Unraveling how digital transformation affects innovation capability in China’s smart manufacturing enterprises

Abstract

In the digital economy, digital transformation has become a crucial driver of innovative capability in smart manufacturing businesses. This study examines the influence of digital transformation on innovation capabilities and its underlying processes within publicly traded smart manufacturing companies in China. Utilizing data from Chinese listed companies spanning 1999–2023, we construct a digital transformation index via text mining. Employing fixed-effects regression models to control for industry, year, and regional heterogeneity, and supplemented by multiple robustness strategies—including substituting the dependent variable with citation count for patent applications, lagged regression, omitted variable control, and placebo tests—we analyze the data. The findings demonstrate that digital transformation significantly enhances the innovative capacity of smart manufacturing enterprises, a result robust to rigorous testing. Additionally, mediation effect analyses indicate that digital transformation indirectly fosters innovation by alleviating financing constraints. Furthermore, heterogeneity analysis reveals that this innovation effect is more pronounced in Eastern regions, non-state-owned enterprises, and technology-intensive sectors, underscoring the moderating role of regional resources, ownership structure, and technological infrastructure. This study provides micro-level empirical evidence on the role of digital transformation in advancing manufacturing innovation and offers strategic insights for policymakers regarding digital infrastructure deployment and tailored guidance for enterprise transformation.

Introduction

The landscape of global manufacturing is undergoing a structural transformation, driven by the integration of digital technologies like artificial intelligence, industrial big data, and cyber-physical systems¹. This transformation extends beyond mere technological upgrades to reshape paradigms of value creation, casting smart manufacturing enterprises as pivotal actors within industrial innovation ecosystems².In China, this transformation is propelled by the synergy between policy and industry, where state-led investments in digital infrastructure and pilot smart factories drive adoption at the enterprise level. The integration of digital technologies is fundamentally reshaping innovation paradigms among smart manufacturing enterprises worldwide³. These entities, characterized by cyber-physical integration and adaptive automation, leverage digital transformation (DT) not merely for operational efficiency but as a strategic catalyst to reconfigure innovation capabilities. These capabilities span both exploitative innovation and exploratory innovation, establishing DT as the cornerstone of sustainable competitive advantage within hyperdynamic industrial ecosystems. National circumstances inevitably mold implementation paths, as evident in emerging economies where the symbiosis between policy and industry accelerates the diffusion of DT⁴, yet the core DT-innovation capability (IC) nexus transcends geographical boundaries.

Contemporary deployments of digital technologies in smart manufacturing manifest across three interdependent dimensions: technological enablers, organizational adaptability, and ecosystem interconnectivity. At the technological level, the integration of digital twins, edge computing, and interoperable platforms facilitates real-time data streaming, which underpins predictive R&D cycles and rapid prototyping. This, in turn, compresses innovation timelines by accelerating the translation of data-driven insights into actionable R&D processes⁵.At the organizational level, such capabilities as absorptive capacity and innovation flexibility enable firms to dynamically reconfigure resources, converting digital inputs into novel outputs even amid risks of imitation⁶.At the ecosystem level, industrial platforms orchestrate cross-border knowledge spillovers and co-creation networks, thereby illustrating how open innovation architectures transform users into active innovation partners⁷. Collectively, these dimensions establish digital technologies as a multi-layered engine for innovation capability evolution.

Digital technologies fundamentally reconfigure the innovation logic of smart manufacturing firms by driving a symbiotic transformation in cognitive frameworks and operational architectures⁸. Where traditional innovation relied on heuristic experience and segmented R&D processes, digital technologies embeds algorithmic rationality into organizational cognition—transforming real-time data streams from cyber-physical systems into predictive intelligence that anticipates market shifts and production constraints. This cognitive shift facilitates proactive innovation orchestration, in which AI-driven simulations preemptively test product iterations within virtual market environments, thereby shortening the traditional trial-and-error cycle⁹. Operationally, interoperable platforms dissolve boundaries between design, manufacturing, and customer engagement, converting linear innovation pipelines into dynamic feedback loops. Production anomalies detected by IoT sensors immediately trigger redesign protocols, while usage pattern analytics directly inform next-generation solutions—effectively merging exploration and exploitation activities into a continuous innovation continuum.

This cognitive shift facilitates proactive innovation orchestration, in which AI-driven simulations preemptively test product iterations within virtual market environments, thereby shortening the traditional trial-and-error cycle¹⁰. Unlike transient technological advantages, digital technologies institutionalizes innovation as an embedded organizational discipline wherein data liquidity enables rapid resource recombination across global R&D networks, architectural flexibility transforms cost constraints into mass customization opportunities, and ecosystem embeddedness leverages distributed co-creation to externalize innovation risks. Consequently, firms transition from developing discrete innovations to cultivating self-reinforcing innovation ecosystems¹¹—where predictive maintenance algorithms autonomously optimize production parameters using real-time performance data, while digital twins simulate emergent consumer behaviors to guide R&D investments¹². This perpetual adaptation capability establishes algorithmic advantage as the new competitive core, rendering innovation an inseparable function of daily operations rather than episodic initiatives.

Literature review

Digital transformation

Propelled by the advancement of the digital economy, digital technologies have become deeply integrated with products and services, fundamentally transforming conventional innovation paradigms¹³. Rooted in digital concepts and thinking, digital technology innovation leverages a modern knowledge system grounded in both digital technology and physical components. This form of innovation enables the creation of innovative outputs by upgrading, enhancing, or fostering emerging entities¹⁴. Yoo et al. (2010) pioneered the definition of digital technology innovation, characterizing it as the process wherein enterprises utilize digital technology to transform existing products, production processes, and business models¹⁵. Fichman et al. (2014) further stressed that digital technology innovation entails enterprises introducing technological and organizational changes via digital technology to adapt to new economic development models¹⁶.

Digital technology innovation focuses on the creation and development of novel digital tools, platforms, or methodologies themselves. It represents breakthroughs in technology, developing a more efficient blockchain protocol, creating advanced IoT sensors, or pioneering a revolutionary cloud architecture. The primary goal of digital technology innovation is technological advancement and capability expansion. Conversely, digital transformation represents a strategic, organization-wide process wherein existing or emerging digital technologies are leveraged to fundamentally alter organizational operations, customer value delivery, and market competitiveness. This process necessitates profound modifications to business models, operational processes, organizational culture, and customer experiences. Digital transformation is widely recognized as a catalyst for change across diverse contexts, particularly within the business sphere, and exerts influence on all facets of human life through technological applications. The concept of digital transformation must be distinguished from digitization (sometimes termed “digitalization” in scholarly literature)¹⁷. Crucially, digital transformation denotes organizational changes arising from digital technologies, while digitization refers to converting information from analog to digital formats and automating processes using information technologies¹⁸. The term “transformation” implies the organizational capability to comprehend and implement requisite actions in response to novel technologies, distinct from mere incremental change¹⁹.

Innovation is indispensable for perpetuating a firm’s core competitive advantage and serves as a principal engine of national economic growth²⁰. Against the backdrop of the rise of digital technologies anchored in artificial intelligence and big data, DT has emerged as a critical strategic imperative for traditional industrial firms to leverage new opportunities²¹. This trend is rooted in the reality that innovation enabled by DT has become the core driver of growth for traditional industries in the new economic era. Amid the advancement of digital technologies grounded in artificial intelligence and big data, DT has become a pivotal strategic priority for traditional industrial firms seeking to capitalize on emerging opportunities²².DT exerts a positive impact on firm operations, encompassing corporate environmental responsibility²³, green innovation²⁴, stock price crash risk reduction²⁵, and acquisition cost reduction²⁶.

Enterprise’s innovation capability

Innovation is widely recognized as a critical determinant of organizational sustainability, while also enhancing performance and cultivating global competitiveness²⁷. While innovation is widely recognized as a critical driver of organizational sustainability, performance enhancement, and global competitiveness, enterprises often struggle to translate innovation efforts into competitive advantages due to inherent challenges. Enterprise innovation confers competitive advantages²⁸,owing to the inherent attributes of long development cycles, high risk, and substantial capital investment in innovative projects, firms’ innovation output remains relatively low²⁹. Innovation activities are characterized by high risks and significant information asymmetry, which impairs firms’ ability to secure financing support for such initiatives, while elevated financing costs further deter their investment in innovation³⁰. Lower financing costs liberate resources for increased investment in innovation activities, consequently boosting firms’ innovation performance³¹.

Innovation constitutes the core driver of corporate development, thereby establishing intellectual capital (IC) as a vital capability for firm growth. Furthermore, IC functions as a critical enabler of organizational development, particularly in the context of sustainable growth among smart manufacturing firms³². IC manifests as specialized assets that integrate technology, products, intellectual assets, tacit knowledge, accumulated experience, and organizational competencies³³. IC encompasses the skills and knowledge essential for effectively absorbing, mastering, and improving current technologies, as well as for creating new innovations. R&D activities serve as a cornerstone of technological innovation within firms, constituting critical intangible investments in innovative endeavors³⁴.

Enterprises’ IC is shaped by a combination of internal and external determinants. Tax policies, for instance, significantly promote DT of private enterprises, consequently enhancing their innovation capital³⁵. Furthermore, research indicates that the liberalization of foreign investment policies notably improves firms’ environmental, social, and governance (ESG) performance, implying that a conducive policy milieu fosters holistic corporate advancement³⁶. Externally, government-led industrial support³⁷, intellectual property protection³⁸, green finance³⁹, and regional collaborative innovation policies have been found to serve as critical determinants. Internally, elements including R&D investment intensity⁴⁰, organizational structure flexibility⁴¹, and an innovation-oriented corporate culture⁴² exert direct influence on firms’ innovation drive and outcomes. Furthermore, a firm’s innovation capability is reflected not only in the quantity of innovation resources invested or outputs generated, but more critically in the quality and impact of its innovative outcomes. Therefore, the academic community has increasingly adopted quality-oriented metrics such as patent citation counts to measure an enterprise’s capacity to generate breakthrough and high-impact innovations. Patents serve as a significant indicator for measuring innovation capability, particularly in terms of innovation quality and its dynamic evolution⁴³. Among the commonly used proxies for assessing innovation capability, patent forward citations have been widely adopted as a standard metric⁴⁴. Specifically, forward citations—which track the number of times a patent is cited within a defined period—function as a sophisticated yet practical quantitative indicator, capable of capturing both the novelty of a technology and its market acceptance. As such, they enable a comprehensive evaluation of the two critical dimensions of innovation capability: technological evolution and market value⁴⁵.

Smart manufacturing is regarded as a cornerstone of the next industrial revolution. Unlike other fields, the inherent conditions for process innovation within smart manufacturing differ from those of product innovation, thereby necessitating tailored approaches for innovation processes in smart manufacturing⁴⁶. Hall et al. (2009) studied the relation between firm size, R&D intensity and investment in equipment, and the likelihood of product and process innovation⁴⁷. Zeng et al. (2010) examined the relationship dynamics among firm-to-firm collaboration, partnerships with intermediary institutions, and alliances with research organizations⁴⁸. Lassen and Laugen (2020) investigated the impact of both internal and external collaborative efforts on innovation magnitude⁴⁶. Collectively, prior studies have offered empirically validated insights into the innovation behavioral patterns and capabilities inherent within Smart Manufacturing Enterprises. However, relevant studies remain merely adequate, necessitating heightened focus on comprehensive investigations in this domain.

The impact of digital transformation on innovation capability

Fundamentally, enterprise innovation constitutes the collective output of all organizational members. Innovation may be conceptualized as the integration of diverse forms of knowledge accumulation⁴⁹. Within digital environments, enhanced innovation depth and openness substantially boost the growth outcomes of new ventures. Digital leadership leverages digital platforms to enhance corporate innovation performance. This occurs primarily because digital leadership positively correlates with platform digital capability, which in turn exhibits a positive correlation with innovation performance. Consequently, digitally-enabled leadership serves as a driver of enhanced firm innovation performance⁵⁰. Furthermore, business model innovation acts as a mediating factor in this relationship, whereas the breadth of innovation openness exerts minimal influence on new ventures. Additional studies demonstrate that firms stimulate innovation via approaches like open network innovation implementation, which subsequently promotes organizational management innovation and augments firm-level human capital⁵¹. Comprehensive DT drives innovation more effectively than the mere application of digital technologies. Given that enterprise innovation is constrained by environmental factors and regional disparities, the impact intensity of DT on innovation performance exhibits variation. The internationalization process subtly influences corporate innovation performance, and firms integrate innovation strategies with their internationalization plans. Analysis of high-tech enterprise samples reveals that enterprise DT effectively stimulates innovation. During internationalization strategy implementation, DT persistently exerts a significant positive effect on firm innovation⁵².

Regarding operational pathways, DT effectively mitigates financing limitations and lowers capital costs, consequently fostering increased innovation investment. From an information asymmetry standpoint, DT diminishes financing constraints originating from information gaps and concurrently lowers financing costs⁵⁰. Additionally, by leveraging digital technologies, DT improves information matching efficacy, thus substantially reducing financing risks and significant losses associated with unsuccessful financing attempts⁵³. From a cost-efficiency viewpoint, digital technologies decrease operational expenditures, curtail resource wastage, and optimize resource allocation, consequently lowering innovation investment risks and incentivizing higher investment levels⁵⁴. Simultaneously, DT promotes knowledge spillovers and restructures the allocation of innovation factors, leading to enhanced innovation efficiency. The adoption of information technology serves as a pivotal driver of innovation efficiency gains. By leveraging digital platforms, firms can lower the costs of technological innovation, break down inter-organizational knowledge barriers, and drive efficiency enhancements⁵⁵. Concurrently, the integration of advanced digital technologies in real-economy sectors optimizes production factor allocation, enhances capital-labor coordination, and boosts the efficiency of innovation output⁵⁶.

Extant literature has established a comprehensive understanding of how DT fundamentally enhances IC through multifaceted mechanisms—notably by alleviating financing constraints, reconfiguring knowledge management systems, and activating dynamic organizational capacities. However, this robust body of research exhibits a critical gap regarding smart manufacturing enterprises, where the unique operational ecosystem imposes distinct innovation imperatives. Unlike conventional settings, smart manufacturing enterprises face technology-physical system interdependencies, legacy integration complexities, and specialized innovation requirements that fundamentally alter DT-IC dynamics. Current frameworks inadequately address these contextual specificities, particularly in mediating mechanisms like adaptive resilience thresholds and domain-specific capability metrics. Consequently, while DT’s universal impact on IC is well-documented, its manifestation in smart manufacturing contexts remains underexplored, necessitating targeted investigations into sector-specific pathways, constraints, and performance moderators unique to Industry 4.0 environments.

Theoretical analysis and research hypotheses

Theoretical hypothesis

The accelerated progression of next-generation information technologies, marked by deep economic integration of innovations including cloud computing, big data, the Internet of Things, artificial intelligence, and mobile communications, enables technological innovation to progressively lower expenses related to data collection, processing, and analysis. Consequently, organizations gain enhanced capacity to efficiently harvest and utilize high-value data for production management optimization⁵⁷. To take the continuous innovation of digital technologies as a case in point, this has accelerated the widespread adoption and application of the Industrial Internet. Such advancement can reduce enterprises’ costs associated with production, operation, and innovation, while enhancing the production efficiency, economic returns, and innovation efficiency within the manufacturing sector⁵⁸. Secondly, DT empowers innovation entities to effectively utilize digital technologies for digital innovation in areas such as products and production processes through these aspects: alleviating the temporal and spatial constraints in the matching of innovation resources¹⁵, increasing the capacity of data transmission and storage⁵⁹, enriching the categories of data acquisition, enhancing the level of integration of innovation resources⁶⁰, and improving R&D decision-making capabilities⁶¹. Accordingly, investment in information technology facilitates smart manufacturing enterprises in developing digital infrastructure and applying advanced technological tools. This helps enterprises better identify potential market opportunities, optimize product design, and improve operational efficiency, thereby enhancing the performance level of digital innovation in smart manufacturing enterprises.

Digital transformation is essentially a systematic change in which enterprises reconstruct the logic of value creation through digital technologies. At its core, it involves the deep integration of digital technologies with business scenarios to develop new digital business models and optimize value creation processes, ultimately enhancing the efficiency of value creation and distribution within enterprises⁴⁹. This transformation not only reshapes the operational environment of enterprises but also makes the development of adaptive and innovative capabilities through the absorption and generation of new knowledge a strategic priority for organizations⁶². Moreover, it provides a core driving force for the cultivation of innovative capabilities. The dynamic capabilities theory offers a critical theoretical perspective for analyzing the intrinsic relationship between these elements.

As a core theory explaining how enterprises sustain competitive advantages in dynamic environments, dynamic capabilities theory emphasizes the integration, creation, and reconfiguration of internal and external competitive resources or advantages to respond to and proactively shape rapidly changing environments⁶³. From a knowledge-based perspective, the essence of dynamic capabilities lies in an enterprise’s ability to acquire, integrate, and reconfigure knowledge resources. Digital transformation precisely provides the technological vehicle and implementation context for the realization of these capabilities. In the process of digital transformation, technologies such as artificial intelligence and cloud computing break down traditional information barriers within organizations⁶⁴, enabling the efficient flow and real time sharing of internal and external knowledge. This significantly enhances the efficiency with which enterprises absorb external new knowledge and transform internal existing knowledge. Such capabilities for knowledge absorption and transformation are the foundational prerequisites for the formation of innovative capabilities. As the core of enterprise competitiveness, innovative capability refers to the ability of an enterprise to mobilize various resources, integrate knowledge and skills, and consistently generate new ideas, products, or services⁶⁵. By empowering the enhancement of dynamic capabilities, digital transformation constructs a complete pathway for the cultivation of innovative capabilities. On one hand, digital transformation drives the upgrading of resource integration models, enabling enterprises to rapidly aggregate internal and external innovation elements such as technology, talent, and capital through digital platforms, thereby providing a resource foundation for innovation activities. On the other hand, the reconstruction of business processes driven by digital technologies reduces the cost of trial and error in innovation and accelerates the innovation cycle from idea generation to prototype development and market validation. This shifts innovative capabilities from sporadic outputs to systematic cultivation.

Therefore, digital transformation does not directly equate to the enhancement of innovative capabilities. Instead, it establishes a transmission mechanism that moves from digital technology empowerment to dynamic capability upgrading and finally to innovative capability cultivation, thereby providing the necessary environmental support, resource guarantees, and process momentum for the formation and strengthening of innovative capabilities. As the core link between the two, the practical implementation of dynamic capabilities in digital contexts is the key to translating digital initiatives into tangible innovation outcomes. Based on this analysis, the following hypothesis is proposed:

H1

DT has a significant positive impact on the average IC of enterprises.

Different regions exhibit significant disparities in economic conditions and production environmental conditions. Such disparities may also influence the impact of enterprise DT on enterprise IC. While traditional theory suggests that pioneering regions like Eastern China are more likely to unleash the innovation potential of DT due to well-developed digital infrastructure⁶⁷, unbalanced development theory provides an explanation for the “reverse advantage” of latecomer regions. Compared with the Eastern region, the Central and Western regions have a relatively lower starting point in DT, thus possessing greater room for development and potential in this regard. Enterprises in the Eastern region are relatively more mature in DT, with less pronounced disparities in infrastructure, technological level, and other aspects. Consequently, the impact of DT on enterprise IC may be less significant than in the Central and Western regions. Based on this analysis, the following hypothesis is proposed:

H2

The promoting effect of DT on enterprise IC is significantly stronger in Central and Western regions than in the Eastern region.

In China, state-owned enterprises (SOEs) constitute the cornerstone of all sectors within the national industry⁶⁷. Leveraging their advantages in resource endowments, SOEs have gained a first-mover advantage in the scale of digital investment and play a pivotal role in the national DT⁶⁸. Yet, their administrative governance structures may undermine the effectiveness of such transformation. Compared to private or foreign firms, SOEs generally carry more substantial policy burdens⁶⁹. Chinese SOEs also shoulder heightened social duties and obligations, implying that normative pressures impose stronger limitations on them than on other organizational forms. SOEs are mandated to discharge both political duties and societal responsibilities, including public welfare initiatives, environmental sustainability efforts, and poverty reduction programs. These characteristic patterns of action lead SOEs to encounter greater cognitive constraints and exhibit stronger behavioral inertia relative to other organizations, which consequently constrains their progress and DT⁷⁰. In contrast, non-SOEs primarily secure resources through their own endeavors within market competition⁷¹. To establish a foothold in fierce market competition, these enterprises must continuously enhance their innovation capabilities, enabling them to adapt swiftly to the external environment and thus pursue sustainable development. Notably, by virtue of their flexible decision-making mechanisms, non-SOEs exhibit greater efficiency in DT⁷²—evidenced, for instance, by their prompt adoption of new digital marketing models and digital production management systems. Based on this analysis, the following hypothesis is proposed:

H3

The strength of DT’s promoting effect on IC is greater for non-SOEs than for SOEs.

The impact of DT varies significantly across different types of enterprises. Technology-intensive manufacturing firms typically possess abundant knowledge capital and human resources, enhancing their capacity for DT implementation⁷³. For instance, in navigating the relationship between digital technologies and global market competitiveness, Huawei has leveraged DT to strategically align operational efficiency enhancements with optimized market positioning, thereby generating significant value creation in international markets. Conversely, non-technology-intensive enterprises rely heavily on low-cost labor, exhibit limited digital system adoption, and face constraints in technological assimilation. These firms encounter substantial upfront investment costs during initial DT stages, operate under relatively simplistic investment decision-making mechanisms, and struggle to achieve effective integration of digitalization. Meanwhile, high-technology enterprises depend more heavily on drivers related to organizational flexibility and dynamic capabilities. These dynamic capabilities afford high-technology firms greater adaptability to external driving factors⁷⁴. In contrast, low-technology enterprises rely predominantly on drivers associated with stability-oriented capabilities. Consequently, DT exerts a more pronounced impact on technology-intensive firms. Based on this analysis, the following hypothesis is proposed:

H4

Compared with non-technology-intensive firms, DT exerts a more pronounced positive effect on the IC of technology-intensive enterprises.

Research design

Sample selection

This study selects A-share listed companies in China from the CSMAR database for the period 1999–2023 as the initial sample. Merging the DT dataset with the IC dataset based on firm name and year, we retain only observations consistent across datasets. After screening, 1087 valid samples of smart manufacturing enterprises are obtained, forming the basis for empirical analysis. The primary rationale for selecting the 1999–2023 period is fundamentally rooted in the core objective of our study: to unravel the dynamic and evolving influence of digital transformation on innovation capability. Conceptualizing DT not as a binary event but as a protracted, multi-stage process of technological adoption and organizational adaptation necessitates a long-term observational window.

Model construction

To verify the above analysis, this study employs a multidimensional fixed-effects regression model to examine the impact of DT on enterprise innovation. The baseline regression equation is as follows:

$In(Cit{2_{i,t}})={\alpha _0}+{\beta _1}D{T_{i,t}}+{\gamma _t}{X_{i,t}}+{\mu _i}+{\lambda _t}+{\eta _j}+{\varepsilon _{i,t}}$

Where: $In(Cit{2_{i,t}})$ represents IC of smart manufacturing enterprises,$D{T_{i,t}}$represents digital transformation,${X_{i,t}}$is a vector of control variables,${\gamma _t}$ is the coefficient vector for the control variables, $\mu _{i}$ represents province fixed effects,${\lambda _t}$represents year fixed effects(controlling for time-varying macro shocks), ${\eta _j}$represents industry fixed effects(where $j$ denotes the industry of firm $i$ , controlling for industry-invariant traits), ${\varepsilon _{i,t}}$ is the random error term. To ensure data stability, relevant measurement data are logarithmically transformed.

To verify the pathway through which DT affects IC, we introduce the mediating variable $Me{d_{i,t}}$ and employ the three-step mediation method, establishing the following regression equations. Additionally, building on this foundation, the Bootstrap method was employed to conduct mediation effect testing, thereby enhancing the robustness of the mediation analysis.

Regression of the mediator on DT:

$Me{d_{i,t}}={a_0}+{a_1}D{T_{i,t}}+y{X_{i,t}}+{u_i}+{\eta _j}+{\varepsilon _{i,t}}$

Regression of IC on DT and the mediator:

$\ln (Cit{2_{i,t}})={\theta _0}+{\theta _1}D{T_{i,t}}+{\theta _2}Me{d_{i,t}}+{\gamma _t}{X_{i,t}}+{\mu _i}+{\eta _j}+{\varepsilon _{i,t}}$

Where: ${\theta _0}$ are intercept terms, $\theta _{1}$ represents the direct effect coefficient of DT, $\theta _{2}$ represents the coefficient of the mediator’s effect, ${\eta _j}$ represents industry fixed effects, where $j$ denotes the industry to which firm $i$ belongs (controlling for time-invariant industry characteristics). Other variables and parameters are defined as in the baseline model.

Variables definition

(1) Dependent variable (IC of smart manufacturing enterprises): In academic research, IC of smart manufacturing enterprises is a multidimensional concept. This study focuses on its quality dimension, measured by the number of patent citations (Cit2). The rationale is that patent citation frequency directly reflects industry recognition, technological radiation power, and commercial application potential of technological achievements. Highly cited patents often contain more breakthrough ideas or solve common industry technical challenges, providing a more precise measure of the “quality” of IC than patent application counts alone. To meet the distributional assumptions of the econometric model, the natural logarithm of the patent citation count (LnCit2) is used, making the indicator more compatible with the normality assumption of linear regression error terms. Data for this indicator were derived from the raw dataset encompassing innovation capability, quality, and efficiency of listed companies. After sample screening (focusing on smart manufacturing firms) and cross-dataset matching (integration with DT data), it is incorporated into the analysis, ensuring consistency and validity of observations.

(2) Core explanatory variable (DT): Leveraging the 1999–2023 China A-share listed company dataset, this study constructs a text mining-based index. The methodology focuses on Management Discussion and Analysis sections of annual reports to develop a tripartite lexicon covering digital technology application, business process change, and organizational capability upgrade domains, incorporating over 50 core keywords. Following preprocessing steps including word segmentation and noise removal, keyword density is calculated as keyword frequency divided by total words per annual report. This standardization controls for document length heterogeneity, yielding the Digital Transformation index. Higher values indicate stronger strategic prioritization of digital and collaborative transformation initiatives.

(3) Mediating variable: Operating Expense Ratio (OER): Leveraging the Operating Expense Ratio as a mediating variable, this metric quantifies cost efficiency in core firm operations. Within the digital transformation to innovation capital pathway, OER functions as a crucial cost-resource transmission mechanism. It manifests through two interconnected dynamics. Firstly, digital technologies potentially compress operating expenses via process automation and precision marketing, enabling resource reallocation toward innovation-driven efficiency improvements. Conversely, initial-phase technological deployments such as digital platform development may temporarily elevate expense ratios, demonstrating a strategic investment dilemma characterized by the dynamic equilibrium between short-term cost expenditures and long-term innovation outcomes. These combined effects delineate digital transformation’s complex operational cost implications. The study derived OER data from annual financial statements of Chinese listed companies, incorporating industry fixed effects to control for sectoral variations including inherently elevated sales expenditures in service industries to ensure metric robustness in mediation effect testing.

(4) Control variables: This study selects the following control variables to isolate the influence of other factors on the core relationship:

Relative Capital Accumulation rate (RCA): Measures the growth rate of a firm’s capital stock relative to industry or economic benchmarks. It evaluates how efficiently an enterprise expands its productive assets compared to peers over time.

Capital Intensity (CAP): Quantifies capital investment per unit of output, calculated as total capital assets divided by total revenue or production volume. Higher values indicate greater reliance on physical assets rather than labor for value creation.

Total Assets (TA): Quantifies organizational scale using total book value of assets as the primary metric, reflecting a firm’s resource endowment and financial capacity. This measure is typically logarithmically transformed in empirical models to mitigate skewness and enable elasticity interpretation.

Fixed Assets to Total Assets Ratio (FATAR): Computes the proportion of immovable, long-term productive assets (PP&E) relative to a firm’s total asset base. This ratio reflects strategic commitments to tangible operational infrastructure.

Intangible Assets to Total Assets Ratio (IATAR): Measures intellectual capital and non-physical resources (patents, software, brand value) as a percentage of total assets. Elevated ratios signal knowledge-driven competitive advantages in innovation-centric industries.

Institutional Investors’ Shareholding Percentage (IISP): Measures the proportion of a firm’s total shares held by institutional investors. This metric reflects institutional monitoring intensity and influences corporate governance effectiveness through active engagement and voting power.

Shareholdings in Other Financial Institutions (SOFI): Quantifies a firm’s equity investments in banks, insurance companies, or other regulated financial entities. Such cross-holdings facilitate strategic synergies but may introduce systemic risk exposure and regulatory complexity.

Independent Directors’ Network Centrality (IDNC): Assesses an independent director’s influence within interlocking directorate networks using graph theory metrics. Higher centrality enhances information access and resource acquisition capabilities but may reduce board independence due to reputation concerns.

Board Size (BS): Indicates the total number of directors serving on a firm’s governing board. Optimal sizing balances diverse expertise against coordination costs, with larger boards potentially diluting monitoring effectiveness while expanding advisory resources.

Empirical analysis

Descriptive statistics

According to Table 1, the mean value of enterprise IC (LnCit2) is 0.536 with a standard deviation of 0.283, indicating significant variation in IC among sample firms. The core explanatory variable, digital transformation (DT), has a mean of 0.036 and a standard deviation of 0.054, reflecting divergence in the degree of DT across enterprises. Among financial characteristics: The mean RCA is 2.451 with a large standard deviation of 2.048, showing significant differences in capital accumulation levels among firms. The mean CAP is 22.307 with a relatively small standard deviation of 1.229, suggesting industry capital intensity is relatively stable. The mean TA is 0.033 (std. dev. 0.031), indicating relatively small size differences among sample firms. The mean FATAR is 0.536 (std. dev. 0.283), reflecting heterogeneity in fixed asset allocation. The mean IATAR is 0.036 (std. dev. 0.054), revealing divergence in intangible asset investment.

Table 1 Descriptive statistics.

Correlation analysis

Table 2 presents the Pearson correlation coefficient matrix for the main variables. The correlation matrix reveals several statistically significant relationships among key variables. IC demonstrates a strong positive association with SA (β = 0.292) while exhibiting significant negative correlations with total assets (TA) (β = -0.273) and FATAR (β = -0.07). DT shows positive linkages with TA (β = 0.380) and IATAR (β = 0.155), but negative relationships with CAP (β = -0.076), SA (β = -0.102), and FATAR (β = -0.157). Notably, RCA correlates positively with both CAP (β = 0.377) and TA (β = 0.061), reflecting capital-intensive expansion patterns. The robust inverse correlation between SA and IATAR (β = -0.145) suggests potential strategic trade-offs between agility and intangible resource allocation.

Table 2 Correlation analysis. ^*p < 0.1, ^**p < 0.05, ^***p < 0.01.

Baseline regression

Table 3 presents the baseline regression results. The baseline regression models systematically evaluate DT 's impact on IC while progressively addressing confounding factors. Model (1) establishes the baseline relationship with only the core explanatory variable (DT), revealing a statistically significant positive effect (β = 0.3247, p < 0.1). When control variables in Model (2), the DT coefficient increases substantially to 1.0063 (p < 0.01). Model (3) incorporates industry fixed effects to control for sectoral heterogeneity, maintaining DT’s statistical significance (β = 0.7817, p < 0.1) while reducing its magnitude by 22.3% versus Model (2). This attenuation demonstrates that sector-specific characteristics partially mediate DT’s innovation returns. The fully specified Model (4) with year and region fixed effects yields the most reliable estimate: DT retains strong significance (β = 0.4731, p < 0.01) after neutralizing temporal trends and geographical variations, confirming its robust positive effect on IC accounts for 47.31% of a standard deviation improvement in innovation capital. These results support Hypothesis 1.

Table 3 Baseline regression results. ^*p < 0.1, ^**p < 0.05, ^*** p < 0.01.

Robustness tests

Substituting dependent variable with citation count for patent applications

To test the robustness of the baseline conclusion, the dependent variable is replaced with Citation Count for Patent Applications. Results in Table 4 reveal a consistently positive relationship between DT and innovation impact across progressively controlled specifications. The baseline Model (1), containing only the core explanatory variable, shows a significant DT effect (β = 0.8649, p < 0.05), confirming DT’s fundamental association with innovation quality even in parsimonious specification. When control variables are introduced in Model (2), DT exhibits a substantially enhanced coefficient (β = 2.5380, p < 0.01), suggesting that organizational characteristics like asset structure and intensity function as significant suppressors of DT’s true innovation impact. Model (3) incorporates industry fixed effects, maintaining DT’s strong significance (β = 2.3054, p < 0.01), demonstrating that sectoral technological regimes do not diminish DT’s positive role in innovation capability. The comprehensive Model (4) with year and region fixed effects preserves DT’s robust significance (β = 1.4994, p < 0.01), confirming DT’s persistent positive association with innovation quality after accounting for temporal trends, geographical innovation disparities, and industrial contexts. This trajectory is reinforced by the consistently negative TA coefficients (Model 4: β = −0.0606, p < 0.05) confirming smaller firms’ advantage in innovation efficiency, while the significantly negative IATAR coefficient (β = −1.1676, p < 0.05) suggests strategic reallocation toward quality-focused innovation in intangible-intensive firms. The dramatic R-squared increase from 0.1506 to 0.5673 underscores the critical importance of accounting for macro-contextual influences in innovation studies.

Table 4 Robustness test (Replacing dependent variable with citation count for patent applications). ^*p < 0.1, ^**p < 0.05, ^***p < 0.01.

Lagging core variable by one period

To mitigate reverse causality endogeneity, DT is lagged by one period for robustness testing (Table 5). The baseline Model (1) reveals no immediate effect (β = 0.0282, ns), indicating innovation outcomes require assimilation periods beyond the current fiscal cycle. Introducing capital dynamics and asset structure controls in Model (2) unveils a robust delayed positive relationship (β = 0.6475, p < 0.01), suggesting prior-year digital initiatives significantly enhance current innovation capacity after accounting for firms’ capital allocation patterns and resource constraints. This effect attenuates to insignificance in Model (3) with industry fixed effects (β = 0.3814, ns), revealing that sectoral adoption cycles mediate the transformation-innovation pathway. Crucially, the comprehensive Model (4) incorporating industry, year, and region fixed effects confirms DT_(t-1)’s persistent influence (β = 0.3941, p < 0.01). This final specification indicates that a one-unit increase in prior-year DT intensity corresponds to a 0.394 standard deviation improvement in current innovation capital, even after controlling for capital accumulation efficiency (RCA), production factor intensity (CAP), organizational scale constraints (TA), and strategic asset allocations. The significant negative coefficient of intangible assets (IATAR: β = −0.5007, p < 0.05) further suggests knowledge-intensive firms experience delayed innovation returns from digital investments, while the persistent negative TA coefficients reinforce smaller firms’ agility in converting technological investments into innovation outcomes.

Table 5 Robustness test (lagging core explanatory variable by one period). ^*p < 0.1, ^**p < 0.05, ^***p < 0.01.

Controlling for omitted variable bias

To alleviate omitted variable bias, we progressively include variables related to corporate governance and external monitoring, specifically IISP, SOFI, IDNC, and BS. Results in Table 6 reveals that DT significant positive impact on IC across progressively controlled specifications. Model (1) establishes the baseline relationship (β = 0.3247, p < 0.1), which substantially amplifies in Model (2) after incorporating capital dynamics and asset structure controls (RCA, CAP, TA, FATAR, IATAR), yielding a stronger DT coefficient (β = 1.0063, p < 0.01) that reveals conventional determinants suppress DT’s true innovation impact, particularly organizational scale constraints (TA: β = −0.0860, p < 0.01). Model (3) introduces governance controls, further enhancing DT’s effect (β = 1.1620, p < 0.01) while uncovering critical moderating mechanisms: institutional monitoring intensity (IISP: β = 0.0008, p < 0.1) positively influences innovation, whereas financial cross-holdings (SOFI: β = −0.1626, p < 0.1) and networked directors (IDNC: β = −0.1239, p < 0.01) exhibit significant innovation-inhibiting effects, indicating governance complexities constrain digital innovation. The attenuation in Model (4) with industry fixed effects (β = 0.9688, p < 0.01) demonstrates sectoral characteristics mediate approximately 16.6% of DT’s gross effect. Crucially, Model (5) with comprehensive spatiotemporal controls (industry/year/region FE) delivers the most reliable estimate (β = 0.5375, p < 0.01), confirming DT’s persistent innovation-enhancing effect after accounting for capital allocation patterns, asset composition trade-offs, governance structures, and contextual heterogeneity. Notably, the significantly negative IATAR coefficient (β = −0.5966, p < 0.01) in the final specification reveals intensifying tension between intangible assets and digital innovation outcomes, while IDNC maintains its negative influence (β = −0.0572, p < 0.01), suggesting board network centrality impedes digital innovation assimilation even after rigorous bias mitigation.

Table 6 Robustness test (Controlling for omitted variable bias). ^*p < 0.1, ^**p < 0.05, ^***p < 0.01.

Placebo test

To rule out interference from sample selection bias or random factors, a placebo test is conducted. DT indicator is randomly reassigned to sample firms using a random permutation algorithm (repeated 1,000 regressions), constructing a probability distribution of placebo estimated coefficients. The results show that the placebo estimated coefficients form a symmetrical bell-shaped distribution, peaking near zero (Fig. 1). This indicates that if the impact of DT were merely due to random factors, its coefficients should fluctuate around zero with high probability. However, the actual coefficient of DT in the baseline regression significantly deviates from the core interval of this distribution, lying in a tail region of extremely low probability density. This result demonstrates that the positive impact of DT on IC is not coincidental but reflects a genuine causal effect.

Fig. 1

Placebo test result.

Mechanism test

Given that SA represents the absolute value of financing constraints, higher SA magnitudes correspond to greater severity of financing constraints. The mechanism test (Table 7) substantiates that DT enhances IC primarily through alleviating financing constraints, with SA serving as a significant mediating channel. Model (1) confirms DT’s substantial reduction effect on financing constraints (β = −0.4575, p < 0.01). Model (2) establishes the innovation-inhibiting nature of financing constraints (SA: β = −0.1550, p < 0.01). Crucially, Model (3) demonstrates that while DT maintains a strong direct innovation effect (β = 0.7810, p < 0.01), the significant attenuation from Model (2)’s coefficient (0.7810 vs. the 1.1620 coefficient in prior specifications without SA) reveals that approximately 32.8% of DT’s total innovation impact operates indirectly through financing constraint mitigation. This mediation pathway remains robust after controlling for capital dynamics and asset structure, with the negative FATAR-IC relationship (β = −0.2394, p < 0.01) further confirming that fixed asset-intensive firms experience compounded innovation constraints beyond financing limitations. The results collectively establish that digital transformation functions as a dual innovation enabler: directly enhancing innovation capacity while indirectly creating financial flexibility through improved information symmetry and risk mitigation.

Table 7 Mechanism test (Causal steps approach). ^*p < 0.1, ^**p < 0.05, ^***p < 0.01. Building on the causal steps approach, bootstrap analysis with 500 replications confirms SA serve as a statistically significant mediating channel through which DT enhances IC. Results in Table 8 reveals that the indirect effect of DT via SA reduction is estimated at 0.174 (95% CI: [0.026, 0.380]), with bias-corrected confidence intervals excluding zero - indicating robust mediation at α = 0.05. This represents 17.3% of dt’s total innovation effect. Simultaneously, the direct effect remains substantial at 0.832 (95% CI: [0.347, 1.314]), demonstrating DT stimulates innovation through both SA-mediated financial flexibility and independent technological capability pathways. The trivial bias estimate (0.006) and narrow bootstrap standard error (0.090) further validate the mediation effect’s reliability, Establishing SA reduction as a complementary rather than competitive transmission mechanism.

Table 8 Mechanism test (Bootstrap approach). ^*p < 0.1, ^**p < 0.05, ^***p < 0.01.

Heterogeneity test

Building upon the core specifications and robustness tests, we conduct heterogeneity analyses across three dimensions: Regional Dimension, Enterprises Ownership Dimension, and Industry Technology Dimension. Detailed results are presented in Table 9.

Regional dimension

The regional heterogeneity analysis reveals significant divergence in digital transformation’s innovation impact across China’s major economic zones. In the Western region, DT demonstrates an exceptionally strong positive effect on innovation capability (β = 12.3855, p < 0.05), where each unit increase in digital transformation corresponds to a 12.4 standard deviation improvement in IC - likely reflecting leapfrogging opportunities in underdeveloped technological ecosystems and substantial government subsidies for digital infrastructure. Conversely, Eastern China exhibits a more moderate yet highly significant relationship (β = 0.5588, p < 0.01), indicating mature innovation systems yield consistent but diminishing marginal returns from digital investments. Central China shows no statistically significant DT-IC relationship (β = −1.3372, ns), suggesting transitional economic structures may experience capability-development bottlenecks that inhibit digital innovation absorption. Notably, CAP functions as a critical innovation driver in the West (β = 0.2033, p < 0.01) but not elsewhere, while intangible assets generate opposing regional effects: significantly negative in the East (IATAR: β = −0.6916, p < 0.01) versus positive in Central China (β = 2.7816, p < 0.05), implying fundamentally different knowledge-absorption capacities across regional innovation regimes. These findings collectively establish that digital transformation’s innovation returns are contingent upon regional development stages, with the strongest effects emerging in technologically lagging but policy-supported Western provinces, stable returns in advanced Eastern hubs, and neutral effects in transitional Central zones facing structural assimilation challenges.

Enterprises ownership dimension

The ownership-based heterogeneity analysis reveals significant differential impacts of DT on IC between SOEs and Non-SOEs. While DT significantly enhances innovation in both ownership types, its effect magnitude is substantially stronger in Non-SOEs (b = 0.5935, p < 0.05) compared to SOEs (β = 0.2598, p < 0.05), indicating private firms generate over twice the innovation return per unit of digital investment. This differential performance stems from distinct innovation pathways: SOEs demonstrate greater efficiency in capital accumulation (RCA: β = 0.0075, p < 0.01) but suffer significant innovation drag from capital intensity (CAP: β = −0.0196, p < 0.01) and intangible assets (IATAR: β = −0.8681, p < 0.05), suggesting bureaucratic inertia in technology assimilation. Conversely, Non-SOEs exhibit superior agility in leveraging digital transformation despite scale disadvantages (TA: β = −0.0472, p < 0.05), though their innovation is constrained by fixed asset investments (FATAR: β = −0.1948, p < 0.05). The results establish that ownership structure fundamentally mediates DT innovation returns, with Non-SOEs achieving greater efficiency in converting digital investments into innovation outcomes due to streamlined decision-making and market-driven incentives, while SOEs face systemic inefficiencies in technological absorption despite superior resource endowments.

Industry technology dimension

The technology intensity-based heterogeneity analysis reveals fundamentally divergent innovation pathways for digital transformation (DT) across industry types. In technology-intensive sectors, DT exhibits a robust positive impact on innovation capability (β = 0.4495, p < 0.01), indicating each unit increase in digital transformation yields a 0.45 standard deviation IC improvement - demonstrating the complementary relationship between digital infrastructure and technological innovation ecosystems. Conversely, non-technology-intensive industries show no statistically significant DT effect (β = −1.0781, ns), suggesting digital investments fail to generate innovation returns without underlying technological absorptive capacity. The analysis further identifies distinct innovation constraints: technology-intensive firms experience significant innovation drag from tangible asset commitments (FATAR: β = −0.1810, p < 0.05) and intangible asset intensity (IATAR: β = −0.5619, p < 0.05), reflecting resource allocation tensions in innovation-driven environments. Non-technology-intensive sectors display uniformly insignificant coefficients across control variables, indicating conventional innovation determinants lose explanatory power without technological foundations. These results establish that digital transformation serves as an innovation catalyst exclusively in technology-intensive contexts, where it synergizes with specialized knowledge bases and innovation ecosystems, while non-technology-intensive industries require foundational capability development before realizing digital innovation benefits.

Table 9 Heterogeneity test. ^*p < 0.1, ^**p < 0.05, ^***p < 0.01.

Conclusions and discussions

Conclusion of the study

This empirical study demonstrates that DT significantly enhances the IC of smart manufacturing enterprises in China, corroborating the global consensus on digital technologies reshaping innovation paradigms⁶⁶. Critical contextual heterogeneities emerge: DT’s innovation-enabling effect proves substantially stronger in Eastern and Western regions, whereas Central China exhibits attenuated outcomes due to fragmented digital infrastructure and legacy industry dominance. non-SOEs exhibit heightened responsiveness to DT through agile decision-making and market-driven mechanisms, while SOEs remain constrained by institutional inertia despite resource advantages. Technology-intensive firms leverage digital-technological complementarities to elevate IC, whereas non-technology-intensive enterprises encounter cost barriers and capability gaps. Crucially, the study reveals financing constraints as a core mediator—DT unlocks innovation capital by alleviating information asymmetry and operational risks, a mechanism particularly salient in transitional economies where the policy-finance nexus shapes innovation trajectories.

Theoretical implications

This research advances both DT and smart manufacturing enterprise literature while extending pertinent theoretical frameworks. Regarding its contribution to existing scholarship, the investigation uncovers that DT bolsters IC within smart manufacturing firms through direct mechanisms and the mediating effect of financing constraint alleviation. This finding significantly broadens the theoretical comprehension of how digital technologies reconfigure innovation processes in technology-intensive manufacturing environments. Theoretically, the identified heterogeneities across regions, ownership structures, and industries furnish more granular perspectives for comprehending DT’s impacts. Contrary to the “late-mover advantage” hypothesis for late-developing regions, the stronger effects observed in China’s eastern and western regions underscore the importance of aligning DT with local resource endowments. Similarly, the differential responses of non-state-owned enterprises and technology-intensive enterprises highlight the necessity of integrating institutional and capability perspectives in research on DT.

Policy implications

From a practical standpoint, this work delivers implementable frameworks for policymakers and smart manufacturing enterprises to strategically manage their DT journeys. Policymakers should prioritize regionally tailored digital infrastructure strategies that account for developmental disparities: Eastern regions must capitalize on mature innovation ecosystems by deepening digital-industrial integration⁷⁵, while Western regions should leverage latecomer advantage dividends through targeted resource allocation to overcome infrastructure bottlenecks and accelerate innovation capability convergence⁷⁶. Central regions require focused interventions to resolve dual constraints of fragmented digital foundations and protracted traditional industry transitions, necessitating cross-regional coordination mechanisms to unlock latent innovation potential. Concurrently, ownership-specific policy frameworks are critical—for non-SOEs, dismantling institutional barriers will enable their agile market responsiveness to capture innovation opportunities⁷⁷, whereas SOEs need governance restructuring to streamline decision hierarchies⁶⁹, mitigate administrative inertia in digital projects, and align initiatives with market-driven innovation imperatives.

Furthermore, enterprises can harness data-driven operational architectures to alleviate financing constraints⁷⁸—enhancing information transparency through real-time analytics broadens funding access and reduces capital costs, thereby securing stable innovation investment. And enterprises should further adopt differentiated transformation pathways aligned with intrinsic attributes: technology-intensive enterprises ought to embed digital tools like industrial big data and digital twins into R&D workflows to accelerate knowledge refinement and innovation deployment, while non-technology-intensive counterparts should utilize policy subsidies or collaborative models to offset initial transformation costs, proactively developing integration scenarios for digital technologies and conventional production systems⁷⁹ to prevent inefficient resource allocation.

Limitations and future work

This study also has limitations. The sample is restricted to Chinese A-share listed companies, limiting generalizability to non-listed or international firms. The DT index, built through text mining of annual reports, may not fully capture the depth of technological integration, such as actual deployment of IoT or AI systems. While financing constraints are identified as a key mediator, other pathways like organizational culture or knowledge spillovers require further exploration. Future research could expand the sample for global comparisons, refine DT measurement with objective metrics, and investigate interactions between institutional factors and DT in shaping IC.