A Study of the Non-Linear Impact of Climate Policy Uncertainty on Enterprises’ Technological Innovation Based on China’s Industrial Enterprise Digital Peer Effect

Abstract

In the digital and low-carbon economy context, climate policy uncertainty’s (CPU) influence on corporate technological innovation has attracted increasing attention. However, prior studies mainly focused on the negative cost effects or singular incentive impacts of CPU on corporate innovation, with limited exploration of its non-linear effects, especially in the new scenario of the integrated development of digital and low-carbon economies. This study fills this gap by using a comprehensive dataset of China’s A-share listed companies from 2007 to 2023 and employing a fixed-effects model. It investigates how CPU impacts corporate technological innovation through the lens of digital peer effects within the same industry and region. The findings reveal a dual “inverted U-shaped” effect of CPU on corporate innovation capabilities: moderate CPU stimulates innovation via increased government subsidies, while excessive uncertainty exacerbates financing constraints, inhibiting innovation. The present study also identifies the significant moderating role of digital peer effects in mitigating the negative impacts of CPU, enhancing innovation compensation, and bolstering firms’ climate risk resilience. Notably, state-owned enterprises and manufacturing firms demonstrate superior innovation capabilities and risk resistance. This study provides new insights into understanding CPU’s impact on corporate innovation and offers valuable references for policy formulation and corporate strategy development.

1. Introduction

As the influence of global climate change intensifies, shifting towards a low-carbon economy has emerged as a shared objective among governments and businesses globally. Against this backdrop, the formulation and implementation of environmental policies have emerged as crucial tools for driving corporate technological innovation [1]. However, climate policy uncertainty (CPU) may exert complex impacts on industrial corporate technological innovation. CPU encompasses the uncertainty and fluctuation in the timing, extent, and strength of climate-related policies. This uncertainty can arise from changes in government regulations, shifts in policy priorities, or ambiguity in the implementation of environmental measures [2]. On one hand, CPU may negatively affect corporate technological innovation by increasing decision-making costs. On the other hand, it may also promote corporate innovation compensation effects through government policy incentive mechanisms [3]. While the manifestation of this dual effect may vary significantly across different countries and regions, its core mechanism possesses a certain degree of universality.

As a representative of a transitional economy, China is witnessing increasingly prominent resource and environmental constraints in the current era of a low-carbon economy. To address climate change and ensure the sustainable use of resources, the Chinese government has implemented a suite of measures aimed at fostering a low-carbon economy. These measures span across various domains such as industrial regulation, market mechanisms, financial incentives, and pricing strategies. These policies, while aiming to guide enterprises towards sustainable practices, also introduce elements of uncertainty that can influence corporate decision-making and innovation strategies. Enterprises are urgently required to meet their “dual carbon” targets, and numerous climate policies are bringing new risks and opportunities to enterprises. The economic growth in this new era is shifting towards a model that emphasizes innovation, coordination, environmental sustainability, openness, and inclusiveness. At the micro level, for high-pollution and highly technology-intensive industrial enterprises, resource and environmental regulation issues are of vital importance [4]. Chinese industrial enterprises are currently facing both digital transformation and the achievement of “dual carbon” goals. On one hand, digital technologies are clean and low-carbon, and digital transformation is a viable path for enterprises to achieve their “dual carbon” goals, with new technology research and development receiving substantial government subsidies [5]. On the other hand, under the requirements of the “dual carbon” goals and low-carbon climate policies, the economy and energy structure are undergoing in-depth adjustments [5], and companies face increasingly stringent environmental regulations which have a significant impact on corporate financing, operations, and other aspects.

At the same time, with the ongoing progress in science and technology, informatization, and globalization, technological innovation has emerged as a crucial driver for corporate sustainable profitability and enhanced competitiveness [6]. Digital technologies, including AI technology, big data, and so on, are gradually being perfected and utilized by enterprises. Digital transformation has become a vital pathway for enterprises worldwide to cope with environmental regulations and enhance competitiveness, and its role is becoming increasingly prominent. Digital technologies can not only improve corporate production efficiency but also help enterprises better adapt to the uncertainty of climate policies by optimizing resource allocation and reducing carbon emissions. In particular, on a global scale, the digital peer effect can provide new impetus for corporate technological innovation by reducing informational disparities and lowering decision-making costs. Although the manifestation of this digital-based collaborative effect may vary across different countries and industries, its role in promoting global corporate technological innovation is of significant international value.

In this study, we choose to focus on China. As a major developing country and one of the largest carbon emitters globally, China has implemented a range of proactive measures to tackle climate change and drive the transition to a low-carbon economy. The implementation of these policies has not only profoundly influenced China’s economic development but also offered valuable insights for the formulation and implementation of global climate policies. At the same time, China is currently in a phase of rapid digital economy growth. The extensive application of digital technologies across various industries presents both new opportunities and challenges for enterprises [7]. In addition, data from Chinese listed companies are highly available and representative, providing a reliable empirical basis for the study. By focusing on China, we can examine the impact of CPU on corporate technological innovation more thoroughly and offer targeted suggestions for policymakers and businesses. This study examines the relationship between climate policy and corporate digitalization from the perspective of the digital peer effect. It is significant for improving corporate operational efficiency, reducing corporate decision-making costs, and thereby promoting industrial technological progress. It also provides a high-quality development path towards intelligent and digital new-quality productive forces. This study seeks to investigate how CPU affects technological innovation within the current landscape of digital and low-carbon economies. Previous studies have often focused on the environmental protection perspective [8,9] or solely on digital transformation [10,11]. This study will delve deeper into the relationship between CPU and industrial technological innovation by examining the digital peer effect and uncovering the underlying connections between them. By deeply exploring the peer effect of technological innovation in industrial enterprises, it helps to reveal the internal laws of technological innovation activities. At the same time, this study, which explores the issue from the perspective of climate policy, provides a technological innovation perspective for interpreting China’s “dual carbon policy” and also supplements and improves the research on technological innovation in industrial enterprises. Technological innovation in enterprises will lead the evolution of high-quality productive forces in the current environment. Future high-quality development in the industrial sector will be oriented towards the application of new-quality productive forces [12], with digitalization as the main path. The conclusions of this study provide strong support for promoting corporate technological innovation and sustainable development and provide strategic recommendations for the implementation and high-quality development of technological innovation in various industries in China.

2. Theory, Literature Review, and Hypothesis Development

2.1. CPU and Technological Innovation in Enterprises

From the perspective of a low-carbon economy, the academic community has been engaged in continuous and extensive discussions on the relationship between environmental policies and corporate technological innovation. In China, under the “dual carbon” goals, climate policy is an important environmental policy. Climate policy is an important tool for promoting corporate innovation. However, the risks triggered by its uncertainty may also become an obstacle to corporate development. On one hand, Chen’s study suggests that CPU can change the risk preferences of corporate management, and such risk preferences have an incentive effect on corporate innovation inputs and outputs [13]. CPU brings more risky decision-making costs. Both Dai and Wang’s research have found that CPU will exacerbate corporate financing constraints [14,15], leading to a reduction in the investment obtained by enterprises, which is not conducive to the financing needed for new corporate technology research and development. Liu [16] believes that a good financial situation is the basis for high technological innovation capabilities. For industrial enterprises, technological innovation is fundamentally a high-risk, high-reward long-term investment. Increased financing constraints can negatively affect the financial health of enterprises [17]. Moreover, from the input–output theory perspective, there exists a mutually dependent relationship between production and consumption across various sectors of the national economy, as well as within sectors, enterprises, and international development organizations [18]. Financing for corporate technological innovation will promote the continuous increase in corporate innovation inputs. Bouchmel’s [19] research on multinational enterprise data found that intensified external financing constraints will reduce corporate innovation inputs. Therefore, from the perspective of financing constraints, within a certain range, an increase in CPU will reduce corporate technological innovation capabilities. Based on this, this study proposes the transmission mechanism of “Climate Policy Uncertainty–Financing Constraints–Corporate Innovation Capabilities”.

On the other hand, when CPU reaches a certain high level, it will prompt local governments to pay more attention to supporting corporate new technology research and development [20]. At the same time, enterprises are more inclined to actively cater to subsidy policies and improve existing production technologies and product designs. From the perspective of external policy environment, Du’s research [15] found that CPU will affect the intensity of local government environmental policies. Local governments’ incentive policies for corporate technological innovation, such as R&D subsidies, will be adjusted based on CPU. Lai’s research [21] directly pointed out that the introduction of low-carbon policies has increased local government subsidies for corporate technology research and development. Faced with uncertain climate policies, local governments’ emphasis on corporate subsidies will to some extent help improve corporate technological innovation capabilities. Based on this, this study proposes that when CPU reaches a certain level, a “Climate Policy Uncertainty–Government Subsidies–Corporate Innovation Capabilities” transmission mechanism will emerge.

In summary, this study puts forward the following hypotheses:

Hypothesis H1a:

CPU will affect technological innovation capabilities in enterprises through financing constraints.

Hypothesis H1b:

CPU will affect technological innovation capabilities in enterprises through government subsidies.

On one hand, CPU generates cost effects that negatively impact corporate innovation. Apart from the perspective of financing constraints, from the corporate perspective, Ren’s research has demonstrated that CPU can lead to a decline in total factor productivity within firms [22], a phenomenon particularly pronounced in labor-intensive industrial enterprises. Syed’s extensive study on U.S. CPU [23] has shown that it can cause a sharp drop in the revenue of energy companies in the near future. However, according to Porter’s hypothesis and the theory of innovation compensation, appropriate environmental regulations can prompt firms to engage in more innovative activities, which in turn enhance productivity, thereby creating an innovation compensation effect. Regarding green innovation, Xu’s research has identified an “inverted U-shaped” relationship between environmental policies and eco-friendly technological advancements in the industrial sector [24]. This implies that there exists a critical point of optimal environmental regulation intensity. Before reaching this critical point, environmental regulations drive green technological innovation, and climate policies have a similar effect. Bouri’s research [25] has also confirmed that CPU can have positive effects on the energy market, primarily manifested in increased government support in response to climate crises. Within a certain range, CPU can lead to greater government support, and the innovation compensation effect generated can exceed the cost effect before reaching the critical point, thereby promoting corporate technological innovation capabilities. However, excessive CPU can bring significant risks, where the cost effect outweighs the innovation compensation effect, negatively impacting corporate technological innovation capabilities. Therefore, CPU exhibits an “inverted U-shaped” impact.

Based on the above discussion, this study formulates the following hypothesis:

Hypothesis H2.

CPU will have an “inverted U-shaped” impact on corporate technological innovation capabilities.

2.2. Mechanism of Digital Peer Effect

The concept of the peer effect first emerged in the fields of education and psychology, referring to the phenomenon where individuals make decisions by referring to the characteristics and behaviors of other individuals within a reference group, and such decision-making behavior is based on rational analysis [26]. Recently, the concept of the peer effect has gained widespread application in studies related to corporate production, operational activities, and innovation. Matray’s study [27] found that firms in the same region influence each other’s level of innovation. Wang’s research found that corporate technological innovation decisions are significantly influenced by peer effects [28], which reduces corporate decision-making costs. Therefore, in the corporate decision-making process, the peer effect, as an important social psychological phenomenon, has significantly influenced technological innovation activities in industrial enterprises. Liu [29] believes that in the industrial sector, technological innovation activities among enterprises often exhibit interrelated and mutually influential trends. For industrial enterprises, technological innovation is a necessary path to achieve sustainable profitability and enhance new-quality productive forces. However, the high costs associated with technological innovation increase the operational risks of independent decision-making by enterprises, leading them to pay more attention to the innovation activities of firms within the same industry or region. Enterprises in the same industry have similar technological needs and market challenges, and enterprises in the same region are affected by similar policies. Based on the principle of profit maximization, enterprises adjust their strategic decisions according to the technological levels of peer enterprises. Kou’s research [30] has confirmed that this is particularly evident in technological innovation, thus giving rise to the digital peer effect. Thus, the study of the peer effect in the field of technological innovation holds significant practical importance.

It is worth pointing out global observations in terms of the challenges of innovative sustainability-oriented companies. Chomac [6] believes that pharmaceutical companies are investing more in innovation to comply with the requirements of sustainable development. Prado [31] pointed out that sustainability is considered a crucial factor for the short-term, medium-term, and long-term survival of businesses. Corporate managers’ decisions are influenced by government’s sustainable development requirements, like climate policies, especially in the field of technology research and development. Digitalization represents a technological revolution for industrial enterprises in the era of the digital economy. For industrial enterprises, the application of digital technologies not only enhances productivity but also brings benefits in energy conservation and emission reduction. Digital technologies are clean and pollution-free. Faced with uncertain climate policies and strict environmental regulations, the impact of technologies on the technological progress of industrial firms will be given more attention by managers. The digital peer effect in enterprises stems from the imitation of digital technology-related decisions made by firms within the same industry or geographical area. Wang’s research [20] has shown that under uncertain climate policies, information asymmetry is exacerbated, and corporate risk management costs increase. Corporate decision-makers, in order to mitigate risks, are more willing to rely on the existing decisions of peer enterprises when investing in digital technologies. Meanwhile, Wang et al. [32] found that digitalization and its peer effect can optimize the business model, management, and production processes of target enterprises, thereby enhancing their ability to withstand climate risks. Corporate managers will pay more attention to this in their decision-making. Therefore, the digital peer effect will enhance enterprises’ ability to withstand climate risks, mitigate the decision-making risks and costs generated by CPU, and facilitate corporate decision-makers in making more appropriate technological innovation decisions, thereby promoting the improvement in corporate innovation capabilities.

Based on the above discussion, this study formulates the following hypotheses:

Hypothesis H3a:

The digital peer effect within the same industry will mitigate the cost effect of CPU on target enterprises’ technological innovation and enhance their risk resistance capabilities.

Hypothesis H3b:

The digital peer effect within the same region will mitigate the cost effect of CPU on target enterprises’ technological innovation and strengthen their capacity to withstand risks.

To sum up, Figure 1 outlines the conceptual framework of this research, elucidating the influence of CPU on firms’ technological innovation capabilities and the moderating role of the digital peer effect. The framework highlights the non-linear relationship between CPU and corporate technological innovation capabilities, capturing the dual nature of CPU’s impact. Specifically, moderate levels of CPU can boost corporate innovation by increasing government subsidies, which in turn provide financial backing for R&D activities. Conversely, excessive uncertainty can intensify financing constraints, resulting in higher costs and diminished investment in innovation. This dual effect is depicted by the inverted U-shaped curve, with the optimal level of uncertainty situated at the curve’s peak.

3. Research Methodology

3.1. Data, Valuables, and Models

3.1.1. Data

Considering the availability of data and the historical context of the digital transformation of Chinese enterprises, the sample period for data analysis in this study spans from 2007 to 2023, focusing on A-share listed companies across various industrial sectors in China. In the A-share market, companies labeled as ‘ST’ are those that are found to have abnormalities in their financial status and other aspects after being reviewed by government departments. To prevent the potential adverse effects of these data on our findings, we exclude ST companies, resulting in a final sample of 3773 industrial listed companies from 31 provinces with a total of 20,738 annual observations. The data for this study are primarily sourced from the CSMAR database, partially sourced from annual reports of A-share listed companies. Referring to the Company Law of the People’s Republic of China, companies are required to accurately disclose financial information. Thus, the corporate data utilized in this study can be deemed genuine and reliable. Furthermore, to maintain uniformity in the statistical methodologies and terminology applied throughout mainland provinces, Macao, Taiwan, and Hong Kong have been omitted from the analysis. These regions possess unique economic and administrative frameworks that may not correspond with the mainland’s data gathering and reporting protocols.

3.1.2. Dependent Variable

Enterprise Technological Innovation Capability (Inn): In prior research, metrics such as R&D expenditure [33], the number of patent applications [34], and the number of authorized invention patents [7] have been utilized to reflect the degree of corporate technological innovation. This study opts to construct a composite index of enterprise technological innovation capability (Inn) by employing a weighted aggregation of multiple indicators. Among various methods for determining weights, the entropy weight method has gained widespread application due to its ability to objectively calculate the weights of indicators [35,36,37]. In line with this, the present study employs the entropy weight method to assess the technological innovation capability of enterprises, using the number of patent applications, the number of granted invention patents, and the R&D expenditure ratio as the basis. The detailed calculations for the entropy weight method are provided in Appendix A.1.

3.1.3. Independent Variable

Climate Policy Uncertainty (CPU): Drawing on the research by Ma et al. [38], we construct the index of climate policy uncertainty using the MacBERT deep learning algorithm, a method widely employed by scholars in recent studies [39,40]. This algorithm is based on over 1.7 million articles from mainstream Chinese newspapers such as the People’s Daily, Economic Daily, Guangming Daily, China News Service, and Science and Technology Daily, to develop a provincial-level index of CPU. We thus calculate the CPU of every province in China. Figure 2 illustrates the CPU for each province in China from 2007 to 2023, showing a clear upward trend in uncertainty over the years.

3.1.4. Moderating Variables

Digital Peer Effect: The digital peer effect is a combination of the concepts of digital transformation and the peer effect [41]. Firstly, a digital transformation index is constructed:

Digital Transformation Index (DTI): Following the approach by Wu et al. [42], this study constructs a digital transformation index (DTI) for enterprises. Initially, a corporate digitalization lexicon was developed based on the national policy semantic framework. We manually selected key national-level digital economy policy documents from the websites of the Central People’s Government and the Ministry of Industry and Information Technology. Relevant keywords pertaining to corporate digitalization were identified. Through Python word segmentation and manual verification, we extracted corporate digitalization-related terms that appeared with a certain frequency or higher to compile the corporate digitalization lexicon. This lexicon was subsequently incorporated into the “jieba” word segmentation library of the Python 3.12.0 software package. We then calculated the frequency of digitalization-related terms in the annual reports. After extracting the frequency of each keyword in the annual reports of each listed company for each year, we used the sum of these frequencies as a proxy for the degree of digitalization. For simplicity, this metric was multiplied by 100. A higher digital value signifies a greater extent of corporate digitalization.

Digital Peer Effect: To investigate the moderating impact of the peer effect, this study draws on the methodology used by Huo et al. [43]. Firms in the same industry as the focal enterprise, or those registered in the same province, are defined as industry peers and regional peers, respectively. The average digital transformation index (DTI) of these peer groups is used as an indicator to measure the digital peer effect within the same industry (IDPE) and the digital peer effect within the same region (RDPE).

3.1.5. Mediating Variables

Financing Constraint (Sa): There is no fixed indicator for measuring the financing constraints faced by enterprises. Scholars have employed various methods in the past. Among them, the SA index proposed by Hadlock [44] has been widely used in academia due to its ability to more intuitively reflect external financing constraints of enterprises [17,45]. This study employs the SA index to assess corporate financing constraints. The SA index is calculated as follows:

$$\mathrm{S}\mathrm{A}=-0.737\times \mathrm{S}\mathrm{i}\mathrm{z}\mathrm{e}+0.043\times \mathrm{S}\mathrm{i}\mathrm{z}{\mathrm{e}}^2-0.04\times \mathrm{A}\mathrm{g}\mathrm{e}$$

Government Subsidies (GSs): These are measured by the total amount of government subsidies received by the enterprise in a given year. Government subsidies are one of the macro-control tools of the government and can directly reflect the degree of support that enterprises receive from local governments [46]. Government subsidies play a significant role in promoting technological innovation in enterprises [47]. They are crucial for corporate technology research and development.

3.1.6. Control Variables

Referring to the research experience of numerous researchers in the field of industrial corporate technological innovation [48,49], this study considers control variables from aspects such as firm size, expenses, and equity structure. The source of variables set for this study are presented in

Table 1. Variable name and source.

Type	Name	Abbreviation	Data Source
Dependent Valuables	Enterprise Technological Innovation Capability	Inn	Constructed by using the entropy weight method. Patent and R&D are from CSMAR database
Independent Valuables	Climate Policy Uncertainty	CPU	Constructed by using MacBERT deep learning algorithm
Mediating Valuables	Financing Constraint	Sa	Size and age are from CSMAR database
	Government Subsidy	GS	CSMAR database
Moderating Valuables	The digital peer effect of firms in the same industry	IDPE	Annual reports of A-share listed companies
	The Digital Peer Effect of Firms in the Same Region	RDPE	Annual reports of A-share listed companies
Control Valuables	Enterprise Scale	Scale	CSMAR database
	Asset–Liability Ratio	Lev	CSMAR database
	Sales Period Fee Ratio	Fee	CSMAR database
	Separation Rate of Ownership and Control	Oc	CSMAR database
	Board of Directors Composition	Bd	CSMAR database
	Financial Performance	Roa	CSMAR database
	Nature of Ownership	SOE	CSMAR database

below:

3.2. Model Specification

To verify the relationship between CPU and industrial corporate technological innovation, drawing on the research of previous scholars [50], we initially conducted a Hausman test (see Appendix A.2), which indicated that a fixed-effects model is appropriate. Based on the test results, we opted for a fixed-effects model. Consequently, this study constructs the baseline regression model (1) as follows:

$${Inn}_{i,t}={\alpha }_{10}+{\alpha }_{11}{CPU}_{i,t}+{\alpha }_{12}{CPU}_{i,t}^2+\sum {\beta }_{1k}{Controls}_{i,t}+\sum Zone+\sum Industry+{\delta }_{1,t}+{\gamma }_{1,i}+{\epsilon }_{1,i,t}$$

(1)

where Inn represents enterprise technological innovation capability, CPU denotes the CPU, Controls are all the control variables, Zone is the regional dummy variable, Industry is the industry dummy variable, ${\mathsf{\delta }}_{\mathrm{t}}$ represents the time effect, ${\mathsf{\gamma }}_{\mathrm{i}}$ represents the firm-specific effect, and ${\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}}$ is the random disturbance term. To investigate the potential non-linear impact of CPU on enterprise technological innovation capability, the quadratic term of CPU was introduced into model (1). After that, we conducted a RESET test (see Appendix A.3) which proved that our quadratic model does not suffer from the omission of higher-order terms. We also conducted a White test (see Appendix A.3), which suggested that there is no heteroscedasticity in the model.

To test the mediating effects of financing constraints and government subsidies, and to verify the proposed transmission mechanisms of “Climate Policy Uncertainty–Financing Constraints–Corporate Innovation Capability” and “Climate Policy Uncertainty–Government Subsidies–Corporate Innovation Capability”, we constructed the mediation effect models (2)–(5) to assess the role of financing constraints using the three-step method referring to Jiang’s study [51]:

$${Inn}_{i,t}={\alpha }_{20}+{\alpha }_{21}{Sa}_{i,t}+\sum {\beta }_{2k}{Controls}_{i,t}+\sum Zone+\sum Industry+{\delta }_{2,t}+{\gamma }_{2,i}+{\epsilon }_{2,i,t}$$

(2)

$${Sa}_{i,t}={\alpha }_{30}+{\alpha }_{31}{CPU}_{i,t}+\sum {\beta }_{3k}{Controls}_{i,t}+\sum Zone+\sum Industry+{\delta }_{3,t}+{\gamma }_{3,i}+{\epsilon }_{3,i,t}$$

(3)

$${Inn}_{i,t}={\alpha }_{40}+{\alpha }_{41}{GS}_{i,t}+\sum {\beta }_{4k}{Controls}_{i,t}+\sum Zone+\sum Industry+{\delta }_{4,t}+{\gamma }_{4,i}+{\epsilon }_{4,i,t}$$

(4)

$${GS}_{i,t}={\alpha }_{50}+{\alpha }_{51}{CPU}_{i,t}+\sum {\beta }_{5k}{Controls}_{i,t}+\sum Zone+\sum Industry+{\delta }_{5,t}+{\gamma }_{5,i}+{\epsilon }_{5,i,t}$$

(5)

where Sa represents corporate financing constraints, and GS denotes government subsidies.

To examine how the digital peer effect influences the relationship between CPU and technological innovation in industrial firms, this study constructs the moderating effect models (6) and (7) as follows:

$$\begin{array}{l}{Inn}_{i,t}={\alpha }_{60}+{\alpha }_{61}{CPU}_{i,t}+{\alpha }_{62}{CPU}_{i,t}^2+{\alpha }_{63}{IDPE}_{i,t}+{\alpha }_{64}{CPU}_{i,t}\times {IDPE}_{i,t}+\\ {\alpha }_{65}{CPU}_{i,t}^2\times {IDPE}_{i,t}+\sum {\beta }_{6k}{Controls}_{i,t}+\sum Zone+\sum Industry+{\delta }_{6,t}+{\gamma }_{6,i}+{\epsilon }_{6,i,t} \end{array}$$

(6)

$$\begin{array}{l}{Inn}_{i,t}={\alpha }_{70}+{\alpha }_{71}{CPU}_{i,t}+{\alpha }_{72}{CPU}_{i,t}^2+{\alpha }_{73}{RDPE}_{i,t}+{\alpha }_{74}{CPU}_{i,t}\times {RDPE}_{i,t}+\\ {\alpha }_{75}{CPU}_{i,t}^2\times {RDPE}_{i,t}+\sum {\beta }_{7k}{Controls}_{i,t}+\sum Zone+\sum Industry+{\delta }_{7,t}+{\gamma }_{7,i}+{\epsilon }_{7,i,t}\end{array}$$

(7)

where IDPE represents the digital peer effect among firms in the same industry, and RDPE represents the digital peer effect among firms in the same region. To assess the non-linear moderating effect, referring to Lin’s theoretical study [52], we introduced the cross terms CPU × IDPE and CPU × RDPE as well as the quadratic cross terms CPU² × IDPE and CPU² × RDPE into the model.

3.3. Descriptive Statistics and Correlation

To eliminate the impact of outliers, referring to Li’s study [53], we applied winsorization to all continuous variables at the 1% level. Subsequently, descriptive statistics were performed on all variables, with the results presented in

Table 2. Data description.

Valuables	N	Mean	sd	Min	Max
Inn	20,737	2.457	4.661	0.251	34.280
GS	20,737	0.377	0.875	0.000	6.185
Sa	20,737	3.866	0.255	3.202	4.508
CPU	20,737	2.653	0.804	1.296	4.708
IDPE	20,737	10.430	9.639	0.157	35.320
RDPE	20,737	15.680	7.837	0.718	36.610
Scale	20,737	91.750	212.7	4.042	1582.569
Lev	20,737	0.390	0.193	0.0536	0.894
Fee	20,737	0.327	0.978	0.0152	8.357
Oc	20,737	4.409	7.022	0.000	28.290
Bd	20,737	0.558	0.185	0.198	0.941
Roa	20,737	0.027	0.035	−0.082	0.143

. The findings show that the minimum value of corporate innovation capability (Inn), as calculated by using the entropy weight method, is 0.251, the maximum value is 34.28, and the standard deviation is 4.661. This suggests that all of 3773 companies we selected have varying degrees of technological innovation performance, and after weighting the various innovation indicators with information entropy, the data tend to be smooth and are suitable for empirical research. The mean value of the IDPE is 10.43, which is slightly lower than the mean value of the RDPE at 15.68, indicating that the digital peer effect in the same region has a greater impact. The variables Scale and Oc exceed 1, which suggests that there are variations in the size and governance structure among the sample firms. Meanwhile, the standard deviations of the other control variables are all below 1, indicating relatively minor differences among them.

To preliminarily examine the relationship between CPU and Inn, a correlation coefficient test was conducted for the main variables of the model, and the correlation coefficient matrix is presented in

Table 3. Pairwise correlations.

Valuab-les	Inn	GS	Sa	CPU	IDPE	RDPE	Scale	Lev	Fee	Oc	Bd	Roa
Inn	1.000
GS	0.234 ***	1.000
Sa	−0.226 ***	−0.114 ***	1.000
CPU	−0.146 ***	−0.036 ***	0.189 ***	1.000
IDPE	−0.057 ***	0.018 **	−0.013 *	0.107 ***	1.000
RDPE	−0.158 ***	0.067 ***	0.133 ***	0.170 ***	0.315 ***	1.000
Scale	0.213 ***	0.630 ***	−0.193 ***	0.025 ***	−0.086 ***	0.085 ***	1.000
Lev	0.076 ***	0.234 ***	0.095 ***	0.003	−0.083 ***	−0.032 ***	0.293 ***	1.000
Fee	0.029 ***	0.056 ***	0.063 **	−0.055 ***	−0.016 **	0.027 ***	0.059 ***	0.103 ***	1.000
Oc	0.008	0.03 ***	0.023 ***	−0.019 ***	−0.069 ***	−0.117 ***	0.038 ***	0.052 ***	0.009	1.000
Bd	0.056 ***	0.045 ***	0.058 ***	−0.029 ***	−0.121 ***	−0.100 ***	0.079 ***	0.095 ***	−0.009	0.173 ***	1.000
Roa	0.014 *	0.014 **	−0.127 ***	−0.189 ***	−0.059 ***	−0.047 ***	−0.021 ***	−0.393 ***	−0.115 ***	0.032 ***	−0.002	1.000

Note: *** p < 0.01, ** p < 0.05, and * p < 0.1.

. The results indicate a strong correlation between the dependent and independent variables in the regression model. For instance, Inn is significantly positively correlated with GS and significantly negatively correlated with Sa. Conversely, CPU is significantly negatively correlated with GS and significantly positively correlated with Sa. These findings provide empirical support for verifying the mediating effects described in Hypotheses H1.a and H1.b. However, whether an “inverted U-shaped” relationship exists between Inn and CPU, as proposed in Hypothesis H2, further empirical testing is required. Additionally, the correlation coefficients among all control variables are below 0.3, suggesting that the selection of control variables is effective and that the endogeneity issue among these variables is relatively minor. Further, we performed the VIF test for all of the variables (see Appendix A.4,

Table A1. VIF test.

	(1)	(2)
Valuables	VIF	1/VIF
GS	1.70	0.589
Sa	1.16	0.864
CPU	1.12	0.891
IDPE	1.17	0.858
RDPE	1.16	0.864
Scale	1.87	0.535
Lev	1.38	0.727
Fee	1.03	0.969
Oc	1.05	0.955
Bd	1.06	0.941
Roa	1.28	0.782

), with the VIF values for all of the variables being small (<2), thus avoiding the problem of multicollinearity.

4. Empirical Research

4.1. Benchmark Regression Results

Table 4. Benchmark model regression results.

	(1)	(2)	(3)	(4)	(5)
Valuables	Inn	Inn	Inn	Inn	Inn
CPU	1.623 ***	1.986 ***	1.512 ***	0.768 **	0.741 **
	(4.65)	(6.47)	(4.42)	(2.56)	(2.49)
CPU2	−0.216 ***	−0.321 ***	−0.202 ***	−0.107 **	−0.104 **
	(−3.84)	(−6.38)	(−3.66)	(−2.23)	(−2.17)
Scale	0.005 ***	0.005 ***	0.005 ***	−0.002 ***	−0.002 ***
	(32.60)	(34.81)	(32.84)	(−5.60)	(−5.50)
Lev	0.215	0.237	0.397 **	1.747 ***	1.518 ***
	(1.28)	(1.41)	(2.35)	(6.14)	(5.35)
Fee	0.079 ***	0.102 ***	0.103 ***	0.216 ***	0.215 ***
	(2.75)	(3.59)	(3.59)	(6.28)	(6.30)
Oc	−0.003	−0.007 *	−0.002	−0.020 ***	−0.018 **
	(−0.76)	(−1.68)	(−0.54)	(−2.60)	(−2.25)
Bd	−0.204	0.070	0.069	0.085	0.055
	(−1.29)	(0.44)	(0.44)	(0.26)	(0.17)
Roa	−0.234	0.314	0.901	−3.662 ***	−3.881 ***
	(−0.26)	(0.35)	(1.00)	(−3.50)	(−3.73)
SOE	0.154 **	0.379 ***	0.393 ***	0.431 **	0.428 **
	(2.10)	(5.28)	(5.32)	(2.21)	(2.21)
Constant	12.913 ***	10.303 ***	10.851 ***	11.511 ***	0.875 *
	(19.82)	(14.67)	(15.12)	(18.82)	(1.76)
N	20,737	20,737	20,737	20,737	20,737
Adjusted R-squared	0.282	0.304	0.313	0.228	0.618
Firm				YES	YES
Industry		YES	YES		YES
Province	YES		YES		YES
Time	YES	YES	YES	YES	YES

Note: t-statistics are in parentheses. *** p < 0.01, ** p < 0.05, and * p < 0.1. Variables have been logarithmically transformed. “Yes” for “Firm”, “Industry”, “Province”, and “Time” means that our fixed-effects model controls for firm, industry, provincial, and time fixed effects.

displays the regression results for the benchmark model (1). Generally, studies account for firm and time fixed effects. Columns (1)–(4) in Table 4 illustrate various combinations of controlling for firm, industry, regional, and time fixed effects, whereas Column (5) includes all four fixed effects simultaneously. The coefficients for CPU in Columns (1)–(5) are all significantly positive, at 1.623, 1.986, 1.512, 0.768, and 0.741, respectively. Meanwhile, the coefficients for CPU² are all significantly negative, at −0.216, −0.321, −0.202, −0.107, and −0.104, respectively. The coefficients of the core explanatory variables are significant at the 5% level. Moreover, the goodness of fit (Adjusted R-squared) in Column (5) is significantly improved compared to the other scenarios. Taking Column (5) as an example, we observe an inverted U-shaped relationship between CPU and Inn, with the critical point of the model at CPU = 3.56 (see Appendix A.5, Figure A1a). This indicates that before CPU reaches a critical point, it enhances enterprise technological innovation capability. However, once it exceeds the critical point, the cost effect outweighs the compensatory effect of innovation. Thus, Hypothesis H2, which posits that CPU has an inverted U-shaped impact on enterprise technological innovation capability, is preliminarily confirmed.

4.2. The Impact Mechanism of CPU on Industrial Enterprise Technological Innovation

4.2.1. The Mediating Effects of Government Subsidies and Financing Constraints

Columns (1)–(4) in

Table 5. Mediating effect mechanism tests.

	(1)	(2)	(3)	(4)
Valuables	Inn	Sa	Inn	GS
Sa	−9.415 ***
	(−40.97)
GS			0.270 ***
			(5.85)
CPU		0.004 ***		0.030 **
		(5.13)		(2.41)
Controls	YES	YES	YES	YES
N	20,737	20,737	20,737	20,737
Adjusted R-squared	0.127	0.265	0.042	0.103
Firm	YES	YES	YES	YES
Industry	YES	YES	YES	YES
Province	YES	YES	YES	YES
Time	YES	YES	YES	YES

Note: t-statistics are in parentheses. *** p < 0.01 and ** p < 0.05. Variables have been logarithmically transformed. “Yes” for “Firm”, “Industry”, “Province”, and “Time” means that our fixed-effects model controls for firm, industry, provincial, and time fixed effects.

show the regression results for models (2)–(5). In Column (1), the coefficient of financing constraints (Sa) is significantly negative at the 1% level, with a value of −9.415. In Column (2), the coefficient of CPU is significantly positive at the 1% level, at 0.004. This indicates that CPU significantly increases the financing constraints faced by firms, thereby raising the costs of technological innovation and reducing their innovation capabilities. Thus, Hypothesis H1a is supported. In Column (3), the coefficient of government subsidies (GSs) is significantly positive at the 1% level. In Column (4), the coefficient of CPU is significantly positive at the 5% level. This suggests that CPU leads to increased government subsidies for firms, which in turn generates an innovation compensation effect that boosts technological innovation. Thus, Hypothesis H1b is supported.

CPU affects corporate technological innovation through two mediating variables, namely government subsidies and financing constraints, generating both cost effects and innovation compensation effects. When CPU is low, the compensation effect brought about by increased government subsidies outweighs the cost effect of financing constraints. However, when CPU is high, the financing constraints faced by firms are significantly increased, and the resulting cost effect outweighs the innovation compensation effect, having a negative impact on enterprise technological innovation capability.

Moreover, in Columns (1)–(3), the coefficients of the SOE dummy variable are consistently positive. Here, SOE = 1 signifies that the firm is a state-owned enterprise (SOE). This finding implies that the technological innovation capability of SOEs is more affected by financing constraints and government subsidies. In Column (4), the SOE coefficient is negative, suggesting that CPU has a relatively lesser impact on the government subsidies received by SOEs.

4.2.2. The Moderating Effects of the Digital Peer Effect

To avoid the endogeneity issue that may arise from the digital peer effect, i.e., the IDPE and RDPE may be affected by the target enterprises, we will use the past IDPE and RDPE for the model regression. Specifically, we subject both the IDPE and RDPE to a prior period. Columns (1) and (2) in

Table 6. Moderating effect mechanism tests.

	(1)	(2)
Valuables	Inn	Inn
CPU	3.066 ***	3.767 ***
	(21.71)	(20.53)
CPU2	−0.555 ***	−0.639 ***
	(−19.66)	(−16.68)
IDPE	0.085 ***
	(6.46)
IDPE × CPU	−0.110 ***
	(−7.02)
IDPE × CPU2	0.015 ***
	(5.53)
RDPE		0.240 ***
		(13.10)
RDPE × CPU		−0.160 ***
		(−10.30)
RDPE × CPU2		0.026 ***
		(8.15)
Controls	YES	YES
N	20,444	20,444
Adjusted R-squared	0.117	0.189
Firm	YES	YES
Industry	YES	YES
Province	YES	YES
Time	YES	YES

Note: t-statistics are in parentheses. *** p < 0.01. Variables have been logarithmically transformed. “Yes” for “Firm”, “Industry”, “Province”, and “Time” means that our fixed-effects model controls for firm, industry, provincial, and time fixed effects.

display the regression results for models (6) and (7), respectively. Upon incorporating the digital peer effect of firms in the same industry (IDPE) and the digital peer effect of firms in the same region (RDPE), the relationship between CPU and enterprise technological innovation capability (Inn) retains its “inverted U-shaped” pattern in both columns. This finding further corroborates Hypothesis H2.

Considering the impact mechanisms of the IDPE and RDPE separately, in Column (1), the coefficients of the interaction terms IDPE×CPU and IDPE × CPU² are both significant at the 1% level, which are −0.110 and 0.015. This indicates that the digital peer effect of firms in the same industry significantly moderates the “inverted U-shaped” relationship between CPU and Inn. The coefficient of the quadratic interaction term represents the change in the curvature of the “inverted U-shaped” curve (see Appendix A.5, Figure A1b). This suggests that the digital peer effect in the same industry alleviates the negative cost effect of CPU on Inn after the critical point, shifts the critical point to the right, indicating that firms have stronger climate risk resistance capabilities, and raises the overall curve, indicating an improvement in enterprise technological innovation capability.

In Column (2), the coefficients of the interaction terms RDPE × CPU and RDPE × CPU² are both significant at the 1% level, which are −0.160 and 0.026. This indicates that the digital peer effect of firms in the same region also significantly moderates the “inverted U-shaped” relationship between CPU and Inn (see Appendix A.5, Figure A1c). The moderating effects of the RDPE are similar to those of IDPE, as they can both mitigate the cost effect of CPU on industrial corporate technological innovation and enhance firms’ climate risk resistance capabilities. Thus, Hypotheses H3.1 and H3.2 are supported.

Moreover, comparing Column (1) and Column (2) shows that the coefficient of RDPE × CPU² is larger than that of IDPE × CPU². Since the information among decision-makers of firms in the same region is relatively more transparent compared to that of firms in the same industry, the digital peer effect of firms in the same region has a greater impact on the target firm than that of firms in the same industry. Compared to the IDPE, the RDPE shifts the critical point of the “inverted U-shaped” curve further to the right, meaning that it can more effectively enhance firms’ climate risk resistance capabilities (see Appendix A.5, Figure A1d). In Columns (1) and (2), the coefficients of the SOE dummy variable are significantly positive, implying that state-owned enterprises (SOEs) possess superior technological innovation capabilities relative to non-SOEs when both the digital peer effect and CPU are considered.

4.3. Robustness Test

4.3.1. Instrumental Variable Method

To verify the robustness of the regression results, this study first employs the instrumental variable method. Climate policies have a significant emission reduction effect [54], and thus their uncertainty is largely related to regional carbon emissions. Therefore, this study uses regional carbon emissions (CF) as an instrumental variable for CPU. Considering the impact of macroeconomic factors, this study also includes regional GDP as a macroeconomic control variable. The regression analysis is performed on the benchmark model (1), with the results presented in Column (1) of

Table 7. Robustness test results.

	(1)	(2)	(2)	(3)	(4)
Valuables	Inn	Inn	L.Inn	L2.Inn	Inn
CF	2.098 ***
	(6.23)
CF2	−0.285 ***
	(−8.30)
CPRI		0.279 ***
		(4.59)
CPRI2		−0.003 ***
		(−5.14)
CPU			1.066 ***	1.176 ***	−0.503 *
			(3.44)	(3.07)	(−1.79)
CPU2			−0.141 ***	−0.148 **	0.070
			(−2.91)	(−2.52)	(1.51)
GDP	0.815 **	0.233 ***
	(2.09)	(12.30)
Controls	YES	YES	YES	YES	YES
N	20,737	20,737	15,540	11,998	20,737
Adjusted R-squared	0.620	0.507	0.667	0.674	0.228
Firm	YES	YES	YES	YES	YES
Industry	YES	YES	YES	YES	YES
Province	YES	YES	YES	YES	YES
Time	YES	YES	YES	YES	YES

Note: t-statistics are in parentheses. *** p < 0.01, ** p < 0.05, and * p < 0.1. Variables have been logarithmically transformed. “Yes” for “Firm”, “Industry”, “Province”, and “Time” means that our fixed-effects model controls for firm, industry, provincial, and time fixed effects.

. In addition, we refer to Guo’s study [55] and construct the Climate Physical Risk Index (CPRI) as our instrumental variable as a way of demonstrating that Inn is influenced by climate policy risk-related factors.

After incorporating firm, industry, provincial, and time fixed effects, it is evident that the coefficient of CF² is significantly negative at −0.285, while the coefficient of CF is significantly positive at 2.098. Similarly, in Column (2), the coefficient of CPRI² is significantly negative at −0.003, and the coefficient of CPRI is significantly positive at 0.279. All of these coefficients are significant at the 1% level. Consequently, the “inverted U-shaped” relationship between CPU and enterprise technological innovation capability remains valid. The findings of this study are robust.

4.3.2. Lagged Regression

Considering that the influence of CPU on firms may have a lagged effect [56], this study conducts regression analyses on the benchmark model (1) with the dependent variable, enterprise technological innovation capability (Inn), lagged by one (L.Inn) and two periods (L2.Inn), respectively. The regression outcomes are displayed in Table 7, with Column (3) presenting the results for Inn lagged by one period and Column (4) for Inn lagged by two periods. These results are consistent with those obtained without lagging Inn.

In the regression analysis where Inn is lagged by one period (Column (3)), the coefficients for CPU and CPU² are both significant at the 1% level, with values of 1.066 and −0.141, respectively. For the regression with Inn lagged by two periods (Column (4)), the coefficient of CPU remains significant at the 1% level, at 1.176, while the coefficient of CPU² is significant at the 5% level, at −0.148. These results further confirm the robustness of the study’s conclusions.

4.3.3. Placebo Test

To further verify that the impact mechanism of the digital peer effect between CPU and industrial corporate technological innovation is not caused by external unobservable factors, this study conducts a placebo test. Specifically, for each year, all firms are randomly assigned to their respective regions and industries to create “pseudo-industries” and “pseudo-regions” while maintaining the total number of firms in the same year and industry, as well as the same year and region, consistent with the original data. The benchmark model (1) is then re-estimated with these placebo variables. The regression results are shown in Column (5) of Table 7.

The results show that the coefficient of CPU² is not significant, indicating no “inverted U-shaped” relationship between CPU and Inn. Specifically, the coefficient of CPU² is not negative. This suggests that the “inverted U-shaped” impact of CPU on industrial corporate technological innovation capability is not driven by external unobservable factors. Thus, the robustness of the study’s conclusions is further validated.

4.4. Heterogeneity Analysis

This study finds that CPU affects industrial corporate technological innovation levels through government subsidies and financing constraints. However, different enterprises in China are also significantly affected by climate policies to varying degrees. For instance, the benchmark regression analysis initially revealed that state-owned enterprises (SOE = 1) exhibit stronger technological innovation capabilities than non-state-owned enterprises (SOE = 0) when influenced by CPU and the digital peer effect. To delve deeper into the disparities among various enterprises, this study performs heterogeneity analysis based on whether the enterprise is part of the manufacturing sector.

The results of the heterogeneity analysis for model (1) are presented in

Table 8. Analysis of industrial heterogeneity.

	(1)	(2)
Valuables	Inn	Inn
CPU	0.634 **	2.548 **
	(2.07)	(2.14)
CPU2	−0.089 *	−0.339 *
	(−1.81)	(−1.81)
Controls	YES	YES
N	19,485	1239
Adjusted R-squared	0.598	0.857
Firm	YES	YES
Industry	YES	YES
Province	YES	YES
Time	YES	YES

Note: t-statistics are in parentheses. ** p < 0.05 and * p < 0.1. Variables have been logarithmically transformed. “Yes” for “Firm”, “Industry”, “Province”, and “Time” means that our fixed-effects model controls for firm, industry, provincial, and time fixed effects.

. Column (1) shows the regression results for manufacturing firms, while Column (2) displays the results for non-manufacturing firms. Upon comparing the coefficients of CPU and CPU², it is evident that both manufacturing and non-manufacturing firms demonstrate an “inverted U-shaped” relationship between technological innovation capability and CPU, with CPU² coefficients of −0.089 and −0.339, respectively. The “inverted U-shaped” curve of non-manufacturing firms has a greater curvature (meaning a larger absolute value of the coefficient), indicating that they are more significantly affected by CPU. In contrast, manufacturing firms are affected more gently by climate policies, with their critical point shifted to the right relative to non-manufacturing firms, suggesting that manufacturing firms have stronger climate risk resistance capabilities.

The heterogeneity analysis conducted in this study can assist governments in formulating differentiated policies based on firm characteristics or selecting representative firms for climate policy pilots. Additionally, this type of heterogeneity analysis can help firms of different types determine their development strategies at various stages of digitalization.

5. Conclusions

5.1. Discussion: Contribution and Innovation

This study finds that under the influence of CPU, state-owned enterprises (SOEs) exhibit stronger technological innovation capabilities. Non-SOEs need to pay more attention to the risks associated with CPU. Further subgroup analysis by industry reveals that in the industrial sector, manufacturing firms experience a more gradual impact of CPU compared to other industries. Manufacturing firms also demonstrate stronger climate risk resistance capabilities. In the current context of the digital economy and “dual carbon” environmental regulations, technology-intensive high-tech manufacturing firms, which are less affected by uncertain climate policies, should play a dominant role in achieving digital high-quality development through technological innovation.

Compared to past scholars’ studies on the linear influence of CPU, such as Bouri’s research [25] on the positive influence of CPU and Syed’s research [23] on its negative impact, our study considers the non-linear impact of CPU, specifically its “inverted U-shaped” effect, on industrial corporate technological innovation. This addresses a gap in the academic literature concerning the effects of CPU. Previous scholars have conducted numerous studies on the peer effect of enterprises. For instance, Kaustia [57] investigated the peer effect in corporate stocks by focusing on listed companies on the New York Stock Exchange. However, their research rarely touched on the peer effect of digitalization among firms in the same industry or region. For example, when Johnson [58] examined corporate digital innovation, he only considered the impact within the company’s internal departments, without involving the peer effect from other external enterprises. Additionally, while past studies have rarely examined the digital peer effect or treated it as an explanatory variable [41], we consider it a moderating variable, investigating its moderating effects, which is of innovative value.

5.2. Research Conclusions

Drawing on the existing literature related to CPU and digital transformation, this study examines the impact mechanism of CPU on industrial corporate technological innovation from the perspective of the digital peer effect. Taking China as an example, the research findings are as follows: (1) CPU exerts a negative cost effect on industrial corporate technological innovation, manifested in increased financing constraints for firms, while also generating a positive innovation compensation effect, reflected in higher government subsidies received by firms. Beyond a certain threshold, the cost effect surpasses the compensation effect of innovation, resulting in an overall “inverted U-shaped” impact on industrial corporate technological innovation capability. (2) The digital peer effect of firms in the same region or industry can effectively mitigate the “inverted U-shaped” impact of CPU on industrial corporate technological innovation, enhancing firms’ climate risk resistance capabilities. The regional peer effect is more significant than the industry peer effect, indicating that more detailed studies should be conducted in the future.

5.3. Research Implications

The implications for corporate decision-making and government policy derived from this study are as follows:

5.3.1. The Inverted U-Shaped Impact of CPU on Industrial Corporate Technological Innovation and Coping Strategies

Industrial firms need to profoundly understand the negative impact of CPU on industrial technological innovation, including increased costs of risky decision-making and financing constraints. Firms should address these challenges by establishing flexible strategic planning and risk management systems. Despite policy uncertainty, firms should also recognize the incentive mechanisms embedded in policies and how to leverage policy orientation to promote technological innovation. Firms can reduce the costs of technological innovation and enhance their innovation capabilities through government R&D subsidies, tax incentives, and other policy tools. From a long-term perspective, firms should acknowledge the potential innovation compensation effect brought about by CPU and, through continuous investment in technological innovation, ultimately achieve increased productivity and reduced costs.

5.3.2. The Role of the Digital Peer Effect in Technological Innovation and Its Application

Industrial firms should value the positive role of the peer effect in technological innovation. By collaborating and communicating with enterprises in the same industry or region, they can share best practices and technological innovation outcomes, thereby reducing the uncertainty and costs associated with technological innovation. In the face of CPU under the “dual carbon” goal, firms should accelerate their digital transformation, utilize digital technologies to improve production efficiency, mitigate environmental impacts, enhance their climate risk resistance capabilities, and meet policy requirements for sustainable development.

5.3.3. The Interplay Between Policy Formulation and Corporate Strategic Adjustment

When formulating climate policies, governments should consider the influences of policy uncertainty on the technological innovation capabilities of different firms. They should provide more guidance and support to help firms adapt to policy changes while encouraging technological innovation. Policies and strategies should be differentiated based on regional and industry-specific characteristics. Industrial firms, on the other hand, should flexibly adjust their strategic planning in response to changes in the external environment, especially CPU. This includes the direction and pace of technological innovation to enhance climate risk resistance and maintain competitive advantages.

5.4. Future Prospects

This study examines the impact mechanisms of CPU and the digital peer effect on industrial corporate technological innovation, providing new theoretical and empirical support for understanding the relationship between CPU and industrial technological innovation. However, this study has certain limitations. The conclusions of this study are based only on relevant data from the Chinese region and apply to the Chinese government and enterprises. At the same time, this study has limitations in the choice of time frame and is intended for policymakers only. This study also has limitations in the selection of the sample, focusing on industrial firms. With the intensification of global climate change and the acceleration of digital transformation, there are many areas for improvement in future research. For example, case study analysis methods can be used to further validate our empirical research results. This study focuses on Chinese enterprises, but the impact of CPU on technological innovation may vary depending on the policy environment, industrial structure, and firm characteristics of different countries or regions. Future research can be extended to other countries or regions to explore the heterogeneous impacts under different policy backgrounds, thereby providing more universal references for global climate policy-making. In addition, digital transformation is not only a tool for enterprises to improve production efficiency but also an important pathway to achieving green technological innovation. Future research can further explore how digital technologies can directly or indirectly promote green technological innovation by optimizing energy management, reducing carbon emissions, and so on, especially in terms of their application effects in different industries and types of enterprises. In terms of the time span, this study mainly focuses on the short-term impact of CPU on corporate technological innovation, but its long-term effects may be more complex. Future research can use dynamic panel models or event study methods to analyze the influences of CPU on firms’ long-term strategic adjustments (such as technology route selection and industrial chain layout).