Environmental regulation and labor demand: Evidence from China's air pollution prevention and Control Action Plan

Abstract

In response to environmental regulatory pressure from the government, enterprises – as the main employers of labor – adjust their production-related choices and may alter their demand for labor. Numerous researches have probed into how environmental regulations affect the labor demand, which has emerged as a significant concern in the discourse around the adoption of environmental regulation policies. Nonetheless, many researches predominantly concentrates on the evaluation on the effects of environmental regulations on economic and environmental levels. Moreover, rare attention has been paid to how environmental regulations affect social activities, especially in terms of labor demands. As a result, the Air Pollution Prevention and Control Action Plan (APPCAP) is adopted as a “quasi-natural experiment” and a difference-in-differences model is employed to analyze the effects of the APPCAP on labor demands of companies. Hence, the 2008–2020 panel data are considered for the 3,949 Chinese A-share listed enterprises. Furthermore, this paper probes deeply into the underlying mechanisms based on mediation models. The following findings are thus concluded: (1) The labor demands of enterprises can be remarkably increased by APPCAP; and the result is still persuasive even though the endogeneity issues are taken into account. (2) APPCAP improves enterprise labor demand through the output and factor substitution effects. (3) As evidently revealed by heterogeneity analysis, the APPCAP could significant positive affect the labor demand size in state-owned companies, large-scale companies, new companies, and companies in polluting industries. (4) The APPCAP strikingly boosts the demand for labor force with high skills. Nevertheless, it exerts little influence on the demand for labor force with low skills and company salary levels. Therefore, the government must continue to steadfastly implement environmental regulatory policies but adopt different policies based on enterprise characteristics. Overall, this study provides micro-level experimental evidence for more in-depth understanding of how environmental policies affect labor market, which is particularly important for actively resolving social employment problems and exploring new growth points in enterprise employment.

1. Introduction

Environmental degradation has evolved into the principal barrier holding back the high-quality growth of global economy. As a consequence, effective environmental governance is one of the focuses of existing academic research. Given an increasingly prominent contradiction between environmental pollution and high-quality economic development [1], the Chinese policy-makers have confronted with a dilemma in how to realize a balance between the pursuit of economic development and the consideration for environmental protection when bringing forth environmental schemes [2]. Environmental regulations not only enhance the environment and robust economic development but also pose new problems and challenges to changes in employment. Against the background of both domestic and international instability and uncertainty, the Chinese government has always placed employment at the top of the stability on the six fronts and security in the six areas. The Chinese government is facing the pressing task of addressing the problem of achieving consistent employment growth. This is crucial for sustaining social stability, fostering a peaceful society, and advancing the national economy towards a high-quality stage of development. The 20th National Congress of the Chinese Communist Party's report advocated to “deeply promote the prevention and control of environmental contamination, continue to fight well in the defense of blue sky, clear water and pure land” and emphasized “implementing the employment priority strategy”. Hence, how can we harmonize and achieve the two development goals mentioned above, and how can we achieve a mutually beneficial outcome between environmental regulation and employment? Answers to such questions are particularly important for resolving social employment problems and exploring new growth points for enterprise employment.

In recent years, environmental regulations in China have become increasingly stringent. However, the consequences of its implementation has not been fully analyzed. The Air Pollution Prevention and Control Action Plan (APPCAP) executed by the State Council of China in 2013 has become the strictest policy ever adopted to eliminate air contamination and encourage eco-friendly economic expansion. As the first comprehensive action plan in China to address prominent environmental issues, it is particularly noteworthy that APPCAP has the strongest intensity and execution in reducing air contamination when compared to all previous policies and attempts. The requirements for air quality in key areas have thus achieved optimal results. The effective implementation and enforcement of air contamination control and prevention measures within China's national and local governments provide valuable insights for continuously improving air quality and efficiently resolving climate change issues. The APPCAP serves as a significant measure in China's efforts to reduce air contamination and enhance environmental quality. It can be regarded as a distinct quasi-natural experiment to systematically evaluate how environmental regulations affect the economy, society and environment. Nonetheless, most of the focus of this study is how to evaluate the impact of environmental regulations on the environment and economy, rather than systematically investigating how environmental regulations affect social activities like labor demands. Coincidentally, a thorough investigation on this impact is crucial for achieving China's social and economic stability as well as high-quality sustainable growth. As a result, this study preliminarily concentrates on studying the APPCAP's effects on labor demands of companies. Our research not only enhance comprehension of the correlation between environmental regulation and employment, but also provide specific policy recommendations for attaining a mutually beneficial outcome between environmental regulation and employment.

Among the many causes of contamination, companies’ industrial activities are the primary factor leading to environmental deterioration. As a consequence, it is necessary for companies to undertake the leading role in environment protection, which is also the predominant target for environmental regulation. However, due to the nature of public goods in the environment, companies are not enthusiastic about environmental regulation. The environmental regulation is predominantly intended to internalize the negative externalities of companies and force them to allocate part of their labor and capital for contamination reduction in the form of laws and regulations, technical standards, environmental taxes, emission charges and tradable emission rights to achieve the goal of environmental protection. As a result, as the main employer of labor, will companies adjust their production decisions and change their demand for labor when facing environmental regulatory pressure from the government?

To investigate how environmental regulations affect employment not only has captured extensive attention for a long time, but also has brought about a major concern in the discussion about the implementation of laws regarding environmental regulation. At present, multitudinous existing studies have been reported to investigate how environmental regulations affect employment. Nevertheless, there was no consensus among these researchers. There are three primary perspectives. First, environmental regulations do not promote overall employment growth. Prior to this, experts held the belief that environmental regulations would exert pressure on firms to decrease emissions and manage pollution, resulting in elevated production and governance expenses. This, in turn, could weaken their competitive advantage, resulting in reduced enterprise size and decreased labor demand. These factors contribute to the negative employment scale effects of environmental regulations [[3], [4], [5], [6], [7], [8]]. Second, some scholars oppose the view that environmental regulations reduce employment. They contend that in order to adhere to updated and more stringent environmental standards, businesses must employ individuals to develop and upkeep pollution control apparatus or modify manufacturing methods to diminish waste creation. The resulting increase in labor demand within enterprises could foster employment growth [[9], [10], [11]]. Finally, some scholars argue that at China's present stage, particular conditions and foundations are necessary to achieve both environmental protection and employment. As a result, it is unclear how environmental regulations affect the employment of regulated companies. As illustrated by the Porter Hypothesis, well-formulated environmental regulations could potentially contribute to sci-tech innovation within enterprises. However, determining the employment effect resulting from changes in production technology is challenging due to variations in the time cycle of technology accumulation. Shea's [12] study demonstrated that an increase in the technology level (measured by R&D spending and patent applications) could initially boost employment but influence negatively over a longer period of time.

The impact of technical advancement on innovation as a whole remains ambiguous. From an industry perspective, the impact of environmental regulations on employment effects cannot be definitively determined due to differences in energy intensity [5,13]. From a regional perspective, variations in environmental regulatory standards lead to the spatial transfer of employment. Labor mobility between environmentally regulated and non-regulated areas may have an unpredictable effect on employment [5]. These differing perspectives highlight the need for more investigation into the correlation between environmental regulations and the demand for labor in businesses.

After the APPCAP was announced by the State Council of China in 2013, numerous scholars have conducted researches on the consequences brought by this policy and evaluated its economic and environmental implications comprehensively. With regard to the evaluation on the effects of the APPCAP on environment, researchers have confirmed that air quality has been dramatically improved by the implementation of this program [[14], [15], [16], [17], [18], [19]]. The improvement in air quality has resulted in significant health benefits [[17], [18], [19]]. Regarding the assessment of the economic impacts of the APPCAP, some scholars have found that the APPCAP has improved economic benefits [20], green innovation levels [21], total factor productivity [22], and the industrial green total factor productivity [23]. When it comes to the APPCAP's effects on labor demands, Li et al. [24] hold a view that the influence of APPCAP on labor demand is time-varying. Specifically, labor demand first declined and subsequently rebounded between 2013 and 2019. In the sector of clean industry, the APPCAP has remarkably improved the employment opportunities. Liu et al. [2] believe that the APPCAP promotes technological progress, improves labor productivity, and significantly reduces labor demand in the manufacturing industry.

Based on the above literature, there is insufficient research on the effect of APPCAP on labor demand, and no consensus has been reached. Hence, on the basis of the microdata about A-share listed companies between 2008 and 2020, how APPCAP affects labor demands of companies is investigated from three perspectives, effectiveness, heterogeneity, and influence mechanisms. Instrumental variables estimation is used to address the endogenous problem.

In this article, we make the following main contributions. First and foremost, in terms of employment, thorough research has been carried out to explore how the APPCAP influences the labor demands in companies. As proven by evidence from China, environmental regulations can exert certain impact on labor demands. Such evidence has conspicuously broadened the viewpoints of researches on the influential factors of labor demands and vigorously facilitated researches on the environmental regulations' social effects. The second objective is to investigate the theoretical mechanism by which the APPCAP affects the demand for labor in companies. Additionally, we aim to empirically examine the presence of both output and factor substitution effects, which are crucial for achieving the dual benefits of environmental governance and employment growth. Third, in order to probe deeper into the complicated impacts of the APPCAP on companies' labor demands, this paper specifically looks into the heterogeneity of the policy's impacts on labor demands of companies with different ownership levels, for enterprises of different ages and sizes, and clean and polluting industries. To a certain degree, we have expanded the research on the heterogeneity of the environmental regulations' impacts on different types of enterprises while providing a reference for government departments to design more reasonable and targeted environmental regulation policies.

Subsequently, this paper are organized in the following structure: Section 2 provides a theoretical basis and puts forward the research hypothesis; Section 3 presents the methodology and relevant data; and the empirical results are displayed in Section 4, which comprises the results obtained from benchmark regression analysis; Section 5 exhibits the analytical results for heterogeneity; mechanism analyses are demonstrated in Section 6 comprehensively; and Section 7 is comprised by the discussion and conclusion.

2. Theoretical basis and research hypothesis

2.1. Conceptual framework

According to the partial equilibrium model of the production function used by Brown and Christensen [25], enterprise development comes with quasi-fixed costs based on external constraints. Berman and Bui [26] follow the assumption of a local static equilibrium model and believes that enterprises allocate resources to quasi-fixed elements during production, such as emission reduction investments that meet environmental regulations. Investments in these quasi-fixed factors is not only determined by the principle of cost minimization but also subject to the external constraints of the enterprise. We draw on the approach of Berman and Bui [26] and consider environmental governance investments generated by enterprises to comply with the APPCAP as a quasi-fixed factor and other production costs, such as labor and raw materials, as variable factors.

Assuming that the enterprise follows the cost minimization principle when deciding factor inputs, its production factors include M variable and N quasi-fixed factors.

The function of enterprise variable is defined below:

$$CV=F\left(Y,P_1,\cdots P_j,Q{Z}_1,\cdots ,Q{Z}_K\right)$$ (1)

As suggested by the above equation, CV symbolizes the enterprise's variable cost; Y is the enterprise's output; P_j (j = 1,2, …,J) represents the price of the jth variable element; and Z_k (k = 1,2, …,K) is the nth type of the quasi-fixed factor input. Based on the Shephard's lemma, the demands for the variable factor labor L can be defined as the function of the enterprise's output, the price of variable factor, and the input of quasi-fixed factor. On this basis, the labor factor demand function can be obtained according to equation (1):

$$L=\alpha +{\rho }_Y\times Y+\sum \begin{array}{c}J\\ j=1\end{array}{Y}_j{p}_j+\sum \begin{array}{c}K\\ k=1\end{array}{\beta }_k{Z}_k$$ (2)

Among them, β depends on whether the environmental governance investment and labor brought about by the APPCAP have substitutive or complementary relationships. If R represents the APPCAP, $\mu$ represents the APPCAP's marginal impact on labor demand, and $\delta$ represents the impact of other factors besides environmental regulatory policy, then equation (2) can be simplified as:

$$L=\delta +\mu R$$ (3)

Simultaneously calculate the first-order partial derivative of R on both sides of equation (3) and obtain:

$$\frac{dL}{dR}={\rho }_Y\frac{dY}{dR}+\sum \begin{array}{c}J\\ j=1\end{array}{Y}_j\frac{d{P}_j}{dR}+\sum \begin{array}{c}K\\ k=1\end{array}{\beta }_k\frac{d{Z}_k}{dR}=\mu$$ (4)

According to Berman and Bui [26], if the factor market is competitive and large-scaled, environmental regulation has a minimal impact on the pricing of other variable variables, while the second term in formula (4) is close to zero. Under such a condition, equation (4) can be transformed into:

$$\frac{dL}{dR}={\rho }_Y\frac{dY}{dR}+\sum \begin{array}{c}K\\ k=1\end{array}{\beta }_k\frac{d{Z}_k}{dR}$$ (5)

In accordance with equation (5), there are two effects that influence the APPCAP's impact (R) on labor demand (L): the output effect is the first item on the equal sign's right side, indicating that the APPCAP impacts labor demand through enterprise output. Implementing the APPCAP would additionally increase the environmental regulatory costs of companies, resulting in decreased production costs. This may lead to the reduction in the production capacity and scale of companies, with dY/dR＜0, which results in decreasing labor demand. On the contrary, implementing the APPCAP negatively impacts companies' technological innovation. companies could decrease expenses by using cleaner production methods, increase the size of their operations, and have a favorable influence on labor demand, namely, when dY/dR＞0 [26]. It is also recognized as the factor substitution effect stemmed from the APPCAP. The effects of APPCAP on labor demands predominantly rests with the balance between the effect of factor substitution and the effect of output.

2.2. Theoretical hypotheses

According to the examination of the mathematical model provided above, the implementation of the APPCAP changes the costs of companies, affects their behavior, and subsequently impacts the demand for labor via the effects of output and factor substitution. In the subsequent section, the APPCAP's effects on enterprises' labor demands are further analyzed on the basis of mathematical models.

2.2.1. The output effect

The effect of output originates from the APPCAP's effects on labor demand by virtue of environmental decision-making and output scale. As the main polluters of the environment, enterprises face strict supervision from government departments, which further affects their production activities and ultimately changes their demand for labor. Implementing environmental regulatory policies like the APPCAP raises production costs for businesses, resulting in higher product prices. This, in turn, reduces consumer demand and forces companies to decrease production, leading to a decrease in labor demand. After implementing the APPCAP, enterprises will face more stringent emission standards. Influenced by externality theory, the simplest response of companies is to reduce emissions by lowering production, which ultimately lessens their labor demands. This measure, however, is extremely detrimental to their development from a long-term perspective. However, environmental regulations could also achieve employment growth for enterprises. The execution of environmental regulation policies such as the APPCAP could improve the overall environment in which enterprises are located, promote the transfer of environmentally preferable labor from other regions to local employment and, thus, provide enterprises with a more extensive range of labor choices. They are more inclined to adopt scale expansion decisions and encourage enterprises to absorb more labor. In addition, implementing the APPCAP requires enterprises to attract more employees to operate pollution control equipment, further enhancing their labor demand.

On the other hand, enterprises proactively comply with environmental regulating regulations, such as the APPCAP, in response to government directives. they do this by investing in clean manufacturing equipment and using green production technologies to minimize pollution. As suggested by the Porter Hypothesis, environmental regulations drive companies to improve and innovate their products and technologies while offsetting the possible cost increase caused by such regulations. Consequently, this leads to a surge in market demand for products and an expansion in the output of enterprises, ultimately resulting in an upsurge in labor demand by these enterprises. Therefore, reasonable environmental regulations could help enterprises innovate and offset, to a certain extent, the adverse consequences of rising production costs [27]. In addition, enterprises may also use environmental regulations to enhance their core competitiveness and compete for market share with innovative products, which could better meet the current consumer demand for product diversification. In the process of environmental regulation and market consolidation, enterprises could widen the gap with potential competitors, achieving the goal of occupying more market share and even exploring new markets. In this way, enterprises further augment their need for workers. In summary, implementing the APPCAP affects productivity of businesses and influences labor demand.

2.2.2. The factor substitution effect

As revealed by the effects of factor substitution, companies enhance investments in environmental protection because of the APPCAP's influence, so as to abide by environmental regulations, which thereby affects labor demand. Under the power of the APPCAP, enterprises are bound to face stricter government environmental regulations, forcing them to increase their investments in pollution control measures, such as acquiring state-of-the-art manufacturing equipment and using either pre-existing or newly created clean production technologies. The investments of enterprises in environmental regulations predominantly involve two aspects, end-of-pipe treatment and production procedure improvements. The end-of-pipe treatment denotes the treatment of pollutants by enterprises to reduce emissions after and before pollution is generated. The necessary links, such as equipment installation, operation, and maintenance required for terminal governance, extends the production chain of the enterprise, thereby increasing the demand for labor. For example, some exhaust gas equipment technologies are end-of-life pollution control technologies, and human resource investments need to be increased in operation and monitoring. This technology could also convert by products from production into commodities, improving the company's profitability and creating more job opportunities [28]. Production process improvements refer to reducing pollution emissions by improving production processes, adopting greener equipment and technologies, and other means. Optimizing the production process of an enterprise can enhance its production efficiency. At the same time, traditional capital and labor production factors could also be effectively replaced; however, doing so may reduce the enterprise's labor demand. Furthermore, as China's environmental regulations continue to become more stringent, enterprises' demand for clean production technology and equipment has increased, and the price of capital has decreased. However, due to a reduction in the demographic dividend, the labor cost in China has also shown an upwards trend [6], making it easier to replace labor as a factor of production. In addition, the government generally provides financial support to enterprises engaged in environmental management or those with the capacity for environmental management. This encourages enterprises to acquire environmentally friendly manufacturing equipment, introduce or independently develop green production technologies, enhance production effectiveness, and reduce labor demand.

In summary, the APPCAP may influence the demand for labor in businesses by impacting both the output and factor substitution impacts. On the basis of the aforementioned analysis, this paper comes up with the research hypotheses in the following.

Hypothesis 1

APPCAP changes enterprise costs and impacts labor demand.

Hypothesis 2

APPCAP may affect enterprise labor demand through the output and factor substitution effects.

3. Data and methodology

3.1. Data collection and sample

In consideration of the feasibility of this research, as well as the implementation time and lag of APPCAP, this paper chooses the research interval of 2008–2020 and the research sample of micro-data related to Chinese A-share listed enterprises. The primary data are processed in the following procedures: ① listed companies marked by “*ST” and “ST” in the stock abbreviations are excluded; ② listed companies with a “*ST” and “ST”, “Terminated listing”, or “Suspended listing” listing status are excluded; ③listed companies with serious missing data are excluded; ④ the method of linear interpolation is adopted to fill in certain missing data of those individual listed companies. After the above processing, the research sample of this article includes 3,949 listed companies and a total of 28,942 observations. The source of variables for all listed companies are originated from the China Stock Market & Accounting Research Database (CSMAR). The descriptive statistical results of main variables are shown in Table 1. The reason why A-share listed enterprises are chosen as the research sample is that A-share listed enterprises, compared to state-owned companies, are not necessarily required to bear the political responsibility of offering job opportunities. Analyzing the influence of the APPCAP on labor demand by using A-share listed firms as samples yields more efficient results. This article is based only on the dataset from 2008 to 2020 for analysis; therefore, we cannot capture more recent possible impacts. Furthermore, using data from A-share listed Chinese companies may not fully reflect the labor demand of Chinese companies.

Table 1.

Summary statistics of variables.

variables	Obs	Mean	Std. Dev.	Min	Max
labor	28,942	7.6321	1.3436	2.0794	13.2227
lperwage	28,942	9.4249	1.2672	−2.1994	15.6567
firmage	28,942	2.8118	0.3909	0	4.1431
soe	28,942	0.3659	0.4817	0	1
size	28,942	22.1713	1.4929	15.5773	31.1379
lev	28,942	0.4322	0.2187	0.0071	3.2213
ltax	28,942	17.1903	1.8520	3.8670	25.1720
tobinq	28,942	2.0895	2.7239	0.1528	259.1459
roa	28,942	0.0419	0.0747	−1.1296	0.8796
ser	28,942	0.0805	0.9693	−0.2091	164.3244
lwage	28,942	11.1659	0.4553	9.2317	12.1283
lpgdp	28,942	11.3379	0.5703	8.3885	13.0557
lpop	28,942	6.4627	0.6986	3.4003	8.1362
ind	28,942	1.5227	1.0121	0.0943	5.3482
lrsc	28,942	17.0725	1.1092	12.5080	18.8865

3.2. Measures

3.2.1. Dependent variable

According to the research conducted by Liu et al. [8] and Gray et al. [29], the dependent variable is enterprise's labor demand in this paper, which is calculated as a logarithm for the sum of employees in the company for the present year.

3.2.2. Independent variable

We conduct a quasi-natural experiment using the APPCAP as an exogenous impact. If a company was in 57 high target cities of the APPCAP in 2013, the value is 1; otherwise, it is 0.

3.2.3. Control variables

• (1)At the level of enterprise, the control variables comprise the salary level of enterprise (lperwage), which is calculated as the logarithm of the average employee salary. Based on the labor supply and demand theory, salary could reflect the labor cost in the production of enterprise, whereas the salary level is in inverse proportion to the labor demand. Increasing salary will leads to the decrease in the enterprises' labor demand. The companies' age (firmage) is calculated by the logarithm of the total number of years since the enterprise was listed. The property rights of enterprises (soe) are determined by the ultimate controller's nature. With the ultimate controller of a state-owned unit, it will be a state-owned company; or it will be a non-state-owned company. The enterprise size (size) is calculated as the logarithm of the total assets of this enterprise. In general, for most industries, an enterprise with a larger size will have a greater labor demand and absorb more employment. Debt to asset ratio (lev) is an important indicator that reflects enterprise's capital structure, debt level, and financial risk. Operators generally choose an appropriate debt-to-asset ratio based on reducing financial risks and fully utilize borrowed funds for investment in equipment, labor, and raw materials. Hence, the debt-to-asset ratio can impose indirect impacts on the labor demand input from companies. The debt-to-asset ratio in this paper is calculated through the ratio of total liabilities to total assets. The Income tax (ltax) may influence enterprises' labor demands in terms of two aspects. Firstly, the increasing corporate income tax will elevate the capital use cost, decrease enterprises' profits whereas squeeze out the application of the labor production factors by companies. Secondly, with the increasing corporate income tax, the production will be expanded, and labor input will be enhanced as well [11]. Due to the presence of non-positive numbers in corporate income tax expenses, this article adds 1 to them and performs logarithmic processing. Enterprise growth capability (tobinq). An enterprise with strong growth ability has a greater potential for scale expansion and a higher demand for labor input. Tobin Q is one of the indicators used to measure an enterprise's long-term growth ability. This indicator is adopted in this paper to reflect how the labor demand of an enterprise is affected by its growth ability. The rate of return on total assets (roa) in this paper is reflected by the ratio of net profit to total assets. The ratio of selling expense (ser) refers to the sales expenses incurred by a company to obtain unit operating income. This indicator reflects the level of marketing efficiency of the enterprise. Ren et al. [11] found that a low selling expense ratio implies high marketing efficiency and great potential for improving business performance, which could encourage companies to increase their workforce. The ratio of selling expense in this paper is represented by the ratio of selling expenses to operating income.
• (2)The control variable at city level. This paper, by referring to the study conducted by Liu et al. [2] and Li et al. [30], simultaneously controls characteristic variables that might influence the labor demands of companies at city level, comprising the wage level of the city (lwage), which is demonstrated by the logarithm of the urban in-service employees' average salary; The level of economic development (lpgdp) is represented by the logarithm of per capita GDP; the size of population (lpop) is defined as the logarithm of a city's total population by the end of the year; the industrial structure (ind) is calculated by the output value ratio of the tertiary sector to that of the secondary sector; and the level of consumption (lrsc) is defined as the logarithm of overall retail sales of consumer products in the society. It is noteworthy that all the urban data originates from the China City Statistical Yearbook in the past years.

3.3. Methodology

3.3.1. Baseline model

Based on the APPCAP issued by the Chinese government in 2013, the concentration of the inhalable particulate matter (PM₁₀) in Chinese cities at or above the prefecture level by 2017 decreased by over 10 % compared with that in 2012. Moreover, in such cities, the total number of days with good or moderate urban air quality increases as well yearly by year. Compared with those in 2012, the concentrations of fine particulate matter (PM_2.5) in regions such as the Pearl River Delta, the Yangtze River Delta, and Beijing-Tianjin-Hebei in 2017 declined by about 15 %, 20 %, and 25 %, respectively. However, the APPCAP did not specify specific compliance tasks for other pollutants and cities outside the abovementioned areas. Due to the varying implementation intensity of the APPCAP nationwide, we consider the three regions that are the focus of the APPCAP, namely, Beijing-Tianjin-Hebei, Yangtze River Delta, and Pearl River Delta, as high target cities, whereas other cities are considered low target cities.

By adopting the execution the APPCAP in 2013 as a quasi-natural experiment and the issuing time of the policy as the time node, this paper selected the high-target cities as the experimental group, while the control group was constituted by low-target cities. According to the method put forward by Liu et al. [2] and Li and Chen [22], a DID model was constructed as follows:

$$labo{r}_{ijt}={\beta }_0+{\beta }_1\cdot city\times tim{e}_{ijt}+{\beta }_2\cdot X_{it}+{\beta }_3\cdot Z_{ij}+{\rho }_j+{\tau }_t+{\epsilon }_{ijt}$$ (6)

In the above equation, i symbolizes the enterprise; j refers to the city of the enterprise's location; and t represents the year. The dependent variable labor stands for the enterprise's labor demand and is calculated as the logarithm of the total number of employees within the enterprise in the current year. city × time, as a dummy variable, is employed to imply whether the city of the enterprise's location was a high-target city for the APPCAP in 2013; and when the city of the company's location is in a high-target area, the value is 1 or otherwise it is 0. X represents the enterprise's control variable over time; Z symbolizes the control variable of the city level over time; ρ_i represents the enterprise's fixed effects; τ_t is the time fixed effect; and ε_ijt denotes the error term allowing for enterprise -level clustering.

As a specialized approach in economics, the DID model is adopted to measure how effective the policy implementation is. Through this approach, policies are typically considered as exogenous variables that have an impact on the research object, addresses the endogeneity issues that arise when policies are used as explanatory variables. Hence, the approach is applied extensively for analyzing the policy. The DID model is simple and effective and can avoid the endogeneity problem of policies as independent variables and obtain unbiased estimates of policy effects.

To ensure the precision of the DID model, a key condition is that changes in public policy must be exogenous or not correlated with the regression equation's error term. As a result, a major threat to the rationality of applying the DID model for policy assessment is that tested subjects might not be assigned to the control group or the treatment group randomly. As it is the government that assigns the pilot cities of the APPCAP, it is somewhat mandatory to be a pilot city. In this sense, the selection of a city as a pilot city relies on whether this city is included in the national planning instead of the city itself. Hence, the APPCAP is exogenous; and no endogeneity issues exist.

3.3.2. Parallel trend test for the DID method

The benchmark regression model for the DID method which was structured earlier reflects the average impact of the environmental regulation on labor demands of enterprises. When exploring the causal relation between labor demands of enterprises and environmental regulations, the parallel trend hypothesis should be satisfied. Before the APPCAP was implemented, the labor demands of enterprises in high-target cities were similar to those of low-target cities. Referring to the methods of Beck and Levkov [31], this research constructed a parallel trend test model in the following:

$$labo{r}_{ijt}={\beta }_0+{\beta }_1{D}_{ijt}^{-4}+\dots +{\beta }_{11}{D}_{ijt}^7+{\beta }_{12}{X}_{it}+{\beta }_{13}\cdot Z_{ij}+{\rho }_j+{\tau }_t+{\epsilon }_{ijt}$$ (7)

In this equation, D_ijt denotes the dummy variable of the APPCAP. The value will be 1 if the city j, which is the place of location of the enterprise i, is one of the high target cities in the APPCAP in the year t; otherwise, it is 0. D_ijt⁻⁴ is equal to one for all years before a city becomes a high target city for the APPCAP for 4 years or more; D_ijt⁷ is equal to one for all years after a city becomes a high target city for the APPCAP. The settings for the other variables are the same as in equation (6).

3.3.3. Robustness test

• 1.Urban placebo test. To further verify that the enhancement in the labor demands of enterprise is spawned from the execution of the APPCAP instead of being affected by other factors which are unpredictable. A placebo test was conducted in this research by referring to the practices in relevant literature [32]. In an effort to randomize APPCAP's effects on specific regions, we randomly select experimental and control groups from all samples; subsequently, repeated the sampling for 500 times, and estimated the benchmark Model (6) repeatedly to obtain 500 coefficient estimates on the APPCAP's impacts on the labor demands of enterprises, so as to ensure that no other factors influence the execution of the APPCAP.
• 2.Time placebo test. In order to further identify whether the estimated results in this paper are robust, this research conducted a time placebo test by referring to the approach of Topalova [33]. The implementation time of APPCAP wasincreased by 3, 4 and 5 years, and samples of pseudo-policy implementation time were constructed and represented by city × time－^false3, city × time－^false4, and city × time－^false5, separately. Equation (6) is used to perform regression.
• 3.Sample data screening. Since there might be extreme values of research samples that could possibly impose certain influence on the benchmark regression results, a 1 % and 5 % winsorized treatment was performed on the dependent variable (labor) prior to the application of the benchmark Model (6) for regression.
• 4.Propensity score matching-difference in difference (PSM-DID). With the purpose to minimize the effects from the systematic difference between low- and high-target cities in the APPCAP on the estimation results of DID method, the approach of PSM-DID is thus adopted to explore the APPCAP's impact on the labor demands of enterprises. The precondition of this approach is that the change trend of the control group and the experimental group should be the same before the APPCAP was implemented without any remarkable difference. If enterprises in high-target cities within the scope of the APPCAP have a greater demand for labor than do enterprises in other cities, the effectiveness of the benchmark regression results is greatly compromised. Hence, the PSM-DID approach is employed in this paper for the robustness test, so as to mitigate the possibility of the self-selection issue within the change trend of the control group and the experimental group, which would give rise to bias in regression results. We use the k-nearest neighbor matching method (k = 3) . Regression is performed using Equation (6).
• 5.Eliminating the impact of other environmental regulation policies.

In order to confirm whether the increasing labor demands of enterprises is stemmed from the APPCAP instead of being affected by other policies, we aim to ensure the accuracy of the benchmark estimates by controlling for other major environmental policies during the sample period. Through the sorting and collection of related policy documents, it has been found that two environmental policies might have influenced the labor demands of enterprises in the sampling period. The first is the Environmental Protection Law of the People's Republic of China (new Environmental Protection Law), which was implemented in 2015 and may impact enterprise labor demand. The legal responsibilities of economic entities, such as the government, enterprises, and individuals, in environmental pollution supervision, environmental protection, and prevention are clarified in the law known as the most severe environmental protection law historically, Second, three batches of the pilot programs for low-carbon cities, regions and provinces are carried out by the National Development and Reform Commission of China in 2010, 2012 and 2017. China has continuously promoted green and low-carbon transformation for many years, causing various industries to face transformation and upgrading. Employers must face more intense job competition, resulting in job changes or even job elimination, which has a particular impact on enterprise labor demand. We include pseudo-variables for these policies in the benchmark regressions to avoid interference from other policies in the implementation of APPCAP. envirpolicy reflects whether the city of the enterprise’ location executed the new Environmental Protection Law in 2015; and if the city did, envirpolicy would be 1 and otherwise it would be 0. LCPC indicates if the city of the enterprise’ location was a low-carbon pilot city in that year. If the city was, LCPC would be 1 and otherwise it would be 0. To eliminate the interference of those two policies on results, Model (6) included envirpolicy and LCPC.

• 6.IV estimation.

Reverse causality and omitted variables might cause endogenous problems when APPCAP's effects on labor demands are analyzed. We use each city's annual ventilation coefficient (VC_it) as the instrumental variable for APPCAP. The ventilation coefficient is constructed as VC_it = WS_it × BLH_it, where BLH_it and WS_it represent the height of boundary layer and wind speed, separately. The original data concerning the height of atmospheric boundary layer and wind speed source from the ERA-INTERIM latitude and longitude grid meteorological data issued by ECMWF, which are then parsed into prefecture-level city data further so that they can be used directly via ArcGIS.

Due to the two considerations below, ventilation coefficient is employed as the instrumental variable. Firstly, when the ventilation coefficient value is larger, the air mobility will be stronger; and the possibility of leading to air contamination like haze is smaller [34]. This city is less likely to be a high target for APPCAP, which could satisfy the correlation assumption of the instrumental variables. Nevertheless, the ventilation coefficient is related to the height and wind speed of the atmospheric boundary layer. In contrast, geographical and weather factors could influence the height and the wind speed of the atmospheric boundary layer, which are not directly correlated with labor demands and could satisfy the exogenous hypothesis of the efficient instrumental variables [35].

4. Empirical results

4.1. Baseline regression estimation

Based on the baseline Model (6), the APPCAP's impact on enterprise labor demand is assessed using two regression specifications, and the results are presented in Table 2. Time and firm fixed effects were controlled for in Column (1), and city time's coefficient is 0.1091 at the 1 % significance level. After the control variables at the enterprise and city levels were added and the results are reflected in Column (2), the coefficients of the core independent variable city × time were still credible at the 1 % significance level. Regardless of whether control variables are added, the APPCAP has promoted increased enterprise labor demand. Worth noting is that, in Column (2), the promotional effect of environmental regulations on enterprise labor demand is smaller than the policy effect of environmental regulations as reflected in Column (1), indicating that the control variables selected in this article indeed have an impact on enterprise labor demand. Therefore, compared to enterprises not located in the high-target cities of the APPCAP, enterprise's labor demand in the high-target cities of the APPCAP has significantly increased. When facing environmental regulations, listed companies have achieved a double dividend of environmental governance and an expansion of labor demand, which is consistent with Hypothesis 1.

Table 4.

Robustness test of APPCAP's impact on enterprise labor demand.

• 6.
  IV estimation. We select the ventilation coefficient as the instrumental variable to alleviate endogeneity problems. Table 5 shows the results of IV estimation. Column (2) lists the IV estimation result of the first stage. The higher the ventilation coefficient is, the less likely the city is to become a high target of the APPCAP, which is also consistent with the expectations of this chapter. The F statistic values of the first-stage regression results were higher than the critical value of the Stock-Yogo weak instrumental variable test (8.96), indicating that the instrumental variable (VC) used in this paper does not have a weak instrumental variable problem. The Kleibergen‒Paap rk LM statistic passed the significance test of 1 %, indicating no problem of insufficient identification of instrumental variables. The Hansen J overidentification test shows no overidentification problem of the instrumental variables, indicating that the tool variables of the air flow coefficient are effective. The second-stage regression result of the IV estimation is shown in column (3). The estimated coefficient of city × time is significantly positive, indicating that the impact of the APPCAP on labor demand is significantly positive. The positive impact of implementing APPCAP on labor demand still exists after considering endogeneity issues, indicating that the estimation results of the benchmark regression in this article are robust. The estimated coefficient obtained by the instrumental variable method is higher than that obtained by the benchmark regression, which indicates that the positive impact of APPCAP on labor demand may be underestimated due to the existence of endogeneity problems.

	(1)	(2)	(3)	(4)	(5)
	1 % winsorized	5%winsorized	PSM	Excluding policy interference
city × time	0.0632***	0.0456**	0.0640***	0.0675***	0.0591**
	(0.0235)	(0.0214)	(0.0242)	(0.0241)	(0.0170)
Control Variables	yes	yes	yes	yes	yes
Constants	−4.2476***	−3.5146***	−4.4184***	−4.2898***	−4.2337***
	(1.0654)	(0.8954)	(1.1717)	(1.1499)	(1.1226)
Firm FE	yes	yes	yes	yes	yes
Year FE	yes	yes	yes	yes	yes
envirpolicy	no	no	no	yes	no
LCPC	no	no	no	no	yes
Observations	28,942	28,942	27,597	28,942	28,942
R2	0.9187	0.9180	0.9272	0.9282	0.9287

Table 2.

Baseline regression estimation.

	(1)	(2)
	labor	labor
city × time	0.1091***	0.0638***
	(0.0308)	(0.0239)
lperwage		−0.2100***
		(0.0143)
firmage		0.1420**
		(0.0614)
soe		0.1160**
		(0.0467)
size		0.6310***
		(0.0199)
lev		0.2300***
		(0.0673)
ltax		0.0219***
		(0.0053)
tobinq		0.0047**
		(0.0021)
roa		−0.0584
		(0.0895)
ser		0.2820*
		(0.1520)
lwage		−0.1240
		(0.0839)
lpgdp		−0.0021
		(0.0278)
lpop		−0.0086
		(0.0289)
ind		0.0232
		(0.0223)
lrsc		0.0186
		(0.0410)
Constants	7.6037***	−4.2050***
	(0.0114)	(1.1260)
Firm FE	yes	yes
Year FE	yes	yes
Observations	28,942	28,942
R2	0.8818	0.9290

Note: Standard errors in parentheses. ***、**and * represents p < 0.01, p < 0.05, and p < 0.1, respectively. The empirical results are clustered at the enterprise level. These values are the same in the tables below and will not be repeated.

4.2. Parallel trend test result

We obtained the results of the parallel trend test using equation (7). The results and 95 % confidence interval are shown in Fig. 1, which have been adjusted for enterprise-level clustering. The left side of the dashed line in Fig. 1 represents the period before implementing the APPCAP, and the right side of the dashed line represents the period after implementing APPCAP. Fig. 1 shows that the estimated coefficients for each period before the implementation of APPCAP are not significant, indicating that the difference between pilot and non-pilot city enterprises before policy implementation is not significant. On the right side of the dashed line, the difference between the treatment and control groups gradually began to emerge during the period after the implementation of the APPCAP, and enterprises' labor demand in treatment group showed a significant upwards trend. In summary, the research sample in this article satisfies the parallel trend assumption of DID and implementing APPCAP may have promoted improvements in enterprise employment.

Fig. 1.

Parallel trend test result.

4.3. Robustness test results

• 1Urban placebo test result.

The kernel density and p value distribution of the 500 estimated pseudo regression coefficients are shown in Fig. 2. The solid black line is the estimated coefficient distribution of randomly selected pseudo-city samples, the blue scatter points represent p values of estimated coefficients, and the horizontal dashed line is the 10 % significance level. From Fig. 2, on the one hand, most regression coefficients fall within the range of [−0.01, 0.015], and the corresponding p values are mostly not significant. On the other hand, the estimated value of the city × time coefficient in the benchmark regression is 0.0638, which is significantly different from the city × time coefficient in the pseudo-sample regression. This finding is statistically consistent with the expectations of the placebo experiment, indicating that the randomly constructed urban sample does not affect enterprise labor demand. Therefore, the positive and significant impact of the APPCAP on enterprise labor demand is not caused by unobservable factors.

• 2.Time placebo test. The results of the time placebo test are shown in Table 3. The results show that the estimated coefficients of city × time－^false3, city × time－^false4, and city × time－^false5 are not significant. This finding indicates that when the implementation of the APPCAP is pushed forward, there is no systematic difference in the time trend of labor demand for enterprises in high-target and low-target cities, suggesting the effectiveness of the APPCAP in promoting enterprise labor demand.

Fig. 2.

Placebo test result.

Table 3.

Time placebo test.

• 3.
  Sample data screening. Columns (1) to (2) of Table 4 show the winsorized results, indicating that the coefficient estimation of city × time is still significantly positive and does not differ significantly from the baseline estimation.
• 4.
  PSM-DID. Column (3) of Table 4 show the PSM-DID results. The estimated value of PSM-DID for city × time is 0.0640, which passes the significance test at the 1 % level and has little difference from the results of the baseline regression. This finding shows that the estimation results of this paper are robust.
• 5.
  Eliminating the impact of other environmental policy. The results of eliminating the impacts of other environmental policies are shown in Columns (4) to (5) of Table 4. The estimated results after excluding the interference of these two policies are consistent with the benchmark regression results.

	(1)	(2)	(3)
	labor	labor	labor
city × time -false3	0.0248
	(0.0220)
city × time -false4		0.0422
		(0.0381)
city × time -false5			−0.1041
			(0.158)
Control Variables
Constants	−4.3676***	−2.8911*	−4.1896***
	(0.6575)	(1.5995)	(0.6423)
Firm FE	yes	yes	yes
Year FE	yes	yes	yes
Observations	28,942	28,942	28,942
R2	0.929	0.929	0.929

5. Heterogeneity analysis

On average, implementing the APPCAP can effectively promote the expansion of enterprise labor demand. However, due to the heterogeneity of enterprises themselves, the APPCAP has a differentiated influence on labor demand according to different enterprise characteristics. The discussion of the heterogeneity of enterprise characteristics helps clarify the internal mechanism through which the APPCAP affects enterprise labor demand from other perspectives.

5.1. Heterogeneity of enterprise ownership

Enterprise ownership is divided based on the nature of the ultimate controller of listed company. If the entity that has the highest level of control is owned by the government, then the enterprise is categorized as a state-owned enterprise. Otherwise, it is a non-state-owned enterprise. After dividing the research sample into state-owned and non-state-owned enterprises, regression analysis was conducted using Model (6). The results of the heterogeneity of labor demand in state-owned and nonstate-owned enterprises under the APPCAP are shown in Columns (1) to (2) of Table 6. The results indicate that APPCAP significantly promotes the expansion of the labor demand scale in state-owned enterprises. However, it does not significantly impact the labor demand scale of nonstate-owned enterprises. This indicates differences in the implementation effects of the APPCAP for enterprises with different ownership systems. One potential reason may be that China's state-owned enterprises have close ties with the government, which provides them with better policy support [36]. When facing environmental regulations, state-owned enterprises can obtain more resources, support, and internal information from the government, which can effectively manage the negative impact of rising environmental regulation costs. Meanwhile, state-owned enterprises actively assume the social responsibility of providing employment opportunities and stabilizing the job market to enhance their reputations or help them strive for a more favorable external business environment. Therefore, when state-owned enterprises face environmental regulatory pressure, they expand the scale of labor demand. However, nonstate-owned enterprises do not have the advantages of policies and resources, making their impact on labor demand insignificant.

Table 5.

IV estimation.

	(1)	(2)	(3)
	labor	city × time	labor
	Benchmark Result	First stage	Second stage
city × time	0.0638***		0.1293***
	(0.0239)		(0.0313)
VC		−0.1522***
		(0.0463)
Control variables	yes	yes	yes
Firm FE	yes	yes	yes
Year FE	yes	yes	yes
Observations	28,942	28,942	28,942
Kleibergen-Paap rk LM statistic		10.723***
Cragg-Donald Wald F statistic		915.622
Kleibergen-Paap rk Wald F statistic		10.799
Stock-Yogo Critical value：15 % maximal IV size		8.96
Hansen J statistic		0.000

Table 6.

Heterogeneity of the impact of APPCAP on enterprise labor demand.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	state-owned	non-state-owned	clean industry	pollution industry	new enterprise	old enterprise	small-medium scale	large scale
city×time	0.0882**	0.0317	0.0214	0.0497*	0.0882**	0.0264	0.0275	0.1985**
	−0.0359	−0.0309	−0.0452	−0.0284	−0.0364	−0.0284	−0.0233	−0.0924
Control variables	yes	yes	yes	yes	yes	yes	yes	yes
Constants	−3.5449***	−3.9902***	−6.0733***	−5.1034**	−2.9371	−4.0229***	−4.2285***	11.9756
	−1.3031	−1.4795	−1.1555	−2.143	−1.7901	−1.4417	−1.158	−14.6284
Firm FE	yes	yes	yes	yes	yes	yes	yes	yes
Year FE	yes	yes	yes	yes	yes	yes	yes	yes
Observations	10,520	18,327	12.48	7229	14,372	14,570	28,552	384
R2	0.9359	0.9218	0.9417	0.9467	0.9169	0.9446	0.9223	0.9745

5.2. Heterogeneity of cleaning and pollution industry

Due to the different regulatory efforts of environmental regulations on cleaning and polluting industries, environmental regulations may have a differentiated impact on the labor demand of enterprises in these two industries. According to the research of Tong et al. [37], we use the industry codes to divide the research sample into clean and polluting industries and regression Model (6) to test the impact of the APPCAP on enterprise labor demand in the clean and polluting industries. The regression results in Columns (3) to (4) of Table 6 indicate that APPCAP significantly increased the labor demand of enterprises in the polluting industry but did not significantly affect the labor demand of enterprises in the cleaning industry. The reason may be that enterprises in polluting industries face greater environmental regulatory pressure and are subject to stricter environmental regulations and penalties. Therefore, compared to those in clean industries, enterprises in polluting industries have stronger urgency in controlling pollution, which has a more significant impact on their scale of the labor demand.

5.3. Heterogeneity of enterprise establishment years

Generally, the longer a company is established, the more capable it is to manage various risks and impacts. However, the longer a company is established, the more likely it is that the structure of its internal departments is too complex, which may result in insufficient flexibility in decision making. We continue to examine the possible heterogeneous impact of the APPCAP on the establishment of enterprise labor demand over time. We divide the sample based on the median of the company's establishment years and consider those above the median number of the company's establishment years as old enterprises, and those below the median of the company's establishment years as new enterprises. The new and old enterprises are used for separate regressions using Model (6). The heterogeneity analysis results of the company's establishment years are shown in Columns (5) and (6) of Table 6. The results show that APPCAP significantly and positively affects the labor demand of new enterprises but has no significant effect on the labor demand of old enterprises.

The reason may be that new enterprises adjust their production methods promptly to manage the additional costs of environmental regulations due to their more flexible production decisions. On the other hand, old enterprises already have relatively mature internal labor markets. In contrast, the development and expansion of new enterprises require hiring more employees for production, which significantly impacts the scale of labor demand.

5.4. Heterogeneity of enterprise scale

Enterprises of different sizes have different resources and abilities to respond to environmental regulation, which may lead to differences effect of environmental regulations on enterprise labor demand. We divide the research sample into large-scale and small-scale and medium-scale enterprises according to the third percentile standard of enterprise size and use each subsample to regress Model (6) separately. The heterogeneity results of enterprise size are shown in Columns (7) to (8) of Table 6. The results show that the APPCAP significantly and positively impacts the labor demand of large-scale enterprises but does not substantially affect that of small-to medium-scale enterprises. The reason may be that compared to small-to medium-scale enterprises, larger enterprises have scale advantages in terms of capital, technology, and equipment. Therefore, large-scale enterprises are prone to completing their green transformation through technological innovation and by creating more job opportunities. Therefore, the impact of large enterprises on labor demand is more significant and positive. Compared to small-to medium-scale enterprises, large-scale enterprises have more frequent production activities, resulting in more pollution emissions, and are more susceptible to the attention of environmental protection departments. The impact of pollution control pressure on the scale of labor demand is more significant.

6. Mechanism analysis

We verify that the APPCAP significantly impacts the enterprise labor demand through benchmark regression and a variety of robustness tests. So, what mechanism will APPCAP affect the labor demand of enterprises? The theoretical mechanism analysis above shows that the APPCAP may affect the enterprise labor demand through the output effect and factor substitution effect. Therefore, we use the output effect and factor substitution effect as the dependent variables and the ventilation coefficient mentioned above as the instrumental variable for the regression to identify the causal relationship between the independent variable city × time and the mechanism variable as accurately as possible. We also test the possible mechanisms through which the APPCAP affects labor demand.

The mechanism test results of the output effect are shown in column (1) of Table 7. This study refers to Liu et al. [2] and uses enterprise revenue (income) to represent the output effect. The result in Column (1) shows that the coefficient of city × time is significantly positive, indicating that the implementation of APPCAP has significantly improved enterprise output, and an increase in enterprise output also means that enterprises adopt scale expansion decisions, driving them to absorb more labor. At this point, enterprises need to absorb more labor for pollution control, thereby further increasing their labor demand. In summary, APPCAP increases the demand for labor through the mechanism of output effect, verifying Hypothesis 2.

Table 7.

Causal mechanism effect of APPCAP on enterprises labor demand.

	(1)	(3)
	income	enir_invest
city × time	0.0025**	0.0095**
	(0.4581)	(0.0042)
Control variables	yes	yes
Constants	−0.0340	0.6453***
	(0.0330)	(0.1818)
Firm FE	yes	yes
Year FE	yes	yes
Observations	28,942	28,942
R2	0.8936	0.8282
First stage F-statistic	10.799	10.799

Column (2) of Table 7 is the mechanism test of the factor substitution effect. When enterprises face environmental regulatory pressure, due to the increase in resource prices, they face the substitution of capital factors and labor production factors. At this time, enterprises judge and weigh investments in pollution control and the input of labor production factors. Under the heavy pressure of environmental regulation, enterprises with substandard production and emission standards are punished and may even be forced to shut down. This is why enterprises carry out green improvements in their production processes or introduce pollution treatment equipment to meet pollution emission standards, which are all manifestations of their investments in environmental governance. Therefore, this section selects environmental governance investments as the proxy variable of the factor substitution effect and test whether the APPCAP changes the scale of labor demand by influencing enterprise environmental governance investments. Due to the lack of statistics on the total amount of environmental governance investments used by enterprises in the data on listed companies, according to the approach of Li et al. [30], we use the ratio of fixed assets, intangible assets, and other long-term assets purchased and constructed by enterprises to total assets to measure environmental governance investments of enterprises (enir_investment). The results in Column (2) of Table 7 indicate that the coefficient of city × time is significantly positive, indicating that the implementation of the APPCAP has significantly increased enterprise investments. To gain a competitive advantage in the market, enterprises keenly capture the demand signals for green development in the market and increase the demand for labor by emphasizing clean production. Therefore, the APPCAP increases labor demand through the factor substitution effect, which verifies Hypothesis 2.

7. Conclusion and discussion

7.1. Conclusion

Through the application of the micro-data related to A-share listed enterprises between 2008 and 2020, the implementation of the APPCAP is considered as a quasi-natural experiment in this research. By adopting the DID approach to assess how environmental regulation affects enterprises’ labor demands, this paper makes the following conclusions. First of all, the APPCAP can positively affect and conspicuously increase the labor demands of enterprises. This conclusion still holds true for a sequence of tests on robustness. Secondly, as suggested by the mechanism test, the APPCAP could facilitate the labor demands of enterprises by virtue of the effects of factor substitution and output. Additionally, the influence of APPCAP on business labor demand is heterogeneous due to differences in enterprise characteristics. The implementation of the APPCAP could beneficially and substantially influence the labor demand of state-owned companies, new enterprises, large-scale enterprises, and firms in polluting sectors. Finally, the APPCAP significantly increases the demand for highly qualified workers, but it does not have a major influence on the demand for low-skilled workers or on firm salaries level.

In view of aforementioned facts, this paper brings forth the following policy implications:

Above all, Chinese government should necessarily maintain unwavering commitment to executing the policies of environmental regulation. In some sense, proper environmental regulation can enhance enterprises’ labor demands by the effects of output and factor substitution, which thereby boosts employment in an effective manner. For one thing, it remains essential for all levels of Chinese governments to invest more time and efforts in environmental protection. Besides, it is an inevitable task for governments to prevent contamination and ameliorate the environment by virtue of environmental regulations. For another, the benefits of environmental regulation are not restricted to environmental quality improvement. Through a variety of mechanisms, it enhances the demand for labor in firms and serves as an efficient strategy to accomplish pollution control and provide stable employment in China. However, it should also be pointed out that environmental regulation is a kind of government regulation, and the government should always maintain a cautious attitude in the process of implementing regulations. Adjusting regulatory policies too quickly or too high may be counterproductive. Therefore, the government should formulate more detailed and rigorous environmental policies, and actively improve the ecological environment through moderately strict environmental governance policies to ensure that there are laws to follow. The government must not only implement relevant policies and regulations, but also ensure strict law enforcement, guiding the accelerated transformation of national economic development toward a more green and sustainable direction.

Second, it is necessary to clarify the positioning of environmental policies. Since the features of the industry for an enterprise and the nature of the enterprise will influence the environmental regulations' effects on the enterprises’ labor demands, the nature of relevant enterprises and various industries to which the enterprises fall into should be attached with more consideration while formulating policy tools for environmental regulation, so as to reduce the degree to which environmental policies negatively affect the labor markets while fully taking advantage of the positive influence brought by the environmental regulations on the labor demands.

Finally, by implementing the APPCAP, the overall demand of enterprises for high-skilled labor have increased ultimately; nevertheless, the demands for labor force with low skill are not influenced significantly.

This suggests that environmental regulations have had an employment crowding out effect on low-skilled workers. This emphasizes the significance of human capital in fostering sustainable green economic development. While advocating for economic transformation and upgrading, government departments should also focus on the transformation and upgrading of the labor skills structure. Targeted support and relief should be provided for low-skilled labor affected by environmental regulations, and reemployment training should be provided for unemployed and low-skilled populations. Simultaneously, it is necessary to improve institutional guarantee measures, solve the unemployment problem resulting from the environmental regulations, and optimize the employment environment.

7.2. Discussion and future

We investigate the impact of the APPCAP on the demand for labor in enterprises. The research primarily focuses on assessing the influence of APPCAP on environmental and economic aspects, and insufficient emphasis has been placed on researching the impact of environmental regulations on social activity, specifically, the demand for labor. Therefore, we use microdata obtained from 3,949 Chinese A-share listed organizations, covering the time frame of 2008–2020. Assessing the influence of APPCAP on labor demand inside enterprises by using the DID model. We find that APPCAP significantly increases enterprise labor demand. Li [24] found that APPCAP can augment the workforce need in the cleaning industry, which to some extent supports the results of this study.

Although we delve into the detailed impact of the APPCAP on labor demand, there are also some limitations. In subsequent studies, we might broaden our inquiry to include two facets. First, China's environmental regulatory policy is now in a nascent stage and lacks a comprehensive and standardized market. Furthermore, the absence of data consistency and completeness has resulted in insufficient study at the prefectural-level city level. In the future, we may expand our focus to include major urban centers inside our country, in order to make a meaningful contribution towards the development of meticulously designed environmental regulations. Second, due to limitations in the research data, we only consider the influence of the APPCAP on the employment requirements of Chinese A-share listed enterprises. Therefore, subsequent research can use survey data from other types of enterprises to identify more accurately the policy effects of the APPCAP, ensuring the objectivity and accuracy of the research conclusions.

Data availability statement

Data will be made available on request.

Funding

Beijing Wuzi University Youth Research Fund Program (2024XJQN03)

Additional information

No additional information is available for this paper.

CRediT authorship contribution statement

Fen Qin: Writing – original draft, Data curation. Zhe Liu: Writing – review & editing, Methodology.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.