Digital Policy Quality and Enterprise Innovation: The Case of China’s Big Data Comprehensive Pilot Zone

Abstract

In the context of promoting sustainable development and innovative growth, few studies have examined the signaling role of digital policy texts and its effects on enterprise innovation. Focused on the pilot project of China’s National Big Data Comprehensive Pilot Zone (NBCPZ), this article applies the modified PMC index model to measure the quality of the 81 digital policies issued by the NBCPZ from 2016 to 2022 and uses a regression model to assess the impact of the digital policies’ quality on the enterprise innovation levels. The results show that the quality of digital policies released by NBCPZ in China shows temporal fluctuation and regional heterogeneity. High-quality digital policies positively promote enterprise innovation, and this relationship exhibits heterogeneity based on enterprise and industry characteristics. High-quality digital policies can enhance enterprises’ innovation level by optimizing the regional data innovation environment and enhancing the willingness of enterprises to innovate with data. These findings reveal the potential effects of digital policy in fostering sustainable enterprise practices and driving enterprise innovation capabilities.

1. Introduction

An accelerated surge in the digital economy has changed the focus of industrial policy, making digital transformation a priority initiative for industrial policy support in developing countries [1]. Digital policies have emerged as a form of innovation policy in response to the digital revolution [2]. Well-designed digital policy frameworks can yield benefits for enterprise innovation [3,4]. These policies promote enterprise innovation by optimizing the allocation of innovation elements [5,6]. They transcend spatial–temporal constraints, reduce transaction costs [7], and foster the influx of innovative elements into sectors characterized by intensive knowledge application. Additionally, digital policies enhance the innovation ecosystem and stimulate the innovation vitality of enterprises [8]. Specifically, digital policies provide a better public service and innovation environment for enterprises by establishing the standard system, improving intellectual property protection, strengthening the digital infrastructure, and fostering synergies among government, industry, academia, and research [9,10]. However, the alignment of digital policies with the developmental characteristics and innovation cycles of enterprises is crucial. Insufficient continuity, stability, and predictability in policies may hinder the innovation willingness of enterprises [11,12]. Policies lacking efficacy impede market maturation, thereby weakening the impetus of enterprise innovation. Overreliance on simplistic or repetitive policy instruments can lead to misallocated innovation resources and inefficiencies. This poses the policy challenge of facilitating data utilization and technology adoption to counteract low absorption rates in the innovation process [13]. Therefore, the formulation of good digital policies is imperative for spurring enterprise innovation [14].

Policy quality stands as a cornerstone determinant of policy effectiveness [15,16], underscoring the importance of the policy design that integrates objective setting, focal area identification, technical facilitation, regulatory frameworks, and audience understanding [17,18]. Such a comprehensive approach is essential to ensure the efficacy and impact of policies, thereby enhancing enterprise innovation performance [19,20]. Most existing studies employ the difference-in-difference (DID) method to establish a causal relationship between digital policies and innovation performance. For example, Wei et al. (2023) contend that big data policies increase the “multiplier effect” of data applications, consequently fostering technological innovation in urban manufacturing [21]. Han and Mao (2023) analyzed panel data from Chinese listed companies to verify that the implementation of the NBCPZ and broadband China strategy strengthens enterprise intelligent transformation and green innovation [22]. Although these studies have contributed valuable insights, it remains challenging to develop truly effective policy optimization feedback for governments. The binary treatment of policy implementation in prior models risks information loss, which cannot fully reflect the policy intentions and effects. Furthermore, a gap exists in exploring the intricate mechanisms through which digital policy quality influences enterprise innovation. Therefore, this study aims to bridge this gap by examining whether the quality of digital policies can stimulate enterprise innovation and exploring the channels through which such influence is exerted.

This study selects the digital policy introduced by China’s National Big Data Comprehensive Pilot Zone (NBCPZ), primarily because these zones were established by the Chinese government as pioneering areas to advance its digital agenda. In recent years, the governments in these pilots have dedicated considerable efforts to the research and formulation of digital policies, unleashing a cascade of policies at various tiers annually [23]. Generally, policy text serves as a carrier of policy information, not only providing accessible records of the evolution and structure of the policy system but also fulfilling blueprint planning and governance functions. On the one hand, it conveys the government’s aims and strategies concerning the application of data as a crucial resource. On the other hand, it reflects the local administrative perspective on data-centric methodologies and the efficiency of policy implementation. These policies act as guiding principles for China’s digital economy expansion, not only offering comprehensive guidance and standards for its development but also transmitting clear signals from the government, thereby clarifying the direction of digital economy growth. Therefore, the digital policies launched by the NBCPZ government hold representativity and effectively reflect the overall quality of China’s digital economy policies [21].

This article selects 81 digital policy documents released by the government from 2016 to 2022. The modified policy modeling consistency (PMC) index model is used to calculate the consistency of policy texts as an indicator of policy quality and evaluate its impact on enterprise innovation. The possible contributions of this article are emphasized in the following aspects.

Firstly, while previous studies have investigated policy effectiveness through various methods, such as questionnaire analyses, fuzzy evaluation approaches, and expert interviews [24], they tend to concentrate on policy instruments or the broader policy context, neglecting a deep analysis of the intricate complexity of policy texts themselves. Our study places greater emphasis on examining the correlation between policy text characteristics and policy effectiveness and innovatively applies the modified PMC index model, a rigorous quantitative tool that directly assesses the quality of digital policy texts—a dimension that has received limited attention. By doing so, we bridge the gap between policy formulation and enterprise innovation performances, offering empirical evidence on how the textual design in digital policy documents can serve as a critical signal influencing corporate behavior and innovation strategies.

Secondly, a large body of previous literature validates various positive factors that influence enterprise innovation, ranging from financial resources and organizational structures to market conditions and regulatory frameworks [25,26,27]. However, a notable gap exists in the literature concerning the role of digital policy quality. Our study concentrates on this unexplored determinant, contributing to the understanding of the factors influencing enterprise innovation by incorporating digital policy quality into the analysis of the micro-innovation effects of digital policies.

Thirdly, while previous studies recognize the signaling role of policies in influencing enterprise decision-making, our study uniquely integrates this perspective into the context of digital policy and innovation [28]. We illuminate how the quality of digital policies, as communicated through policy texts, serves as a potent signal that actively guides and incentivizes firms’ innovation strategies. This perspective not only enriches the theoretical discourse on policy impacts but also broadens the application of signaling theory, a dimension that has not been systematically explored in previous literature.

2. Policy Background and Theoretical Hypothesis

2.1. Policy Background and Research Region

In the context of the digital economy, major economies, such as the United States, have made strides through the formulation of national data strategies and the reinforcement of domestic data legislation, thereby positioning data as a crucial strategic resource. Similarly, China has unveiled the “Action Plan for Promoting the Development of Big Data”, aligning with global efforts. Since September 2015, it has progressed with establishing National Big Data Comprehensive Pilot Zones (NBCPZs) in regions including Guizhou, the Beijing–Tianjin–Hebei area, Shenyang, Guangdong, Chongqing, Shanghai, Henan, and Inner Mongolia.

2.2. Theoretical Hypotheses

Institutional theory posits that institutions (including policies and regulations) shape organizational behavior and outcomes [29]. High-quality policies create a stable and predictable environment that encourages investment in innovative activities. Within this framework, the PMC index, as a measure of policy quality, directly relates to the institutional environment that fosters or inhibits innovation. Innovation systems theory highlights the interconnections of various actors (enterprises, universities, governments) and factors (policy, infrastructure, finance) in driving innovation [30]. Digital policies play a central role in coordinating these elements, with their quality influencing the efficiency and effectiveness of the innovation system. These two theories provide complementary perspectives that emphasize the critical role of policy quality in shaping organizational behavior and promoting enterprise innovation, respectively. In general, high-quality policies integrate both long-term and short-term goals and exhibit multidimensional coordination involving technology, finance, and mechanisms. However, the research findings regarding the impact of policy quality on enterprise innovation are inconclusive. Some scholars have argued that the potential information in policy text might hinder enterprises’ innovative performance [31], as enterprises are often confronted with a large number of fragmented policies issued by local governments. These policies are well-intentioned but may stifle innovation by providing uncoordinated guidance to enterprises. Moreover, policy effectiveness is inherently susceptible to variations in implementation contexts and regional discrepancies, introducing policy deviations that distort resource allocation and encourage enterprises’ opportunistic behavior [32]. Conversely, other scholars advocate that well-designed, high-quality policies by the government can play a constructive role in stimulating the technological innovation of enterprises [33,34]. This study posits that policy attention to data elements can improve policy accuracy and clarity, thereby creating a superior policy environment. In a supportive institutional environment, the synergy of sound data elements and advanced information and communication technologies can promote resource optimization, reduce information asymmetry, and lower transaction costs, thereby stimulating innovation [35,36]. However, if policy formulation only focuses on the technology while neglecting the crucial aspects, such as financial support, the maturation of the data elements market may be impeded, which burdens enterprises with substantial costs tied to digital technology’s development, counterproductive to fostering an innovative environment. In other words, only by achieving a harmonious integration of the multiple dimensions of data element advancement within policy discourse can the effectiveness of the policy be ensured and exert a favorable influence on the innovation performance of enterprises. Therefore, this article proposes the following hypothesis:

H1: 

High-quality digital policies can promote enterprise innovation performance.

The objective of the digital policy is to guide and regulate the development of regional digital transformation. High-quality digital policy texts are crucial in enhancing the region’s digital infrastructure, fostering a pool of digital experts, and refining comprehensive digital service ecosystems. Such enhancements are vital for cultivating a conducive environment for digital innovation, thereby significantly enhancing the innovative performance of enterprises. The construction of digital infrastructure provides support for enterprise digital transformation and innovative applications. The cultivation of digital talents provides intellectual resources for innovation, and a perfect digital service system reduces the transaction costs of enterprise innovation [37]. Collectively, these policy-driven advancements serve as pivotal foundations for the elevation of enterprise innovation performance.

H2: 

High-quality digital policies can promote enterprise innovation performance by optimizing the regional data innovation environment.

Signal theory, initially proposed by American economist Spence (2002) [38], is used to explain how organizations address information asymmetry through signaling mechanisms. According to signaling theory, policy texts reflect the intentions of the policy issue agencies and are a kind of policy signal for the strategic decision-making and production of various economic entities. The effectiveness of these policy signals is inherently linked to the quality of the signal itself, encompassing attributes like clarity, linguistic robustness, and coherence among policy statements. Similarly, digital policies issued by the NBCPZ convey government preferences and expectations to market participants, thus influencing their decision-making processes [3,39]. A high-quality digital policy text effectively reflects the government’s intention to encourage innovation, which will boost their subjective initiative and willingness to pursue digital innovation. Specifically, the clarity of innovation incentives, the certainty of policy implementation, and the consistency of policy benefits in well-designed policies elevate enterprises’ expectations for innovation success, alleviate perceived innovation risks, and foster a greater appetite towards digital innovation. This heightened willingness of enterprises to engage in digital innovation subsequently translates into investments in innovation, ultimately leading to measurable innovation outcomes and propelling overall improvements in enterprise innovation performance. Based on this analysis, we propose the third hypothesis:

H3: 

Higher-quality digital policies promote the innovation performance of enterprises by increasing their willingness to innovate digitally.

3. Research Design

3.1. Methodology

3.1.1. Citation and Modification of the PMC Index Model

The PMC index model is regarded as a quantitative method in policy evaluation within academia [40]. Its function is to assess the text clarity and the content comprehensiveness of the policies to determine whether the text contains guidelines for specific issues. In this article, some indicators of the existing PMC index model are modified to measure the textual quality of digital policies. By comparing and analyzing the average PMC index value of various NBCPZs, this study aims to identify value transmission patterns, deficiencies, and shortcomings in digital policies, offering benchmarks and empirical evidence for policy improvements.

3.1.2. Regression Model

We constructed the following model to explore the impact of the quality of policy text and enterprise innovation

$${EI}_{i,k,t}={\alpha }_0+{\alpha }_1{PMC}_{i,t-1}+{\alpha }_2{Control}_{i,k,t}+u_k+{\delta }_t+{\epsilon }_{i,k,t}$$

In this equation, ${EI}_{i,k,t}$ represents the innovation level of the k enterprise in the i city in year t. ${PMC}_{i,t-1}$ denotes the quality of policy text issued by the government i in the previous year. ${Control}_{i,k,t}$ includes the series of control variables. ${\alpha }_0$ is a constant term, ${\alpha }_1$ and ${\alpha }_2$ are estimated coefficients for the independent variable and the control variables, respectively. $u_k$ is the fixed effect of the enterprise, and ${\delta }_t$ signifies the fixed effect of the year, ${\epsilon }_{i,k,t}$ is the residual term.

3.1.3. Theoretical Background on Methodological Adoption

This study integrates the two aforementioned methods and employs the modified PMC index model to evaluate policy text quality, followed by regression analysis of the impact of digital policy quality on enterprise innovation, which is based on two main considerations.

On the one hand, at the theoretical level, our approach is grounded in a series of studies that apply regression models for evaluating policy effects. These studies generally focus on exploring the impact of textual policy attributes, such as the choice of policy instruments [41], policy volume [42], policy intensity [43], and government attention [44], on enterprise strategy and behaviors. Consequently, they offer profound invaluable insights into how policy quality translates into micro-level enterprise strategic decisions, actions, and performance, thus validating our research methodology. However, the policy assessment measures used in the above literature tend to rely solely on descriptive statistics, which ignores the large amount of information embedded in the policy text. Some studies recommend employing the PMC index model as the approach to address the aforementioned shortcomings, thereby enabling a more detailed evaluation of policy document quality [45]. However, given that the original PMC index model primarily serves as a reference framework for modeling economic policies and is not without its limitations, subsequent literature suggests the necessity of adapting and refining the model to accommodate the specific characteristics of research subjects. Therefore, in this article, we assess the quality of policy texts by the modified PMC index model. The advantage of this method lies in its function to comprehensively assess the quality of digital policy texts through the construction of multivariate indicators. On the other hand, from the functional perspective of the methods, while the PMC index model is a structured method for qualitatively assessing policy design, coherence, and mix, it does not inherently quantify the strength or direction of the relationship between policy quality and innovation performance. The combination of the PMC index with regression analysis is grounded in the necessity for both a robust measurement of policy quality and a statistical method capable of elucidating causal relationships in complex systems. The regression model complements the PMC index model by controlling for confounding variables, enabling an estimation of the net effect of policy quality on enterprise innovation. In addition, by conducting regression analysis, we can explore potential heterogeneities in the relationship between policy and innovation across different enterprise types or industries, as well as the influence mechanisms. This exploration would be significantly more challenging if solely relying on the PMC index assessment.

3.2. Variable and Data Sources

3.2.1. Dependent Variable—Enterprise Innovation (EI)

In this article, we use the logarithm of invention patents and utility model patents plus one to represent the enterprise innovation level. The data source is the CSMAR database.

3.2.2. Independent Variable—PMC Index Value

The PMC index value of each region’s digital policy text, which is the quality of the digital policy, is used as the independent variable in this article. Next, the article elaborates on the measurement of this variable.

(1)

Data and sources

The primary sources include digital policies released by the governments of NBCPZ in China. The collection strategies are as follows. The first step is to select the search keywords, which include “big data”, “digitization”, “digital economy”, and “Internet data”. The second step is to search for relevant documents associated with these keywords through the official portal websites of various ministries, including but not limited to the People’s Government, Development and Reform Commission, and the Economic and Information Technology Commission. To ensure the comprehensiveness and completeness of the samples, we have also augmented the data collection process through two authoritative channels: the Peking University policy document database and the Baidu search tool. The third step involves the screening process for digital policies, adhering to the following criteria: (1) inclusion is reserved for policy formats such as plans, opinions, or notifications, excluding documents categorized as treaties, requests, approvals, or regulations. (2) The timeframe of the document starts in 2016 and extends to 2022. (3) Policy documents with textual content closely related to digital and data development planning are retained. Finally, we select 81 policy documents for further analysis, labeled P1 to P81, as shown in

Table 1. List of policy documents.

ID	Policy Name	Release Date	Release Agency
P1	Beijing Big Data and Cloud Computing Development Action Plan (2016–2020)	2016	Beijing People’s Government
P2	Implementation Plan for the Deepening of the “Internet + Retail” Action in Beijing	2017	Beijing People’s Government
……	……	……	……
P81	Guangdong Province Digital Economy Development Guide 1.0	2022	Guangdong Province Department of Industry and Information Technology

(the specific documents are shown in Appendix A).

(2)

Measurement tools and methods

Firstly, we construct variables based on the content characteristics of digital policy texts, including primary and secondary variables, with values as shown in Formula (1):

$$X~N[0,1]$$

(1)

Secondly, we use the binary method to assign values to secondary variables, as shown in Formula (2):

$$X=\left\{\mathit{XR}:\left[0~1\right]\right\}$$

(2)

Next, the primary variable’s value is then computed by combining it with the secondary variable, as shown in Formula (3):

$$X_t\left({\displaystyle \sum _{j=1}^n\frac{X_{tj}}{T\left(X_{tj}\right)}}\right), t=1, 2, 3$$

(3)

where t is a primary variable and j is a secondary variable.

Formula (4) shows how the PMC index is calculated:

$$\begin{gathered} PMC=[X_1\left({\sum }_{i=1}^5\frac{X_{1i}}{5}\right)+X_2\left({\sum }_{i=1}^3\frac{X_{2k}}{3}\right)+X_3\left({\sum }_{i=1}^5\frac{X_{3l}}{5}\right)+X_4\left({\sum }_{i=1}^7\frac{X_{4m}}{7}\right)+X_5\left({\sum }_{i=1}^3\frac{X_{5n}}{3}\right)+X_6\left({\sum }_{i=1}^5\frac{X_{6o}}{5}\right)+ \\ {X}_7\left({\sum }_{i=1}^5\frac{X_{7p}}{5}\right)+X_8\left({\sum }_{i=1}^5\frac{X_{8q}}{5}\right)+X_9\left({\sum }_{i=1}^5\frac{X_{9r}}{5}\right)] \end{gathered}$$

(4)

where $X_i$ is the primary variable (i = 1, 2, 3, …, 9). $X_{ij}$ represents the total number of secondary variables.

(3)

Establishment of an indicator system for the PMC index

1. Theoretical analysis of PMC indicator construction

In the formulation of digital policies, governments need to ensure the quality of policy documents from both a formulation and implementation perspective [45]. From a decision-making perspective, an effective policy document should comprehensively capture the diverse interests of stakeholders, encompass multiple digital aspects, and address the extensive scope of digital strategy’s influence. Moreover, evaluating digital policy texts necessitates a multidimensional approach, considering policy domains, functions, and content. From an implementation perspective, a high-quality policy text should be explicitly stated and minimize conflicts [46]. Clarity and specificity dictate that digital policy texts concentrate on distinct policy areas, clearly defining functions, policy actors, and recipients, thereby fostering effective execution. Furthermore, minimizing conflicts involves harmonizing long-term visions with short-term objectives, providing greater flexibility for policy adaptation and reform, and ultimately ensuring policy continuity.

2. The modification of the original PMC index model and selection of indicators

The original PMC index model established by Estrada involves 10 primary variables and 50 subvariables, of which the main variables include type, research orientation, source, econometric methodology, research areas, theoretical framework, policy modeling by sector, economics framework, geographic analysis, and paper citation. This study’s PMC index model retains most of the variables of the original model. However, the original model is constructed based on academic papers rather than policy documents, resulting in the fact that some original indicators are not entirely suitable to the content characteristics of digital policies. To ensure the scientific rigor and rationality of the policy evaluation system, we optimize the indicators by deleting and refining them. Specifically, given that the policy texts are issued independently by governments and do not involve inter-policy citations, we exclude the citation indicator (X10) from the original model. Furthermore, since some original indicators, such as research orientation, econometric methods, and geographic analyses, are not consistent with the digital policy content, we follow the modification measures proposed by other scholars [47,48] and replace them with policy evaluation (X5), policy guarantees (X7), and policy focuses (X8) by mining digital policy texts and conducting thematic analysis.

Of these, the policy evaluation (X5) indicator consists of three subindicators. Sufficient basis (X5:1) ensures the credibility of the policy formulation process. Clear goals (X5:2) reflect the expected outcomes and the foundation for the evaluation. Scientific, mature, and feasible (X5:3) ensures that the policy is theoretically sound and practically feasible in practical application. Policy guarantees (X7) ensure that there are solid foundations for the implementation of digital strategies. The institutional system (X7:1) assesses the structural frameworks in place to support and regulate the digital sector. Government subsidy (X7:2) measures the financial incentives provided by the state, which can stimulate innovation and adoption of digital technologies. Digital talent introduction and training (X7:3) are used to evaluate the emphasis on cultivating a skilled digital workforce. Demonstration pilot (X7:4) assesses the presence of experimental initiatives, enabling policy testing and refinement before broader application, thereby minimizing risks and enhancing effectiveness. Lastly, propaganda promotion (X7:5) stands for the efforts in communicating and promoting digital policies to stakeholders and the public, essential for awareness, engagement, and, ultimately, the successful adoption of policy. The policy focus (X8) assesses the depth and maturity of the digital policy as it relates to the focus aspects. Given the important role of data elements in digital development, we include indicator data value (X8:1) to evaluate the extent to which policies prioritize the exploitation and capitalization of data elements. Given the fundamental role of digital infrastructure in fostering a digital economy and society, we need to evaluate whether infrastructure construction (X8:2) is incorporated into digital policies. Innovative application of data (X8:3) is used to assess the promotion of creative applications of data in policies aimed at fostering an innovative environment. As a region exploring new modes of data-driven economic growth, a core mission of the NBCPZ is to lead the intelligent upgrading of industries through technological innovation and application promotion. Consequently, its digital policy text should highlight how to guide and promote this transformation, which is the main reason for the inclusion of indicators of the industrial transformation and upgrading (X8:4). The indicator digital governance and public services (X8:5) is used to measure the level of public service provision and governance mechanisms in the digital domain. In summary, the modified model proposed in this article maintains the integrity of the original PMC index model indicators to the greatest extent and ensures that the policy assessment framework is better aligned with the characteristics of digital policy documents. By introducing crucial dimensions such as policy guarantee and policy focuses, based on the textual characteristics of digital policies, it ensures the accuracy of the findings, thereby comprehensively reflecting the status of digital policies and enabling policymakers to gain a more practical understanding of policy effectiveness.

Finally, the modified PMC index model comprises 9 primary variables and 43 secondary variables, as shown in

Table 2. The setting of policy evaluation variables.

Primary Variables	Secondary Variables	Criteria	Reference	Function
Policy nature (X1)	Prediction (X1:1)	Is there a statement that reflects policy predictability?	Ruiz Estrada (2011) [49]	Measure the ability of policies to anticipate future conditions and provide stability
	Supervision (X1:2)	Whether the policy content involves data regulation		Assess the inclusion of rules for managing and overseeing data usage
	Suggestion (X1:3)	Whether the policy has the recommended content		Identify if the policy offers suggestions for actions
	Description (X1:4)	Whether the policy has descriptive content about the data development environment		Evaluate the explanation of the status of data ecosystems
	Diagnosis (X1:5)	Whether the policy contains a diagnosis of the current data system and the difficulties of policy implementation		Analyze system weaknesses and obstacles
Policy timeliness (X2)	Long term (more than 5 years) (X2:1)	Whether the long-term plan is mentioned	Zhao (2023) [50]	Assess policy alignment with short-, medium-, and long-term objectives
	Medium term (3–5 years) (X2:2)	Whether the medium-term policy plan is mentioned
	Short term (1–3 years) (X2:3)	Whether the short-term policy plan is mentioned
Policy releases agency (X3)	Provincial committee (X3:1)	Whether the policy issuance agency is the provincial committee	Liu (2017) [51]	Assess policy influence based on issuer authority
	Provincial People’s Congress (X3:2)	Whether the policy issuance agency is the provincial People’s Congress
	Provincial government (X3:3)	Whether the policy issuance agency is the provincial government
	Provincial government departments (X3:4)	Whether the policy issuance agency is the provincial government department
	Departments and institutions directly under the government (X3:5)	Whether the policy issuance agency is the departments and institutions directly under the government
Policy type (X4)	Law (X4:1)	Whether the type of policy is a law	Self-built according to sample policy types	Distinguish legal bindings, strategies, operational plans
	Plan (X4:2)	Whether the type of policy is a plan
	Implementation (X4:3)	Whether the type of policy is an implementation program
	Outline (X4:4)	Whether the type of policy is an outline
	Opinion (X4:5)	Whether the type of policy is an opinion
	Regulation (X4:6)	Whether the type of policy is a regulation
	Others (X4:7)	Whether the type of policy belongs to others
Policy evaluation (X5)	Sufficient basis (X5:1)	Whether the basis for policy formulation is sufficient	Liu et al. (2022) [52]	Validate policy text quantity with evidence, clarity, and feasibility checks
	Clear goals (X5:2)	Whether the goals set by the policy are clear
	Scientific, mature, and feasible (X5:3)	Whether the policy-making plan is scientific
Policy domain (X6)	Economy (X6:1)	Whether the policy involves the economic field	Shen et al. (2021) [53]	Highlight policy coverage across economic, tech, social, political, and environmental aspects
	Technology (X6:2)	Whether the policy involves technology
	Society (X6:3)	Whether the policy involves the society field
	Politics (X6:4)	Whether the policy involves the political field
	Environment (X6:5)	Whether the policy involves the institution’s environmental field
Policy guarantees (X7)	Institutional system (X7:1)	Whether the policy involves institutional and regime-related aspects	Dai et al. (2021) [54]	Assess the structural frameworks in place to support and regulate the digital sector
	Government subsidy (X7:2)	Whether the policy involves government subsidy		Measure the financial incentives provided by the state
	Digital talent introduction and training (X7:3)	Whether the policy involves digital talent recruitment and training		Evaluate the focus on cultivating a skilled workforce
	Demonstration pilot (X7:4)	Whether the policy involves a demonstration pilot		Assess the presence of experimental initiatives
	Propaganda promotion (X7:5)	Whether the policy involves propaganda promotion		Stand for the efforts in communicating and promoting digital policies to stakeholders and the public
Policy focuses (X8)	Data value (X8:1)	Whether policy priorities involve data value	Mining results based on text theme	Evaluate the degree to which policies prioritize the exploitation and capitalization of data elements
	Infrastructure construction (X8:2)	Whether policy priorities involve digital infrastructure		Present the fundamental role of digital infrastructure
	Innovative application of data (X8:3)	Whether policy priorities involve innovative application of data		Assess the promotion of creative applications of data
	Industrial transformation and upgrading (X8:4)	Whether policy priorities involve industrial transformation and upgrading		Target modernization of industries through digital integration and optimization
	Digital governance and public services (X8:5)	Whether policy priorities involve digital government and public services		Measure the level of public service provision and governance mechanisms
Policy receptor (X9)	Government (X9:1)	Whether the policy recipient involves the government	Bartolacci et al. (2019) [55]	Clarify intended beneficiaries, affecting policy reception and efficacy
	Enterprise (X9:2)	Whether the policy recipient involves an enterprise
	Scientific research academy (X9:3)	Whether the policy recipient involves a scientific research academy
	Financial institution (X9:4)	Whether the policy recipient involves a financial institution
	The public (X9:5)	Whether the policy recipient involves the public

. The function of each indicator and the basis for its revision are also presented. We assign values to each variable according to the following rules: a value of “1” is assigned if the digital policies under evaluation align with the descriptive criteria of the secondary variables; otherwise, a value of “0” is recorded.

(4)

Establish a multi-input–output table

The multi-input–output table is a robust data analysis framework capable of accommodating substantial datasets and quantifying individual variables across multiple dimensions. Its establishment facilitates a thorough and systematic assessment of digital policies, as exemplified in

Table 3. The multi-input–output table is used in the process of quantitative evaluation of the consistency level of China’s digital policies.

Primary Variables	X1	X2	X3	X4	X5	X6	X7	X8	X9
Secondary variables	X1:1	X2:1	X3:1	X4:1	X5:1	X6:1	X7:1	X8:1	X9:1
	X1:2	X2:2	X3:2	X4:2	X5:2	X6:2	X7:2	X8:2	X9:2
	X1:3	X2:3	X3:3	X4:3	X5:3	X6:3	X7:3	X8:3	X9:3
	X1:4		X3:4	X4:4		X6:4	X7:4	X8:4	X9:4
	X1:5		X3:5	X4:5		X6:5	X7:5	X8:5	X9:5
				X4:6
				X4:7

.

(5)

Measurements and grading of PMC index values

In line with the assessment standards proposed by Ruiz Estrada (2011) [49], we employ the PMC index to calculate the policy quality. Referring to other scholars [56], we have categorized the PMC index values into four levels, ranging from “Poor” to “Perfect”, as detailed in

Table 4. The grading scale for policies is based on the policy modeling consistency index.

PMC Index Score	0–3.99	4–6.99	7–7.99	8–9
Evaluation level	Poor	Qualified	Excellent	Perfect

.

(6)

NBCPZ policy analysis and evaluation

1. Analysis based on a temporal perspective

Figure 1 depicts the evolution of the average PMC index values of policy text in the China NBCPZ from 2016 to 2022. Overall, these values show significant fluctuations, peaking at 6.14 in 2021 and dipping to a low of 5.07 in 2018. The average PMC value of policy texts has remained at the “Qualified” level over the years, indicating that the overall consistency of NBCPZ’s policy text remains less than optimal and characterized by noticeable instability. The possible explanation for this phenomenon lies in the complexity of digital policy formulation, which involves dynamic factors such as regional economic strategies, resource allocation, and the phase of digital economy development.

Figure 2 shows the spatial disparities in the quality of digital policies through the perspective of average PMC index values. Guangdong Province exhibits the highest PMC index value across all regions, closely followed by Beijing, both occupying the top echelon in terms of digital policy PMC index values. This dominance can be attributed to their standing as strong economic provinces, characterized by robust digital resource endowments and a flourishing ecosystem for tech-driven innovations. Following closely in the second tier are regions like Chongqing, Shanghai, and Henan, showing a secondary cluster of policy capabilities. The third tier includes Tianjin, Guizhou, Inner Mongolia, Hebei, and Shenyang. One plausible explanation is that, unlike other economically developed regions, these areas are in the early phase of digital economy development, coupled with a relatively limited reserve of digital resources. Furthermore, the formulation of digital policies is a complex process influenced by various factors such as regional economic strategies and infrastructure availability, leading to varied disparities in policy focus among governments across China, thereby adding to the complexity of China’s digital policy.

Figure 1. The evolution of the PMC index average value.

Figure 1. The evolution of the PMC index average value.

2. Analysis based on a spatial perspective

Figure 2. The spatial characteristics of the PMC average value.

Figure 2. The spatial characteristics of the PMC average value.

3.2.3. Control Variables

To address the issue of omitted variable bias and to better quantify the relationship between digital policy and enterprise innovation, this article selects the following control variables: (1) age is measured by the logarithm of the enterprise’s age. (2) Current ratio (Cr) is calculated by the ratio of current assets to current liabilities, which represents the enterprise’s solvency. (3) Leverage (Lev) is defined as the ratio of total liabilities to total assets. (4) TobinQ is quantified by the ratio of enterprise market value to total assets, serving as a measurement of the enterprise’s overall value. (5) Size is measured by the logarithm of the enterprise’s total assets. (6) Stuff represents the logarithm of the number of employees. (7) Largest shareholder stake (Largest) is measured by the logarithm of the proportion of shares held by the largest shareholder. All data are sourced from the CSMAR database.

3.2.4. Mechanism Variables

The data innovation environment is the first mechanism variable, and this article selects 12 indicators and uses the entropy method for assessment. The details are shown in Appendix B. Regarding the innovative willingness of enterprises, we use the number of digital patent applications of enterprises to represent the propensity to engage in digital innovation.

3.2.5. Variable and Data Sources

The study sample comprises A-share Chinese enterprises listed between 2016 and 2022. To ensure the effectiveness of the results, we have established several criteria for optimizing the data sample: (1) enterprises with an asset–liability ratio over 1 are not included; (2) enterprises exhibiting abnormal operations or facing the risk of delisting are not included; (3) to test the net effect of digital policies, enterprises in the information technology industry are excluded from the analysis; (4) samples with missing data or anomalies are excluded.

4. Empirical Analysis and Findings

4.1. Descriptive Statistics

The descriptive statistical results of the main variables are listed in

Table 5. Descriptive statistics for variables.

Variable Type	Variables	Obs.	Mean	SD	Min	Max
Explained variable	EI	7379.000	1.822	1.704	0.000	6.428
Explanatory variable	PMC	7379.000	5.886	0.841	3.610	7.340
Control variables	Age	7379.000	3.003	0.301	2.197	3.611
	TobinQ	7379.000	1.929	1.229	0.829	8.246
	Cr	7379.000	2.627	2.693	0.352	16.975
	Lev	7379.000	0.425	0.206	0.054	0.894
	Size	7379.000	22.472	1.458	19.966	26.926
	Amount	7379.000	7.715	1.388	4.605	11.712
	Largest	7379.000	3.427	0.481	2.138	4.318
Robustness variable	Inva	7379.000	0.255	0.349	0.000	1.000
Endogenous variables	Urban	7379.000	5251.721	3383.092	1165.000	9466.070
	Indus	7379.000	2.240	1.432	0.906	5.283
Mechanism variables	Dipa	7379.000	0.733	1.236	0.000	8.332
	Dece	7379.000	0.378	0.221	0.046	0.833

Note: Mean is the average value of the variables, SD stands for the standard deviation, Min stands for the minimum value of the variable, and Max stands for the maximum value of the variable.

. The mean value of EI is 1.822, with a variance of 1.704. The average value of the PMC index is 5.886, categorizing it as the “Qualified” level.

4.2. Benchmark Regression Results

The benchmark regression results are shown in

Table 6. Benchmark regression results.

	(1)	(2)	(3)
Variables	EI	EI	EI
L.PMC	0.0437 ***	0.0436 ***	0.0427 ***
	(0.0130)	(0.0130)	(0.0129)
Cr		−0.0036	0.0020
		(0.0112)	(0.0117)
Lev		−0.0389	−0.1642
		(0.1806)	(0.1832)
TobinQ		−0.0317 *	−0.0153
		(0.0179)	(0.0174)
Age			−1.0272 *
			(0.5317)
Size			0.0800
			(0.0623)
Amount			0.2134 ***
			(0.0611)
Largest			0.1383
			(0.0933)
Constant	1.5927 ***	1.6790 ***	0.8744
	(0.0752)	(0.1315)	(2.0916)
Firm-FE	Yes	Yes	Yes
Year-FE	Yes	Yes	Yes
Obs.	7379	7379	7379
R-squared	0.8334	0.8335	0.8355

Note: Standard errors in parentheses, * p < 0.1, *** p < 0.01.

, and firm-fixed effects and time-fixed effects are controlled. Model (1) initially excludes control variables yet reveals a significantly positive coefficient for L.PMC at the 1% significance level. In models (2)–(3), control variables are sequentially integrated into the base model, sustaining a consistently positive trend for the explanatory variables. Specifically, for every 1% increase in the PMC, the innovation level of enterprises increases by 4.27 percent, with significant corresponding coefficients at the 1% level. These empirical results provide validation for the first hypothesis.

4.3. Robustness Check

Given the different potential influences that the regression methods selection, main variables measurement, and variations in the research samples may exert on the regression results, this article conducts the following robustness analysis.

4.3.1. Substituting the Estimation Methods

Referring to relevant literature [57], we further examined the robustness of H1 by changing the method. Specifically, we use the least squares dummy variables (LSDV) model and the feasible generalized least squares (FGLS) model. The results are displayed in columns (1)–(4) of

Table 7. Robustness analysis—research methods.

	(1)	(2)	(3)	(4)
Variables	EI	EI	EI	EI
L.PMC	0.0437 ***	0.0427 ***	0.0433 ***	0.0423 ***
	(0.0147)	(0.0147)	(0.0139)	(0.0139)
Control	No	Yes	No	Yes
Firm-FE	Yes	Yes	Yes	Yes
Year-FE	Yes	Yes	Yes	Yes
Obs.	7379	7379	7379	7379
R-squared	0.8703	0.8722	0.8577	0.8616

Note: Standard errors in parentheses, *** p < 0.01.

. Columns (1) and (3) do not include control variables, whereas columns (2) and (4) incorporate them. The latter results show the estimated coefficients for the L.PMC are 0.0427 and 0.0423, which are significantly positive at the 1% level. This result provides compelling evidence for the robustness of previous findings.

4.3.2. Adjusting Sample Capacity

To address potential sample selection bias in our research results, we re-evaluate the impact of policy quality on enterprise innovation by adjusting the sample size. Firstly, this study excludes some special enterprise samples that had never applied for a patent during the study period, thereby eliminating enterprises with zero patent applications. The results are reported in

Table 8. Robustness analysis.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
Variables	EI	EI	EI	EI	EI	EI	EI	EI
L.PMC	0.0545 ***	0.0540 ***	0.0456 ***	0.0448 ***	0.0442 ***	0.0437 ***	0.0513 ***	0.0509 ***
	(0.0164)	(0.0164)	(0.0098)	(0.0135)	(0.0141)	(0.0141)	(0.0138)	(0.0138)
Control	No	Yes	No	Yes	No	Yes	No	Yes
Firm-FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year-FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Obs.	5680	5680	7041	7041	6431	6431	6950	6950
R-squared	0.7652	0.7696	0.8379	0.8401	0.8341	0.8363	0.8345	0.8366

Note: Standard errors in parentheses, *** p < 0.01.

, columns (1)–(2). Secondly, considering the distinct characteristics of Guizhou Province as the first batch of NBCPZs, we exclude enterprises located in this province. The corresponding results are displayed in columns (3) and (4) of Table 8. The estimated coefficients of L.PMC maintain a positive trend, signifying that the benchmark regression results exhibit robustness at the sample level.

4.3.3. Excluding Alternative Explanations

In this part, we aim to alleviate the possible influence of other relevant policies and exclude alternative explanatory factors as comprehensively as possible. According to existing literature, both the national innovative city pilot and broadband China pilot policies show a substantial impact on enterprise innovation [58,59]. Hence, we separate the effects of these two model policies from our analysis.

(1)

National Innovative City Pilot policies

During the sample period, the National Development and Reform Commission implemented the National Innovation City pilot to support the construction of enterprises’ innovation infrastructure. We exclude enterprises located in the pilot areas, and the results are listed in columns (5)–(6) of Table 8. By controlling for the impact of this policy, the coefficient of the explanatory variable remains significant at 0.437, affirming the robustness of the result.

(2)

Broadband China pilot policies

In the context of global technological advancements, broadband networks play a crucial role in the new wave of information technology development. Against this background, numerous nations, including China, have prioritized broadband network development as a strategic measure to enhance competitiveness in international economic, technological, and industrial domains. In recent years, the Chinese government has also embarked on implementing broadband China pilot policies, aimed at fostering the coordinated construction of Internet data centers to unleash innovative economic benefits. To ensure the accuracy of our research findings and eliminate any potential interference caused by the implementation of this policy, we exclude enterprises located in the broadband China pilot areas. If the estimation coefficient of L.PMC remains statistically significant, it provides further evidence of the reliability of our benchmark regression results. As evident in columns (7) and (8) of Table 8, the explanatory variable continues to demonstrate a statistically significant positive effect at the 1% level, thereby strengthening the robustness of our research findings.

4.4. Endogeneity Check

To further address endogeneity concerns, we conduct a series of analyses targeting potential sources of bias, including the examination of variable measurement bias, the identification of omitted variables, and the utilization of instrumental variables. The specific analysis process is as follows.

4.4.1. Changing Variable Measurement

In the baseline regression, we employ the number of invention patents and utility model patents as proxies for measuring enterprise innovation. In this section, we re-evaluate the explanatory variables and propose the proportion of invention patent applications as the ideal proxy indicator. As presented in columns (1)–(2) of

Table 9. Endogeneity test results.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	Change Variable Measure		Add Fixed Effect		Add Control Variable	GMM	GMM
Variable	Inva	Inva	EI	EI	EI	EI	EI
L.EI						0.4820 ***	0.5259 ***
						(0.0455)	(0.0487)
L.PMC	0.0111 **	0.0114 **	0.0428 ***	0.0422 ***	0.0435 ***	0.0561 ***	0.0469 ***
	(0.0050)	(0.0051)	(0.0130)	(0.0131)	(0.0132)	(0.0174)	(0.0181)
Urban					−0.0001
					(0.0001)
Indus					0.0292
					(0.0678)
Control	No	Yes	Yes	Yes	Yes	No	Yes
Firm-FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year-FE	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Indus-FE	No	No	Yes	No	No	No	No
City-FE	No	No	No	Yes	No	No	No
Obs.	4284	4284	7179	7179	7179	7379	7379
R-squared	0.5780	0.5806	0.8378	0.8355	0.8373
AR (1)						0.0000	0.0000
AR (2)						0.3610	0.2530

Note: Standard errors in parentheses, ** p < 0.05, *** p < 0.01.

, regardless of the chosen measurement method, the estimated coefficient of L.PMC remains statistically significant, effectively addressing the endogeneity concerns.

4.4.2. Omitted Variable Bias

To mitigate the endogeneity issue stemming from omitted variables, we consider the influence of city-time and industry-time effects. Columns (3)–(4) of Table 9 introduce industry-fixed effects and city-fixed effects, respectively. The coefficient of the explanatory variable remains significant at the 1% level, indicating the robustness of our findings. Given the distinct characteristics of different industries and city-level resource endowments that may influence enterprise innovation levels, it is prudent to incorporate control variables at the city level. Therefore, we include urban population and industrial structure as additional variables in the model, and their impact is shown in column (5) of Table 9. This adjustment is justified by the fact that the population agglomeration can provide enterprises with access to talent resources, generate economies of scale, and facilitate knowledge sharing, ultimately stimulating enterprise innovation.

4.4.3. GMM

To eliminate estimation bias resulting from endogeneity issues, this study adopts the generalized method of moments (GMM), which effectively addresses potential endogeneity and heteroscedasticity concerns among variables. The lagged term of the dependent variable is introduced into the regression model. Columns (6)–(7) of Table 9 demonstrate a positive influence of high-quality policies on enterprises’ innovation levels. Furthermore, the AR (1) value is less than 0.1, indicating the presence of first-order autocorrelation. Conversely, the AR (2) value exceeds 0.1, suggesting the absence of second-order autocorrelation. This validates the appropriateness of instrumental variable selection and enhances the credibility of our findings.

4.5. Heterogeneity

We categorize enterprises into distinct subsamples based on their size and industry classification: large enterprises and small enterprises, high-tech industry enterprises and non-high-tech industry enterprises. The findings presented in columns (1) and (2) of

Table 10. Heterogeneity analysis results.

	(1)	(2)	(3)	(4)
	Large Enterprises	Small Enterprises	High-Tech Industry	Non-High-Tech Industry
Variable	EI	EI	EI	EI
L.PMC	0.0531 ***	0.0277	0.0453 **	0.0352 *
	(0.0177)	(0.0193)	(0.0186)	(0.0181)
Control	YES	YES	YES	YES
Firm-FE	YES	YES	YES	YES
Year-FE	YES	YES	YES	YES
Obs.	3559	3820	4002	3377
R-squared	0.8638	0.7868	0.8101	0.8290

Note: Standard errors in parentheses, * p < 0.1, ** p < 0.05, *** p < 0.01.

imply that, for large enterprises, high-quality digital policies significantly contribute to innovation. However, for small enterprises, although the coefficient of the explanatory variable is positive, it is not statistically significant. The possible reason is the challenges small enterprises encounter in implementing policies that drive innovation, potentially due to limited access to data resources, thereby hindering the transformation impact of digital policies within the NBCPZ [60]. At the industry level, as shown in columns (3)–(4) of Table 10, the estimated coefficient of policy variables for the high-tech industry is positive but not statistically significant. On the contrary, the coefficient for the traditional industry group is significantly positive, suggesting that the quality of the pilot policy has effectively fostered the innovation transformation in such enterprises. A plausible explanation is that high-tech industries generally possess potential inherent technological advantages, enabling them to efficiently utilize data elements, which to some extent diminish the marginal utility of the pilot policies. However, traditional industries tend to be constrained by limited data resources. When the government implements a more comprehensive and detailed policy, it increases the supply of data elements, generating stronger incentives for innovation in these enterprises. This prompts them to adjust their strategies and engage in innovation activities, which is reasonable.

4.6. Further Mechanisms, Channel Examination, and Discussion

The theoretical analysis and hypotheses of this study posit digital policies can enhance the innovation level of enterprises by optimizing the data innovation environment. As reported in columns (1)–(2) of

Table 11. Mechanism effect test results.

	(1)	(2)	(3)	(4)
Variable	Dece	EI	Dipa	EI
L.PMC	0.0074 ***	0.0452 ***	0.0231 **	0.0257 **
	(0.0004)	(0.0142)	(0.0096)	(0.0114)
Dece		0.9144 **
		(0.3963)
Dipa				0.7366 ***
				(0.0233)
Control	Yes	Yes	Yes	Yes
Firm-FE	Yes	Yes	Yes	Yes
Year-FE	Yes	Yes	Yes	Yes
Obs.	7379	7379	7379	7379
R-squared	0.9869	0.8361	0.8335	0.8841

Note: Standard errors in parentheses, ** p < 0.05, *** p < 0.01.

, the inclusion of this variable in the empirical model reveals a positive estimation coefficient for the L.PMC. This suggests that high-quality digital policies foster a conducive environment for data innovation, which is greatly beneficial for the application and sharing of enterprise data technology and information, thus emerging as a positive factor in the innovation level of enterprises. The potential mechanisms can be illuminated from two perspectives. On the one hand, well-designed digital policies provide enterprises with a clear data governance blueprint, emphasizing the importance of data quality, privacy safeguards, and adherence to regulatory norms. This fosters a stable and predictable digital ecosystem where enterprises are motivated to engage in research and development activities, guided by consistent data standards and clear regulatory directives. On the other hand, by eliminating barriers to innovation stemming from ambiguous or disorderly regulations, robust digital policies empower enterprises to advance new technologies and business models. The transparency and coherence of these policies facilitate more effective planning, implementation, and scalability of innovative solutions, alleviating concerns related to regulatory uncertainties.

In terms of the second mechanism, the results are shown in columns (3)–(4) of Table 11. The coefficients of Dipa are significantly positive, indicating that the willingness of enterprises to pursue digital innovation serves as a mediating factor in explaining the beneficial influence of digital policies on enterprise innovation. This result underscores the crucial role of digital policies in propelling the enterprise innovation process. With the rapid advancement and widespread adoption of digital technologies, enterprises gradually recognize the imperative of digital transformation. Digital policies play a crucial role in enhancing awareness and comprehension of digital transformation within enterprises through various channels such as promotion, advocacy, and demonstration. Consequently, they stimulate a surge in the eagerness and momentum among enterprises to embrace digital innovation. In conclusion, the hypotheses H2 and H3 in this article have been validated, confirming the significant impact of digital policies on fostering enterprise innovation.

5. Conclusions and Enlightenment

5.1. Conclusions

This article modifies the indicators of the existing PMC index model to evaluate the quality of digital policies and examines its impact of digital policy quality on enterprise innovation, using the digital strategic domain of China’s NBCPZ as a case study. The study indicates that the overall PMC index value of digital policies is relatively low, characterized by significant temporal fluctuations and spatial disparities. The benchmark results indicate a positive correlation between higher-quality digital policies and elevated levels of the enterprises’ innovation.

To ensure the reliability of the research findings, this article incorporates various tests at both the methodological and sample size levels, addressing potential endogeneity issues caused by variable measurements, omitted variables, and instrumental variables. Based on the inherent signaling effect of policy texts, we identify two intermediary variables: the data innovation environment and enterprise digital innovation willingness, which explain the relationship between digital policy quality and enterprise innovation. In addition, heterogeneity analysis indicates that the micro-level innovation effect of digital policies is influenced by the enterprise size and industry, particularly benefiting large-scale enterprises and those in high-tech industries.

5.2. Theoretical Contribution

This study makes a significant contribution to the theoretical comprehension of the intricate interplay between digital policy quality and enterprise innovation, particularly in the context of China’s national big data strategy. Two primary contributions emerge from this investigation.

Firstly, while many previous studies have recognized the important role of policies, such as digital economy pilots, in stimulating enterprise innovation, they have not sufficiently explored the systematic evaluation of typical policies. The study highlights that high-quality digital policies play a crucial role in fostering enterprise innovation. This revelation not only expands the array of factors influencing innovation within enterprises but also enriches the theoretical framework about the confluence of policy and innovation studies. It underscores the importance of a favorable policy environment for fostering sustainable innovation. Secondly, this study explores the signaling effects embedded in policy texts, elucidating the mechanisms that underlie the positive influence of digital policy quality on enterprises’ innovation. Specifically, it identifies two critical paths: the regional data innovation environment and enterprise digital innovation willingness. That clarifies the pathway by which these signals resonate within the enterprises, subsequently impacting innovation outcomes. This insight sheds light on the relationship between policy making and subsequent behavioral responses of enterprises in the innovation domain, thereby contributing to the discussion on sustainable innovation practices.

5.3. Practical Implications

This study offers significant insights for policy formulation and practical execution, particularly for the Chinese government in improving its digital policies. Firstly, it highlights the urgency of developing explicit and effective policy documents. This emphasizes the necessity for governments to prioritize enhancing the quality of digital policies, which may involve a comprehensive review and refinement process. Ensuring consistency and clarity in policy texts, minimizing uncertainty, and furnishing enterprises with stable and predictable guidance on digital progress are fundamental in fostering an environment conducive to innovation. Secondly, incorporating flexibility into digital policies represents a wise approach. On the one hand, it is crucial to establish a system of periodic evaluation of these policies. Regular assessments facilitate the identification of areas for improvement, enable tracking of policy effectiveness, and address emerging challenges in the rapidly evolving digital environment. On the other hand, policymakers should tailor digital policies to accommodate the distinct requirements of different industries, particularly those of large-scale and high-tech industries, which are poised to reap significant benefits from well-designed digital policies. Moreover, the integration of signaling theory and the identification of mediating factors enhance the practical implications of this study. It is crucial for policymakers to assess the data-factor innovation environment and devise initiatives that promote and support data-factor innovation. These initiatives may include establishing collaborative platforms to facilitate research and development in data-related technologies, as well as promoting the adoption of advanced data management practices. Additionally, recognizing the importance of enterprises’ willingness to innovate digitally, policies should strive to cultivate a culture that encourages and supports digital innovation. Governments can incentive this behavior through tax breaks, financial support, and innovation rewards, thus encouraging enterprises to actively engage in digital innovation activities.

5.4. Research Limitations and Further Breakthroughs

This study still has certain limitations. Due to constraints in data availability, this article is confined to Chinese sample data. Consequently, when generalizing the findings to other regions or industries, one must exercise caution, as contextual factors specific to each region may influence the relationship between digital policies and innovation in distinct manners. Care must be taken when generalizing the results to other regions or industries, as different regions may have unique contextual factors that could influence the relationship between digital policies and innovation in distinct ways. To address this limitation, future research could conduct comparative studies on an international scale, taking into account regional disparities and industry-specific patterns. Such an approach would contribute to a more comprehensive understanding of the broader innovation effects of digital policy quality across diverse contexts.