Breaking the Cost Barrier: How Environmental Policy Intensity and Cost Stickiness Shape Green Innovation in China’s Manufacturing Sector

Abstract

Facing a society that has accumulated many ecological and environmental problems, green technology innovation has gradually become an effective way to test the efficiency of environmental policies. This study investigated the impact of environmental policy intensity on green technological innovation within manufacturing enterprises listed in Shanghai and Shenzhen A-shares from 2010 to 2021. This research delved into the micro-level effects of such policies by utilizing data from various levels, including provinces, regions, enterprises, and years, along with a tailored environmental protection dictionary derived from government reports. It introduced the notion of cost stickiness to assess the relationship between regulatory stringency and innovation in the manufacturing sector. This study identified a specific regulatory intensity range within which environmental policies encourage green innovation, highlighting the importance of balanced regulation. This nuanced analysis contributes to a deeper understanding of the interplay between environmental regulation and technological innovation in manufacturing.

1. Introduction

Since the 1990s, China’s remarkable economic expansion has been significantly fueled by the manufacturing industry, notable for its high energy use and emissions [1,2,3]. This growth trajectory, while impressive, has concurrently led to a substantial accrual of ecological and environmental challenges [4,5,6]. China’s 160th place out of 180 in the 2022 GEP Index shows an urgent need for environmental management. The green innovation’s critical role is undeniable. As an instrumental means to counterbalance the ecological impact of industrial growth, green innovation emerges as a key solution for sustainable development [7,8]. Here, the effectiveness of environmental policies becomes evident, shaping the trajectory of ecological conservation and economic sustainability.

However, the impact of these policies on green innovation is uncertain. While intended to foster sustainable practices, the actual impact of such policies on corporate behavior, particularly in terms of innovation, is a subject of extensive debate. Research indicates that while some policies effectively stimulate green innovation, others may inadvertently hinder it [9], especially in sectors characterized by high costs and rigid production methods [10]. Delving deeper into the microenterprise level, the interplay between environmental policy and corporate innovation behavior becomes even more pronounced [11]. For these smaller-scale enterprises, which are highly sensitive to operational efficiency and cost control [12,13], environmental policy shocks are often first manifested in the rigidity characteristic of the cost structure [14]. Therefore, this paper specifically examines how macro-environmental policies influence green innovation behaviors in manufacturing enterprises (ME) by influencing their cost lag phenomenon (i.e., cost stickiness).

Cost stickiness reflects cost asymmetries, which implies that firms may overinvest in factors of production when business volumes increase, without a corresponding reduction in such investment when business volumes decrease. Analyzing cost stickiness reveals how companies allocate resources internally. It opens a window into the ‘black box’ of enterprise operations, revealing how environmental policies at the macro level can indirectly shape corporate strategies at the micro level, particularly in the context of green innovation. This study, therefore, seeks to answer pivotal questions: Do environmental policies increase cost stickiness in firms, consequently affecting their inclination toward green innovation? By exploring this question, this study aimed to provide empirical insights that could guide the formulation of more effective environmental policies. Such an exploration is crucial for understanding the full range of implications these policies have on the regional and national scales.

The potential contributions of this paper are as follows. First, it methodologically advanced the field by constructing a multi-dimensional dataset (provincial–regional–enterprise) and developing an environmental policy dictionary through web crawling of prefecture-level government work reports. Second, it theoretically extended the existing research by introducing the concept of cost stickiness, revealing how environmental regulations inhibit green innovation through increased cost stickiness. Third, it identified a threshold effect of regulatory intensity, demonstrating that moderate policy intervals best promote green innovation in manufacturing, providing empirical evidence for precise environmental policy making. These contributions enhance the understanding of the complex relationship between environmental regulation and green innovation in manufacturing.

The subsequent sections of this paper are organized as follows: Section 2 presents theoretical analysis and research hypotheses. Section 3 details the construction of models, data sourcing, and descriptive statistics. In Section 4, we analyze empirical results and their robustness. Section 5 introduces mechanisms, heterogeneity, and threshold effects, while Section 6 contains the study’s discussion and implications.

2. Literature Review and Research Hypothesis

2.1. Environmental Policy and Innovation

The realm of environmental policy, characterized by the imposition of direct regulatory measures such as emissions standards and mandated technological implementations, is intricately entwined with the phenomenon of eco-innovation. A spectrum of perspectives within the academic literature marks this relationship [15]. A faction within this scholarly discourse posits that the advent of stringent regulatory frameworks catalyzes corporate innovation, compelling firms to embark on innovative ventures to achieve compliance, thereby nurturing eco-innovation growth. This viewpoint finds its theoretical underpinning in the “Porter Hypothesis” [16], which postulates that rigorous environmental regulations can act as a stimulant for enhanced competitiveness and innovation [17,18,19].

In contrast, an alternate scholarly contingent articulates a dissenting viewpoint [20,21,22]. This perspective emphasizes the potential of overly stringent regulations to suffocate creative impulses and impose financial burdens that could impede the innovation trajectory. This body of work accentuates the peril of innovation being narrowly channeled toward mere regulatory compliance, at the expense of broader environmental advancements. Furthermore, an emerging scholarly narrative, encapsulated in the hypothetical studies by Hou et al. (2023) [23] and Y. Zhang et al. (2023) [24], delves into the regional intricacies within the command-and-control regulatory framework, unveiling the temporal variability of impact effects [25]. It is worth noting that excessive regulatory pressure may lead to diminishing returns or even inhibit innovation, especially for manufacturing firms with limited resources [26]. This threshold effect emphasizes the nonlinear nature of policy impacts and the importance of balancing regulatory stringency with firms’ ability to adapt [27,28]. Despite these different insights, inconsistent findings are prevalent in the literature, with some investigations revealing positive impacts on innovation, while others describe negative or neutral outcomes. This dichotomy highlights the need to delve deeper into the impact of macro-level policies on firms’ propensity to innovate at the micro level, with particular emphasis on different industrial contexts and firm capabilities. The emerging consensus highlights the need for a nuanced understanding of the interplay between the regulatory environment and firms’ eco-innovation in order to formulate effective policies. Based on this, this paper proposes the following hypotheses:

H1: 

Coercive environmental policies can inhibit green innovation in manufacturing firms, and there is a threshold effect.

2.2. Environmental Policy, Cost Stickiness and Green Innovation

Pertaining to the manufacturing sector, environmental policies incrementally escalate operational expenditures by instituting novel environmental stipulations and standards [26]. MEs may find themselves compelled to procure costlier equipment or adopt more financially demanding production methodologies to align with environmental mandates [29,30]. Such financial augmentations can increase a business’s fixed expenses, thereby constraining short-term adaptability [31]. Moreover, the advent of environmental mandates is likely to subject enterprises to heightened supervision and scrutiny, thereby amplifying managerial costs [32]. These multifaceted factors coalesce to amplify the cost stickiness within enterprises, subsequently hampering traditional cost-mitigation avenues such as scalability or production efficacy enhancements [33]. This predicament propels MEs toward the exploration of alternative strategies to alleviate the fiscal pressures engendered by environmental policies [34]. The escalation in investments toward specific environmental technologies engenders a progressively sticky cost structure, circumscribing firms’ abilities to curtail expenses, particularly in scenarios characterized by diminishing market demand or swift technological evolution [35]. This scenario potentially ensnares MEs within inefficient or economically unviable production methodologies. The inherent rigidity associated with cost stickiness further impedes firms’ agility in embracing green innovation. The substantial capital allocation toward specific assets may curtail the availability of resources for novel green technological investments or enhancements [36,37]. Moreover, elevated cost stickiness may precipitate lethargic corporate responses to market dynamics or technological advancements.

This is attributed to the fact that the production modalities of manufacturing enterprises undergo significant transformations in response to fluctuations in the intensity of environmental regulations. On the one hand, manufacturing entities are necessitated to integrate new machinery and technologies to revamp the production process in adherence to mandatory environmental mandates. On the other hand, these MEs may augment investments in certain fixed assets, which are inherently challenging to repurpose for alternate uses [30]. Consequently, under the yoke of mandatory regulations, enterprises markedly enhance the specialization of certain asset types to conform to green prerequisites. This strategy is not conducive to swift enterprise adaptability in the face of market dynamics, thereby exacerbating cost stickiness. As the sunk costs associated with independent research and development amplify the cost stickiness of firms, these entities are increasingly incentivized to pursue alternative avenues to fulfill environmental policy requisites [38]. Increased cost stickiness makes firms more reluctant to embrace green innovation, as firms need to balance short-term compliance with long-term growth [39]. Furthermore, the surge in cost stickiness also galvanizes firms toward strategic realignments to accommodate mandatory environmental policy mandates [40]. Enterprises may opt for collaborative ventures with peer entities to share the financial burdens associated with environmental protection apparatuses or research and development outcomes, thereby distributing the costs across multiple stakeholders [41]. This collaborative framework can materialize through joint research and development initiatives, technology exchanges, or collective procurement endeavors [42]. This, however, could potentially precipitate innovation stagnation under the guise of the free-riding phenomenon. According to the aforementioned information, the initial hypothesis is as follows:

H2: 

Environmental policies inhibit the green innovation, which is caused by mandatory environmental policies increasing the cost stickiness of manufacturing firms.

3. Materials and Methods

3.1. Data and Their Sources

In terms of data, according to the theme, the initial research sample of this paper selected explicitly the data of manufacturing companies listed on the Shanghai and Shenzhen stock exchanges from 2010 to 2023. The selection of 2010 as the starting point for this study is attributed to the Chinese government’s heightened emphasis on environmental protection from that year onward, marked by the introduction of numerous policies aimed at environmental conservation and green development for manufacturing. This era is characterized by an intensive phase of policy making related to environmental and green innovation, providing a comprehensive policy backdrop for analyzing the interplay between environmental regulation and green innovation. The manufacturing companies were matched by their industry classification name and the code of the 2012 edition of the China Securities Regulatory Commission. The samples were screened according to the following principles: First, research samples with a risk warning (ST, ST *, and PT) and termination of listing during the listing period were excluded; second, the sample of enterprises with missing key variables was eliminated. Finally, the sample was determined to be 1993 listed manufacturing companies, and a total of 11,603 non-equilibrium panel numbers of company annual sample observations were clustered to 204 cities at the prefecture level and above. The original data were collected from the China Statistical Yearbook, China Energy Statistical Yearbook, National Bureau of Statistics, National Research Network, CSMAR, CNRDS, provincial and municipal statistical yearbooks, social development statistical bulletins, and municipal government work reports.

3.2. Measurement of Constructs

3.2.1. Dependent Variables

Manufacturing enterprises green technology innovation (MGI): Patents are effective indicators to measure technological innovation, and green patents can most directly reflect the output of green technological innovation activities. According to the Green List of International Patent Classification launched by the World Intellectual Property Organization (WIPO), green patents are divided into green invention patents and green utility model patents. According to the research on innovation structure by (W. Li & Zheng, 2016) [43], green invention patents are more innovative than green utility model patents, indicating substantial green technology innovation. Therefore, this paper took the number of green invention patent applications (MGI-inno) in the current year as an indicator to measure the green technology innovation of manufacturing enterprises. The total number of green patent applications (MGI-total) was used as a robustness test.

3.2.2. Independent Variables

The intensity of regional environmental policy (ER): According to the practice of [44], the emphasis on the environment in the work report of the prefecture-level city government can reflect the government’s environmental governance efforts in the current year, and the environmental protection will of the prefecture-level city government can more directly affect the manufacturing enterprises. Therefore, based on the work reports of municipal governments at different levels, this paper selected 27 environmental terms that can more fully reflect the government’s environmental protection from three aspects: “environmental protection goals”, “environmental protection work objects”, and “environmental protection measures” to construct a thesaurus, and used Python 3.13.2 for text analysis to obtain the frequency of environmental words. The intensity of regional environmental policy was measured by the percentage of word frequency of environmental vocabulary and word frequency of the work report of prefecture-level city government (ER-per). In addition to the above measurement methods, in the robustness test, this paper also refers to the practice of [45] to measure the intensity of regional environmental policies and establish the following formula:

$$\begin{gathered} {\mathrm{T}\mathrm{O}\mathrm{T}\_\mathrm{w}\mathrm{o}\mathrm{r}\mathrm{d}\mathrm{s}}_{\mathrm{i},\mathrm{t}}=\sum \left\{\mathrm{k}\mathrm{e}\mathrm{y}\mathrm{w}\mathrm{o}\mathrm{r}\mathrm{d}\mathrm{s}\right\} \\ {\mathrm{E}\mathrm{R}-\mathrm{a}\mathrm{s}\mathrm{i}\mathrm{n}}_{\mathrm{i},\mathrm{t}}=\mathrm{l}\mathrm{n}\left({\mathrm{T}\mathrm{O}\mathrm{T}\_\mathrm{w}\mathrm{o}\mathrm{r}\mathrm{d}\mathrm{s}}_{\mathrm{i},\mathrm{t}}+\sqrt{1+{\left({\mathrm{T}\mathrm{O}\mathrm{T}\_\mathrm{w}\mathrm{o}\mathrm{r}\mathrm{d}\mathrm{s}}_{\mathrm{i},\mathrm{t}}\right)}^2}\right) \end{gathered}$$

(1)

The above equation represents the sum of the frequency of environmental words disclosed in the government work report of the prefecture-level city i in year t. Considering that the word frequency of keywords may be minor or 0 in the annual report of the government work of a prefecture-level city whose time is earlier or data are missing, the inverse hyperbola sine transform was processed, which can quickly eliminate the speed and high-frequency tremor. It can improve the smoothness of the data column and has a good definition when the value is 0. Therefore, another index of regional environmental policy intensity (ER-asin) was obtained, and the higher the value of this index, the greater the intensity of regional environmental policy.

3.2.3. Control Variables

In order to overcome the impact of missing variables as much as possible, this paper refers to previous studies [45,46]. It included macro and micro-level variables that may affect manufacturing enterprises’ green technology innovation performance. The control variables at the firm level included ownership concentration (Lhr), fixed capital density (Fcd), economic growth (Eg), human capital intensity (Si), asset size (Assets), debt size (Debt), and asset–liability ratio (Lev). The control variables at the prefecture and city level included industrial structure (Industry) and economic development level (Rgdp).

Table 1. Variable definitions.

Name	Label	Measurement
Dependent variables	Manufacturing enterprises green technology innovation (MGI-inno)	Number of patent applications for green inventions in that year
Independent variables	Intensity of regional environmental policy (ER-per)	The percentage of word frequency of environmental vocabulary and word frequency of prefecture-level city government work report
Control variables	Enterprise ownership concentration (Lhr)	The proportion of the largest shareholder
	Fixed capital density (Fcd)	The ratio of the total assets of the enterprise at the end of the year to the operating income of the current year
	Economic growth (Eg)	The ratio of GDP of the current year to GDP of the previous year
	Human capital intensity (Si)	The ratio of the number of employees at the end of the year to the operating income of the current year
	Asset size (Assets)	Expressed as the logarithm of total assets
	Debt scale (Debt)	Expressed as the logarithm of total liabilities
	Asset-liability ratio (Lev)	Total liabilities/total assets × 100%
	Industrial structure (Industry)	Ratio of GDP added value of secondary industry to GDP
	Level of economic development (Rgdp)	Expressed as the logarithm of GDP per capita

provides a detailed description of the variables and how they are measured, and our regression model only includes observations available for all relevant variables.

3.2.4. Model Specification

In this study, the following bi-directional fixed effect model was adopted to study the impact of regional environmental policy intensity on manufacturing enterprises’ green technology innovation:

$${\mathrm{M}\mathrm{G}\mathrm{I}}_{\mathrm{j},\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{E}\mathrm{R}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\alpha }}_2{\mathrm{x}}_{\mathrm{i}\mathrm{j},\mathrm{t}}+{\mathsf{\upsilon }}_{\mathrm{j}}+{\mathsf{\omega }}_{\mathrm{t}}+{\mathsf{\nu }}_{\mathrm{j}}{+\mathsf{\epsilon }}_{\mathrm{i}\mathrm{j},\mathrm{t}}$$

(2)

where i represents the region where the manufacturing enterprise is located, j represents the manufacturing enterprise, and t represents the year. The dependent variable is MGI, which represents the level of green technology innovation in manufacturing industry. The independent variable is ER, which represents the intensity of environmental policy in the region where the manufacturing enterprise is located. The matrix vector x represents a set of control variables; ${\mathsf{\upsilon }}_{\mathrm{j}}$ refers to firm fixed effect, ${\mathsf{\omega }}_{\mathrm{t}}$ refers to year fixed effect, ${\mathsf{\nu }}_{\mathrm{i}}$ refers to region fixed effect; and ${\mathsf{\epsilon }}_{\mathrm{i}\mathrm{j},\mathrm{t}}$ is a random interference item. Meanwhile, in the following empirical analysis, the t statistic after heteroscedasticity robust standard error adjustment is used.

Table 2. Descriptive statistics.

Variable	N	Mean	p50	SD	Min	Max
MGI-inno	15,501	1.619	0	8.804	0	417
ER-per	15,476	0.332	0.320	0.332	−1.162	1.585
Lhr	15,501	33.69	31.63	14.53	0	100
Fcd	15,500	1.762	1.562	3.666	0	294.4
Eg	15,464	1.826	1.108	1.538	−1.900	9.500
Si	15,500	1.217	1.023	0.922	0	29.92
Assets	15,500	21.73	21.61	0.981	18.76	26.55
Debt	15,500	20.46	20.42	1.375	15.83	26.07
Lev	15,501	0.333	0.316	0.174	0	1.718
Industry	15,464	29,240	0.474	53,719	0.149	200,278
Rgdp	15,484	11.51	11.58	0.468	8.797	12.22

summarizes the descriptive statistics of variables. The contents of the report include sample size, mean value, variance, maximum value and minimum value, etc. Among them, the mean value of MGI-inno in manufacturing enterprises is 1.619, the maximum value is 417, and the minimum value is 0, indicating that the green technology innovation level of different manufacturing enterprises is uneven and greatly different. The regional environmental policy intensity (ER-per) is characterized by “large mean and small standard deviation”, and the maximum and minimum values are 1.585 and −1.162, respectively, indicating that there are certain differences in the intensity of environmental policy among different regions. Other control variables also have different degrees of difference, which is basically consistent with the existing studies.

4. Empirical Results and Analysis

4.1. Benchmark Model

Table 3. Fixed effects model regression results.

	M (1) MGI-Inno	M (2) MGI-Inno
ER-per	−1.3460 *** (−2.63)	−1.1673 ** (−2.22)
Lhr		−0.0296 (−0.91)
Fcd		−0.0365 ** (−2.24)
Eg		−0.2387 (−1.14)
Si		0.3313 * (1.83)
Assets		6.2892 *** (6.20)
Debt		−2.7068 *** (−4.05)
Lev		1.8874 * (1.71)
Industry		0.0003 (0.53)
Rgdp		−0.7576 (−1.24)
Constant	2.0677 *** (11.73)	−29.1316 *** (−4.87)
Observations	15,441	15,400
Individual fixed effect	Yes	Yes
Year fixed effect	Yes	Yes
Region fixed effect	Yes	Yes
Df_m	1.0000	10.0000
F	6.9384	8.2382
R2	0.4278	0.4321

Notes: (1) *, **, *** represent 10%, 5%, and 1% significance levels, respectively; (2) The figures in () indicate the t-values.

presents the results derived from linear regression analyses assessing the influence of regional environmental policy stringency on manufacturing firms’ green technological innovation capabilities. Model 1, which controls for firm-specific, annual, and regional variations, reveals a regression coefficient of −1.3460 for the intensity of regional environmental policies (ER-per), achieving statistical significance at the 1% level. The robustness of these findings is further affirmed in Model 2, where the inclusion of additional control variables does not alter the significance of the ER-per coefficient (t-value = −2.22). This indicates that an increase in the stringency of environmental policies within a region correlates negatively with its manufacturing enterprises’ green technology innovation levels. Such results lend empirical support to Hypothesis 2. It is crucial to recognize that the negative impact of regional environmental policy stringency on the green technological innovation capabilities of manufacturing firms does not imply an inherent contradiction between environmental regulation and technological innovation. Instead, this finding highlights the nuances and balances that must be considered in the design and implementation of environmental policies. For instance, mitigating the short-term cost pressures of environmental policies through direct financial support and incentives for green innovation can facilitate long-term technological advancement and sustainable development [8]. Additionally, the combined use of mandatory policies with voluntary incentives, such as tax breaks and subsidies, may enhance firms’ willingness and capacity to adopt green technologies [47]. Lastly, encouraging cross-industry and international knowledge sharing and technological collaboration has also proven to be an effective way to promote green innovation [48].

4.2. Endogenous Problems

The endogeneity of variables will make the OLS estimates inconsistent, and there may be a two-way causal relationship between environmental policy and manufacturing enterprises’ green technology innovation. The intensity of regional environmental policy will affect the level of manufacturing enterprises’ green technology innovation, and the average level of enterprises’ green technology innovation will, in turn, affect the adjustment of regional environmental policy intensity. Therefore, the instrumental variables used in this paper overcome the endogeneity problem, and the instrumental variables used in environmental policy are the cross term of provincial annual precipitation and current total energy consumption (IV1), while the independent variable is taken with a lag of one period (IV2).

Table 4. Instrumental variable method.

		M (3) Tool Variables 1	M (4) Tool Variables 2
		ER-per
First stage	IV1 IV2	1.89 × 10−10 *** (8.55)	0.1432 *** (8.83)
		MGI-inno
Second stage	ER-per	−6.7627 ** (−2.24)	−6.0049 *** (−3.21)
	Controls	Yes	Yes
	Year	Yes	Yes
	Id	Yes	Yes
	Kleibergen–Paap rk LM statistics	82.930 *** {0.0000}	168.848 *** {0.0000}
	Kleibergen–Paap Wald rk F statistics	73.068 [16.38]	74.934 [19.93]
	Observations	6949	6072

Notes: (1) **, *** represent 5% and 1% significance levels, respectively; (2) The figures in () indicate the t-values; the figures in [] are the Stock–Yogo critical value at the 10% significance level of the Cragg–Donald Wald F statistic; the figures in {} are the Anderson LM and Hansen J statistic that corresponds to the p-values.

reports the regression results of IV estimation. In the estimation results of the two-stage least square method, the F-value of the first stage is far greater than 10, and the p-value is 0.0000, indicating no weak recognition problem. In addition, according to the statistical results, the p-value of LR statistics of the equation unidentifiable test is 0.0000, which firmly rejects the null hypothesis of “unidentifiable”. In addition, the results of the K-PF test show that the instrumental variables meet the correlation characteristics, and there is no problem with weak instrumental variables. At the same time, the estimation results of the second-stage regression are still significant, indicating that the primary conclusion of this paper is robust and reliable.

4.3. Robustness Tests

4.3.1. Substitution Variable

The dependent variable was substituted into the sum of the number of green invention patent applications and the number of green utility model patent applications (MGI-total), and then regression was carried out; the measurement results for model M (5) are shown in

Table 5. Robustness test.

	M(5)	M(6)	M(7)	M(8)
	MGI-Total	MGI-Inno	MGI-Inno	MGI-Inno
ER-per	−1.8918 ** (−3.67)			−0.6855 *** (−3.38)
ER-asin		−1.2287 ** (−2.15)
L.ER-per			−1.6090 *** (−2.74)
Constant	−55.1810 *** (−6.11)	−27.8274 *** (−4.75)	−33.0067 *** (−4.66)	−19.3236 *** (−7.74)
Observations	15,400	15,441	13,339	15,400
Controls	Yes	Yes	Yes	Yes
Individual/Year/Region	Yes	Yes	Yes	Yes
Df_m	10.0000	10.0000	10.0000	10.0000
F	10.7988	8.2320	6.7150	10.7078
R2	0.5616	0.4316	0.4656	0.6153

Notes: (1) **, *** represent 5% and 1% significance levels, respectively; (2) The figures in () indicate the t-values.

. After replacing the independent variables with the frequency of environmental protection words and adding them together, the inverse hyperbolic sine processing (ER-asin) was carried out, and regression was performed. The results for model M (6) are shown in Table 5.

4.3.2. Lagging

In addition, the potential lag effect of environmental policy intensity (ER) on the level of green technology innovation (MGI) of manufacturing enterprises was also considered. Therefore, this paper adopted the intensity of environmental policy (L.ER-per) with a lag of one phase to conduct regression, and the results for model M (7) are shown in Table 5.

4.3.3. Truncation

In order to avoid the influence of outliers on the regression results, this paper carried out bilateral tail reduction treatment on continuous variables at the levels of 1% and 99% before conducting regression. The measurement results for model M (8) are shown in Table 5.

As shown in the empirical regression results in Table 5, the core conclusion that “the enhancement of environmental policy intensity in the region where manufacturing enterprises are located inhibits the level of green technology innovation of manufacturing enterprises” has not changed, which verifies the robustness of the benchmark regression model above.

5. Further Analysis

5.1. Mechanism Test

The previous research and analysis on the direct effects show that the strengthening of environmental policy intensity in the region where manufacturing enterprises are located will inhibit the level of green technology innovation of manufacturing enterprises. According to the logic of theoretical analysis, environmental policy inhibits the negative impact on green technology innovation by increasing cost stickiness. This paper further explored the mechanism effect.

$${∆\mathrm{l}\mathrm{n}\mathrm{C}}_{\mathrm{j},\mathrm{t}}={\mathsf{\beta }}_0+{\mathsf{\beta }}_1∆\mathrm{l}\mathrm{n}{\mathrm{I}}_{\mathrm{j},\mathrm{t}}+{\mathsf{\beta }}_2\mathrm{l}\mathrm{n}{\mathrm{I}}_{\mathrm{j},\mathrm{t}}\times \mathrm{M}\mathrm{D}+{\mathsf{\beta }}_3{∆\mathrm{l}\mathrm{n}{\mathrm{I}}_{\mathrm{j},\mathrm{t}}\times {\mathrm{E}\mathrm{R}}_{\mathrm{i},\mathrm{t}}\times \mathrm{M}\mathrm{D}+{\mathsf{\beta }}_4{\mathrm{x}}_{\mathrm{i}\mathrm{j},\mathrm{t}}+\mathsf{\upsilon }}_{\mathrm{j}}+{\mathsf{\omega }}_{\mathrm{t}}+{\mathsf{\nu }}_{\mathrm{j}}{+\mathsf{\epsilon }}_{\mathrm{i}\mathrm{j},\mathrm{t}}$$

(3)

where ∆lnC represents the ratio of the current year to the previous year’s operating cost and then takes the natural logarithm, lnI, which represents the natural logarithm of the current year’s operating revenue; ∆lnI represents the ratio of the current year to the previous year’s operating revenue and then takes the natural logarithm, MD, as the dummy variable, and if the current year’s operating revenue decreases compared to the previous year, the declining value is 1, otherwise the value is 0. Other variables were measured by model (1). If the regression coefficient β2 is significantly negative, this indicates that operating costs do not change rapidly according to the rise or fall of operating revenue; that is, there is cost stickiness. If the cross-multiplication coefficient β3 of regional environmental policy intensity (ER), the change in business revenue (∆lnI), and the dummy variable (MD) of revenue decline are all significantly negative, it is proven that the introduction of an environmental policy will increase cost stickiness.

Table 6. Mechanism test.

	M (1) ∆lnC	M (2) ∆lnC	M (3) ∆lnC1	M (4) ∆lnC1
∆lnI	0.9055 *** (61.25)	0.9000 *** (59.13)	0.3961 *** (61.25)	0.3937 *** (59.13)
lnI × MD	−0.0005 *** (−2.65)	−0.0006 *** (−2.86)	−0.0002 *** (−2.65)	−0.0002 *** (−2.86)
∆lnI × ER × MD	−0.1479 ** (−2.15)	−0.1254 * (−1.84)	−0.0647 ** (−2.15)	−0.0548 * (−1.84)
ER	−0.0016 (−0.19)	−0.0066 (−0.70)	−0.0030 (−0.75)	−0.0028 (−0.70)
Constant	0.0308 *** (7.74)	−0.4403 *** (−3.47)	0.0134 *** (7.74)	−0.1926 *** (−3.47)
Observations	13,626	13,592	13,627	13,592
Controls	NO	Yes	NO	Yes
Individual/Year/Region	Yes	Yes	Yes	Yes
Df_m	4.0000	13.0000	4.0000	13.0000
F	5405.6003	1801.8435	5410.0603	1801.8435
R2	0.8896	0.8912	0.8896	0.8912

Notes: (1) *, **, *** represent 10%, 5%, and 1% significance levels, respectively; (2) The figures in () indicate the t-values.

shows the regression results of environmental policy and cost stickiness. The high goodness of regression fit of each column indicates that the model is well set.

In this paper, a progressive regression strategy is adopted. Model M (1) only controls time and individual fixed effects. In model M (2), a set of control variables is added on the original basis. According to the coefficient of the cross term between the ratio of operating revenue and whether the revenue decreases, the results show that for every 1% increase in operating revenue, the operating cost increased by 0.9000%. In comparison, the operating cost decreased by 0.8994% for every 1% decrease in operating revenue; that is, when the revenue decreased, the cost decreased by less than the cost increased when the revenue increased, indicating the phenomenon of cost stickiness. The regression coefficient of the cross-multiplication term for the dummy variable of regional environmental policy intensity, business revenue change, and revenue decline is −0.2014 and significant at the 1% level, which implies the following economic implications: when the region where the manufacturing enterprises are located has strong environmental policies, the operating cost increases by 0.9000% for every 1% increase in operating revenue, while the operating cost decreases by 0.7515% for every 1% decrease in operating revenue, which indicates that environmental policies have a significant positive impact on the stickiness of enterprise costs. The phenomenon of cost stickiness, as indicated by the data, shows that operating costs increase more when revenues increase than when they decrease when revenues fall. This asymmetric cost behavior underscores the inherent inflexibility in the operational cost structure of firms, which is further influenced by environmental policies. The significant negative coefficient for the interaction term involving regional environmental policy intensity, business revenue change, and revenue decline suggests that such policies exacerbate cost stickiness. Specifically, in regions with stringent environmental regulations, the reduction in operating costs during revenue downturns is less pronounced, thereby increasing the stickiness of costs. At the same time, we used the sum of operating expenses and selling expenses instead of operating costs to improve the robustness of the results, which are shown for M (3) and M (4) in Table 6.

The enhanced cost stickiness in manufacturing enterprises due to green innovation can be attributed to several factors rooted in the real-world context. Environmental policies often mandate significant investments in compliance-related infrastructure, technologies, and practices, which represent fixed costs that do not vary directly with revenue fluctuations, thereby contributing to cost stickiness [49]. Additionally, as argued in the study by F. Xu et al. (2023) [49], the adoption of green technologies and practices, while beneficial for long-term sustainability and potentially offering operational efficiencies, may initially increase cost stickiness due to their fixed and sunk nature. This implies that firms under stringent environmental regulations may face higher operational inflexibility, impacting their ability to adapt to revenue downturns efficiently.

However, this phenomenon also reflects a potential strategic behavior where firms invest in environmentally friendly technologies and practices, which, while increasing cost stickiness, may also lead to long-term efficiencies, innovation, and sustainability benefits. This observed cost stickiness, exacerbated by environmental policies, suggests a complex interplay between regulatory compliance and cost management. These policies increase operational inflexibility due to higher fixed and compliance-related costs. The implications of this dynamic are multifaceted:

For policymakers, it emphasizes the need to design environmental regulations that balance the immediate compliance burden with incentives for long-term innovation and sustainability. The goal is to create policies that not only protect the environment but also encourage firms to innovate in ways that improve their competitive position and operational efficiency over time [50].

For firms, it highlights the strategic importance of leveraging environmental policies as a catalyst for adopting sustainable practices and technologies. While these policies may pose short-term challenges to operational flexibility due to increased cost stickiness, they also offer a strategic avenue for firms to invest in sustainability and innovation. By doing so, firms can potentially transform regulatory compliance from a cost burden to a competitive advantage [34,51].

In essence, while environmental policies may pose short-term challenges to operational flexibility due to increased cost stickiness, they also offer a strategic avenue for firms to invest in sustainability and innovation, potentially transforming regulatory compliance from a cost burden to a competitive advantage.

5.2. Analysis of Heterogeneity

5.2.1. Heterogeneity Test Based on Property Rights of Enterprises

To some extent, the property-suitable nature of enterprises will affect the apparent differences in resource allocation and other aspects of enterprises [52]. In state-owned enterprises, the management often sacrifices shareholders’ long-term interests for individuals’ short-term interests. The attribute of state ownership also means that state-owned enterprises need to assume the responsibility of “stabilizing finance” when the economy declines [53], which makes the operational decisions of state-owned enterprises often fail to find optimal solutions. Private enterprises are likelier to make appropriate cost-control decisions based on maximizing corporate benefits. Based on the above analysis, this paper concludes that, compared with state-owned enterprises, the negative impact of environmental policy on green technology innovation by increasing cost stickiness is more significant in non-state-owned enterprises.

Table 7. Heterogeneity test.

	(1) State-Owned Enterprise	(2) Non-State-Owned Enterprise	(3) Provincial Capital	(4) non-Provincial Capital
ER-per	−1.6614 * (−1.82)	−0.6448 *** (−3.17)	−1.6850 *** (−3.65)	−0.2564 (−1.19)
Constant	−23.4588 * (−1.72)	−20.9291 *** (−8.36)	−14.5093 *** (−3.11)	−21.2333 *** (−7.13)
Observations	1410	13,974	5451	9949
Controls	Yes	Yes	Yes	Yes
Individual/Year/Region	Yes	Yes	Yes	Yes
Df_m	10.0000	10.0000	10.0000	10.0000
F	2.5327	10.6866	4.1449	7.9209
R2	0.7102	0.6066	0.6213	0.6117

Notes: (1) *, *** represent 10% and 1% significance levels, respectively; (2) The figures in () indicate the t-values.

shows the results of the grouping regression according to the nature of enterprise property rights. As can be seen from column (1) (2) of Table 7, the coefficients of both state-owned and non-state-owned enterprises are significantly negative under the significance level of 10%, consistent with the benchmark regression results. This indicates that from a macro perspective, the environmental policy level of both state-owned enterprises and non-state-owned manufacturing industries hurts enterprises’ green technology innovation at the prefecture and city level.

5.2.2. Heterogeneity Analysis Based on Whether the Enterprise Is Located in the Provincial Capital

The different administrative levels of the cities in which enterprises are located will also affect the effect of environmental policy intensity on enterprises’ green technology innovation. Provincial capitals collect the most essential resources of their province and are usually the forerunners of national or regional development strategies. They are more sensitive to relevant government policies. Therefore, the listed manufacturing companies whose cities are provincial capitals have higher execution power in the face of mandated environmental policies. The results in columns (3) and (4) of Table 7 show that environmental policies of listed manufacturing companies in provincial capitals inhibit their green technology innovation level. The coefficient is significantly negative at 1% (t-value is −2.64), while the listed manufacturing companies in non-provincial capitals have no significant impact.

5.3. Further Analysis: Reasonable Intensity of Environmental Policies

In order to further study whether there is a nonlinear relationship between environmental policy and green technology innovation in the manufacturing industry, the threshold effect model is adopted for testing by referring to the research of Hansen (1999) [54]. Considering the situation that most of the microscopic data are unbalanced panels, the study of Q. Wang (2015) [55] is referenced to avoid a large number of missing samples. Missing values were filled in by linear interpolation, and variables that can only be integers in the economic sense are rounded. Taking the green technology innovation level of manufacturing enterprises (MGI-inno) and regional environmental policy intensity (ER-per) as threshold variables, different threshold indices were set to investigate the impact of environmental policy on green technology innovation in manufacturing. The model was constructed as follows:

$${\mathrm{M}\mathrm{G}\mathrm{I}}_{\mathrm{j},\mathrm{t}}={\mathrm{a}}_0+{\mathrm{a}}_1{\mathrm{X}}_{\mathrm{i},\mathrm{t}}\times \mathrm{I}({\mathrm{Y}}_{\mathrm{i},\mathrm{t}}\le {\mathsf{\theta }}_1)+{\mathrm{a}}_2{\mathrm{X}}_{\mathrm{i},\mathrm{t}}\times \mathrm{I}({\mathrm{Y}}_{\mathrm{i},\mathrm{t}}\le {\mathsf{\theta }}_2)+\cdots +{\mathrm{a}}_{\mathrm{m}}{\mathrm{X}}_{\mathrm{i},\mathrm{t}}\times \mathrm{I}({\mathrm{Y}}_{\mathrm{i},\mathrm{t}}>{\mathsf{\theta }}_{\mathrm{m}}){\mathsf{\upsilon }}_{\mathrm{j}}+{\mathrm{a}}_{\mathrm{m}+1}{\mathrm{x}}_{\mathrm{i}\mathrm{j},\mathrm{t}}+{\mathsf{\upsilon }}_{\mathrm{j}}+{\mathsf{\omega }}_{\mathrm{t}}+{\mathsf{\nu }}_{\mathrm{j}}+{\mathsf{\epsilon }}_{\mathrm{i}\mathrm{j},\mathrm{t}}$$

(4)

where ${\mathrm{a}}_0$ is the constant term, $\mathsf{\theta }$ is the threshold value, m is the threshold number, ${\mathrm{a}}_{\mathrm{m}}$ is the coefficient to be estimated, $\mathrm{I}$(.) is the indicator function, ${\mathrm{Y}}_{\mathrm{i},\mathrm{t}}$ is the threshold variable, and other variables are measured in accordance with model (1). The Bootstrap self-sampling method was used to sample 300 times, and the tests of three thresholds, double thresholds and single thresholds were carried out successively. As shown in the

Table 8. Threshold regression results.

Threshold Variable	Threshold Value	Threshold	ER-Per
ER-per	0.0954	ER-per ≤ 0.2649	0.4151 (0.48)
	0.7746	0.2649 < ER-per ≤ 0.2708	−0.3269 *** (−2.82)
		ER-per > 0.2708	−0.5299 *** (−6.02)
		Constant	−9.3700 (−3.18)
		Controls	Yes
		Individual/Year/Region	Yes
		Observations	15,476

Notes: (1) *** represent 1% significance levels, respectively; (2) The figures in () indicate the t-values.

, when the level of green technology innovation of manufacturing enterprises (MGI-inno) and the intensity of regional environmental policy (ER-per) are threshold variables, the impact of environmental policy on green technology innovation has a double threshold effect.

According to the estimation results of the model with regional environmental policy intensity (ER-per) as the critical variable, when ER-per is smaller than the first critical value, environmental policy does not hinder green technological innovation but when ER-per crosses the first critical value of 0.0954, environmental policy has a significant negative effect on green technological innovation. And when ER-per crosses the second critical value 0.7746, the negative impact of environmental policy on green technology innovation will be greater.

6. Discussion and Implications

6.1. Findings

In the era of frequent ecological problems, the adoption of green technology innovation by manufacturing enterprises, as the main body of the national economy, is an essential means to achieve a “win–win” situation between macro control and micro subjects. This study used the data of the manufacturing sector listed in Shanghai and Shenzhen A-shares from 2010 to 2021 to construct multiple data sets of provinces, regions, enterprises, and years, and used crawler technology to build an environmental protection related dictionary from the government work reports of prefecture-level cities. This study aimed to examine the effect of environmental policy intensity at the micro level from the perspective of enterprises that are more sensitive to policies. This study advanced the understanding of mandatory environmental policies by applying them at the firm level, thus enriching research in this area.

Previous studies have predominantly examined the relationship between environmental regulation and innovation from a singular cost perspective, often overlooking the complex micro-level dynamics. While the existing literature has extensively explored macro-level policy effects [20,21,22], limited attention has been paid to how cost structures mediate this relationship at the firm level. This study introduced the concept of cost stickiness to investigate understudied micro-aspects, particularly delving into the influence of environmental regulations in manufacturing regions on green technological innovation. The analysis revealed a negative correlation, suggesting that stringent policies may inadvertently hinder innovation by increasing cost stickiness. This study further explored the threshold of environmental regulatory intensity and found that once a specific regulatory intensity interval is exceeded, the envisaged environmental policy increases the degree of negative disincentives to green innovation in manufacturing firms. This nuanced approach offers a deeper insight into the complex dynamics between environmental regulation and technological innovation in the manufacturing sector.

6.2. Managerial Implications

In the context of the manufacturing sector, a pivotal element of the national economy, there is a complex link between green innovation needs and strict environmental policies, particularly within the milieu of China’s forceful regulatory landscape. These policies, designed to propel sustainable development, often impose considerable compliance burdens, especially on small and medium-sized enterprises (SMEs) that are constrained by limited financial, technological, and human resources. The inherent high fixed costs in the manufacturing industry exacerbate the phenomenon of cost stickiness, where fixed and variable costs do not adjust proportionally with fluctuations in production volume or revenue, particularly under stringent regulatory regimes. This cost rigidity ties up resources, making them unavailable for innovation, and makes businesses less nimble in adapting to market and regulatory shifts.

Globally, the challenge of balancing environmental regulation and innovation is not unique to China. This study found that moderate environmental regulation is most effective in driving green innovation. This balance is achieved when the need to comply with regulations is balanced against the resources available for innovation; otherwise, the costs of compliance may hinder innovation efforts. The multifaceted impact of these dynamics highlights the urgent need for a more nuanced approach to environmental regulation, and is also consistent with global efforts to achieve the United Nations Sustainable Development Goals (SDGs).

6.3. Future Research and Limitations

This paper has conducted a series of explorations on the impact of environmental policies on green technology innovation in the manufacturing industry, but there are still limitations. Further studies can be conducted on the following aspects in the future. First of all, although this paper discusses the space level, it does not profoundly study the transmission mechanism in the space network. Secondly, although the measurement of environmental policy in this paper is improved and relatively novel compared with predecessors, the dictionary construction has a particular subjectivity, and the word frequency in this paper is similar to the connotation of “command-type environmental policy”, so there is still a gap in the research on different types of environmental policy. Finally, although this paper focuses on enterprises in the manufacturing sector, the mechanism of high-tech manufacturing should differ from that of non-high-tech manufacturing if used as a research sample. In future research, these aspects can be studied in depth with a view to making a breakthrough in green innovation.