Digital infrastructure construction and the development of new-quality productive forces in enterprises

Abstract

Digital infrastructure is the foundation for the development of digital technologies, providing the necessary production materials for the development of the digital economy and new types of productivity. This study aims to explore the impact of digital infrastructure construction on the new types of productivity of enterprises. Using the “Broadband China” demonstration cities as a quasi-natural experiment, empirical analysis is conducted based on data from A-share listed companies between 2011 and 2023, employing a difference-in-differences approach with multiple time points. The study finds that the “Broadband China” demonstration city pilot significantly promoted the improvement of enterprise new-type productivity, and this conclusion remains robust after several tests. Further analysis of the mechanism reveals that digital infrastructure promotes the development of new-type productivity by accelerating the smart transformation of enterprises, enhancing their level of informatization, and boosting innovation capabilities. Heterogeneity analysis shows that the policy has a more significant effect on improving the new-type productivity of enterprises in sub-provincial cities, private enterprises, and growth-stage enterprises. This research not only provides empirical evidence for the empowerment of enterprise new-type productivity development through digital infrastructure but also enriches the research on how macro policies influence micro-level enterprises. Furthermore, it provides theoretical references and practical guidance for the development of the digital economy and the acceleration of digital technology advancements in other Chinese cities and many developing countries.

Introduction

In the era of the digital economy, information infrastructure has become a strategic resource driving high-quality economic development. Since China launched the “Broadband China” initiative in 2013, network construction centered around demonstration cities has accelerated nationwide. This initiative has not only built high-speed, ubiquitous communication networks but also catalyzed profound transformations in industrial digitalization. As a crucial tool for national informatization, the construction of “Broadband China” demonstration cities, through projects like fiber optic network coverage, 5G base station deployment, and data center expansion, has significantly enhanced urban digitalization capabilities, creating an interconnected, low-latency, and high-reliability information environment for enterprises. This leap in infrastructure development has provided key support for the cultivation of new-type productivity in enterprises. New-type productivity, as an advanced form of productivity driven primarily by technological innovation, hinges on the deep integration of technological, data, and innovation elements. Against this backdrop, it is crucial to investigate how the construction of “Broadband China” demonstration cities promotes the enhancement of new-type productivity in enterprises by reshaping their technological ecosystems, optimizing factor allocation efficiency, and stimulating innovation vitality.

Through the innovative mechanism of a “proactive government + targeted pilot programs,” the Chinese government has provided a demonstrative solution to the informatization challenges faced by developing countries. Under the framework of the “Broadband China” strategy, the government has leveraged policy wisdom to dynamically balance top-level design with grassroots innovation, establishing a full-cycle mechanism characterized by “goal orientation – competitive pilot programs – performance validation – iterative promotion.” In the pilot projects, the government’s role went beyond traditional subsidy-based approaches, constructing a multi-level investment system in which central guiding funds serve as leverage, commercial capital as the main driver, and technical standards as the baseline requirement. The funding for the “Broadband China” pilot policies was sourced through a mix of fiscal appropriations, special funds, and investments from telecom operators. This diversified financing structure eased the fiscal burden on the government while stimulating market investment enthusiasm. The innovative value of this model lies in its creation of a “policy trial-and-error space” for developing countries. By establishing institutional buffer zones around demonstration areas, the model enables a focused resolution of key bottlenecks such as funding gaps, technological adaptation, and institutional frictions, while avoiding the systemic risks associated with large-scale rollouts. This three-stage evolutionary path—“controlled experimentation – value validation – model replication”—offers a replicable policy template for resource-constrained nations. It highlights the critical role of a proactive government in the strategic allocation of resources and demonstrates how gradient development can maximize the value of limited resources. Our research addresses the long-standing neglect of the government’s role in studies of capitalist market economies. It reveals the leading role of the state in driving economic development and expands the theoretical boundaries, thereby deepening our understanding of the mechanisms through which the construction of “Broadband China” demonstration cities contributes to the enhancement of enterprises’ new quality productive forces.

Research on the impact of “Broadband China” demonstration cities on the economy and society can be divided into two main categories. One category is macro-level impacts, such as accelerating economic growth¹ improving urban innovation levels^2,3 enhancing green innovation^4,5 stimulating urban entrepreneurial activity^6,7 reducing carbon emissions⁸ accelerating knowledge diffusion between cities⁹ increasing employment^10,11,12 and promoting social equity¹³. The other category is micro-level research focusing on enterprises, with empirical studies indicating that “Broadband China” demonstration cities improve enterprise innovation capability¹⁴ green innovation capacity¹⁵ reduce economic uncertainty at the enterprise level¹⁶ boost productivity¹⁷ curb financialization¹⁸ accelerate digital transformation¹⁹ and increase the concentration of companies²⁰. At present, research on new-type productivity is mainly theoretical, focusing on its definition²¹ generation logic²²and formation pathways²³. A few empirical studies address the measurement of new-type productivity and its impact on provincial industrial structures or green innovation^24,25. Digital infrastructure, as a foundational condition for the development of the digital economy and AI technologies, is intricately linked to the development of new-type productivity.

Existing studies show that information infrastructure construction has transcended the traditional linear model of production factors, moving towards exponential growth driven by network effects and data empowerment. However, most of the existing literature focuses on the impact of broadband networks on macroeconomic indicators, lacking a deep analysis of the mechanisms by which digital infrastructure affects the formation of new-type productivity at the enterprise level. Particularly under the policy framework of “Broadband China” demonstration cities, how infrastructure upgrades influence enterprise productivity systems through smart transformation, hardware/software investment, and innovation capability restructuring has yet to form a systematic research framework.

The scientific question this paper aims to address is: How does the construction of “Broadband China” demonstration cities promote the improvement of enterprise new-type productivity by reshaping technological ecosystems, optimizing factor allocation efficiency, and stimulating innovation vitality? Based on technological innovation theory and the resource-based view, we construct an analytical path of “infrastructure—technological penetration—capability reconstruction—productivity leap,” and empirically test the effects and mechanisms of demonstration city construction on enterprise new-type productivity using micro-data from listed companies between 2011 and 2023. We particularly focus on the synergy between software iteration and hardware upgrades in the intelligent transformation of enterprises, as well as the process of innovative integration of data elements with traditional production factors. We aim to reveal the internal logic of how new information infrastructure empowers high-quality enterprise development and provide new theoretical explanations and policy insights for the productivity transformation in the digital economy era. This paper uses “Broadband China” demonstration cities as a quasi-natural experiment to study how digital infrastructure construction affects the development of enterprise new-type productivity. It fills research gaps in the field and provides theoretical and practical evidence for the development of digital economies and the promotion of new-type productivity in other Chinese cities and many developing countries.

The marginal contributions of this paper are reflected in the following two aspects: (1) From the perspective of digital infrastructure construction, this paper explores how the construction of “Broadband China” demonstration cities impacts the development of enterprise new-type productivity, providing empirical evidence on how digital infrastructure empowers enterprise new-type productivity development. (2) From the perspectives of intelligent transformation, increased informatization levels, and enhanced innovation capabilities, the paper examines how digital infrastructure construction affects the development of enterprise new-type productivity, broadening the mechanisms through which digital infrastructure influences enterprise productivity development. This enriches the study of how macro policies impact micro enterprises and offers valuable insights for other regions and developing countries to accelerate the development of enterprise new-type productivity.

Policy background and theoretical analysis

Policy background

The “Broadband China” demonstration cities are an important measure in advancing China’s national informatization strategy. The “Broadband China” policy aims to accelerate broadband network construction, improve information and communication technology levels, and promote the development of the digital economy. After the release of the “Broadband China” strategy and implementation plan in 2013, the Ministry of Industry and Information Technology (MIIT), in collaboration with the National Development and Reform Commission (NDRC), selected 120 demonstration cities in three batches from 2014 to 2016. The creation period for these demonstration cities (or city clusters) is set at three years, with active promotion from the national telecommunications industry and relevant local government departments. The goal is to leverage broadband networks to shift China’s economic development model and foster the growth of digital industries and other advanced technological sectors.

The “Broadband China” policy has facilitated the rapid and healthy development of China’s broadband infrastructure, strengthening fiber optic networks and mobile communication networks while improving broadband access speeds. This has provided a solid foundation for enterprise intelligent transformation and increased the likelihood of enterprises achieving smart production and management. On one hand, high-speed, stable broadband networks allow enterprises to transmit data and share information more efficiently, contributing to the automation and intelligence of production processes, thereby improving productivity and product quality. On the other hand, robust network infrastructure has created the conditions for enterprises to adopt emerging technologies such as the Internet of Things (IoT), big data, and artificial intelligence (AI), enhancing the level of informatization within enterprises and driving innovation development.

Theoretical analysis

The core of new quality productive forces lies in achieving a leap in total factor productivity through revolutionary technological breakthroughs and innovative configurations of production factors²⁶. As a national strategic infrastructure initiative, the “Broadband China” policy has systematically promoted the enhancement of enterprises new quality productive forces by reducing information costs, optimizing resource allocation, and accelerating technology diffusion. These mechanisms have facilitated enterprise-level intelligent transformation, improvements in informatization, and strengthened innovation capacity. This paper conducts a theoretical analysis based on three primary mechanism channels, drawing upon economic frameworks such as endogenous growth theory, transaction cost theory, and innovation diffusion theory.

Accelerating enterprise smart transformation

The essence of smart transformation is to deeply integrate data elements with artificial intelligence technologies into production processes, reshaping the production function through “data-driven + algorithm optimization.” The “Broadband China” policy promotes this process through the following channels:

According to endogenous growth theory, economic growth is driven by endogenous technological progress, and the widespread adoption of general-purpose technologies plays a critical role in both economic growth and enterprise development. The public good nature of infrastructure lowers the threshold for enterprises to adopt intelligent technologies. Broadband networks, characterized by non-rivalry and positive externalities, enable enterprises to access intelligent technologies at a marginal cost approaching zero due to their high-speed and wide-coverage features.

From the perspective of endogenous growth theory, the diffusion of general-purpose technologies—such as 5G and the industrial internet—can generate knowledge spillover effects, allowing small and medium-sized enterprises (SMEs) to share the benefits of technological advancement without bearing high R&D costs. Enterprises can more easily adopt emerging technologies such as artificial intelligence and big data to achieve intelligent transformation of production processes²⁷. With AI technologies, firms can implement automated production, intelligent quality inspection, and smart logistics management, thereby improving production efficiency and product quality while reducing costs. Remote equipment control supported by low-latency networks enables SMEs to achieve “cloud-based production,” overcoming physical space constraints and accelerating intelligent transformation to enhance the level of new quality productive forces.

From the perspective of information economics, information plays a vital role in economic decision-making. The flow of data elements optimizes the efficiency of intelligent decision-making. As a “data highway,” the broadband network facilitates the shift from experience-driven to data-driven operations. According to information economics theory, the real-time data analysis capabilities of intelligent systems can reduce information asymmetry in production processes and improve decision accuracy in areas such as inventory management and quality control. Enterprises can leverage these data to conduct precise market analysis and demand forecasting, thereby optimizing product design, production planning, and marketing strategies, ultimately enhancing market competitiveness²⁸. Manufacturing enterprises can integrate supply chain data via industrial internet platforms to improve the accuracy of equipment failure predictions and reduce downtime losses²⁹. Referring to the interventionist logic of New Keynesian economics³⁰ the “Broadband China” policy—driven by relevant government departments—has facilitated the construction of advanced digital infrastructure for enterprises. This infrastructure helps firms better utilize and unlock the value of data elements, strengthens their capabilities in data collection, analysis, and application, and accelerates the development of their new quality productive forces.

Enhancing enterprise informatization level

From the perspective of transaction cost theory, the purpose of an enterprise is to minimize transaction costs. The widespread adoption of fiber-optic and mobile communication networks has brought about unprecedented changes for enterprises. Due to the low cost and broad coverage of these networks, enterprises can collaborate across regions at near-zero cost. The core of enhancing informatization lies in using digital tools to compress both internal and external transaction costs, thus reconstructing the value creation network. According to Coase’s theorem, when the cost of information transmission approaches zero, enterprises can replace certain hierarchical organizational structures with digital platforms, reducing internal agency costs. Broadband networks facilitate information sharing and communication both within different departments of an enterprise and between the enterprise and its external partners³¹. Enterprises can establish informatization management systems to optimize internal processes and facilitate collaboration, thereby improving management efficiency. With the support of broadband networks, enterprises can realize remote work and collaboration, breaking geographical barriers and improving the utilization efficiency of human resources. Employees can collaborate remotely via the internet to complete project tasks, enhancing work efficiency and innovation.

General equilibrium theory emphasizes the importance of market clearing for the efficient allocation of resources. With the development of cloud computing and big data technologies—supported by broadband networks—enterprises are empowered with powerful capabilities to match supply and demand information in real time. This significantly accelerates the pace of market clearing and effectively reduces welfare losses caused by resource misallocation. Through broadband networks, enterprises can engage in e-commerce activities, expand market channels, and reduce marketing costs³². At the same time, they can implement informatized supply chain management, enhancing the responsiveness and flexibility of the supply chain, reducing inventory costs, and improving operational efficiency³³. The use of broadband networks enhances enterprises’ level of informatization, lowers both internal and external transaction costs, and improves resource utilization efficiency. In order to better apply new technologies, enterprises are incentivized to increase their exploration and adoption of such technologies, thereby enhancing their new quality productive forces.

Enhancing enterprise innovation capability

The improvement of innovation capabilities depends on a virtuous cycle of knowledge production, dissemination, and application. The “Broadband China” policy constructs an innovation ecosystem through the following mechanisms:

According to endogenous innovation theory, the accumulation of knowledge capital exhibits nonlinear characteristics, and cross-organizational collaboration can trigger synergistic effects of “1 + 1 > 2.” Collaborative R&D platforms supported by high-speed networks (such as open-source communities and virtual laboratories) break down innovation silos. Broadband networks allow enterprises to more easily access global innovation resources, including technological information, market dynamics, and talent resources. Enterprises can collaborate with domestic and international research institutions and companies to jointly carry out innovative activities, thereby enhancing innovation capabilities³⁴.

According to Schumpeter’s innovation theory, data mining has given rise to new business models such as subscription services and the sharing economy³⁵. The development of broadband networks has driven a transformation in enterprise innovation models, such as crowdsourcing and open innovation. Enterprises can publish innovation needs on network platforms to attract innovators from around the world, pooling ideas and improving innovation efficiency and quality. The user behavior data accumulated by broadband networks has become a new factor of production.

Evolutionary economics’ trial-and-error learning model emphasizes that in the innovation process, enterprises need to continually experiment and explore, and lowering trial-and-error costs is critical to improving innovation efficiency. The emergence of advanced technologies such as virtual reality (VR) and digital twins provides powerful digital learning tools for enterprises. Digital tools transform physical experiments into virtual simulations, reducing the cost of each innovation and significantly improving the R&D return on investment³⁶. This enables enterprises to iteratively improve products at a low cost. The application of broadband networks has enhanced both internal and external collaborative innovation within enterprises. The adoption of new technologies helps reduce innovation costs, improve innovation efficiency, and strengthen innovation capabilities, thereby contributing to the improvement of enterprises’ new quality productive forces.

Synergistic effect analysis of mechanisms

Resource-based theory suggests that an enterprise’s competitive advantage arises from its unique resources and capabilities. Through informatization construction, enterprises can efficiently collect, store, and analyze data, allowing them to more accurately grasp market demand and industry trends, thus providing abundant data resources and strong decision-making support for innovation. In terms of smart transformation, enterprises leverage the data and technologies accumulated through informatization to drive comprehensive smart transformations in production processes and management models. Innovation improvements, in turn, further enhance informatization and smart transformation. The new technologies and methods generated during innovation processes drive the improvement of informatization levels. For example, new data analysis algorithms or AI technologies may be integrated into an enterprise’s information systems to improve the speed and accuracy of information processing. Innovation also drives enterprises to adopt more advanced smart equipment and management concepts, accelerating the smart transformation process. The interplay of smart transformation, informatization, and innovation, based on a resource-based view, creates a dynamic, mutually reinforcing relationship, collectively fostering the development of new-type productivity in enterprises.

In conclusion, the “Broadband China” policy systematically enhances enterprise smart transformation, informatization, and innovation capabilities through three mechanisms: technological embedding, cost optimization, and knowledge spillovers. The economic essence of this policy lies in reducing information friction, expanding the boundaries of technological diffusion, and reconstructing the composition of production factors, achieving a dynamic balance between economies of scale and scope. Based on the above theoretical analysis, the following hypotheses are proposed:

H0: The “Broadband China” demonstration city pilot can promote the improvement of enterprise new-type productivity levels.

H1: The “Broadband China” demonstration city pilot can enhance new-type productivity levels by improving enterprise intelligence.

H2: The “Broadband China” demonstration city pilot can promote new-type productivity improvement by enhancing enterprise informatization levels.

H3: The “Broadband China” demonstration city pilot can drive the development of enterprise new-type productivity by enhancing innovation capabilities.

H4: The “Broadband China” pilot cities can promote the improvement of enterprise new quality productivity through the synergistic effects among the three influencing mechanisms: enterprise intelligence, informatization, and innovation capability.

Empirical tests

Variables and model

To empirically examine the impact of the “Broadband China” demonstration city pilot on enterprises’ new quality productivity, we construct a multi-period difference-in-differences (DID) model based on the staggered approval timelines of the pilot cities. The baseline model is specified as follows:

$$NQ{P_{it}}={\beta _0}+{\beta _1}di{d_{it}}+{\beta _2}Controls+{\mu _i}+{\lambda _t}+{\varepsilon _{it}}$$

(1)

In the model, $NQ{P_{it}}$ is the dependent variable, representing the new quality productivity level of enterprise i in year t. $di{d_{it}}$ is the core explanatory variable: it is assigned a value of 1 if the city where enterprise i is located is selected as a “Broadband China” demonstration city in year t, and 0 otherwise. ${\beta _1}$ is the estimated coefficient of the core explanatory variable. If ${\beta _1}$ is significantly positive, it indicates that the “Broadband China” demonstration city pilot has a positive promoting effect on enterprises’ new quality productivity, and vice versa. If ${\beta _1}$ is insignificant, it suggests no statistically significant causal relationship between the pilot policy and enterprises’ new quality productivity. Thus, ${\beta _1}$ is the focus of this study. $Controls$ represents control variables, ${\mu _i}$ denotes firm fixed effects, ${\lambda _t}$ denotes year fixed effects, and ${\varepsilon _{it}}$ is the error term. All regressions employ robust standard errors clustered at the firm level. See Table 1 for the definitions and calculation methods of all variables.

Dependent Variable: New-Type Productivity. New-type productivity refers to the advanced form of contemporary productivity driven by technological breakthroughs, innovative allocation of production factors, and deep industrial transformation and upgrading. It is fundamentally characterized by the qualitative change in the combination of laborers, labor materials, and labor objects, with an emphasis on the enhancement of total factor productivity (TFP) as its core indicator. Based on the third and fourth technological and industrial revolutions, new-type productivity is anchored on key enhancement points such as informatization, networking, digitalization, intelligence, automation, greening, and efficiency improvement. Referring to the methodology for measuring new-type productivity in enterprises proposed by Song et al.³⁷ this study constructs an indicator system for new-type productivity based on the two-factor productivity theory³⁸. The primary indicators include labor force and production tools. For the labor force indicator, annual reports from enterprises are used to extract data on R&D personnel and manufacturing expenses. For the production tools indicator, annual reports are used to extract data related to R&D expenses and assets. The entropy method is employed to calculate the weights of each indicator, forming the data set for enterprise new-type productivity.

Control Variables: According to the resource-based theory, the resources a firm possesses are key sources of its competitive advantage and performance. Firms with different types and levels of resources will allocate varying amounts of resources to activities related to new quality productive forces, such as technological R&D and equipment upgrading, thereby providing differing material foundations for the enhancement of new quality productive forces. Regarding firm size, we use the logarithm of total assets (size) as a proxy variable. Total assets comprehensively reflect the amount of economic resources under a firm’s control and are a widely used and important indicator of firm size. Compared to other size indicators, such as operating revenue, total assets better represent the firm’s accumulated resource strength over the long term. It includes not only current operational outcomes but also fixed assets such as land, plants, and equipment, as well as intangible assets. These resources are closely related to activities that contribute to the improvement of new quality productive forces. For example, firms with more fixed assets are better positioned to undertake equipment upgrades and technological transformation, thereby providing material support for enhancing new quality productive forces.

Meanwhile, firm lifecycle theory suggests that firms at different stages of development vary in terms of strategic decision-making, resource allocation, and innovation capabilities. Based on these relevant theories and literature, we select a set of control variables. In addition to the logarithm of total assets (size), we include: Operating revenue (sale): This reflects the firm’s performance over a given period. Higher revenues suggest greater financial capacity to invest in activities related to new quality productive forces; Cash flow (cflow): This indicates the firm’s liquidity and financial health. Adequate cash flow ensures sufficient funding for technological R&D and other innovation-related investments; Firm age (age): This represents the development stage of the firm in the market. Firms at different stages may pursue and realize new quality productive forces in different ways; Revenue growth rate (growth): This reflects the firm’s growth momentum. Fast-growing firms are often more motivated and capable of enhancing new quality productive forces; Ownership concentration (top10): This affects the firm’s decision-making mechanisms and governance efficiency, which in turn influence the development of new quality productive forces; Return on assets (roa): This measures the firm’s ability to generate earnings from its assets. A higher ROA indicates more efficient resource utilization, which positively contributes to new quality productive force enhancement; Leverage ratio (lev): This reflects the firm’s debt burden and financial risk. A reasonable level of leverage supports stable development and creates a favorable financial environment for the improvement of new quality productive forces.

Table 1 Definition and calculation of main variables.

This study selects data from Chinese A-share listed companies spanning 2011–2023 (13 years), comprising 29,290 initial observations. After excluding firms in the financial industry, ST/*ST companies, samples with missing key variables, and those with insufficient time-series data, the final dataset is derived from the China Stock Market & Accounting Research Database (CSMAR) and Wind Economic Database (WIND). To mitigate the influence of outliers, all continuous variables are winsorized at the 1% level. Table 2 reports the descriptive statistics of key variables. The New Quality Productivity (NQP) ranges from 0.004 to 2.838, indicating significant heterogeneity in productivity levels across firms, which provides substantial analytical potential for further research.

Table 2 Descriptive statistics of main variables.

Parallel trends and dynamic effects test

The parallel trends test is a key assumption in the Difference-in-Differences (DID) methodology, used to verify whether the trends of the treatment group and the control group were consistent prior to the policy intervention. If the trends are parallel, the results of the policy effect evaluation are more credible. The parallel trends assumption requires that the treatment and control groups have similar development trends before the policy is implemented. This is the most crucial and fundamental assumption of DID. To perform the parallel trends test, data from five periods prior to the policy event are classified as the fifth period. The data from the five periods before and the nine periods after the policy implementation are selected, with the policy implementation year as the baseline. The parallel trends test is then conducted. Figure 1 illustrates the parallel trends test. Prior to the policy intervention, the estimated coefficients are not statistically significant, indicating that there were no systematic significant differences in the development of new-type productivity between the treatment and control groups, confirming the parallel trends assumption. The dynamic model is set as follows:

$$NQ{P_{it}}={\beta _0}+\mathop \sum \limits_{{t= - 5}}^{{t=9}} {\beta _1}polic{y_{it}}+\sum {\beta _2}Controls+\sum {\mu _i}+\sum {\lambda _t}+{\varepsilon _{it}}$$

(2)

Among them, $polic{y_{it}}$ is a set of dummy variables, and $\beta$ is the coefficient of primary interest. It is used to assess whether there is a significant difference in trend changes between the treatment group and the control group. The other variables are consistent with those in Model (1).Table 3 presents the regression results for the dynamic effects test of the policy. From Table 3, it can be seen that in the five periods prior to policy implementation, the coefficients of the core explanatory variables are not significant. However, in the first to sixth periods after the policy implementation, the coefficients of the core explanatory variables are significantly positive, suggesting that the digital infrastructure development effectively promoted the growth of enterprise new-type productivity during the 1–6 years following the policy implementation. After the seventh year and beyond, the positive impact of digital infrastructure development gradually diminishes. By the ninth year after the policy implementation, digital infrastructure development shows a negative effect on enterprise new-type productivity, indicating that the positive impact brought by digital infrastructure is not permanent and needs to be continuously updated according to technological, industrial, and economic developments.

Fig. 1

Results of parallel trend test.

Table 3 Dynamic effect test.

Benchmark regression

Table 4 reports the results of the benchmark regression. Column 1 presents the direct regression results for the explanatory and dependent variables. Columns 2 and 3 show the regression results with fixed effects and control variables included, respectively. Column 4 displays the regression results with both fixed effects and control variables included. From the regression results, it is evident that, both before and after including fixed effects and control variables, the DID coefficient is significantly positive at the 1% level. This indicates that the “Broadband China” pilot cities have a significant positive effect on the new-type productivity of enterprises, supporting the hypothesis H0. The primary reason for this result is that the “Broadband China” pilot cities not only accelerate the intelligent transformation of enterprises by increasing investments in smart-related hardware and software, but also enhance the enthusiasm for innovation within enterprises, thereby promoting the improvement of enterprise new-type productivity.

Table 4 Benchmark regression results.

Robustness check

Endogeneity test

The selection of “Broadband China” pilot cities is not entirely random, which may introduce endogeneity issues and cause bias in the benchmark regression results. To address this, we use the instrumental variable (IV) method to mitigate the endogeneity problem. We choose the topography of each city, specifically the degree of terrain variation, as an instrumental variable. On one hand, the degree of terrain variation is shaped by natural geographical factors and is generally fixed over time during the study period. It is unaffected by the “Broadband China” policy or the economic and social outcome variables being studied, satisfying the exogeneity requirement. On the other hand, the degree of terrain variation significantly influences the construction costs and difficulties of broadband networks. In cities with more varied terrains, such as mountainous areas, laying broadband networks faces more geographical obstacles, leading to higher construction costs and greater difficulty. This limits the speed and coverage of broadband network deployment, resulting in relatively fewer broadband connections. In contrast, in flat cities, broadband network construction is easier and cheaper, facilitating rapid broadband expansion and higher access rates, thereby meeting the relevance requirement. Therefore, the degree of terrain variation is closely related to the broadband access situation in cities. Additionally, we use the number of broadband accesses before the policy implementation as another instrumental variable. As a historical data point, the number of broadband accesses before the policy implementation is not directly influenced by the outcome variables after policy implementation, fulfilling the exogeneity requirement. Furthermore, the pre-policy broadband access number reflects the city’s existing network infrastructure. Governments may increase investments in areas with better infrastructure (or prioritize filling in gaps), making it correlated with the policy implementation intensity, thus satisfying the relevance requirement.

We use the interaction terms of the city’s topography variation and DID, and the interaction terms of city broadband access number and DID as instrumental variables (IV). The regression results are shown in Table 5. Columns 1 and 2 present the two-stage regression results using city topography variation as the instrumental variable, while columns 3 and 4 show the two-stage regression results using city broadband access numbers as the instrumental variable. In columns 1 and 3, the regression results are significantly positive, with F-statistics of 499.53 and 830.71, respectively, indicating a strong correlation between the instrumental variables and the independent variables, and there is no weak instrument problem. In columns 2 and 4, the regression results remain significantly positive, showing that after controlling for potential endogeneity issues, the benchmark regression results remain robust.

Table 5 Endogeneity test regression results.

Placebo test

Although the benchmark regression model includes control variables related to the development of enterprise new quality productivity and accounts for fixed individual and time effects to mitigate biases from omitted variables, some unobserved and uncontrollable factors may still affect the development of enterprise new quality productivity. To further validate the robustness of the benchmark regression results, we adopt a placebo test, following the approaches used by La Ferrara et al.³⁹, Liu and Lu⁴⁰ and others. In this test, we randomly select a sample to create a pseudo-treatment group, while the remaining samples form a pseudo-control group. We then randomly generate the policy implementation time and perform DID regression on the sample. This process is repeated 500 times with random sampling and DID regressions. The results are shown in Fig. 2. The horizontal axis represents the distribution of DID regression coefficients from 500 simulations, while the vertical axis shows the p-values for each regression coefficient. The black fine line indicates the regression coefficients, and the blue dots represent the p-values. The vertical red dashed line on the right marks the benchmark regression coefficient.

From Fig. 2, we observe that the DID regression coefficients from the 500 random simulations are mainly centered around zero, showing a significant difference from the benchmark regression coefficient of 0.0145. Furthermore, most p-values are distributed above the red line at y = 0.1, meaning they are not statistically significant at the 10% level. This suggests that the majority of the random regression coefficients are not significant. The results of this placebo test indicate that the benchmark regression results are not influenced by other potential unobserved variables, thus confirming the robustness of the benchmark results.

Fig. 2

Placebo test.

PSM-DID and entropy balancing method

When constructing the DID model, it is essential that the characteristics of the treatment and control groups are as similar as possible to obtain robust and reliable regression results. To further validate the reliability of the benchmark regression results, we combine propensity score matching (PSM) and DID. First, we use logistic regression to obtain the propensity scores for each sample based on the control variables. Then, we apply the nearest-neighbor matching method to match the treatment group and the control group on a 1:1 basis. After matching, we perform DID regression on the matched sample, and the regression results are shown in Table 6. The PSM-DID coefficient is 0.0137, which is statistically significant at the 10% level. The positive coefficient is consistent with the benchmark regression result, confirming that the benchmark regression results are robust.

Additionally, we use the entropy balancing method for matching. The entropy balancing method adjusts for the first, second, and third moments (variance, mean, and skewness) of the covariates. In selecting the optimal weights, we include not only the covariates but also their quadratic, cubic, and interaction terms to achieve more precise matching between the treatment and control groups. The regression results for the matched sample are shown in column (2) of Table 6. The core explanatory variable coefficient is 0.0098, which is statistically significant at the 10% level. This result is also consistent with the benchmark regression, further verifying the robustness of the benchmark results.

Changing fixed effects

The “Broadband China” pilot cities represent an exogenous policy shock, but the effects of the policy may be influenced by changes in regional economic development levels and industry development trends. To control for differences in regional economic development and industry development imbalances, we re-perform the regression while controlling for time fixed effects, regional fixed effects, and industry fixed effects. The results in column (3) of Table 6 show that after controlling for regional and industry effects, the coefficient of the core explanatory variable remains statistically significant at the 10% level. This indicates that the benchmark regression results are not affected by regional and industry-level factors.

Excluding other policy interference

The smart city pilot program is an important initiative in China to promote urban modernization. It encourages cities to adopt information technology to enhance the intelligence and automation of urban infrastructure and management, improving urban efficiency and sustainability. To examine whether the development of enterprise new quality productivity is influenced by the smart city pilot program, we use a multiple-period DID method with policy dummy variables for the smart city pilot program, which was implemented in 2012, 2013, and 2014. The empirical regression results are shown in column (4) of Table 6. The coefficient for the smart city pilot program dummy variable is not statistically significant, suggesting that the smart city pilot program does not significantly affect the development of enterprise new quality productivity.

In 2015, the Chinese government released the “Vision and Actions on Jointly Building the Silk Road Economic Belt and the 21st-Century Maritime Silk Road.” The Belt and Road Initiative (BRI) entered a substantive construction phase. As of June 2023, China has signed cooperation documents with over 180 countries and international organizations. The BRI covers 18 provinces, autonomous regions, and municipalities in China and has had a broad and profound impact on China’s economic and social development. The BRI has had a positive effect on the high-quality development of cities along the route, as well as on the international expansion of Chinese enterprises. To assess the impact of the BRI on the results, we construct a policy dummy variable for the BRI with 2015 as the implementation year and replace the dummy variable in the benchmark regression model with the BRI variable. The empirical regression results are shown in column (5) of Table 6. The coefficient for the BRI dummy variable is not statistically significant, further confirming that the BRI does not affect the benchmark regression results and reinforcing the reliability of the benchmark regression findings.

Table 6 Robustness test results.

Mechanism testing

The above analysis argues that the construction of “Broadband China” demonstration cities can significantly promote the improvement of the new quality productivity level of enterprises, and this part will verify the impact mechanism of the construction of “Broadband China” demonstration cities on the new quality productivity of enterprises from three aspects: intelligent transformation, informatization, and innovation ability through the construction mechanism test model. The regression model is constructed to test the mechanism as follows:

$$NQ{P_{it}}={\beta _0}+{\beta _1}di{d_{it}}*R{V_{it}}+{\beta _2}di{d_{it}}+{\beta _3}R{V_{it}}+{\beta _4}Controls+{\mu _i}+{\lambda _t}+{\varepsilon _{it}}$$

(3)

where $NQ{P_{it}}$ indicates the level of new qualitative productivity of the ith enterprise in year t. $di{d_{it}}$ is the policy dummy variable, and $R{V_{it}}$ is the moderating variable, which respectively represent enterprise-level intelligent transformation, informatization, and innovation capability in the regression model. The interaction term $di{d_{it}}*R{V_{it}}$ is the core explanatory variable. This interaction helps assess whether a specific mechanism operates under different conditions. If the interaction term is statistically significant, it suggests that the moderating variable influences the relationship between the policy intervention (dummy variable) and the dependent variable, indicating that the moderating variable may be an important mechanism through which the policy exerts its effect. Therefore, ${\beta _1}$ is the estimation coefficient of the core explanatory variable, if ${\beta _1}$ is significantly positive, it means that the moderating variable has a promoting effect on the improvement of the new quality productivity of enterprises in the “Broadband China” demonstration city, otherwise, it means that the moderating variable has a negative effect in it, if ${\beta _1}$ is not significant, it means that the impact of the pilot policy of the “Broadband China” demonstration city on the new quality productivity of the moderating variable has no statistical causal relationship, therefore, ${\beta _1}$ is the core of attention. $Controls$ stands for control variables. Other model settings are consistent with the baseline regression model.

Building on the work of Zhang and Li⁴¹ this study further breaks down the process of intelligent transformation into two components: intelligent technology level (sp) and depth of intelligent technology application (sd). Specifically: Intelligent Technology Level (sp) is defined as: sp = ln (number of AI-related keywords in annual reports of listed companies + 1); Depth of Intelligent Technology Application (sd) is defined as: sd = ln (number of smart business-related keywords in annual reports of listed companies + 1). The informationization level of enterprises is measured by two components: software informationization level (soft) and hardware informationization level (hard). Based on the research of Ho et al.⁴² these are calculated from the financial reports of listed companies, focusing on intangible assets and fixed assets related to artificial intelligence. Software Informationization Level is the proportion of intangible assets related to informationization in total assets. Hardware Informationization Level is the proportion of fixed assets related to informationization in total assets. Enterprise Innovation Level (Innov) is measured by patents, calculated as: Innov = ln(number of invention patents and utility model patents + 1).

Table 7 presents the regression results for five variables: Intelligent Technology Level, Depth of Intelligent Technology Application, Software Informationization Level, Hardware Informationization Level, and Innovation Ability (columns 1–5). The regression results show that the interaction terms between the moderating variables and the policy dummy variable are significantly positive. This indicates that the “Broadband China” demonstration cities not only promote enterprise intelligent transformation and informationization but also enhance innovation capabilities, which in turn boost the new quality productivity levels of enterprises. These findings support the hypotheses H1, H2, and H3, confirming that the “Broadband China” initiative effectively contributes to higher levels of intelligent transformation, informationization, and innovation, all of which positively influence the new quality productivity of enterprises.

Table 7 Mechanism test regression results.

To examine the synergistic effect of the mechanism variables on the development of new quality productivity (NQP) in enterprises, we construct the following model for the mechanism synergy effect test:

$$NQ{P_{it}}={\beta _0}+{\beta _1}di{d_{it}}*IN{T_{it}}*IN{F_{it}}+{\beta _2}di{d_{it}}+{\beta _3}IN{T_{it}}+{\beta _4}IN{F_{it}}+{\beta _5}Controls+{\mu _i}+{\lambda _t}+{\varepsilon _{it}}$$

(4)

In this model: did represents the policy dummy variable, $IN{T_{it}}$ and $IN{F_{it}}$ represent two mechanism variables (Any two influencing mechanisms in the transformation of enterprise intelligence, informatization, and innovation capabilities). The interaction term $di{d_{it}}*IN{T_{it}}*IN{F_{it}}$ is denoted as TII (Triple Interaction Term), which captures the synergy effect between the two mechanism variables in the presence of the policy impact. $IN{T_{it}}$ represents one of the mechanism variables among enterprise intelligent transformation, informatization, and innovation capability, while $IN{F_{it}}$ represents another mechanism variable. If the coefficient of the interaction term is statistically significant, it indicates the presence of a synergistic effect between the two mechanisms. This suggests that the “Broadband China” demonstration cities can enhance enterprises’ new quality productive forces through the coordinated effect of these two mechanism variables. ${\beta _1}$ is the coefficient of the core explanatory variable, and if ${\beta _1}$ is significantly positive, it indicates that the “Broadband China” demonstration cities can promote enterprise new quality productivity through the synergistic effects of the two mechanism variables. Controls represents the control variables, and ${\mu _i}$ and ${\lambda _t}$ are the individual and time fixed effects, respectively, ${\varepsilon _{it}}$ is the error term.

Tables 8 and 9 show the regression results for the interaction term TII. In both tables, the TII coefficients are significantly positive, indicating that there is a complementary synergistic effect between the mechanism variables. This finding confirms that the synergy between the adoption of smart technologies and informatization, the adoption of smart technologies and enterprise innovation, as well as informatization and enterprise innovation, plays a positive role in enhancing enterprise new quality productivity. This supports Hypothesis H4, which suggests that the three mechanisms—smart technology adoption, informatization, and enterprise innovation—work synergistically in a complementary manner to improve enterprise new quality productivity. It proves that the “Broadband China” policy enhances enterprise new quality productivity by influencing the interaction effects between these mechanisms.

Table 8 Mechanism synergy effect test regression results I.

Table 9 Mechanism synergy effect test regression results II.

Heterogeneity analysis

Heterogeneity analysis is a crucial part of understanding internal differences within the research object and enhancing the credibility and policy relevance of the research conclusions. The benchmark regression results typically reflect the “average effect,” but the research objects (such as enterprises, cities, or individuals) may exhibit significant differences. Focusing solely on the overall effect could obscure the unique patterns within specific subgroups and potentially lead to biased conclusions. Therefore, we conduct heterogeneity analysis in three dimensions: city level, ownership structure, and the lifecycle stage of the enterprise, in order to explore the different policy effects in various enterprises and cities.

City level heterogeneity

In China, cities differ in administrative levels, which include directly governed municipalities, sub-provincial cities, prefecture-level cities, and county-level cities. Based on the administrative level of the cities where the listed companies’ headquarters are located, we categorize the enterprises and conduct empirical regressions. The regression results are shown in Table 10. From Table 10, it is evident that the “Broadband China” demonstration cities have a more significant impact on the enhancement of new quality productivity in enterprises located in sub-provincial cities. This is likely because sub-provincial cities can offer better policy support, thereby assisting enterprises in their development. According to Acemoglu & Robinson’s⁴³ “inclusive institutions” theory, sub-provincial cities have a higher administrative rank than typical prefecture-level cities and thus enjoy more advantages in institutional inclusiveness. This advantage is reflected in a higher degree of autonomy in fiscal management, economic planning, and social governance. With these inclusive institutions, sub-provincial cities are better positioned to integrate administrative resources, promote policy implementation, and attract a range of support resources such as funding, talent, and policy support. From the perspective of institutional inclusiveness, sub-provincial cities hold significant economic importance within their respective provinces, and their more inclusive institutional environment allows them to upgrade resource allocation, policy authority, and development potential. Moreover, they play a key role in driving the development of surrounding cities and regions. Compared to prefecture-level cities, sub-provincial cities have advantages in decision-making autonomy, resource aggregation, and regional influence, enabling them to gain a head start in economic development, social governance, and strategic implementation.

Table 10 Regression results of city-level heterogeneity.

Ownership structure heterogeneity

Based on the ownership structure of the enterprises, we categorize the sample into state-owned enterprises, private enterprises, and foreign-invested enterprises, and conduct separate regressions for each category. The regression results are presented in Table 11. From Table 11, it is clear that the “Broadband China” demonstration cities have a more significant impact on the enhancement of new quality productivity in private enterprises. The results indicate that the impact of the “Broadband China” demonstration cities is more pronounced in boosting the new quality productivity levels of private enterprises. Based on the resource-based view⁴⁴ private enterprises have unique agility advantages. Their market-oriented, flat decision-making structure makes the decision process more efficient, enabling them to quickly respond to market demands. Moreover, their flexible capital allocation capabilities provide strong support for technological breakthroughs and industrial upgrades in fields such as artificial intelligence and new energy. State-owned enterprises, on the other hand, are constrained by an administrative management framework, with decision-making processes being more hierarchical. Resource allocation in state-owned enterprises must balance policy objectives and social responsibilities, which limits their ability to quickly respond to market dynamics and technological iterations. Foreign-invested enterprises, while possessing technological advantages, tend to focus their core research and development activities at their headquarters in the home country, with their branches in China mainly focusing on technology application and local adaptation. As a result, they have less autonomy over strategic disruptive innovations.

Table 11 Equity heterogeneity analysis regression results.

Heterogeneity analysis of enterprise life cycle

Referring to the method of Li Yunhe et al.⁴⁵ four indicators—sales revenue growth rate, retention ratio, capital expenditure ratio, and enterprise age—are used to categorize the development stages of enterprises into growth, maturity, and decline stages. The regression coefficient for growth-stage enterprises in Table 12 is significantly positive, indicating that the “Broadband China” pilot cities can significantly promote the improvement of new quality productivity in growth-stage enterprises. Growth-stage enterprises are typically in a market expansion phase, where they need to achieve market share through technological breakthroughs or business model innovations. These enterprises are more willing to allocate a high proportion of resources to research and development, as new quality productivity emphasizes the central role of technological innovation. The high-risk preferences and innovation strategies of growth-stage enterprises align naturally with this goal, allowing them to more quickly transform technology into productivity. In contrast, mature enterprises (such as some state-owned enterprises) are often in a phase of stock-market competition, where they are constrained by price regulations or market share consolidation, resulting in insufficient innovation motivation. They tend to rely on economies of scale rather than technological innovation to maintain profits. Declining enterprises, on the other hand, face significant survival pressures and struggle to sustain long-term research and development investments.

Table 12 Results of heterogeneous regression in the enterprise life cycle.

Conclusion and discussion

The construction of digital infrastructure, exemplified by the “Broadband China” pilot cities, is a key prerequisite and driving force for the development of the digital economy and new quality productivity. This paper examines how digital infrastructure construction affects enterprise new quality productivity development from both theoretical and practical perspectives, using the “Broadband China” pilot cities as a natural experiment. Compared to previous studies, most of which focus on the impact of digital infrastructure on macroeconomic indicators^{1,2,3,4,5,6,7,8,9,10,11,12,13} or analyze its effects on enterprises from a single dimension^{14,15,16,17,18,19,20} this study provides a multi-dimensional and systematic investigation of the formation mechanisms of new quality productivity in enterprises.

The empirical results show that the “Broadband China” policy promotes new quality productivity by improving infrastructure, accelerating enterprise digital transformation, enhancing enterprise informatization levels, and boosting innovation capabilities. Moreover, the mechanisms involved have a synergistic effect. This finding resonates with Acemoglu & Restrepo’s⁴⁶ “Technology-Organization Synergistic Evolution” theory, providing new evidence for the complex mechanisms of digital technology diffusion. After conducting parallel trend tests and a series of robustness checks, the results remain robust. While existing research mainly focuses on the average effect of policies on enterprises, this paper reveals through multi-dimensional heterogeneity analysis that the “Broadband China” policy has a more significant impact on the new quality productivity of enterprises in higher administrative-level cities, private enterprises, and growth-stage enterprises. Based on the theoretical and empirical analysis above, the following policy recommendations are proposed:

First, strengthen targeted policy support. Governments should develop differentiated policies tailored to enterprises in cities of varying administrative levels. For sub-provincial cities, in addition to existing support, special funds such as a “Digital Infrastructure Upgrade Fund” and a “Smart Transformation Fund” should be established. These funds would prioritize projects like the upgrading of intelligent manufacturing equipment and the construction of big data centers. Quantitative targets—such as achieving 80% coverage of smart manufacturing equipment within five years—should be set to further enhance these cities’ advantages in developing new quality productive forces. For prefecture-level and county-level cities, more foundational subsidies should be provided for digital infrastructure construction, including fiber-optic network installation and 5G base station deployment. Target coverage rates for 5G base stations (e.g., reaching 90% coverage within five years) should be set to reduce the digital operational costs for enterprises. Pilot programs for data factor markets (e.g., the Shenzhen Data Exchange model) should be promoted, with clear definitions of data ownership and transaction regulations. For private enterprises, administrative approval processes should be streamlined, and more financial support and tax incentives should be granted in areas such as digital technology R&D and the expansion of application scenarios. These could include raising the percentage of R&D tax credits and increasing subsidies for smart equipment. For state-owned enterprises, reforms in internal management systems should be pursued to simplify decision-making processes and improve responsiveness to digital technology applications. For enterprises at different stages of development, targeted support should be offered. Growth-stage firms should benefit from dedicated innovation incentive funds to stimulate digital R&D investments. Mature firms should be guided to use digital technologies to explore new markets and business models. Declining firms should be supported with digital transformation assistance programs to help them upgrade and transition to new industries.

Second, improve the layout of digital infrastructure development. Investment in digital infrastructure in economically underdeveloped regions and small to medium-sized cities should be increased to reduce the digital divide between regions and cities. Through government investment and the attraction of private capital, the coverage of fiber-optic and 5G networks should be accelerated and overall network quality improved. A regional digital industry coordination mechanism should be built to encourage cooperation between enterprises in developed and underdeveloped regions through digital platforms. Plans for cross-city digital alliances (e.g., modeled after the Yangtze River Delta’s “Digital Integration” framework) should be advanced. Within the framework of the “Western Digital Corridor,” cities like Chengdu-Chongqing and Xi’an should be encouraged to co-construct computing hubs (e.g., under the “East Data, West Computing” project) to enable shared access to technologies, talent, and data, promoting balanced development of new quality productive forces across regions.

Third, promote the development of the data factor market. A sound regulatory framework for the data factor market should be established, with clear definitions of data ownership, transaction standards, and security requirements. A unified data trading platform should be built to facilitate the orderly flow and sharing of data, reducing the cost for enterprises to access data. Governments should accelerate efforts to close the digital divide and address data security risks. A “Digital Inclusion Fund” should be created to support digital training for small and medium-sized enterprises in remote areas (e.g., referencing the EU’s Digital Skills and Jobs Coalition). Enterprises should be encouraged to strengthen data governance, improve data quality and value, and promote deep integration of data with digital technologies. This will help uncover the potential value of data in production, management, and innovation, and enhance the momentum for developing new quality productive forces.

Our study has several limitations: (1) Data coverage bias: We only use data from listed companies, neglecting small and medium-sized enterprises (SMEs) and the informal sector, which raises questions about the external validity of our conclusions. (2) Sample selection bias: Listed companies may be more inclined to invest in digitalization, so relying solely on listed company data could overestimate the policy effect. Future research should incorporate SME data to verify the robustness of the findings. (3) Neglect of negative impacts: We do not discuss the energy consumption or electronic waste associated with digital infrastructure—for example, the high energy use of 5G base stations. There is a lack of quantification regarding the carbon footprint of future digital infrastructure such as 5G base stations. (4) Scope limitation: Our research mainly focuses on the impact of digital infrastructure construction on firms’ new quality productivity improvement, while ignoring its effects on green economic development. (5) Future research directions: Future studies could combine Life Cycle Assessment (LCA) methods to analyze the energy efficiency of digital infrastructure and explore pathways for green transformation.