Int. J.

Production Economics 263 (2023) 108969

Contents lists available at ScienceDirect

International Journal of Production Economics

journal homepage: www.elsevier.com/locate/ijpe

Digital transformation and low-carbon technology innovation in

manufacturing firms: The mediating role of dynamic capabilities
Gangqiang Yang, Yiming Nie, Honggui Li, Haisen Wang *
Economics and Management School, Wuhan University, Wuhan, 430072, China

A R T I C L E I N F O A B S T R A C T

Keywords: Green challenges such as global warming call for deeper low-carbon technology changes in manufacturing firms,
Digital transformation and digital transformation (DT) may play a major role. Most existing studies have investigated the impact of DT
Low-carbon technology innovation as a whole on green technology innovation but lack a dynamic mechanism analysis from a process perspective.
Dynamic capabilities
Therefore, this article proposes a theoretical framework for multistage DT to drive low-carbon technology
Manufacturing firms
innovation (LCTI) in manufacturing firms, which is closely linked to three dynamic capabilities (DCs). Based on
the panel data of A-share manufacturing firms in China from 2011 to 2019, this article finds the following. First,
DT effectively promotes LCTI, and this effect is a “leverage effect” superimposed on existing innovation. Second,
the theoretical mechanism shows that the data analysis, data operation, and data empowerment stages of DT
enhance the absorptive capacity, innovation capacity, and adaptive capacity of enterprises, respectively, which
in turn promote LCTI. Empirical tests support the mediating effects of the first two capabilities; however,
adaptive capacity is validated to be effective only for low-tech firms due to the limited increase in strategic
flexibility of high-tech firms by marginal DT enhancement. Third, the heterogeneity analysis shows that the
promotion effect of DT on LCTI is stronger for carbon reduction and decarbonization technologies, more sig­
nificant for low-emission firms, and stronger for firms in the cohort network. These conclusions add new
microevidence of the environmental benefits of the digital economy and provide new explanations for opening
the “black box” of DT and LCTI based on a process perspective. At the same time, this work can help
manufacturing firms in developing countries follow the low-carbon transformation path by independent inno­
vation and, moreover, reduce the originally enormous amounts of energy consumption and carbon emissions of
these countries and contribute to the response to global warming and the energy crisis.

1. Introduction provide better examples of technological innovation. For example, Haier

Group, the world’s largest home appliance manufacturer, independently
Green challenges such as global warming call for more profound developed a new technology for hydrocarbon (HC) refrigerant and
technological changes (Nordhaus, 2021). As a key element in energy Liquid Blowing Agent (LBA) polyurethane foam, thus allowing them to
saving and emission reduction, low-carbon technology has become an invent a completely environmentally friendly − 150 ◦ C deep
important form of strategic competition for countries to seize the in­ low-temperature refrigerator. Moreover, Sany Group, one of the world’s
dustrial high ground. Developed countries such as the U.S. are taking top-ranked Chinese construction machinery manufacturers, indepen­
increasingly aggressive “technology decoupling” measures against dently developed the “DMOS” engine, which achieves comprehensive
China, such as more targeted technology export controls and investment fuel savings of more than 10% through intelligent control. However, it is
audits, to protect their open-source technology ecosystem. As a major very difficult for most manufacturing firms to develop low-carbon
manufacturing country with the highest carbon emissions worldwide, technologies on their own because environmental technology in­
China’s source of key low-carbon technologies must shift from a long ventions require the dynamic coupling of multidomain technical
history of technological transformation (Wan et al., 2021) to the knowledge (Ardito et al., 2019) and expensive transformation costs
establishment of an autonomous and controllable technological inno­ (Huang et al., 2019), and there is a lack of positive economic returns to
vation system. Fortunately, some large manufacturing enterprises the ecological externalities generated by such technologies (Rennings,

* Corresponding author.
E-mail addresses: [email protected] (G. Yang), [email protected] (Y. Nie), [email protected] (H. Li), [email protected] (H. Wang).

https://doi.org/10.1016/j.ijpe.2023.108969
Received 12 October 2022; Received in revised form 26 June 2023; Accepted 29 June 2023
Available online 1 July 2023
0925-5273/© 2023 Elsevier B.V. All rights reserved.
G. Yang et al. International Journal of Production Economics 263 (2023) 108969

2000). Therefore, many scholars have been working to find the factors helps absorb fragmented, static tacit knowledge and decode it into
that drive low-carbon technology or green technology innovation, intelligent knowledge modules (Zhou et al., 2023), while digital con­
mainly in terms of market demand (Farza et al., 2021), stakeholder nectivity, aggregation and screening help achieve the rapid reconfigu­
pressure (Huang et al., 2016), partners (Arfi et al., 2018), organizations ration of organizational resources in a crisis (Shan et al., 2021). On the
(Zhou et al., 2021), and environmental policies (Zhang et al., 2021). other hand, different dimensions of DCs enhance different components
Moreover, with the Fourth Industrial Revolution, digital trans­ of green technology innovation; for example, absorptive capabilities
formation (DT) has become an important opportunity for firms to create provide mainly information support (Mousavi et al., 2019), innovation
sustainable competitive advantages and break through development capabilities provide practical guidance (Qiu et al., 2020), and adaptive
bottlenecks and technical problems (Qi and Xiao, 2020) and has brought capabilities help develop strategic plans (Dangelico et al., 2017). It can
about new opportunities for manufacturing firms to invent low-carbon be seen that DCs play a major role in connecting the various stages of DT
technologies. At present, scholars are still in the preliminary stage of with the various components of green technology innovation. Unfortu­
investigating DT’s green technology innovation effect, either from a nately, the literature has not yet given a systematic compendium of the
macro perspective to investigate the impact of different digital tech­ connection between the three. Therefore, the application of DC theory
nologies such as big data (El-Kassar and Singh, 2019), internet infor­ provides a new solution with which to open the black box of the process
mation technology (Liu et al., 2013), and blockchain on green of DT for LCTI, which is more helpful for manufacturing enterprises in
innovation or from a micro perspective to consider DT as a whole breaking through the established path dependence in the increasingly
concept, arguing that DT facilitates the decrease in corporate knowledge dynamic environment, stimulating low-carbon innovation through DT,
search costs (Yin and Yu, 2022), the increase in innovation resource and gaining sustainable competitive advantage.
investment and the relaxation of financing constraints (Liu et al., 2023). Based on the panel data of Chinese A-share listed manufacturing
This situation, in turn, enhances green technology innovation. While the firms from 2011 to 2019, this paper explores the effect of DT on LCTI. To
above studies help increase the understanding of the role of DT in pro­ fill the gap in the literature on mechanism research, this article applies
moting green technology innovation, there is still more room for dis­ Dynamic Capacity Theory to provide a dynamic mechanism explanation
cussion. In terms of technology scope, most literature focuses on green from a process perspective. In addition, we discuss the heterogeneity of
technology innovation, while studies focusing on low-carbon technology the effect in further analysis from more in-depth perspectives, such as
innovation (LCTI) against a “dual-carbon” background are very scarce. low-carbon technology type, industry carbon intensity, and firm cohort
In terms of heterogeneity analysis, the current literature presents rather characteristics, hoping to expand our results and form more interesting
common perspectives, such as firm ownership and firm size (Xue et al., conclusions. This article provides empirical evidence from a micro
2022), which have little relevance to and reference value for the topic. perspective to supplement the cross-cutting research on digital econ­
More importantly, the literature lacks a consideration of DT staging omy, low-carbon economy, and enterprise organizational management.
characteristics when discussing internal mechanisms. In fact, enterprises There are two main theoretical contributions: (1) This article finds a new
need a certain transformation cycle for DT to gradually embed digital internal driving factor, digital transformation, based on the current
technology into different aspects, such as knowledge arrangement (Zhou digital economy landscape, which is different from existing studies
et al., 2023), product development (Liu et al., 2020), operation man­ limited to external drivers such as market demand, stakeholder pressure,
agement (Qi and Xiao, 2020), and business models (Nambisan et al., and environmental policies, and is crucial to mitigate climate change
2019), forming multiple layers of progressive and synergistic digital and achieve global sustainable development. (2) This article proposes a
subjects to realize the “information digitalization - process digitalization novel driving mechanism for promoting LCTI in enterprises - the
- business digitalization - industry digitalization - ecological digitaliza­ enhancement of three dynamic capacities (DCs), which differs from
tion” technological change pathway. Therefore, there may be significant existing literature limited to superficial mediating mechanisms such as
stage-specific differences across firms in terms of their transformation financing constraints and information search costs. The proposed
cycles (Cao et al., 2023). In addition, as LCTI is influenced by multiple mechanism represents the most fundamental intrinsic activation mech­
components, such as organizational strategy (Song and Yu, 2018), anism from DT to LCTI. In particular, we find that this mechanism is
innovation resources (Arfi et al., 2018; Yu et al., 2021), supply chain dynamic and phased. Throughout the lifecycle of firms’ DT, the data
management (Li et al., 2022), and open collaboration (Ovuakporie et al., analysis phase gathers knowledge for LCTI by enhancing absorptive
2021), the digitalization scenarios associated with these components capacity, the data operation phase provides guidance for LCTI by
also differ across DT firms. These two aspects suggest that there is a enhancing innovative capability, and the data empowerment phase
staged and scenario matching problem between DT and LCTI (Cao et al., provides strategic support for LCTI by enhancing adaptive capability.
2023), a topic that is scarce in the literature. Therefore, the research on These findings constitute an innovative dynamic mechanism framework
DT and LCTI needs to be supplemented with a more dynamic mechanism for promoting LCTI through DT. This article has significant academic
analysis from the process perspective, which would provide important and practical value. From an academic perspective, the proposed dy­
guidance for how manufacturing firms in various transformation stages, namic mechanism framework inspires further empirical research based
in reality, can utilize their advantages and gradually build an autono­ on a process perspective to explore the future application value of DT.
mous and controllable LCTI system. From a practical standpoint, this article not only offers guidance on how
As the environment becomes increasingly complex and rapidly firms can establish independent and controllable LCTI systems during
changing, dynamic capabilities (DC) become a winning strategy for their respective DT cycles to gain sustainable advantages in global
firms to gain competitive advantage. DC refers to “the ability of a firm to competition, but also contributes to reducing the huge energy con­
integrate, structure, and reconfigure internal and external resources to sumption and carbon emissions of manufacturing enterprises in devel­
respond to a rapidly changing environment” (Teece et al., 1997) and is oping countries. Thus, it makes a valuable contribution towards
divided into different dimensions such as absorptive capacity, innova­ addressing the pressing issues of global warming and energy crises.
tive capacity, and adaptive capacity (Wang and Ahmed, 2007). We The remainder of this paper is organized as follows. Section 2 pre­
believe that DC theory provides a more fitting theoretical guide for sents the literature review and hypothesis development, Section 3 pre­
filling the gaps in the literature on internal mechanisms, as DCs provide sents the research methodology, Section 4 presents the analysis and
a deeper explanation of how firms gain a green competitive advantage results, Section 5 presents further analysis, and Section 6 presents the
from the perspective of higher-order capabilities and that the evolution conclusions and discussion.
of such capabilities arises from DT and from acting on green technology
innovation (Yuan and Cao, 2022). On the one hand, DT enhances
different levels of DCs; for example, digital knowledge orchestration

2
G. Yang et al. International Journal of Production Economics 263 (2023) 108969

2. Literature review and hypothesis development the blockchain consensus mechanism gives banks access to firms’ real
emission information, prevents moral risks such as information manip­
Vial (2019) synthesizes previous research and defines DT as “the ulation, and provides firms with financing services that are both
process by which the combination of information, computing, commu­ accessible and low cost. In summary, DT can help companies overcome
nication, and connectivity technologies triggers significant changes in two major challenges of LCTI, that is, cross-domain knowledge accu­
organizational attributes that improve the organization”. The term mulation and financing constraints. As a result, the following hypothesis
“digital” refers to back-end access to computer and internet technolo­ is proposed:
gies, while “transformation” places more emphasis on innovation and
H1. With all other things remaining equal, DT makes a positive
user value added as the main objectives, thus redefining organizational
contribution to a firm’s LCTI.
strategies and governance mechanisms and building new business
models (Cennamo et al., 2020). DT is a process in which enterprises
build different digital subjects in their respective transformation cycles 2.2. DCs: new intermediary mechanism
and finally perform their overall function through the connection and
aggregation of each subject. Because subjects are progressive, we refer to DC theory is an important theory for explaining firm performance in
them as the different “stages” of DT. In this article, we refer to Jiao the field of management. In a complex and changing environment, if a
et al.’s (2021) stage-by-stage division of enterprise DT, as it is more firm lacks DCs, then its existing market leadership may be lost in a flash
relevant to the context of enterprise DT in China. Specifically, DT can be (Zollo and Winter 2002). However, this theory has been less mentioned
divided into three stages: (1) data analysis stage, with information data in empirical studies in the field of economics; thus, there is plenty of
as the core content, through data collection, intelligent analysis and room for exploration and potential applications. This theory originates
quality monitoring, to realize the transformation of information form from the resource-based view (RBV). Since the RBV does not explain the
and open up “data islands”; (2) data operation stage, with process link between a firm’s resources and its product market, Teece et al.
digitization and business digitization as the core content, to improve (1997) proposed the concept of DCs, i.e., the abilities of a firm to inte­
internal operation efficiency through the flow, sharing and interopera­ grate, construct, and reconfigure internal and external resources to
bility of standardized data with the entire production line and each respond to a rapidly changing environment. In the process of sustain­
business unit within the enterprise; and (3) data empowerment stage, ability, DC helps firms integrate and create internal and external re­
with industrial data and ecological data as the core content, through sources related to environmental protection (Qiu et al., 2020). Wang and
data internalization and reconstruction and open sharing, to realize the Ahmed (2007) identified three dimensions of DC based on the principle
synergy of organizational strategy and the whole industry chain. In this of the “commonality of key features and specificity in details”, and
section, we first elaborate on how DT directly enhances LCTI from the subsequent literature has demonstrated the linkage between each
viewpoint of the literature, and then, by taking the key characteristics dimension and green innovation on this basis. (1) Absorptive capacity,
and technological performance of each DT stage as research objects, we which refers to the ability to identify and assimilate valuable external
construct a theoretical framework—DTx − DCx − LCTIx —to provide a knowledge and integrate it with internal knowledge, helps companies
dynamic mechanism analysis of the DT’s LCTI effect from a process understand customers’ needs and competitors’ R&D progress, which in
perspective. turn provides information for them to support green innovation
(Mousavi et al., 2019). (2) Innovation capability, which refers to the
2.1. DT and LCTI: direct effects ability to combine strategic innovation directions with innovation be­
haviors to develop new products, helps firms fully mobilize internal
In this article, we suggest that DT has a positive direct contribution innovation resources to achieve green technology innovation practices
effect on LCTI. First, DT is beneficial for enterprises in expanding their (Qiu et al., 2020). (3) Adaptive capacity, which refers to the ability to
internal knowledge stock (Yin and Yu, 2022) and building a composite align the organization with the external environment by flexibly
database to provide a knowledge base for green technology research. adjusting strategies and allocating resources, helps companies flexibly
Research has shown that the R&D and application of green innovation adjust their strategies and resource mix to take advantage of new green
often occur when other knowledge resources or assets in the firm’s opportunities (Dangelico et al., 2017). Based on this, this article estab­
innovation network are complementary (Aghion et al., 2019), and the lishes an echo between each DT stage and each DC dimension, consid­
establishment of the internet, Internet of Things and other wise inter­ ering that the data analysis, data operation and data empowerment
action platforms breaks the barriers of space and time, enabling firms to stages focus on developing the absorptive capacity, innovation capacity
disseminate information and absorb knowledge from a wide range of and adaptive capacity of firms, respectively, which enable them to
channels in the resource-sharing network and accumulate knowledge in promote the flow of innovative resources to low-carbon technologies
multiple fields, such as energy-related agriculture, materials, atmo­ and enhance low-carbon competitive advantages in a more effective
spheric environment, and waste recycling and treatment. Therefore, DT interaction with the dynamic environment. The presentation of this
can significantly reduce the reliance of subjects of innovation on tech­ theoretical framework is innovative. Specifically, we visualize the
nology’s existing relevance (especially nongreen relevance) (Santoalha logical framework in Fig. 1.
et al., 2021) and consolidate the knowledge foundation for low-carbon
technology research. Second, DT helps match the supply for green 2.2.1. Mediating role of absorptive capacity
financing with the demand, thereby easing the financial pressure placed First, during the data analysis stage, DT promotes LCTI mainly by
on firms and expanding their green innovation capital (Liu et al., 2023). enhancing absorptive capacity. The key point of the research on green
For firms, on the one hand, digital tools such as energy management technology lies in the organic reorganization of frontier green emerging
systems help firms (especially those operating across regions and sec­ field knowledge and multidomain nongreen knowledge (Castellacci and
tors) conduct accurate carbon audits and monitoring of activities,1 thus Natera, 2013). However, since green technologies are nonproductive
generating high-quality environmental data reports for disclosure to technologies, their applications vary greatly among manufacturing
banks; on the other hand, digital financial analysis tools help firms firms, resulting in a large amount of uneven green information that is
monitor the financial environment and find the lowest-cost form of difficult to identify emerging in the market. Therefore, it is very
green financing (Gong et al., 2021). For the funding provider, relying on important for the high-quality development of LCTI to determine
whether a firm can effectively identify, extract and absorb knowledge in
the complicated data world, i.e., whether it has strong absorption ability
1
Source: " Corporate Net Zero Pathway " published by the United Nations. (Zhou and Li, 2012), which is the focus of the data analysis stage. First,

3
G. Yang et al. International Journal of Production Economics 263 (2023) 108969

Fig. 1. DT and LCTI: Mediation of DC

through a series of complete closed-loop operations such as collection the R&D department to “see” the hidden sources of carbon emission
and extraction, automatic aggregation, preliminary analysis and quality improvement in the production chain. On the other hand, under the
assessment of online and offline metadata information, enterprises can circular feedback mechanism, digital simulation, virtual reality and
break the “data silos” and standardize the noise data from external other digital technologies map the real physical environment into a
competitive product markets, innovation organization platforms (such mathematical model of information space, which can simulate the dy­
as enterprise clusters, school-enterprise cooperation, and strategic alli­ namics of emission reduction in each production process (Chen et al.,
ances (Lu et al., 2021)) and internal processes, thereby building an 2021) and provide real-time feedback for the direct reference of tech­
automatically managed database with greater accuracy and validity nology modification and optimization. In the above process, enterprises
(Bloom et al., 2014). These steps do not require high costs. Second, since have stronger innovation capability and can actively implement the
the data analysis stage is an important fundamental stage of DT, com­ organization’s green strategy by relying on the innovation system of
panies often choose to recruit employees with complex knowledge skills production-learning-research collaboration and drive low-carbon prod­
and learning abilities to quickly codify and emulate the advanced uct innovation with LCTI.
knowledge extracted from the platform (Castellacci et al., 2018). In the
abovementioned process, enterprises are equipped with stronger 2.2.3. Mediating role of adaptive capacity
absorptive capacity to more accurately organize and assimilate prior In the data empowerment stage, DT facilitates LCTI mainly by
knowledge and previous experience of low-carbon technologies from enhancing adaptive capabilities. Adaptive capabilities focus on firms’
massive data, thus facilitating more high-quality low-carbon flexibility to adjust their strategies in a volatile environment, which
technologies. manifests themselves not only in financial strategies aimed at profit
maximization but also in diverse nonfinancial strategies, including
2.2.2. Mediating role of innovation capacity market penetration, customer satisfaction, and sustainability (Ovuak­
In the data operation stage, DT promotes LCTI mainly by enhancing porie et al., 2021). The digital enablement stage has contributed to the
innovation capability. Innovation capability is critical for companies to flexibility of both types of strategies related to LCTI. First, for financial
turn innovation strategies into innovation practices. However, the data strategies, although stakeholders’ expectations of corporate social and
operation stage enables enterprises to build a full-link data center, break environmental value are increasing, the potential benefits from these
the “R&D silo” and establish a collaborative innovation system among demands are still not visible to non-DT firms, making the cost of
production, learning and research, which is conducive to technology developing low-carbon technologies a “black hole” that diverts firms
output. First, for the collaboration of learning and research, firms from optimal production technologies and scale of production and sales.
effectively learn external information in the data analysis stage, while However, the construction of digital ecology can alleviate information
the mechanism of flow and interoperability in the data operation stage asymmetry, and its core function, i.e., the data sharing mechanism
promotes information cooperation with the R&D system and enhances among stakeholders (clients and those upstream and downstream of the
R&D efficiency. Second, for production and research collaboration, on supply chain) within the business ecological network, allows enterprises
the one hand, the advanced measurement infrastructure (sensors, etc.) to know and grasp the potential benefits of market demand (Bajari et al.,
and artificial intelligence technology in the smart grid can dynamically 2019) for low-carbon products and the potential costs of technology
track and predict the carbon emission level in the process of raw ma­ innovation; moreover, through intelligent cost‒benefit analysis models,
terial consumption, production and waste disposal, making it easier for enterprises can develop optimal technology attack strategies to

4
G. Yang et al. International Journal of Production Economics 263 (2023) 108969

maximize profits. Second, for nonfinancial strategies, through the control in waste management, and in the secondary category of carbon
expansion and aggregation of network space, companies move from trading in administrative regulation and design from the list of green
closed to open and gradually form a customer value-oriented organi­ patents applied for by firms and logged the number of the above­
zational philosophy (Qi and Xiao, 2020), which leads their strategic mentioned patent applications to measure firms’ LCTI levels. Next, we
decisions toward the practice of social responsibility, thus allowing calculated the proportion of the low-carbon technology patent appli­
companies to shift their value creation approach from that of pursuing cations of firms to all patent applications (LCTI_prop), which charac­
lower prices to that of contributing to sustainable development. In the terizes the attention paid by firms to low-carbon development during
abovementioned process, the digital empowerment stage enhances the their DT as a check variable. After further discussion, we divided the
adaptability of enterprises, enabling them to develop more scientific and low-carbon patents of sample firms into carbon-free technology, carbon-
flexible LCTI strategies to pursue low-carbon competitive advantages in reducing technology, and carbon-removing technology to examine the
a complex and changing environment. target of technological progress.
Based on the above analysis, this article proposes the following
research hypothesis. 3.2.2. Independent variables
H2. All other things being equal, DT promotes LCTI by enhancing
3.2.2.1. Firms’ degree of DT (DT). Existing studies have focused on two
firms’ DCs. Based on the classification of DCs, we further refine the
methods for measuring this variable: the quantitative descriptive and
above hypothesis into the following three points:
textual analysis methods. The quantitative description method refers to
H2a. DT promotes LCTI by enhancing absorptive capacity. obtaining the specific value of a company’s capital investment or results
in the DT process with the help of specific information channels (Liu
H2b. DT promotes LCTI by enhancing innovation capacity.
et al., 2021), such as questionnaires (Li, 2022), which often face diffi­
H2c. DT promotes LCTI by enhancing adaptive capacity. culty obtaining data, have a small sample size, and usually lack dynamic
change. The text analytics method involves the use of big data tech­
3. Research methodology nology to intelligently analyze the frequency of keywords related to DT
in specific text materials. China enforces a stringent information
3.1. Sample and data disclosure system for listed companies, which requires them to publish
annual reports on the websites of the Shanghai Stock Exchange and
Considering that the digital technology level and digital economy Shenzhen Stock Exchange for each fiscal year. These reports provide an
scale in China developed rapidly after 2010 (Yuan et al., 2021), and account of the company’s true operational status over the past year,
China implemented a lockdown policy to deal with the COVID-19 during including key information such as basic details, major accounting data
2020–2022 by restricting factor flow and enterprise production, we and financial indicators, stock and bond issuances, board reports,
select 2011–2019, when China’s economy is running relatively nor­ management discussions and analysis, among others. The China Secu­
mally, as the research window, and choose Shanghai and Shenzhen rities Regulatory Commission regularly reviews and examines the
A-share listed manufacturing firms as the research sample and excluded disclosure of annual reports by companies. Against the background of
them according to the following criteria: (1) firms in the financial in­ China’s strict disclosure system for listed companies, the annual report
dustry; (2) ST, PT and insolvent firms; and (3) firms with missing rele­ has become an important public document presenting the effectiveness
vant variables. Finally, we obtained 4868 firm-year observations and strategic direction of the company’s operation and is therefore a
containing 1124 listed manufacturing firms. There were two main data popular text for scholars both at home and abroad to measure DT (Zhai
sources: (1) firm innovation data mainly from the global patent database et al., 2022). This article selects the text analysis method. Based on Wu
(IncoPat) and (2) data on other firm characteristics, with financial data et al. (2021), we captured the frequency of keywords related to “un­
from the China Stock Market & Accounting Research (CSMAR) database derlying technology use” and “technology practice use” in the annual
and firm-related annual report data from the official websites of the reports of listed firms and calculated their logarithms as a proxy for the
Shenzhen and Shanghai Stock Exchanges. To reduce the influence of degree of firms’ DT. We use Python software and Java PDFbox library to
outliers, all microlevel continuous variables in this paper were subject to crawl the frequency of key words related to “underlying technology use”
1% and 99% tail reductions. and “technology practice use” in the annual reports of listed companies,
and logarithmically process them as a proxy for the degree of DT, and
the specific structured feature words are shown in Table 1.
3.2. Measures
3.2.3. Mediating variables
3.2.1. Dependent variable
3.2.3.1. Three types of DCs. Most existing studies measuring DCs have
3.2.1.1. Firms’ LCTI level (LCTI). Patent grant criteria have better sta­ been based on questionnaire data, but cross-sectional data cannot reflect
bility and objectivity, and the patent number is a more reliable indicator the evolution of firms’ DCs during the sample period. Therefore, we
that can reflect the innovation level. There are two main standards for constructed a panel data structure using firm characteristic variables to
distinguishing low-carbon technology patents: the International Patent measure the three DCs.
Classification (IPC) Green List and the Cooperative Patent Classification
(CPC)–Y02 classification. Since the patent entries of the State Intellec­ (1) Innovation capacity (innov). We draw on the standardization of
tual Property Office do not contain the relevant codes of the Y02 clas­ annual R&D investment intensity and percentage of technicians
sification (Zhu et al., 2019), this paper chooses the IPC Green List as the to quantify the innovation capability of the firm, which is
reference basis. This search entry was launched by the World Intellectual calculated by the following equation. where innov indicates
Property Organization (WIPO) in 2010 to classify green patents based on innovation capability, RI indicates the intensity of enterprise
the United Nations Framework Convention on Climate Change and is the R&D investment, and RP indicates the percentage of enterprise
most widely used and credible classification standard in academia technicians. We have two reasons for taking this approach. First,
today. this measure is consistent with the DC theory definition of inno­
To extract green innovations with “low-carbon” attributes, we vation capability and has some empirical support; e.g., Yang et al.
extracted patents in the primary category of alternative energy pro­ (2020) argues that the higher the metric is, the better the firm is
duction and energy savings, in the secondary category of pollution

5
G. Yang et al. International Journal of Production Economics 263 (2023) 108969

Table 1 capacity as a mediating mechanism and that educational level is

Keywords of DT. the most commonly used basis for distinguishing human resource
Major Category Minor Key words quality (Liu et al., 2018).
Category (3) Adaptive capacity (adapt). Adaptive capacity refers to strategic
Underlying Artificial Speech recognition, face recognition, flexibility, and Rindova and Kotha (2001) links the strategic
Technology Use Intelligence biometrics, intelligent robotics, artificial dimension to the resource perspective, arguing that strategic
intelligence, machine learning, natural flexibility and the firm’s flexibility to allocate available resources
language processing, business intelligence, are aligned. On this basis, Nadkarni and Narayanan (2007) and
deep learning, autonomous driving, identity
verification, image understanding,
Fombrun and Ginsberg (1990) use the coefficients of variation in
intelligent data analysis, semantic search sample firms’ annual R&D intensity (R&D expenditure/operating
Blockchain Digital currency, federated chain, revenue), capital intensity (capital expenditure/operating reve­
decentralization, smart contracts, nue), and advertising intensity (advertising expenditure/operat­
distributed computing, consensus
ing revenue) to measure their adaptive capacity. In addition, to
mechanism, bitcoin
Cloud Internet of Things, cloud computing, ensure that the indicators are consistent with the direction of
Computing streaming computing, cognitive computing, adaptive capacity, we use the negative value of the calculated
information physical systems, graph coefficient of variation; thus, the larger the value of the adjusted
computing, converged architectures, in- coefficient of variation is, the stronger the adaptive capacity. The
memory computing, EB-class storage, green
computing, billion-level concurrency,
specific calculation process is shown in equation (2), where adapt
multiparty secure computing, brain-like indicates the adaptive capacity of the firm; σ indicates the stan­
computing dard deviation of annual R&D, capital, and advertising intensity
Big Data Credit collection, data mining, virtual of the sample firms; and mean indicates the average of the three
reality, mixed reality, augmented reality,
expenditure intensities.
data visualization, heterogeneous data, text
mining, storage large data σ
adapt = − (2)
Technology E-commerce, mobile payment, mobile internet, internet mean
Practice Use finance, mobile internet, fintech, digital finance, smart
customer service, open banking, smart agriculture, smart 3.2.4. Control variables
investment advisor, digital marketing, smart marketing, smart To improve the precision of the study, we added a series of control
home, third party payment, industrial internet, netlink, smart
variables, such as firm maturity (lnage, the natural logarithm of the
wear, smart transportation, smart grid, internet medical,
business to business (B2B), business to consumer (B2C), number of years of business establishment), firm size (size, the natural
unmanned retail, online to offline (O2O), smart environmental logarithm of the total assets), equity concentration (own, percentage of
protection, smart medical, smart energy, near field shareholding of the company’s largest shareholder), firm growth
communication (NFC) payment, smart culture and tourism,
(growth, total assets growth rate), operating capacity (far, ratio of net
consumer to business (C2B), consumer to consumer (C2C),
quantitative finance, Fintech, digital staff, digital officer
fixed assets to total assets), profitability (roa, ratio of net income to total
assets), and social wealth creativity (tobinq, ratio of enterprise market
Note: This word base is summarized from the classic literature on DT and Chi­ value to replacement capital).
nese digital economy policy documents.
The settings and data structure of each variable are shown in Table 2.
The mean values of digits and patents are low, indicating that the average
able to translate strategic innovation directions into innovative level of low-carbon innovation output of manufacturing firms in China
behavior and thus develop new products. Second, this measure is during 2011–2019 was low and that their DT was still in the primary
consistent with the mechanism analysis framework of this paper; stage. Fig. 2 depicts the variance in and correlation between the vari­
i.e., DT’s promotion toward industry-university-research collab­ ables. The lengths of the arrows represent the variance, and the angles
oration is an important manifestation of innovation capability as between the arrows indicate how correlated the variables are—the
a mediator. Related scholars have also shown through empirical closer the angle is to 90◦ and 270◦ indicates less correlation, while the
tests that there is a significant positive correlation between en­ closer the angle is to 0◦ and 180◦ indicates absolute positive and nega­
terprise R&D investment (funds and personnel) and tive correlations. Both patents and digits vary widely and show a strong
industry-university-research collaboration innovation perfor­ positive correlation, which initially demonstrates the promotion effect
mance (Li et al., 2019). of DT on LCTI.
XRI − minRI XRP − minRP
innov = + (1)
maxRI − minRI maxRP − minRP
3.3. Model

(2) Absorptive capacity (absorb). In this paper, we use the percentage 3.3.1. Baseline regression model
of employees with graduate degrees and above in enterprises to Based on a two-way fixed effects model, we constructed the
measure absorptive capacity. There are two reasons for this. First, following model to test the effect of DT on LCTI, where LCTIit indicates
this measure is consistent with the definition of absorptive ca­ the level of LCTI of firm i in year t, and DTit is the degree of DT of firm i in
pacity in DC theory. Cohen and Levinthal (1990) mentioned that year t. Xit is a series of control variables containing other variables that
“the ability to assess and utilize external knowledge depends may affect the level of LCTI. δt indicates time fixed effects, θI indicates
largely on the level of prior knowledge”, and enterprises with industry fixed effects, and εit indicates the error term.
more educated employees have a more specialized knowledge
base and stronger learning ability and thus can absorb external LCTIi,t = α0 + α1 DTi,t− 1 + αc Xi,t + δt + θI + εit (3)
knowledge more quickly and transform it into effective imitation Since it takes time to effectively integrate digital technologies into
behavior, which has also been empirically tested by scholars in the core business and management mechanisms of firms, we treat the DT
the context of graded education in China (Huang et al., 2013). variable with a one-year time lag. In addition, for more robust results, by
Second, this measure is consistent with our theoretical and default, all regression equations take the robust standard error-adjusted
analytical framework, namely, that DT’s renewal and optimiza­ t-statistic of cluster clustering to individuals. According to the previous
tion of human capital is an important manifestation of absorptive theoretical analysis, if the coefficient α1 of DTi,t− 1 is significantly

6
G. Yang et al. International Journal of Production Economics 263 (2023) 108969

Table 2
Descriptive statistics of the variables.
Variable Variable Names N Mean S.D. Min. Max.

Dependent Variables Low-Carbon Technology Innovation Level LCTI 4804 0.134 0.476 0.000 4.700
Low-Carbon Technology Preference LCTI_prop 4804 2.15 0.250 0.000 12.000
Independent Variables Degree of Digital Transformation DT 4804 0.145 0.122 0.000 0.548
Mediating Variables Innovation Capability innov 3534 0.222 0.197 0.00016 1.724
Absorption Capacity absorb 4804 3.453 5.578 0.000 51.880
Adaptive Capacity adapt 4650 − 0.994 0.296 − 1.732 − 0.028
Control Variables Firm Maturity age 4804 2.763 0.373 0.693 3.951
Firm Size size 4804 22.230 1.203 19.550 27.150
Equity Concentration own 4804 32.964 14.364 3.000 89.990
Firm Growth growth 4540 0.159 0.477 − 0.707 15.890
Operating Capacity far 4804 0.207 0.126 0.001 0.700
Profitability roa 4804 0.042 0.064 − 0.290 0.223
Social Wealth Creativity tobinq 4804 2.199 0.649 0.000 3.296

significance can be measured using the Sobel test. The relationship

∑
among the above parameters is α1 = α′ + 3j=1 βj γ j .

4. Analysis and results

4.1. Direct effect of DT

4.1.1. Regression results

Table 3 reports the test results of Hypothesis 1. The regression co­
efficients of DT in columns (1)–(2) are all significantly positive, which
shows that the DT of firms effectively promotes low-carbon technolog­
ical innovation and makes them pay more attention to low-carbon
development, which is consistent with Hypothesis 1. In addition, to
further investigate whether digitalization promotes firms’ innovation in
only the low-carbon field, we use the natural logarithm of the total
number of firms’ patent applications in the current year to characterize
the level of general innovation (innovation) and substitute it as the
dependent variable in equation (3) to examine the competing relation­
ship between firms’ LCTI and general innovation under the digitaliza­
tion scenario. The regression results are shown in column (3), and the
key coefficients are significantly positive, indicating that DT promotes
firms’ total innovation; thus, the LCTI effect is manifested as a “leverage
Fig. 2. Two-dimensional plot of the variance of and correlations among effect” superimposed on existing technology innovations rather than a
core variables.
“crowding-out effect” on other technology innovations. Moreover, we
use the fixed asset investment rate to characterize enterprise investment
positive, then it indicates that DT significantly promotes LCTI in (invest) and the firm cash asset ratio to characterize cash flow (cash),
manufacturing firms, which is consistent with Hypothesis 1; conversely, which are substituted into the model as dependent variables, as shown
a nonsignificant coefficient indicates no significant promotion effect. in columns (4) and (5). Both key coefficients are significantly positive,
indicating that digital technology application is conducive to firms
3.3.2. Intermediation effect model overcoming financing difficulties and weakening resource constraints,
Based on the discussion in Chapter 2, we empirically test the inter­ providing additional investments to expand innovation, which indicates
mediary mechanism between DT and LCTI from a DC perspective. Since that digital technology application is beneficial for firms by helping
a total of three variables—innovation capability, absorptive capability, them overcome financing difficulties and weakening resource con­
and adaptive capability—are to be tested simultaneously, this problem is straints and by providing liquidity support for additional investments to
a parallel multiple mediating effect problem. To this end, we use a one- expand innovation and, thus, achieve the “leveraging effect” with low-
dimensional, parallel multiple mediator model to portray the mecha­ carbon technology.
nism path. Specifically, we introduce recursive equations for testing
based on equation (3), and the specific model is as follows: 4.1.2. Endogeneity problems
LCTIi,t = α0 + α1 DTi,t− 1 + αc Xi,t + δt + θI + εit (4) According to the empirical results in 4.1.1, firms’ DT enhances their
LCTI output. However, this finding may face endogeneity challenges,
medi,j,t = β0,j + βj DTi,t− 1 + βc,j Xi,t + δt + θI + εi,j,t (5) and thus, we attempt to identify the net effect through an instrumental
variable (IV) approach. Since geographic location is an exogenous var­
∑
3 iable from the economic system, we use the spherical distance between
LCTIi,t = γ 0 + α′DTi,t− 1 + γ j medi,j,t + αc Xi,t + δt + θI + εit (6) the firm’s city and the nearest coastal port and the spherical distance
j=1
between the firm’s city and the province’s capital city (Zhang et al.,
2020) as IVs for the degree of firm digitalization. For the first IV, on the
In the above equations, medi,j,t indicates the mediating variables (j =
one hand, the closer a firm is to the coastal port, the more advantageous
1,2,3) corresponding to innov, absorb and adapt, respectively. α1 is the
it is for the firm to participate in trade competition, obtain advanced
total effect of digitalization on LCTI, and α′ is the direct effect. α1 − α′ is experience and frontier technology, and be willing to adopt digital
the overall mediating effect; βj γj are individual mediating effects whose

7
G. Yang et al. International Journal of Production Economics 263 (2023) 108969

Table 3
Direct effect of DT.
VARIABLES (1) (2) (3) (4) (5)

LCTI LCTI_prop innovation invest cash

DT 0.281*** 9.556* 115.056*** 0.028** 0.043***

(2.74) (1.87) (3.06) (1.96) (3.16)
lnage 0.017 0.725 46.350*** − 0.002 0.007
(0.39) (0.66) (3.31) (-0.30) (1.42)
size 0.120*** 0.452 86.111*** 0.1123 − 0.010***
(5.65) (1.31) (20.65) (7.87) (-6.93)
own − 0.001 0.011 − 1.321*** − 0.000* 0.000**
(-1.09) (0.47) (-4.27) (-1.83) (2.56)
growth 0.017 0.480 − 29.339*** 0.010*** − 0.015***
(0.49) (0.58) (-3.28) (2.98) (-4.57)
far − 0.200** − 3.582 − 91.377** 0.436*** − 0.258***
(-2.49) (-1.23) (-2.35) (29.56) (-18.42)
roa − 0.205 0.361 151.733** 0.130*** 0.369***
(-1.25) (0.11) (2.15) (4.87) (14.53)
tobinq − 0.048 − 0.343 − 21.159** − 0.040*** − 0.013***
(-1.56) (-0.33) (-2.32) (-12.54) (-4.02)
Observations 4540 4540 4540 4526 4540
R-squared 0.158 0.022 0.128 0.279 0.283
Controls Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes

Note: *, **, and *** represent significance at the 10%, 5% and 1% levels, respectively; t values are in parentheses.

behavior to enhance its core competitiveness in the market and trade

Table 4
and improve its operational efficiency; on the other hand, cities adjacent
Endogeneity test.
to seaports tend to have a better industrial organization structure and
higher economic development level, and their digital economies are VARIABLES (1) (2) (3) (4)

more likely to be incubated and activated (Bai and Zhang, 2021), which first second first second
provides a more favorable growth environment for the firm to make DT LCTI DT LCTI
digital technology changes. For the second IV, provincial capitals tend to
DT 1.762** 2.150**
have the highest concentration of digital resources and the most (2.20) (2.08)
advanced digital technologies in the province, and thus, the closer the dist_seaport − 0.001***
location of the firm is to the provincial capital, the more it can be (-5.18)
affected by the radiation effect of the provincial capital city’s digital dist_capital − 0.003***
(-4.16)
economy and the more conducive it is to the DT of the firm. Based on
Observations 4316 4316 4314 4314
this, we follow the “National Coastal Port Layout Plan” issued by the R-squared 0.082 0.081
Ministry of Transportation, select ten coastal ports from north to Identifiable Test (P 27.670*** 19.812***
south—Dalian Port, Tangshan Port, Tianjin Port, Qingdao Port, Lia­ value) (0.00) (0.00)
nyungang Port, Shanghai Port, Ningbo Port, Xiamen Port, Guangzhou Weak IV test 29.532 17.300
(16.38) (16.38)
Port and Qinzhou Port—and calculate the spherical distance among Controls Yes Yes Yes Yes
them according to the latitude and longitude of each port and the Year FE Yes Yes Yes Yes
prefecture-level city where the firm is located. In addition, we calculate Industry FE Yes Yes Yes Yes
the spherical distance between each provincial capital city and the city Note: The above table uses the K-Paap Rk LM statistic for the identifiability test,
where the firm is located based on their latitudes and longitudes. In with the p value of the corresponding statistic in the lower brackets, and the
addition, to satisfy the time-varying characteristics of the IVs in the Cragg–Donald Wald F statistic for the weak instrumental variable test, with the
panel data, we calculate the cross-multipliers of the two types of Stock–Yogo weak ID test critical values at the 10% significance level in the lower
spherical distances with the year of the sample, denoted as dist_seaport brackets.
and dist_capital, respectively, and use two-stage least squares (2SLS) to
test the model for endogeneity. coefficients estimated by 2SLS. Fig. 3 shows the interval variation in the
The results are shown in Table 4. As indicated by the results in col­ DT coefficient under these UCI assumptions. The horizontal axis repre­
umns (1) and (3), there is a strong correlation between the two IVs and sents the tolerance of the deviation from “strict exogeneity” for both
the endogenous explanatory variables; as indicated by the results of the types of IVs, and the vertical axis represents the estimated interval of the
weak IV test, the Cragg–Donald Wald F statistic in each column is greater DT coefficient. As seen from the figure, the upper interval of the DT
than the Stock–Yogo critical values at the 10% significance level; i.e., coefficient rises as the degree of exogeneity decreases, and the interval is
both instrumental variables pass the weak instrumental variable test. always positive; thus, the impact of DT on LCTI is still significantly
For each column, the K-Paap Rk LM statistic for the identifiability test positive, even after considering the “nonstrict exogeneity” of the IVs.
rejects the original hypothesis at the 1% significance level; i.e., the IVs
satisfy identifiability. The second-stage regression results in columns (2) 4.1.3. Robustness tests
and (4) indicate that the effect of digitalization on LCTI remains robust
after using natural geographic variables as IVs. However, to be more 4.1.3.1. Change the metric of the DT level. First, the digital keywords
rigorous, we draw on Conley et al. (2012) to relax the “strict exogeneity” crawled in this paper can be divided into five dimensions: artificial in­
assumption of the instrumental variables and apply the Union of Con­ telligence, blockchain, cloud computing, big data, and digital technol­
fidence Intervals (UCI) method for additional tests. We obtain resolution ogy application. Considering that there are differences in the total
intervals of (0.188, 7510.311) and (0.803, 59.853), which cover the number of frequencies of feature words in different dimensions, there

8
G. Yang et al. International Journal of Production Economics 263 (2023) 108969

Fig. 3. Confidence intervals for the two types of IVs under the UCI assumptions.

may be errors when directly summing up the number of word fre­ strategic disclosure behaviors, we refer to Yuan et al. (2021) and retain
quencies in each dimension; thus, this paper standardizes the indicators only the sample of firms that have not been punished by the China Se­
of the five dimensions by the yearly deviation to eliminate the dimen­ curities Regulatory Commission (CSRC) or the Stock Exchange for in­
sion. These indicators are then summed to characterize DT levels and are formation disclosure during the sample period and re-examine such
denoted as DT_sd. Second, textual information about DTs in firms’ samples. In addition, we retain only the sample of listed companies with
annual reports may not represent the true level of these transformations. excellent or good information disclosure assessment results from the
In this article, we refer to Zhang et al. (2021) and use the proportion of Shenzhen Stock Exchange and re-examine such samples. However, the
the digital-technology-related part of total intangible assets in the carbon trading market’s pilot policies may also cause shocks to the re­
year-end intangible asset line items disclosed in the notes of the firms’ sults; therefore, we exclude those firms located in Beijing, Tianjin,
financial reports as the characterization variable of DT, which is recor­ Shanghai, and Guangdong after 2013, Hubei and Chongqing after 2014,
ded as DT_asset. Third, since macroscopic digital economy growth and and Fujian after 2016 according to when each pilot province and city
microscopic digital technology application growth in China basically started the pilot and perform regression tests on the remaining samples.
coincide, we refer to Zhao et al. (2020) to measure the digital economy The regression results from these three exclusions are shown in columns
development level in terms of both internet development and digital (5)–(7) of Table 5 and are all significantly positive at the 1% confidence
financial inclusion in the city where the firm is located, use it to replace level, which again verifies the robustness of the results.
digits, and note it as DT_city. The results are shown in columns (1)–(3) of
Table 5. The regression coefficients for the above three measures are all
significantly positive at the 1% confidence level, which proves the 4.2. Mediating effects of DCs
robustness of the core findings of 4.1.1.
4.2.1. Regression results
Table 6 shows the results of the tests based on the multiple mediating
4.1.3.2. Replace the regression model. Since the patent data have a left
effects model. In the first step, we calculate the overall mediating effect
truncated tail and the sample data of low-carbon patents used in this
of the three DCs as 0.015 based on the results in columns (1) and (5). The
article have a zero-value stacking phenomenon, we adopt a panel tobit
coefficient in column (5) is significantly lower than that in column (1),
model for supplementation. The regression results are shown in column
which proves the significance of the overall mediating effect. This result
(4) of Table 5. The coefficient is 0.11 and significant at the 10% level,
shows that it is reasonable to simultaneously use innovation capacity,
which proves that the core findings of 4.1.1 are still relatively robust
absorptive capacity and adaptive capacity as mediating variables. The
after considering the errors brought about by the censored variables to
second step involves estimating and testing individual mediating effects.
the model.
According to the results, the mediating effects via innovation capacity
and absorptive capacity are 0.034 and 0.03, respectively, and the z
4.1.3.3. Exclude part of the sample. First, to exclude the effect of firms’
values of the Sobel test relative to these effects are 2.737 and 2.223,

Table 5
Robustness tests.
VARIABLES (1) (2) (3) (4) (5) (6) (7)

LCTI LCTI LCTI LCTI LCTI LCTI LCTI

DT 1.054* 0.300*** 0.276*** 0.326***

(1.90) (5.01) (3.46) (5.38)
DT_sd 0.125***
(3.57)
DT_asset 0.219***
(4.12)
DT_city 0.037***
(3.74)
Observations 3745 4536 4269 4540 4475 2409 3864
R-squared 0.170 0.158 0.165 0.160 0.203 0.167
Controls Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes

9
G. Yang et al. International Journal of Production Economics 263 (2023) 108969

Table 6 firms, which means that although the adaptive capacity of low-tech
Mediating role of DCs. manufacturing firms plays a significant mediating role, there is no
VARIABLES (1) (2) (3) (4) (5) mediating effect of adaptive capacity for high-tech manufacturing firms
in the larger sample; therefore, the mediating effect of adaptive capacity
LCTI innov absorb adapt LCTI
under full sample regression is not significant. In summary, DCs play a
DT 0.286*** 0.205*** 5.995*** 0.046 0.154** partially mediating role in the facilitative effect of DT on LCTI, mainly
(4.83) (5.39) (4.01) (0.87) (2.33)
innov 0.167***
through the mediating role of innovation and absorptive capabilities,
(2.81) while adaptive capability is significant only for low-tech manufacturing
absorb 0.005*** firms.
(2.79)
adapt 0.011
5. Further analyses
(0.40)
Observations 4540 3352 4540 4407 3352
R-squared 0.159 0.479 0.209 0.155 0.159 5.1. Targeting of the low-carbon innovation effect
Controls Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes There are different characteristics of low-carbon technologies, and
Industry FE Yes Yes Yes Yes Yes
firms can choose to embed different types of low-carbon technologies in
various aspects of production and processing according to their needs to
which shows that the mediating effects of innovation capacity and finally achieve the source, process and end controls of CO2 emissions.
absorptive capacity are significant at the 1% and 5% confidence levels, Among them, the source control technology represented by clean energy
verifying H2(a) and H2(b). However, the Sobel test z value for adaptive technology is on the key path that enables firms to fundamentally reduce
capacity was − 0.239, indicating that the mediating effect of adaptive carbon emissions; thus, it is important to clarify the relative contribution
capacity was not significant, and H2 (c) was not verified. of DT to such an innovation to achieve a real zero-carbon future; how­
ever, the question has not been answered in existing studies. Based on
4.2.2. Contribution of technological properties this, we decompose the low-carbon technology patents of sample firms
Technology attributes may be able to explain the nonsignificant in a descending way: the first category involves zero-emission technol­
mediation effect of adaptive capacity, and this section attempts to pro­ ogy, which includes mainly nuclear energy, wind energy, solar energy,
vide empirical evidence. Specifically, there are significant differences in biomass and other renewable energy technologies, and this type of
adaptive capacity among firms with different technological attributes. technology has no carbon emissions as its fundamental feature and be­
High-tech firms, which use high tech as the basic condition for estab­ longs to source control. The second category involves carbon reduction
lishment and growth, require higher-quality human resources, tech­ technologies that improve efficiency, such as the clean and efficient
nology quality, financial reserves, and higher information sensitivity utilization of coal, exploration and development of oil and gas resources
and may have formed more stable development strategies and mature and coalbed methane. These technologies belong to process control. The
and perfect regulatory mechanisms. Thus, there is limited space for third category is carbon removal technologies, mainly carbon dioxide
digitalization to enhance the adaptive capacity of such firms. However, capture and storage (CCS), which realize the capture, transportation and
low-tech firms are prone to the “inertia trap” caused by path dependence geological storage of carbon elements and belong to end-of-pipe control.
because they have been in the traditional growth stage for a long time. According to the technology patent classification number information of
Because their adaptive capacity has more room for improvement, the sample firms, we distinguish the number of patent applications for
mediation mechanism of the adaptive capacity of DT may still hold in carbon-free, carbon reduction, and carbon removal technologies
such firms. Therefore, in this article, we further classify sample firms (recorded as carbon_free, carbon_reduc, and carbon_rem, respectively) and
into high- and low-technology manufacturing firms based on the defi­ standardize them to examine the targeting of the low-carbon innovation
nition of high- and low-technology industries in the World KLEMS effect.
database and Zhu et al.’s (2020) matching work using the industrial There are two findings from the results in Table 8. First, DT signifi­
industry classification criteria of the National Bureau of Statistics. We cantly enhances firms’ carbon-free technological innovation. Lin (2022)
then test the individual mediating effects of adaptive capacity for each of shows that digitalization helps mitigate the instability of new energy
the two types of sample firms. sources; for example, an intelligent grid can ensure a stable power
Table 7 shows that the Sobel test z value of the adaptive capacity supply under a large proportion of wind PV grid connections, which
mediating effect is 1.487 for high-tech firms and 3.015 for low-tech provides the necessary guarantee for enterprises to develop new energy

Table 7
Mediating effects of adaptive capacity by technological properties.
VARIABLES (1) (2) (3) (4) (5) (6)

High-tech firms Low-tech firms

LCTI adapt LCTI LCTI adapt LCTI

DT 0.401*** − 0.039 0.274*** 0.156* 0.355*** − 0.221*

(5.70) (-0.69) (3.47) (1.59) (2.78) (-1.78)
innov 0.110 0.340**
(1.62) (2.20)
absorb 0.004** 0.012**
(2.15) (2.12)
adapt − 0.027 0.060
(-0.78) (1.43)
Observations 3191 3122 2281 1349 1285 1071
R-squared 0.156 0.096 0.157 0.181 0.272 0.184
Controls Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes

10
G. Yang et al. International Journal of Production Economics 263 (2023) 108969

Table 8 Therefore, based on the “List of Listed Industries for Environmental

LCTI targets: Carbon free, carbon reduction and removal. Verification and Management” issued by the Ministry of Environmental
VARIABLES (1) (2) (3) Protection in 2008 and the “Notice of the General Office of the National
Development and Reform Commission on the Key Work for the Launch
LCTI_free LCTI_reduc LCTI_rem
of the National Carbon Emission Trading Market” issued by the National
DT 0.389*** 0.538*** 0.613*** Development and Reform Commission (NDRC) in 2016, we identify 13
(3.18) (4.30) (4.82)
Observations 4540 4540 4540
manufacturing subsectors, such as the paper and paper products in­
R-squared 0.174 0.157 0.123 dustry, pharmaceutical manufacturing industry, chemical fiber
Controls Yes Yes Yes manufacturing industry, and nonferrous metal smelting and rolling
Year FE Yes Yes Yes processing industries, as high-carbon industries.2 Based on equation (3),
Industry FE Yes Yes Yes
we conduct separate regressions on the samples of high- and low-carbon
firms to explore whether DT has a heterogeneous effect on LCTI under
technologies. Second, digitalization has also significantly increased the the different nature of carbon emissions.
level of innovation in both carbon reduction and removal technologies, The results are shown in Table 10. Columns (1)–(4) show that the
with these technologies being significantly more effective than effects of digitalization are significantly positive for the sample of low-
carbon-free technologies. The possible reasons for this are as follows. carbon firms but not for that of high-carbon industries, indicating that
First, limited by special energy endowment conditions, China has long the direct effect of DT on LCTI does not hold for all firms. For low-carbon
formed a coal-led energy consumption structure, and the new energy companies, which have a stronger sense of social responsibility, a
development force is insufficient. As shown in Table 9, in 2020, the cleaner development philosophy and a more solid innovation base to
proportion of coal energy consumption in China was as high as 56.6%, support them, DT has been effective; however, for high-carbon com­
the proportion of natural gas was only 8.2%, and there was still a large panies that truly need to stimulate clean innovation, on the surface, DT
gap between developed countries in Europe and the United States in does not work. For a thorough analysis, we focus on the intensity of
terms of the cleanliness of energy. Under this energy-use structure, the environmental regulations, an important factor that pushes the envi­
low-carbon technology progress of Chinese firms is biased toward the ronmental innovation behavior of high-polluting firms, given that some
process control and end control of traditional energy. Second, in recent studies have found a combined effect of digital technology and the
years, as the energy supply market has become more competitive, the traditional drivers of green innovation in the hope of exploring the
world has paid more attention to energy security issues, and environ­ combined effect of the DT of high-carbon firms and environmental
mental pollution has become more serious. Firms face energy supply governance policies in their regions. First, we refer to Li and Zou (2018),
constraints and, therefore, prefer to develop energy-saving and who used the comprehensive index method to measure the intensity of
emission-reducing equipment and technologies to ease these constraints environmental regulations (envir), standardized the three waste emis­
(He and Wang, 2015). Third, China’s new energy industry is still sions (industrial wastewater, industrial SO2, and industrial soot) in the
immature, and the core technology of new energy is relatively backward province where the enterprise is located, applied the entropy value
or even missing compared to that internationally, resulting in Chinese method to determine the index weights, and calculated the compre­
firms relying on the overseas introduction of new energy technology hensive index of environmental regulations based on the weights and
rather than independent R&D (Luo et al., 2015). Therefore, although standardized values; the higher the comprehensive index is, the stricter
digitalization has made a significant contribution to all three types of the environmental regulation in the firm’s location. Second, a regression
technological innovations, it favors innovation enhancement in terms of model containing the DT of firms (DT) and envir and their interaction
improving fossil-energy-use efficiency and carbon sequestration, while terms is constructed, as shown in equation (7). If the coefficient of the
the innovation enhancement effect of clean energy technologies that interaction term γ3 is significantly positive, then it indicates that for
fundamentally reduce carbon emissions is weaker. high-carbon firms, DT can still have an effect on the inverse effect of
low-carbon technological innovation by regulating environmental
regulation intensity.
5.2. Heterogeneity of carbon emission intensity
LCTIi,t = γ0 + γ 1 DTi,t− 1 + γ 2 envirc + γ3 DTi,t− 1 ∗ envirc + γ4 Xi,t + δt + θI + εit
(7)
The decarbonization of the industrial structure is an important focus
of the low-carbon economy. Compared with other firms, firms in high- Substituting the sample data into the model, as reflected in column
carbon industries produce more intensive carbon emissions either (5), the coefficient of the interaction term is significantly positive,
directly or indirectly in the production process (Liu and Zhao, 2017). To indicating that the marginal effect of environmental regulation on LCTI
fundamentally optimize the carbon structure of the manufacturing in­ increases as the level of DT increases; i.e., DT positively moderates the
dustry, it is necessary to promote the low-carbon behavior of backward effect of environmental regulation on LCTI in high-carbon
high-carbon firms as a key objective. Therefore, it is important to firms. Although the core coefficient in column (6) with the share of
examine whether the low-carbon technological innovation effects of DT LCTI as the explanatory variable is not significant, the regression results
differ between high- and low-carbon industries. However, there are few under the technology classification in columns (7)–(9) further comple­
studies on this issue, and a consistent view has not yet been formed. ment the robustness of the effect. Although the DT behavior of firms has
a limited effect on enhancing low-carbon innovation, it has a significant
Table 9
Structure of energy consumption in major countries (regions) in the world in
2020 (%).
2
High-carbon industries include agriculture and food processing industry,
Country (region) Coal Oil Natural Gas Others
food manufacturing, wine, beverage and refined tea manufacturing, textiles,
China 56.6 19.6 8.2 15.7 leather, fur, feathers and their products and footwear, paper and paper prod­
United States 10.5 37.1 34.1 18.3 ucts, petroleum processing, coking and nuclear fuel processing industry,
Europe 12.2 33.8 25.2 28.8 chemical raw materials and chemical products manufacturing, pharmaceutical
Germany 15.2 34.8 25.8 24.4 manufacturing, chemical fiber manufacturing, rubber and plastic products in­
Japan 26.8 38.1 22.1 12.9
dustry, nonmetallic mineral products industry, ferrous metal smelting and
India 54.8 28.2 6.7 10.3
rolling processing industry, nonferrous metal smelting and rolling processing
Data source: BP World Energy Statistics Yearbook industry, and metal products industry.

11
G. Yang et al. International Journal of Production Economics 263 (2023) 108969

Table 10
Heterogeneity analysis: Differences in carbon emission characteristics.
VARIABLES (1) (2) (3) (4) (5) (6) (7) (8) (9)

Low-carbon industry High-carbon industry

LCTI LCTI_prop LCTI LCTI_prop LCTI LCTI_prop LCTI_free LCTI_reduc LCTI_cs

DT 0.360*** 11.671* − 0.054 0.020 − 0.218*** − 0.584 − 0.633*** − 0.396** − 0.395**

(2.89) (1.86) (-0.76) (0.06) (-2.63) (-0.95) (-2.89) (-2.32) (-2.42)
envir − 0.019 − 0.076 − 0.072** − 0.035 − 0.043*
(-1.60) (-0.84) (-2.25) (-1.20) (-1.81)
DT*envir 0.021*** 0.078 0.052** 0.027* 0.035**
(2.64) (1.31) (2.44) (1.78) (2.23)
Observations 3437 3437 1103 1103 1103 1103 1103 1103 1103
R-squared 0.163 0.021 0.047 0.018 0.053 0.020 0.088 0.063 0.048
Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes Yes Yes Yes

“multiplier” effect on the government’s environmental governance; i.e., relatively weak digital infrastructure, such as that for information fa­
for high-carbon firms with the same regulatory intensity, those with a cilities (Zhu and Wang, 2021), and the cost of DT for firms is both high
high degree of digitalization can take the lead in seizing technological and risky; if there exist individuals (firms) who are similar to them in the
advantages to cope with environmental regulations and enhance green social network (i.e., peers) who have implemented DT, then they will
competitiveness. Digital technology is important for China in achieving actively learn and imitate to avoid the trial-and-error costs of digital
a low-carbon industrial structure and the construction of a high-quality, technology in their own production (Ni and Liu, 2021) and R&D activ­
ecological civilization in the new era. ities; thus, the digital technology in the peer group operates more effi­
ciently. On the other hand, competition among peer firms intensifies
their external market pressure, forcing them to adopt low-carbon tran­
5.3. Peer effect sition strategies and inducing low-carbon technological innovation be­
haviors. In contrast, nonpeer firms lack social network references that
The peer effect is an important perspective in the study of business can be imitated and face weaker competitive pressures; thus, the pro­
behavior. Individuals’ decisions are influenced by not only their own motion effect of DT on the LCTI behavior of such firms is weaker.
characteristics and needs but also the behavior of groups with similar Therefore, under the combined effect of information learning and
characteristics, i.e., “peers” (Mas and Moretti, 2009; Sacerdote, 2001). competition theories, DT has a stronger effect on the LCTI behavior of
The mechanism of the peer effect is derived from information learning peer firms compared to nonpeer firms.
theory, which suggests that firms learn from others’ decisions to reduce
information acquisition costs, and competition theory, which suggests 6. Discussion
that firms imitate to maintain competitive parity (Lieberman and Asaba,
2006; Smallwood and Conlisk, 1979). Existing studies have confirmed 6.1. Conclusions
the peer effect in terms of corporate investment and financing decisions
(Dougal et al., 2015), capital structure choices (Leary and Roberts, The research goals of this article are (1) to discuss the promotion
2014), etc., but there is a lack of research on firms’ digitalization effects of DT on LCTI; (2) to analyze the mechanism of the above effect
behavior. Further in-depth analysis from the perspective of peer effects based on DC theory and to discuss the mediating role of absorptive ca­
is beneficial for firms to make full use of cohort network relationships to pacity, innovation capacity, and adaptive capacity in each stage of DT;
build a good ecology for a low-carbon transition strategy. and (3) to further discuss the heterogeneity of the effect based on three
Therefore, in this article, the sample is divided into two groups with perspectives: technology category, industrial carbon intensity and en­
and without cohorts according to whether three or more firms in the terprise cohort characteristics. Through the empirical analysis of 1124
same city and in the same two-digit industry have implemented DT. manufacturing firms in China, this paper draws several conclusions.
Based on equation (3), regressions are conducted for each of the two First, DT effectively promotes LCTI, and the effect is a “leveraging
groups, and the results are shown in Table 11. DT has a significant effect” rather than a “crowding-out effect”, which is superimposed on
positive contribution effect on LCTI for both types of firms; the coeffi­ existing innovation. This finding echoes Loebbecke and Picot’s (2015)
cient is generally larger for firms with cohorts but not significant for idea that DT has enlivened corporate innovation dynamics in general.
either LCTI_prop. The results suggest that there is a significant peer effect Second, the theoretical mechanism shows that the data analysis, data
of DT on LCTI, and we attempt to explain this effect using information operation, and data empowerment stages of DT enhance the absorptive
learning and competition theories. On the one hand, China still has a

Table 11
Heterogeneity analysis: Peer effect.
VARIABLES (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Peer firms Nonpeer firms

LCTI LCTI_prop LCTI_free LCTI_reduc LCTI_cs LCTI LCTI_prop LCTI_free LCTI_reduc LCTI_cs

DT 0.355** 14.554 0.469 0.671** 0.737** 0.286*** 4.581* 0.387* 0.505** 0.460**
(2.45) (1.57) (1.63) (2.23) (2.48) (2.77) (1.87) (1.78) (2.45) (2.26)
Observations 2275 2275 2275 2275 2275 4540 2265 2265 2265 2265
R-squared 0.180 0.036 0.199 0.175 0.147 0.159 0.019 0.156 0.149 0.099
Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
Industry FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

12
G. Yang et al. International Journal of Production Economics 263 (2023) 108969

capacity, innovation capacity, and adaptive capacity of enterprises, which DCs are established and maintained within the digital economy’s
respectively, in turn promoting LCTI. Empirical tests support the context. Furthermore, it underscores the need for future empirical
mediating effects of the first two capabilities; however, adaptive ca­ research to explore the application value of DT from a process
pacity is validated to be effective only for low-tech firms due to the perspective.
limited increase in the strategic flexibility of high-tech firms by DT’s
marginal enhancement. Third, heterogeneity analysis leads to inter­ 6.3. Managerial insights
esting conclusions: the enhancement effect of DT on LCTI is stronger for
carbon reduction and decarbonization technologies; i.e., it is stronger In an increasingly volatile global market competition, Chinese
for process control and end-of-pipe control technologies than for source manufacturing companies are in urgent need of a low-carbon technology
control technologies, which is related to factors such as the tight global system driven by independent innovation, and the findings of this article
supply and demand of traditional energy sources (He and Wang, 2015), suggest that DT provides assistance for this purpose. Managers should
the lack of key technologies in new energy areas in developing countries target the development of different DCs at each stage of the enterprise
(Luo et al., 2015), and other factors. DT has a direct effect on LCTI only DT cycle. In the data analysis stage, managers should focus on devel­
for low-carbon emitting firms, and for high-carbon emitting firms, it oping absorptive capability, actively using big data platforms and arti­
positively moderates the promotion effect of environmental regulation ficial intelligence analysis for data collection, identifying beneficial
on low-carbon innovation, thereby extending Porter’s hypothesis. external knowledge (such as the latest green technology and R&D
Furthermore, the promotion effect of DT on LCTI is stronger for firms in experience of competitors) and coding it automatically, and building
the cohort network, which fully supports the necessity of building a teams with rapid learning capability for absorbing external knowledge
digital innovation network for the low-carbon transformation of effectively. In the data operation stage, managers should pay attention
manufacturing firms. to cultivating innovation capability, building a full-chain data center,
especially strengthening the linkage between green energy monitoring
6.2. Theoretical contributions and the waste emission real-time monitoring platform in the production
line and R&D department so that the R&D department can develop the
This article aims to complement the existing literature on digital appropriate technology based on the real-time emission data from the
transformation (DT) and green technology innovation by introducing a production plant and enhance innovation capability. In the data
new internal driving factor, namely, DT, based on the current digital empowerment stage, managers should focus on building an intelligent
economy landscape. While prior research has extensively examined cluster ecology, understanding the preferences of various stakeholders
green technology innovation and enterprise DT separately, this article through digital networks, and adjusting strategies and optimizing value
highlights their virtuous connection. On the one hand, literature sug­ creation paths more intelligently and flexibly in the process of aligning
gests that green technology innovation may significantly increase the with the external environment. In particular, low-tech manufacturing
production and operation costs of enterprises (Ardito et al., 2019; Huang companies should pay more attention to the construction of this stage.
et al., 2019), mainly driven by external policy push (Zhang et al., 2021) Through the above path, managers should take a moment to realize the
and demand pull (Farza et al., 2021). However, there is a lack of three-dimensional transformation of enterprise digitalization and iter­
research on internal factors that activate the green innovation dynamics. ative innovation of low-carbon technology.
On the other hand, DT has transformative impacts on various aspects of In addition, manufacturing enterprises in developing countries
corporate operations, organization, strategy, and culture (Qi and Xiao, should fully recognize that carbon-free technology is the key technology
2020), effectively reducing search costs, replication costs, trans­ with which to achieve a zero-carbon economy and the core technology
portation costs, tracking costs, and verification costs (Goldfarb and most valued and protected by developed countries and should make full
Tucker, 2019). This precisely bridges the difficulties encountered in use of the dividends of DT to conduct independent R&D to achieve the
low-carbon technological innovation. Previous literature on DT and source control of carbon emissions and gain key competitive advantages.
green technological innovation has predominantly focused on the macro High-carbon enterprises such as those in the agro-processing and textile
level, such as the application of regional information communication industries need to shift their mindset to recognize that environmental
technologies (Cecere et al., 2014) and the development of regional innovation can offset environmental management costs, improve prof­
digital finance (Feng et al., 2022a, 2022b), with discussions regarding itability (Chan et al., 2016), and autonomously promote the linkage of
micro-level corporate DT behaviors being relatively scarce. Thus, the digital strategies and green technology innovation to gain competitive­
first theoretical contribution of this article is discovering the positive ness under increasing environmental regulations. Finally, firms should
correlation between DT and LCTI. It reveals that DT can internally drive fully incorporate the information and experience from the cohort
LCTI, providing impetus for firms to invest in DT and offering avenues network into their own digital strategy making and implementation
for reducing the costs associated with low-carbon innovation. processes to play on its reference value.
This article delves into the core mechanism underlying DT’s driving
force - the enhancement of DCs. While prior research on the relationship 6.4. Research limitations and future directions
between DT and green innovation has mainly focused on the benefits
resulting from digital technology adoption, such as reduced knowledge This article exposes the dynamic process mechanism of DT via DCs to
search costs (Yin and Yu, 2022) and alleviated financing constraints (Liu excite LCTI and draws conclusions with some theoretical value and
et al., 2023), these mechanisms lack explanatory power. For instance, it practical insight, but there are still limitations. Therefore, it is necessary
remains uncertain whether a company would necessarily increase its to continue to expand and deepen the research in this area. First, gov­
research and development efforts in low-carbon technologies after ernment regulations may affect the regression results. Although the
acquiring cheaper funds and knowledge. Therefore, drawing on the rapid expansion of China’s digital economy was reflected mainly after
framework of dynamic capabilities, we elucidate how DT stimulates a 2010, the Chinese government issued several policies on the digital
firm’s low-carbon innovation vitality. Specifically, we contend that DT economy and environmental protection during 2011–2019, which may
facilitates firms’ effective interaction with the dynamic market envi­ interfere with firms’ DT and LCTI behaviors and thus have an impact on
ronment, thereby providing knowledge, guidance and strategy for LCTI. the regression results that is difficult to quantify and exclude. Second,
Moreover, building on extensive literature (Rowley, 2007; Battistoni the green technology innovation of enterprises may also be influenced
et al., 2023; Vial, 2019), we propose a phased dynamic mechanism by the public connections between countries, between governments and
framework, which represents a relatively rare perspective. This frame­ enterprises, and between the public and enterprises. For example, in
work provides a novel explanation for the micro processes through terms of country-to-country connections, countries should have

13
G. Yang et al. International Journal of Production Economics 263 (2023) 108969

“common but differentiated” responsibilities for addressing climate Chen, X., Hu, D., Cao, W., Liang, W., Xu, X., Tang, X., Wang, Y., 2021. Path of digital
technology promoting realization of carbon neutrality goal in China’s energy
change, and China, as a responsible power, has taken the initiative to
industry. Bull. Chin. Acad. Sci. 36 (9), 1019–1029.
propose a “dual-carbon” target and set incentives for green technology Cohen, W.M., Levinthal, D.A., 1990. Absorptive capacity: a new perspective on learning
innovation. In terms of connections between government and enter­ and innovation. Adm. Sci. Q. 35 (1), 128–152.
prises, enterprises with stronger political ties tend to have sufficient Conley, T.G., Hansen, C.B., Rossi, P.E., 2012. Plausibly exogenous. Rev. Econ. Stat. 94
(1), 260–272.
innovation capital and social responsibility; in terms of connections Dangelico, R.M., Pujari, D., Pontrandolfo, P., 2017. Green product innovation in
between the public and enterprises, public opinion monitors and regu­ manufacturing firms: a sustainability-oriented dynamic capability perspective. Bus.
lates enterprises’ emission behavior. The effects caused by these factors Strat. Environ. 26 (4), 490–506.
Dougal, C., Parsons, C.A., Titman, S., 2015. Urban vibrancy and corporate growth.
could not be effectively removed in this study. Finally, this research J. Finance 70 (1), 163–210.
shows that the role of DT is highly potential for the future global El-Kassar, A.N., Singh, S.K., 2019. Green innovation and organizational performance: the
response to green challenges and sustainable development; however, we influence of big data and the moderating role of management commitment and HR
practices. Technol. Forecast. Soc. Change 144, 483–498.
have examined green technology innovation only as a change in envi­ Farza, K., Ftiti, Z., Hlioui, Z., Louhichi, W., Omri, A., 2021. Does it pay to go green? The
ronmental behavior, while the contributions of DT to environmental environmental innovation effect on corporate financial performance. J. Environ.
outcomes such as carbon performance, green economic growth, and Manag. 300, 113695.
Feng, H., Wang, F., Song, G., Liu, L., 2022a. Digital transformation on enterprise green
global warming mitigation and mechanisms have yet to be explored. innovation: effect and transmission mechanism. Int. J. Environ. Res. Publ. Health 19
(17), 10614.
Feng, S., Zhang, R., Li, G., 2022b. Environmental decentralization, digital finance and
Funding
green technology innovation. Struct. Change Econ. Dynam. 61, 70–83.
Fombrun, C.J., Ginsberg, A., 1990. Shifting gears: enabling change in corporate
This work was supported by the Major Program of the National Fund aggressiveness. Strat. Manag. J. 11 (4), 297–308.
of Philosophy and Social Science of China (CN) (grant number Goldfarb, A, Tucker, C., 2019. Digital economics. J. Econ. Lit. 57 (1), 3–43.
Gong, Q., Ban, M., Zhang, Y., 2021. Blockchain, enterprise digitalization and supply
18ZDA040). chain finance innovation. Manag. World 37 (2), 22–34+23.
He, X., Wang, Z., 2015. Energy biased technology progress and green growth
transformation—an empirical analysis based on 33 industries of China. China Indust.
Econ. 32 (2), 50–62.
Declaration of competing interest Huang, X.X., Hu, Z.P., Liu, C.S., Yu, D.J., Yu, L.F., 2016. The relationships between
regulatory and customer pressure, green organizational responses, and green
The authors have no conflicts of interest to declare. innovation performance. J. Clean. Prod. 112, 3423–3433.
Huang, Y.P., Liu, Y., Wu, Y.Q., Li, W.B., 2013. Can corporate digital transformation
promote input-output efficiency? Econ. Res. J. 48 (4), 94–105.
Data availability Huang, Z., Liao, G., Li, Z., 2019. Loaning scale and government subsidy for promoting
green innovation. Technol. Forecast. Soc. Change 144, 148–156.
Jiao, H., Yang, J., Wang, P., Li, Z., 2021. Research on data-driven operation mechanissm
Data will be made available on request.
of dynamic capabilities based on analysisof digital transformation process from the
data lifecycle management. China Indust. Econ. (11), 174–192.
Acknowledgements Leary, M.T., Roberts, M.R., 2014. Do peer firms affect corporate financial policy?
J. Finance 69 (1), 139–178.
Li, L., 2022. Digital transformation and sustainable performance: the moderating role of
We would like to thank the editor and the anonymous referees for market turbulence. Ind. Market. Manag. 104, 28–37.
their valuable comments on this paper. Li, L., Shan, S., Dai, J., Che, W., Shou, Y., 2022. The impact of green supply chain
management on green innovation: a meta-analysis from the inter-organizational
learning perspective. Int. J. Prod. Econ. 250, 108622.
References Li, P., Li, M., Chen, W., 2019. Analysis of enterprise R&D investment and collaborative
innovation performance of industry-university-research. Stat. Decis. 35 (2),
183–185.
Aghion, P., Hepburn, C., Teytelboym, A., Zenghelis, D., 2019. Path dependence,
Li, H., Zou, Q., 2018. Environmental regulations, resource endowments and urban
innovation and the economics of climate change. In: Fouquet, R. (Ed.), Handbook on
industry transformation: comparative analysis of resource-based and non-resource-
Green Growth. Edward Elgar Publishing, Cheltenham, UK, pp. 67–83.
based cities. Econ. Res. J. 53 (11), 182–198.
Ardito, L., Petruzzelli, A.M., Ghisetti, C., 2019. The impact of public research on the
Lieberman, M.B., Asaba, S., 2006. Why do firms imitate each other? Acad. Manag. Rev.
technological development of industry in the green energy field. Technol. Forecast.
31 (2), 366–385.
Soc. Change 144, 25–35.
Lin, B.Q., 2022. China’s high-quality economic growth in the process of carbon
Arfi, W.B., Hikkerova, L., Sahut, J.M., 2018. External knowledge sources, green
neutrality. Econ. Res. J. 57 (1), 56–71.
innovation and performance. Technol. Forecast. Soc. Change 129, 210–220.
Liu, C.J., Zhao, X.M., 2017. Does strong porter hypothesis have industrial
Bai, P.Y., Zhang, Y., 2021. Digital economy, declining demographic dividends and the
heterogeneity?—the perspective from segmentation of industrial carbon intensity.
rights and interests of low-and medium-skilled labor. Econ. Res. J. 56 (5), 91–108.
China Popul., Resour. Environ. 27 (6), 1–9.
Bajari, P., Chernozhukov, V., Hortaçsu, A., Suzuki, J., 2019. The impact of big data on
Liu, H., Ke, W., Wei, K.K., Hua, Z., 2013. The impact of IT capabilities on firm
firm performance: an empirical investigation. AEA Pap. Proc. 109, 33–37.
performance: the mediating roles of absorptive capacity and supply chain agility.
Battistoni, E., Gitto, S., Murgia, G., Campisi, D., 2023. Adoption paths of digital
Decis. Support Syst. 54 (3), 1452–1462.
transformation in manufacturing SME. Int. J. Prod. Econ. 255, 108675.
Liu, S., Yan, J., Zhang, S., Lin, H., 2021. Can corporate digital transformation promote
Bloom, N., Garicano, L., Sadun, R., Van Reenen, J., 2014. The distinct effects of
input-output efficiency? Manag. World 37 (5), 170–190+113.
information technology and communication technology on firm organization.
Liu, X., Liu, F., Ren, X., 2023. Firms’ digitalization in manufacturing and the structure
Manag. Sci. 60 (12), 2859–2885.
and direction of green innovation. J. Environ. Manag. 335, 117525.
Cao, Y., Li, X., Hu, H.L., Wan, G.Y., Wang, S.Y., 2023. How does digitalization drive the
Liu, Y., Xie, K., Deng, H., 2020. Data-driven R&D transformation of new product: a case
green transformation in manufacturing companies? An exploratory case study from
study from the perspective of adaptive change in organizational practices. Manag.
the perspective of resource orchestration theory. Manag. World 39 (3), 96-112+126
World 36 (3), 164–183.
+113.
Liu, Z.Y., Li, H.Z., Hu, Y.Y., Li, C.H., 2018. Human capital structure upgrading and
Castellacci, F., Consoli, D., Santoalha, A., 2018. Technological Diversification in
economic growth: a reconsideration of disparities among China’s eastern central and
European Regions: the Role of E-Skills. Centre for Technology, Innovation and
western regions. Econ. Res. J. 53 (3), 50–63.
Culture, University of Oslo, Oslo, Norway.
Loebbecke, C., Picot, A., 2015. Reflections on societal and business model transformation
Castellacci, F., Natera, J.M., 2013. The dynamics of national innovation systems: a panel
arising from digitization and big data analytics: a research agenda. J. Strat. Inf. Syst.
cointegration analysis of the coevolution between innovative capability and
24 (3), 149–157.
absorptive capacity. Res. Pol. 42 (3), 579–594.
Lu, R., Zhou, Y., Ding, Y., Zhou, D., Fen, X., 2021. Enterprise innovation network:
Cecere, G., Corrocher, N., Gossart, C., Ozman, M., 2014. Technological pervasiveness and
tracing, evolution and research prospect. Manag. World 37 (1), 217–233+214.
variety of innovators in Green ICT: a patent-based analysis. Res. Pol. 43 (10),
Luo, L., Zhu, S., Zou, Z., 2015. Key technical challenges of Chinese new energy strategy
1827–1839.
and its countermeasures. J. Quant. Tech. Econ. 32 (2), 113–128+143.
Cennamo, C., Dagnino, G.B., Di Minin, A., Lanzolla, G., 2020. Managing digital
Mas, A., Moretti, E., 2009. Peers at work. Am. Econ. Rev. 99 (1), 112–145.
transformation: scope of transformation and modalities of value co-generation and
Mousavi, S., Bossink, B., Van Vliet, M., 2019. Microfoundations of companies’ dynamic
delivery. Calif. Manag. Rev. 62 (4), 5–16.
capabilities for environmentally sustainable innovation: case study insights from
Chan, H.K., Yee, R.W.Y., Dai, J., Lim, M.K., 2016. The moderating effect of
high-tech innovation in science-based companies. Bus. Strat. Environ. 28 (2),
environmental dynamism on green product innovation and performance. Int. J.
366–387.
Prod. Econ. 181, 384–391.

14
G. Yang et al. International Journal of Production Economics 263 (2023) 108969

Nadkarni, S., Narayanan, V.K., 2007. Strategic schemas, strategic flexibility, and firm Xue, L., Zhang, Q., Zhang, X., Li, C., 2022. Can digital transformation promote green
performance: the moderating role of industry clockspeed. Strat. Manag. J. 28 (3), technology innovation? Sustainability 14 (12), 7497.
243–270. Yang, L., He, X., Gu, H.F., 2020. Top management team’s experiences, dynamic
Nambisan, S., Wright, M., Feldman, M., 2019. The digital transformation of innovation capabilities and firm’s strategy mutation: moderating effect of managerial discretion.
and entrepreneurship: progress, challenges and key themes. Res. Pol. 48 (8), 103773. Manag. World 36 (6), 168-188+201+252.
Ni, K.J., Liu, X.Y., 2021. Digital transformation and enterprise growth logic and practice Yin, S., Yu, Y., 2022. An adoption-implementation framework of digital green knowledge
of China’s capital market. Econ. Manag. 43 (12), 79–97. to improve the performance of digital green innovation practices for industry 5.0.
Nordhaus, W.D., 2021. The Spirit of Green. Princeton University Press, Princeton, NJ. J. Clean. Prod. 363, 132608.
Ovuakporie, O.D., Pillai, K.G., Wang, C., Wei, Y., 2021. Differential moderating effects of Yu, C.H., Wu, X., Zhang, D., Chen, S., Zhao, J., 2021. Demand for green finance:
strategic and operational reconfiguration on the relationship between open resolving financing constraints on green innovation in China. Energy Pol. 153,
innovation practices and innovation performance. Res. Pol. 50 (1), 104146. 112255.
Qi, Y., Xiao, X., 2020. Transformation of enterprise management in the era of digital Yuan, B., Cao, X., 2022. Do corporate social responsibility practices contribute to green
economy. Manag. World 36 (6), 135–152+250. innovation? The mediating role of green dynamic capability. Technol. Soc. 68,
Qiu, L., Jie, X., Wang, Y., Zhao, M., 2020. Green product innovation, green dynamic 101868.
capability, and competitive advantage: evidence from Chinese manufacturing Yuan, C., Xiao, T.S., Geng, C.X., Sheng, Y., 2021. Digital transformation and division of
enterprises. Corp. Soc. Responsib. Environ. Manag. 27 (1), 146–165. labor between enterprises: vertical specialization or vertical integration. China
Rennings, K., 2000. Redefining innovation — eco-innovation research and the Indust. Econ. (9), 137–155. https://doi.org/10.19581/j.cnki.
contribution from ecological economics. Ecol. Econ. 32 (2), 319–332. ciejournal.2021.09.007.
Rindova, V.P., Kotha, S., 2001. Continuous “morphing”: competing through dynamic Zhai, H., Yang, M., Chan, K.C., 2022. Does digital transformation enhance a firm’s
capabilities, form, and function. Acad. Manag. J. 44 (6), 1263–1280. performance? Evidence from China. Technol. Soc. 68, 101841.
Rowley, J., 2007. The wisdom hierarchy: representations of the DIKW hierarchy. J. Inf. Zhang, J., Ouyang, Y., Ballesteros-Pérez, P., Li, H., Philbin, S.P., Li, Z., Skitmore, M.,
Sci. 33 (2), 163–180. 2021. Understanding the impact of environmental regulations on green technology
Sacerdote, B., 2001. Peer effects with random assignment: results for dartmouth innovation efficiency in the construction industry. Sustain. Cities Soc. 65, 102647.
roommates. Q. J. Econ. 116 (2), 681–704. Zhang, X., Yang, T., Wang, C., Wan, G.H., 2020. Digital finance and household
Santoalha, A., Consoli, D., Castellacci, F., 2021. Digital skills, relatedness and green consumption: theory and evidence from China. Manag. World 36 (11), 48–63.
diversification: a study of European regions. Res. Pol. 50 (9), 104340. Zhao, T., Zhang, Z., Liang, S.K., 2020. Digital economy, entrepreneurship, and high-
Shan, Y., Xu, H., Zhou, L., Zhou, Q., 2021. Digital and intelligent empowerment: how to quality economic development: empirical evidence from urban China. Manag. World
form organizational resilience in crisis? An exploratory case study based on forest 36 (10), 65–76.
cabin’s turning crisis into opportunity. Manag. World 37 (3), 84–104+107. Zhou, K.Z., Li, C.B., 2012. How knowledge affects radical innovation: knowledge base,
Smallwood, D.E., Conlisk, J., 1979. Product quality in markets where consumers are market knowledge acquisition, and internal knowledge sharing. Strat. Manag. J. 33
imperfectly informed. Q. J. Econ. 93 (1), 1–23. (9), 1090–1102.
Song, W., Yu, H., 2018. Green innovation strategy and green innovation: the roles of Zhou, M., Govindan, K., Xie, X., Yan, L., 2021. How to drive green innovation in China’s
green creativity and green organizational identity. Corp. Soc. Responsib. Environ. mining enterprises? Under the perspective of environmental legitimacy and green
Manag. 25 (2), 135–150. absorptive capacity. Resour. Pol. 72, 102038.
Teece, D.J., Pisano, G., Shuen, A., 1997. Dynamic capabilities and strategic management. Zhou, X., Ye, W.P., Li, X.C., 2023. Digit-intellectualized knowledge orchestration and the
Strat. Manag. J. 18 (7), 509–533. evolution of organizational dynamic capabilities: a case study of Xiaomi technology.
Vial, G., 2019. Understanding digital transformation: a review and a research agenda. Manag. World 39 (1), 138–157.
J. Strat. Inf. Syst. 28 (2), 118–144. Zhu, H.L., Wang, C.J., 2021. Industry digitalization against the strategic background of
Wan, P., Yang, M., Chen, L., 2021. How do environmental technology standards affect the new development paradigm: theory and countermeasures. Finance Trade Econ.
the green transition of China’s manufacturing industry—a perspective from 42 (3), 14–27.
technological transformation. China Indust. Econ. (9), 118–136. Zhu, J., Fan, Y., Deng, X., Xue, L., 2019. Low-carbon innovation induced by emissions
Wang, C.L., Ahmed, P.K., 2007. Dynamic capabilities: a review and research agenda. Int. trading in China. Nat. Commun. 10 (1), 4088.
J. Manag. Rev. 9 (1), 31–51. Zhu, M., Zhang, L.M., Peng, D.J., 2020. The transformation of China’s industrial
Wu, F., Hu, H.Z., Lin, H.Y., Ren, X.Y., 2021. Enterprise digital transformation and capital structure and its potential economic growth rate. Soc. Sci. China 11, 149–171+208.
market performance: empirical evidence from stock liquidity. Manag. World 37 (7), Zollo, M., Winter, S.G., 2002. Deliberate learning and the evolution of dynamic
130–144+110. capabilities. Organ. Sci. 13 (3), 339–351.

15