Digital infrastructure construction and corporate innovation efficiency: evidence from Broadband China Strategy

The development of multiple technologies, including big data, the Internet of Things and cloud computing, are driving the modern economy into a phase of digital evolution (Piepponen et al., 2022). Digitalisation is not only having a profound impact on business strategy, management and operations (Truant et al., 2021), it is also changing the ways in which value is created, delivered and captured (Li et al., 2023). The digitalization of enterprises is context-dependent, which in turn is closely related to the social, political and competitive environments (Appio et al., 2021). Therefore, governments always take various measures to provide room for companies to digitally develop. Digital infrastructure serves as a crucial channel for governments to boost the development of the digital economy (Wang & Shao, 2024), promoting the use of various digital resources, having the potential to disrupt existing businesses that are slow to adapt to technological changes (Vial, 2019), and accelerating the improvement of the digital literacy in society (Zhou, 2022). The rapid development of digital infrastructure is creating unprecedented opportunities for companies to improve their innovation efficiency, underpinning the vigorous growth of China’s economy (Ding et al., 2021).

Among the various economic consequences of improved digital infrastructure, corporate innovation holds substantial academic value. The role of innovation in enabling firms to achieve long-term competitive advantages and promote national economic growth is increasingly recognized (Yue, 2022). According to the Global Innovation Index 2023, China ranks twelfth globally in innovation capacity among top thirty innovative nations. As the key decision-makers within the national innovation strategy system (Zhong et al., 2022), enterprises with high innovation efficiency not only enhance their competitive advantage(Qiu et al., 2023), but also drive sustainable development (Zhong & Ren, 2024), and contribute to overall economic growth (Fitzgerald et al., 2021). Simultaneously, with the growing importance of ecological concerns and sustainable development, green innovation has attracted considerable academic interest. Research has demonstrated that both digital transformation (Han et al., 2024) and the development of digital infrastructure (Guo et al., 2024) play pivotal roles in fostering green innovation of companies.

In view of this, we hypothesize that the Broadband China strategy enhances corporate innovation efficiency, with financing constraints and human capital playing moderating roles. This hypothesis was strongly supported by our analysis of Chinese-listed companies from 2010 to 2022. The results reveal key insights: the development of digital infrastructure significantly improves innovation efficiency. Mechanism analysis shows the intensity of financing constraints negatively moderates this innovation impact, while the structure of human capital positively moderates it. Additionally, the impact on innovation varies across firms, with the most pronounced effects observed in non-state-owned, non-high-tech, and noneastern region enterprises.

Compared to previous studies, the contributions of our paper are threefold. First, we systematically examine the economic impact of the Broadband China Strategy from the microenterprise perspective. While much of the existing research has focused on the strategy’s macroeconomic benefits, such as promoting low-carbon development (Ding et al., 2024), improving city-level innovation ability (Liu et al., 2024), and increasing employment (Jin et al., 2023), our study fills a gap by exploring its impact on firm-level innovation efficiency. Second, we expand the scope of research on the factors affecting corporate innovation efficiency. While many studies have demonstrated the critical effect of the digital economy on corporate innovation (Benitez et al., 2022; Zhuo & Chen, 2023), we use a multi-period Difference-in-Differences (DID) model, and provide a deeper understanding of how digital infrastructure influences corporate innovation efficiency. Third, we examine the internal mechanisms and boundaries through which the Broadband China Strategy impacts corporate innovation efficiency. By analyzing how financing constraints and human capital play moderating roles, our study reveals the micro-level pathways through which macro-level policies influence corporate innovation. This deepens our understanding of the effects of policy interventions and their limits. Furthermore, our research offers practical implications, providing valuable guidance for both governmental entities and corporations on improving innovation efficiency through enhanced institutional frameworks and digital infrastructure construction.

Digital infrastructure construction

Current academic research on digital infrastructure can be categorized into three primary areas. First, scholars are investigating the theories and characteristics of digital infrastructure. Koutsikouri et al. (2018) proposed an initial classification of four growth strategies aimed at extending infrastructure to accommodate future service demands within its operational scope. Fürstenau et al. (2019) identified parallel, competitive, and spanning processes in the evolution of digital infrastructure. Ratner and Plotnikof (2022) explored the relationship between digital technologies and organizations, conceptualizing digital data infrastructures as partial connections. In empirical research, some academics have developed index systems to measure the comprehensive development of digital infrastructure. Using the entropy weight method, Fan et al. (2022) created a four-dimensional index system that evaluates digital infrastructure. Similarly, Tang and Yang (2023) employed an entropy-based methodology to assess the overall development level of China’s digital infrastructure, considering construction extent, usage costs, and coverage rates. As research deepens, increasing attention is being paid to the effects of digital infrastructure construction. Studies have shown that digital infrastructure enhances total factor productivity (Tang & Zhao, 2023), facilitate low-carbon development (Hu et al., 2023), and increase labor income shares (Wang et al., 2023). Meanwhile, research on digital infrastructure is gaining a global perspective. Schade and Schuhmacher (2022) utilized data from 46 countries to examine how digital infrastructure affects individual’s likelihood of starting new venture, while Osei (2024) explored the role of digital infrastructure in fostering innovation across African.

Corporate innovation efficiency

The existing literature on the determinants of corporate innovation efficiency primarily focuses on both macro and micro-level factors. At the macro level, research has explored the influence of economic policy uncertainty (Zhou et al., 2023), technological advancements such as artificial intelligence (Wang et al., 2024), monetary policy (Yang et al., 2024), and market competition (Huang, 2023), among others. At the micro level, studies have examined corporate ownership structures (Wu et al., 2022), trends in financialization (Zhu et al., 2023), corporate social responsibility (Yuan et al., 2023), and CEO characteristics (Zhang et al., 2022), as key determinants. While previous research has recognized the substantial impact of digitalization on corporate innovation activities (Usai et al., 2021; Li et al., 2023; Wu & Li, 2024), the relationship between digital infrastructure construction and corporate innovation efficiency remains largely unexplored. Could digital infrastructure serve as a catalyst for enhancing innovation efficiency in Chinese firms? Investigating this nexus has the potential to significantly deepen current scholarly understanding and offer invaluable insights for both academics and policymakers aiming to promote high-quality development in China.

Policy background

Broadband China Strategy. As an auspicious external policy, the Broadband China Strategy provides strategic guidance and methodical implementation to accelerate the expansion of China’s broadband infrastructure and enhance the innovation capabilities of its enterprises. Therefore, we zoom in on beefing up Chinese digital infrastructure within the context of this strategy. In the early 2000s, China faced challenges with its broadband infrastructure (Zhen et al., 2015) and encountered high costs associated with Internet usage (Zhou et al., 2022). The Broadband China Strategy, tested in 117 cities from 2014 to 2016, bolsters the groundwork for digital infrastructure construction. In 2013, the China’s State Council issued a program to execute the Broadband China Strategy, recognizing broadband as a strategic component of national public infrastructure for the first time. The initial focus of the strategy was on the popularization of broadband from 2014 to 2015, followed by optimization from 2016 to 2020. The strategy’s objectives are divided into two primary areas: maximizing the potential of wireless and wired broadband networks, with a focus on critical sectors such as education, healthcare, energy conservation, public safety, household decoration, transportation, and environmental protection; and fostering business innovation through the development and deployment of various technologies, including mobile Internet, cloud computing, and e-commerce. The policy has accomplished its intended goals in terms of its accomplishments (Liu & Wang, 2019). As of 2020, the development of China’ broadband infrastructure has approached the standards of developed countries (Zhang, 2021).

Theoretical mechanisms

Digital infrastructure involves broadband networks, data centers and various digital platforms, etc., with the advantages of open sharing, spatial ubiquity, and resource expansibility (Tang & Zhao, 2023). According to the resource base view (RBV), the innovative development of enterprises requires a large number of tangible and intangible resources, which can be provided both internally and externally (Mella, 2024). Digital infrastructure is vital in accumulating and agglomerating the elements of innovation, acting as a conduit for information and knowledge capital (Wang et al., 2023). Digital infrastructure can generate new value networks and service systems in the physical world, thus enhancing the vitality of corporate innovation. According to the innovation diffusion theory, innovation can spread through different channels over time among the members of a social system (Ho, 2022). Digital infrastructure can reduce information asymmetry and facilitates the flow of knowledge across firms, fostering innovation (Sun et al., 2023). In terms of internal R&D efficiency, businesses can respond to customer demands in real-time and remotely manage operational costs, risks, and uncertainties through intelligent supply chains and online platforms (Autio et al., 2018), fostering open-source technology and cross-disciplinary innovation, thereby increasing innovation efficiency. In terms of external cooperation and innovation efficiency, the construction of digital infrastructure effectively overcomes geographical and temporal constraints on information exchange, increases technology and knowledge spillovers, reduces information search and transmission costs, and improves the efficiency of technology exchange and R&D cooperation among enterprises (Forman & Zeebroeck, 2012). Moreover, in regions with well-developed digital infrastructure, the speed of information transmission at very large scales has increased significantly, and the sharing of the latest R&D results and the spillover of cutting-edge knowledge have been greatly improved and accelerated by building intelligent platforms (Nylén & Holmström, 2019). Therefore, we propose the following:

Hypothesis 1: Digital infrastructure construction positively impacts corporate innovation efficiency.

Innovation activities represent a form of investment that is both risky and capital intensive, characterized by significant financial outlays and a long R&D cycle. Any significant innovation requires long-term capital investment (Hsu et al., 2014), and the lack of capital limits firms’ innovative behavior. Existing literature has thoroughly investigated the impact of financing constraints on corporate innovation endeavors and performance (Hall et al., 2016; Guariglia & Liu, 2014). On the one hand, enterprises facing serious financing constraints have a lag effect in the process of strategic transformation, and the lack of funds delays the output of innovation, which is not conducive to the enhancement of the innovation efficiency. When enterprises face a more relaxed financing environment, they can obtain sufficient development funds from the external capital market at a smaller cost, and the capital demand for innovation and R&D investment is satisfied, the enterprise may choose to carry out high-risk and high-potential innovation projects to improve innovation efficiency (Campello et al., 2010; Zhao & Zhang, 2023). On the other hand, firms with high financing constraints face greater operational risks, and managers may hide negative information to alleviate their financing pressures, which ultimately leads to deeper information asymmetry, exacerbates the adverse selection problem (Bae et al., 2021), and is more likely to have a negative impact on their innovation efficiency. Therefore, we propose the following:

Hypothesis 2: Financing constraints have a negative moderating effect on the relationship between digital infrastructure construction and corporate innovation efficiency.

Endogenous growth theory states that human capital is a centerpiece factor influencing innovation (Izushi, 2008). As one of the most important strategic resources, human capital plays a vital role in driving scientific research, technological innovation, knowledge dissemination, and the transformation of achievements (Wen et al., 2023). The moderation of human capital on digital infrastructure development and corporate innovation efficiency is mainly realized through internal and external pathways. From an internal perspective, human capital with advanced information and digital science technology promotes innovation efficiency because of its unique innovative power, which is the subjective dynamic effect. Specifically, as carriers of technological knowledge, people with a certain level of education have a higher ability to acquire, transform and utilize knowledge (Acemoglu & Autor, 2010). The expansion of human capital can increase the opportunities for firms to obtain more highly skilled labor and enable them to adopt cutting-edge technologies (Che & Zhang, 2018). Using advanced human capital not only boosts firms’ ability to absorb and reinvent technologies (Kong et al., 2022) but also motivates them to improve their innovation efficiency. From an external perspective, human capital can bring about external agglomeration and spillover effects, and externally promote innovation efficiency. The exchange and agglomeration of talents make the interaction across regions and enterprises more frequent and efficient, reduce the cost of acquiring technical information and management experience. This multiplier effect accelerates the dissemination and migration of knowledge, and produce knowledge spillover effect (Moretti, 2021), which is conducive to accelerating the research and development of new knowledge, new technologies and new products. Human capital also has a social network effect (Ning et al., 2016), which can help enterprises gain timely insight into promising technological frontiers and trends from the outside and improve the efficiency of enterprise innovation. Therefore, we propose the following:

Hypothesis 3: Human capital have a positive moderating effect on the relationship between digital infrastructure construction and corporate innovation efficiency.

Econometric model

Multi-period DID model

Given that the tendency of using multiple linear regression methods to prioritize correlation over causation in the development of indicators for digital infrastructure construction, as well as the potential for endogeneity issues within such regressions, we utilized the DID model for our investigation. Leveraging the Broadband China Strategy as a quasi-natural experiment, we specifically implemented a multi-period DID model to account for differential counts of cities approved as Broadband China demonstration cities. Following Moser and Voena (2012)’s instruction and guidance from other researchers, the multi-period DID regression model was constructed. Equation 1 tests the direct effect of digital infrastructure construction on corporate innovation efficiency, and Eq. 2 examines the moderating effects of financing constraints and human capital on digital infrastructure construction on innovation efficiency.

where Inoeff_it represents innovation efficiency for corporation i in year t; the interaction term did_it is a dummy variable indicating the implementation of the Broadband China Strategy, and it represents the strategy implementation effect, with specific setting rules described below; M_it stands for the moderating variables, which are financing constraints and human capital; X_it symbolizes additional control variables; γ_t indicates the year-fixed effect; μ_i denotes the corporate-fixed effect; and ε_it is a term for random error. The robust standard errors were also clustered at the corporate level during the estimation process. To a certain extent, the model successfully accounted for the differences and temporal patterns between the treatment and control groups.

Parallel trend test

A series of assumptions must be fulfilled before processing DID models, with the parallel trend assumption being the most pivotal. This assumption posits that, in the absence of policy interventions, the development trend of innovation efficiency for corporations in the treatment group would coincide with that of the control group. We drew on Beck et al. (2010) and adopted an event study approach to assess parallel trends, scrutinizing the dynamic effects of the Broadband China Strategy with the following model setup:

where did is the dummy variables, did^k (when k is negative) indicates whether corporation i is in the k^th year before the implementation of the Broadband China Strategy at time t; D^k (when k is positive) indicates whether corporation i is in the k^th year prior to the execution of the Broadband China Strategy at time t. If the coefficient λ_k (when k is negative) is statistically insignificant, then the parallel trend test is deemed passed.

Variable selections

Explained variable

Corporate innovation efficiency (Inoeff) based on a DEA model is the dependent variable. The DEA model uses a mathematical planning method to evaluate the relative effectiveness between the input and output decision-making units (DMUs), which can avoid the inaccuracy of the measurement due to the large fluctuation of the data (Carayannis et al., 2016; Tran, 2020). Therefore, we adopted the natural logarithm of R&D expenses as the innovation input variable and the natural logarithm of the total number of the three patent types as the innovation output variable (including inventions, utility models and designs), utilizing the DEA method to measure innovation efficiency (Li et al., 2023).

Core explanatory variable

The independent variable under examination is the implementation of the Broadband China Strategy (did). This variable is derived from the multiplication of the treatment group dummy variable (group) and the policy implementation time dummy variable (time). Group i symbolizes corporate dummies (assigned a value of 1 if the city in which corporation i is located is affected by the strategy, or 0 otherwise); time t is a temporal dummy variable (assigned a value of 1 for the year when the Broadband China pilot city program inception and for subsequent years, or 0 otherwise).

Control variables

To mitigate the influence of extraneous factors and to facilitate an unbiased assessment of the policy’s impact, control variables were incorportated. Firm size, a determinant of innovation behavior, is measured by the logarithm of total assets (Horbach, 2008). Financial constraints, which can impede innovation efficiency, are controlled through financial indices such as return on assets, financial leverage, and cash flow ratio (Chava et al., 2017; Wang, 2023). Next, we added firm age and growth capacity by incorporating the logarithm of the establishment time (Zhang et al., 2023) and the growth rate of primary business revenue (Wang et al., 2019), respectively. Moreover, shareholders and the CEO significantly promote corporate innovation efficiency (Lewellyn & Muller-Kahle, 2012). Hence, we controlled the shareholding concentration, the board size, and the combination of the positions of shareholders and CEOs. Finally, we also controlled for the aggregate stock of patents obtained by firms to respond to the concern that more innovation was due to richer prior knowledge (Capolupo et al., 2024).

Mechanism variables

Firstly, to quantify the corporate financing constraints, we selected the KZ index, followed by Kang et al. (2017). A larger KZ index represents a higher degree of financing constraints. Secondly, given the well-documented correlation between educational level and individual comprehensive quality, we examined the human capital by assessing the educational composition within the workforce. Drawing upon the methodology proposed by Call et al. (2017), we quantified the level of employee education by considering the proportion of employees who hold a bachelor’s degree or higher in relation to the overall employee population.

Data and statistical description

The sample for our research includes companies listed on the Shanghai and Shenzhen Stock Exchanges spanning the period from 2010 and 2022. The dataset underpinning this analysis is sourced from the China Stock Market and Accounting Research (CSMAR) Database. The list of Broadband China pilot cities is based on the official list established by the Ministry of Industry and Information Technology and the National Development and Reform Commission in 2014, 2015, and 2016. Data matching was conducted utilizing textual information pertaining to the registered locations of the listed companies and the prefecture-level cities engaged in the Broadband China pilot city initiative. In adherence to research conventions, we excluded the ST and PT anomalies, the financial and insurance samples, and the samples with missing main variables. Additionally, to mitigate the influence of outliers, all continuous variables underwent a trimming process at the 1% level. Consequently, the skewed panel data comprised 24,263 legitimate observations, including 79 industries and 3773 companies.

Table 1 presents the descriptive statistics of the dataset. The average inoeff stands at 0.364, with a standard deviation of 0.150. There is significant variation in corporate innovation efficiency, thereby substantiating the academic significance of our research. The mean value of the did variable is 0.611, signifying that the Broadband China Strategy encompasses more than half of the enterprises within our sample. In addition, the descriptive statistical data for the control variables exhibit a reasonably dispersed distributed.

Parallel trend test

Figure 1 implies that the policy variable coefficients are insignificant before the implication of the Broadband China Strategy, suggesting no statistically significant difference in corporate innovation efficiency between the treatment and control groups prior to the strategy implementation. Furthermore, the estimated coefficients demonstrate a significant positive trend within 4 years of strategy implementation. This indicates that adopting the Broadband China Strategy significantly enhances corporate innovation efficiency during the performance period. To sum up, the hypothesis of parallel trends has been confirmed. However, the Broadband China Strategy has only a short-term policy effect and lacks sustainability, as evidenced by the fact that the confidence intervals of the coefficients of DID contain 0 starting from the fifth period following the strategy’s adoption.

This figure depicts the dynamic impact of the pre- and post-terms of the pilot policy on corporate innovation efficiency. This figure is covered by the Creative Commons Attribution 4.0 International License. Reproduced with permission of NAME; copyright © NAME, all rights reserved.

Full size image

Benchmark model result

To precisely evaluate the overall effect of the Broadband China Strategy, a multi-period DID model (Eq. 1) was utilized to finalize our baseline regression analysis. The coefficient estimates of did are all positive and statistically significant at the 1% critical level (see Table 2), under the step-by-step method of adding control variables. This suggests that digital infrastructure construction has improved innovation efficiency in Chinese firms, supporting Hypothesis 1. Digitization plays a crucial role in promoting technological innovation (Ahn, 2020). Hence, the enhancement of digital infrastructure construction under the guidance of Broadband China Strategy is expected to become the driving force for enterprises to improve innovation efficiency.

Robustness test

PSM-DID regression

The determination of the cities designated for the Broadband China pilot is not arbitrary; rather, these cities are typically characterized by robust economic expansion and considerable potential for innovation. The aforementioned process of selection may introduce a preference bias within the treatment group, potentially yielding a consistent bias within the dataset. To tackle the possible endogeneity issue, we drew on the methodology framework developed by He et al. (2023) and employed the PSM approach for examination. The basic principle of PSM is to transform multidimensional covariates into one-dimensional propensity scores, and then find firms Y’ in the control group that have the same or similar scores as firms Y in the experimental group based on the propensity scores to match, so as to make the sample observed data close to the data of randomized experiments. The problem of bias due to self-selection of observable variables is solved, and the difference in the impact of the external policy shock on the experimental and control groups is the net effect of the policy.

In concrete terms, we firstly performed Probit regression on the Broadband China Strategy and the factors that may affect the corporate innovation efficiency to calculate the propensity score of the sample enterprises. Secondly, according to the different matching principles, we constructed a control group of samples of enterprises in the pilot region that have similar characteristics with those in the non-pilot region. Ten control variables were selected to serve as corresponding variables. Subsequently, three distinct matching techniques were used to pair the treatment and control groups on an annual basis, including 1:1 nearest neighbor matching, radius matching, and kernel matching. Ultimately, a reevaluation of the matched results was conducted. Table 3 shows that under a trio of matching techniques, the estimated coefficients (did) are 0.011, 0.009 and 0.013, respectively, and statistically significant at 1%. This shows that the corporate innovation efficiency in non-pilot regions is not only significantly different from that of firms in pilot regions before pairing, but the difference remains significant after controlling for differences in other characteristics. The estimations detailed supra are concordant with the baseline regression outcomes, suggesting that the results remain robust following the mitigation of the bias associated with sample self-selection.

Placebo test

To mitigate the likelihood of estimation bias arising from unobservable confounders, a placebo test was conducted to ascertain the robustness of the DID results, in accordance with the methodology delineated by Chetty et al. (2009). Specifically, we introduced new DID variables by randomly selecting pilot cities and adjusting the timing of the policy and repeated the random selection regression 1000 times, thereby executing a counterfactual test of the baseline regression results. Figure 2 shows that the estimated coefficients of the pseudo-did under the random treatment are distributed infinitely close to 0 and closely approximate the standard normal distribution. The vast majority of the pseudo-did estimated coefficients are accompanied by p-values that exceed 0.10, signifying that none of the estimated coefficients meet the statistical significance at the 10% threshold. Moreover, the estimated coefficients of the baseline regression fall outside the distribution of the pseudo coefficients. Therefore, the baseline regression results are not affected by the omission of variables or by random factors, reinforcing the validity of our results.

This figure shows the coefficients, P-values, and kernel density curves obtained by 1000 regressions by randomly selecting pilot cities and adjusting the timing of the policy. This figure is covered by the Creative Commons Attribution 4.0 International License. Reproduced with permission of NAME; copyright © NAME, all rights reserved.

Full size image

Excluding the effect of other policies

In addition to the Broadband China Strategy, China is proactively enforcing additional pilot policies aimed at nurturing innovation within enterprises. A policy directly contiguous with this initiative is the Smartcity Pilot Policy, executed by the Ministry of Housing and Urban-Rural Development in 2012. The two policies share considerable overlap in terms of their chronology and intended goals, culminating in significant enhancements in the efficacy of corporate innovation processes (Qiu, 2023). The coefficient of did remains significantly positive (see column (1) of Table 4), so after excluding the interference of smart city policy, the Broadband China Strategy can still significantly boost corporate innovation efficiency. Moreover, we altered the sample interval from 2012 to 2018 to eliminate the confounding influence of other policy interferences, and the findings continue to exhibit robustness (see column (2) of Table 4).

Heterogeneous treatment effects

Heterogeneity treatment effects across time and groups within a multi-period DID framework may engender a negative weighting issue, attributable to the inconsistent timing of treatment exposure among the samples. This can result in the generation of biased estimates (Goodman-Bacon, 2021). To address this, we adopted the DID_M estimator, as proposed by de Chaisemartin and D’ Haultfoeuille (2020), to diagnose the potential for heterogeneous treatment effects within our baseline regression. First, the test results derived from the twowayfeweights command indicate that the aggregate of positive weights is 1.0423, while the sum of negative weights is -0.0423, implying that the heterogeneity treatment effects may not exert a substantial impact on the results of the benchmark regression. Additionally, the DID_M estimator yields an average treatment effect of 0.007 for policy switching, and the assessment of dynamic effects further supports the robustness of our conclusions (Fig. 3).

The figure presents the robust DID_M estimator obtained by the weighted average with dynamic treatment effects. This figure is covered by the Creative Commons Attribution 4.0 International License. Reproduced with permission of NAME; copyright © NAME, all rights reserved.

Full size image

Analysis of influencing mechanism

The influence mechanism of digital infrastructure construction on corporate innovation efficiency includes the moderating roles of financing constraints and human capital. We conducted moderated effects tests using interaction terms (see Table 5). The results in column (1) illustrate that the estimated coefficient of the interaction term between digital infrastructure construction and financing constraints (KZ×did) is -0.008, and is significant at the 1% level, indicating that financing constraints significantly weaken the relationship between digital infrastructure development and enterprise innovation efficiency, supporting Hypothesis 2. Furthermore, the interaction term HC×did in column (2) is 0.019 and significant at the 5% level. So the optimization of human capital has a significant positive moderating effect in the process of building digital infrastructure to improve the innovation efficiency of enterprises, thereby affirming Hypothesis 3.

Heterogeneous analysis

The efficacy of any policy is not uniformly distributed across all individuals, and the innovation effect may vary among different companies. Our research delved deeper into the differential impacts of digital infrastructure construction on innovation efficiency, with a specific focus on ownership heterogeneity, technology endowment heterogeneity, and regional heterogeneity.

Ownership heterogeneity effect

Compared with non-SOEs, SOEs have significant advantages in terms of resource accessibility, innovation capacity, and so on (Genin et al., 2021; Poczter, 2017), and their digital infrastructures are built on a solid foundation. We categorized companies into SOEs and non-SOEs based on their ownership. In column (1) of Table 6, the estimated coefficient of did reveals no statistical significance among the sampled SOEs. Conversely, the estimated coefficient is 0.01 and statistically significant at the 1% level in the sample of non-SOEs, as shown in column (2). The empirical p-values obtained by Bootstrap method further validate the statistical significance of the above differences. This suggests that the impact of the Broadband China Strategy on non-SOEs is more pronounced in terms of enhancing corporate innovation efficiency.

Technology endowment heterogeneity

We categorized companies into high-tech and non-high-tech enterprises based on the 2012 industry classification standard of the Securities and Futures Commission (SFC) and the “High and New Technology Fields with National Key Support”. The data presented in Table 6, columns (3) and (4), indicate that digital infrastructure construction significantly boosts corporate innovation efficiency in non-high-tech enterprises, while the effect in high-tech enterprises is less pronounced. This is due to the fact that high-tech enterprises with extensive digital infrastructure construction exhibit higher innovation efficiency within their respective industries. The adoption of the Broadband China Strategy has addressed the deficiency in digital infrastructure construction within non-high-tech companies, thereby facilitating an increase in enterprise innovation efficiency. This conclusion is in alignment with the Yuan’s (2023).

Regional heterogeneity

From a geographical perspective, the innovation efficiency of Chinese companies exhibits pronounced spatial heterogeneity. There are significant differences between the eastern and non-eastern companies in terms of economic development, infrastructure development, talent pool and technology accumulation. Therefore, our research classified enterprises into two distinct categories based on their respective locations in China’s eastern and non-eastern regions. Empirical findings indicate that the digital infrastructure construction selectively enhances the innovation efficiency of enterprises located in the non-eastern region (Table 6). These results can be primarily attributed to the fact that the previous digital infrastructure in the eastern region is more advanced and established (Ma & Lin, 2023), the strategy is more effective in assisting areas that are relatively underdeveloped.

By processing the data from Chinese listed firms from 2010 to 2022, we harness the Broadband China Strategy as a quasi-natural experiment to investigate the impact of digital infrastructure construction on corporate innovation efficiency via the multi-period DID model. The main findings derived from our analysis are as follows: Firstly, there is a discernible facilitative effect of digital infrastructure construction on enhancing corporate innovation efficiency. And the parallel trend test shows the Broadband China Strategy has only a short-term policy effect and lacks sustainability. Secondly, the mechanism analysis indicates that this innovation impact is positively moderated by human capital and adversely moderated by financing constraints. This conclusion in turn enriches the moderating role of these two mechanistic variables in terms of studies from Zhang and Li (2023). Thirdly, it is pertinent to acknowledge that the innovation-enhancing effects of the Broadband China Strategy are more pronounced among non-state-owned enterprises, non-high-tech enterprises, and enterprises in the non-eastern region. In contrast to the research of Yang et al. (2022), we argue that the policy effects of the Broadband China Strategy are timely assistance. Our findings demonstrate that the Broadband China Strategy on fostering innovation within firms, thereby fulfilling its intended objectives. Overall, our research offers valuable insights for enterprises seeking to augment their innovation efficiency, providing micro-level evidence for the innovation effects of digital infrastructure.

The Chinese government should continue to implement the Broadband China Strategy, and cumulative long-term effects of the policy though expanding the scope of pilot Broadband China demonstration cities and broadband network coverage. Based on the heterogeneous effects of pilot policies on corporate innovation efficiency, the Chinese government should emphasize the balanced development of the economies between developed and underdeveloped enterprises, increase investment in and support for new infrastructure in disadvantaged companies, and give full play to the role of the digital infrastructure centered on the Broadband China Strategy in empowering high-quality economic development. This will assist dismantle barriers to the flow of innovative resources and knowledge within markets, foster cross-regional and cross-sectorial innovation cooperation among enterprises, and bolster their collective innovative capacity. Pilot cities should summarize in depth the successful experiences and shortcomings in the process of policy implementation and accelerate the construction of a new generation of information infrastructure. It is imperative to coordinate the planning of regional and industrial layouts for digital infrastructure construction, promote the integration and open sharing of data resources, and stimulate the innovation potential of enterprises. Non-pilot cities should formulate relevant policies as soon as possible, attract regional enterprises to actively participate in network infrastructure construction through financial support and talent policies, and bridge the gap as soon as possible.

Enterprises should pay close attention to the dynamics of relevant national policies, actively cooperate with national pilot policies, increase investment in innovation activities, and improve their own innovation efficiency. Specifically, enterprises should intensify their recruitment of highly skilled personnel, enhance the professional knowledge and digital literacy of current employees through on-the-job training and education, and actively refine their human capital. Also, enterprises should harness diverse financing mechanisms to expand their financing channels, effectively address the financing requirements necessary for engaging in innovative activities, and thus provide a solid financial foundation for enterprise digital innovation.

Our research contribution needs to be acknowledged, along with the constraints, which delineate avenues for future inquiry. Firstly, owing to the constraints of data accessibility, our focus has been narrowed exclusively to patent-based innovations, excluding other types of innovation such as those related to processes, organization structures, and business models. Future research could expand the scope by incorporating these additional dimensions of innovation, leveraging new datasets or methodologies to capture a more holistic view of innovation. In addition, economic and social transformation, as well as technological advancement, are inherently fraught with risk and challenges; the cycle of “technological revolution-financial bubble-crash” has iteratively played out throughout history (Perez, 2003). We encourage future research further investigate the economic consequences of such cycles. Furthermore, our findings are based on the context of Chinese digital infrastructure development and corporate innovation efficiency. To enhance the generalizability of our results, future research could explore similar relationships in different countries and regions, taking into account variations in digital infrastructure, economic policies, and cultural factors. Finally, our heterogeneity analysis is preliminary and could be further expanded. Given the multifaceted characteristics of firms, future research could delve deeper into the diverse factors that influence innovation efficiency, such as firm size, industry sector, and technological intensity. By conducting a more nuanced heterogeneity analysis, researchers can uncover new insights into how different firms respond to technological advancements and changes in digital infrastructure.