Digital Innovation and Cost Stickiness in Manufacturing Enterprises: A Perspective Based on Manufacturing Servitization and Human Capital Structure

Abstract

This paper examines the effect of digital innovation on cost stickiness in manufacturing firms, focusing on the underlying mechanisms and contextual factors. Using data from Chinese A-share listed manufacturing firms from 2012 to 2023, we find that, first, for each one-unit increase in the level of digital technology, the cost stickiness index of enterprises decreases by an average of 0.4315 units, primarily through digital process innovation and digital business model innovation, whereas digital product innovation does not exhibit a statistically significant impact. Second, manufacturing servitization and the optimization of human capital structure are identified as key mediating mechanisms. Digital innovation promotes servitization by transitioning firms from product-centric to service-oriented business models, thereby reducing fixed costs and improving resource flexibility. It also optimizes human capital by increasing the proportion of high-skilled employees and reducing labor adjustment costs. Third, the effect of digital innovation on cost stickiness is found to be heterogeneous. Firms with high financing constraints benefit more from the cost-reducing effects of digital innovation due to improved resource allocation efficiency. Additionally, mid-tenure executives are more effective in leveraging digital innovation to mitigate cost stickiness, as they balance short-term performance pressures with long-term strategic investments. These findings contribute to the understanding of how digital transformation reshapes cost behavior in manufacturing and provide insights for policymakers and firms seeking to achieve sustainable development through digital innovation.

1. Introduction

The manufacturing industry has long been a cornerstone of national economic development, driving economic growth, generating employment, and enhancing international competitiveness. However, in the 21st century, China’s population is aging rapidly, leading to the decline in its traditional demographic dividend. Rising costs of labor, land, and other production factors, coupled with increasing trade protectionism, stricter environmental regulations, and the restructuring of global industrial chains, have significantly compressed profit margins in the manufacturing sector. These economic and environmental pressures collectively challenge the long-term sustainability of traditional manufacturing models, forcing enterprises to transition toward more resilient and socially responsible modes of operation. With raw material costs constituting an increasing proportion of total expenses, industry-wide average gross profit margins have declined, and even minor cost fluctuations can be amplified through leverage effects, causing substantial profit volatility. Given these circumstances, manufacturing enterprises face unprecedented cost pressures and management challenges [1]. To maintain competitiveness and achieve high-quality development, firms must implement life cycle cost management that enhances the flexibility of their cost control strategies. In the process of enterprise cost management, the phenomenon of cost stickiness has become increasingly prominent. Cost stickiness refers to the asymmetric adjustment of costs, whereby the decrease in costs when business activity declines is less pronounced than the increase in costs when business activity rises. This asymmetry amplifies profit volatility in the face of demand fluctuations and external shocks, thereby increasing operational risk. Consequently, effectively reducing cost stickiness and enhancing the agility of cost adjustments have become critical issues for the high-quality development of the manufacturing sector. Meanwhile, with the rapid advancement of digital technologies, particularly the widespread adoption of big data, cloud computing, artificial intelligence, and the Internet of Things, manufacturing enterprises are experiencing a “digital dividend”. Firms should fully leverage the opportunities brought by digital transformation through emerging technologies, shifting from the traditional “demographic dividend and resource-driven” model to a “digital dividend and innovation-driven” strategy. By developing an intelligent and collaborative cost management system, firms can enhance the resilience of their value chains and successfully navigate the profound transformations reshaping the industry.

Digital technology is a key driver of the ongoing scientific and industrial revolutions, permeating every aspect of production, consumption, and distribution. By reshaping the allocation of production factors, digital technology facilitates the transition from a traditional factor-driven economy to an innovation- and data-driven economy, fostering new pathways for economic development. The 14th Five-Year Plan emphasizes the need to integrate the digital economy with the real economy, support the transformation and upgrading of traditional industries, stimulate the emergence of new industries and business models, and strengthen the drivers of economic growth. Similarly, Made in China 2025 identifies intelligent manufacturing as a strategic priority, advocating for the digitalization, networking, and automation of the manufacturing sector while promoting the deep integration of the industrial internet, big data, and artificial intelligence with manufacturing processes. To align with the national strategy for innovation-driven development, the manufacturing industry must actively pursue digital transformation, engage in independent technological innovation, and leverage emerging technologies and business models to modernize traditional industries. However, while data have become a fundamental factor of production in the digital economy, scientific and technological innovation—particularly digital innovation—should not be regarded as an abstract concept developed independently of traditional factors such as capital, materials, and labor. A strong cost-awareness mindset is essential for achieving cost-effective innovation.

Early studies established that cost stickiness reflects deliberate managerial decisions influenced by optimistic expectations of managers, adjustment cost, and agency conflict. In earnings forecasting models, cost stickiness has been shown to affect the accuracy of forecasts made by managers, analysts, and investors [2]. This occurs because firms with sticky costs exhibit greater earnings volatility, which increases the likelihood of forecasting errors by managers. Since managerial forecasts are a key source of information for analysts and investors [3], inaccuracies caused by cost stickiness can cascade into their forecasting behavior as well. The impact of cost stickiness on capital markets is multifaceted. First, firms with sticky costs exhibit lower dividend payouts [4]. This practice enables them to maintain stable cost commitments and consistent dividend distributions when confronted with future revenue declines, thereby safeguarding operational continuity and shareholder value stability. Second, cost stickiness may hinder timely adjustments to cost structures during revenue downturns [5], delaying the market’s recognition of a firm’s true financial condition and causing lagged stock price responses. This information asymmetry complicates investor decision making and reduces the efficiency of resource allocation in capital markets. More critically, when firms maintain rigid cost structures over extended periods and fail to adjust them during financial stress or systemic shocks, they face an elevated risk of sharp stock price declines or even crashes [6]. Thus, cost stickiness is not merely an internal management issue but can evolve into a significant factor affecting capital market stability. Nonetheless, cost stickiness may also have positive implications. For instance, at the macroeconomic level, it can enhance the accuracy of unemployment forecasts [7].

Regarding the drivers of cost stickiness, previous studies have explained the mechanisms of its variation in terms of three main drivers: agency conflicts, adjustment costs, and managers’ optimistic expectations. Supply chain network centrality has also been identified as an important factor affecting cost stickiness. Firms that are centrally located in the supply chain network tend to face higher adjustment costs and agency conflicts, which limit their ability to reduce costs during economic downturns [8]. However, this effect can be mitigated by strong internal controls and external audits, which highlights the importance of internal governance mechanisms for cost stickiness. For example, internal governance of the top management team can reduce managerial discretion and agency-based cost stickiness [9]. In addition to the internal environment, the external environment in which a firm operates affects its propensity to incur sticky expenditures. When firms face significant policy risk exposure, signaling the possibility of a downturn, managers reduce their expectations of future demand and future adjustment costs, thus reducing their cost stickiness [10]. In contrast, environmental investment growth (EIG) increases cost stickiness because firms with high asset specialization and high stakeholder pressure are less likely to reduce costs as revenues decline, and they need to maintain higher levels of costs to sustain environmental commitments or demonstrate long-term responsibility [11]. In addition, external competition can be a powerful factor. Both tariff-induced competition in product markets and banking competition in financial markets increase cost stickiness [12,13].

Recently, a growing body of literature has examined how digital factors affect cost stickiness, highlighting various mechanisms by which digital transformation reshapes firms’ cost behavior. A consistent finding across the various studies is that digitization—whether through internal transformation, customer-side innovation, or smart manufacturing—tends to reduce cost stickiness by improving operational agility and decision quality. Specifically, internal digital transformation plays a critical role in reducing cost stickiness by lowering adjustment costs and curbing managerial over-optimism [14]. Similarly, customer digital transformation has spillover effects on suppliers, curbing their cost stickiness by mitigating agency problems and reducing adjustment frictions [15], but its impact is constrained by external oversight and competitive pressures. Smart manufacturing and digital manufacturing further contributes to reducing cost stickiness by improving resource allocation and information processing capabilities [16]. Interestingly, fully integrated smart manufacturing has a greater dampening effect on cost stickiness than collaborative models, suggesting that deep digital integration is more effective in transforming cost behavior.

In summary, these studies reveal two core drivers of cost stickiness: managers’ expectations of future demand and agency costs. Digital transformation induces more timely cost adjustments by reducing managers’ over-optimism; however, digital tools improve operational transparency and traceability, effectively reducing agency costs and thus mitigating cost stickiness. While existing studies have addressed the impact of digitization on cost stickiness, there remains significant scope for further exploration. First, most of the current literature treats enterprise digitization as a single variable, which may obscure the heterogeneous effects of digitization on cost stickiness across different contexts. Second, limited research has examined the mechanisms through which digital innovation influences cost stickiness in manufacturing firms, particularly with respect to the unique characteristics of the manufacturing sector. Accordingly, we make the following marginal contributions. First, we evaluated the impact of digital innovation on enterprise cost stickiness by distinguishing among three dimensions: digital product innovation, digital process innovation, and digital business model innovation. Second, we contextualized this impact within the development of the manufacturing industry in the era of the digital economy, examining how digital technology promotes servitization and optimizes the structure of human capital. Third, we revealed the heterogeneous role played by different tenure periods of executives and high or low financing constraints based on managerial characteristics and firm financing capabilities.

2. Hypothesis Development

2.1. Direct Impact of Digital Innovation on Cost Stickiness

Digital innovation refers to the development of new products and the transformation of production processes, organizational structures, and business models through the integration of four key technologies: information, computation, connectivity, and communication [17]. Based on this definition, the impact of digital innovation on cost stickiness can be examined along three dimensions.

First, digital product innovation refers to the development of smart products and the enhancement of existing goods through the application of advanced digital technologies. Unlike traditional physical products, digital products often integrate software and data-driven services, forming a hybrid offering that reduces reliance on tangible resources. For instance, smart manufacturing enterprises increasingly deliver value through digital ecosystems, where the marginal cost of delivering digital services approaches zero. This shift reduces fixed costs associated with physical production, thereby lowering the degree of cost stickiness. Furthermore, the integration of digital features enhances firms’ responsiveness to market demand fluctuations, enabling more agile cost adjustments.

Second, digital process innovation focuses on the optimization and redesign of internal and external business processes through the use of digital technologies. Traditional cost stickiness is often driven by managerial optimism, in which managers retain redundant resources due to the overestimation of future demand. Digital technologies, such as artificial intelligence (AI), the Internet of Things (IoT), and big data analytics, address this issue by transforming decision-making processes from intuitive to data-driven processes. For example, predictive analytics can generate real-time insights into market dynamics, automating inventory adjustments and improving resource allocation. Additionally, digital process innovation enhances production efficiency by streamlining workflows and reducing operational redundancies, which collectively mitigate cost stickiness.

Third, digital business model innovation involves redefining how firms create, deliver, and capture value by leveraging digital platforms and ecosystems. Digital transformation has revolutionized consumer interactions, shifting purchasing behaviors from offline to online platforms. This transformation reduces enterprises’ dependency on physical infrastructure and labor-intensive processes, simultaneously lowering fixed costs and improving resource flexibility. Moreover, innovative pricing models, such as subscription services and on-demand payment systems, allow firms to align resource allocation more closely with demand fluctuations, thereby minimizing the rigidity of cost structures.

In summary, digital innovation enhances firms’ ability to adapt to changes in operating environments by reducing fixed costs, improving resource allocation, and fostering operational agility. These improvements directly address the underlying drivers of cost stickiness, including adjustment costs and managerial optimism. Therefore, we propose the following hypothesis:

H1. 

Digital innovation in products, processes, and business models mitigates cost stickiness in manufacturing firms.

2.2. Indirect Impact of Digital Innovation on Cost Stickiness

2.2.1. The Mediating Role of Manufacturing Servitization

Manufacturing servitization transformation refers to the gradual shift of manufacturing enterprises from traditional product-centered business models towards providing integrated solutions and value-added services [18]. In the digital economy era, digital development can directly drive enterprise servitization transformation [19]. Digital technology promotes manufacturing enterprises to transform from physical goods product deliverers to value co-creators integrating physical products and services.

First, the premise of manufacturing servitization lies in quantifying the value of intangible services provided. Digital technology resolves the difficulty of service pricing by converting the service’s invisible value into tradable assets through data servitization. Second, after business model innovation transfers value chain dominance to consumers, manufacturing enterprises develop comprehensive services such as information consulting, operational maintenance, system integration, and financial leasing, further expanding the scope of enterprise servitization. Additionally, big data itself is a service; the application of digital technology generates massive data. Manufacturing enterprises’ private data can serve the enterprise itself, while publicly available data can be sold to third-party institutions to develop related information services, broadening the enterprise’s service boundaries.

Regarding manufacturing servitization and enterprise cost stickiness, existing research has confirmed a negative relationship between them [20]. From the perspective of adjustment costs, the “service”-related software introduced by manufacturing servitization reduces the proportion of hardware in enterprises, thereby reducing fixed costs. From the perspective of management’s optimistic expectations, enterprises expanding into the service industry attract a large number of high-quality human resources. These human resources from the service industry maintain closer contact with customers over the long term, therefore effectively obtaining customer demand information and adjusting supply situations in a timely manner [21]. From the agency problem perspective, under the manufacturing servitization context, enterprises and customers tend to establish long-term cooperative relationships, and customers as stakeholders also become one of the enterprise’s external governance subjects, which directly reduces management’s discretionary space, thus helping to suppress resource idleness caused by imperialist motives. Accordingly, it is evident that digital innovation contributes to promoting manufacturing servitization transformation, thereby suppressing manufacturing enterprise cost stickiness. Therefore, the following hypothesis is proposed:

H2. 

Digital innovation mitigates manufacturing firms’ cost stickiness by driving manufacturing servitization.

2.2.2. The Mediating Role of Human Capital Structure

From the perspective of the resource-based view, sustainable competitive advantage is derived from unique and inimitable resources and capabilities, with human capital regarded as a critical strategic asset. However, the mere possession of skilled employees is insufficient; rather, the continuous reconfiguration of human capital structures in response to technological change is regarded as dynamic capabilities essential for sustainable development. Digital innovation is considered as a primary catalyst for the development of such dynamic capabilities, fundamentally transforming workforce management and ultimately altering cost structures.

First, traditional structured labor is being replaced by digital tools, and the adoption of digital technologies such as artificial intelligence in manufacturing requires new skill sets [22]. As a result, companies hire professionals for high-tech positions, while the replacement of low-skilled jobs encourages the remaining employees to shift toward higher-value activities, thus improving the overall quality of existing human capital. This transformation is consistent with the resource-based view, which emphasizes resources that are unique, scarce, and difficult to imitate. Moreover, as the marginal adjustment cost of automated production equipment approaches zero, frequent adjustments of low-skilled positions during periods of poor performance are no longer required, which significantly reduce the adjustment costs associated with hiring and dismissing employees. Second, according to person–organization fit theory, organizational effectiveness is maximized when individual abilities and needs are aligned with organizational requirements [23], and the application of digital technologies can effectively enhance the fit between employees and organizations. On the one hand, data-driven recruitment processes are enabled by big data and algorithms, allowing for more accurate identification of the skilled talent required by organizations, thus shortening recruitment cycles, reducing hiring risks, and minimizing sunk costs in the recruitment process. On the other hand, flexible employment models have been facilitated by digital technologies, enabling the establishment of short-term, flexible contractual relationships with high-skilled talent through the gig economy and remote collaboration platforms. As a result, labor input can be rapidly adjusted in response to demand fluctuations without incurring the fixed costs of long-term employment or the adjustment costs associated with layoffs.

In summary, from the perspectives of the resource-based view and person–organization fit theory, digital innovation not only enhances the uniqueness of human capital, but also significantly reduces the adjustment costs of human resources in response to external environmental changes by optimizing employee–organization fit and increasing employment flexibility. Therefore, the following hypothesis is proposed:

H3. 

Digital innovation mitigates manufacturing firms’ cost stickiness by optimizing the human capital structure.

The theoretical mechanism of the impact of digital innovation on cost stickiness in manufacturing is shown in Figure 1.

3. Methodology

3.1. Data Collection

We selected manufacturing firms listed on China’s A-share market from 2012 to 2023 as the research sample. The manufacturing sector is defined according to the China Securities Regulatory Commission (CSRC) industry classification (Industry Codes: C13–C43). The specific variable processing procedure is as follows: (1) samples from non-manufacturing industries are excluded; (2) to reduce the impact of outliers on the research results, samples labeled as ST or *ST are excluded in accordance with CSRC regulations and stock exchange listing rules; (3) all continuous variables are winsorized at the 1st and 99th percentiles to eliminate the influence of extreme values. Ultimately, a total of 22,475 samples is obtained. Data on digital innovation are obtained from the WinGo Finance Text Data Platform, information on manufacturing servitization is collected from companies’ official websites, and the remaining data are sourced from the CSMAR database.

3.2. Variable Selection

3.2.1. Cost Stickiness

Cost stickiness (STICKY) is measured based on the methodology proposed by Weiss [2], which is specified as follows:

$$\begin{array}{c}{∆\mathrm{C}\mathrm{O}\mathrm{S}\mathrm{T}}_{\mathrm{i},\mathrm{t}}=\left({\mathrm{S}\mathrm{A}\mathrm{L}\mathrm{E}}_{\mathrm{i},\mathrm{t}}-{\mathrm{E}\mathrm{A}\mathrm{R}\mathrm{N}\mathrm{I}\mathrm{N}\mathrm{G}}_{\mathrm{i},\mathrm{t}}\right)-\left({\mathrm{S}\mathrm{A}\mathrm{L}\mathrm{E}}_{\mathrm{i},\mathrm{t}-1}-{\mathrm{E}\mathrm{A}\mathrm{R}\mathrm{N}\mathrm{I}\mathrm{N}\mathrm{G}}_{\mathrm{i},\mathrm{t}-1}\right)\\ {∆\mathrm{S}\mathrm{A}\mathrm{L}\mathrm{E}}_{\mathrm{i},\mathrm{t}}={\mathrm{S}\mathrm{A}\mathrm{L}\mathrm{E}}_{\mathrm{i},\mathrm{t}}-{\mathrm{S}\mathrm{A}\mathrm{L}\mathrm{E}}_{\mathrm{i},\mathrm{t}-1}\\ {\mathrm{S}\mathrm{T}\mathrm{I}\mathrm{C}\mathrm{K}\mathrm{Y}}_{\mathrm{i},\mathrm{t}}={\log \left({\displaystyle \frac{∆COST}{∆SALE}}\right)}_{\mathrm{i},\mathsf{\omega }}-{\log \left({\displaystyle \frac{∆COST}{∆SALE}}\right)}_{\mathrm{i},\mathsf{\tau }}\mathsf{\omega },\mathsf{\tau }\in \left\{\mathrm{t},\dots ,\mathrm{t}-3\right\}\end{array}$$

(1)

where ΔCOST and ΔSALE represent the changes in operating costs and operating revenues, respectively, for enterprise i in a given quarter. The variable is denoted as $\mathsf{\tau }$ when operating revenue increases and $\mathsf{\omega }$ when it decreases. Following the existing literature, we define operating costs as the difference between operating revenue and operating profit from the income statement. In the standard Weiss model, the estimated value of cost stickiness is typically negative—the smaller the value, the greater the degree of cost stickiness. Therefore, we use the absolute value of the results obtained from model (1) and define STICKY as the measure of cost stickiness.

3.2.2. Digital Innovation

Following previous research, digital innovation is measured based on the frequency of relevant keywords in firms’ annual reports. According to the classification of digital innovation keywords by Zheng [24], we identified three dimensions of digital innovation. The specific keywords are listed in

Table 1. Keywords for digital innovation.

Dimension	Definition	Keywords
Digital Product Innovation	Developing new products through digital technologies	Internet of Things (IoT), Smart Home, Artificial Intelligence, (AI), Application Software, Software Programs, Software Platforms, Software Systems, Smart Terminal Devices, Smart Hardware, Wearable Smart Devices, Mobile Healthcare, Mobile Payment, Intelligent Security Systems, Smart Services, Intelligent Transportation Systems (ITS), Smart Devices, Smartphones, Intelligent Driving, Smart Vehicles, Biometric Recognition, Image Recognition, Virtual Reality (VR), Face Recognition, Human-Computer Interaction (HCI), Autonomous Driving, Robots, Semantic Understanding, Smart TVs, Smart Cities, Online Education, Smart Watches, Smart Communities, Smart Grids, Operating Systems (OS), Application Systems.
Digital Process Innovation	Application of digital technologies to improve or restructure existing innovation process frameworks	Digital Process Innovation, Intelligentization, Automation, Integration, Systematization, Data Mining, Machine Learning, Neural Networks, Intelligent Algorithms, Smart Technologies, Digital Simulation, Smart Manufacturing, Blockchain, Supply Chain, Information Management, Information Systems, Management Systems, Technology Platforms, Cloud Technology, Cloud Computing, Management Platforms, Simulation Technology, System Management, Data Management, Integration Systems, Modularization, Unmanned/Autonomous Operations, Automatic Control, Information Services, Smartization, Control Systems.
Digital Business Model Innovation	Application of digital technologies to improve or reshape existing business models	Digitalization, Electronification, Networking, Informatization, Cloud Services, Cloud, Cloud Platform, Big Data, Data Resources, Information Resources, Data Platform, Business Intelligence (BI), Cloud Strategy, Cloud Applications, Cloud Architecture, Cloud Infrastructure, Cloud Migration, Cloudification, Platformization, Service Platform, E-commerce, Digital Marketing, Internet Plus, Massive Data, User Profiling, Modernization, Online Platform, Triple-Play Convergence, Virtualization, Business Model, Online Mall, Electronic Trading, Internetization, Operations Platform.

. Specifically, the degree for each innovation type—digital product innovation (Product), digital process innovation (Process), and digital business model innovation (Business)—is measured as the proportion of their respective keywords to the total word count of the annual report. The proportion of the total frequency of these three innovation types to the total word count of the annual report is used to indicate the firm’s overall level of digital innovation, and these indicators are multiplied by 100 to avoid excessively small values. The specific keywords used are listed in Table 1. Additionally, for robustness testing, we also adopt an alternative proxy for digital innovation by using the number of digital invention patents.

3.2.3. Control Variables

Based on the relevant theory and literature, we control for several firm-level characteristics that have been shown to have a significant impact on cost stickiness:

(1)

Size: It is generally believed that larger firms have more complex bureaucratic hierarchies and higher adjustment costs, which lead to greater cost stickiness.

(2)

Lev: The asset–liability ratio reflects a firm’s financial leverage. Firms with high leverage are more likely to adjust their cost structures promptly under financial pressure to reduce financial risk, thereby affecting cost stickiness.

(3)

Growth: Managers in high-growth firms tend to be more optimistic about the future. Even if current revenues temporarily decline, they may view this as a short-term fluctuation and be reluctant to cut valuable resources, resulting in significantly higher cost stickiness.

(4)

Mfee: The administrative expense ratio reflects the cost control ability and management efficiency of a firm’s management. Firms with higher administrative expenses may exhibit greater cost stickiness.

(5)

Mshare: The higher the managerial ownership, the more aligned the interests of management and shareholders. To protect their own wealth, managers are more motivated to cut redundant costs when revenues decline, thus reducing cost stickiness.

(6)

TOP1: Major shareholders have stronger ability and motivation to monitor management and prevent them from refusing to cut costs in order to maintain personal power or build corporate empires. This supervisory pressure typically reduces cost stickiness.

(7)

Big4: Engaging a Big Four accounting firm for auditing subjects the firm to stricter external supervision and higher quality financial reporting. Such external governance pressure encourages managers to make more economically rational cost decisions, which may reduce cost stickiness.

The following

Table 2. Definition of variables.

Variable Type	Variable Symbol	Variable Name	Measurement Method
Explained variable	STICKY	Degree of cost stickiness	Calculated based on the WEISS model.
Explanatory variable	DIG	Digital innovation	The proportion of digital innovation keywords to the total number of words in the annual report.
	Product	Digital product innovation
	Process	Digital process innovation
	Business	Digital Business Model Innovation
Mechanism variables	Service	Manufacturing servitization	The product of service depth and number of service types
	Undergraduate	Proportion of employees with a bachelor’s degree or above	The proportion of employees with a bachelor’s degree or above to the total number of employees.
	Technology	Proportion of technical personnel	The proportion of employees with professional technical titles to the total number of employees.
	Research	Proportion of R&D personnel	The proportion of R&D personnel to the total number of employees.
Control variables	Size	Firm size	The natural logarithm of total assets.
	Lev	Asset–liability ratio	Total liabilities/total assets.
	Growth	Operating revenue growth rate	The year-on-year growth rate of operating revenue.
	Mfee	Administrative expense ratio	Administrative expenses/operating revenue
	Mshare	Managerial ownership ratio	The ratio of shares held by directors, supervisors, and senior executives to the total number of shares.
	Top1	Ownership concentration	The number of shares held by the largest shareholder divided by the total number of shares.
	Big4	Audited by a Big Four accounting firm (yes/no)	1 for the auditor is from one of the Big Four accounting firms, 0 for not.

is a definition of variables:

3.3. Descriptive Statistics

The results of the descriptive statistics are presented in

Table 3. Descriptive statistics.

Variables	N	Mean	Median	SD	Min	Max
STICKY	22,475	0.733	0.426	0.855	0.00489	4.505
DIG	22,475	0.0682	0.0417	0.0820	0	0.572
Product	22,475	0.0131	0.00192	0.0298	0	0.221
Process	22,475	0.0354	0.0224	0.0422	0	0.277
Business	22,475	0.0182	0.0116	0.0220	0	0.228
Service	5513	1.270	1.099	0.432	0.693	2.773
Undergraduate	21,573	0.253	0.205	0.175	0.0290	0.927
Technology	22,265	0.193	0.155	0.134	0.00997	0.865
Research	18,180	15.75	13.40	10.78	0.360	70.90
Size	22,475	22.08	21.92	1.182	19.48	26.45
Lev	22,475	0.395	0.386	0.191	0.0381	0.934
Growth	22,475	0.243	0.105	0.728	−0.926	12.46
Mfee	22,475	0.0826	0.0684	0.0622	0.00695	0.766
Mshare	22,475	0.163	0.0411	0.204	0	0.706
Top1	22,475	0.328	0.306	0.140	0.0760	0.758
Big4	22,475	0.0473	0	0.212	0	1

. The mean value of STICKY is 0.733, indicating that cost stickiness is a common issue among manufacturing firms in China. The maximum and minimum values are 4.505 and 0.00489, respectively, suggesting substantial variation in the degree of cost stickiness across firms. The minimum value of DIG is 0, indicating that some enterprises are not currently engaged in digital innovation. This highlights the need to further enhance the digital innovation capacity of manufacturing firms. The standard deviation of DIG is 0.082, reflecting substantial differences in digital innovation levels among enterprises. An analysis of the mean and median word frequencies for the three types of digital innovation indicates that most digital innovation activities are concentrated in process innovation. The median of the manufacturing servitization indicator is lower than the mean, indicating that the level of servitization is concentrated at a relatively low level for most companies, while a few companies exhibit a very high degree of servitization. The descriptive statistics of the control variables are generally consistent with findings from the existing literature.

3.4. Empirical Strategy

We chose the two-way fixed-effects model for this study to examine the effect of digital innovation on cost stickiness. We estimated the following regression model:

$${\mathrm{S}\mathrm{T}\mathrm{I}\mathrm{C}\mathrm{K}\mathrm{Y}}_{\mathrm{i},\mathrm{t}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{D}\mathrm{I}\mathrm{G}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\alpha }}_2{\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{t}}+\gamma {\mathrm{F}\mathrm{i}\mathrm{r}\mathrm{m} \mathrm{F}\mathrm{E}}_i+\mathsf{\lambda }{\mathrm{Y}\mathrm{e}\mathrm{a}\mathrm{r} \mathrm{F}\mathrm{E}}_{\mathrm{t}}+{\epsilon }_{i,t}$$

(2)

where ${\mathrm{S}\mathrm{T}\mathrm{I}\mathrm{C}\mathrm{K}\mathrm{Y}}_{\mathrm{i},\mathrm{t}}$ represents the cost stickiness of firm i in year t; ${\mathrm{D}\mathrm{I}\mathrm{G}}_{\mathrm{i},\mathrm{t}}$ is the degree of digital innovation for firm i in year t; ${\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}\mathrm{s}}_{\mathrm{i},\mathrm{t}}$ are a set of control variables.

To account for unobserved time-specific shocks, we included year fixed effects (${\lambda }_t$) in all regression models; we also incorporated firm fixed effects (${\gamma }_i$) to absorb time-invariant firm-specific heterogeneity. ${\epsilon }_{i,t}$ is the random error.

4. Empirical Results and Analysis

4.1. Foundational Regression Results

To verify the impact of digital innovation on cost stickiness, our study conducts a stepwise regression based on Model (2). The regression results are presented in

Table 4. Digital innovation and cost stickiness.

	(1)	(2)	(3)	(4)	(5)	(6)
	STICKY	STICKY	STICKY	STICKY	STICKY	STICKY
DIG	−0.2987 ***	−0.5192 ***	−0.4315 **
	(−3.959)	(−7.000)	(−2.470)
Product				0.4301
				(0.988)
Process					−0.8749 **
					(−2.574)
Business						−1.5776 ***
						(−3.377)
Size		0.0062	−0.0246	−0.0303	−0.0253	−0.0246
		(0.843)	(−1.250)	(−1.546)	(−1.285)	(−1.256)
Lev		−0.1609 ***	0.2699 ***	0.2664 ***	0.2684 ***	0.2650 ***
		(−3.738)	(3.643)	(3.592)	(3.618)	(3.580)
Growth		−0.0958 ***	−0.0633 ***	−0.0634 ***	−0.0633 ***	−0.0635 ***
		(−11.244)	(−6.797)	(−6.820)	(−6.796)	(−6.824)
Mfee		2.4968 ***	2.2331 ***	2.2509 ***	2.2346 ***	2.2363 ***
		(16.925)	(9.669)	(9.686)	(9.656)	(9.700)
Mshare		0.0067	−0.0627	−0.0624	−0.0620	−0.0617
		(0.196)	(−0.754)	(−0.750)	(−0.745)	(−0.742)
Top1		−0.2080 ***	−0.4451 ***	−0.4401 ***	−0.4433 ***	−0.4419 ***
		(−4.399)	(−4.046)	(−3.983)	(−4.028)	(−4.021)
Big4		−0.0086	0.0950	0.0960	0.0942	0.0947
		(−0.267)	(1.275)	(1.284)	(1.265)	(1.274)
Year	No	Yes	Yes	Yes	Yes	Yes
Firm	No	No	Yes	Yes	Yes	Yes
Constant	0.7548 ***	0.5810 ***	1.1820 ***	1.2682 ***	1.1954 ***	1.1789 ***
	(82.025)	(3.627)	(2.732)	(2.937)	(2.768)	(2.735)
N	22,475	22,475	22,087	22,087	22,087	22,087
adj. R2	0.001	0.045	0.107	0.107	0.108	0.108

Notes: ***, ** denote significance at 1% and 5% levels, t-statistics in parentheses, based on standard errors clustered by firm.

. Column (1) does not include the control variables, while Columns (2) and (3) sequentially include control variables, year fixed effects, and firm fixed effects. We observe a strong negative correlation between digital innovation and cost stickiness at a 1% level of significance based on the above three columns, with values of −0.2987, −0.5192, and −0.4315. These results indicate that digital innovation helps mitigate cost stickiness in manufacturing firms, which is consistent with Hypothesis 1. To further examine the effects of different types of digital innovation, our study conducts regressions for the three innovation dimensions separately. The results are reported in Columns (4) to (6). The inhibitory effect of digital product innovation on cost stickiness is not statistically significant, indicating that its direct impact may be more pronounced in terms of revenue growth and market share expansion rather than cost reduction. While digital product innovation can enhance product value and market competitiveness, its typically long development cycles, substantial investment requirements, and uncertainties regarding market adaptation limit its ability to directly influence firms’ cost structures and adjustment capabilities in the short term. In contrast, both digital business model innovation and digital process innovation demonstrate significant negative effects on cost stickiness. Notably, digital business model innovation exhibits a stronger inhibitory effect compared to digital process innovation. This difference can be explained by the distinct focus of each innovation type: digital process innovation primarily targets the intelligent transformation of internal processes—such as production, management, and supply chains—thereby improving dynamic resource allocation and operational efficiency, which helps reduce certain adjustment costs. However, these improvements are often localized within the enterprise, mainly contributing to partial optimization of the cost structure and short-term flexibility. Conversely, digital business model innovation—through platformization, servitization, and ecosystem development—fundamentally transforms the logic of value creation and delivery. This enables firms to flexibly adjust their product and service portfolios in response to market demand and facilitates the externalization and sharing of resources, thereby exerting a more substantial and comprehensive effect on reducing cost stickiness.

4.2. Robustness Tests

To test the robustness of the main findings, our study conducts the following robustness checks.

First, the study uses an alternative measure of digital innovation. Given that patent application volume reflects a firm’s level of innovation to some extent, we also use the presence of digital technology keywords in patent application documents to determine a firm’s engagement in digital economy-related innovation [25].

Second, the study adjusts the sample period. To account for the potential impact of economic fluctuations in 2020, the sample period is adjusted to exclude this year. We re-estimate the regression using data from 2012 to 2019.

Third, the study applies alternative clustering and fixed effects. While the baseline regression clusters standard errors at the firm level, this robustness test adjusts the clustering dimension to the industry-year level. In addition, industry-year fixed effects are included in the model.

The results of these robustness tests are presented in

Table 5. Robustness tests results.

	(1)	(2)	(3)	(4)
	Alternative Variable Substitution	Sample Period Adjustment	Multi-way Fixed Effects	Clustering at the Industry-Year Level
	STICKY	STICKY	STICKY	STICKY
Patent	−0.0203 **
	(−2.187)
DIG		−0.4362 *	−0.5527 ***	−0.4702 **
		(−1.907)	(−3.066)	(−2.423)
Size	−0.0186	−0.0771 ***	−0.0275	−0.0245
	(−0.933)	(−2.671)	(−1.342)	(−1.580)
Lev	0.2659 ***	0.2587 **	0.2314 ***	0.2697 ***
	(3.604)	(2.486)	(3.023)	(3.338)
Growth	−0.0633 ***	−0.0346 ***	−0.0598 ***	−0.0633 ***
	(−6.803)	(−3.488)	(−6.400)	(−6.099)
Mfee	2.2332 ***	2.0445 ***	2.4126 ***	2.2333 ***
	(9.582)	(7.381)	(10.309)	(9.696)
Mshare	−0.0568	−0.0719	−0.0198	−0.0625
	(−0.683)	(−0.538)	(−0.237)	(−0.734)
Top1	−0.4422 ***	−0.3937 ***	−0.3906 ***	−0.4453 ***
	(−4.020)	(−2.609)	(−3.475)	(−4.610)
Big4	0.0967	0.1410	0.0806	0.0952 *
	(1.299)	(1.260)	(1.085)	(1.734)
Year	Yes	Yes	Yes	Yes
Firm	Yes	Yes	Yes	Yes
Ind-Year	No	No	Yes	No
Constant	1.0540 **	2.2686 ***	1.2278 ***	1.1810 ***
	(2.406)	(3.618)	(2.736)	(3.590)
N	22,085	12,245	22,085	22,087
adj. R2	0.107	0.104	0.109	0.108

Notes: ***, **, * denote significance at 1%, 5%, and 10% levels, t-statistics in parentheses.

. The significance of the digital innovation variable remains consistent across alternative indicators, sample periods, and clustering/fixed effects specifications, thus confirming the robustness of the main findings.

4.3. Endogeneity Issues

4.3.1. Sample Selection Bias: Propensity Score Matching (PSM) Method

To address the endogeneity issue arising from sample selection bias, we adopt the propensity score matching (PSM) method, following previous studies. First, firms with a digital innovation index (DIG) above the median are classified as the treatment group, while the remaining firms constitute the control group. Second, the control variables used in this study are selected as matching variables, and 1:1 nearest-neighbor matching is applied. The average treatment effect on the treated (ATT) for firms engaging in digital innovation yields a t-value of −3.25, which is statistically significant at the 1% level. The results of the PSM balance test are illustrated in Figure 2. After matching, differences in control variables between the treatment and control groups are statistically insignificant, with standardized bias values below 10%. The PSM successfully balances observable characteristics between the groups, confirming that the main conclusions remain robust after accounting for sample selection bias. The regression results after matching are shown in Column 2 of

Table 6. Endogenous issues results.

	(1)	(2)
	Controlling for Omitted Variables	PSM
	STICKY	STICKY
DIG	−0.4435 **	−0.6844 **
	(−2.509)	(−2.386)
Size	−0.0211	−0.0309
	(−1.033)	(−0.991)
Lev	0.2484 ***	0.3073 ***
	(3.351)	(2.824)
Growth	−0.0642 ***	−0.0931 ***
	(−6.859)	(−6.116)
Mfee	1.9679 ***	2.5941 ***
	(7.588)	(7.722)
Mshare	−0.0597	0.0931
	(−0.718)	(0.669)
Top1	−0.4626 ***	−0.4380 ***
	(−4.241)	(−2.642)
Big4	0.0974	0.0082
	(1.305)	(0.070)
EInt	0.0103
	(0.619)
AInt	0.0644 *
	(1.898)
Opportunism	−0.0423 **
	(−1.986)
Constant	1.7158 ***	1.2775 *
	(3.560)	(1.878)
Year	Yes	Yes
Firm	Yes	Yes
N	22,043	9731
adj. R2	0.108	0.123

Notes: ***, **, * denote significance at 1%, 5%, and 10% levels, t-statistics in parentheses.

, confirming that our conclusions remain robust.

4.3.2. Omitted Variable Bias: Controlling for Omitted Variables

To mitigate the potential influence of omitted variables, we incorporate additional controls: asset intensity (net fixed assets divided by operating income), employee intensity (number of employees at year-end divided by operating income), and managerial opportunism (natural logarithm of total compensation of the top three executives). The results reported in Column (1) confirm the robustness of the main findings.

5. Mechanism Analysis

Theoretical analysis suggests that digital innovation promotes the servitization of manufacturing and optimizes the structure of human capital, which in turn helps reduce cost stickiness in manufacturing firms. Based on this, and drawing on established practices, the following model is employed to test these two mechanisms, after confirming the validity of the corresponding mechanism variables:

$${Mediator}_{i,t}={\beta }_0+{\beta }_1{DIG}_{i,t}+{\beta }_2{Controls}_{i,t}+\sum Year+\sum Firm+{\epsilon }_{i,t}$$

(3)

The variable ${Mediator}_{i,t}$ represents the mechanism through which digital innovation may affect cost stickiness. It includes the following two mechanisms.

5.1. Level of Manufacturing Servitization

The level of servitization is calculated as the product of service depth and the number of service types [26]. The number of services refers to the types of services provided by an enterprise, classified based on the national industry classification standard (GB/T 4754-2011) [27]. The classification is determined by identifying the types of service activities conducted by each enterprise during the current year. Service depth is assessed using a qualitative scoring system based on the complexity or added value of the services: a score of 1 for basic services, 2 for intermediate services, and 3 for services with high complexity or added value.

5.2. Optimization of Human Capital Structure

It is measured using three indicators: First, the proportion of employees with a bachelor’s degree or above, calculated as the number of such employees divided by the total number of employees. Second, the proportion of technical personnel, defined as the share of employees with professional or technical titles. Third, the proportion of R&D personnel, measured as the number of employees engaged in research and development relative to the total workforce.

The regression results of the established model (3) are detailed in

Table 7. The mediating effect of the above two mechanisms on the DIG-STICKY relationship.

	(1)	(2)	(3)	(4)
	Service	Undergraduate	Technology	Research
DIG	0.5920 ***	0.1222 ***	0.1574 ***	7.5780 ***
	(2.824)	(3.750)	(4.737)	(3.357)
Size	0.0552 **	0.0128 ***	0.0073 **	0.2149
	(2.347)	(4.002)	(2.262)	(0.815)
Lev	0.1401 **	−0.0433 ***	−0.0181 *	−3.1393 ***
	(1.994)	(−4.475)	(−1.916)	(−4.032)
Growth	0.0095	0.0062 ***	0.0035 ***	−0.0896
	(1.519)	(4.124)	(2.606)	(−1.135)
Mfee	0.2077	0.1659 ***	0.0543 **	0.8941
	(1.193)	(5.791)	(2.315)	(0.492)
Mshare	−0.0813	0.0288 ***	0.0150	1.2314 *
	(−1.071)	(2.631)	(1.367)	(1.686)
Top1	−0.2008 *	0.0039	−0.0083	−0.4521
	(−1.650)	(0.187)	(−0.406)	(−0.358)
Big4	−0.0368	−0.0000	−0.0067	−1.3792 **
	(−0.705)	(−0.002)	(−0.862)	(−2.217)
Constant	−0.0030	−0.0434	0.0235	11.5395 **
	(−0.006)	(−0.616)	(0.329)	(1.994)
Year	Yes	Yes	Yes	Yes
Firm	Yes	Yes	Yes	Yes
N	5306	21,165	21,869	17,750
adj. R2	0.740	0.888	0.810	0.879

Notes: ***, **, * denote significance at 1%, 5%, and 10% levels, t-statistics in parentheses.

. Results in Column (1) demonstrate a statistically significant positive coefficient for DIG (β = 0.592, t > 2.58), indicating that digital innovation significantly enhances the servitization level of manufacturing (H2 supported). Drawing on existing research, the resource allocation effects of digital transformation in manufacturing can help alleviate cost stickiness [28], while digital innovation mitigates cost stickiness by facilitating the industry’s transition toward a service-oriented model. The consistently positive coefficients of DIG in Columns (2)–(4) (β = 0.1222, 0.1574, and 7.578, respectively, t > 2.58) reveal its role in upgrading human capital structure. Collectively, these findings suggest that digital innovation reduces cost stickiness by optimizing human capital structure. Skilled labor and R&D teams improve operational flexibility, lowering managerial inertia in cost adjustments [29]. For instance, AI-driven predictive maintenance reduces unplanned downtime costs, diminishing the upward rigidity of maintenance expenses.

6. Heterogeneity

6.1. Heterogeneity of Executive Tenure

Strategic decisions made by executives at different career stages directly influence the trajectory of corporate development. Our study investigates the impact of digital innovation on cost stickiness across varying CEO tenures, based on the heterogeneity of executive characteristics. Executive tenure is measured as the natural logarithm of the number of months the CEO has held office and is categorized into three groups—short, medium, and long tenure—based on the lower and upper quartiles of the tenure distribution. Grouped regression results, reported in Columns (1) to (3) of

Table 8. Heterogeneity analysis results.

	(1)	(2)	(3)	(4)	(5)
	long-career executives	mid-career executives	short-career executives	high financing constraints	low financing constraints
	STICKY	STICKY	STICKY	STICKY	STICKY
DIG	−0.3053	−0.6138 **	−0.2812	−0.5893 *	−0.2756
	(−0.831)	(−2.232)	(−0.672)	(−1.866)	(−1.029)
Size	−0.0655	−0.0286	0.0600	−0.0105	−0.0522
	(−1.230)	(−1.091)	(1.327)	(−0.281)	(−1.613)
Lev	0.3600 *	0.2332 **	0.1122	0.4137 ***	0.2353 *
	(1.864)	(2.266)	(0.622)	(3.217)	(1.937)
Growth	−0.1118 ***	−0.0615 ***	−0.0584 ***	−0.0952 ***	−0.0441 ***
	(−4.898)	(−4.662)	(−3.307)	(−6.761)	(−3.433)
Mfee	2.1620 ***	2.0875 ***	2.8642 ***	2.9023 ***	2.1023 ***
	(4.432)	(6.402)	(5.914)	(5.919)	(6.849)
Mshare	0.1720	−0.1411	0.1965	−0.1411	0.0025
	(0.731)	(−1.249)	(0.856)	(−0.893)	(0.020)
Top1	−0.0623	−0.5291 ***	−0.8369 ***	−0.4722 **	−0.5992 ***
	(−0.218)	(−3.234)	(−3.409)	(−2.373)	(−3.346)
Big4	0.1355	0.1545 *	−0.0674	0.1035	0.0035
	(0.591)	(1.660)	(−0.522)	(0.968)	(0.039)
Constant	1.9046	1.3418 **	−0.5931	0.8183	1.8180 **
	(1.621)	(2.336)	(−0.607)	(0.996)	(2.568)
Year	Yes	Yes	Yes	Yes	Yes
Firm	Yes	Yes	Yes	Yes	Yes
N	6741	12,021	5472	12,510	12,385
adj. R2	0.130	0.104	0.175	0.169	0.148

Notes: ***, **, * denote significance at 1%, 5%, and 10% levels, t-statistics in parentheses.

, show that the coefficient of digital innovation is significantly negative only for firms with medium-tenure CEOs, while it is statistically insignificant for the short- and long-tenure groups. This phenomenon is mainly attributed to the heterogeneity of executives’ career motivation. First, newly appointed executives, although motivated to implement high-risk strategic changes due to performance pressures and career ambitions, may avoid core process reengineering because performance evaluations often emphasize short-term financial metrics. Second, long-tenured executives tend to adhere to traditional management practices, which can diminish organizational innovation capacity and reduce business exploration efforts [30]. In contrast, mid-tenure executives are better positioned to balance short-term performance with long-term strategic investment. The capital accumulated during earlier stages of their tenure provides them with the confidence to invest in innovation and effectively reduce cost adjustment delays.

6.2. Heterogeneity of Financing Constraints

Previous studies have shown that cost stickiness tends to be more pronounced in firms with lower financing constraints [31], as reduced constraints increase managerial confidence in the external environment and encourage excessive risk-taking behavior. Accordingly, this paper selects financing constraints as a moderating variable to examine whether the effect of digital innovation on cost stickiness differs across varying levels of financial pressure. The absolute value of the SA index is used as a proxy for financing constraints, where a larger value indicates greater financial pressure. Based on the median SA index, firms are divided into two groups: high and low financing constraints. The last two columns of Table 8 report the results of the subgroup regressions. The coefficient of digital innovation is significantly negative in the high financing constraint group but statistically insignificant in the low-constraint group. This suggests that when firms experience less financial pressure, managers may become overly optimistic and disregard insights from big data analytics, thereby weakening the role of digital innovation in mitigating cost stickiness.

7. Conclusions and Implications

7.1. Conclusions

Our study investigates the impact of digital innovation on cost stickiness in China’s A-share listed manufacturing firms from 2012 to 2023. The findings demonstrate that digital innovation—spanning product, process, and business model dimensions—significantly mitigates cost stickiness. However, the effects are heterogeneous across innovation types: while digital process innovation and business model innovation exhibit strong inhibitory effects on cost asymmetry, digital product innovation shows limited statistical significance, likely due to its inherently high R&D costs and delayed market adaptation. The analysis further uncovers two critical mediating pathways. First, digital innovation accelerates manufacturing servitization, transitioning firms from rigid product-centric operations to agile service-oriented models, thereby reducing asset specificity and unplanned downtime costs. Second, it optimizes human capital structure by increasing the proportion of skilled technical and R&D personnel, which enhances decision-making agility and mitigates managerial inertia in cost adjustments. These mechanisms collectively address the mechanism pathway of how digitization reshapes cost behavior. Our analysis uncovers context-specific heterogeneous effects. The cost stickiness reduction effect is amplified in firms facing high financing constraints, as digital innovation alleviates information asymmetry and strengthens external financing capacity. Additionally, the effect peaks when executives are in their mid-career stage, reflecting career incentive dynamics. Mid-career managers, driven by reputational motives and risk appetite, prioritize digital adoption to signal competence, whereas late-career managers exhibit conservatism due to risk aversion.

7.2. Implications

7.2.1. Managerial Implications

Several important insights for business practice are provided by these findings.

First, investments in digital process and business model innovation should be prioritized by manufacturing firms seeking to enhance cost flexibility and resilience, as these dimensions have been found to be most effective in reducing cost stickiness. The adoption of digital tools and platforms that enable real-time monitoring, agile resource allocation, and data-driven decision making can benefit firms by improving their capacity to respond to market fluctuations and external shocks.

Second, the strategic value of servitization and human capital optimization is highlighted by the identified mediating mechanisms. The potential of digital innovation to streamline cost structures and reduce managerial inertia can be further unlocked by accelerating the transition toward service-oriented business models and investing in the recruitment and development of skilled technical and R&D personnel.

Third, practical guidance for tailoring digital transformation strategies is offered by the heterogeneous effects observed with respect to financing constraints and managerial career stages. Digital innovation should be leveraged by firms with limited access to external capital to enhance transparency and strengthen their financing capacity. Meanwhile, the pivotal role of mid-career executives in driving digital adoption should be recognized by boards of directors and senior management teams, and incentive structures that align managerial interests with long-term innovation objectives should be considered.

7.2.2. Policy Implications

Several policy recommendations are proposed to further advance the digital transformation and sustainable development of China’s manufacturing sector.

First, policy should be directed toward advancing the digital transformation of manufacturing enterprises, with a particular focus placed on supporting process digitalization and business model innovation. For instance, the strategic leveraging of the industrial internet, the Internet of Things, and platform economy models can facilitate the optimization of resource allocation and reduce dependence on fixed costs.

Second, the servitization of manufacturing should be promoted, thereby guiding firms in their transition from traditional product manufacturing to integrated “product + service” models. The provision of intelligent and customized services enables greater operational flexibility; furthermore, a deep integration of producer services with manufacturing can contribute to improved cost elasticity.

Third, the optimization of human capital structure is recognized as a critical indirect mechanism through which digital innovation exerts its effects. Through the implementation of targeted subsidies and skills training programs, employees’ digital capabilities can be bolstered. Concurrently, efforts should be made to attract high-end digital technology talent, thereby strengthening firms’ innovation and execution capabilities.