The Effect of the National Specially Monitored Firms Program on Water-Polluting Firms’ Green Total Factor Productivity

Abstract

Since 2007, the National Specially Monitored Firms (NSMF) program has been a key instrument in the Chinese government’s environmental regulation efforts and a crucial approach for achieving sustainable development in China. There is limited literature examining its effect on green total factor productivity (GTFP). This study investigates the effect of this policy on water-polluting firms’ GTFP using pollution discharge data from Chinese industrial firms and employing a time-varying difference-in-differences model. The findings illustrate that (1) the NSMF program significantly enhances the GTFP of water-polluting firms, primarily by promoting technological progress; (2) the NSMF program advances water-polluting firms’ GTFP through three channels: alleviating financing constraints, improving human capital, enhancing pollution treatment technologies. This study provides empirical evidence on the effectiveness of the NSMF program, offering valuable insights for the formulation of command-and-control environmental regulations and the pursuit of sustainable social development in the future.

1. Introduction

Existing studies highlight that a healthy ecological environment is essential for sustainable socio-economic development as well as the enhancement in residents’ living standards [1,2,3]. Maintaining the balance of ecosystems is contingent upon preserving unpolluted water resources [4]. However, water pollution has emerged as a significant ecological challenge, placing immense strain on aquatic ecosystems globally [5]. The issues are particularly acute in developing countries, where water quality and drinking water safety are pressing concerns [6]. As the largest developing nation, China faces severe water pollution challenges, largely driven by its industrial sector’s high levels of pollution and energy consumption [7]. This has resulted in issues such as eutrophication, contamination by heavy metals, and the presence of organic compounds [8,9], all of which have serious implications for food safety and the health of Chinese residents [9].

Since 1979, the Chinese government has implemented a series of water pollution control policies aimed at improving water quality in both urban and rural areas [10,11]. These policies include the “Two Control Zones” policy [12,13], the River Chief System [14], the National Specially Monitored Firms (NSMF) program [15], and the Water Ecological Civilization City Pilot initiative [6], among others. These efforts have led to significant advancements in controlling water pollution. Moreover, research indicates that China’s regulatory policies on water pollution not only facilitate reductions in corporate emissions, but also influence green total factor productivity (GTFP) and technological progress within firms. For instance, Huang et al. [16] demonstrated that environmental taxes on water pollutant emissions effectively stimulate green innovation in firms. Yang et al. [6] found that the Water Ecological Civilization City Pilot Policy fosters green technological innovation among regional firms.

Among the various environmental regulations introduced by the Chinese government, the NSMF program stands out as a key command-and-control measure. Unlike phased approaches to pollution control, the NSMF program has been continuously implemented since 2007. Each year, the Ministry of Ecology and Environment (MOE) compiles a list of industrial firms ranked by their pollutant emissions. The NSMF program includes firms whose emissions collectively represent 65% of the total industrial emissions, and these firms are subject to real-time monitoring of their emissions [15,17]. The NSMF program requires monitored firms to reduce their pollutant emissions, compelling them to invest more resources in cleaner production and pollution control. This, in turn, leads to an optimized resource allocation and upgraded input structures, potentially influencing the total factor productivity of these firms.

Investigating the effect of the NSMF program on the GTFP of Chinese industrial firms is of paramount importance. China faces severe pollution challenges [18] and is committed to achieving sustainable development. Therefore, evaluating whether the NSMF program, as a key environmental regulation measure, has effectively promoted green development within firms is essential. Additionally, clarifying the fundamental mechanisms through which the NSMF program influences firms’ GTFP is valuable for elucidating how command-and-control environmental measures can advance corporate sustainability. Insights from this study could also be beneficial for other countries dealing with significant environmental and economic challenges.

Current research on the NSMF program primarily addresses effects on monitored firms’ pollutant emissions [15,19], social responsibility [20], innovation capacity [21], and investor confidence [22]. However, there is a notable gap in the literature regarding the effect of the NSMF program on the GTFP of monitored firms. Huang et al. [23] employed the Stochastic Frontier Analysis method to evaluate the green technology efficiency of National Specially Monitoring (NSM) thermal power plants, but did not investigate the policy’s effect on green technology efficiency more broadly. Thus, the effect of the NSMF program on firms’ GTFP, which reflects the program’s ability to foster sustainable development, warrants further exploration.

This study addresses this gap by calculating the Malmquist–Luenberger (ML) index and its decomposition indices using firm-level data from 2004 to 2013 to assess GTFP and technological progress [24,25,26]. Subsequently, a time-varying difference-in-differences (DID) model is utilized to assess the effect of the NSMF program on the GTFP of water-polluting firms. This study focuses on the following two key questions: (1) Does the NSMF program enhance the GTFP of water-polluting firms? (2) If so, through which mechanisms does the NSMF program facilitate these effect improvements?

This study makes two significant contributions. First, it is the first to investigate the effect of the NSMF program on water-polluting firms’ GTFP, thereby expanding the literature on the effects of command-and-control environmental regulations on firms’ sustainability in developing economies. The empirical insights provided could inform the evolution of such regulations in developing countries and offer practical guidance for achieving sustainable development.

Second, this study investigates the mechanisms through which the NSMF program affects water-polluting firms’ GTFP. It finds that the policy promotes GTFP growth primarily by alleviating firms’ financing constraints, enhancing human capital, and improving pollution treatment technologies. Understanding these mechanisms offers deeper insights into firms’ responses and enables a more accurate assessment of the policy’s effects.

The structure of this study is organized as follows: Section 2 provides the policy background; Section 3 reviews the relevant literature and formulates the research hypotheses; Section 4 describes the dataset, model, and variables; and Section 5 reports the empirical findings. Section 6 explores the underlying mechanisms, while Section 7 offers concluding remarks.

2. Policy Background

Environmental regulation in China encompasses command-and-control, market-incentive, and voluntary approaches [27]. The NSMF program stands out as a pivotal command-and-control environmental regulation [17]. Initiated by the Chinese Environmental Protection Administration (EPA) (renamed as Ministry of Ecology and Environment after 2008) in 2007 under the “Opinions on Strengthening and Improving Environmental Statistics”, this program aims to collect timely information on pollutant emissions from highly polluting firms, facilitating strengthened law enforcement and supervision, thereby promoting corporate pollution management [19].

The list of the first batch of NSMF is divided into three categories: water-polluting firms, air-polluting firms, and sewage treatment plants. The initial NSMF list, released in 2007, identified water-polluting firms based on their industrial emissions of chemical oxygen demand and ammonia nitrogen in 2005, prioritizing those whose emissions collectively represent 65% of the total industrial emissions. By 2016, the monitored water-polluting firms had decreased from 3115 in 2007 to 2660, as firms successfully met emission reduction targets and were subsequently removed from the list [15]. The NSMF program expanded over time to include firms emitting heavy metals (2012), large-scale livestock and poultry operations (2013), and hazardous waste facilities (2015), reflecting China’s evolving ecological priorities. Subsequently, provincial environmental departments took over list publication from the MOE after 2016.

Under the NSMF, firms are mandated to install real-time emission monitoring systems linked to a national network, ensuring the continuous collection of accurate pollutant discharge data [21]. Monthly supervision and on-site inspections are conducted to verify data integrity and system functionality, with findings reported to local governments to determine pollution taxation standards for monitored firms [15]. These stringent regulatory measures underscore the heightened environmental oversight imposed on NSMF-listed entities. Exploring the effect of such regulatory intensity on the GTFP of monitored firms warrants detailed investigation.

3. Literature Review and Research Hypotheses

Some scholars, grounded in the neoclassical “cost theory”, argue that the implementation of environmental regulations escalates production costs and diminishes firms’ capacity for technological innovation [28]. These regulations impose additional environmental management costs, such as the procurement or upgrading of pollution control equipment and the payment of sewage fees, thereby diverting resources from productive activities to environmental management or emission reduction efforts, leading to potential productivity losses [29]. However, Porter and Linde [30] challenged this neoclassical perspective through their well-known “Porter Hypothesis”. They contend that moderate environmental regulation can stimulate innovation and enhance productivity, thereby partially or fully offsetting the increased costs and reduced profits caused by such regulations. This, in turn, can lead to an overall increase in firms’ total factor productivity (TFP). As environmental regulation intensity and pollution control costs rise, firms are driven to increase R&D investment and pursue green technological innovations in areas such as resource input, energy consumption, and emission reduction, ultimately lowering pollution control costs [31]. Compared to paying directly for pollution emissions, technological innovation offers higher expected returns to firms [32], as it not only reduces pollutant emissions but also improves productivity and product quality, thereby enhancing firms’ competitiveness and increasing output. Consequently, the competitive advantage gained through productivity improvements via R&D and innovation can offset profit reductions caused by higher environmental management costs, ultimately boosting corporate TFP.

Other scholars suggest that the positive effect of environmental regulation on firms’ TFP improvement arises not only from technological advancement but also from optimizing production resource allocation efficiency [33]. On the one hand, the enactment of environmental regulation policies, such as environmental and energy taxes, promotes the marketization of non-renewable energy prices like for coal and oil, reducing potential price distortions and curbing excessive energy and resource consumption due to low costs. This, in turn, facilitates TFP enhancement through the upgrading of factor structures. On the other hand, increased environmental regulation intensity compels firms to either install end-of-pipe pollution treatment equipment or pursue green technological innovations in production to reduce emissions, thereby increasing the demand for high-skilled talent in research and development [31]. By enhancing human capital to replace high energy consumption and capital investment and continuously improving factor resource allocation efficiency, firms can ultimately promote TFP growth. The positive effect of environmental regulation on firms’ TFP enhancement has been corroborated by several studies [34,35,36].

When firms face environmental regulation, the cost of pollution control is included in the inputs to the TFP measure, while the reduction in undesirable output is not accounted for. This approach may underestimate a firm’s actual TFP and overestimate the negative effect of environmental regulation on firm productivity. GTFP, which incorporates pollution emissions into the TFP measure, provides a more accurate assessment. Industrial production and pollution emissions are often intertwined, yet the existing literature assessing the effect of environmental regulations on firm productivity typically considers only inputs and economic outputs, neglecting pollution emissions as an undesirable output. This uneven treatment of desired versus undesired outputs may skew the evaluation of production performance and social welfare, potentially resulting in flawed policy recommendations [37,38]. Therefore, to accurately assess the productivity effects of environmental regulation, the GTFP indicator should be used instead of the traditional TFP measure.

Numerous studies have explored the effect of environmental regulations on firms’ GTFP. Zhang et al. [39] and Zhang et al. [40] found that environmental regulation can significantly increase firms’ GTFP. Li and Chen [41] and Liu and Cui [42] concluded that while environmental regulation may negatively impact firms’ GTFP in the short term, in the long term, such regulation can foster a “win-win” situation for both firms’ competitiveness and environmental protection [41]. Cheng and Kong [43] further indicated that both command-and-control and market-incentive environmental regulations can enhance firms’ GTFP, and that a combination of policies is more conducive to GTFP growth than a single policy.

Based on this analysis, we propose Hypothesis 1:

H1. 

The NSMFprogram significantly enhancesthe GTFP of monitored firms.

It is crucial to understand the mechanisms through which the NSMF program affects the GTFP of monitored firms. Prior research indicates that firms can improve their GTFP through external resources, such as government subsidies and bank credits [21], as well as internal resources, including human capital and pollution treatment technologies [44,45,46]. Therefore, the NSMF program may enhance monitored firms’ GTFP by alleviating financing constraints, improving human capital, and inducing an increased use of pollution treatment technologies.

First, the implementation of the NSMF program may facilitate firms’ access to government subsidies and alleviate financing constraints, thereby contributing to GTFP growth. Following the release of the NSMF list, China’s EPA mandated that local environmental departments prioritize these firms for pollution control measures and allocate dedicated funding programs, thereby enhancing firms’ access to government subsidies and bank credits. The inflow of external funds alleviates financing constraints on firms’ green technological innovation [47], promotes technological progress, and increases the likelihood of GTFP growth.

Second, the NSMF program may drive GTFP growth by improving firms’ human capital. As noted, NSMF firms receive priority for advanced technology demonstration projects, facilitating talent acquisition and technology exchanges, thereby boosting their human capital. Human capital is a critical determinant of firms’ technological innovation capabilities [44]. By enhancing human capital to replace high energy consumption and capital investment and continuously improving factor resource allocation efficiency [31], the NSMF program could serve as a powerful driver for enhancing their GTFP.

Third, the NSMF program, as a classic command-and-control environmental regulatory policy [17], is designed to strengthen pollution treatment among monitored firms. Effective pollution reduction requires firms to upgrade their pollution treatment technologies, including both end-of-pipe and source control technologies [46,48,49,50]. End-of-pipe treatment involves converting pollutants into more manageable substances before discharge [45]. While end-of-pipe technologies typically require fewer inputs and have a lower technological threshold [51], they are often preferred by firms with limited capital or R&D capabilities [52]. In contrast, source control technologies aim to reduce pollutant generation at the production stage by minimizing reliance on polluting inputs like energy or raw materials and improving resource efficiency for cleaner production [53]. Source control methods depart from the traditional “pollute first, treat later” approach by preventing pollution through cleaner production practices [54], leading to more effective emission reductions [46]. The adoption of both end-of-pipe and source control technologies promotes cleaner production, potentially enhancing firms’ GTFP.

We propose Hypotheses 2a, 2b, and 2c based on the preceding analysis.

H2a. 

The NSMF program enhances monitored firms’ GTFP by alleviating their financing constraints.

H2b. 

The NSMF program enhances monitored firms’ GTFP by improving their human capital.

H2c. 

The NSMF program enhances monitored firms’ GTFP by inducing an increased use of pollution treatment technologies.

The research framework diagram of this study is illustrated in Figure 1.

4. Model, Data, and Variables

4.1. Model

In this paper, we examined the effect of the NSMF program, announced by China’s EPA (renamed as China’s MOE after 2008), on the GTFP of affected firms from 2007 to 2013. To accurately capture the program’s effect, we employed a time-varying DID model. This approach is particularly suited to our analysis because the NSMF program was implemented incrementally, with new firms continuously added to the monitoring list over the period from 2007 to 2013. As a result, the timing of the policy shock varied across treated firms. The time-varying DID can accurately capture the timing of the policy effect on treated firms, whereas the traditional DID model assumes a uniform policy implementation time, which does not apply to this study. Additionally, the time-varying DID model, which includes multiple treatments, plausibly alleviates concerns that contemporaneous trends influenced by factors unrelated to the treatment of interest may confound the estimated treatment effect in traditional DID models with a single treatment period [55]. The model is specified as follows:

$$Y_{it}={\beta }_0+{\beta }_1{Policy}_{it}+\gamma X_{it}+{\theta }_t+{\partial }_i+{\epsilon }_{it}$$

(1)

In Equation (1), the dependent variable $Y_{it}$ represents the GTFP of firm I in year t. ${Policy}_{it}$ is a binary indicator that denotes whether firm I is included in the NSMF program in year t. $X_{it}$ denotes a vector of control variables. ${\theta }_t$ are year fixed effects and ${\partial }_i$ are firm fixed effects. ${\epsilon }_{it}$ is a randomized error term. The coefficient ${\beta }_1$ captures the average treatment effect of the policy.

4.2. Data

Following Brandt et al. [56], we constructed a panel dataset by merging the 2004–2013 Annual Survey of Industrial Firms (ASIF) and Environmental Survey and Reporting (ESR) datasets in China. This dataset integrates enterprise financial data with pollution discharge information, focusing on nine heavily polluting industries characterized by significant wastewater discharge and robust sample sizes. The nine industries include agricultural and sideline food processing (13); food manufacturing (14); wine, beverage, and refined tea manufacturing (15); textile (17); paper and paper products (22); pharmaceutical manufacturing (27); non-metallic mineral products (31); ferrous metal smelting and calendaring (32); and non-ferrous metal smelting and calendaring (33), with the numbers in parentheses representing their respective two-digit industry codes. It is important to note that data for the non-metallic mineral products industry (31) are missing for 2012, and for the non-ferrous metals smelting and calendaring industry (33) for 2007 and 2008. The ASIF dataset encompasses all state-owned industrial enterprises as well as private industrial enterprises with annual sales exceeding CNY 5 million, making it the most comprehensive firm-level database in China [57]. The ESR dataset provides detailed firm-level emission data for approximately 85% of China’s major pollutants [58]. Both datasets are collected, audited, published, and maintained by the National Bureau of Statistics of China (NBS) in collaboration with other ministries, including the MOE. The high quality of these data has been widely recognized and utilized in academic research [15,57,58,59]. Additionally, we obtained economic and pollution emission data at the provincial level from the China Research Data Service Platform (CNRDS) and integrated this information into our final dataset.

In our data preparation, we excluded firms with missing or negative emissions for chemical oxygen demand (COD) and ammonia nitrogen (AN), as well as those with missing or negative values for gross output value, average annual balance of net fixed assets, and intermediate inputs. Firms with fewer than eight employees were also excluded. To ensure sufficient observations for the policy year and years preceding it within the treatment group, we excluded firms lacking complete data for these periods and those with fewer than two observations before implementation. Additionally, control group firms with only a single year of data were excluded.

4.3. Variables

4.3.1. Measurement of Changes in GTFP

First, this paper employs the Malmquist–Luenberger (ML) index in a fixed parametric form based on the directional distance function [60] to measure changes in GTFP. The ML index, a non-parametric approach using Data Envelopment Analysis, is well suited for evaluating firm productivity as it simultaneously accounts for both desirable and undesirable outputs without the need for shadow price information. This method provides a more precise measure of GTFP changes over time. Additionally, we utilized the production efficiency (EC) index and technical progress (TC) index, both of which are derived from the decomposition of the ML index. The detailed steps of this methodology are outlined in Appendix A. Using 2004 as the base year, we set the GTFP index for firms in that year to 1, thereby normalizing each year’s GTFP relative to the base year’s value.

Polluting firms operate as multi-input, multi-output production units, producing both desired and undesired outputs. To calculate the ML index, we used three inputs: capital (K), labor (L), and intermediate inputs (M). Capital (K) is defined as the average annual balance of firms’ net fixed assets, measured in thousands of CNY, adjusted using the geometric mean of the price index for fixed asset investment at both regional and national levels. Labor (L) is quantified by the number of employees at each firm. Intermediate inputs (M) are defined as price-deflated intermediate inputs, using the 1998 base period, and measured in thousands of CNY. The desired output is gross industrial output (Y), which is expressed as price-deflated gross industrial output, also based on the 1998 base period, in thousands of CNY. The price deflator for both intermediate inputs and industrial output is the ex-factory price index for industrial goods.

This study specifically targets water-polluting firms. During the study period, emission data for ammonia nitrogen (AN) and chemical oxygen demand (COD) were the most comprehensive, making these two pollutants the key undesired outputs for water-polluting enterprises. Both pollutants are measured in kilograms, following the methodology established by Fujii and Managi [61] and Li et al. [62]. Descriptive statistics of input and output variables are provided in

Table 1. Descriptive statistics of input and output variables.

Variables	Observations	Mean	Standard Deviation	Minimum	Maximum
K	13,063	164,753.22	1,350,379.73	97.47	65,983,720.00
L	13,063	727.93	2909.51	9	135,438
M	13,063	268,756.53	1,409,551.71	211.96	37,354,608.00
Y	13,063	336,466.26	1,808,792.18	358.06	49,469,208.00
AN	13,063	11,266.97	78,935.98	0.10	3,700,167.00
COD	13,063	216,746.14	1,253,300.95	25.60	44,072,750.90

.

4.3.2. Control Variables

We included both firm-level and regional-level control variables. At the firm level, the control variables encompass the previous year’s pollutant emissions, specifically ammonia nitrogen (AN_1) and chemical oxygen demand (COD_1). These emissions are critical reference indicators for determining a firm’s inclusion in the NSMF program and have a direct effect on the GTFP, production efficiency, and technological progress of the firms.

The level of regional economic development significantly influences firm input factors, such as capital and labor, as well as outputs like gross production value, which in turn influence GTFP, production efficiency, and technological progress. Moreover, differences in pollution emissions among industrial firms across regions with varying levels of economic development and pollution may influence their inclusion in the NSMF program. Although the NSMF program is not a region-specific policy, it cannot be entirely considered exogenous at the regional level. Therefore, it is crucial to include provincial-level control variables to reduce omitted variable bias. This paper considers three provincial-level control variables: GDP per capita (Per_GDP), regional COD emission intensity (Reg_CI), and regional industrial structure, which reflected in the proportion of secondary industry value added to the gross regional product (Indu_sec). In addition, industry-year fixed effects are incorporated to account for unobserved time-varying factors specific to each industry. Descriptive statistics for the baseline regression variables are presented in

Table 2. Descriptive statistics of variables in baseline regressions.

Variables	Observations	Mean	Standard Deviation	Minimum	Maximum
GTFP	13,063	1.0499	0.1611	0.5344	1.9258
EC	13,063	1.1084	0.2188	0.5004	1.9978
TC	13,063	0.9629	0.1266	0.5580	1.6254
Policy	13,063	0.0899	0.2861	0	1
AN_1	13,063	12,061.6417	83,211.5185	0.1000	3,700,167.0000
COD_1	13,063	234,687.9747	1,277,137.1092	6.0000	38,620,072.0000
Per_GDP	13,063	11.7690	8.8982	1.1574	57.6950
Reg_CI	13,063	0.0012	0.0018	0.0000	0.0152
Indu_sec	13,063	50.1318	5.5880	22.3000	61.5000

.

4.3.3. Mechanistic Variables

To evaluate financing constraints, we employed the SA index (SA), developed by Hadlock and Pierce [63], calculated using the following formula: ${SA}_{i,t}=–0.737\times {SIZE}_{i,t}+0.043\times {SIZE}_{i,t}^2–0.04\times {AGE}_{i,t}$. In this equation, ${SIZE}_{i,t}$ denotes firm size, measured as the logarithm of total assets, while ${AGE}_{i,t}$ indicates firm age, calculated as the difference between the observation year and the incorporation year. A more negative SA index reflects higher financing constraints. Additionally, government subsidies can mitigate firms’ financing constraints to some extent. We measured government subsidies (Subsidy) as the ratio of government subsidies to operating revenues. To proxy human capital, we utilized the logarithm of main business wages (lnWage) and the logarithm of per capita main business wages (lnWage_av). Regarding pollution treatment technologies, we differentiated between end-of-pipe and source control technologies. Firms’ engagement in end-of-pipe treatment was assessed using the removal rates of COD and AN, where the removal rate is calculated as the amount of pollutant removed divided by the amount produced, denoted by COD_rate and AN_rate, respectively. For evaluating technological progress in source control treatment, we employed the logarithmic forms of COD and AN generation per unit of output, denoted as lnCOD_prin and lnAN_prin, respectively. Descriptive statistics for the mechanistic variables are provided in

Table 3. Descriptive statistics of mechanistic variables.

Variables	Observations	Mean	Standard Deviation	Minimum	Maximum
SA	12,565	−3.3873	0.7084	−18.8173	−0.0376
Subsidy	11,032	0.0031	0.0143	0	0.3272
lnWage	7973	8.0637	1.6965	0	14.9824
lnWage_av	7973	2.6148	0.6366	0	5.1985
COD_rate	10,873	0.6532	0.3383	0	0.9996
AN_rate	9560	0.3476	0.3682	0	0.9998
lnCOD_prin	9060	3.0590	1.8525	0.0065	12.4413
lnAN_prin	7756	0.7731	0.9188	0	7.2895

.

5. Empirical Results

5.1. Baseline Regression Results

Table 4. Baseline regression results.

	(1)	(2)	(3)	(4)	(5)	(6)
Dependent Variables:	GTFP	GTFP	EC	EC	TC	TC
Policy	0.0188 ***	0.0163 ***	0.0040	−0.0039	0.0165 ***	0.0183 ***
	(0.0060)	(0.0063)	(0.0076)	(0.0080)	(0.0038)	(0.0038)
Control variables	NO	Yes	NO	Yes	NO	Yes
Firm FE	Yes	Yes	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Industry × Year FE	Yes	Yes	Yes	Yes	Yes	Yes
Observations	13,016	13,016	13,016	13,016	13,016	13,016
R2	0.662	0.662	0.740	0.741	0.822	0.822

Note: Firm FE, Year FE, and Industry × Year FE denote firm fixed effects, year fixed effects, and industry-year fixed effects, respectively. *** denotes 1% level of significance. Heteroskedasticity-robust standard errors are provided in parentheses. Yes and NO indicate whether such variables are controlled.

presents the estimation results derived from the time-varying DID model. Specifically, columns (1) through (6) detail the impact of the NSMF program on the GTFP of water-polluting firms (columns (1)–(2)), production efficiency (columns (3)–(4)), and technological progress (columns (5)–(6)), respectively. Columns (1), (3), and (5) display results without control variables, while columns (2), (4), and (6) present results with the inclusion of control variables. To reduce the influence of outliers on the regression results, variables are trimmed at the 1% and 99% levels.

The results indicate that the coefficients measuring the effect of policy on GTFP and TC are significantly positive, whereas the effect on EC is not statistically significant, regardless of the inclusion of control variables. When control variables are included, the GTFP and TC of the monitored firms increased by 0.0163 and 0.0183, respectively, compared to firms not subject to the NSMF program. These findings imply that the NSMF program primarily enhances the GTFP of water-polluting firms through technological progress. Thus, Hypothesis 1 is confirmed. The baseline regression results align with the findings of Zhang et al. [39] and Zhang et al. [40], affirming that environmental regulation can significantly enhance firms’ GTFP.

5.2. Parallel Trend Tests

A crucial assumption for the validity of the DID model is that the pre-intervention trends in the outcome variables are parallel between the treated and control groups. Following the approach outlined by Ai et al. [64], we performed a parallel trend test and conducted a dynamic effects analysis.

$$Y_{it}={\beta }_0+{\sum }_{t=-3}^4{\beta }_t{Policy}_{it}+\gamma X+{\theta }_t+{\partial }_i+{\epsilon }_{it}$$

(2)

In this equation, ${Policy}_{it}$ indicates whether firm I is included in the monitoring list during period t. Here, t > 0 denotes the period after policy implementation, t < 0 denotes the period preceding policy implementation, and t = 0 denotes the policy implementation year. For t < 0, if the null hypothesis ${\beta }_t$ = 0 is not rejected, it suggests that there is no significant difference in the pre-policy trends between the control and treatment groups. We set the year immediately preceding the policy implementation (t = −1) as the baseline year for the parallel trend test, as illustrated in Figure 2. Figure 2A depicts the effect of the policy on firms’ GTFP, while Figure 2B illustrates its effect on firms’ TC.

The results reveal that none of the coefficients for the variable Policy are significant statistically before the implementation of the NSMF program. This finding indicates that, prior to the policy’s enactment, there are no significant differences in the trends of GTFP and TC between policy-affected and non-policy-affected firms, thereby supporting the parallel trend assumption. After the policy’s implementation, the TC of the policy-affected firms exhibited a continuous increase over the subsequent three years, while the GTFP began to improve one year after the policy was enacted. This pattern suggests that the NSMF program has a lagged effect on firms’ GTFP, whereas its effect on firms’ technological progress is more immediate. One plausible explanation is that, following the imposition of environmental regulations, firms can swiftly achieve technological progress, for instance, by adopting advanced production technologies. However, translating these technological advancements into improved production efficiency and GTFP typically requires more time [35], justifying the observed lag between the policy implementation and the enhancement in GTFP.

5.3. Placebo Tests

To address concerns about potential omitted variable bias affecting the observed effect of the NSMF program on the GTFP and technological progress of water-polluting firms, we conducted a placebo test following the methodology outlined by Chetty et al. [65]. This test involves creating a dummy policy shock to validate the robustness of our findings. We began by randomly selecting a subset of firms from the full dataset to serve as the new treatment group. For this group, we subsequently assigned a random time period to serve as the dummy policy shock period. A dummy variable corresponding to this dummy policy shock was created, and we performed regressions using Equation (1) on these randomized datasets. To ensure the robustness of the results, this procedure was repeated 500 times. The estimation results for these placebo tests are presented in Figure 3. Panel (A) illustrates the effect of the dummy policy shock on firms’ GTFP, while Panel (B) shows its effect on firms’ TC.

In Figure 3, the mean values of the estimated coefficients from the placebo tests are close to zero, with the corresponding p-values generally exceeding 0.10. These results indicate that the observed effects of the NSMF program on the enhancement in GTFP and technological progress in water-polluting firms are unlikely to be due to random chance, thereby supporting the robustness of our findings.

5.4. Robustness Tests for Heterogeneous Treatment Effects

Estimating staggered adoption designs by the traditional two-way fixed effects model may result in biased outcomes [66,67,68]. This bias occurs because units that have already received treatment can inadvertently serve as control groups for units treated later, due to the varying timing of treatment [68]. Such variations can distort the treatment effect weights and, in some cases, even produce “negative weights” [66,67]. The “negative weight” issue is a key reason why traditional two-way fixed effects models may not accurately estimate the effects in staggered adoption scenarios. In a time-varying DID model, the estimation represents a weighted average of multiple traditional DID estimates. The presence of “negative weights” skews these individual weights from their true values, resulting in biased regression estimates.

To address potential heterogeneous treatment effects and assess whether the “negative weight” problem introduces significant estimation bias, we began by calculating the percentage of negative weights in our baseline regression. To obtain robust estimates of the average treatment effect, we subsequently applied alternative methods, including the DIDM model [67], the DID imputation model [66], and the two-stage DID model [69]. The percentage of negative weights and the results from these alternative methods are presented in

Table 5. The percentage of negative weights and the robust estimates’ results.

	(1)	(2)	(3)
Dependent Variables:	GTFP	EC	TC
Percentage of negative weights	0.34%	0.34%	0.34%
DIDM	0.0204 **	−0.0108	0.0395 ***
	(0.0097)	(0.0145)	(0.0081)
DID imputation	0.0225 ***	0.0071	0.0184 ***
	(0.0086)	(0.0114)	(0.0053)
Two-stage DID	0.0216 ***	0.0056	0.0189 ***
	(0.0087)	(0.0115)	(0.0053)

Note: Significance levels are denoted by ** and ***, corresponding to 5% and 1%, respectively, with heteroskedasticity-robust standard errors provided in parentheses.

.

The results show that in the baseline regression samples, the proportion of negative weights is low, indicating that significant heterogeneous treatment effects are unlikely. The average treatment effect coefficients for GTFP and TC are significantly negative and closely align with the baseline regression results, whereas the coefficients for EC remain insignificant. This indicates that the bias arising from heterogeneous treatment effects is minimal in the baseline regressions.

5.5. Mitigating the Influence of Penecontemporaneous Policies

The Chinese government implemented several additional environmental regulatory policies alongside the NSMF program. These policies could have significantly influenced firms’ GTFP and technological progress. To mitigate potential estimation bias arising from these contemporaneous policies, we excluded firms affected by three major policies, the emission fee rates, the Thousand Firms Energy Conservation Action, and the Cleaner Production Standard, as detailed in Appendix B,

Table A1. Introduction to the quasi-simultaneous policy.

Formulation	Policy Introduction	Regulated Business or Industry
Emission Fee Rates	In 2007, the Chinese government issued the “Notice of Printing and Distributing Comprehensive Work Plans for Energy Conservation and Emission Reduction”. Mandated by the principle of compensating cost governance, it required all regions in China to increase emission fee rates. Specifically, the notice established sulfur dioxide (SO2) emission fee rates at CNY 1.26 per kilogram over a three-year period. However, it did not specify an adjustment standard for chemical oxygen demand (COD) emission fee rates. However, in the short term, only Jiangsu Province implemented the increase in SO2 and COD emission fee rates.	Industrial firms in Jiangsu Province
Thousand Firms Energy Conservation Action	In 2006, the Chinese government initiated supervision and management of 1008 independent accounting firms across nine key energy-consuming industries, focusing on firms of a specified minimum size. As of 2004, the comprehensive energy consumption of these firms reached or exceeded 180,000 tons of standard coal, which collectively represented 33% of the country’s total energy consumption and 47% of industrial energy consumption.	Firm listing
Cleaner Production Standard	In 2006, the government promulgated cleaner production standards for eight industries, regulating the utilization of production technologies and equipment, pollutant emissions, and waste recycling practices.	Manufacturing of beer, edible vegetable oil, textiles, sugar cane, electrolytic aluminum, nitrogen fertilizer, steel, and basic chemical raw materials.

. The results of this adjustment are presented in

Table 6. Regression results excluding contemporaneous policy disturbances.

	Emission Fee Rates		Thousand Firms Energy Conservation Action
	(1)	(2)	(3)	(4)
	GTFP	TC	GTFP	TC
Policy	0.0176 ***	0.0179 ***	0.0167 ***	0.0184 ***
	(0.0065)	(0.0039)	(0.0064)	(0.0038)
Observations	12,021	12,021	12,823	12,823
R2	0.663	0.820	0.661	0.819
	Cleaner Production Standard		Exclude All Concurrent Policies
	(5)	(6)	(7)	(8)
	GTFP	TC	GTFP	TC
Policy	0.0184 **	0.0168 ***	0.0200 **	0.0175 ***
	(0.0085)	(0.0054)	(0.0091)	(0.0055)
Observations	8024	8024	7460	7460
R2	0.663	0.800	0.663	0.800
Control variables	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Industry × Year FE	Yes	Yes	Yes	Yes

Note: Firm FE, Year FE, and Industry × Year FE denote firm fixed effects, year fixed effects, and industry-year fixed effects, respectively. Significance levels are denoted by ** and ***, corresponding to 5% and 1%, respectively, with heteroskedasticity-robust standard errors provided in parentheses. Yes indicates that such variables are controlled.

.

Table 6 demonstrates that all estimated coefficients remain significantly positive and are consistent with the baseline regression results. This indicates that the observed improvements in firms’ GTFP and technological progress can be attributable to the NSMF program rather than the effect of other contemporaneous environmental policies.

6. Mechanism Analysis

6.1. Financing Constraint Alleviation

We assessed the effect of the NSMF program on water-polluting firms’ access to government subsidies and its role in alleviating financing constraints. The results in columns (1) and (2) of

Table 7. Mechanism test results (funding acquisition and level of human capital).

	(1)	(2)	(3)	(4)
Dependent Variables:	Subsidy	SA	lnWage	lnWage_av
Policy	0.0001	0.0264 ***	0.3325 ***	0.1159 ***
	(0.0003)	(0.0075)	(0.1020)	(0.0415)
Control variables	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Industry × Year FE	Yes	Yes	Yes	Yes
Observations	10,381	13,013	7601	7601
R2	0.639	0.986	0.735	0.684

Note: Firm FE, Year FE, and Industry × Year FE denote firm fixed effects, year fixed effects, and industry-year fixed effects, respectively. *** denotes 1% level of significance. Heteroskedasticity-robust standard errors are provided in parentheses. Yes indicates that such variables are controlled.

reveal that the coefficient for government subsidies (Subsidy) is not statistically significant, whereas the coefficient for the SA index (SA) is significantly positive. This suggests that the NSMF program does not directly provide subsidies but instead supports technological advancement by alleviating financing constraints, thereby improving GTFP. Thus, Hypothesis 2a is confirmed. Fang et al. [21] similarly found that the NSMF program enhances the innovation capabilities of monitored firms through the alleviation of financing constraints, though they did not explore the role of government subsidies. This indirect approach is preferable, as direct subsidies to high-polluting firms could be perceived as condoning pollution. Thus, mitigating financing constraints is a more appropriate strategy for supporting these firms. However, the specific mechanisms through which the government alleviates financing constraints—such as influencing banks to increase credit availability—remain unclear and warrant further investigation.

6.2. Human Capital

The regression results assessing the impact of the NSMF program on firms’ human capital are displayed in Table 7, columns (3) and (4). The coefficients in these columns are significantly positive, indicating that the NSMF program effectively enhances human capital within water-polluting firms, thereby contributing to improvements in GTFP, which supports Hypothesis 2b. These findings align with the results of Tang et al. [31] that environmental regulations incentivize firms to advance their human capital.

6.3. Pollution Treatment Technologies

The estimated results of the impact of the NSMF program on firms’ adoption of end-of-pipe and source control technologies are presented in columns (1)–(2) and (3)–(4) of

Table 8. Mechanism test results (pollution treatment technologies).

	(1)	(2)	(3)	(4)
Dependent Variables:	COD_Rate	AN_Rate	lnCOD_Prin	lnAN_Prin
Policy	0.0409 ***	0.0601 ***	−0.1799 ***	−0.1125 *
	(0.0157)	(0.0180)	(0.0677)	(0.0580)
Control variables	Yes	Yes	Yes	Yes
Firm FE	Yes	Yes	Yes	Yes
Year FE	Yes	Yes	Yes	Yes
Industry × Year FE	Yes	Yes	Yes	Yes
Observations	10,908	9269	8980	7305
R2	0.659	0.680	0.842	0.744

Note: Firm FE, Year FE, and Industry × Year FE denote firm fixed effects, year fixed effects, and industry-year fixed effects, respectively. Significance levels are denoted by * and ***, corresponding to 10% and 1%, respectively, with heteroskedasticity-robust standard errors provided in parentheses. Yes indicates that such variables are controlled.

, respectively. The coefficients for both types of technologies are significantly negative, suggesting that the NSMF program enhances GTFP by inducing an increased use of both end-of-pipe and source control treatment among water-polluting firms, which validates Hypothesis 2c. The results align with the findings of Omri and Bel Hadj [45] and Wen et al. [46], which indicate that environmental regulations drive firms to implement pollution treatment technologies. The further exploration of the specific pollution treatment technologies adopted by the monitored firms is warranted.

7. Conclusions

This study utilizes the NSMF program as a quasi-experiment and employs a time-varying DID model to assess the policy’s effect on the GTFP of water-polluting firms. Additionally, the analysis investigates the mechanisms through which the policy promotes GTFP growth. The key findings are as follows:

(1) The baseline regression results indicate that the NSMF program significantly enhances the GTFP growth of water-polluting firms, with technological progress being the primary driver of this effect. This finding supports the “Porter Hypothesis”, indicating that moderate environmental regulation can stimulate innovation and enhance productivity, thereby partially or fully offsetting the increased costs caused by such regulations. Furthermore, the results imply that command-and-control environmental regulations, such as the NSMF program, can effectively promote both GTFP growth and technological advancement in developing countries like China. Thus, pollution abatement and high-quality development are not mutually exclusive objectives for regulated firms. These insights are valuable for shaping future command-and-control environmental policies and advancing sustainable development in regulatory frameworks.

(2) The mechanism analysis reveals that the NSMF program enhances GTFP growth among water-polluting firms primarily by alleviating financing constraints, improving human capital, and expediting the adoption of pollution treatment technologies. The EPA directs local governments to prioritize firms listed in the NSMF program for targeted funding aimed at pollution control. This funding alleviates financing constraints, thereby equipping firms with additional resources to either upgrade their production processes or invest in pollution abatement technologies. Additionally, firms included in the NSMF program receive priority for advanced technology demonstration projects, which supports talent acquisition and strengthens human capital. This prioritization also facilitates technological exchanges with universities, research institutes, and high-tech firms, thereby promoting the adoption of more advanced end-of-pipe and source control technologies. This result underscores the importance of accounting for the incentive effects on various GTFP growth mechanisms when designing policy interventions.

(3) This study extends the existing literature by examining the effect of command-and-control environmental regulations on sustainable development within developing economies. By assessing the effects of the NSMF program—an illustrative example of such regulations—on the GTFP of Chinese firms, we provided empirical evidence that enriches the understanding of how these regulations impact sustainable development. The NSMF program can enhance firms’ GTFP while constraining their pollution emissions, making it an important policy tool for achieving a win–win situation in both economic and environmental performance. Our findings offer significant insights for the future design and implementation of similar regulations in developing countries and propose practical strategies for firms and policymakers to foster sustainable development.

The findings of this study have several important policy implications. First, the findings underscore the significant potential of command-and-control environmental regulations, such as the NSMF program, to advance sustainable development by promoting GTFP growth among firms. Governments can implement command-and-control regulations akin to the NSMF program, compelling firms to allocate greater internal resources toward green production and energy conservation. Such regulations contribute to the overarching goal of sustainable development for both firms and society. To further incentivize sustainable development and increase the effectiveness of these regulations, the government should introduce complementary policies, such as establishing exemplary cases and creating reward funds. These initiatives would provide financial incentives to firms that demonstrate excellence in green development. These supplementary measures can further motivate firms to adopt and sustain sustainable practices.

Second, the diverse pathways to achieving GTFP growth highlight the necessity for governments to tailor incentives to various contexts. To more effectively alleviate firms’ financing constraints, the governments could supplement traditional subsidies and bank credits with targeted funds and tax reductions as rewards for firms that successfully reduce pollution, which would incentive firms’ GTFP growth and technological progress. To strengthen firms’ human capital, the government can facilitate talent acquisition and development by creating online platforms for communication between firms and universities and organizing exchange events between firms and academic institutions. For advancing pollution treatment technologies, targeted funding, low-interest loans for acquiring pollution control equipment, and subsidies for clean energy or environmentally friendly materials are recommended.

Third, the NSMF program’s effectiveness is contingent upon its management by municipal ecological and environmental authorities, who currently exercise significant autonomy. There is a risk that local authorities may lower pollution emission standards or reduce constraints to promote local economic development, potentially undermining the program’s effectiveness and leading to delayed emission reductions and GTFP growth. To mitigate this issue, future reforms to the existing environmental regulatory framework could be explored by building on the current system of vertical management for monitoring and supervision and enforcement by sub-provincial environmental agencies. By further integrating these regulatory bodies within central government departments, enforcement capabilities could be significantly strengthened.

This study is subject to several limitations. First, the unavailability of data on firms’ pollution emissions beyond 2013 precludes an assessment of the NSMF program’s effect on GTFP after this period. Future research should address this gap as post-2013 data become available. Second, the focus on heavily polluting industries leaves the effect of the NSMF program on less polluting sectors unexplored, an area that merits further investigation. Finally, while this study concentrated on command-and-control regulations, future research could investigate the effect of market-incentive and voluntary environmental regulations on GTFP.