Technological Forecasting & Social Change 189 (2023) 122329

Contents lists available at ScienceDirect

Technological Forecasting & Social Change

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

Theorizing the relationship between the digital economy and firm

productivity: The idiosyncrasies of firm-specific contexts
Zhe Sun a, b, Liang Zhao c, Puneet Kaur d, e, f, Nazrul Islam g, *, Amandeep Dhir f, h, i
a
Center for China Public Sector Economy Research, Jilin University, China
b
Economics School, Jilin University, China
c
Department of Marketing, Innovation, Strategy and Operations, University of Leicester School of Business, UK
d
Department of Psychosocial Science, University of Bergen, Norway
e
Jaipuria Institute of Management, Lucknow, Uttar Pradesh, India
f
Optentia Research Focus Area, North-West University, Vanderbijlpark, South Africa
g
Centre of Fintech, Royal Docks School of Business and Law, University of East London, UK
h
Department of Management, School of Business & Law, University of Agder, Norway
i
Norwegian School of Hotel Management, University of Stavanger, Stavanger, Norway

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

Keywords: With the rise of emerging economies such as China, the research environment for the digital economy (DE) has
Digital economy changed significantly. However, our understanding of the productivity impact of DE development in Chinese
Firm productivity firms remains in its infancy. The idiosyncrasies of the firm-specific contexts are closely related to further research
U-shaped relationship
on the this topic. As a baseline, we hypothesize a U-shaped DE-firm productivity (FP) relationship. We analyze
Firm-specific context
Chinese firms
the idiosyncratic influences of firm size and locality on the DE–FP relationship. The findings, which are based on
a sample of Chinese firms from 2016 to 2019, show that (a) the U-shaped DE–FP relationship applies to Chinese
firms; (b) this relationship is moderate for large firms, substantially steeper for medium firms, and inverted for
small firms; (c) the U-shaped DE–FP relationship for eastern region firms is moderate, while the U-shaped
relationship for central region firms is steep, but the transition is incomplete, and western region firms have
experienced increasing productivity since the early stage of DE development. This study offers an alternative
approach to understanding Chinese firms’ strategic choices in DE development and provides a more nuanced
explanation for the productivity paradox by emphasizing the significance of the firm-specific context. In this way,
the study captures the sophisticated and constantly evolving relationships between DE and FP for heterogeneous
Chinese firms.

1. Introduction term “productivity paradox” to describe the fact that the rapid devel­
opment of ICT-related industries has not brought about a significant
According to the 2016 G20 Digital Economy Development and increase in productivity. Several subsequent studies have evaluated the
Cooperation Initiative, the digital economy (DE) is a collection of eco­ validity of the productivity paradox in developed countries, such as the
nomic activities in which digitized knowledge and information serve as United States, the United Kingdom, and Japan. Dewan and Kraemer
essential factors of production, the modern information network func­ (2000) and Stanley et al. (2018) concluded that the paradox had van­
tions as a key carrier, and the effective use of information and ished in developed countries, while Lin and Shao (2006) observed its
communication technology (ICT) is a significant driving force for effi­ persistence.
ciency improvement and ecological sustainability. Notably, DE has Most studies on the productivity paradox have taken place in
become increasingly important in increasing global productivity. developed countries. However, the research environment has undergone
Many empirical studies have examined the relationship between the remarkable changes with the rise of emerging economies, such as China.
efficient use of ICT and productivity. Unfortunately, these attempts have DE development has emerged as a prominent strategy and a key driver of
yielded inconsistent, if not conflicting, results. Solow (1987) coined the high-quality growth in China (Pan et al., 2022), with significant

* Corresponding author.
E-mail address: [email protected] (N. Islam).

https://doi.org/10.1016/j.techfore.2023.122329
Received 15 September 2021; Received in revised form 7 December 2022; Accepted 20 December 2022
Available online 18 January 2023
0040-1625/© 2023 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Z. Sun et al. Technological Forecasting & Social Change 189 (2023) 122329

productivity implications (Cai and Zhang, 2015; Li and Wu, 2020; Sun context is China, which is undoubtedly the most active economy in the
et al., 2012) and opportunities to understand the productivity conse­ world in terms of DE development. In fact, China’s DE development has
quences for Chinese firms, both theoretically and empirically. The overtaken that of the United Kingdom and Japan, and it now ranks
growing relevance of DE in today’s competitive environment has second in the world. The China Academy of Information and Commu­
prompted research on the digitization strategies of Chinese firms nications Technology (2021) published a white paper on the develop­
(Heshmati and Kumbhakar, 2011; Lee et al., 2011; Liu et al., 2014; Sun ment of China’s DE, estimating that the volume of DE reached 39.2
et al., 2017). As more Chinese firms undergo digital transformation, trillion yuan in 2020 and accounted for 38.6 % of the country’s GDP; this
understanding the effects of digitization initiatives is vital. The various represented an increase of 3.3 trillion yuan from the previous year. We
strategies employed by Chinese firms in the DE era may impact pro­ predict, moreover, that China’s position in the global DE will continue to
ductivity, but the extent of this impact is unclear. This constitutes a improve because of its significant latecomer advantage. Therefore, it is
significant omission in the existing research. necessary to develop a deeper understanding of the role that Chinese
More importantly, scholars have recently acknowledged the need to firms play in the competitive DE environment.
identify the idiosyncrasies of firm-specific contexts in considering firms’ This research is structured into the following sections. Section 2
digitization strategies (e.g., Lin and Shao, 2006). Consistent with this proposes the theoretical framework and develops the hypotheses; Sec­
viewpoint, Lin and Shao (2006) reported that firm size influences the tion 3 describes the methodology, including the sample information and
payoffs of digitization. It is, however, empirically uncertain whether DE variables; Section 4 analyses the empirical results; and Section 5 dis­
is more beneficial or detrimental to large, medium, or small firms (Sun cusses the findings and provides conclusions from the research.
et al., 2017). In addition, a firm’s locality influences its digitization
strategy as well as its potential to improve efficiency (Pan et al., 2022). 2. Theoretical framework and hypotheses development
We do not yet know whether the productivity paradox persists or van­
ishes in Chinese regions at various stages of DE development (Chen and 2.1. Firms’ digitization strategies: digitized versus non-digitized
Xie, 2015).
Previous studies have assumed firm-specific contexts to be “mere At a basic level, firms can choose between two types of digitization
background” information in addressing Chinese firms’ digitization strategies: digitized and non-digitized. These two types of strategies vary
choices. They have thus largely ignored the importance of firm size and in five aspects. According to the framework in Table 1, the first char­
firm locality in shaping those choices. This is a significant lacuna in the acteristic of a digitized strategy is that firms using it believe their value
literature because while all firms place a premium on DE growth, firms creation derives from heterogeneous production and services; hence,
vary greatly by firm size and locality. Without a thorough knowledge of they focus on offering high-quality heterogeneous products and services
the effects of size and locality on firms’ strategic choices and produc­ (Porter and Heppelmann, 2016). The second critical feature of a digi­
tivity, it becomes challenging to comprehensively analyze the produc­ tized strategy is that firms rely on their extremely proactive and inno­
tivity outcomes of Chinese firms’ digitization strategies. vative employees to engage in creative work (Frynas et al., 2018). Third,
Our analysis addresses these gaps by merging the literature on DE, a digitized strategy necessitates a high degree of diversification for firms
firms’ strategic choices, and firm-specific contexts. More explicitly, we to run efficiently. In other words, firms must implement diversification
extend previous research to Chinese firms and incorporate the idiosyn­ to realize the complex economy of scope. The fourth characteristic of a
crasies of firm-specific contexts into the DE–FP model. As previously digitized strategy is that it is often organized into a small-world network
indicated, inconsistent findings in the ICT–productivity relationship with a flat structure (Cui et al., 2015; Porter and Heppelmann, 2016).
prompted us to conduct this study. We also respond to the increasingly The last characteristic of a digitized strategy is that a firm’s resource
acknowledged need to rethink conventional wisdom regarding the allocation is defined by network-based mechanisms in which the supply
ICT–productivity relationship in the context of Chinese firms. We found and demand sides are linked inside a network; these might include one-
that the DE–FP relationship is curvilinear, meaning that the productivity to-one targeted marketing, contextual pricing, and partnership (Cheng
paradox persists at the early stage of DE development but gradually et al., 2018; Xie et al., 2016).
fades as DE progresses. While prior research on the productivity paradox Firms with a non-digitized strategy, on the other hand, focus on
has implied a linear relationship (e.g., Dewan and Kraemer, 2000; Lin producing a large volume of homogeneous goods and services (Ace­
and Shao, 2006), we propose a non-linear relationship (e.g., Chen and moglu and Restrepo, 2018). In general, firms that pursue a non-digitized
Xie, 2015). strategy believe that their homogeneous goods and services create value.
Our study was also motivated by the need to examine the role of firm- In a non-digitized firm, the organization is structured bureaucratically in
specific idiosyncrasies in firms’ strategic choices and productivity in the a pyramid form (Ward et al., 2005). Employees work in a standardized
development of the DE in China. Prior studies have largely neglected manner, with managers determining their work’s style and substance.
these endeavors. We include firm-size and firm-locality variables to Managers expect employees to be highly specialized in order to work
investigate the impact of firm-specific contexts. Our study demonstrates better with machines and gain economies of scale from homogeneous
that a moderate curvilinear relationship is particularly evident in firms
in China’s eastern region and in large firms, whereas the curvilinear
relationships are steep in central region firms and in medium firms. Table 1
Meanwhile, the U shape is inverted for small firms, and western region Digitization strategies of firms: Differences between non-digitized and digitized
firms have experienced increasing productivity since the early stage of strategies.
DE development. Non-digitized strategy Digitized strategy
Using an unbalanced panel data set of 10,739 observations on Chi­ 1. Orientation Quantity oriented: Quality oriented:
nese listed firms for the 2016–2019 period, we tested the influence of production and services production and services
firm size and firm locality on the DE–FP relationship. We then applied homogeneity heterogeneity
the China “Internet Plus” Digital Index to measure the most recent de­ 2. Labor force Standard, reactive Creative, active
3. Technological High specialization for the High diversification for the
velopments in the DE–FP relationship. In addition, we used the Haus­
paradigm economy of scale economy of scope
man test to determine whether a productivity paradox exists in China 4. Organizational Pyramid-shaped, Small-world network, flat
based on the panel data fixed effect model and illustrate DE’s impact on management bureaucracy structure
FP in a systematic manner. We also conducted sub-sample analyses by 5. Resource Market mechanism based on Network mechanism based
firm size and locality to empirically examine the DE–FP relationship allocation price competition on quality competition

across firms of varying sizes or from various regions. Our empirical Source: Own elaboration.

2
Z. Sun et al. Technological Forecasting & Social Change 189 (2023) 122329

production. In this way, non-digitized firms are able to participate in the increased productivity.
market mechanism of resource allocation based on price competition for Third, the flow of data resources among different industries has
their homogeneous goods and services. become a typical example in the new DE business format, blurring the
boundaries between industries (Cui et al., 2015). Cross-border opera­
2.2. The choices of firms’ digitization strategies in the context of DE tions allow firms to distribute resources more efficiently and achieve
development economies of scope. Manufacturing firms might leverage big data to
provide customers with tailored value-added services and shift from
Using the framework described above for firms’ digitization strategy, homogeneous to heterogeneous manufacturing depending on customer
we next investigate how firms choose between alternative digitization experiences. Service firms might also evolve from single-service sup­
strategies. To address this question, we analyze the cost structures of the pliers to multi-service providers and platforms. As a traditional real
two types of strategy at various stages of DE development as well as the estate middleman, Lianjia has evolved into one of the industry’s first
associated productivity advantages. models of a cross-border operator using digital technology. Lianjia
DE development includes well-developed digital technology, digital launched an online real estate service platform, “Lianjia Online” (later
infrastructure, digital finance, and digital government. Assuming that called lianjia.com), in 2010. In 2013, barely four years after the Internet
DE development is exogenous, the fixed cost for firms to use it will be platform was founded, lianjia.com accounted for 24 % of Lianjia’s total
minimal once it is established. In other words, advanced DE develop­ income. This example demonstrates that firms may overcome industry
ment allows firms to adopt a digitized strategy at a minimal fixed cost barriers in the context of DE, enabling the efficient and effective allo­
and a minimal and declining marginal cost. Specifically, the integration cation of resources across industries and the achievement of economies
of digital technology enables the low-cost implementation of a flat and of scope (Dyer et al., 2018).
flexible organizational structure with small-world networks and moti­ Furthermore, due to the development of DE, firms can achieve high
vated employees (Bloom et al., 2016). In Haier’s “ren dan he yi,” for productivity through network-based resource allocation mechanisms.
example, the terms “ren” and “dan” refer to employees and user orders, For example, during a product’s R&D stage, firms can collect user data,
respectively. Simply put, “ren dan he yi” indicates that a firm breaks its such as transaction information and evaluation feedback, to analyze
objectives down into user orders and assigns those orders to the consumer needs and establish a precise connection between product
appropriate employees who are accountable for them. The management innovation and consumer demand. This information, in turn, provides a
department evaluates employees’ performance by assessing the fulfill­ critical reference for firms to assess market demand and potential
ment of each order. “Ren dan he yi” is a creative model that allows firms business risks (He et al., 2019). In contrast, R&D staff cannot precisely
to become platforms for integrating global resources. When they are comprehend the actual needs of customers in the conventional closed
trusted and empowered, employees can fully unleash their potential, R&D paradigm; therefore, innovation failure often results (Cheng et al.,
achieve self-growth, and create value for users. Through zero-distance 2018). DE has successfully overcome this challenge and significantly
employee–user interactions, moreover, firms can better understand reduced the trial-and-error costs and innovation risks associated with
customers’ personalized demands and provide them with heterogeneous R&D (Johnson et al., 2017). Meanwhile, a new-generation ICT-based
goods (Frynas et al., 2018). Notably, embedding DE facilitates the for­ digital collaborative innovation platform has created an open innova­
mation of small-world networks both inside and outside of firms. For tion environment (Mikalef and Pateli, 2017). This platform allows pro­
instance, instant office communication software, such as WeChat, spective players to participate in a firm’s R&D and innovation processes
DingTalk, and Lark, enable rapid communication between various de­ (Varian, 2010). The use of a digital collaborative innovation platform
partments within firms and between the firm and the external envi­ can considerably reduce the search costs for R&D resources and effi­
ronment (Bauer et al., 2015). This significantly improves ciently integrate dispersed R&D resources (Cui et al., 2015).
communication efficiency and lowers information transmission costs. The Xiaomi ecosystem is an example of a supplier-side network-
More importantly, a digitized strategy encourages firms to focus on based mechanism of resource allocation. The Xiaomi ecosystem is built
diversification and achieve economies of scope at a low cost through on its IoT development platform, which is an open platform with
cross-boundary operations. DE development, moreover, makes this extensive support. All Xiaomi ecosystem firms as well as third-party
easier. First, the development of new-generation ICT provides firms with developers interested in connecting their hardware products to the
digital R&D tools that allow them to attain product diversification at Xiaomi platform are welcome to access and use these services. Such
extremely low marginal costs. For example, digital simulation systems firms can maximize their profits by using the open platform and its
and virtual reality technology (Graetz and Michaels, 2018) allow firms derived low (or even zero) entrance costs. From the demand-side
to simulate physical entities, which enables the R&D of diversified va­ network-based mechanism of resource allocation, a digital platform
rieties in complex application scenarios (Zhang et al., 2018). The integrating various firms’ business information has arisen, bridging the
development of new-generation ICT also minimizes resource losses information divide between lenders and borrowers. Gathering necessary
associated with repetitive mold development, thus lowering R&D costs information and effectively determining the true status of firms from
(Vaccaro et al., 2011). traditional banking institutions is challenging (Brown et al., 2009). Such
Second, the development of the DE helps firms to achieve a more institutions are often hesitant to lend when extensive information
precise and efficient manufacturing process while reducing production asymmetry exists. One of the implications is that firms face severe
costs. Firms can integrate dispersed production data into a unified financing constraints, thus limiting investment in R&D and innovation
production management system using digital technologies, such as big (Howell, 2015).
data, artificial intelligence, and the Internet of Things (Frynas et al., The use of digital technology expands opportunities for dissemi­
2018). Decentralized and intermittent states of R&D, design, nating information and increases the efficiency of obtaining informa­
manufacturing, and quality control can be transformed into continuous tion. Meanwhile, the growth of digital finance has intensified
and integrated states via DE advancement (Bloom et al., 2016). This has competition in the financial sector and expanded external financing
the potential to significantly increase the efficiency and precision of channels. The emergence of this novel business format has significantly
production. For example, the Midea Group engages in data exchanges improved the matching efficiency of credit resources and successfully
with suppliers by utilizing its supplier collaboration cloud. Suppliers resolved the issue of financing constraints. For example, Sesame Credit
gain real-time access to Midea’s current inventory and manufacturing is a private credit scoring and loyalty program system developed by
schedule. Meanwhile, Midea has access to suppliers’ material inventory Alibaba Group. Through its own e-commerce platforms, such as Taobao,
and logistical data. The production arrangement can thus be tailored to Alipay, and Tmall, Alibaba Group has accumulated rich data on small
each manufacturing line and hour, resulting in accelerated delivery and and micro enterprises in the fields of trade, distribution, logistics, and

3
Z. Sun et al. Technological Forecasting & Social Change 189 (2023) 122329

customer evaluation. These data have been used to determine the Digitized
creditworthiness of small and micro firms. Sesame Credit delivers
dependable credit investigation services on this premise. Credit audits, 2 1
which formerly required a significant amount of time, effort, and offline
operations, have been steadily phased out, significantly increasing Low cost
High cost
firms’ financing efficiency. Additionally, governments can utilize the
digital information platform to assess firms’ innovation potential and
facilitate the efficient allocation of innovation subsidies, tax incentives, DE development
and other policy resources (Chen and Xie, 2015).
Nevertheless, firms may, at some point, encounter a “digital vac­
uum” (Stieglitz et al., 2016). During the DE transition, the application of Low cost High cost
the traditional non-digitized strategy declines, while the new digitized 3
strategy is not yet effective, resulting in a digital vacuum. We argue that 4
this digital vacuum often occurs when DE development is at a moderate
stage (neither excessively low nor extremely high; Brynjolfsson and Hitt,
Non-digitized
2000). At this point, firms create certain internal digital infrastructure
and bear the expense of doing so endogenously. Compared to the early Fig. 1. Cost structure of the digitization strategy at various stages of DE
stage of DE development, firms have a reduced fixed cost at the mod­ development.
erate stage. However, firms sometimes bear significant marginal costs Source: Own elaboration.
associated with compatibility difficulties (Dyer et al., 2018; Ward et al.,
2005). For example, employees must adjust to a new work pattern that fixed and marginal costs would be minimal. The fourth quadrant is the
shifts them from reactive to proactive work, although many of them are most unique, with firms opting for a non-digitized strategy at an
unfamiliar with these changes (Ward et al., 2005). advanced stage of DE development. In this scenario, non-digitized firms
It is increasingly critical for employees to gain new skills and ways of would face strong competition from digitized firms, forcing them to
thinking (Stieglitz et al., 2016; Xie et al., 2016) to adapt to the use of a devote considerable human and physical resources to improving their
digitized strategy. Hence, firms must, therefore, increase investment in homogeneous goods and services and thus increasing costs.
professional training for employees and recruit new employees (Matt Considering the various cost structures of digitization strategies at
et al., 2015; Watson, 2017). Meanwhile, the emergence of new tech­ different stages of DE development, the productivity (cost) advantage of
nologies poses significant problems and threats to network and data a non-digitized strategy makes it the most logical choice for firms at the
security (Kim et al., 2015). As a result, it imposes additional expecta­ early stage of DE development. However, at the moderate stage of DE
tions on employees’ data availability cognition (Wagner et al., 2012) as development, firms might opt to endure the painful transition from a
well as their knowledge regarding relevant laws and ethics (Agarwal non-digitized to a digitized strategy to better compete in the future when
et al., 2010). Consequently, labor costs increase (Acemoglu and DE development progresses to an advanced stage. Because firms are
Restrepo, 2018; Borland and Coelli, 2017; Frey and Osborne, 2017). likely to encounter personnel and organizational compatibility diffi­
Additionally, firms may fail to develop the necessary organizational culties, productivity plummets at this stage. When DE reaches a mature
coordination to capitalize on the benefits associated with the DE level, the digitized strategy becomes the most sensible choice due to its
(Arvanitis and Loukis, 2009). Typically, when firms implement digital remarkable productivity benefits.
technology as a new power, they must systematically adjust their Fig. 2 illustrates the productivity advantages of digitized and non-
manufacturing, commercial, and organizational processes to achieve digitized strategies. The thick curve depicts the theoretical relation­
compatibility (Porter and Heppelmann, 2016). Internal communication ship between DE and FP as DE develops. In sum, the relationship be­
via a digital platform may impair the flexibility and openness of the tween the DE and FP will be curved (U-shaped).
information processing and transmission processes, resulting in infor­ Thus, we proposed the following hypothesis:
mation distortion and further impairing decision-making efficiency
(Pesch and Endres, 2019). In sum, personnel transformation and Hypothesis 1. A U-shaped relationship exists between the develop­
compatibility issues that emerge during the moderate stage of DE ment of DE and FP.
development make it difficult to realize the payoffs of the DE while also
increasing fixed and marginal costs when firms opt for a digitized
strategy.
In comparison, when DE development is in its infancy and firms opt
for a digitized strategy, they must construct these digital infrastructures
entirely on their own. The fixed cost thus becomes endogenous, making
digitization expensive for firms and ultimately affecting productivity. In
such cases, it is more productive for firms to maintain a non-digitized
strategy (i.e. homogeneous production and services) rather than a
digitized one. Based on the above, Fig. 1 presents a quadrant diagram
illustrating the cost structures of digitized and non-digitized strategies at
various stages of DE development. The horizontal axis depicts the stages
of DE development, while the vertical axis shows firms’ digitization
strategies. In the first quadrant, at an advanced stage of DE develop­
ment, firms would opt for a digitized strategy with exceptionally low
fixed and marginal costs. Firms in the second quadrant, meanwhile,
would adopt a digitized strategy at the early stage of DE development.
The fixed and marginal costs would be high for firms in this scenario
because the cost of developing digital technology and infrastructure
must be internalized. In the third quadrant, firms would select a non- Fig. 2. Productivity advantage of digitized and non-digitized strategies.
digitized strategy at the early stage of DE development. Here, both the Source: Own elaboration.

4
Z. Sun et al. Technological Forecasting & Social Change 189 (2023) 122329

2.3. The idiosyncrasies of firm-specific contexts must plan ahead for a seamless transition. As a result, large firms may
endure a slower decline in productivity throughout the DE development
As previously stated, we argue that firms will experience a digital process as well as a lengthier period of digital vacuum during the
vacuum when transitioning from a non-digitized to a digitized strategy. transition. In sum, the U-shaped link between the development of the DE
However, the duration of this digital vacuum moment will vary across and FP may be moderate for large firms.
firms in different contexts, depending on the difficulty they encounter in Medium firms, on the other hand, may experience fewer compati­
resolving compatibility issues during the transition to and development bility issues. As a result, medium firms are likely to face a shorter period
of the DE (Howell, 2015). Thus, the firm-specific context is critical. In of digital vacuum. They may complete the transition more quickly than
this paper, we concentrate on two critical firm-specific contexts: firm large firms, allowing for a longer ascending productivity process. In
locality and firm size. sum, the U-shaped link between DE and FP may be substantially steeper
for medium firms.
2.3.1. Firm locality In contrast, small firms are the most adaptable (Krishnan et al., 2015)
China is an expansive country characterized by distinct DE devel­ because they can respond swiftly by taking advantage of the DE while
opment stages in its eastern, central, and western regions. The eastern avoiding the complicated compatibility issues. Meanwhile, because the
region is the most developed in terms of the DE, and eastern firms are complexity of digital technology and infrastructure is not as critical for
more advanced in their digital transformation than are those in the most small firms, they may introduce a low-cost application of digital
central and western regions (Pan et al., 2022). As the first mover in the technology at the early stage of DE development. Such a digitized
digital transformation, however, eastern firms have less expertise with strategy will dramatically and rapidly increase small firms’ productivity.
transition and compatibility concerns. Due to their inexperience and the For example, the growth of e-commerce trading platforms and mobile
need for more complicated digital infrastructure to support their oper­ payment alleviates market segmentation caused by space–time separa­
ations, firms in the eastern region may experience a lengthier produc­ tion while also widening the sales channels available to small firms (Bell
tivity decline. The link between the DE and FP, in turn, may assume a et al., 2012; Brynjolfsson and Hitt, 2000). When, as a result, the mar­
moderate U-shaped structure in this scenario. ginal cost of capacity expansion is minimized, economies of scale occur
Firms in the central region, while at a less developed DE stage, may (Hsieh and Klenow, 2009). For example, following the outbreak of the
benefit from a latecomer advantage when transitioning to a digitized COVID-19 pandemic, Guangdong Xinbao Electrical Appliances began
strategy. By developing their digital strategy later than firms in the focusing on the e-commerce model not only by utilizing social media
eastern region, they may experience a shorter digital vacuum (Bonfa­ influencers to promote sales via live streaming but also by experiment­
delli, 2002). However, firms in the central region require sophisticated ing with other online channels, such as new retail and new marketing, to
digital technology and infrastructure to enable their digital trans­ reach more consumers. Through its use of e-commerce, Xinbao Electrical
formation. Therefore, they may choose to maintain their existing non- Appliances has consistently built brand recognition, reduced market
digitized strategy to preserve comparatively higher but rapidly drop­ reaction time, and achieved a net interest rate of 15 %. However, as the
ping productivity. Hence, the U-shaped relationship between the DE and DE evolves and more firms adopt a digitized strategy, small firms will be
FP may be steeper for firms in China’s central region. forced to compete more vigorously, resulting in a decline in their pro­
Firms in the western region, on the other hand, are the least devel­ ductivity. The U-shaped relationship between DE and FP in small firms,
oped (Pan et al., 2022). The intricacy of the DE is thus not as important similar to firms in the western region, may thus become inverted.
to them as it is to their counterparts in the eastern and central regions. As We present the following hypotheses based on the above argument:
a result, they can benefit from implementing DE within the organization
Hypothesis 3a. A moderate U-shaped relationship exists between the
at a very low cost throughout the early stages of DE development.
DE and FP for large firms.
Meanwhile, the development of DE has decreased the cost of inter-
regional resource mobility, thereby accelerating the migration of Hypothesis 3b. A steep U-shaped relationship exists between the DE
labor, money, and other resources from developed to developing regions and FP for medium firms.
(Du and Zhang, 2021). Consequently, firms in the western region, as
Hypothesis 3c. An inverted U-shaped relationship exists between the
latecomers, may capitalize on DE development to significantly and
DE and FP for small firms.
rapidly enhance FP by avoiding errors and learning from their peers.
Thus, rather than experiencing a digital vacuum, western firms may
3. Research design
experience rapid productivity growth by adopting a digitized strategy.
When firms in the eastern and central regions complete the trans­
3.1. Sample and data
formation and market rivalry intensifies, however, firms in the western
region are likely to experience declining productivity.
To empirically test the proposed U-shaped relationship, we collected
We present the following hypotheses based on the above argument:
a large set of data from multiple sources. Specifically, we chose the firms
Hypothesis 2a. A moderate U-shaped relationship exists between the listed in China’s A-share market, including the Shanghai Stock Exchange
DE and FP for firms in the eastern region. and Shenzhen Stock Exchange, as our sample. The firm-level informa­
tion was from the China Stock Market & Accounting Research database
Hypothesis 2b. A steep U-shaped relationship exists between the DE
and Wind, while the city-level information was from China City Statis­
and FP for firms in the Central region.
tical Yearbook. To ensure the validity of the data, we eliminated the
Hypothesis 2c. An inverted U-shaped relationship exists between the following firms from the sample: (a) firms that had suffered consecutive
DE and FP for firms in the Western region. losses (marked as ST or *ST); (b) firms in the finance industry; (c) firms
with a serious lack of information; (d) firms with a registered address
2.3.2. Firm size that did not match the city-level data.
Firm size matters significantly in resolving compatibility difficulties. Due to limited data availability on the DE, we restricted the sample
By virtue of their complicated bureaucracy, large firms confront greater period to 2016–2019. To reduce the influence of extreme observations,
compatibility challenges in terms of employees, business processes, and we winzorized all continuous variables at 1 %. Ultimately, we obtained a
culture (Du and Zhang, 2021). Given similar stages of DE development, total of 10,739 observations.
large firms find it more difficult than small firms and are, therefore,
hesitant to shift from a non-digitized to a digitized strategy. Large firms

5
Z. Sun et al. Technological Forecasting & Social Change 189 (2023) 122329

3.2. Variables 3.3. Model specification

3.2.1. Dependent variable To investigate the impact of DE on the productivity of Chinese firms,
To measure FP, we used the natural logarithm of total factor pro­ we established the following panel model for empirical research:
ductivity (Tfp). We employed the Wooldridge method to measure total
Tfpi,t = α + β1 Digitaldexc,t + β2 Digitaldex2c,t + γEi,t + δCc,t + μh + μt + εi,t
factor productivity at the firm level and adopted the Levinsohn–Petrin
(LP) method and the LP method with Ackerberg–Caves–Frazer correc­
where i denotes individual firms, c denotes cities, and t denotes time in
tion (LP–ACF) for the robustness test. The LP method takes an inter­
years. Tfpi,t is the productivity of firm i in year t; Digitaldexc,t is the digital
mediate product input as the proxy variable, which ensures that the
index of the city c where the firm was registered in year t; Digitaldex2c,t is
selection of the proxy variable is flexible according to data availability
introduced as an exponential square term to determine whether a
(Levinsohn and Petrin, 2003). When using the LP method to estimate the
nonlinear relationship exists between the DE and FP; Ei,t and Cc,t
total factor productivity, the free variables and proxy variables must be
represent the control variables at the firm and city level, respectively; μh
independent of one another, or the collinearity between the estimation
is the industry fixed effect of the model; μt is the year fixed effect of the
coefficients may impair the consistency of the estimation results
model; and εi,t is the error term of the model. In addition, we clustered
(Ackerberg et al., 2015). The LP–ACF method remedies this problem.
the robust standard errors at the city level to mitigate the estimation
Wooldridge (2009) suggested a one-step estimation technique based on
result deviation caused by the correlation of firms within the city.
the generalized method of moments. It not only eliminates the potential
identification difficulty of the LP method but also yields a robust stan­
4. Empirical results
dard error when considering sequence correlation and
heteroscedasticity.
4.1. Full-sample regression analysis
In the specific calculation, according to Giannetti et al. (2015),
Guariglia et al. (2011), and Krishnan et al. (2015), the total output (Y) is
Table 2 presents the descriptive statistics and correlations of all
measured as the main business income of the firm, the labor input (L) is
variables in the model. None of the correlation coefficients among the
measured as the cash paid to and for employees, the capital input (K) is
control variables and independent variables were high. The results also
measured as the net fixed assets, and the intermediate product input (M)
suggest that multicollinearity was not a concern in the samples. The full-
is calculated as the “operating cost + sales expense + administrative
sample regression results of the panel data fixed effect model through
expense + financial expense − depreciation of fixed assets − depletion
the Hausman test are shown in Table 3. Model 1 contains only control
of oil and gas assets − depreciation of productive biological assets −
variables and acts as the baseline model. In Models 2 and 3, we added
cash paid to and for employees.”
the independent variable (Digitaldex) and the square of the independent
variable (Digitaldex2) stepwise to the baseline model. Model 2 analyzes
3.2.2. Independent variable
the linear relationship between the development of the DE and FP. The
We used the China “Internet Plus” Digital Index (Digitaldex) to
regression coefficient of Digitaldex was negative at the significance level
measure the development of the DE. The Tencent Research Institute has
of 5 %, indicating that the development of the DE inhibits the produc­
released the China “Internet Plus” Digital Index since 2016. Since 2019,
tivity growth of Chinese firms to some extent. Model 5 examines the
however, the index has been termed the Digital China Index. The index
nonlinear relationship between the development of the DE and FP. It
is compiled based on big data from Tencent, JD.com, DiDi, meituan,
shows that the coefficient of Digitaldex remained negative at the signif­
Ctrip, Pinduoduo, dianping.com, and other digital platforms, using a
icance level of 1 %, while the coefficient of Digitaldex2 was positive at the
wide variety of indicators, such as cloud consumption, number of cloud
significance level of 5 %. Among the control variables, the coefficient of
virtual machines, flow rate of content delivery network, and storage of
size was positive at the significance level of 1 %, and the coefficients of
continuous data protection as well as WeChat and QQ data. Meanwhile,
pft and gdp were negative at the significance levels of 1 % and 5 %,
using indicators such as mobile payment, e-commerce, WeChat urban
respectively, while the coefficients of age, loar, contl, and popu were
services, and WeChat subscriptions of government administration, the
insignificant. The regression results reveal a U-shaped relationship be­
index also depicts the development level of digital finance and digital
tween the development of the DE and FP.
government in cities. Therefore, we utilized the China “Internet Plus”
Notably, the value of the turning point of the U-shaped curve can be
Digital Index to match the firm data. In doing so, we were able to
calculated from the regression results. The turning-point values of the U-
determine the development stage of the DE from various perspectives,
shaped curve shown in Models 3, 4, and 5 are 29.7, 27.7, and 28.4,
including digital infrastructure, digital finance, and digital government
respectively. All three fall within the interval of the independent vari­
(Du and Zhang, 2021; IMF, 2018).
able [0.0462, 35.7336], meaning that the relationship between the
development of the DE and FP was primarily nonlinear during the
3.2.3. Control variables
sample period (i.e., 2016–2019). Therefore, we selected the nonlinear
We included a set of firm- and city-level control variables with the
model containing Digitaldex2 for empirical analysis. The value of the
potential to influence the dependent variable Tfp (Guariglia et al.,
turning point of the U-shaped curve indicates that when the value of
2011). Among the firm-level control variables, we measured firm size
Digitaldex was lower than the turning point value of 28.4, the develop­
(size) as the natural logarithm of the total assets of a firm, firm age (age)
ment of the DE inhibited FP. In contrast, when the Digitaldex value
as the natural logarithm of the number of years since the firm was
exceeded 28.4, the development of the DE promoted FP.
established, the proportion of fixed assets in total assets (pft) as the ratio
Based on the regression results, we used Stata 16.0 to fit the U-sha­
of net fixed assets to total assets, the asset–liability ratio (loar) as the
ped curve. In Fig. 3, the solid line indicates the impact of the develop­
ratio of total liabilities to total assets, and equity concentration (contl) as
ment of the DE on FP during the sample period, while the dotted line
the proportion of the largest shareholder in all shares. Among the city-
represents a possible future trend in the impact of the development of
level control variables, we measured urban economic development
the DE on FP. The value on the horizontal axis represents the level of DE
level (gdp) as the natural logarithm of the GDP of the city where the firm
development. A smaller value indicates low levels of DE development,
had registered and urban permanent population (popu) as the natural
and a higher value suggests higher levels. The value on the vertical axis
logarithm of the permanent population of the city where the firm had
is FP. A smaller value implies a lower FP, and a higher value represents a
registered.
higher FP. Consistent with Hypothesis 1, between 2016 and 2019, the
productivity of Chinese firms first declined and then increased with the

6
Z. Sun et al. Technological Forecasting & Social Change 189 (2023) 122329

Table 2
Descriptive statistics and correlations between study variables.
Variables Observations Mean Std. Dev. Min Max 1 2 3 4 5 6 7 8 9

1. Tfp 10,739 12.9845 1.0027 10.9255 15.8319 1.000

2. Digitaldex 10,739 7.6954 9.3959 0.0462 35.7336 0.088 1.000
3. size 10,739 22.2661 1.3272 19.9349 26.3050 0.815 0.065 1.000
4. age 10,739 5.3886 0.3042 4.5539 5.9789 0.135 − 0.009 0.168 1.000
5. pft 10,739 0.1959 0.1507 0.0020 0.6663 − 0.080 − 0.203 0.089 0.045 1.000
6. loar 10,739 0.4149 0.1999 0.0610 0.8913 0.510 0.035 0.543 0.170 0.051 1.000
7. contl 10,739 33.6769 14.4517 8.4468 73.0561 0.180 0.027 0.185 − 0.032 0.091 0.060 1.000
8. gdp 10,739 18.1206 1.1205 15.2074 19.6049 0.075 0.702 0.043 − 0.045 − 0.253 0.016 0.043 1.000
9. popu 10,739 6.4366 0.6644 4.5799 8.1242 0.061 0.387 0.072 0.010 − 0.130 0.005 0.043 0.676 1.000

Note. *** p < 0.01, ** p < 0.05, * p < 0.1.

Table 3
Regression results for the full sample.
Variable Model 1 Model 2 Model 3 Model 4 Model 5

Digitaldex − 0.0012** − 0.0050*** − 0.0054*** − 0.0047***

(− 2.0911) (− 2.8082) (− 3.3276) (− 2.8235)
2
Digitaldex 0.0001** 0.0001*** 0.0001**
(2.1958) (2.7102) (2.1685)
size 0.4376*** 0.4377*** 0.4379*** 0.4380***
(18.5742) (18.5478) (18.4668) (18.5720)
age 0.1079 0.1181 0.1140 0.1122
(0.9052) (0.9853) (0.9502) (0.9482)
pft − 1.1249*** − 1.1240*** − 1.1250*** − 1.1234***
(− 11.0790) (− 11.0896) (− 11.1021) (− 11.0704)
loar 0.0110 0.0118 0.0115 0.0125
(0.2092) (0.2227) (0.2148) (0.2363)
contl − 0.0004 − 0.0004 − 0.0004
(− 0.2073) (− 0.2163) (− 0.2171)
gdp − 0.0256*** − 0.0256*** − 0.02321**
(− 3.0883) (− 2.7287) (− 2.4570)
popu − 0.0568 − 0.0323 − 0.0144
(− 1.0143) (− 0.5680) (− 0.2570)
Constant 3.6097*** 3.3978*** 14.2283*** 3.5055*** 3.2748***
(4.7031) (4.4179) (204.1064) (5.1603) (4.2711)
Observation 10,739 10,739 10,739 10,739 10,739
R-squared 0.37 0.37 0.17 0.37 0.37
Year fixed effect Yes Yes Yes Yes Yes
Industry fixed effect Yes Yes Yes Yes Yes

Note. Robust t-statistics in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

development of the DE. Digitaldex2. The results of Model 3 indicate that the development of the
DE first inhibited and then promoted FP during the sample period.
4.2. Robustness analysis Therefore, the findings of the full-sample regression analysis remain
valid when considering the potential endogeneity issues in the model.
4.2.1. Endogeneity test
We applied an instrumental variable estimation method to avoid 4.2.2. Robustness test
potential endogeneity bias in the model. Consistent with Du and Zhang We also investigated the robustness of our initial empirical results.
(2021), we divided the cities where the sample firms were registered First, we used the LP method and the LP–ACF method to examine the
into three categories: municipalities directly under the central govern­ robustness of the measurement of the dependent variable (Tfp). The
ment, sub-provincial cities, and non-sub-provincial cities in the same regression results for Models 1–4 are presented in Table 5.
province. The China “Internet Plus” Digital Index of cities in the same Second, we replaced the independent variable Digitaldex with the
category (excluding the city where the sample firm was registered) was Peking University Digital Inclusive Financial Index (Dfdex) to examine
used as an instrumental variable. We chose this instrumental variable the robustness of the measurement of the independent variable (Digi­
primarily because Chinese cities within the same category typically taldex; Du and Zhang, 2021). This index is jointly prepared by the
exhibit significant similarities in terms of their economic development, Institute of Digital Finance Peking University and the Ant Group. It
population density, and administrative authority. In addition, their DE objectively and comprehensively reflects the actual development of
development levels are strongly correlated. digital financial inclusion in China by creatively establishing an index of
Table 4 presents the endogeneity test results. Model 1 shows the digital financial inclusion from the perspective of innovative Internet
regression results of the first stage of the two-stage least square (2SLS). finance. The regression results appear in Models 5 and 6 of Table 5. The
The estimated coefficient of the instrumental variable is positive at the 1 results suggest that after replacing the dependent and independent
% significance level, and the value of the first-stage F-statistic is variables, the U-shaped relationship between the development of the DE
529.554, thus exceeding the critical value at a significance level of 1 %. and FP persisted in China during the sample period. This is consistent
This means that the instrumental variable is effective, and the weak with the full-sample regression results.
instrumental variable issue does not exist. The results for the second- Third, to examine the robustness of the estimation method, we
stage 2SLS for Models 2 and 3 are presented in Table 4, before and tightened the fixed effects by including the city fixed effect in the
after the introduction of the square term of the endogenous variable regression model. We clustered the robust standard errors at the firm

7
Z. Sun et al. Technological Forecasting & Social Change 189 (2023) 122329

Fig. 3. U-shaped relationship between the development of DE and FP.

4.3. Sub-sample regression analysis

Table 4
Regression results for the endogeneity test.
We conducted a sub-sample regression analysis on the sample firms
Model 1 Model 2 Model 3 by locality and size to comprehensively investigate the impact of DE
Digitaldex Tfp Tfp development on the productivity of Chinese firms based on firm-specific
Digitaldex − 0.0050* − 0.0489** characteristics.
(0.0028) (0.0209)
Digitaldex2 0.0012** 4.3.1. Sub-sample regression analysis by firm locality
(0.0006) We divided the sample into three sub-samples based on firm locality:
Instrumental 0.6416***
variable (0.02788)
firms in the eastern region, firms in the central region, and firms in the
size 0.2017*** 0.0748*** 0.0737*** western region. The eastern region includes Beijing, Shanghai, Tianjin,
(0.5510) (0.0052) (0.0053) Liaoning, Hebei, Shandong, Jiangsu, Zhejiang, Fujian, Guangdong, and
age − 1.0943*** − 0.0471*** − 0.0515*** Hainan; the central region includes Heilongjiang, Jilin, Shanxi, Henan,
(0.1912) (0.0180) (0.0181)
Anhui, Hubei, Hunan, and Jiangxi; and the western region includes
pft − 0.9946** − 1.5081*** − 1.5280***
(0.4511) (0.0449) (0.0464) Guangxi, Yunnan, Guizhou, Sichuan, Chongqing, Shaanxi, Gansu,
loar − 0.1822 0.4788*** 0.4872*** Ningxia, Inner Mongolia, Xinjiang, Qinghai, and Tibet. The results for
(0.3495) (0.0351) (0.0359) Models 1, 2, and 3 are presented in Table 6 for the eastern, central, and
contl 0.0014 0.0003 0.0004 western regions, respectively.
(0.0041) (0.0004) (0.0004)
gdp 5.2979*** 0.0057 0.1060**
The regression results in Model 1 show that the coefficient of Digi­
(0.9488) (0.0187) (0.0477) taldex is negative at a significance level of 1 %, while the coefficient of
popu − 3.2427*** 0.0708 − 0.0355** Digitaldex2 is positive at a significance level of 1 %, indicating a U-sha­
(0.1236) (0.567) (0.0138) ped relationship between DE development and FP in the eastern region.
Constant − 71.5213*** − 0.0185 − 3.2724***
According to the regression results for Models 2 and 3, however, the
(2.1401) (0.0120) (0.7518)
First-stage F-statistics 529.554 development of the DE in the central and western regions has no sig­
Observations 10,739 10,739 10,739 nificant impact on FP.
R-squared 0.66 0.44 0.43 Based on the regression results for firms in the eastern, central, and
Year fixed effect Yes Yes Yes western regions, we fitted the curves in detail in Fig. 4a, b, and c,
Industry fixed effect Yes Yes Yes
respectively. Fig. 4a depicts the eastern region. It shows a moderate U-
Note. Robust standard errors in parentheses. shaped curve, in which the turning point value for Digitaldex is 25.9. The
***
p < 0.01. shape and the turning point support Hypothesis 2a. Meanwhile, DE
**
p < 0.05. development and FP exhibit a steep U-shaped relationship for firms in
*
p < 0.1.
the central region. The turning point value for Digitaldex is 10.9, and the
maximum value is 9.3968, suggesting that firms in the central region
level. The regression results for Models 7 and 8 are presented in Table 5. remain on the left side of the steep U-shaped curve and have not
Finally, we winzorized all continuous variables at 5 % to alleviate the completed the transition. Although the regression result does not sup­
impact of extreme values on the regression results. The results for port Hypothesis 2b, the curve depicted in Fig. 4b partly supports our
Models 9 and 10 are presented in Table 5; these suggest that the full- argument. The regression results also fail to support Hypothesis 2c for
sample regression results remain robust, thus supporting Hypothesis 1. firms in the western region. Fig. 4c shows that for these firms, DE
development and FP do not exhibit an inverted U-shaped relationship.

8
Z. Sun et al. Technological Forecasting & Social Change 189 (2023) 122329

Table 5
Regression results for the robustness test.
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Model 7 Model 8 Model 9 Model 10
(LP) (LP) (LP–ACF) (LP–ACF) (Dfdex) (Dfdex) (City-fixed (City-fixed (5 % (5 %
effect) effect) winsorization) winsorization)

Digitaldex − 0.0012** − 0.0047*** − 0.0009* − 0.0048*** − 0.0012* − 0.0049** − 0.0012* − 0.0048**

(− 2.0884) (− 2.8313) (− 1.9262) (− 2.6431) (− 1.8434) (− 2.1754) (− 1.9052) (− 2.3074)
Digitaldex2 0.0001** 0.0001** 0.0001* 0.0001*
(2.1787) (2.1267) (1.7378) (1.7260)
Dfdex − 0.0005*** − 0.0061***
(− 2.9485) (− 3.6453)
2
Dfdex 0.00001***
(3.4328)
size 0.4301*** 0.4304*** 0.0614** 0.0617** 0.4383*** 0.4390*** 0.4379*** 0.4382*** 0.4289*** 0.4291***
(18.3116) (18.3360) (2.1080) (2.1234) (18.7943) (18.7823) (16.7726) (16.7805) (20.3188) (20.3221)
age 0.1125 0.1065 − 0.1744 − 0.1809 0.1240 0.1451 0.1168 0.1100 0.1991* 0.1948*
(0.9367) (0.8992) (− 1.1635) (− 1.2249) (1.0212) (1.1877) (0.8175) (0.7739) (1.7616) (1.7212)
pft − 1.1340*** − 1.1334*** − 1.1697*** − 1.1690*** − 1.1205*** − 1.1195*** − 1.1212*** − 1.1207*** − 1.0618*** − 1.0622***
(− 11.1929) (− 11.1737) (− 10.6540) (− 10.6490) (− 11.0155) (− 11.0153) (− 11.0924) (− 11.0787) (− 10.9104) (− 10.8937)
loar 0.0104 0.0112 − 0.0789 − 0.0781 0.0109 0.0132 0.0142 0.0150 0.0410 0.0414
(0.1981) (0.2119) (− 1.2994) (− 1.2919) (0.2061) (0.2501) (0.2186) (0.2313) (0.7808) (0.7856)
contl − 0.0004 − 0.0004 − 0.0005 − 0.0005 − 0.0003 − 0.0004 − 0.0004 − 0.0004 − 0.0008 − 0.0008
(− 0.2235) (− 0.2242) (− 0.2603) (− 0.2608) (− 0.1980) (− 0.2320) (− 0.2412) (− 0.2446) (− 0.4479) (− 0.4597)
gdp − 0.0254*** − 0.0230** − 0.0179* − 0.0153 − 0.0195** − 0.0211*** − 0.0287*** − 0.0266*** − 0.0197* − 0.0174*
(− 2.7165) (− 2.4444) (− 1.8227) (− 1.5750) (− 2.4278) (− 2.6269) (− 3.1650) (− 2.9273) (− 1.9444) (− 1.7432)
popu − 0.0323 − 0.0143 − 0.0250 − 0.0055 − 0.0505 − 0.0155 − 0.0771 − 0.0508 0.0059 0.0151
(− 0.5702) (− 0.2576) (− 0.4970) (− 0.1119) (− 0.9434) (− 0.3050) (− 0.7830) (− 0.5098) (0.0995) (0.2611)
Constant 3.3298*** 3.2067*** 0.4000 0.2661 4.2638*** 3.8818*** 3.6896*** 4.3023*** 3.5926*** 2.7259***
(4.3120) (4.1639) (0.3384) (0.2275) (5.2050) (5.0613) (3.2465) (3.5884) (4.5365) (3.7219)
Observations 10,739 10,739 10,739 10,739 10,739 10,739 10,739 10,739 10,739 10,739
R-squared 0.36 0.36 0.10 0.11 0.37 0.37 0.37 0.37 0.36 0.36
Year fixed Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
effect
Industry Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes
fixed effect
City fixed YES YES
effect

Note. Robust t-statistics in parentheses.

***
p < 0.01.
**
p < 0.05.
*
p < 0.1.

Table 6
Regression results for the sub-sample by firm locality and firm size.
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6
(Eastern) (Central) (Western) (Large) (Medium) (Small)

Digitaldex − 0.0055*** − 0.0157 0.0071 − 0.0077** − 0.0009 0.0015

(− 3.3803) (− 1.2900) (0.4891) (− 2.5661) (− 0.2028) (0.3816)
2
Digitaldex 0.0001*** 0.0007 0.0002 0.0001* 0.00003 − 0.0001
(2.7355) (0.6956) (0.2623) (1.8581) (0.5167) (− 0.9331)
size 0.4303*** 0.4388*** 0.4712*** 0.3826*** 0.4956*** 0.2996***
(13.3094) (13.0114) (9.0596) (11.8088) (10.6582) (3.8653)
age 0.1552 0.0185 0.0962 0.3187* 0.5133 − 0.6299
(1.4576) (0.0332) (0.2079) (1.7088) (1.3822) (− 1.5768)
pft − 1.1647*** − 1.0933*** − 1.0239*** − 1.1314*** − 1.1392*** − 1.2546***
(− 9.8188) (− 7.7205) (− 3.1794) (− 7.6381) (− 8.5627) (− 6.2890)
loar 0.0079 0.1295 − 0.0534 − 0.2331** − 0.0310 0.2623*
(0.1222) (1.0902) (− 0.4071) (− 2.3090) (− 0.2377) (1.8232)
contl 0.0008 − 0.0037** − 0.0001 − 0.0026** 0.0016 − 0.0091**
(0.3206) (− 2.2224) (− 0.0446) (− 2.3021) (0.4337) (− 2.1723)
gdp − 0.0206* − 0.0177 − 0.0396 − 0.0244 − 0.0354** 0.0060
(− 1.8620) (− 1.1137) (− 1.1877) (− 1.5815) (− 2.3479) (0.2376)
popu 0.0212 0.1820** − 0.6654*** − 0.0576 − 0.2393 0.0473
(0.3217) (2.5721) (− 2.7166) (− 0.9321) (− 0.8962) (0.5297)
Constant 3.1242*** 2.8378 6.9384* 4.1037*** 1.2465 9.0039***
(3.3121) (1.0968) (1.9875) (3.0552) (0.3805) (3.9763)
Observations 7645 1699 1395 5717 3160 1862
R-squared 0.35 0.50 0.32 0.28 0.24 0.26
Year fixed effect Yes Yes Yes Yes Yes Yes
Industry fixed effect Yes Yes Yes Yes Yes Yes

Note. Robust t-statistics in parentheses.

***
p < 0.01.
**
p < 0.05.
*
p < 0.1.

9
Z. Sun et al. Technological Forecasting & Social Change 189 (2023) 122329

Fig. 4. a. Relationship between DE and FP in the eastern region of China.

b. Relationship between DE and FP in the central region of China.
c. Relationship between DE and FP in the western region of China.

However, an upward trend is evident, indicating that firms in the the western region have not yet reached the peak of FP, suggesting that
western region can boost their productivity at the early stage of the DE they will continue to benefit from DE development.
via their latecomer advantage. Notably, Fig. 4c indicates that firms in

10
Z. Sun et al. Technological Forecasting & Social Change 189 (2023) 122329

Fig. 4. (continued).

4.3.2. Sub-sample regression analysis by firm size to conceptualize and investigate the productivity consequences of DE
We divided the firms into three sub-samples based on size: large, development in Chinese firms. We analyze 10,739 firm-level observa­
medium, and small; the regression results for Models 4, 5, and 6 in tions from 2016 to 2019 and find that, for Chinese firms in general, DE
Table 6 pertain to the regression results for large, medium, and small development and FP have a nonlinear U-shaped relationship. Initially,
firms, respectively. DE development is associated with a decrease in FP. However, in later
The results in Model 4 reveal that the coefficient of Digitaldex is stages, DE development is associated with an increase in FP. In other
negative at a significance level of 5 %, while the coefficient of Digitaldex2 words, Chinese firms are likely to experience a decline in productivity if
is positive at a significance level of 10 %. This indicates a U-shaped they opt for a digitized strategy early in the process of DE development.
relationship between the development of the DE and FP for large firms. However, they will reap the benefits later.
The results in Models 5 and 6 show that the development of the DE had In the case of ICT investment, our results confirm the hypothesis of a
an insignificant impact on the productivity of medium-sized and small U-shaped relationship; linear (either upward or downward) curves are
firms, respectively. merely sub-stages of the main U-curve hypothesis. It is especially
Based on the regression results for large, medium, and small firms, fascinating that our findings, which are based on data from Chinese
we fitted the curves in detail in Fig. 5a, b, and c, respectively. Fig. 5a firms, may also explain the contradictory results of earlier studies on ICT
depicts a U-shaped curve in which the turning point value for Digitaldex investment and the productivity paradox. Specifically, we propose that
is 31.3. This indicates that large firms require a lengthy period of researchers take into account the DE development stages and firms’
preparation to complete the digital transition. Therefore, the results strategic choices. Indeed, these factors account for the curvilinear nature
support Hypothesis 3a. Fig. 5b depicts a U-shaped curve with a turning of the relationship between the DE and FP relationship. Thus, our
point value of 9.4, indicating that medium-sized firms take a relatively findings contribute to a more nuanced understanding of the productivity
shorter period to prepare to alter their digitization strategies. Although paradox, indicating that arguments derived from earlier investigations
the regression result is insignificant, Fig. 5b thus partly supports Hy­ of firms in developed countries may also be relevant to Chinese firms.
pothesis 3b. Fig. 5c shows an inverted U-shaped relationship with a Second, our findings have substantial implications for research into
turning point value of 9.2. This suggests that small firms enjoy a very how firm idiosyncrasies shape the relationship between the DE and FP in
short-term increase in productivity but encounter a subsequent decline. China. We build a conceptual model by illustrating how DE development
Although this regression result is again insignificant, Fig. 5c partly and two distinct firm-specific factors (firm size and locality) impact the
supports Hypothesis 3c. productivity outcomes of Chinese firms. While previous studies have
revealed the DE’s broad influence on Chinese firms (e.g., Pan et al.,
5. Discussion and conclusions 2022; Sun et al., 2017), there has been little research on which types of
firms are more affected.
5.1. Theoretical implications Our findings also emphasize the importance of the firm-specific
context in the complex and constantly shifting relationship between
Our analysis contributes to studies on both the general DE–FP rela­ the DE and FP in heterogeneous firms. While firms do not benefit equally
tionship and the idiosyncratic effects of firm size and locality on the from DE development, our analysis advances research in this area by
DE–FP relationship for Chinese firms. First, our study is among the first uncovering the firm-specific contexts underpinning DE development in

11
Z. Sun et al. Technological Forecasting & Social Change 189 (2023) 122329

Fig. 5. a. U-shaped relationship between DE and FP of large firms in China.

b. U-shaped relationship between DE and FP of medium firms in China.
c. The U-shaped relationship between DE and FP of small firms in China.

China. Thus, we provide more thorough knowledge on the impact of the compatibility-related difficulties and technological complexities. Our
firm-specific context on firms’ choice of digitized or non-digitized stra­ findings suggest that the benefits of DE development are not consistent
tegies to improve productivity. We contribute to the literature on the across firms. Our study thus answers the call for a more nuanced
DE–FP relationship and the productivity paradox by demonstrating that explanation of the productivity paradox, which states that the existence
the relationship is shaped by different digital vacuums arising from of the productivity paradox depends on firm-specific contexts.

12
Z. Sun et al. Technological Forecasting & Social Change 189 (2023) 122329

Fig. 5. (continued).

Third, we propose an alternative interpretation of the strategic recognize how diverse firm-specific contexts influence the productivity
choice to digitize at the various stages of the DE development process. outcomes of their digitization initiatives. Our findings on these out­
We explain how the cost structures of non-digitized and digitized stra­ comes for heterogeneous firms in China should assist in better under­
tegies differ at the early and advanced stages of DE development. The standing the critical role that a more flexible organizational structure
study finds that firms often opt for a strategy based on its relative cost and latecomer advantage play in overcoming disadvantages and facili­
advantages. When DE development is in its infancy, firms will choose a tating successful digitization efforts. Examples include small and me­
non-digitized strategy because of the high fixed costs and rising mar­ dium firms and those in the central and western regions, that, due to
ginal costs required to internalize the advantages of a DE. In contrast, their flexibility and latecomer advantage, are more likely to benefit from
once DE development has advanced, firms are likely to adopt a digitized digitization efforts than are large firms and those from the eastern re­
strategy due to the low fixed costs and diminishing marginal costs gion; the latter should prioritize organizational management and capi­
associated with its implementation. Firms suffer from a digital vacuum talize fully on their first-mover advantage.
during the transition from a non-digitized to a digitized strategy when
compatibility concerns arise and productivity plummets. Therefore, we 5.2. Policy implications
propose a novel theoretical framework that adds value to existing
research in respect of strategic digitization; the literature currently fo­ Our findings also have substantial policy implications. First, under­
cuses on the technical and innovation aspects of the relationship be­ standing the productivity implications of digitization strategies can help
tween DE and FP. Chinese authorities to promote the DE more effectively. The findings
demonstrate the explicit heterogeneity in productivity among firms of
5.1.1. Managerial implications varying sizes. Specifically, small and medium firms reap the benefits of
Our results have significant managerial implications for Chinese DE development more quickly than do large firms. Thus, it is critical for
firms and those in emerging economies. First, our research offers Chinese authorities to explore multiple policy alternatives for sustaining
essential insights to address questions confronting Chinese firms: Is a DE development and thereby lower the cost of internalizing the
digital strategy indeed the best choice? If so, how can firms make the advantage of DE development for Chinese firms.
transition from a non-digitized to a digitized strategy? Our findings Second, our findings indicate that there is significant regional het­
suggest that a non-digitized strategy is better for firms at the early stage erogeneity in the process of DE development in China. We argue that
of DE development, while a digitized strategy may be better for firms at preferential policies are important to accelerate the construction of
the advanced stage of DE. However, these firms are more likely to digital infrastructure in the under-developed central and western re­
endure a digital vacuum during the transition. Chinese firms must be gions and to support the coordinated and high-quality development of
patient and confident and remain aware of this potential vacuum to reap regional DE.
the full advantages of a digitized strategy. Although there are potential
advantages to adopting such a strategy, firms must address compatibility
5.3. Limitations and future research
concerns as efficiently as possible in this period. Firms in emerging
economies should thus attempt to learn from successfully digitized
Future research on the DE, particularly in the context of emerging
firms. In doing so, they may be able to shorten the duration of the digital
economies, should work to address the limitations of our study. First, our
vacuum.
focus on a sample of Chinese firms may limit the generalizability of our
Second, our findings reveal that firms in emerging economies should
findings to other emerging economies. Because technological

13
Z. Sun et al. Technological Forecasting & Social Change 189 (2023) 122329

environments and marketplaces differ significantly across countries, Cai, Y., Zhang, J., 2015. The substitution and pervasiveness effects of ICT on China’s
economic growth. Econ. Res. J. 12, 100–114.
Chinese firms may not adequately reflect their counterparts in other
China Academy of Information and Communications Technology, 2021. White paper on
emerging economies. Our empirical results fail to consider this hetero­ China’s Digital Economy Development, Beijing. http://www.caict.ac.cn/english/res
geneity and should thus be interpreted cautiously. Future studies should earch/whitepapers/202104/t20210429_375940.html.
thus include data on firms from other emerging and developed econo­ Chen, S.Y., Xie, Z., 2015. Is China’s e-governance sustainable? Testing Solow IT
productivity paradox in China’s context. Technol. Forecast. Soc. Chang. 96, 51–61.
mies to enhance the validity of the findings. Cheng, Y., Tao, F., Xu, L., Zhao, D.M., 2018. Advanced manufacturing systems:
Second, data limitations led us to focus solely on two types of firm- supply–demand matching of manufacturing resource based on complex networks
specific factors: size and locality. However, additional factors may and Internet of Things. Enterp. Inf. Syst. 12 (7), 780–797.
Cui, T.R., Ye, H., Teo, H.H., Li, J.Z., 2015. Information technology and open innovation:
explain firm contexts and hence influence the DE–FP relationship a strategic alignment perspective. Inf. Manag. 52 (3), 348–358.
(Wamba et al., 2017). Consequently, future research should consider a Dewan, S., Kraemer, K.L., 2000. Information technology and productivity: evidence from
model that includes the impact on the DE–FP relationship of various country-level data. Manag. Sci. 46, 548–562.
Du, C.Z., Zhang, Y., 2021. The mechanism of the impact of digital economic development
firm-specific factors, such as digital capabilities, financing constraints, on the growth of enterprise productivity. Secur. Mark. Her. 2, 41–51 (in Chinese).
and ownership diversity. Finally, due to data availability issues, we only Dyer, J.H., Singh, H., Hesterly, W.S., 2018. The relational view revisited: a dynamic
cover data from 2016 to 2019, which may limit the robustness of our perspective on value creation and value capture. Strateg. Manag. J. 39 (12),
3140–3162.
findings to some extent. Future research can overcome this problem with Frey, C.B., Osborne, M.A., 2017. The future of employment: how susceptible are jobs to
a more long-term study. computerisation? Technol. Forecast. Soc. Chang. 114 (1), 254–280.
Frynas, J.G., Mol, M.J., Mellahi, K., 2018. Management innovation made in China:
Haier’s Rendanheyi. Calif. Manag. Rev. 61 (1), 71–93.
CRediT authorship contribution statement Giannetti, M., Liao, G., Yu, X., 2015. The brain gain of corporate boards: evidence from
China. J. Financ. 70 (4), 1629–1682.
Zhe Sun: Conceptualization, Methodology, Data Curation, Formal Graetz, G., Michaels, G., 2018. Robots at work. Rev. Econ. Stat. 100 (5), 753–768.
Guariglia, A., Liu, X.X., Song, L., 2011. Internal finance and growth: microeconometric
analysis, Writing – Original Draft
evidence on Chinese firms. J. Dev. Econ. 96 (1), 79–94.
Liang Zhao: Conceptualization, Supervision, Data Curation, Writing He, J.N., Fang, X., Liu, H.Y., Li, X.D., 2019. Mobile app recommendation: An
– Original Draft, involvement-enhanced approach. MIS Q. 43 (3), 827–849.
Heshmati, A., Kumbhakar, S.C., 2011. Technical change and total factor productivity
Review & Editing
growth: the case of Chinese provinces. Technol. Forecast. Soc. Chang. 78 (4),
Puneet Kaur: Conceptualization, Research Method & Validation, 575–590.
Review & Editing Howell, A., 2015. “Indigenous” innovation with heterogeneous risk and new firm
Nazrul Islam: Supervision, Conceptualization, Research Method & survival in a transitioning Chinese economy. Res. Policy 44 (10), 1866–1876.
Hsieh, C.T., Klenow, P.J., 2009. Misallocation and manufacturing TFP in China and
Validation, Review & Editing India. Q. J. Econ. 124 (4), 1403–1448.
Amandeep Dhir: Supervision, Conceptualization, Research Method IMF, 2018. Measuring the Digital Economy. IMF Policy Papers, Washington, D.C.
& Validation, Review & Editing Johnson, J.S., Friend, S.B., Lee, H.S., 2017. Big data facilitation, utilization, and
monetization: exploring the 3Vs in a new product development process. J. Prod.
Innov. Manag. 34 (5), 640–658.
Data availability Kim, W.G., Lim, H., Brymer, R.A., 2015. The effectiveness of managing social media on
hotel performance. Int. J. Hosp. Manag. 44, 165–171.
Krishnan, K., Nandy, D.K., Puri, M., 2015. Does financing spur small business
The authors are unable or have chosen not to specify which data has productivity? Evidence from a natural experiment. Rev. Financ. Stud. 28 (6),
been used. 1768–1809.
Lee, S., Xiang, J.Y., Kim, J.K., 2011. Information technology and productivity: empirical
evidence from the Chinese electronics industry. Inf. Manag. 48 (2–3), 79–87.
Acknowledgments Levinsohn, J., Petrin, A., 2003. Estimating production functions using inputs to control
for unobservables. Rev. Econ. Stud. 70 (2), 317–341.
This research was funded by the National Social Science Fund of Li, Q., Wu, Y.R., 2020. Intangible capital, ICT and sector growth in China. Telecommun.
Policy 44 (1), 101854.
China (Grant number: 2020CJY032), and Philosophy and Social Science Lin, W.T., Shao, B.M., 2006. The business value of information technology and inputs
Research Innovation Team Project of “Chinese Path to Modernization substitution: the productivity paradox revisited. Decis. Support. Syst. 42 (2),
Road and New Form of Human Civilization” of Jilin University (Grant 493–507.
Liu, T.K., Chen, J.R., Huang, C.J., Yang, C.H., 2014. Revisiting the productivity paradox:
number: 2022CXTD26).
a semiparametric smooth coefficient approach based on evidence from Taiwan.
Technol. Forecast. Soc. Chang. 81, 300–308.
References Matt, C., Hess, T., Benlian, A., 2015. Digital transformation strategies. Bus. Inf. Syst. Eng.
57 (5), 339–343.
Mikalef, P., Pateli, A., 2017. Information technology-enabled dynamic capabilities and
Acemoglu, D., Restrepo, P., 2018. The race between man and machine: implications of
their indirect effect on competitive performance: findings from PLS-SEM and fsQCA.
technology for growth, factor shares, and employment. Am. Econ. Rev. 108 (6),
J. Bus. Res. 70 (1), 1–16.
1488–1542.
Pan, W., Xie, T., Wang, Z., Ma, L., 2022. Digital economy: an innovation driver for total
Ackerberg, D.A., Caves, K., Frazer, G., 2015. Identification properties of recent
factor productivity. J. Bus. Res. 139, 303–311.
production function estimators. Econometrica 83 (6), 2411–2451.
Pesch, R., Endres, H., 2019. The marginal utility effect of formalization for digital
Agarwal, R., Gao, G.D., DesRoches, C., Jha, A.K., 2010. The digital transformation of
product innovation. Acad. Manag. Proc. https://doi.org/10.5465/
healthcare: current status and the road ahead. Inf. Syst. Res. 21 (4), 796–809.
AMBPP.2019.14834abstract.
Arvanitis, S., Loukis, E.N., 2009. Information and communication technologies, human
Porter, M.E., Heppelmann, J.E., 2016. How smart, connected products are transforming
capital, workplace organization and labour productivity: a comparative study based
companies. Harv. Bus. Rev. 94 (1–2), 24-24.
on firm-level data for Greece and Switzerland. Inf. Econ. Policy 21 (1), 43–61.
Solow, R., 1987. We’d better watch out. In: New York Times Book Review, p. 36.
Bauer, W., Hämmerle, M., Schlund, S., Vocke, C., 2015. Transforming to a hyper-
Stanley, T.D., Doucouliagos, H., Steel, P., 2018. Does ICT generate economic growth? A
connected society and economy–towards an “Industry 4.0”. Procedia Manuf. 3,
meta-regression analysis. J. Econ. Surv. 32 (3), 705–726.
417–424.
Stieglitz, N., Knudsen, T., Becker, M.C., 2016. Adaptation and inertia in dynamic
Bell, D.R., Choi, J., Lodish, L., 2012. What matters most in internet retailing. MIT Sloan
environment. Strateg. Manag. J. 37 (9), 1854–1864.
Manag. Rev. 54 (1), 27–33.
Sun, L., Zheng, H., Ren, R., 2012. The contribution of ICT to China’s economic growth:
Bloom, N., Sadun, R., Van Reenen, J., 2016. Management as a technology? NBER
evidence from industrial panel data analysis. J. World Econ. 2, 3–25.
working paper no. 22327. https://www.nber.org/papers/w22327.
Sun, Z.J., Hou, J., Li, J.Z., 2017. The multifaceted role of information and
Bonfadelli, H., 2002. The internet and knowledge gaps—a theoretical and empirical
communication technology in innovation: evidence from Chinese manufacturing
investigation. Eur. J. Commun. 17 (1), 65–84.
firms. Asian J. Technol. Innov. 25 (1), 168–183.
Borland, J., Coelli, M., 2017. Are robots taking our jobs? Aust. Econ. Rev. 50 (4),
Vaccaro, A., Brusoni, Veloso, F.M.S., 2011. Virtual design, problem framing, and
377–397.
innovation: an empirical study in the automotive industry. J. Manag. Stud. 48 (1),
Brown, J.R., Fazzari, S.M., Petersen, B.C., 2009. Financing innovation and growth: Cash
99–122.
flow, external equity and the 1990s R&D boom. J. Financ. 64 (1), 151–185.
Varian, H.R., 2010. Computer mediated transactions. Am. Econ. Rev. 100 (2), 1–10.
Brynjolfsson, E., Hitt, L.M., 2000. Beyond computation: information technology,
Wagner, E.L., Newell, S., Kay, W., 2012. Enterprise systems projects: the role of liminal
organizational transformation and business performance. J. Econ. Perspect. 14 (4),
space in enterprise systems implementation. J. Inf. Technol. 27 (4), 259–269.
23–48.

14
Z. Sun et al. Technological Forecasting & Social Change 189 (2023) 122329

Wamba, S.F., Gunasekaran, A., Akter, S., Ren, S.J.F., Dubey, R., Childe, S.J., 2017. Big books, and industry reports. Moreover, he also authors popular articles for FinTech web
data analytics and firm performance: effects of dynamic capabilities. J. Bus. Res. 70, media such as CrowdfundInsider and Techfoliance.
356–365.
Ward, J., Hemingway, C., Daniel, E., 2005. A framework for addressing the
Puneet Kaur is an Associate Professor at Department of Psychosocial Science, University
organisational issues of enterprise systems implementation. J. Strateg. Inf. Syst. 14
of Bergen, Norway. Her research appears in Journal of Retailing and Consumer Services,
(2), 97–119.
International Journal of Information Management, Computers in Human Behaviour, Interna­
Watson, H.J., 2017. Preparing for the cognitive generation of decision support. MIS Q.
tional Journal of Hospitality Management, and Information Technology & People among
Exec. 16 (3), 153–169.
others.
Wooldridge, J.M., 2009. On estimating firm-level production functions using proxy
variables to control for unobservables. Econ. Lett. 104 (3), 112–114.
Xie, K., Wu, Y., Xiao, J.H., Hu, Q., 2016. Value co-creation between firms and customers: Nazrul Islam is Professor of Business & Director of Research Degrees, and Associate Di­
the role of big data-based cooperative assets. Inf. Manag. 53 (8), 1034–1048. rector for Centre of FinTech at Royal Docks School of Business and Law, University of East
Zhang, Y.F., Ma, S.Y., Yang, H.D., Lv, J.X., Liu, Y., 2018. A big data driven analytical London, UK. He holds a PhD in innovation management. His research interest focuses on
framework for energy-intensive manufacturing industries. J. Clean. Prod. 197, interdisciplinary fields: the management of technology; technological transformation; the
57–72. emergence and growth of disruptive and digital technology-based innovation; and SMEs
business sustainability. His research was published in the leading international journals,
and he has complemented his peer reviewed journal efforts with three books. Prof Islam’s
Zhe Sun is an Associate Professor at the Economics School, Jilin University. She holds a
research received awards including the ‘Brad Hosler Award for Outstanding Paper’ from
Ph.D. in entrepreneurship and innovation from the University of Amsterdam and an MSc
USA; and the ‘Pratt & Whitney Canada Best Paper Award’ from Canada. Prof Islam serves
from Jilin University. Her research interests revolve around international entrepreneur­
on the board of directors for Business and Applied Sciences Academy of North America. He is
ship and specific-context innovation, with focus on the innovation strategy of Chinese
an Associate Editor for Technological Forecasting & Social Change, Department Editor for
enterprises within and outside China. She has published in journals such as Research in
IEEE Transactions on Engineering Management, and Editor-in-Chief of International Journal of
International Business and Finance, Technological Analysis and Strategic Management,
Technology Intelligence and Planning.
and Chinese Management Studies. Currently, she is serving as the Director Assistant for the
Centre for China Public Sector Economy Research.
Amandeep Dhir is a Professor of Research Methods at University of Agder, Norway. He is
also a visiting professor at Norwegian School of Hotel Management, University of Sta­
Liang Zhao is an Assistant Professor in Marketing Strategy at the School of Business,
vanger, Norway. His research appears in Technological Forecasting and Social Change,
University of Leicester, UK. He holds a PhD in entrepreneurship and innovation. His
Journal of Retailing and Consumer Services, International Journal of Information Management,
research interests include marketing strategy, digital marketing, entrepreneurial
Computers in Human Behaviour, Computers in Industry, International Journal of Hospitality
financing, digital innovation and innovation strategy with special expertise in crowd­
Management, Journal of Cleaner Production, Food quality and preferences, Appetite, Infor­
funding research. He has been granted external funding such as the Norwegian Research
mation Technology & People, Australasian Marketing Journal, and Enterprise Information
Council (NRC) and the Chinese Social Science Foundation (CSSF). His research outputs
Systems among others.
include journal articles (including JBR, BH, BJM, CMS, RBF, etc.), book chapters, edited

15