprise patent applications.

As for the control variables, all are directly sourced or computed from the China Industrial Enterprises Database, with the exception of fiscal decentralization data, which comes from the China Statistical Yearbook, and the marketization index from the China National Economic Research Institute.

The data used in this paper are distributed in several databases, so matching and merging are required before use. Following Brandt et al. (2012), the industrial enterprise database is merged for different years by enterprise code and enterprise name, and then it is matched and merged with the customs import/export database by enterprise name. Finally, the merged data are then merged with other required data.

Characteristic facts

This paper focuses on investigating the relationship between the dynamic evolution of environmental decentralization and enterprise export DVAR. It presents scatter plots to illustrate the correlation between administrative decentralization, enforcement decentralization, and monitoring decentralization and the average export DVAR of enterprises within each province, as shown in Fig. 1. The analysis reveals an inverse relationship between administrative decentralization and enterprise export DVAR, although the nonlinearity of this correlation is somewhat subtle. Furthermore, there is a discernable inverted U-shaped nonlinear relationship between enforcement decentralization and enterprise export DVAR, as well as a U-shaped nonlinear relationship between monitoring decentralization and enterprise export DVARFootnote 1.

Fig. 1
figure 1

Scatter plot of environmental decentralization versus provincial average export DVAR.

Full size image

Empirical results and analysis

Baseline empirical results

Table 1 displays the baseline regression results for the impact of environmental decentralization on the export DVAR of enterprises. Columns (1) and (2) detail the effects of overall environmental decentralization on enterprise export DVAR. In column (1), the coefficient for the linear term of the total environmental decentralization variable is 0.029, significant at the 1% level, suggesting a positive contribution of total environmental decentralization to the DVAR of enterprise exports. The linear and quadratic coefficients of the central variables are 0.200 and −0.103, respectively, both significant at the 1% level. The U-test yields a t-value of 4.726, indicating an inverted U-shaped nonlinear relationship between total environmental decentralization and enterprise export DVAR, with a turning point at 0.975. Out of approximately 395,382 valid samples from the study period, 373,240 are situated to the left of this turning point. This suggests that in 94.4% of provinces and regions, overall environmental decentralization is conducive to enhancing the export DVAR. These findings lend support to Hypothesis 1.

Table 1 Baseline regression results.
Full size table

The paper delves into the effects of various types of environmental decentralization on enterprises’ export DVAR. Columns (3),(4), (5),(6), and (7),(8) present the impacts of administrative decentralization, inspecting decentralization, and monitoring decentralization on the export DVAR of enterprises, respectively. When considering only the linear terms of the variables, administrative and monitoring decentralization positively influence the export DVAR of enterprises. However, the impact of enforcement decentralization is not statistically significant. The inclusion of quadratic terms in the respective models yields linear and quadratic coefficients in columns (4), (6), and (8) of −0.054 and 0.039, 0.017 and −0.009, −0.022 and 0.012, respectively, all statistically significant at the 10% level or below. The associated U-test t-values are 3.189, 1.728, and 3.432, respectively, which challenge the initial “linear effect” hypothesis, instead suggesting a U-shaped non-linear effect of administrative and inspecting decentralization on the export DVAR of enterprises. Conversely, an inverted U-shaped non-linear effect is observed for monitoring decentralization. Further analysis indicates that the inflection points for administrative, inspecting, and monitoring decentralization are 0.689, 0.928, and 0.960, respectively. The proportions of samples to the left of these inflection points are 71.5, 78.2, and 74.0%, respectively. This suggests that, within the corresponding ranges, administrative and inspecting decentralization are not favorable for enhancing export DVAR, while monitoring decentralization appears to promote improvement in export DVAR.

This paper posits that the variability observed in the outcomes is intimately connected to the nuances of environmental affairs and the institutional context in which they operate. Firstly, the positive impact of overall environmental decentralization on export DVAR suggests a need for further decentralization to refine environmental governance in China. Environmental administrative functions typically encompass issues like local environmental regulations and planning, which necessitate a comprehensive understanding of the political, economic, and social landscapes within the jurisdiction. Increased administrative decentralization could leverage local informational advantages, thereby fostering a robust competitive mechanism. Current levels of administrative decentralization fall significantly short of the optimal inflection point, indicating a need to empower local governments further in environmental matters. Furthermore, environmental inspecting functions, which include law enforcement, inspections, and related activities, benefits from greater decentralization. This not only mitigates risks to local economic growth but also has the potential to disrupt inefficient hierarchical management structures, thus diminishing institutional risks and enhancing export DVAR. Lastly, environmental monitoring functions, which involve overseeing regional environmental governance data, come with substantial financial and technical demands. Greater decentralization in monitoring entrusts more responsibility to local governments, which may adversely affect the quality of monitoring data due to less stringent oversight. The difficulty in establishing “genuine constraints” and incentives at the local level could be detrimental to improving export DVAR.

Among the control variables, FD and AGE notably hinder the growth of enterprise export DVAR, suggesting that an increase in local fiscal autonomy and the aging of enterprises negatively affect their export DVAR. Conversely, higher levels of market-oriented development and SOE enhance enterprise export DVAR. This implies that regions with more advanced market mechanisms foster better performance in enterprises’ export DVAR, as does the status of being a state-owned enterprise. The coefficients for the other control variables do not yield robust or significant outcomes.

Analysis of the heterogeneity in exporters’ trade patterns

This paper investigates whether the impact of environmental decentralization on enterprise export DVAR varies according to different trade patterns, specifically examining ordinary trade, processing trade, and mixed trade.

Table 2 presents the regression analyses concerning how environmental decentralization influences the trade pattern heterogeneity manifested in enterprise export DVAR. Columns (1)–(3) detail the effects of administrative decentralization on the export DVAR for enterprises engaged in diverse trade patterns. The linear and quadratic coefficients of the administrative decentralization variables are −0.044 and 0.032 for ordinary trade, −0.234 and 0.124 for processing trade, and 0.032 and −0.016 for mixed trade, respectively. The administrative decentralization coefficients are statistically significant at the 1% level for both ordinary and processing trade enterprises, but they are not significant for mixed trade enterprises. The U-test results are statistically significant for ordinary trade enterprises, with t-values of 3.365 for ordinary trade and 2.395 for processing trade, both reaching statistical significance. These findings suggest a U-shaped non-linear effect of administrative decentralization on the export DVAR of enterprises engaged in ordinary and processing trades, with no discernible impact on mixed trade enterprises. Additionally, the calculated inflection points are 0.686 for ordinary trade and 0.943 for processing trade. Correspondingly, the proportion of samples falling to the left of these inflection points is 26.8% for ordinary trade and 14.3% for processing trade, indicating that during the sample period, administrative decentralization negatively affected the export DVAR of 73.2% of ordinary trade enterprises and 85.7% of processing trade enterprises.

Table 2 Regression results of trade pattern heterogeneity.
Full size table

Columns (4)–(6) present the regression findings on the impact of monitoring decentralization on the export DVAR for companies with distinct trade practices. The linear and quadratic coefficients of the monitoring decentralization variables are −0.022 and 0.009, 0.141 and −0.068, 0.032 and −0.012 for each respective column. Importantly, the coefficients of these core variables are statistically significant solely in the samples of ordinary and processing trade enterprises. The U-test yields t-values of 2.134 for ordinary trade and 3.296 for processing trade enterprises, with both figures demonstrating statistical significance. These results indicate that monitoring decentralization has a U-shaped non-linear impact on the export DVAR of ordinary trade enterprises, and conversely, an inverted U-shaped non-linear impact on the export DVAR of processing trade enterprises. Further examination reveals inflection points at 1.233 for ordinary trade and 1.043 for processing trade. Additionally, 30.8% of ordinary trade and 14.3% of processing trade samples are located to the left of these inflection points. These percentages suggest that during the sampling period, monitoring decentralization acted to suppress the export DVAR of 69.2% of ordinary trade enterprises, while it positively influenced the export DVAR of 85.7% of processing trade enterprises.

Columns (7)–(9) detail the regression outcomes assessing the influence of inspecting decentralization on the export DVAR for businesses with diverse trade patterns. The linear and quadratic coefficients for the inspecting decentralization variables are −0.046 and 0.026 for ordinary trade, −0.103 and 0.036 for processing trade, and 0.013 and −0.006 for mixed trade, respectively. The correlation coefficients are statistically significant at the 1% level for ordinary and processing trade enterprises. The U-test t-values for the respective sub-samples are 5.152 for ordinary trade and 2.692 for processing trade, with both being statistically significant at the 1% level. This signifies that inspecting decentralization has a U-shaped nonlinear impact on the export DVAR of both ordinary and processing trade enterprises. Moreover, the inflection points are identified at 1.446 for ordinary trade and 1.072 for processing trade. Additionally, 34.0% of the ordinary trade samples and 13.3% of the processing trade samples are situated to the left of these inflection points. This distribution suggests that during the examined period, inspecting decentralization had a detrimental effect on the export DVAR of 66.0% of ordinary trade enterprises and 86.7% of processing trade enterprises.

The results of the influence mechanism

Exploring the mechanisms that explain the impact of environmental decentralization on enterprises’ export DVAR, as well as delineating the connections between environmental decentralization and enterprises’ export DVAR, is beneficial. In pursuit of this objective, this study investigates the transmission channels through which environmental decentralization modulates enterprises’ export DVAR. In this paper, we adopt the stepwise regression approach as proposed by Baron and Kenny (1986) to scrutinize the mediating effect, where the focus lies on the coefficient product test in the mediating analysis. Nonetheless, it is not uncommon in empirical applications to encounter situations where the coefficient product is significant, whereas the sequential test is not (Judd and Kenny, 1981; MacKinnon et al. 2002). Consequently, should the sequential test render any of the independent or mediating variables insignificant, it becomes imperative to further probe the coefficient product through the Sobel (1982) method to ascertain the presence of a mediating effect.

The econometric model relevant to the testing of mediating effects is structured as follows:

$$\begin{array}{ll}DVA{R}_{ikt}={\alpha }_{0}+{\beta }_{1}E{D}_{pt}+{\beta }_{2}E{D}_{pt}^{2}+{\gamma }_{1}F{D}_{pt}+{\gamma }_{2}AG{E}_{it}+{\gamma }_{3}MARKE{T}_{pt}\\ \qquad\qquad\quad+\,{\gamma }_{4}HH{I}_{pt}+{\gamma }_{5}SAL{E}_{it}+{\gamma }_{6}F{S}_{it}+{\gamma }_{7}SO{E}_{it}+{\gamma }_{8}SU{B}_{it}+{\phi }_{k}+{\eta }_{t}+{\varepsilon }_{ikt}\end{array}$$
(12)
$$\begin{array}{ll}{Z}_{ikt}={\alpha }_{0}+{\beta }_{1}E{D}_{pt}+{\beta }_{2}E{D}_{pt}^{2}+{\gamma }_{1}F{D}_{pt}+{\gamma }_{2}AG{E}_{it}+{\gamma }_{3}MARKE{T}_{pt}\\ \qquad+\,{\gamma }_{4}HH{I}_{pt}+{\gamma }_{5}SAL{E}_{it}+{\gamma }_{6}F{S}_{it}+{\gamma }_{7}SO{E}_{it}+{\gamma }_{8}SU{B}_{it}+{\phi }_{k}+{\eta }_{t}+{\varepsilon }_{ikt}\end{array}$$
(13)
$$\begin{array}{ll}DVA{R}_{ikt}={\alpha }_{0}+{\beta }_{1}E{D}_{pt}+{\beta }_{2}E{D}_{pt}^{2}+{\delta }_{1}{Z}_{ikt}+{\gamma }_{1}F{D}_{pt}+{\gamma }_{2}AG{E}_{it}+{\gamma }_{3}MARKE{T}_{pt}\\ \qquad\qquad\quad+\,{\gamma }_{4}HH{I}_{pt}+{\gamma }_{5}SAL{E}_{it}+{\gamma }_{6}F{S}_{it}+{\gamma }_{7}SO{E}_{it}+{\gamma }_{8}SU{B}_{it}+{\phi }_{k}+{\eta }_{t}+{\varepsilon }_{ikt}\end{array}$$
(14)

Where \(E{D}_{pt}\) denotes the environmental decentralization index. \({Z}_{ikt}\) denotes the mechanism variable, based on the conclusions drawn in the combing derivation section above, including cost markup, \({\varpi }_{ikt}\), and R&D innovation, \(R{D}_{it}\).

Table 3 displays the regression analysis of how environmental decentralization influences enterprises’ export DVAR. Column (1) provides the regression outcomes for the impact of environmental decentralization on enterprises’ export DVAR, revealing that the coefficients for both the linear and quadratic terms of the environmental decentralization variable are statistically significant at the 1% level. Column (2) indicates that environmental decentralization has a notably negative effect on enterprises’ cost markup. Conversely, the results in column (3) hold statistical significance at the 1% level after incorporating cost markup as a mediating variable, demonstrating that environmental decentralization does indeed exert an influence on enterprises’ export DVAR through cost markup. This finding substantiates Hypothesis 2. The findings presented in column (4) show that environmental decentralization has a considerable impact on enterprises’ R&D innovation.

Table 3 Regression results of influence mechanism.
Full size table

In contrast, even after introducing the mediating variable of R&D innovation in column (5), the linear and quadratic coefficients of the environmental decentralization variable remain statistically significant. This persistence suggests that environmental decentralization also exerts an impact on enterprises’ export DVAR via the channel of R&D innovation, lending further support to the validity of Hypothesis 2. Column (6) incorporates both mediating variables—cost markup and R&D innovation—into the model concurrently. The correlation coefficients exhibit no significant change, maintaining consistency with previous findings in terms of the direction and statistical significance. To ensure robustness, this paper utilizes the Sobel (1982) method to examine the influence mechanism. This method ascertains the mediation effect by evaluating the significance of the coefficient product term. The Sobel test produces z-values of −27.61 and 3.46, respectively, both decisively rejecting the null hypothesis that the coefficient product term is zero. This robust conenterpriseation implies that both cost markup and R&D innovation are indeed mediating variables through which environmental decentralization influences enterprises’ export DVAR.

Robustness test results

Endogeneity problem

The primary focus of this study is the provincial-level environmental decentralization degree index, with the enterprise-level export DVAR serving as the dependent variable. The likelihood of reverse causality in this context is minimal. However, due to the differentiation in environmental governance across provinces, which is not strictly exogenous, it becomes imperative to conduct further tests for potential endogeneity. In empirical research, endogeneity typically manifests in three ways: firstly, as omitted variable bias, where unobserved factors that may influence the explanatory variables are inadvertently included in the model’s error term. This study mitigates this issue by controlling for year and industry fixed effects. The second form is sample selection bias, which occurs when the sample is comprised of a non-random subset of observations. This is addressed using the two-step method proposed by Heckman (1979). The third form is simultaneity or reverse causality, where the dependent and independent variables reciprocally affect each other. To address this, the study employs the two-stage least squares (2SLS) method with instrumental variables.

Selecting appropriate instrumental variables is critical for the 2SLS method. This paper initially chooses the lagged levels and quadratic terms of the core independent variables as instrumental variables. Furthermore, it constructs new instrumental variables using Lewbel’s (2012) heteroskedasticity-based identification method. This approach transcends the traditional instrumental variable estimation’s exclusion restriction requirement, offering greater flexibility in constructing instrumental variables.

Indicator and model robustness

In the robustness test for the indicators, this paper replaces the previously used environmental decentralization index with the year-end count of environmental protection agencies within each province and those under the central government. This alternative measure serves to test the robustness of the environmental decentralization indicator. For the robustness test of the model, considering that the dependent variable, enterprise export DVAR, is typically based on imputed data, this study employs the Tobit model to assess model robustness, taking into account the bounds of imputation.

Table 4, columns (1), (2), present the regression outcomes from employing Heckman’s (1979) two-step method to evaluate sample selection bias. The inverse Mills ratio (IMR) noted in column (2) is −0.415 and achieves statistical significance, suggesting the presence of sample selection bias within this research. Nonetheless, the coefficients for both the linear and quadratic terms of the environmental decentralization variables in column (1) align with the benchmark results, indicating that sample selection bias does not distort the study’s identification.

Table 4 Robustness regression results.
Full size table

Columns (3), (4) detail the regression findings from applying the 2SLS method to test for reverse causality. The correlation results show that environmental decentralization has an inverted U-shaped relationship with enterprise export DVAR, corroborating the benchmark findings. Additionally, the validity of the instrumental variables is assessed using Kleibergen and Paap’s (2006) criteria. The Kleibergen-Paap RK LM statistic conenterprises the identifiability of the instrumental variables, while the Kleibergen-Paap RK Wald F statistic demonstrates that the instrumental variables do not suffer from weak identification issues.

Columns (5), (6) display the regression results for the robustness of indicators and of the model, respectively. The consistency in the direction and significance of the linear and quadratic coefficients for the environmental decentralization variables, when compared with the baseline model findings, conenterprises the robustness of both the indicators and the model.

Conclusions and implications

China’s environmental landscape continues to be challenging, and given the public’s growing demand for environmental quality, selecting a viable environmental governance model and establishing a well-founded environmental decentralization system are crucial. These measures are not only pivotal for relevant authorities to effectively carry out environmental protection duties and deliver public environmental services but also form the structural foundation required to navigate through China’s present environmental governance complexities. This paper investigates the influence of environmental decentralization on enterprises’ export DVAR by utilizing a comprehensive dataset spanning from 2000 to 2014, which integrates information from the Chinese Industrial Enterprises Database, the Customs Database, and the Chinese Environmental Yearbook. The research results indicate that overall environmental decentralization has an inverted U-shaped effect on enterprises’ export DVAR. The conclusion has significant policy implications, suggesting that there exists a threshold effect in the impact of environmental decentralization. This finding is consistent with the study by Jiang et al. (2023). In 94.4% of the regions, overall environmental decentralization contributes to the improvement of enterprises’ export-related DVAR, which aligns closely with the findings of studies on environmental factors (Fang et al. 2022; Sun et al. 2023). Administrative decentralization and monitoring decentralization, on the whole, have a U-shaped effect on enterprises’ export DVAR, with 71.5 and 78.2% of enterprises falling into the inhibitory range, consistent with studies that found negative effects (Udeagha and Muchapondwa, 2023). Supervisory decentralization has an inverted U-shaped effect on enterprises’ export DVAR, with 74.0% of enterprises in the promotion range. In terms of heterogeneous effects, administrative decentralization has a U-shaped effect on export DVAR for general trade and processing trade enterprises, with 73.2% of general trade enterprises and 85.7% of processing trade enterprises in the inhibitory range. Supervisory decentralization has a U-shaped effect on export DVAR for general trade enterprises, with 69.2% of general trade enterprises in the inhibitory range. Supervisory decentralization has an inverted U-shaped effect on export DVAR for processing trade enterprises, with 85.7% of processing trade enterprises in the promotion range. Monitoring decentralization has a U-shaped effect on export DVAR for general trade and processing trade enterprises, with 66.0% of general trade enterprises and 86.7% of processing trade enterprises in the inhibitory range. Mechanism analysis results show that cost markups and research and development innovation are mediating variables linking environmental decentralization and enterprises’ export DVAR. On one hand, environmental decentralization reduces enterprises’ export DVAR through cost markups, while on the other hand, it enhances enterprises’ export DVAR through research and development innovation. This conclusion is consistent with related studies (Chakraborty and Chatterjee 2017; De Loecker and Van Biesebroeck, 2018; Jiang et al. 2023; Duan and Xia, 2022). Robustness analysis considers endogeneity, indicator robustness, and model robustness, and the core conclusion of this study remains robust.

The findings of this study hold significant implications for policy development and refinement. A balanced approach to environmental and administrative decentralization, alongside enhanced monitoring, can improve the export profitability of enterprises. Promptly advancing the development and reform of China’s environmental decentralization system stands out as a key strategy to boost the profitability of Chinese export businesses. In the current context of a decelerating export growth rate in China, widespread international trade protectionism, and escalating trade disputes among nations, fostering China’s green and sustainable growth, as well as promoting high-quality foreign trade development, will be crucial for future policy adjustments.