Command-driven vs. market-oriented environmental regulations: impacts on high-quality development of manufacturing industry

The manufacturing industry is a key area of environmental regulation. However, whether command-driven and market-oriented environmental regulations exert heterogeneous impacts on the high-quality development of the manufacturing industry (HQDM) remains underexplored. This study treats the command-driven low-carbon city pilot policy and the market-oriented carbon emissions trading pilot policy as “quasi-natural experiments”. Firm-level data of listed manufacturing enterprises spanning 2003–2021, it adopts the double machine learning method to evaluate the influence of heterogeneous environmental regulations on the HQDM. The findings show that the low-carbon city pilot policy significantly inhibits the HQDM, whereas the carbon emissions trading pilot policy significantly promotes it. The effect of market-oriented environmental regulation on the HQDM is primarily achieved through the mechanism of technological innovation. In regions where both the low-carbon city pilot policy and the carbon emissions trading pilot policy are implemented, both command-driven and market-oriented regulations boost the HQDM, signifying a synergistic effect between them. Further heterogeneity analysis shows that the results for eastern and western areas, state -owned firms, and technology-intensive manufacturing sectors align with the baseline regression results. The conclusions of this study provide important references for the selection of carbon reduction policies, formulating differentiated emission reduction measures.

1 Introduction

Manufacturing stands as the mainstay of the national economy. Over time, China has evolved into the world’s foremost manufacturing powerhouse (Guo and Sun, 2023). However, products produced by the manufacturing industry are often carbon-intensive (Naegele and Zaklan, 2019), making it a major source of carbon emissions (Cao et al., 2021). As a result, the manufacturing industry has become a key field for environmental regulation and low-carbon transformation.

Environmental regulation plays a crucial role in balancing economic development and environmental protection (Ai et al., 2020; Liu et al., 2021). In recent years, curbing carbon emissions has evolved into a shared global objective. As the world’s top energy consumer and carbon emitter (Fan et al., 2016), During the 75th United Nations General Assembly, China pledged to reach carbon peak by 2030 and attain carbon neutrality by 2060, demonstrating its commitment as a responsible major power to advancing low-carbon transformation. A series of regulatory measures have been implemented, such as the implementation of low-carbon city pilot policy (LCCP, hereinafter referred to as such) and carbon emissions trading pilot policy (CETP, hereinafter referred to as such). the LCCP are typical command-driven environmental regulation, while the CETP are typical market-oriented ones. Their implementation marks a shift in China’s carbon emission regulations from purely command-driven to a combination of command-driven and market-oriented approaches.

An urgent question is whether carbon emission regulations promote the HQDM. How do command-driven and market-oriented environmental regulations ¹ differ in their effects on the HQDM? Which type of environmental regulation can more effectively balance low-carbon transition and high-quality development?

To address the above issues, this study uses the command-driven LCCP and market-driven CETP as quasi-natural experiments. Based on microdata from manufacturing listed firms spanning 2003–2021,it applies double machine learning to assess how heterogeneous environmental regulations impact the HQDM. Specifically, What are the similarities and differences in their transmission mechanisms? Do regions with both policies perform better than those with a single policy, resulting in a “1 + 1>2″synergy effect?Scientifically answering these questions and clarifying the economic effects and mechanisms of different low-carbon transition methods have significant theoretical implications. Moreover, in the context of global warming, examining the impact of command-driven and market-oriented carbon emission regulations on the HQDM holds considerable practical significance. Empirically, the LCCP hampers the HQDM, whereas the CETP bolsters it—implying market-oriented regulations outperform administrative ones.

This paper potentially makes two marginal contributions: Firstly, it investigates the distinct impacts of command-driven and market-oriented environmental regulations on the HQDM by leveraging China’s LCCP and CETP as quasi-experimental settings. Unlike previous studies that primarily used proxy variables to measure heterogeneous environmental regulations, this approach avoids subjective bias in the selection of measurement indicators. Moreover, it enriches the research on the economic effect of environmental regulation by considering both LCCP and CETP together. Secondly,It employs the double machine learning (DML) method to identify the policy effects of heterogeneous environmental regulations. the DML is particularly advantageous in high-dimensional settings with complex covariate structures, which can avoids the pre-selection of control variables (Yang J. et al., 2020; Zhang et al., 2022; Bodory et al., 2022; Farbmacher et al., 2022). This differs from previous methods that primarily relied on approaches such as DID, PSM-DID, triple difference (DDD), and spatial DID (SDID).

2 Literature review

The study primarily covers literature on the economic effects of command-drived and market - oriented environmental regulations, as well as methods for evaluating policy effectiveness.

Scholars have extensively investigated the economic effects of environmental regulation, but academia has yet to reach a consensus, resulting in two opposing hypotheses. The first is the “compliance cost” theory proposed by neoclassical economics, which contends that environmental regulations raise firms’ production and operational costs, hampering economic growth (Gray, 1987; Lanoie et al., 2011). The second is the “innovation compensation” hypothesis proposed by Porter, which believes that appropriate environmental regulation not only avoids negative impacts on the economy but can also promote technological innovation in enterprises and stimulate economic growth (Porter and Linde, 1995). Some scholars’ research supports this perspective (Zhang et al., 2011; Zhou and Tang, 2021; Chen et al., 2022; Zheng et al., 2023). Environmental regulations are generally classified into command-drived and market-oriented regulation. This paper will next review the economic effects of these two types of environmental regulations.

2.1 The economic effects of command-drived and market-oriented environmental regulations

Existing studies on command-drived environmental regulation focus on evaluating specific policies, using quasi-natural experiment designs and methods such as DID,DDD and PSM-DID. For instance, Zhang and Zhao (2023) found sulfur dioxide emission control policies enhanced firms’ technological innovation; Shao et al. (2024) showed the Top 10,000 Energy - Intensive Enterprises Initiative reduced pollutant emissions; Cai et al. (2016) revealed the Two-Zone Emission Control Policy lowered FDI. Regarding low-carbon pilot policies, Chen et al. (2021) and Qiu et al. (2021) demonstrated positive impacts on firms'TFP and cities'GTFP. Basu et al. (2025) demonstrated that under the policy frameworks of the U.S. Superfund Program and the Clean Air Act, relevant environmental regulations significantly reduced the average pollution exposure levels among the elderly population.

The aforementioned studies emphasize the positive socioeconomic effects of command-driven environmental regulations, other research has uncovered their negative impacts. For instance, Carril-Caccia and Milgram Baleix (2024) found that stricter environmental regulations (ER) reduce countries’ attractiveness to foreign investors; Benatti et al. (2024) observed that tightened ER negatively affects the productivity growth of high-polluting firms. Agarwal et al. (2019) used the NOx Budget Trading Program (NBP) as a quasi-natural experiment, finding that housing markets in regulated regions with high manufacturing density were sluggish. Zeng et al. (2023) noted that stricter environmental law enforcement significantly reduced the TFP of high air-polluting enterprises, and Huang et al. (2022) further found that tightened environmental supervision lowered manufacturing firms’ productivity. Still other studies have reported no significant effects: Alpay and Kerkvliet (2020) found U.S. pollution regulations had no impact on food manufacturing productivity; Wang et al. (2018) observed water quality regulations exerted no notable effect on the productivity of surviving firms; Lange and Redlinger (2019) found oil sector regulatory policies did not significantly affect oil and gas drilling and production speeds. Evidently,the socioeconomic effects of command-driven environmental regulations remain uncertain, as they may be positive, negative, or non-significant.

Market-oriented environmental regulations research mainly assesses emissions trading and carbon taxes. Dechezleprêtre (2016) found the EU Emissions Trading Scheme (EU ETS) spurred innovation in carbon - mitigation technologies; Naegele and Zaklan (2019) detected no carbon leakage triggered by the EU ETS; Clarkson et al. (2015) linked carbon allowance shortfalls to reduced firm value; Martin et al. (2014) noted carbon taxes had no significant effect on manufacturing revenue. For carbon emissions trading policies, Fan et al. (2016) highlighted efficiency improvements, while Cao et al. (2021) found no impact on coal-fired power plants’ coal efficiency. Similarly, studies by domestic and international scholars indicate that market-oriented environmental regulations often yield significant economic effects.

In comparative studies of heterogeneous regulations, proxy variables are commonly employed (e.g., Yang and Green, 2024, using fiscal expenditure ratios for command-driven regulation and pollution fee ratios for market-oriented), which are not only subjective but also inherently susceptible to measurement bias. Additionally, existing literature examines low-carbon initiatives and carbon trading separately, lacking simultaneous analysis of their differential effects. This paper addresses these gaps by treating two policies—the command-driven LCCP and market-oriented CETP—with unified goals—as quasi-natural experiments, avoiding measurement subjectivity to better explore their differential impacts.

2.2 Methods for evaluating policy effectiveness

The methods for evaluating the effects of environmental regulation mentioned above primarily use difference methods such as DID (Zhang and Zhao, 2023; Shao et al., 2024; Qiu et al., 2021), DDD (Cai et al., 2016), and PSM-DID (Chen et al., 2021). These traditional policy evaluation methods share the same limitation: they have restrictions on the dimensions of control variables.

Currently, machine learning methods are rapidly permeating the field of econometrics and have also been increasingly adopted for event evaluation. Chernozhukov et al. (2018) pioneered the Double Machine Learning (DML) for event evaluation. Subsequently, Yang J. et al. (2020), Zhang et al. (2022), Bodory et al. (2022), and Farbmacher et al. (2022) applied DML in policy evaluation. the DML excels in high - dimensional variable screening, as it circumvents multicollinearity pitfalls inherent in traditional methods when dealing with high - dimensional control variables (Zhang et al., 2022).

3 Theoretical mechanisms

Firstly, the paper analyzes the influence of LCCP and CETP on the HQDM. The specific analysis is as follows:The LCCP constrains pollution emissions from local enterprises by establishing greenhouse gas emission data statistics and management systems and implementing a target-based responsibility system. These measures will directly increase the environmental management costs for businesses. Meanwhile, administrative orders can act as a market signal, leading to an excessive concentration of resources and potential resource misallocation. Furthermore,the LCCP may overly focus on the carbon reduction outcomes, lacking a comprehensive consideration of economic impacts, which could inhibit the growth of the manufacturing industry. Therefore, We posit that the LCCP could dampen manufacturing'TFP and impede the HQDM.

In contrast, carbon trading’s theoretical foundation of carbon emissions trading lies in Coase’s property rights theory. The CETP allocates initial free allowances via historical emissions or carbon intensity. Under the framework of total emission control, carbon emission rights are treated as tradable commodities, enabling market exchanges to incentivize firms to internalize carbon externalities. This mechanism facilitates Pareto - efficient resource allocation. Thus, this paper suggests that CETP can enhance manufacturing’s TFP and promote its high-quality development.

When a region implements both the CETP and the LCCP simultaneously, the two policies generate a positive synergistic effect on the HQDM; more precisely, after the CETP is formulated,the LCCP’s impact on the HQDM shifts from negative to positive. The formation of this synergistic effect stems primarily from two aspects: first, the combination of the two low-carbon policies can strengthen the publicity and guidance for low-carbon development, deepen low-carbon awareness, and thereby enhance the overall policy implementation effect; second, it can construct a “government-guided, market-led” collaborative governance mechanism—the CETP plays a core role in guiding the optimal allocation of carbon emission rights, while the LCCP serves an auxiliary function to address market failures. Meanwhile, the CETP provides market entities with more emission reduction options, which can alleviate the potential “one-size-fits-all” issue caused by LCCP’s administrative intervention, reduce its strong impact on market entities, and ultimately enable the LCCP to promote the HQDM while achieving effective emission reduction. In conclusion, in regions where the CETP and the LCCP are implemented simultaneously, these two types of environmental regulations exert a synergistic effect on the HQDM, with both contributing to the improvement of the HQDM.

Secondly, this paper analyzes the influence mechanisms of LCCP and CETP on the HQDM from the perspectives of innovation effects and resource effects. The specific analysis is as follows.

3.1 Innovation effect

Innovative activities, characterized by long cycles, high investment, and significant risks (Gustavo, 2011), are susceptible to government fiscal incentives and support policies (Borghesi et al., 2015; Montmartin and Herrera, 2015). Both the LCCP and the CETP exert dualistic impacts on the innovation effect of HQDM.

From the “compliance cost perspective”, negative effects manifest as follows:As a command-drived environmental regulation,The LCCP pursues emission reduction by adjusting the energy structures, enhancing energy efficiency, and promoting energy conservation. It also focuses on establishing a low-carbon production system and advocating for low-carbon production methods. These measures will directly increase enterprises’ compliance costs. In carbon emission trading,if enterprises have insufficient quotas and need to purchase from the market. such carbon trading expenditures can crowd out R&D investment, weakening firms’ technological innovation capacity. The increase in costs may crowd out investment in technological innovation, leading to a decline in the innovation capabilities of enterprises.

From the “innovation compensation perspective”, the positive impact is manifested as follows: In the face of cost pressures, enterprises’ investment in innovation may not decrease but may instead increase (Bu et al., 2019). Firms can mitigate policy - induced cost burdens via technological innovation, generating an innovation - offset dynamic that bolsters enterprises’ TFP. In addition, productivity improvement also depends on enterprises’ ability to apply existing innovations and efficient technologies (Albrizio et al., 2017; Daron et al., 2006). The LCCP fosters a collaborative ecosystem for low - carbon tech scaling, facilitating knowledge spillovers and technology diffusion. This action will encourage companies to learn advanced technologies from their peers and expand the use of high-tech innovations (Wang et al., 2018). Unlike command-drived environmental regulation, carbon emission trading’s distinctive edge lies in capping total carbon emissions yet enabling firms to monetize surplus allowances (Clarkson et al., 2015), thus creating enduring economic incentives for technological innovation.

3.2 Resource effect

The LCCP exerts both negative and positive impacts on resource allocation for the HQDM.

In terms of negative effects, as a command-driven environmental regulation, its administrative intervention may cause excessive resource allocation to low-carbon industries and sectors. Driven by the policy’s low-carbon orientation and carbon reduction assessment pressures, enterprises might over-invest in non-productive emission reduction equipment (He et al., 2020), leading to resource misallocation due to excessive resource shifts to low-carbon sectors. Moreover, drawing on the rent-seeking theory (Murphy et al., 1993) and the “grabbing hand” theory of government (Andrei and Vishny, 1994), enterprises may proactively cater to government demands to gain local government support or even allocate resources under government “domination”, which further exacerbates resource misallocation.

On the positive side, improved factor utilization efficiency is reflected in the flow of production factors from inefficient to efficient sectors or enterprises. For individual enterprises, both policies encourage increased investment in low-carbon, high-efficiency sectors, and clean energy use while reducing input in high-carbon, inefficient sectors and fossil fuel consumption (Chen et al., 2021), thereby enhancing internal resource allocation efficiency. At the industry level, the LCCP imposes stringent emission and technical benchmarks, creating industry barriers and raising entry thresholds for manufacturing. It also increases competitive pressure on enterprises with high carbon reduction costs, forcing some to exit. This leads to resource concentration in enterprises with lower carbon reduction costs and higher productivity, promoting overall industry productivity growth (Cheng et al., 2019).

The CETP bolsters firms’ emission - reduction flexibility by providing clear market price signals (Albrizio et al., 2017). Specifically, market trading reallocates carbon emission rights among enterprises, production factors are redirected from inefficient firms to their high - efficiency counterparts. This optimizes quota distribution, enabling cost - minimal emission reduction (Montgomery, 1972). Notably, the resource effect of carbon emission rights depends on market liquidity. Given China’s carbon market remains in a nascent phase with incomplete mechanisms, it may not yet have a significant resource allocation effect.

In summary, the LCCP may cause resource misallocation (inhibiting HQDM), optimize resource allocation (promoting HQDM), or have no significant impact (with coexisting effects). The CETP can optimize resource allocation to boost HQDM, but its resource allocation effect may be insignificant in the early stage due to immature market mechanisms. The actual policy effects require further verification.

In summary, The influence mechanism of heterogeneous environmental regulations on HQDM are shown in Figure 1.

Figure 1

Figure 1. The influencet mechanism of heterogeneous environmental regulations on the HQDM.

4 Methodology

4.1 Policy background, sample selection, and data sources

In September 2020, during the 75th UN General Assembly, China announced its 2030 carbon peak and 2060 carbon neutrality targets, emphasizing a commitment to advance low-carbon transitions across economic and social spheres. In response, initiatives like LCCP and CETP have emerged as key policy tools.

The LCCP rolled out in three phases: the first in July 2010 (5 provinces, e.g., Guangdong, Liaoning; 8 cities, e.g., Tianjin, Chongqing); the second in November 2012 (Hainan Province and 28 cities, e.g., Beijing, Shanghai, Guangzhou); and the third in January 2017 (45 cities, e.g., Wuhai, Shenyang, Dalian). While specific goals differ marginally across phases, all aim to curb greenhouse gas emissions and pilot green, low - carbon development models. Key strategies include forging low - carbon industrial ecosystems, optimizing energy structures, advancing energy - efficient buildings and low - carbon mobility.

The CETP started in June 2013 with Shenzhen, subsequently expanding to six regions (e.g., Tianjin, Chongqing). Fujian joined in December 2016, making it the eighth pilot. The national carbon market began trading in July 2021. The policy implementation involves determining covered enterprises via historical emissions or carbon intensity benchmarks, allocating initial free allowances, and allowing enterprises to trade allowances in the market. Pilot markets have a 40%–60% carbon emissions coverage rate (Ma et al., 2023) and operate independently with no cross-market trading.

The research sample comprises A-share listed firms on Shanghai and Shenzhen exchanges, spanning 2003–2021. Including 2021 arises from the July 2021 launch of the national carbon market, with pilot markets coexisting during the transition period and policy effects being delayed (Ma et al., 2023). Abnormal samples (e.g., ST/*ST companies, those with significant data gaps) were excluded, and missing data were filled using linear interpolation. Data were sourced from CSMAR database and the China City Statistical Yearbook. The policy time nodes are set as 2010, 2013, and 2017 for the three low-carbon pilot batches; for carbon trading, Shenzhen starts in 2013, Beijing, Tianjin, Shanghai, Chongqing, Hubei, and Guangdong in 2014, and Fujian in 2017.

4.2 Research method

This study applies Double Machine Learning (DML) to assess how heterogeneous environmental regulations influence manufacturing high-quality development (HQDM). Unlike traditional nonparametric regression methods, DML accommodates high - dimensional nuisance functions with numerous covariates, obviating the need for pre - selecting control variables (Yang G. et al., 2020; Zhang et al., 2022; Bodory et al., 2022; Farbmacher et al., 2022).

Referring to Robinson (1988) and Chernozhukov et al. (2018), the following partial linear model is established.

$Y_{it}={\beta }_1{P}_{it}+{\mathit{g}}_{\mathbf{0}}\left(X_{it}\right)+{\mathit{U}}_{it},\mathbf{E}\left[{\mathit{U}}_{it}|X_{it},P_{it}\right]=\mathbf{0}(1)$

$P_{it}={\mathit{m}}_{\mathbf{0}}\left(X_{it}\right)+{\mathit{V}}_{it},\mathbf{E}\left[{\mathit{V}}_{it}|X_{it}\right]=\mathbf{0}(2)$

In the equation, $Y_{it}$ is the outcome variable; $P_{it}$ as both the explanatory variable and a policy dummy variable; X represents control variables. ${\mathit{g}}_{\mathbf{0}}\left(X_{it}\right)$ 、 ${\mathit{m}}_{\mathbf{0}}\left(X_{it}\right)$ are unknown functions estimated via machine learning. ${\mathit{U}}_{it}$ and ${\mathit{V}}_{it}$ are error terms with zero conditional means. Equation 1 is the main equation, and Equation 2 is the auxiliary equation.

By using machine learning to predict ${\mathit{g}}_{\mathbf{0}}\left(X_{it}\right)$ , we obtain its estimated value ${\widehat{g}}_0\left(X_{it}\right)$ , which is then substituted into Equation 1 to derive Equation 3:

$Y_{it}-{\widehat{g}}_0\left(X_{it}\right)={\beta }_1{P}_{it}+{error}_{it}(3)$

Then, using ordinary least squares (OLS), we obtain the OLS estimator ${\beta }_1$ for ${\widehat{\beta }}_1$

${\widehat{\beta }}_1={\left(\frac{1}{n}{\displaystyle \sum _{i=1}^n}{P_{it}}^2\right)}^{-1}\frac{1}{n}{\displaystyle \sum _{i=1}^n}{P}_{it}\left[Y_{it}-{\widehat{g}}_0\left(X_{it}\right)\right](4)$

Substituting $Y_{it}={\beta }_1{P}_{it}+{\mathit{g}}_{\mathbf{0}}\left(X_{it}\right)+{\mathit{U}}_{it}$ from Equation 1 into Equation 4 yields Equation 5:

$\begin{array}{c}{\widehat{\beta }}_1={\left(\frac{1}{n}{\displaystyle \sum _{i=1}^n}{P_{it}}^2\right)}^{-1}\frac{1}{n}{\displaystyle \sum _{i=1}^n}{P}_{it}\left[{\beta }_1{P}_{it}+{\mathit{g}}_{\mathbf{0}}\left(X_{it}\right)+{\mathit{U}}_{it}-{\widehat{g}}_0\left(X_{it}\right)\right]\\ ={\beta }_1+{\left(\frac{1}{n}{\displaystyle \sum _{i=1}^n}{P_{it}}^2\right)}^{-1}\frac{1}{n}{\displaystyle \sum _{i=1}^n}{P}_{it}{\mathit{U}}_{it}+{\left(\frac{1}{n}{\displaystyle \sum _{i=1}^n}{P_{it}}^2\right)}^{-1}\frac{1}{n}{\displaystyle \sum _{i=1}^n}{P}_{it}\left[{\mathit{g}}_{\mathbf{0}}\left(X_{it}\right)-{\widehat{g}}_0\left(X_{it}\right)\right]\end{array}(5)$

Rearranging terms gives:

$\sqrt{n}\left({\widehat{\beta }}_1-{\beta }_1\right)=\underset{=a}{\underbrace{{\left(\frac{1}{n}{\displaystyle \sum _{i=1}^n}{P_{it}}^2\right)}^{-1}\frac{1}{\sqrt{n}}{\displaystyle \sum _{i=1}^n}{P}_{it}{\mathit{U}}_{it}}}+\underset{=b}{\underbrace{{\left(\frac{1}{n}{\displaystyle \sum _{i=1}^n}{P_{it}}^2\right)}^{-1}\frac{1}{\sqrt{n}}{\displaystyle \sum _{i=1}^n}{P}_{it}\left[{\mathit{g}}_{\mathbf{0}}\left(X_{it}\right)-{\widehat{g}}_0\left(X_{it}\right)\right]}}(6)$

In the equation, the term “a” adheres to a normal distribution centered at zero. Substituting $P_{it}={\mathit{m}}_{\mathbf{0}}\left(X_{it}\right)+{\mathit{V}}_{it}$ which is from Equation 2 into the “b” part of Equation 6 yields Equation 7:

$\mathbf{b}={\left(\mathbf{E}{P_{it}}^2\right)}^{-1}\frac{1}{\sqrt{n}}{\displaystyle \sum _{i=1}^n}{m}_0\left(X_{it}\right)\left[{\mathit{g}}_{\mathbf{0}}\left(X_{it}\right)-{\widehat{g}}_0\left(X_{it}\right)\right]+0_p\left(1\right)(7)$

Due to regularization bias in machine learning, the convergence rate of ${\mathit{g}}_{\mathbf{0}}\left(X_{it}\right)-{\widehat{g}}_0\left(X_{it}\right)$ to 0 is typically slower than that of $\frac{1}{\sqrt{n}}$ , with its convergence rate denoted as $n^{-{\phi }_g}$ , i.e., ${\phi }_g<\frac{1}{2}$ . Since $m_0\left(X_{it}\right)$ ≠ 0, as $n$ → $\infty$ , the magnitude of “ $\mathbf{b}$ ”, which is $\sqrt{n}{n}^{-{\phi }_g}$ → $\infty$ . Consequently, even with large samples, $\sqrt{n}\left({\widehat{\beta }}_1-{\beta }_1\right)$ will not converge to 0 in probability, and ${\widehat{\beta }}_1$ will not converge to ${\beta }_1$ . This results in a bias in the estimation of ${\beta }_1$ . To correct this estimation bias, orthogonalization methods need to be introduced.

Specifically, first use machine learning to predict the residuals ${\mathit{V}}_{it}$ of auxiliary Equation 2 again to obtain their estimates ${\widehat{V}}_{it}$ . Then, treat ${\widehat{V}}_{it}$ as an instrumental variable for $P_{it}$ to finally obtain the estimate ${\beta }_1$ for ${\widehat{\beta }}_1$ , as shown in Equation 8.

${\widehat{\beta }}_1={\left(\frac{1}{n}{\displaystyle \sum _{i=1}^n}{{\widehat{V}}_{it}}^2\right)}^{-1}\frac{1}{n}{\displaystyle \sum _{i=1}^n}{\widehat{V}}_{it}\left[Y_{it}-{\widehat{g}}_0\left(X_{it}\right)\right](8)$

We then obtain Equation 9:

$\sqrt{n}\left({\widehat{\beta }}_1-{\beta }_1\right)=\underset{=a}{\underbrace{{\left(\frac{1}{n}{\displaystyle \sum _{i=1}^n}{{\widehat{V}}_{it}}^2\right)}^{-1}\frac{1}{\sqrt{n}}{\displaystyle \sum _{i=1}^n}{\widehat{V}}_{it}{\mathit{U}}_{it}}}+\underset{=b}{\underbrace{{\left(\frac{1}{n}{\displaystyle \sum _{i=1}^n}{{\widehat{V}}_{it}}^2\right)}^{-1}\frac{1}{\sqrt{n}}{\displaystyle \sum _{i=1}^n}\left[m_0\left(X_{it}\right)-\widehat{m_0}\left(X_{it}\right)\right]\left[{\mathit{g}}_{\mathbf{0}}\left(X_{it}\right)-{\widehat{g}}_0\left(X_{it}\right)\right]}}(9)$

In the equation, the term “a” adheres to a normal distribution centered at zero and $\left[m_0\left(X_{it}\right)-\widehat{m_0}\left(X_{it}\right)\right]$ and $\left[{\mathit{g}}_{\mathbf{0}}\left(X_{it}\right)-{\widehat{g}}_0\left(X_{it}\right)\right]$ represent the prediction biases of machine learning concerning $m_0\left(X_{it}\right)$ and ${\mathit{g}}_{\mathbf{0}}\left(X_{it}\right)$ . Let the convergence rate of $\left[m_0\left(X_{it}\right)-\widehat{m_0}\left(X_{it}\right)\right]$ is $n^{-{\phi }_m}$ , then the convergence rate of $\left[{\mathit{g}}_{\mathbf{0}}\left(X_{it}\right)-{\widehat{g}}_0\left(X_{it}\right)\right]$ is $n^{-{\phi }_g}$ . The convergence rate of the interaction term of the two error terms $\left[m_0\left(X_{it}\right)-\widehat{m_0}\left(X_{it}\right)\right]\left[{\mathit{g}}_{\mathbf{0}}\left(X_{it}\right)-{\widehat{g}}_0\left(X_{it}\right)\right]$ is $n^{-\left({\phi }_m+{\phi }_g\right)}$ which is faster than that of a single prediction bias, either $n^{-{\phi }_m}$ or $n^{-{\phi }_g}$ . In this case, it is also faster than $\frac{1}{\sqrt{n}}$ , meaning ${\phi }_m+{\phi }_g>\frac{1}{2}$ . As $n$ → $\infty$ , the magnitude of ${}^{″}\mathbf{b}^{″}$ approaches $\sqrt{n}{n}^{-\left({\phi }_m+{\phi }_g\right)}$ →0, causing $\sqrt{n}\left({\widehat{\beta }}_1-{\beta }_1\right)$ to converge to 0 in probability, making ${\widehat{\beta }}_1$ an unbiased estimator of ${\beta }_1$ . In summary, predicting ${\mathit{g}}_{\mathbf{0}}\left(X_{it}\right)$ and ${\mathit{V}}_{it}$ , via double machine learning (DML) circumvents the dimensionality curse induced by high - dimensional controls, yielding an unbiased estimator of ${\beta }_1$ .

DML boosts estimation precision via cross - fitting: by partitioning the sample, one subset trains models to predict ${\mathit{g}}_{\mathbf{0}}\left(X_{it}\right)$ and residuals ${\mathit{V}}_{it}$ , (generating ${\widehat{g}}_0\left(X_{it}\right)$ and ${\widehat{V}}_{it}$ ), while the other subset conducts regression. For partially linear models where covariate functional forms are unknown, DML yields unbiased treatment effect estimates.

4.3 Variable selection

4.3.1 Dependent variable

In this paper, the dependent variable is the indicator of high-quality development of the manufacturing industry (HQDM), which is measured by total factor productivity (TFP).

4.3.2 Explanatory variable

Key policy indicators—Tlowc (LCCP) and Tctrade (CETP)—act as dummy variables. Drawing on prior research (Demir et al., 2022; Teng et al., 2022; Filippini et al., 2020; Oberfield, 2013; Yang J. et al., 2020), controls cover firm - (micro) and city - level (macro) dimensions. Firm - level controls (14 variables) capture financial attributes (e.g., Age, Size, Rate), while city - level metrics (16 items) span regional growth (Gdp), government spending (Gov), and market conditions (Mkr). Variable definitions and summary statistics are presented in Tables 1, 2.

Table 1

Table 1. Key variable definitions.

Table 2

Table 2. Descriptive statistics of key variables.

5 Empirical results

5.1 Benchmark regression tests

Drawing on partial linear model insights, we partition the sample into a 1:4 split (machine learning prediction vs. regression estimation). Following Yang G. et al. (2020), gradient boosting outperforms other algorithms, so we adopt it for policy assessment. Results appear in Table 3.

Table 3

Table 3. Impact of heterogeneous environmental regulations on the HQDM.

In baseline regressions (Columns 1 and 3), Tlowc exhibits a statistically significant negative effect on listed manufacturers’ TFP (5% level), whereas Tctrade shows a positive effect (1% level). When adding interaction and quadratic control terms (Columns 2 and 4):Tlowc’s coefficient is −0.0386 (1% significance) implies a 0.0386 average TFP reduction in pilot regions relative to non-pilots;Tctrade’s coefficient is 0.0975 (1% significance), which indicates the average TFP of relevant listed manufacturing firms is 0.0975 higher than that of the reference group.

These results suggest command-driven environmental regulations outperform command-driven measures in fostering manufacturing high - quality development (HQDM).

5.2 Robust tests

5.2.1 Sample screening

To address outliers and special samples, we employ trimming and exclude central municipality observations. Tables 4, 5 show Tlowc remains negatively significant and Tctrade is significantly positive, It is consistent with the baseline findings and confirming result robustness.

Table 4

Table 4. Sample trimming estimation results.

Table 5

Table 5. Estimation results after removing municipalities.

5.2.2 Resetting the double machine learning model

This study adopts three methods for robustness checks: adjusting the sample split ratio, algorithm replacement, and model substitution. Specifically, following Chernozhukov et al. (2018), the sample split ratio is modified from 1:4 to 1:3. For the algorithm replacement, the gradient boosting (gradboost) algorithm is replaced with a stacked regression algorithm—this algorithm assesses policies through a weighted combination of multiple machine learning methods, including gradient boosting, random forests (rf), lasso (lassocv), ridge regression (ridgecv), and neural networks (nnet). Furthermore, the model shifts from a local linear model to a more general interactive model.

Results in Table 6 show Tlowc remains significantly negative, while that of Tctrade stays significantly positive, validating baseline result robustness.

Table 6

Table 6. Estimation results for resetting the double machine learning model.

5.2.3 Excluding the impact of simultaneous implementation of two environmental regulations

The implementation timelines for the LCCP and the CETP overlap across multiple periods, and both environmental regulations may significantly impact the TFP of the manufacturing industry. To avoid the issue of omitting important variables when evaluating the policies separately, this paper will further conduct robustness analysis using subsamples. we divide the sample into four policy-assessment groups, with robustness test results in Table 7.

Table 7

Table 7. Estimation results for excluding the impact of simultaneous implementation of two environmental regulations.

First, Remove cities with the CETP, leaving only low-carbon pilot cities as the treatment group (results in Table 7, columns (1)–(2)).

Second, Remove cities with the LCCP, leaving only carbon emissions trading cities as the treatment group (results in columns (3)–(4)).

Third, we remove cities covered by both policies, and then take cities covered by only one of the two policies as the treatment group (results in columns (5)–(6)).

Finally, Remove cities with only one policy, including only cities with both policies in the treatment group (results in columns (7)–(8)).

From the subsample empirical results in columns (1) to (6), it confirm baseline robustness. Columns (7) and (8) show that Tlowc’s coefficient remains significantly positive, indicating that the LCCP significantly promotes the HQDM. The coefficient for Tctrade is positive but not significant, implying the CETP have no notable impact on HQDM.

One plausible reason is that cities implementing both low-carbon policies are more conducive to creating an environment for carbon reduction, increasing companies’ acceptance and motivation for carbon reduction, and generating a synergistic effect between the two policies. Consequently, the LCCP positively contributes to the HQDM. However,due to the implementation of the LCCP, companies are required to reduce carbon emissions, which decreases their carbon quota trading in the carbon market. This weakens the impact of the CETP, making its effect on HQDM insignificant.

6 Mechanism and heterogeneity analysis

6.1 Mechanism analysis

6.1.1 Innovation effect

This paper uses R&D intensity (R&D expenditure/operating revenue) as a proxy for technological innovation (Zhou and Tang, 2021). Given the lack of R&D expenditure data prior to 2009, the mechanism test focuses on the second and third pilot policy batches, with corresponding results provided in Table 8.

Table 8

Table 8. Estimation results of mechanism testing.

Table 8, Columns 1–2 assess the 2012 and 2017 LCCP’s effects on technological innovation. Results indicate that, after adding the two policy variables, whether controlling for interaction and quadratic terms or not,the Tlowc’s coefficient is insignificant. This indicates that the LCCP has no significant impact on technological innovation in manufacturing. In contrast, Tctrade’s positive significance suggests the CETP spurs technological innovation. Consistent with prior mechanism logic, the policy’s innovation-inducing effect outweighs its crowding-out effect, driving technological innovation.

6.1.2 Resource effect

This study proxy resource allocation efficiency with capital allocation efficiency (Chen et al., 2021), adopting the “investment level and investment opportunity sensitivity” model to assess heterogeneous environmental regulations’ impact. The model is specified in Equations 10–12:

$A_{it}{=P}_{it}\times {Roa}_{it}(10)$

${Invest}_{it}={\beta }_1{A}_{it}+{\mathit{g}}_{\mathbf{0}}\left(X_{it}\right)+{\mathit{U}}_{it},\mathbf{E}\left[{\mathit{U}}_{it}|X_{it},A_{it}\right]=\mathbf{0}(11)$

$A_{it}={\mathit{m}}_{\mathbf{0}}\left(X_{it}\right)+{\mathit{V}}_{it},\mathbf{E}\left[{\mathit{V}}_{it}|X_{it}\right]=\mathbf{0}(12)$

Here, ${Invest}_{it}$ represents the firm’s investment level, ${Roa}_{it}$ is the return on assets, which captures the firm’s investment opportunities. $P_{it}$ is the policy dummy variable. The term $A_{it}$ is the interaction term of $P_{it}$ and ${Roa}_{it}$ , and the definitions of other variables are the same as in models (1) and (2).

The empirical results, presented in columns (5) to (8) of Table 8, indicate insignificant coefficients for Tlowc and Tctrade. This suggests that the LCCP and the CETP do not have a significant impact on firms’ capital allocation efficiency.

Regarding the non-significant impact of the LCCP on firms’ capital allocation efficiency, potential explanations can be derived from mechanism analysis. As a command-driven environmental regulation, the LCCP may cause resource misallocation: government administrative intervention, coupled with firms’ proactive compliance to obtain local government support, ultimately leads to excessive resource reallocation toward low-carbon sectors. At the same time, it can also optimize factor allocation—both within individual firms and across industries—by driving factors to flow from inefficient sectors or enterprises to efficient ones. These two opposing effects, which exert negative impacts and positive impacts on resource allocation for the HQDM, likely offset each other, ultimately resulting in the LCCP’s non-significant influence on capital allocation.

The CETP also exerts no significant impact on firms’ capital allocation efficiency. A plausible explanation, similarly as elaborated in the mechanism analysis, lies in that CETP—a market-based environmental regulation—can optimize resource allocation and reduce carbon emissions through market price signals of carbon emission rights, but this depends on the maturity of market mechanisms. Since China’s carbon emission trading market is still in the initial stage of development, characterized by underdeveloped market mechanisms, and its pilot carbon markets operate independently of one another with no cross-market carbon quota trading permitted—resulting in its limited market coverage—these are the reasons why the CETP fails to generate a significant resource allocation effect. Meanwhile, Carbon emission rights are quasi-public goods, and the carbon market is a market with quasi-public goods (i.e., carbon emission rights) as its core trading instrument. Given that the non-rival nature of quasi-public goods requires government intervention in supply via“total quantity control”and the definition of exclusive property rights relies on administrative rules, the carbon market cannot meet the core assumptions of a perfectly competitive market—namely,“free entry and exit, homogeneous products, and complete information.“It thus does not qualify as a perfectly competitive market, and its resource allocation efficiency is also constrained.

6.2 Heterogeneity analysis

This paper has demonstrated that the command-driven LCCP inhibits HQDM, while the market-oriented CETP promotes it. Given policy effects differ across firms by ownership and technological intensity, we further explore heterogeneity in these dimensions.

6.2.1 Firm ownership

This paper stratify firms by ownership—specifically state-owned enterprises (SOEs) and non-state-owned enterprises (nonSOEs)— for regression, with results in Table 9.

Table 9

Table 9. Estimation results of heterogeneity in Enterprise’s ownership.

Table 9 Columns 1–3 show Tlowc remains significantly negative at the 1% level, regardless of whether other interaction terms and quadratic terms are controlled, indicating the LCCP hinders the HQDM of SOEs.Meanwhile, Tctrade’s coefficient is significantly positive at the 5% level, suggesting CETP boosts HQDM of SOEs. In Columns 4–6, both coefficients are insignificant, indicating that, neither the LCCP nor the CETP exerts a significant impact on the HQDM of non-SOEs.

The possible reason for the heterogeneous effects is that SOEs exhibit a heightened commitment to social responsibilities. and are more proactive in responding to policies. As a result, they implement national policy decisions more effectively, leading to significant policy effects. Conversely, non-SOEs are more focused on profits and have weaker enforcement of environmental regulations, resulting in less significant policy effects.

6.2.2 Technical attributes

Based on 2012 CSRC industry classification, manufacturing is categorized into labor-intensive, technology-intensive, and capital-intensive sectors.

Columns (2), (4), and (6) of Table 10 report results controlling for interaction and quadratic terms. Only Tlowc in Column 4 is negatively significant, meaning the LCCP solely exerts a significant negative impact solely on technology-intensive manufacturing, while it has no significant impact on labor-intensive or capital-intensive manufacturing.

Table 10

Table 10. Estimation results of heterogeneity in technical attribute.

Columns (1) and (2) of Table 10 show that the Tctrade’s coefficient is significantly negative, suggesting that the CETP inhibits the TFP of labor-intensive manufacturing industries. Columns (3) and (4) reveal that the Tctrade’s coefficient is significantly positive, indicating that the CETP promotes the TFP of technology-intensive manufacturing industries. In contrast, column (6) shows that the Tctrade’s coefficient is insignificant, implying that the CETP has no significant promoting effect on the TFP of capital-intensive manufacturing industries. A plausible explanation for these findings is that labor-intensive manufacturing relies heavily on labor rather than technology or equipment and exhibits relatively low profitability. Consequently, compared with capital-intensive and technology-intensive enterprises, labor-intensive manufacturing industries face greater difficulties in improving TFP through technological innovation.

7 Conclusions and policy recommendations

Carbon emissions have become a major challenge facing the globe (Wu et al., 2025), Carbon emission reduction is a core global issue related to humanity’s sustainable development. Carbon regulations encompass diverse instruments such as command-driven and market-oriented mechanisms, among which carbon markets have been established and implemented in countries like the UK and the US,accumulating phased practical experience. This paper focuses on practical cases of China’s low-carbon policies—specifically the command-driven LCCP and the market-oriented CETP. This study empirically examines their impact on the HQDM and the underlying mechanisms. The analysis is based on microdata of listed manufacturing firms spanning from 2003 to 2021, employing the double machine learning method,and reaches the following research findings.

1. The LCCP significantly inhibits the HQDM, whereas the CETP notably promotes it. Findings stay robust across robustness tests.

2. In regions where both policies are implemented simultaneously, they both enhance the HQDM, demonstrating a synergistic effect.

3. Further mechanism analysis reveals that the CETP can significantly boost technological innovation in listed manufacturing firms yet exerts no notable influence on resource allocation efficiency. Thus, the improvement in TFP of these manufacturing firms is primarily driven by the enhancement of technological innovation. In contrast, the LCCP has no significant influence on either technological innovation or resource allocation efficiency.

4. Heterogeneity tests shows that the regression results for the eastern and western regions, SOEs, and technology-intensive manufacturing are similar to the baseline results. However, the CETP significantly hinders HQDM in the central region and labor-intensive manufacturing. The LCCP has no significant effect on non-SOEs or capital-intensive manufacturing.

Drawing on China’s practical experiences, this paper has arrived at the aforementioned research findings and puts forward the following policy recommendations—all in an effort to offer valuable insights for the selection and formulation of carbon emission reduction approaches globally, particularly among developing countries.

7.1 Strengthen the precision of command-driven policies to avoid a “one-size-fits-all” approach

The LCCP exert an overall inhibitory effect on the HQDM. This may be attributed to the strong administrative intervention imposed on the manufacturing industry during LCCP implementation—excessive intervention, to a certain extent, constrains the HQDM. Based on this, when formulating command-driven policies such as LCCP, it is necessary to prioritize the precision of policy objectives: fully take into account the development stages and actual capacity of different industries, formulate differentiated carbon reduction targets, and accompany them with targeted support measures. This will help avoid inappropriate inhibition on the development of the manufacturing industry caused by “one-size-fits-all” policy implementation.

7.2 Prioritize market-oriented environmental regulations as the core to establish a “government-guided, market-led” collaborative governance mechanism

The paper finds that, unlike LCCP, which significantly hinders HQDM, the CETP significantly promotes it. In regions with “dual carbon” policies, both command-driven and market-oriented regulations enhance HQDM, indicating a synergistic effect. It is important to focus on developing market-oriented environmental regulations and improving related mechanisms to better enhance the synergy between the market and government in environmental governance.

The market-oriented environmental regulations, such as carbon emission trading and carbon tax, can be adopted as core tools, while command-driven environmental regulations serve only as supplementary means to address market failures. Some developing countries may first establish basic rules for low-carbon transition (e.g., setting industry-specific emission reduction targets) through command-driven environmental regulations, and then gradually introduce market mechanisms to avoid long-term reliance on administrative intervention. Additionally, In the process of carbon market development, efforts should be made to promote the legislation of the Carbon Market Management Regulations, while formulating rules for the registration, settlement and compliance of carbon allowances; a “Special Fund for Carbon Market Development” should be established to support the exploration and development of carbon market trading mechanisms.

7.3 Strengthen carbon market construction to activate resource allocation functions

Optimize the carbon market to facilitate efficient resource allocation. Mechanism analysis indicates that the market’s allocative function remains underdeveloped, which may be attributed to insufficient market liquidity and inadequate trading activity in carbon quotas. Therefore, refining carbon market trading mechanisms and supporting systems, expand the market’s coverage to include more industries and participants, and enhance a unified national carbon market. This will provide institutional support to stimulate carbon market trading and promote efficient resource flow.

It is recommended to promote the expansion of the carbon market in phases: the first phase incorporates high-emission industries such as power, steel, and cement; the second phase extends to moderate energy-consuming industries including chemicals and non-ferrous metals; the third phase gradually covers the entire manufacturing sector. Meanwhile,the participation threshold for small and medium-sized enterprises (SMEs) should be lowered to strengthen their capacity to integrate into the market. Additionally, support should be extended to the technological upgrading of carbon trading platforms, the development of real-time carbon emission monitoring systems for enterprises, and the establishment of a “Carbon Market Information Disclosure Platform”, while third-party institutions should be introduced to conduct data verification.

7.4 Implement differentiated carbon emissions trading policy based on enterprise attributes

Heterogeneity analysis indicates that command-driven LCCP generally inhibit the HQDM. In contrast, market-oriented CETP significantly promotes such development in state-owned enterprises (SOEs), technology-intensive, and capital-intensive manufacturers, while notably inhibiting it in non-state-owned and labor-intensive manufacturing. Therefore, differentiated carbon quota standards should be considered for manufacturing enterprises with different attributes.

For instance, capital-intensive enterprises—characterized by low dependence on advanced technology and equipment and limited potential to achieve efficiency improvements through technological innovation—should be allocated more carbon quotas. Additionally, Supervision and regulation of non-state enterprises should be strengthened to enhance their proactive response to policies, and a “Hierarchical Management System for Carbon Compliance of Non-SOEs” should be established: For enterprises with delayed carbon compliance, their participation in government project bidding shall be restricted. Additionally, It shall build a low-carbon technology sharing platform and organize technical exchanges, so as to facilitate the green transition of small and medium-sized enterprises.

Data availability statement

Publicly available datasets were analyzed in this study. This data can be found here: The research data was sourced from the CSMAR database (https://data.csmar.com/) and the “China City Statistical Yearbook”.

Author contributions

CL: Conceptualization, Data curation, Investigation, Methodology, Software, Writing – original draft. KW: Funding acquisition, Project administration, Resources, Supervision, Validation, Writing – review and editing. HL: Formal Analysis, Project administration, Validation, Writing – review and editing.

Funding

The author(s) declare that financial support was received for the research and/or publication of this article. This research was funded the Key Project of the National Social Science Foundation of China: Theoretical and Empirical Research on Measurement, Influencing Factors, and Performance of Comprehensive Reduced Utilization of Natural Resources (CRUNR) (No. 22AGL027), Shanghai Social Science Planning Project “Theoretical, Measurement, and Impact Mechanisms and Policy Research on Multi Low Efficiency Land Use Reduction (MLELR)” (2023ZGL003) and Shanghai Planning and Natural Resources Bureau Project “Research on Implementation Strategies and Models for Reducing Inefficient Construction Land for State-owned Enterprises” (Ghzy2023001), and the Graduate Innovation Fund of Shanghai University of Finance and Economics (CXJJ-2023-359).

Conflict of interest

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Generative AI statement

The author(s) declare that no Generative AI was used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

¹ The heterogeneity of environmental regulations studied in this paper refers to the command-driven low-carbon city pilot policy and the market-oriented carbon emissions trading pilot policy. Low-carbon city pilot policy and carbon emissions trading policy are two different types of environmental regulations.

References

Acemoglu, D., Aghion, P., Bursztyn, L., and Hemous, D. (2012). The environment and directed technical change. Am. Econ. Rev. 102 (1), 131–166. doi:10.1257/aer.102.1.131

PubMed Abstract | CrossRef Full Text | Google Scholar

Agarwal, S., Deng, Y., and Li, T. (2019). Environmental regulation as a double-edged sword for housing markets: evidence from the NO budget trading program. J. Environ. Econ. Manag. 96, 286–309. doi:10.1016/j.jeem.2019.06.006

CrossRef Full Text | Google Scholar

Ai, H. S., Hu, S. L., Li, K., and Shao, S. (2020). Environmental regulation, total factor productivity, and enterprise duration: evidence from China. Bus. Strategy Environ. 29 (6), 2284–2296. doi:10.1002/bse.2502

CrossRef Full Text | Google Scholar

Alpay, E., Kerkvliet, B. J., and Buccola, S. (2020). Productivity growth and environmental regulation in Mexican and U.S. food manufacturing. Am. J. Agric. Econ. 84 (4), 887–901. doi:10.1111/1467-8276.00041

CrossRef Full Text | Google Scholar

Basu, S., Ketcham, J. D., and Kuminoff, N. V. (2025). Environmental regulation, residential sorting, and pollution exposure among senior Americans. J. Environ. Econ. Manag. 133, 103211. doi:10.1016/j.jeem.2025.103211

CrossRef Full Text | Google Scholar

Benatti, N., Groiss, M., Kelly, P., and Lopez-Garcia, P. (2024). Environmental regulation and productivity growth in the euro area: testing the porter hypothesis. J. Environ. Econ. Manag. 126, 102995. doi:10.1016/j.jeem.2024.102995

CrossRef Full Text | Google Scholar

Bodory, H., Huber, M., and Lafférs, L. (2022). Evaluating (weighted) dynamic treatment effects by double machine learning. Econ. J. 25 (3), 628–648. doi:10.1093/ectj/utac018

CrossRef Full Text | Google Scholar

Cai, X., Lu, Y., Wu, M., and Yu, L. (2016). Does environmental regulation drive away inbound foreign direct investment? Evidence from a quasi-natural experiment in China. J. Dev. Econ. 123, 73–85. doi:10.1016/j.jdeveco.2016.08.003

CrossRef Full Text | Google Scholar

Cao, J., Ho, M. S., Ma, R., and Teng, F. (2021). When carbon emission trading meets a regulated industry: evidence from the electricity sector of China. J. Public Econ. 200, 104470. doi:10.1016/j.jpubeco.2021.104470

CrossRef Full Text | Google Scholar

Carril-Caccia, F., and Milgram Baleix, J. (2024). Impact of environmental regulation on M&As in the manufacturing sector. J. Environ. Econ. Manag. 127, 103038. doi:10.1016/j.jeem.2024.103038

CrossRef Full Text | Google Scholar

Chen, H., Guo, W., Feng, X., Wei, W., Liu, H., Feng, Y., et al. (2021). The impact of low-carbon city pilot policy on the total factor productivity of listed enterprises in China. Resour. Conservation Recycl. 169 (2), 105457. doi:10.1016/j.resconrec.2021.105457

CrossRef Full Text | Google Scholar

Chen, Y. P. V., Zhuo, Z., Huang, Z., and Li, W. (2022). Environmental regulation and ESG of SMEs in China: porter hypothesis re-tested. Sci. Total Environ. 850, 157967. doi:10.1016/j.scitotenv.2022.157967

PubMed Abstract | CrossRef Full Text | Google Scholar

Chernozhukov, V., Chetverikov, D., Demirer, M., Duflo, E., Hansen, C., Newey, W., et al. (2018). Double/Debiased machine learning for treatment and structural parameters. Econ. J. 21 (1), C1–C68. doi:10.1111/ectj.12097

CrossRef Full Text | Google Scholar

Clarkson, P. M., Li, Y., Pinnuck, M., and Richardson, G. D. (2015). The valuation relevance of greenhouse gas emissions under the european union carbon emissions trading scheme. Eur. Account. Rev. 24 (3), 551–580. doi:10.1080/09638180.2014.927782

CrossRef Full Text | Google Scholar

Dechezleprêtre, R. C. A. A., and Dechezleprêtre, A. (2016). Environmental policy and directed technological change: evidence from the European carbon market. Rev. Econ. Statistics 98 (1), 173–191. doi:10.1162/rest_a_00470

CrossRef Full Text | Google Scholar

Demir, F., Hu, C., Liu, J., and Shen, H. (2022). Local corruption, total factor productivity and firm heterogeneity: empirical evidence from Chinese manufacturing firms. World Dev. 151, 105770. doi:10.1016/j.worlddev.2021.105770

CrossRef Full Text | Google Scholar

Fan, Y., Wu, J., Xia, Y., and Liu, J. Y. (2016). How will a nationwide carbon market affect regional economies and efficiency of CO2 emission reduction in China? China Econ. Rev. 38, 151–166. doi:10.1016/j.chieco.2015.12.011

CrossRef Full Text | Google Scholar

Farbmacher, H., Huber, M., Lafférs, L., Langen, H., and Spindler, M. (2022). Causal mediation analysis with double machine learning. Econ. J. 25 (2), 277–300. doi:10.1093/ectj/utac003

CrossRef Full Text | Google Scholar

Filippini, M., Geissmann, T., Karplus, V. J., and Zhang, D. (2020). The productivity impacts of energy efficiency programs in developing countries: evidence from iron and steel firms in China. China Econ. Rev. 59, 101364. doi:10.1016/j.chieco.2019.101364

CrossRef Full Text | Google Scholar

Gray, W. B. (1987). The cost of regulation: OSHA, EPA and the productivity slowdown. Am. Econ. Rev. 77 (5), 998–1006. doi:10.1016/0038-0121(88)90025-0

CrossRef Full Text | Google Scholar

Gugler, K., Szücs, F., and Wiedenhofer, T. (2024). Environmental policies and directed technological change. J. Environ. Econ. Manag. 124, 102916. doi:10.1016/j.jeem.2023.102916

CrossRef Full Text | Google Scholar

Guo, J., and Sun, Z. (2023). How does manufacturing agglomeration affect high-quality economic development in China? Econ. Analysis Policy 78, 673–691. doi:10.1016/j.eap.2023.04.007

CrossRef Full Text | Google Scholar

Huang, B., Punzi, M. T., and Wu, Y. (2022). Environmental regulation and financial stability: evidence from Chinese manufacturing firms. J. Bank. and Finance 136, 106396. doi:10.1016/j.jbankfin.2021.106396

CrossRef Full Text | Google Scholar

Lange, I., and Redlinger, M. (2019). Effects of stricter environmental regulations on resource development. J. Environ. Econ. Manag. 96, 60–87. doi:10.1016/j.jeem.2019.04.006

CrossRef Full Text | Google Scholar

Lanoie, P., Laurent-Lucchetti, J., Johnstone, N., and Ambec, S. (2011). Environmental policy, innovation and performance: new insights on the porter hypothesis. J. Econ. and Manag. Strategy 20 (3), 803–842. doi:10.1111/j.1530-9134.2011.00301.x

CrossRef Full Text | Google Scholar

Liu, M. D., Tan, R. P., and Zhang, B. (2021). The costs of “blue sky”: environmental regulation, technology upgrading, and labor demand in China. J. Dev. Econ. 150, 102610. doi:10.1016/j.jdeveco.2020.102610

CrossRef Full Text | Google Scholar

Martin, R., de Preux, L. B., and Wagner, U. J. (2014). The impact of a carbon tax on manufacturing: evidence from microdata. J. Public Econ. 117, 1–14. doi:10.1016/j.jpubeco.2014.04.016

CrossRef Full Text | Google Scholar

Naegele, H., and Zaklan, A. (2019). Does the EU ETS cause carbon leakage in European manufacturing? J. Environ. Econ. Manag. 93, 125–147. doi:10.1016/j.jeem.2018.11.004

CrossRef Full Text | Google Scholar

Oberfield, E. (2013). Productivity and misallocation during a crisis: evidence from the Chilean crisis of 1982. Rev. Econ. Dyn. 16 (1), 100–119. doi:10.1016/j.red.2012.10.005

CrossRef Full Text | Google Scholar

Porter, M., and Linde, C. (1995). Toward a new conception of the environment-competitiveness relationship. J. Econ. Perspect. 9 (4), 97–118. doi:10.1257/jep.9.4.97

CrossRef Full Text | Google Scholar

Qiu, S., Wang, Z., and Liu, S. (2021). The policy outcomes of low-carbon city construction on urban green development: evidence from a quasi-natural experiment conducted in China. Sustain. Cities Soc. 66, 102699. doi:10.1016/j.scs.2020.102699

CrossRef Full Text | Google Scholar

Robinson, P. M. (1988). Root-n-consistent semiparametric regression. Econometrica 56 (4), 931–954. doi:10.2307/1912705

CrossRef Full Text | Google Scholar

Shao, S., Xu, L., Yang, L., and Yu, D. (2024). How do energy-saving policies improve environmental quality: evidence from China’s top 10,000 energy-consuming enterprises program. World Dev. 175, 106466. doi:10.1016/j.worlddev.2023.106466

CrossRef Full Text | Google Scholar

Teng, Z., Xu, S., and Zhang, S. (2022). Are export tax rebates patronage for Chinese firms? An analysis of productivity. China Econ. Rev. 75, 101844. doi:10.1016/j.chieco.2022.101844

CrossRef Full Text | Google Scholar

Wang, C., Wu, J., and Zhang, B. (2018). Environmental regulation, emissions and productivity: evidence from Chinese COD-Emitting manufacturers. J. Environ. Econ. Manag. 92, 54–73. doi:10.1016/j.jeem.2018.08.004

CrossRef Full Text | Google Scholar

Wu, J., Yu, Z., and Wang, R. (2025). The impact of public participatory environmental regulation on carbon emission intensity: a policy text analysis. Front. Environ. Sci. 13, 1534066. doi:10.3389/fenvs.2025.1534066

CrossRef Full Text | Google Scholar

Xu, L., Yang, L., Li, D., and Shao, S. (2023). Asymmetric effects of heterogeneous environmental standards on green technology innovation: evidence from China. Energy Econ. 117, 106479. doi:10.1016/j.eneco.2022.106479

CrossRef Full Text | Google Scholar

Yang, J., Green, C. J. L., Gao, L., and Yu, S. (2024). Green transformation and manufacturing value chain climbing: the moderating role of environmental regulations. Appl. Econ. 57, 2726–2741. doi:10.1080/00036846.2024.2331037

CrossRef Full Text | Google Scholar

Yang, G., Huang, X., Huang, J., and Chen, H. (2020). Assessment of the effects of infrastructure investment under the belt and road initiative. China Econ. Rev. 60, 101418. doi:10.1016/j.chieco.2020.101418

CrossRef Full Text | Google Scholar

Yang, J., Chuang, H., and Kuan, C. (2020). Double machine learning with gradient boosting and its application to the big N audit quality effect. J. Econ. 216 (1), 268–283. doi:10.1016/j.jeconom.2020.01.018

CrossRef Full Text | Google Scholar

Zeng, J., Zhou, Q., and Yang, D. (2023). Estimating the economic impact of intensifying environmental regulation in China. Environ. Resour. Econ. 86 (1-2), 147–172. doi:10.1007/s10640-023-00791-5

CrossRef Full Text | Google Scholar

Zhang, M., and Zhao, Y. (2023). Does environmental regulation spur innovation? quasi-Natural experiment in China. World Dev. 168, 106261. doi:10.1016/j.worlddev.2023.106261

CrossRef Full Text | Google Scholar

Zhang, C., Liu, H., Bressers, H. T. A., and Buchanan, K. S. (2011). Productivity growth and environmental regulations - accounting for undesirable outputs: analysis of China's thirty provincial regions using the malmquist–luenberger index. Ecol. Econ. 70 (12), 2369–2379. doi:10.1016/j.ecolecon.2011.07.019

CrossRef Full Text | Google Scholar

Zhang, Y., Li, H., and Ren, G. (2022). Quantifying the social impacts of the London night tube with a double/debiased machine learning based difference-in-differences approach. Transp. Res. Part A Policy Pract. 163, 288–303. doi:10.1016/j.tra.2022.07.015

CrossRef Full Text | Google Scholar

Zheng, H., Wu, S., and Zhang, Y. (2023). Environmental regulation effect on green total factor productivity in the yangtze river economic belt. J. Environ. Manag. 325, 116465. doi:10.1016/j.jenvman.2022.116465

PubMed Abstract | CrossRef Full Text | Google Scholar

Zhou, L., and Tang, L. (2021). Environmental regulation and the growth of the total-factor carbon productivity of China's industries: evidence from the implementation of action plan of air pollution prevention and control. J. Environ. Manag. 296, 113078. doi:10.1016/j.jenvman.2021.113078

PubMed Abstract | CrossRef Full Text | Google Scholar