Digital Trade Networks: Multinational Enterprises and Digital Regulation

Abstract

The influence of multinational enterprises on digital trade is growing significantly, resulting in profound transformations within global production networks. In this paper, we establish a comprehensive global digital trade network using the OECD-AMNE database and conduct empirical analysis to examine the impact of digital regulations and network structure. Our findings reveal domestic and international digital regulations' collaborative influence on countries' network participation. Moreover, we observe that stringent domestic restrictions impede exports. Additionally, our results highlight the positive role of reciprocal network structures in fostering trade relationships. By employing a value-added decomposition approach, we uncover that digital service trade restrictions hinder the export of value-added trade when multinational enterprises act as upstream value-added providers. However, downstream producers experience comparatively lesser impact. Furthermore, we identify that disparities in digital service regulation between home and host countries significantly influence the digital service output of multinational enterprises. Additionally, differences in global rules have implications for investment relationships. To sum up, we have put forward policy suggestions that are specifically designed for the newly adopted digital commerce systems. These suggestions highlight the significance of maintaining harmony between digital accessibility and safeguarding the country's economic interests.
JEL Classification: F14; F42;

Full text

Digital Trade Networks: Multinational Enterprises
and Digital Regulation
Tianding Zhang (  ordin@126.com )
Wuhan University
Tong Gong
Wuhan University
Research Article
Keywords: multinational enterprises, digital services, digital regulations, trade networks, social network
analysis
Posted Date: May 25th, 2023
DOI: https://doi.org/10.21203/rs.3.rs-2971841/v1
License:   This work is licensed under a Creative Commons Attribution 4.0 International License. 
Read Full License
Additional Declarations: No competing interests reported.

1 / 38
Digital Trade Networks: Multinational Enterprises and Digital
Regulation
Tianding Zhang1, Tong Gong1
1. Economics and Management School, Wuhan University, Wuhan, China;
Corresponding author:
Tianding Zhang, Email: ordin@126.com, Economics and Management School, No.16,
Luojiashan, Wuhan University, Wuhan, P.R. China 430072
Funding
This work was supported by National Natural Science Foundation of China
[71673205].

2 / 38
About the authors:
Tianding Zhang is a professor in the Department of World Economy of Economics and
Management School at Wuhan University, where he has been since 2008. From 2018 he
served as Department Chair. From April to August 2018, he was a visiting professor at
Oldenburg University. During 2014–2015 he was a visiting scholar at Columbia University.
During 2007–2008 he was an exchange student by the Sino-France Doctoral College Program
at the University of Auvergne (Université d'Auvergne). He received a B.E. from Wuhan
University of Technology in 2000, and an M.E. from Wuhan University in 2005. He received
his Ph.D. in Economics from Wuhan University in 2008. His research interests span open
macroeconomics and finance, commodities market, and financial economics. Much of his
research work has been on improving the understanding, forecasting and performance on
China’s macroeconomy, industrial policy and firm. He has published more than 80 referred
academic journal papers in China, and has been conducting research projects funded by the
National Natural Science Foundation of China, the Ministry of Education of China, and other
institutions. He has published 2 academic monographs and 1 textbook.
Tong Gong is a Ph.D. student majoring in international economics studied in Economics and
Management School at Wuhan University.
Declaration of interests
☒ The authors declare that they have no known competing financial interests or personal
relationships that could have appeared to influence the work reported in this paper.
Data availability statement
The data supporting this study’s findings are available from the corresponding author upon
request.

3 / 38
Digital Trade Networks: Multinational Enterprises and Digital Regulations
Abstract: The influence of multinational enterprises on digital trade is growing significantly,
resulting in profound transformations within global production networks. In this paper, we
establish a comprehensive global digital trade network using the OECD-AMNE database and
conduct empirical analysis to examine the impact of digital regulations and network structure.
Our findings reveal domestic and international digital regulations' collaborative influence on
countries' network participation. Moreover, we observe that stringent domestic restrictions
impede exports. Additionally, our results highlight the positive role of reciprocal network
structures in fostering trade relationships. By employing a value-added decomposition
approach, we uncover that digital service trade restrictions hinder the export of value-added
trade when multinational enterprises act as upstream value-added providers. However,
downstream producers experience comparatively lesser impact. Furthermore, we identify that
disparities in digital service regulation between home and host countries significantly
influence the digital service output of multinational enterprises. Additionally, differences in
global rules have implications for investment relationships. To sum up, we have put forward
policy suggestions that are specifically designed for the newly adopted digital commerce
systems. These suggestions highlight the significance of maintaining harmony between digital
accessibility and safeguarding the country's economic interests.
Keywords: multinational enterprises; digital services; digital regulations; trade networks; social
network analysis
JEL Classification: F14; F42;

4 / 38
1. Introduction
Multinational enterprises (MNEs) are critical players in the global industrial landscape, serving
as network hubs in the global value chain (Cadestin et al. 2018). In today's digital economy,
large and emerging digital companies around the world are actively engaging in cross-border
business, establishing digital multinational enterprises worldwide to meet the growing
demand for digital products. This has given rise to a digital trade network dominated by MNEs.
Digital trade refers to digitalized service trade. MNEs and their affiliates are important
conduits for overseas digital service trade, enabling them to gain more business opportunities
in global production and operations. Moreover, digital trade plays a significant role in the
international exchange of goods, services, capital, and technological knowledge. According
to OECD data, the proportion of intermediate and final goods exports by MNEs' overseas
subsidiaries in the service industry has continued to rise from 2000 to 2014. The development
of the digital trade network facilitates domestic and international economic circulation and
achieves better resonance of the digital economy, service industry development, and the real
economy sector.
However, digital trade has also changed the business model of MNEs' international operations
and the economic impact of foreign multinational subsidiary companies on host countries.
For the economic development of host countries, the impact of digital MNEs is not limited to
direct investment in industry or the creation of employment opportunities. Rather, it indirectly
affects production efficiency through investment, which can support the host country's digital
development (UNCTAD 2017). In addition, cross-border production and division of labor in
the digital economy sector are crucial for the upgrading and transforming traditional
industrial sectors in the host country and national competitiveness. Ignoring digital
transformation may cause traditional industries in the host country to lose international
competitiveness, leading to the transfer of industries to other countries (!!! INVALID
CITATION !!! ). Therefore, MNEs worldwide are competing to invest in the digital economy,
recognizing its importance as a critical link connecting the global digital value chain.
The rapidly developing digital economy has brought new opportunities for MNEs. However,
digital barriers and restrictions limit the free flow of production factors, thereby affecting the
configuration of global production networks. However, the World Trade Organization's (WTO)
shaping of the digital trade rule framework needs to catch up in the context of the rapid
development of the digital economy and new-generation information technology. Therefore,
bilateral or multilateral free trade agreements (FTAs) have become the main driving force for
promoting global digital rule convergence. Since 2000, more newly added regional FTAs have
included digital economy provisions, reflecting the degree of rule convergence in the digital
economy (Smeets 2021).
For international competition in the field of digital trade rules, the "American template" and
the "European template" have each become integrated, and the "alliance" of rules has led to
a fragmented situation in global digital trade governance (Li and Zhang 2022). The "American
template" has introduced a series of second-generation digital trade rules, such as "cross-
border data free flow," which poses significant challenges to global digital trade governance

5 / 38
(Peng and Ouyang 2020). The fragmentation of global digital governance is becoming
increasingly severe, and this has become a problem that countries urgently need to address.
On the one hand, as pioneers of the digital economy, major developed economies in the
West have launched strategic games and resource struggles around the digital network.
"Long-arm jurisdiction" has intensified conflicts over data jurisdiction, and digital economy
barriers have made data flow between them difficult, resulting in market fragmentation. On
the other hand, the "digital divide" between developed and developing economies is also
deepening, further hindering digital trade's deep development (OECD 2021).
Due to the concentration of export targets for digital trade by MNEs, the layout of digital
trade networks dominated by MNEs exhibits prominent heterogeneous development
characteristics (UNCTAD 2019). With the deepening development of the digital economy, the
cross-border flow of factors penetrates various stages of the production process. Therefore,
the correlation between upstream and downstream stages will significantly impact the
production and trade activities of MNEs. At the same time, the endogenous network structure
also plays a vital role in forming network relationships. This has led to more vital
interdependence among economies in the digital economy era, and countries are showing
an intensified trend of competition over data sovereignty and trade interests. In this realistic
background, MNEs' digital trade can be viewed as a game of trade interests under the global
value chain division of labor. Therefore, actively participating in the international governance
of digital trade can effectively expand international cooperation in the digital economy,
thereby better integrating and reconstructing digital value chains. To this end, this paper
focuses on the macroscopic perspective of the overall network dimension. It constructs a
digital trade network dominated by MNEs, examining the factors influencing the formation of
such a network from the viewpoint of digital regulation and endogenous network structure.
This paper presents several marginal contributions compared to existing research. Firstly, it
focuses on multinational corporations (MNCs) as the main body and hub that connect the
digital value chain. Their digital trade networks reflect various interest games in the era of
digital globalization. Unlike previous studies that mainly explore digital trade between
different countries, this paper provides a detailed analysis of the digital trade networks
controlled by MNEs. By narrowing the research scope to the level of MNEs based on company
ownership, and examining the relationship between digital regulatory differences and MNC-
dominated digital trade networks from the perspective of value-added trade, this paper
provides valuable insights. Secondly, the networked characteristics of the digital economy
have prompted researchers to adopt a holistic perspective in grasping the value-added
network of MNEs. Therefore, this paper follows the pioneering approach of Mele (2017) and
Hsieh et al. (2022) to construct a network model, focusing on structural variables to
supplement the shortcomings of other methods in obtaining network structural
characteristics. To distinguish the different roles played by MNEs in the digital value chain,
this paper also decomposes the value-added trade of MNEs in the digital industry. This
approach can identify the production relations between MNEs and domestic enterprises and
explore MNC-dominated digital trade networks based on different value chain activities.
Thirdly, while the global rules of digital trade have developed significantly in recent years, the

6 / 38
role of domestic digital regulations cannot be ignored. Therefore, this paper constructs
domestic and external regulatory integration variables. From the dual perspectives of
domestic policy regulation and international rule integration, this paper enriches the research
level of digital regulation. This research provides a new solution for participating in global
digital governance and finding a balance between digital openness and national economic
security.
The paper is structured as follows: Part II is a literature review; Part III discusses MNC-
dominated digital trade networks and their topological characteristics; Part IV focuses on the
model, variables, and measurements; Part V presents the empirical research results, Part VI
outlines further research, and Part VII concludes with policy recommendations.
2. Literature Review
The global economy has transitioned into the digital economy era, with digital trade emerging
as a critical component of international trade. Digital service trade, in particular, has gained
prominence, as it refers to services provided through information and communication
networks (UNCTAD 2015). In the development of information and communication technology,
digital transformation has become a priority, especially in the service industry. As service
globalization grows, economies of scale and scope in the digital service industry have become
increasingly apparent (Capello et al. 2022). MNEs play a crucial role in the digital service trade,
with information and communication technology companies accounting for most of the top
100 MNEs worldwide (Bolwijn et al. 2018). The digital value chain is based on the production,
operation, and sales of MNEs' overseas subsidiaries. According to OECD research, the output
of foreign subsidiaries of MNEs has continued to rise from 2000 to 2014, with MNEs
accounting for 33% of global output in 2014 and the output of foreign subsidiaries accounting
for 12% of global output. In the digital economy era, cross-border service trade shows a trend
of fragmentation and personalized development (Cadestin et al. 2021). The interconnection
between MNEs is impacted by digital technology, resulting in more flexible production
network configurations (Alcácer et al. 2016; Strange et al. 2022; Zhan 2021). As a result, the
market structure and operating model have undergone significant changes, posing
challenges to the traditional economic order.
Digital trade policies have a crucial impact on the production network configuration of MNEs
worldwide. However, the WTO framework needs more unified rules and regulations for this
new form of trade, and the imbalance in the development of global digital trade rules is
becoming increasingly apparent. Therefore, MNEs must adhere to domestic and global digital
regulations. To measure the regulatory measures of various countries on digital service trade,
Ferencz (2019) constructed a digital service trade restriction index from five policy areas, such
as electronic transactions and infrastructure construction. Meanwhile, van der Marel and
Ferracane (2021) found that data flow restrictions significantly negatively impact trade,
particularly in data-intensive service industries, while Drynochkin et al. (2023) highlights digital
trade restrictions mainly affect the export of digital services through trade costs. Fortunately,
improving network facilities can effectively reduce the negative impact of digital service trade
restrictions, primarily when the network infrastructure is well constructed and good network

7 / 38
soft power helps overcome digital service barriers (Suh and Roh 2023).
The increasing complexity of global digital trade rules based on digital FTAs poses a significant
challenge to conducting a comprehensive analysis of their economic impact. To address this,
Burri and Polanco (2020) developed the TAPED database, which provides a detailed
classification and rating of digital clauses in regional trade agreements established after 2000,
enabling more in-depth research into digital trade rules. However, it is essential to stress that
the focus of digital FTAs differs due to varying interest demands behind digital rules. More
than the coverage and depth of digital FTAs are needed to comprehensively reflect the
differences and integration of regulations. Therefore, research on digital rules for digital FTAs
should focus more on the heterogeneity of digital clauses. Cross-border data flow and
localization are the primary sources of conflict in existing cross-border digital rules. Data flow
and the ability to transform data into digital intelligence have become essential for enterprises
to acquire international competitiveness, making the clauses related to data flow particularly
significant (Elsig and Klotz 2021). Spiezia and Tscheke (2020) focused on the impact of
international data agreements on bilateral trade flows of goods and services using the gravity
model. Their study concluded that international data agreements promote cross-border
transactions by enhancing cooperation under national regulatory frameworks.
Zhou and Chen (2020) researched the trade effects of representative American digital trade
rules, finding that signing agreements covering American digital rules significantly promotes
bilateral digital trade flows. However, the promoting effect of American digital rules on digital
trade is affected by the level of bilateral internet development, and the effect diminishes as
internet popularity increases. In another study, Liu et al. (2021) constructed a template
similarity index to compare the "American template" and the "European template" in FTA
digital rules. They found that although the "American template" has higher requirements for
the free flow of cross-border data, it does not significantly promote digital service trade
because its rules do not meet the actual needs of most countries. These studies highlight that
while global digital rules' integration will positively impact digital trade, rule integration, and
heterogeneity may offset the effects of the rules. Therefore, promoting both domestic
regulations of digital trade and multilateral, regional digital trade rules is necessary for digital
rule governance to implement the positive role of digital rules in promoting digital trade.
As global digital rules become more complex and a competitive and complementary pattern
emerges in global digital trade, researchers need to study digital trade from a macro
perspective that considers the overall network dimension (Amador and Cabral 2017). Trade
network analysis has been widely used in current economic research to present the
connection and dependence between countries better. Lv et al. (2021) conducted a systematic
and comprehensive study of the topological properties of the global digital service trade
network using social network analysis methods. They examined network centrality and
community change and constructed a multi-dimensional evaluation framework. In recent
years, China has shown a trend toward approaching the core region, mainly reflected in the
gradual diversification of trading partners.

8 / 38
Social network analysis provides new perspectives for the study of digital trade. This approach
transforms trade relationships into edges connecting nodes in the network. MNEs dominating
the digital trade network can be seen as network hubs connecting the digital value chain.
Furthermore, the intermediate inputs of digital trade establish close upstream and
downstream relationships, which influence the formation of the overall network relationship
through network relationship games. To study the formation of the MNEs-dominated digital
trade network, researchers use the Exponential Random Graph Model (ERGM), which views
the global structure of the network as the result of the joint action of network structure, node
attributes, and external environment (Lusher et al. 2012).
Previous research has extensively studied the formation mechanism of trade networks using
the ERGM. These studies have shown that network structure significantly impacts the
formation of trade networks. However, a more in-depth exploration of the factors influencing
the MNEs-dominated digital trade network needs to be explored. As economic entities
worldwide become increasingly interconnected in the era of the digital economy, the
formation of the digital trade network is more susceptible to the impact of the overall network
structure. This is an important starting point for the present study.
3. Characteristics of the topology structure of digital trade networks dominated by
MNEs
3.1 Data and Network Construction
The Analytical AMNE database, provided by the OECD-AMNE, contains input-output tables
from 2005 to 2016, which encompass input-output data from 59 countries worldwide and
the rest of the world (Row) (Cadestin et al. 2018). The input-output table distribution form is
similar to a general input-output table and includes intermediate goods, final goods, and
value-added sections. However, it differs in that the table classifies the input and use of
intermediate goods according to enterprise ownership into MNEs (F) and domestic
enterprises (D). This data table allows for the extraction and separate study of input-output
data of multinational companies belonging to foreign countries. Table 1 depicts the
distribution form of the input-output table.
Table 1 Example of OECD input-output table
Intermediate use
Final goods use
Total
outp
ut
1
2
G
1
2
G
D
F
D
F
D
F
Intermedi
ate
input
1
D
11
DD
Z
11
DF
Z
12
DD
Z
12
DF
Z
1G
DD
Z
1G
DF
Z
11
D
Y
12
D
Y
1G
D
Y
1
D
X
F
11
FD
Z
11
FF
Z
12
FD
Z
12
FF
Z
1G
FD
Z
1G
FF
Z
11
F
Y
12
F
Y
1G
F
Y
1
F
X
2
D
21
DD
Z
21
DF
Z
22
DD
Z
22
DF
Z
2G
DD
Z
2G
DF
Z
21
D
Y
22
D
Y
2G
D
Y
2
D
X
F
21
FD
Z
21
FF
Z
22
FD
Z
22
FF
Z
2G
FD
Z
2G
FF
Z
21
F
Y
22
F
Y
2G
F
Y
2
F
X
G
D
1G
DD
Z
1G
DF
Z
2G
DD
Z
2G
DF
Z
GG
DD
Z
GG
DF
Z
1G
D
Y
2G
D
Y
GG
D
Y
G
D
X
F
1G
FD
Z
1G
FF
Z
2G
FD
Z
2G
FF
Z
GG
FD
Z
GG
FF
Z
1G
F
Y
2G
F
Y
GG
F
Y
G
F
X
Value Added
1
D
V
1
F
V
2
D
V
2
F
V
G
D
V
G
F
V
Total output
1
D
X
1
F
X
2
D
X
2
F
X
G
D
X
G
F
X

9 / 38
Notes: Organized by the author.
This paper is founded on the research framework developed by Zhu et al. (2022), which utilizes
input-output data to differentiate between domestic and MNEs. The value-added exports of
each country are decomposed using Equation (1). Here,
V
is the diagonalized value-added
coefficient matrix, B is the Leontief inverse matrix, and Y is the final output matrix. V is classified
as either domestic
D
V
or multinational
F
V
based on ownership. L is the local Leontief
inverse matrix, which reflects the domestic industry relationships of each country.
E
A
is the
matrix of import input coefficients, which can be divided into two parts: domestic enterprises
and MNEs.
=+
+
++
+ + +
EE
D D D D D F D F
EE
F D F D F F F F
V BY V LY V L A B Y V LY V LA BY
V LY V LA B Y V LY V LA BY
(1)
This paper aims to highlight the value that MNEs bring to the digital industry by focusing on
the digital content sector. To achieve this, we used a final product matrix approach that only
considers the final output data of the digital economy industry
digital
Y
, with the data of other
industries set to zero. By doing so, we can calculate the added value trade of MNEs in the
digital industry of the host country, as shown in equation (2). Measuring digital trade is
challenging due to its penetration into various industries (Liu et al. 2021), which is why we
focused on typical digital content industries for this paper. We selected four service industries
- publishing, audiovisual and broadcasting activities, telecommunications, IT and other
information services, and financial and insurance activities - based on the National Bureau of
Statistics "Statistical Classification of Digital Economy and Its Core Industries (2021)" and the
OECD's industry digital intensity data (Calvino et al., 2018) for our analysis.
=digital
FF
V A V B Y
(2)
This paper aims to establish a benchmark network for MNEs' value-added activities in the
digital industry of host countries. In addition, this paper examines the division of labor in the
value chain among multinational companies by analyzing networks based on the differences
between multinational and domestic companies. Equations (3) to (6) represent pure domestic
value chains, domestic-multinational value chains, multinational-domestic value chains, and
multinational-multinational value chains, respectively.
+
digital E digital
DD D D D D
VA V LY V LA BY
=
(3)
digital E digital
DF D F D F
VA V LY V LA BY
=+
(4)
digital E digital
FD F D F D
VA V LY V LA BY
=+
(5)
+
digital E digital
FF F F F F
VA V LY V LA B Y
=
(6)
This paper carefully considered various variables and ultimately chose to use value-added
data from 47 countries. The data was aggregated by country, resulting in a 47x47 value-

10 / 38
added matrix. The inherent consistency between the structure of international trade and social
network structures was considered, as countries are connected through trade and exporting
companies are linked to foreign importing companies (Chaney 2016). Using the calculated
value-added matrix, the study constructed a digital trade network that is mainly dominated
by MNEs
G
：
( ( ), )
t t t t
G V X E=
. Among them, t corresponds to the sample network period
from 2014 to 2016,
t
V
representing the collection of economic entities in the network,
t
X
representing the economic attributes of the economic entities, and
t
E
representing the
dependency relationships in the network.
Social network analysis divides research objects into binary and multivalued networks based
on whether network relationships have weights, and into directed and undirected networks
based on whether the relationships have directions. This paper focuses on the calculated
incremental network G, which is a directed multivalued network. Although existing research
on multivalued networks tends to focus more on descriptive analysis, parameter estimation
for multivalued network models is still in the development stage (Krivitsky 2012; Lusher et al.
2012). Therefore, to facilitate analysis, this paper converts the multivalued network into a
binary network using a widely used method of extracting network backbone structures (Zhou
et al. 2016). This process involves binarizing the multivalued matrix using a threshold value (t),
which is determined using Equation (7). When the incremental value is greater than the
threshold value t, the network relationship is assigned a value of 1, otherwise it is assigned a
value of 0. This method ensures network sparsity while preserving the main relationships in
the network. Following the Pareto principle, the network threshold value is set to the top 20%
of all network values. Specifically, all incremental value relationships are arranged in
descending order, and only the top 20% of network relationships are retained.
1
ijt ijt
ijt ijt
E VA t
GE VA t
= 


==  


(7)
3.2 Topological Features
This paper presents an analysis of the distribution characteristics of the digital trade network,
which is dominated by MNEs. The aim of this paper is to eliminate external factors, such as
geographical distance, through visual analysis. By using a specific force-directed algorithm,
this paper transforms network associations into pure digital trade associations. Figure 1
illustrates that the digital trade network dominated by MNEs displays a non-uniform "center-
edge" distribution feature. Developed economies occupy the center of the network, while
small and developing economies are distributed around the center countries on the periphery
of the network. From a network perspective, this paper reflects that developed economies,
such as the United States and Germany, occupy the network center and dominate most
relationships worldwide in 2005 and 2016. The size of the nodes in the network represents
the number of relationships that the network nodes initiate. It is evident that there is a
significant status gap between center countries and edge countries.

11 / 38
Figure 1: Digital trade network dominated by MNEs
Note: Drawn by the author.
The visualized network illustrates the imbalance in network distribution. Additionally, this
paper calculates the outdegree centrality, indegree centrality, betweenness centrality, and
PageRank centrality of all sample countries in the digital trade network dominated by MNEs.
These indicators quantitatively measure the central position of nodes from different
perspectives. Degree centrality reflects the central position of nodes in the network based on
the number of direct connections they have established. This paper measures outdegree
centrality and indegree centrality separately, based on the directed relationships in the
network, which reflect the number of network relationships sent and received. Betweenness
centrality measures a node's control ability over other nodes in the network and reflects the
node's intermediary bridge position in the network. PageRank centrality simultaneously
considers the number and quality of adjacent nodes from the perspective of intermediary
status. When a node is connected to a node with higher centrality in the network, i.e., there
is a digital trade relationship, its PageRank centrality also increases. The utility of a node in
the network is also influenced by indirect relationships, and PageRank centrality reflects the
characteristic of "birds of a feather flock together" in the digital trade network. Therefore, this
algorithm has been widely used in the field of international trade to measure the trade
network characteristics of nodes (Lv et al. 2021).
Table 2 presents the top five ranked economies based on various centrality measures for
sample countries in 2005 and 2016. The table shows that major developed economies such
as the United States, Germany, and the United Kingdom have consistently dominated the
centers of the digital trade network, which is primarily led by MNEs. This dominance has
remained unchanged over time. The United States has always been the top-ranked economy
in terms of degree centrality, intermediary centrality, and PageRank centrality, indicating its
absolute central position in the digital trade network.
The changing centrality of China reflects its journey from the periphery to the center of the
digital trade network. In 2016, China's degree centrality entered the top five. As an "absolute
central hub" of the digital trade network dominated by MNEs, China's PageRank centrality

12 / 38
reached 0.058, ranking third after the United States and Germany. China's continuous efforts
to open its market, promote the industrialization and digitization of industries, and develop
the digital economy have attracted numerous MNEs, which has transformed China into a hub
for the interaction of the digital trade network dominated by MNEs.
Table 2 Analysis of the centrality of digital trade networks dominated by MNEs
Year
out-degree
centrality
in-degree
centrality
betweenness
centrality
PageRank
centrality
2005
UK（37）
US（38）
US（430.01）
US（0.123）
Germany（35）
Germany/UK（30）
UK（285.01）
UK（0.069）
US/ Netherlands
（30）
France（24）
Germany
（230.14）
Germany
（0.067）
Switzerland/Ireland
（29）
Japan（21）
France（50.96）
France
（0.050）
France（23）
Netherlands（17）
Netherlands
（45.20）
Japan
（0.049）
2016
UK（37）
US（38）
US（472.26）
US（0.122）
Germany（35）
Germany（27）
UK（248.05）
Germany
（0.061）
US（30）
UK（26）
Germany
（242.01）
China
（0.058）
Switzerland（27）
Switzerland/Japan
（24）
Switzerland
（105.77）
India
（0.055）
Ireland（24）
China/France（23）
India（67.05）
Japan
（0.054）
Notes: Calculated and sorted by the author, centrality is in brackets.
4. Models, Variables, and Measurements
4.1 Models
Traditional economic models assume that individuals are anonymous and interactions are
centralized, as reflected in general equilibrium analysis and imperfect competition models in
economic theory. However, social economy phenomena challenge these assumptions, such
as contagion or resilience in economics, political and social governance, cooperation patterns
between different countries, industries or enterprises, and the diffusion of ideas or products.
These phenomena typically involve a large number of objectively existing individuals who
interact with a small group of individuals, and the interacting groups have overlapping and
cross-cutting boundaries across different communities.
With the continuous development of game theory and the enrichment and improvement of
economic theoretical models, researchers can now model and analyze self-organized
network relationships based on game theory (Jackson, 2003). This paper presents a structural
model of digital trade networks that reflects the typical characteristics of economic and social
phenomena and provides a theoretical mechanism analysis based on economic structural
models for subsequent empirical research.

13 / 38
Let
 
= 1,2, ,n

be a set of individuals in the digital trade network, representing various
participating economies in the network. Each economy has a set of endogenous attributes

by class S, which can be variables such as economic size (e.g. GDP) and internet development
level. Let vector
i
X
represent the set of these attributes, and let
 
12
, , , S
i i i i
X
  
=
be
the matrix that collects all attribute sets into
 
12
, , ..., n
X X X X=
.
The digital trade network dominated by MNEs can be represented by a
nn
binary
adjacency matrix
GG
after binarization. When the element of the adjacency matrix is
=1
ij
g
, it means that there is a network relationship sent from to in the network, and when
=0
ij
g
it means that the relationship does not exist. Network relationships are bidirectional.
All elements on the diagonal are 0, i.e.
0
ii
g=
. Let the formation of the time
t
network be
denoted as
t
g
, for the convenience of analysis,
t
ij
g
represent the formation of the
relationship between
i
and
j
at the time
t
, and
t
ij
g−
represent all network relations at
the time except
t
ij
g
.
4.1.1 Preferences and Utility Functions
In a digital trade network, the primary goal of each node is to maximize its utility by building
connections. Metcalfe's Law states that the value of individuals in a network increases with
the number of connected individuals, making the number of connections a crucial factor in
determining utility. However, reciprocity also plays a significant role in the digital economy's
collaborative nature, where data sharing and deepening communication and cooperation
between supply and demand sides are essential for new business development models.
Furthermore, the law of diffusion of innovation affects utility as the first generation of products
or services entering the market can automatically gain 50% of the market share, leading to
monopolistic competition in the global digital economy. In the digital trade network, MNEs
dominate, with only a few large players truly participating in digital trade competition.
However, the Matthew effect causes an imbalanced distribution in the network, leading to a
phenomenon of "the winner takes all." Thus, the utility of individuals in the digital trade
network is determined by their direct and indirect relationships in the network. This
relationship can be expressed using the following formula:
( )
( )
1 1 1 ,
, ; ( , )
u m v
n n n n
i ij ij i ij i j ij ij ji ir ij jr
j j j r i j
U g X g u X h X X m g g g g

   
  
= = = 
= + + +
   
(8)

14 / 38
Based on Equation (8), it can be determined that the utility of network nodes participating in
a digital trade network can be divided into the following components:
(1) Direct utility of node association
+



u
ij ij
u
. This encompasses the net utility derived from
establishing relationships between nodes, which includes the sum of benefits and costs. The
size of this utility depends on the attributes of the nodes
i
X
and the assortative relationship
between them
( , )
ij
h X X
. Nodes with specific attributes are more likely to form network
connections. The direct utility of nodes is influenced by endogenous attributes such as
domestic digital regulations and digital infrastructure. Digital trade restrictions can impact the
cost of participating in the network. The higher the domestic digital regulations faced, the
higher the compliance costs required to participate in the digital trade network dominated
by MNEs, thus reducing the net benefits of participation.
Assumption 1: In the digital trade network dominated by MNEs, domestic regulations
governing digital technologies in the host country can hinder the establishment of network
connections.
In addition to the influence of endogenous attributes, direct utility is also affected by
homophily, which means that when both sides of the node have certain similar attributes,
they are more inclined to establish digital trade relationships. Let
( ) ( )
,,=−
i j i j
h X X V c X X
,
V
is the utility obtained by the node establishing association,
and
( )
,
ij
c X X
is the cost paid. In a digital trade network dominated by MNEs, assortativity
can play a crucial role in reducing trade costs. Countries that have not reached a consensus
on digital rules may face greater restrictions and coordination costs when it comes to digital
localization and market access conditions. As a result, they may experience higher costs than
benefits in digital trade. Let

represent the global digital rule variable, when the two parties
have reached a consensus on digital trade rules

=
ij
, the cost of digital trade at this time
is
c
; When the two parties have not reached an agreement on digital trade regulation,


ij
, the cost of digital trade at this time is
C
.Assume the cost function as shown in
formula (9), assuming that reaching a consensus on digital trade regulations will help reduce
trade costs, ie
0c V C  
.
( )
,ij
ij
ij
c
cC

 
=


=


(9)
Assumption 2: In the digital trade network dominated by MNEs, the establishment of digital
trade agreements plays a promoting role in establishing network connections between nodes.

15 / 38
(2) The mutual benefit of node connections

m
ij
m
. Assuming that the utility derived from
mutual relationships is symmetric, that is, from the perspective of reciprocity, the utility
obtained by both parties from this mutual relationship is equal, as shown in Equation (10).
Reciprocity is a general principle followed in international trade. When nodes are connected
to each other, they can achieve complementary advantages in resource factors through
mutual exports, thereby generating additional utility. Therefore, nodes are more likely to form
reciprocal relationships in the network.
( ) ( )
, ; , ; ,
i j m j i m
m X X m X X for all i j Z

=
(10)
Assumption 3: The digital trade network, dominated by MNEs, has a reciprocal effect in which
nodes within the network are more inclined to establish mutually beneficial relationships.
(3) Indirect utility of node associations


v
ir
. The utility of establishing network associations
may be influenced by other associations with which the node is connected. If


v
ir
is positive,
there is a positive externality, meaning that nodes are more likely to establish associations
with nodes that have more connections. Since the network is a directed network, nodes'
sending and receiving relationships will produce externalities outside the network. This paper
measures the degree to which nodes send and receive relationships externally using the terms
"expansiveness" and "convergence," respectively.
Assumption 4: Due to the presence of externalities in the network, the distribution of
relationships in the digital trade network, which is dominated by MNEs, is likely to exhibit a
skewed distribution.
After assuming the utility of the nodes, the utility of the network formation can be converted
into a specific utility function, and the factors affecting the network formation can be divided
into two categories: node attributes and network structure. At this point, the profit function
of all participants in the digital trade network can be mapped to a potential function
( )
,;

G g X
, and the network formation process can be regarded as a potential game process
(Monderer and Shapley 1996). The function is expressed as follows:
( ) ( )
( ) ( )
11
1 1 1 ,
, ; ( ( ) ( , ))
nn
ij ij u i ij i j
ij
n n n n n
ij ji ij m ij jr ir v
i j i i j r i j
G g X g u X h X X
g g m g g v

   

==
=  = = 
= + +
+

  
(11)
The potential function serves the purpose of analyzing the equilibrium process of a network
without the need to trace the actions of each participant involved. Rather, it enables the
acquisition of relevant information through the function, thereby allowing for the expression
of the change in utility of nodes in different networks through the function form (Monderer
and Shapley 1996). This serves as the foundation for analyzing network games involving
numerous participants.

16 / 38
4.1.2 Network equilibrium process
As the adjustment of digital trade networks is always in dynamic development, MNEs can
arbitrarily choose to construct network relationships in their global production network layout.
In order to obtain a closed solution for the network, we assume that the establishment of a
relationship between economic entity i and j is not affected by their existing relationship, and
the probability of establishing any group relationship in the network is positive, i.e.
( ) ( )
11
, , , , 0
tt
i j ij i j
g X X g X X

−−
−
=
. The above assumption mainly ensures that any
equilibrium network has a positive probability of being reached.
The process of network equilibrium in digital trade, dominated by MNEs, can be viewed as
each participant maximizing their own utility by establishing network relationships. It is
assumed that all network participants possess complete information, which includes an
awareness of the overall network structure and the attributes of other participants. Prior to
establishing a relationship with another participant, external shocks are considered, which can
influence a participant's preference. These random shocks simulate external factors that could
potentially affect the utility of the network node. For instance, if the target node experiences
an economic crisis, resulting in a decrease in demand, it may impact the utility of new network
relationships for other nodes. As a result, the network remains in a constant state of flux.
Participant i only has the motivation to establish a relationship with j if the utility of
establishing a new relationship in the current network configuration
1
−
−
t
ij
g
is greater than or
equal to the utility in the original state, as shown in equation (12).
( ) ( )
11
10
1, , ; 0, , ;
t t t t
i ij ij t i ij ij t
U g g X U g g X
   
−−
−−
= +  = +
(12)
So, a series of networks
01
( , ,... )
t
g g g
formed by online games constitutes a Markov chain
with a steady distribution. Because the process of network formation is traversable and the
Markov chain satisfies the Detailed Balance Condition (Butts, 2009; Mele, 2017), the network
formation game will ultimately converge to a unique steady distribution
( )
,;gX

.
( ) ( )
( )
exp , ;
,; exp , ;
G g X
gX GX


 



=


G
(13)
Assuming that the utility function's parameters are linear, the equilibrium solution of the game
formed by the digital trade network can be expressed through the exponential random graph
model (ERGM) (Mele, 2017). By building a model, structural parameters can be estimated
based on observed networks at specific time points, thereby identifying factors that affect
network formation and change. This paper establishes an ERGM model for digital trade
networks dominated by MNEs, as shown in Equation (14). Here,

represents estimated
parameters and
( )
,gX

represents model estimates, which in this paper include a series of

17 / 38
variables such as digital trade rules and network reciprocity structure. The specific variable
settings are described in the following text.
( ) ( )
( )
exp ,
,; exp ,
gX
gX X


   




=



G
(14)
ERGM is a graph model specifically designed for network data. In the model, the connections
within the network are considered as an output, while the endogenous attributes of each
network node and the network structure contribute to explaining and predicting the
probability of a connection occurring (Harris 2014). In the Exponential Random Graph Model
(ERGM), the model estimates represent the structural organization that constitutes the
network, allowing us to infer the process of network formation. Each subtype of network
effects in the network has independent explanatory power for the occurrence of relationships
in the observed network (Lusher et al. 2012). Therefore, the Exponential Random Graph Model
will help us understand the formation and dynamics of digital trade network structures.
4.2 Variables and Measurement
4.2.1 Core explanatory variable
There are significant differences in the level of domestic regulations among countries, and
the global digital rules reflect complex features. The regulation of the digital trade network
dominated by MNEs in the global economy comes from both domestic policy restrictions and
external institutional environments. On the one hand, due to the differences in the level of
economic development and core economic interests among countries, there are significant
differences in digital trade barriers between countries. On the other hand, the establishment
of global digital rules is full of various hidden barriers due to the game of interests among
various parties. Therefore, this paper studies the different forms of digital regulations from
two perspectives: domestic and international.
(1) Domestic digital regulations
Based on the Digital Service Trade Restrictiveness Index of the OECD, this paper constructs a
domestic restriction index for digital trade (DSTRI). The value of this index ranges from 0 to 1,
with higher values indicating lower openness to digital trade investment. The Digital Service
Trade Restrictiveness Index covers 50 countries, identifies and quantifies the barriers that
affect the digital service trade of various countries, and comprehensively analyzes five policy
areas: infrastructure connectivity, electronic transactions, payment systems, intellectual
property, and other areas. These restrictions or barriers bring additional costs to foreign
companies' entry and reduce the efficiency of the service industry. More importantly, as an
indicator reflecting the openness of the digital service industry, the evaluation framework of
this index focuses on domestic regulatory policies and does not consider the impact of FTA
clause regulations, so the Digital Service Trade Restrictiveness Index can be regarded as the
level of domestic regulation for the digital service industry. Since the network in this paper is
a directed network, when examining the impact of digital service trade restrictions, this paper
distinguishes between sender effects and receiver effects, and examines the impact on the
sending and receiving of network relationships.
(2) International digital rules

18 / 38
Due to the serious lag in WTO's shaping of the digital trade rules framework, there is a lack
of norms for dealing with digital trade. At the same time, global FTAs are experiencing rapid
development, and digital trade rules are generally included in new regional trade agreements.
Therefore, this paper uses digital FTAs to reflect the degree of international rule integration
at the level of the digital economy, and depicts the global digital governance pattern from
the perspective of external regulation. Due to the relatively limited scope of digital trade
agreements and the differences in rules, the inconsistency of global digital rules makes it
difficult to effectively integrate them, and the collaborative and shared nature of the digital
economy has led to increasingly serious "fragmentation" problems in global digital
governance. As more FTAs are signed and rules become increasingly detailed, the
inconsistency of digital rules will hinder the normal occurrence of digital trade.
The TAPED database of the University of Lucerne in Switzerland (Burri and Polanco 2020)
provides all free trade agreements (FTA) with digital economy clauses since 2000, which reflect
the level of rule integration in the digital economy (Liu et al. 2021). This paper establishes a
covariate matrix of rule integration (FTA) based on the TAPED database: when a digital trade
agreement is established between two economies, the relationship variable between them is
set to 1 in the network, otherwise it is set to 0. All regional free trade agreements have been
converted into a "country pair" format to construct the covariate matrix. As the commitments
and coverage of different FTAs are not the same, the impact of rule integration on coverage
and depth cannot be identified solely through digital agreements. Therefore, this paper
further examines the impact of regulatory depth and scope on the digital trade network based
on different provisions. In terms of global digital rules, cross-border data flows, as the main
carrier of the digital economy, are the foundation for promoting the global development of
the digital economy. The impact of digital flow restrictions on trade has received the most
attention (Azmeh et al. 2020). As the "open" faction represented by the United States in the
new regional trade agreements has put forward higher requirements for the free flow of
cross-border data, while the other side represented by the European Union has adopted strict
regulations for cross-border data flow, there are significant differences in the digital trade
rules related to cross-border data flow. This paper classifies all FTAs based on whether they
include the provision of "free flow of cross-border data" in the trade agreement text and
compiles a data flow rule covariate (dataflow) for cross-border data flow provisions.
Figure 2 shows the development of global digital free trade agreements (FTA). The left graph
displays the distribution of digital FTA communities. In this paper, the main feature vector
algorithm (Newman 2006) was used to measure the relationship communities of digital FTA,
and the community division followed the principle that nodes within the community have
high similarity while nodes outside the community have low similarity. The clustering analysis
results show that although the global digital rules are complex, some economies have formed
relatively tight relationship communities within a certain regional scope. Due to significant
differences in legal supervision and interest demands, the global digital rules have naturally
divided into network communities with closer internal relations.
The right graph constructs a deep association diagram of digital FTA based on the frequency

19 / 38
of digital economy-related terms appearing in the trade agreement text. Generally speaking,
the higher the frequency of related terms in the agreement, the broader the coverage of the
corresponding provisions and the more extensive the regulation of the digital economy.
Overall, the development of global digital rules mainly shows two trends: first, regionalization
and modularization are deepening. Although the global digital FTA is developing rapidly, and
more and more countries are signing digital FTAs, due to the co-construction of the digital
economy and differences in data rules and demands between Europe and the United States,
the global digital rule system is seen as a competition for leadership over global digital rules
between European and American countries, forming corresponding "American-style
templates" and "European-style templates," making the overall integration of digital rules a
problem. Second, there are significant differences in the development stages among various
digital FTAs. Due to differences in the development level of the digital economy and the
starting point of digital governance among countries, the problem of the "digital divide"
between developed and developing countries has emerged. In recent years, a series of
regional trade agreements containing second-generation digital trade rules, represented by
the Trans-Pacific Partnership (TPP), has increased the differences and divide of global digital
rules.
Figure 2 Development status of digital FTA
Notes: Drawn by the author.
4.2.2 Endogenous Structural Variables
Based on theoretical mechanisms analysis, network structural characteristics influence the
formation and dynamics of digital trade networks. The network structure can offer numerous
benefits in understanding economic behaviors (Jackson 2014). To highlight the impact of the
endogenous structure of MNEs-dominated digital trade networks, the empirical model in this
paper also includes network configurations such as the number of edges and reciprocity to
characterize the structural characteristics of digital networks, mainly including the following
aspects:
(1) The role of the number of edges （
edges
）is similar to the intercept term in a linear
regression model.
(2) Reciprocity （
reciprocity
）mainly examines the reciprocal utility between network nodes.
As trade relationships are usually two-way, this indicator measures the possibility of B also

20 / 38
pointing to A when there is a relationship from A to B.
(3) Expansion (gwodegree)， also known as geometric weighted out-degree, is used to
characterize the distribution trend of network nodes in relation to relationship sending.
(4) Convergence (gwidegree), also known as geometric weighted in-degree, is used to
characterize the distribution trend of network nodes in relation to relationship receiving.
(5) Sender effect （
sender
）is used to measure the tendency of nodes with certain attributes
to send more relationships externally.
(6) Receiver effect （
receiver
）is used to measure the tendency of nodes with certain attributes
to receive more relationships internally.
4.2.3 Control Variables
In terms of control variables, this paper controls for other variables that may affect digital
trade networks, mainly including the following variables:
(1) ICT Development Index (IDI). The ICT Development Index from the International
Telecommunication Union is used to measure the basic situation of digital infrastructure. As
digital technology is the foundation and transmission medium of digital trade, it is an
important factor affecting MNEs-dominated digital trade networks. The improvement of
bilateral internet development levels will increase the possibility of the two countries
conducting cross-border digital service trade (Meltzer 2015).
(2) Market size (lnGDP). Actual GDP of each country is used to measure market size, and the
data is from the World Bank database.
(3) Bilateral geographic distance (dist). As one of the basic variables of gravity models, bilateral
geographic distance also affects cross-border e-commerce (Kim et al. 2017).
(4) Whether there is a common language (comlang). If two countries share a common
language, it can effectively reduce communication costs at the enterprise level, thereby
promoting the development of digital trade.
(5) Whether the territories are adjacent (contig).
(6) Whether there is a colonial history (colony).
(3)-(6) are all observable trade costs, and the data is from the CEPII database. The expressions
for all network configurations and other independent variables are shown in Table 3.
Table 3 Network configuration and assumptions
Variables
Expressions
Configuration
Meaning
edges
,
ij
ij
y

Constant term
gwodegree
( )
1
0
n
j out
j
j
e d x

−−
=

Expansion
gwidegree
( )
1
0
n
j in
j
j
e d x

−−
=

Convergence
reciprocity
,
ij ji
ij
yy

Reciprocity
sender.lnGDP
,
ij i
ij
y


Sender effect
sender.DSTRI
sender.IDI

21 / 38
receiver.lnGDP
,
ji i
ij
y


Receiver effect
receiver.DSTRI
receiver.IDI
dist
,
ij ij
ij
yx

Network nesting
relationship
comlang
FTA
contig
colony
Notes: Organized by the author.
5. Empirical Research Results
5.1 Results of the Baseline Model Regression
Table 4 reports the results of the baseline regression in column (1). With respect to the core
explanatory variable, the index of digital service trade restrictions, which reflects the degree
of domestic digital regulation, is significantly negative in the sender effect. This indicates that
the higher the level of digital service trade restrictions in an economy, the lower the
probability of its participation in cross-border digital service exports by multinational
companies. Due to the high level of digital service trade restrictions, non-tariff trade barriers
make it difficult for digital service factors to flow freely, which increases operating costs and
makes it difficult for multinational companies to enter and layout, thereby affecting the
occurrence of export relationships among nodes in the network. In contrast, the receiver effect
of the index of digital service trade restrictions is 0.030 and statistically insignificant, which
indicates that digital service trade restrictions do not affect network relationships in terms of
reception. This is mainly because the overseas subsidiaries of multinational companies largely
serve the demand of their home country, and are therefore less affected by other factors. The
estimated coefficient of the global digital FTA network (FTA) is 0.660, indicating that nodes
that have signed digital FTAs are more likely to establish network connections, which suggests
that global digital rules have a positive promoting effect on the development of multinational
company-dominated digital trade networks.
Regarding the network structure variables, the estimated coefficient of reciprocity is 1.391,
and it passes the 1% significance level test, which indicates that reciprocal network
relationships promote the formation of network relationships. For the expansion and
convergence coefficients that measure the overall distribution trend of the network, only the
expansion coefficient is negative and significant, while the estimated coefficient of
convergence has not passed the significance test. The above results indicate that in terms of
sending network relationships, the network as a whole has a high skewness distribution, which
means that most nodes only send relationships to a few cooperating nodes. This skewness
distribution indicates that the global digital trade network of multinational companies is
dominated by developed economies, producing and exporting services to a few developed
economies, reflecting the "Matthew Effect" of the development of digital trade networks.
Under such a network distribution, the central nodes of the network are more likely to send
new relationships.
The estimated coefficients of the relevant control variables are consistent with expectations.
The coefficients of market size (lnGDP) and ICT development index (IDI) are significantly
positive, indicating that bilateral market size and digital technology development level both
promote the formation of multinational digital trade networks. The coefficient of the

22 / 38
geographic distance network covariate (dist) is -0.210, indicating that spatial distance still has
an impact on digital trade under globalization. The coefficients of the common language
(comlang), geographic adjacency (contig), and colonial relationship (colony) network
covariates are all significantly positive, consistent with the predictions of standard theory.
To examine the impact of restrictive measures on multinational digital trade networks in
different policy areas, this paper further empirically investigates sub-indicators of the DSTRI
index, including electronic transactions (EA), infrastructure (IC), intellectual property rights
(IPR), cross-border payment systems (PS), and other (other), and replaces the digital services
trade restriction index with the five indicators to construct different models. The results are
shown in columns (2) to (6) of Table 4. Except for electronic transactions, the sender effects
of other indicators remain negative. Among them, the effect of intellectual property rights
protection is significant, reaching -15.387. Since digital services trade involves underlying
technologies such as big data and is a key area of intellectual property protection, the degree
of restrictions in the intellectual property rights area has the most obvious effect on
multinational digital trade networks dominated by MNEs.
Table 4 Baseline regression results
(1)
(2)
(3)
(4)
(5)
(6)
total
EA
IC
IPR
PS
other
digital regulation
sender.DSTRI
-2.642 **
11.134 *
-2.165 *
-15.387 *
-7.164
-3.576
(1.070)
(5.724)
(1.299)
(8.316)
(6.387)
(4.330)
receiver.DSTRI
0.030
-5.345
-0.023
10.486
-4.995
-1.186
(1.147)
(6.916)
(1.429)
(9.067)
(7.102)
(5.479)
FTA
0.660 ***
0.647 ***
0.676 ***
0.601 ***
0.620 ***
0.647 ***
(0.167)
(0.170)
(0.163)
(0.172)
(0.162)
(0.164)
endogenous
structural
variables
edges
-57.151 ***
-56.924 ***
-56.950 ***
-57.134 ***
-57.519 ***
-57.136 ***
(4.226)
(4.410)
(4.182)
(4.441)
(4.270)
(4.533)
reciprocity
1.391 ***
1.397 ***
1.383 ***
1.394 ***
1.386 ***
1.385 ***
(0.274)
(0.270)
(0.268)
(0.277)
(0.266)
(0.277)
gwodegree
-4.036 ***
-4.064 ***
-4.015 ***
-3.819 ***
-4.142 ***
-4.079 ***
(0.471)
(0.479)
(0.474)
(0.472)
(0.487)
(0.494)
gwidegree
0.236
0.383
0.270
0.216
0.350
0.315
(0.652)
(0.653)
(0.655)
(0.659)
(0.633)
(0.681)
Control variables
sender.lnGDP
1.298 ***
1.134 ***
1.244 ***
1.283 ***
1.232 ***
1.228 ***
(0.201)
(0.199)
(0.196)
(0.205)
(0.196)
(0.217)
receiver.lnGDP
3.381 ***
3.432 ***
3.379 ***
3.323 ***
3.443 ***
3.407 ***
(0.254)
(0.264)
(0.249)
(0.260)
(0.262)
(0.288)

23 / 38
sender.IDI
0.110 **
0.222 ***
0.153 ***
0.179 ***
0.160 ***
0.168 ***
(0.054)
(0.048)
(0.048)
(0.044)
(0.051)
(0.047)
receiver.IDI
0.069
0.049
0.064
0.083 *
0.046
0.058
(0.060)
(0.051)
(0.052)
(0.047)
(0.058)
(0.053)
colony
0.626
0.727 *
0.670
0.650
0.668
0.634
(0.410)
(0.416)
(0.430)
(0.411)
(0.416)
(0.423)
contig
1.078 ***
1.079 ***
1.102 ***
1.070 ***
1.003 ***
1.048 ***
(0.346)
(0.341)
(0.345)
(0.340)
(0.340)
(0.354)
dist
-0.210 ***
-0.212 ***
-0.207 ***
-0.210 ***
-0.217 ***
-0.211 ***
(0.025)
(0.025)
(0.026)
(0.026)
(0.025)
(0.025)
comlang
1.376 ***
1.450 ***
1.439 ***
1.383 ***
1.384 ***
1.340 ***
(0.226)
(0.235)
(0.230)
(0.238)
(0.224)
(0.228)
AIC
1019.503
1020.110
1021.638
1020.982
1022.010
1024.297
BIC
1104.684
1105.292
1106.820
1106.164
1107.191
1109.479
Notes: *, **, *** indicate that the estimated coefficients are significant at the 10%, 5% and 1%
levels, respectively, and the standard errors are in brackets.
5.2 Robustness Test
In the study of social networks, robustness testing is mainly based on the construction of the
network. Currently, the construction of the trade network is generally based on the extraction
of the network backbone (Zhou et al. 2016). This approach ensures the sparsity of the network
while preserving the main relationships within it. However, as the selection of the threshold
value is subjective, this paper re-examined the empirical results through the modification of
the network construction method in the robustness analysis. In order to further examine the
impact of cross-border data flow rules on digital trade networks dominated by MNEs and to
avoid bias caused by variable measurement, this paper also conducted a robustness analysis
by changing the measurement method of the core variables. In addition, this paper used
TERGM to examine the dynamic changes of the network. All robustness tests are shown in
Table 5.
(1) Substituting the Dependent Variable: Changing the Network Threshold
Although the selection of the threshold value in this paper follows the general principle of
network construction, the subjectivity of the threshold setting may still lead to biased results.
In order to ensure the robustness of the results, this paper re-built the network with a
threshold value of 10% and re-tested it, as shown in column (1) of Table 5. It was found that
the results were basically consistent with the baseline model.
(2) Substituting the Core Independent Variable: Cross-border Data Flow
As data flow is the main carrier of digital economic activities, cross-border data flow has
become a key issue in the game of European and American digital trade rules (Ferracane et
al. 2020). Therefore, cross-border data flow rules have also become an important factor
affecting digital trade. This paper used the Digital Trade Restrictiveness Index (DTRI) of the
European Centre for International Political Economy (ECIPE) DTE database (Ferracane et al.,
2018) to replace the index of digital service trade restrictions. This index reflects a country's
legislation and law enforcement on digital trade and has the same range of values as the
index of digital service trade restrictions, which is within the range of [0,1]. A higher value
indicates a higher degree of closure of the digital economy. In terms of global digital rules,

24 / 38
the cross-border data free flow clause (dataflow) in the text of trade agreements was used as
a substitute variable for the digital FTA network covariate. When two countries sign a trade
agreement that includes cross-border data flow, their relationship in the network is defined
as 1, and 0 otherwise. The empirical results are shown in column (2) of Table 5. It was found
that the basic conclusions remained unchanged after replacing the original index with the
DTRI and the cross-border data flow clause.
5.3 Dynamic Network Analysis: Research Based on Longitudinal Models
In the ERGM model, when the node attribute is not completely exogenously given, it can still
be used as a predictor variable of the network model. However, more caution needs to be
taken in identifying causal relationships (Lusher et al. 2012). As the information provided by
cross-sectional data on which the ERGM model is based is limited, this paper measured
longitudinal data and established a multi-period network model to solve the path
dependence problem of network evolution. The Temporal Exponential Random Graph Model
(TERGM) was used to re-test the model, with the stability time trend variable (stability) added
to test the time-dependency effect. Stability reflects the trend of the network remaining
unchanged in period t+1 compared to period t. Compared to ERGM, TERGM examines
dynamic networks by treating multiple-period networks as the object of study.
Table 5 Robustness analysis
（1）
（2）
（3）
Change threshold
Change core
explanatory variables
TERGM
Digital regulation
sender.DSTRI
-2.169 *
-3.867 ***
(1.253)
(0.979)
receiver.DSTRI
-2.073
-0.521
(1.475)
(0.564)
sender.DTRI
-4.783 ***
(1.052)
receiver.DTRI
0.561
(0.773)
FTA
0.436 **
0.120 ***
(0.187)
(0.037)
dataflow
0.430 **
(0.177)
Endogenous
structural variables
edges
-58.052 ***
-61.713 ***
-37.424 ***
(5.976)
(4.871)
(0.508)
reciprocity
1.529 ***
1.472 ***
1.092 ***
(0.373)
(0.282)
(0.033)
gwodegree
-4.138 ***
-3.513 ***
-2.846 ***
(0.493)
(0.488)
(0.691)

25 / 38
gwidegree
-0.245
0.508
-0.479 **
(0.597)
(0.634)
(0.222)
Control variable
sender.lnGDP
1.027 ***
1.743 ***
0.826 ***
(0.267)
(0.246)
(0.131)
receiver.lnGDP
3.747 ***
3.412 ***
2.348 ***
(0.391)
(0.283)
(0.093)
sender.IDI
0.049
0.047
-0.007
(0.061)
(0.051)
(0.055)
receiver.IDI
0.019
0.089
0.005
(0.082)
(0.054)
(0.017)
colony
-0.174
0.573
1.216 ***
(0.528)
(0.417)
(0.319)
contig
0.958 **
1.092 ***
0.379 *
(0.396)
(0.338)
(0.215)
dist
-0.241 ***
-0.231 ***
-0.131 ***
(0.036)
(0.026)
(0.015)
comlang
1.465 ***
1.256 ***
0.573 ***
(0.264)
(0.232)
(0.104)
stability
2.863 ***
(0.037)
AIC
600.302
1001.799
BIC
685.484
1086.980
Notes： *, **, *** indicate that the estimated coefficients are significant at the 10%, 5% and 1%
levels, respectively, and the standard errors are in brackets.
5.4 Heterogeneity Analysis
5.4.1 Heterogeneity analysis based on value-added
Firstly, we compare the domestic enterprise network with the cross-border company network
under the benchmark study. The results in column (1) of Table 6 show that under the domestic
enterprise digital trade network, the sender and receiver effects estimated coefficients of
DSTRI are -0.542 and -1.062, respectively, and both are not significant. The estimated
coefficient of FTA is positive and significant, but the coefficient value is smaller than that of
the FTA coefficient in the benchmark regression. This result indicates that the export of value-
added by domestic enterprises will not be affected by the restrictions on domestic digital
services trade, and the influence of global digital rules will be even smaller.
Under the D-F value chain, the value-added network traces the use of domestic enterprise-
created value-added by cross-border enterprises in domestic final and export product
production activities in the digital industry, reflecting the forward relationship between cross-
border companies and upstream domestic enterprises. At this time, the sender effect of DSTRI
is positive but not significant, indicating that domestic digital regulations do not have a

26 / 38
suppressive effect on D-F value chain production.
Under the F-D and F-F value chains, the value-added network traces how the value-added
created by cross-border companies is used in domestic enterprise and cross-border
enterprise domestic final and export product production activities in the digital industry,
reflecting the backward relationship of cross-border companies. At this time, the sender effect
of DSTRI is -2.228 and -2.748, respectively, and the coefficients are negative and significant,
indicating that domestic digital regulations have a suppressive effect on F-D and F-F value
chain production.
In summary, we found that in digital value-added trade, the digital service trade restrictions
of the host country mainly hinder the export of value-added by cross-border companies
located upstream. It can be said that domestic digital regulations mainly hinder the value-
added created by cross-border companies, rather than affecting the value-added created by
domestic enterprises.
Table 6 Heterogeneity analysis based on added value
（1）
（2）
（3）
（4）
Domestic
enterprises
D-F
F-D
F-F
Digital regulation
sender.DSTRI
-0.542
1.764
-2.228 **
-2.748 **
(1.216)
(1.115)
(1.070)
(1.079)
receiver.DSTRI
-1.062
0.268
-0.862
-0.009
(1.380)
(1.189)
(1.237)
(1.141)
FTA
0.546 ***
0.818 ***
1.005 ***
0.629 ***
(0.180)
(0.176)
(0.174)
(0.163)
Endogenous
structural variables
edges
-65.216 ***
-65.852 ***
-62.208 ***
-55.698 ***
(5.205)
(5.085)
(4.818)
(4.284)
reciprocity
2.298 ***
1.364 ***
1.279 ***
1.459 ***
(0.299)
(0.277)
(0.279)
(0.272)
gwodegree
-4.075 ***
-3.021 ***
-4.502 ***
-3.726 ***
(0.610)
(0.546)
(0.495)
(0.474)
gwidegree
0.345
-0.264
0.200
0.143
(0.744)
(0.683)
(0.672)
(0.649)
Control variable
sender.lnGDP
2.064 ***
2.251 ***
1.653 ***
1.277 ***

27 / 38
(0.272)
(0.244)
(0.218)
(0.201)
receiver.lnGDP
3.272 ***
3.027 ***
3.465 ***
3.285 ***
(0.311)
(0.274)
(0.279)
(0.257)
sender.IDI
0.068
0.264 ***
0.148 ***
0.116 **
(0.060)
(0.059)
(0.055)
(0.054)
receiver.IDI
0.025
0.034
-0.010
0.053
(0.067)
(0.062)
(0.061)
(0.060)
colony
-0.534
0.397
0.654
0.640
(0.458)
(0.434)
(0.446)
(0.412)
contig
1.244 ***
1.189 ***
1.533 ***
1.039 ***
(0.363)
(0.347)
(0.369)
(0.346)
dist
-0.173 ***
-0.191 ***
-0.192 ***
-0.209 ***
(0.023)
(0.024)
(0.024)
(0.026)
comlang
1.753 ***
1.246 ***
1.276 ***
1.319 ***
(0.243)
(0.225)
(0.231)
(0.231)
AIC
833.449
963.778
965.186
1037.659
BIC
918.631
1048.960
1050.368
1122.841
Notes：*, **, *** indicate that the estimated coefficients are significant at the 10%, 5% and 1%
levels, respectively, and the standard errors are in brackets.
5.4.2 Heterogeneity Analysis Based on Digital Industry
In order to examine the differences in the impact of digital regulation on different industries
and to ensure the robustness of the research results, this paper analyzes the value-added
networks for each industry. Table 7 reports the empirical results of constructing networks
based on the value-added of four industries. In the IT and other information service industry,
and the financial and insurance industry, the sender effect of DSTRI is still significantly negative,
while in the publishing, audio and video broadcasting industry, and the telecommunications
industry network, although the sender effect of DSTRI is negative, it is not statistically
significant. The estimated coefficients of FTA are positive in all networks, indicating that global
digital regulation promotes cross-border digital value-added trade.
Table 7 Heterogeneity analysis based on digital industry
（1）
（2）
（3）
（4）
Publishing
audiovisual and
broadcasting
Telecommunications
IT and other
information
services
Finance and
insurance

28 / 38
Digital regulation
sender.DSTRI
-1.202
-0.876
-2.184 **
-4.312 ***
(1.050)
(0.988)
(1.014)
(1.019)
receiver.DSTRI
0.546
-0.663
1.244
-1.552
(1.177)
(1.069)
(1.120)
(1.092)
FTA
0.943 ***
0.743 ***
0.527 ***
0.463 ***
(0.163)
(0.166)
(0.163)
(0.156)
Endogenous
structural
variables
edges
-61.957 ***
-53.458 ***
-39.134 ***
-47.748 ***
(4.541)
(4.012)
(3.407)
(3.653)
reciprocity
1.394 ***
0.468 *
1.226 ***
1.609 ***
(0.289)
(0.253)
(0.255)
(0.265)
gwodegree
-3.692 ***
-3.499 ***
-4.187 ***
-3.163 ***
(0.499)
(0.459)
(0.441)
(0.450)
gwidegree
0.077
-0.865
-1.846 ***
0.203
(0.638)
(0.619)
(0.517)
(0.652)
Control variable
sender.lnGDP
1.364 ***
1.654 ***
0.914 ***
1.211 ***
(0.215)
(0.195)
(0.180)
(0.186)
receiver.lnGDP
3.632 ***
2.824 ***
2.198 ***
2.763 ***
(0.267)
(0.226)
(0.196)
(0.226)
sender.IDI
0.165 ***
0.100 **
0.035
0.153 ***
(0.053)
(0.048)
(0.050)
(0.049)
receiver.IDI
0.081
-0.100 *
0.220 ***
-0.107 *
(0.059)
(0.054)
(0.061)
(0.055)
colony
0.778 *
0.923 **
0.052
-0.018
(0.412)
(0.425)
(0.412)
(0.403)
contig
1.254 ***
1.694 ***
1.718 ***
0.983 ***
(0.332)
(0.350)
(0.379)
(0.318)
dist
-0.177 ***
-0.211 ***
-0.143 ***
-0.149 ***
(0.023)
(0.024)
(0.021)
(0.021)

29 / 38
comlang
1.022 ***
0.820 ***
1.545 ***
1.138 ***
(0.235)
(0.228)
(0.237)
(0.205)
AIC
1010.060
1164.755
1086.908
1141.018
BIC
1095.242
1249.937
1172.090
1226.199
Notes：*, **, *** indicate that the estimated coefficients are significant at the 10%, 5% and 1%
levels, respectively, and the standard errors are in brackets.
6. Further Research: The Impact of Digital Service Regulation Heterogeneity on Cross-
border Digital Output of MNEs
Through the previous research, we found that both internal digital regulation and external
digital rules have an impact on the digital trade network of MNEs. However, the trade network,
as a reflection of the digital trade relationship of MNEs, only depicts the export distribution
of MNEs and does not provide information on their sales in the host country. OECD data
shows that over 60% of the output of foreign subsidiaries is sold domestically in the host
country (Cadestin et al., 2018), which is obviously missing from the analysis of the trade
network. Additionally, the trade network cannot explain the relationship between MNEs and
their home and host countries. MNEs often face a choice between trade and investment when
organizing cross-border production, and the main problem lies in differences in institutions
and regulations. Restrictions on digital service trade hinder the output of MNEs in the digital
service industry. When the differences in domestic digital regulations are small, the integration
of external rules will not generate additional costs due to regulatory differences, which
provides convenience for the production of MNEs. The integration of digital rules will lead
MNEs to choose to provide services in the international market by establishing subsidiaries
locally. The output of these foreign subsidiaries in the service industry is defined as "mode 3"
service trade, which refers to the form in which suppliers provide services by establishing
commercial entities locally (Andrenelli et al., 2018).
In order to further examine the impact of digital regulation on the digital service output of
MNEs, this paper constructs a digital service output correlation network of MNEs based on
the output data of MNEs and the relationship between the home country and the host country.
The research is also based on the digital services output of 47 economies. Service industry,
build network M：
( ) ( )
( , , )
t t t t t
M V X W X E=
, where
t
V
and
t
W
represent the home and
host countries of MNEs, respectively. Similar to the construction of a digital trade network,
the threshold is also set at 20%, that is, all output relationships are arranged in descending
order, and only the top 20% of network relationships are retained. When the total output of
the digital service industry of MNEs located in a country is greater than the threshold value,
the network relationship between the home country and the host country is marked as 1,
otherwise it is marked as 0. For example, when the digital service output of a multinational
company affiliated to the United States in China is greater than the threshold value n
(
USA CH N
output n
), the network relationship between the United States and China in the

30 / 38
network M is marked as 1 (
=1
USA CHN
E
), otherwise it is 0 as following:
1
ijt ijt
ijt ijt
E output n
ME output n
= 


==  


(15)
Due to the cross-border production of MNEs, they face dual regulation, restrictions, and
supervision from both their home country and the host country, which undoubtedly increases
the cost of exporting for businesses. UNCTAD (2017) pointed out that many countries' digital
development strategies do not fully meet investment needs, which adds additional costs to
the integration of external digital rules. Therefore, the distribution of output-related networks
is more sensitive to institutional barriers and regulatory differences between the home
country and the host country. The main purpose of this paper is to further examine the impact
of differences in digital service trade restrictions on cross-border production. This paper uses
the OECD's digital STRI heterogeneity indices to construct the digital service trade restriction
difference network covariate (DSTRIH), which measures the difference in the level of digital
service trade regulation between bilateral economies and assigns certain weights to different
policy areas. The higher the degree of regulatory difference between two countries in digital
services, the larger the index. The OECD constructed the index based on two principles:
"answer" and "score". The score-based scoring principle further incorporates an examination
of market structure on the basis of the answer principle. This paper uses the score-based
digital STRI heterogeneity index for empirical research. In addition, following the method of
Wang Lan (2022), the World Bank's WGI database is used to measure the strength of each
country's legal system and regulatory intensity using the Rule of Laws indicator, and the
difference in the Rule of Laws index (absdiff.rol) is used as a substitute indicator for regulatory
policy heterogeneity index, as shown in column (2) of Table 8.
Figure 3 visualizes the digital service trade regulation heterogeneity index among the 47
sample countries using a heat map. It can be seen that the regulatory differences in digital
service restrictions between economies are relatively obvious, forming a "gap" between
different economies. Due to the lower level of digital service trade restrictions in developed
countries and the higher level of digital service restrictions in developing countries, the
differences between developed countries are smaller than those between developed
countries and developing countries.

31 / 38
Figure 3 Heterogeneity of digital service regulation
Notes：Drawn by the author.
Through the previous analysis of the global digital FTA community, we found that the
distribution of the digital value chain shows a clear regional distribution. Due to the
collaborative and shared nature of the digital economy, countries have formed an invisible
"digital rule gap", and multinational companies will be affected by differences in digital
regulations when conducting global production layout. Therefore, multinational companies
may be more inclined to lay out their production networks within the same digital rules. Table
8 shows the composition of each community in the digital FTA network. Based on the
community analysis using the characteristic vector algorithm, the digital FTA network is
divided into three major communities. Community 1 is mainly composed of members of the
original Trans-Pacific Partnership (TPP), mainly from Asia-Pacific countries; Community 2 is
mainly composed of EU member states and some South American countries; and Community
3 includes China, South Korea, and other countries. The TAPED database categorizes all digital
FTAs according to the degree of template similarity, into "American-style templates",
"European-style templates", and other types of templates. It can be observed that the
distribution of community members in the network coincides with the distribution of
members in American-style and European-style templates. To examine the impact of
modularization of global digital rules on the digital service output networks of multinational
companies, this paper added the homophily analysis of the network community variable. The
community variable is used to mark the community in which the network is located, and
homophily analysis reflects whether nodes in the network tend to choose nodes in the same
community to establish relationships. In this case, the community is used as a substitute
variable for the global digital FTA network to examine the impact of differences in global
digital rules. The results are shown in column (3) of Table 9.

32 / 38
Table 8 Composition of Digital FTA Network Community Members
community
Countries
community 1
United States, Canada, Mexico, Japan, Australia, New Zealand, Turkey,
India, Indonesia, Malaysia, Chile
community 2
Austria, Belgium, Czech Republic, Germany, Denmark, Spain, Estonia,
Finland, France, United Kingdom, Greece, Hungary, Ireland, Italy, Lithuania,
Luxembourg, Latvia, Netherlands, Poland, Portugal, Slovakia, Slovenia,
Sweden, South Africa, Brazil, Argentina, Colombia
community 3
China, South Korea, Costa Rica, Saudi Arabia, Iceland, Switzerland, Norway
Notes: Organized by the author.
The results in column (1) of Table 9 show that the homogeneity coefficient of digital service
trade restrictions (DSTRIH) is -2.256, with a negative coefficient that passes a 1% significance
level test. This indicates that the greater the regulatory differences between countries in digital
services, the less multinational production there will be for multinational companies. This is
because the greater the differences in institutional regulation, the more difficult it is to achieve
regulatory consistency among member countries through FTAs (Peng Yu et al., 2021). Non-
tariff barriers between countries impose additional costs on multinational companies to
comply with different regulatory rules, thereby hindering the development of digital trade
networks (Liu Hongkui, 2020). In column (2), after replacing the variable for regulatory
differences, the estimated coefficient for legal regulatory differences (absdiff.rol) is -0.295,
which is negative and also indicates that regulatory differences will hinder the multinational
production of digital services by multinational companies. In column (3), the homogeneity
coefficient for network communities is 0.619, which is significantly positive at the 1% level,
indicating a strong tendency for countries within the same community to establish network
connections. Since the production network layout of digital multinational companies needs
to consider upstream and downstream relationships, nodes belonging to the same
community have more third-party relationships in common, which will help multinational
companies' multinational production and investment activities, thereby showing a positive
homogeneity coefficient for network communities.
In terms of network structure variables, the reciprocity coefficient of the network is not
significant, indicating that reciprocal structures do not promote the establishment of network
relationships. This is mainly because the vast majority of digital multinational companies are
from developed countries, and developed economies dominate most of the relationships in
the network. Therefore, the network is not symmetrical in terms of relationships. The
coefficients of gwodegree and gwidegree are both negative and significant, indicating that
the relationships in the network are directed towards relatively fewer economies, and the
network is more concentrated in distribution, showing a higher degree of skewed distribution.
In addition, the estimation results of each control variable in columns (1)-(3) remain consistent
without significant changes. The sender and receiver effects of lnGDP and IDI are both positive,
indicating that both market size and network infrastructure improvements will enable nodes

33 / 38
to send and receive more associative relationships in the network, which will help
multinational companies' multinational production activities in the digital service industry. In
terms of network control variables, territorial adjacency and colonial relationship networks are
both significant and positive, which is intuitive, as geographic proximity and historical colonial
relationships will increase the probability of forming multinational production relationships
for digital multinational companies. Similarly, the covariant network coefficient of common
languages is positive, indicating that the same language background will promote the
multinational production relationships of digital multinational companies. The coefficient of
geographic distance is negative, indicating that geographic distance will inhibit the
occurrence of multinational production relationships for MNEs in the digital service industry.
Table 9 Empirical results of digital service output association network of MNEs
(1)
(2)
（3）
Digital regulation
DSTRIH
-2.256 ***
-1.859 **
(0.806)
(0.825)
absdiff.rol
-0.295 ***
(0.108)
FTA
0.414 ***
0.350 **
(0.159)
(0.168)
community
0.619 ***
(0.162)
Endogenous
structural variables
edges
-53.545 ***
-52.456 ***
-54.318 ***
(3.670)
(3.648)
(3.800)
reciprocity
0.058
0.077
0.019
(0.254)
(0.251)
(0.257)
gwodegree
-2.248 ***
-2.254 ***
-2.254 ***
(0.697)
(0.668)
(0.738)
gwidegeree
-2.470 **
-2.560 **
-2.506 **
(1.073)
(1.037)
(1.138)
Control variable
sender.lnGDP
2.470 ***
2.397 ***
2.495 ***
(0.176)
(0.174)
(0.186)
receiver.lnGDP
1.755 ***
1.707 ***
1.755 ***
(0.178)
(0.171)
(0.175)
sender.IDI
0.380 ***
0.422 ***
0.406 ***
(0.047)
(0.042)
(0.045)
receiver.IDI
0.075
0.093 **
0.099 **
(0.047)
(0.042)
(0.045)

34 / 38
colony
1.024 ***
1.121 ***
1.033 ***
(0.379)
(0.393)
(0.373)
contig
1.185 ***
1.113 ***
1.209 ***
(0.325)
(0.314)
(0.327)
dist
-0.085 ***
-0.096 ***
-0.081 ***
(0.019)
(0.018)
(0.018)
comlang
0.775 ***
0.720 ***
0.808 ***
(0.228)
(0.233)
(0.230)
AIC
1301.713
1302.682
1292.137
BIC
1381.216
1382.185
1371.640
Notes: *, **, *** indicate that the estimated coefficients are significant at the 10%, 5% and 1%
levels, respectively, and the standard errors are in brackets.
7. Main Conclusions and Policy Recommendations
The digital trade of multinational enterprises (MNEs) is reflective of the trend of digital
globalization. This paper examines the digital trade network and digital production network
led by MNEs and draws the following conclusions: Firstly, the analysis of network
characteristics reveals that the distribution of the digital trade network led by MNEs displays
a polarization trend, with most countries having relatively concentrated export destinations,
which reflects the uneven development of digital services. The formation of the digital trade
network led by MNEs is influenced by the endogenous structure of the network. The mutually
beneficial structure of the network promotes the formation of network relationships, while
overall preference dependence determines the "center-edge" structure of the network.
Secondly, in the analysis of the formation mechanism of the digital trade network led by MNEs,
the sender effect of the digital services restriction index is significantly negative, indicating
that the degree of domestic regulation of digital services limits the export of digital services
by MNEs. This means that strict digital service trade restrictions in a country are not conducive
to the production of digital services by MNEs in that country. In the analysis of the
heterogeneity of value added, this paper found that when MNEs act as providers of upstream
value added, restrictions on digital service trade hinder the export of value added. However,
when MNEs act as producers of downstream final products, restrictions on digital service
trade have no impact.
Thirdly, global digital rules represented by digital Free Trade Agreements (FTAs) will promote
the digital trade of MNEs, and cross-border data flow policies as an important component of
global digital rules will have a significant impact on the digital trade network, especially for
MNEs that frequently use cross-border digital elements. Their operations will inevitably
involve cross-border access and transmission of data, making the requirements for cross-
border data flow higher.
Lastly, further research found that the difference in domestic digital service regulation
between host countries and home countries is not conducive to the production of digital
services by MNEs. From the perspective of network analysis, the divergence of global digital
rules has led to the division of global digital rules into several network communities, which
makes it easier for countries under the same digital rules community to establish cross-border

35 / 38
production relationships.
The cross-border production of MNEs has significant implications for the economic
development of both host and home countries. Digital MNEs, in particular, combine digital
capital with local data to provide digital services to host countries, thereby promoting digital
development and indirectly impacting the production efficiency of host countries. Thus,
actively attracting direct investment from digital MNEs is advantageous for host countries. In
the era of digital economy, MNEs from countries with different digital trade rules than the
American and European templates use new generation information technology to optimize
the global market, improve core competitiveness, and efficiently connect with the
international circulation. This paper offers the following policy implications:
Firstly, national policies should coordinate at the international level to effectively integrate
and implement digital rules. Given the different national interests, there is a significant gap in
the depth of international digital rules between countries. Countries need to balance free
trade with national network security when formulating corresponding rules. Digital latecomer
economies can make up for the shortcomings in global multinational corporation’s trade by
formulating effective policies to better integrate into the digital trade network led by MNEs.
Secondly, in addition to domestic regulations and international rules, the structure and level
of digital infrastructure determine the balance of the digital trade network led by MNEs.
Imbalance in its development leads to the Matthew Effect of "the rich getting richer,"
providing central economies in the network with more significant advantages. Developing
economies can participate in global digital governance and find a balance between digital
openness and national economic security. They should leverage their potential market size
and continue participating in the network as a central hub.
Overall, this paper highlights the importance of cross-border production by digital MNEs for
the economic development of host and home countries. Furthermore, it offers policy
implications for digital latecomer economies to better integrate into the digital trade network
led by MNEs and find a balance between digital openness and national economic security.

36 / 38
References:
Alcácer J, Cantwell J, Piscitello L (2016) Internationalization in the information age: A new era for
places, firms, and international business networks? Journal of International Business Studies
47(5):499-512. https://doi.org/10.1057/jibs.2016.22
Amador J, Cabral S (2017) Networks of Value-added Trade. The World Economy 40(7):1291-
1313. https://doi.org/https://doi.org/10.1111/twec.12469
Azmeh S, Foster C, Echavarri J (2020) The International Trade Regime and the Quest for Free
Digital Trade. International Studies Review 22(3):671-692. https://doi.org/10.1093/isr/viz033
Bolwijn R, Casella B, Zhan J (2018) International Production and the Digital Economy. In: van
Tulder R, Verbeke A, Piscitello L (eds) International Business in the Information and Digital Age,
Emerald Publishing Limited, pp 39-64
Burri M, Polanco R (2020) Digital Trade Provisions in Preferential Trade Agreements: Introducing
a New Dataset. J Int Econ Law 23(1):187-220. https://doi.org/10.1093/jiel/jgz044
Cadestin C, Backer KD, Desnoyers-James I, Miroudot S, Ye M, Rigo D (2018) Multinational
enterprises and global value chains. In: OECD Science, Technology and Industry Working Papers,
No 2018/05. OECD, Paris
Cadestin C, Jaax A, Miroudot S, Zürcher C (2021) Multinational enterprises and intangible capital.
https://doi.org/doi:https://doi.org/10.1787/6827b3c9-en
Capello R, Lenzi C, Panzera E (2022) The rise of the digital service economy in European regions.
Industry and Innovation:1-27. https://doi.org/10.1080/13662716.2022.2082924
Chaney T (2016) Networks in International Trade. In: Bramoullé Y, Galeotti A, Rogers BW (eds)
The Oxford Handbook of the Economics of Networks. Oxford University Press, pp 754–775
Drynochkin A, Fedorchenko A, Habarta A, Kaveshnikov N, Komkova E, Kostyunina G,
Morozenskaya E (2023) World Economy Major Trends: International Economic Integration. In:
Bulatov A (ed) World Economy and International Business: Theories, Trends, and Challenges,
Springer International Publishing, Cham, pp 93-122
Elsig M, Klotz S (2021) Data Flow-Related Provisions in Preferential Trade Agreements: Trends
and Patterns of Diffusion. pp 42-62
Ferencz J (2019) The OECD Digital Services Trade Restrictiveness Index.
https://doi.org/doi:https://doi.org/10.1787/16ed2d78-en
Ferracane MF, Kren J, van der Marel E (2020) Do data policy restrictions impact the productivity
performance of firms and industries? Rev Int Econ 28(3):676-722.
https://doi.org/https://doi.org/10.1111/roie.12467
Harris JK (2014) Building a Useful Exponential Random Graph Model. An Introduction to
Exponential Random Graph Modeling, SAGE Publications, Inc., Thousand Oaks
Thousand Oaks, California
Hsieh C-S, König MD, Liu X (2022) A Structural Model for the Coevolution of Networks and
Behavior. The Review of Economics and Statistics 104(2):355-367.
https://doi.org/10.1162/rest_a_00958
Jackson MO (2014) Networks in the Understanding of Economi c Behaviors. Journal of Economic
Perspectives 28(4):3-22. https://doi.org/10.1257/jep.28.4.3
Kim TY, Dekker R, Heij C (2017) Cross-Border Electronic Commerce: Distance Effects and Express
Delivery in European Union Markets. International Journal of Electronic Commerce 21(2):184-
218. https://doi.org/10.1080/10864415.2016.1234283

37 / 38
Krivitsky PN (2012) Exponential-family random graph models for valued networks. Electronic
Journal of Statistics 6(none):1100-1128. https://doi.org/10.1214/12-EJS696
Li B, Zhang S (2022) Research on the development path of China's digital trade under the
background of the digital economy. Journal of Internet and Digital Economics 2(1):1-14.
https://doi.org/10.1108/JIDE-10-2021-0010
Liu B, Zhen Y, Li X (2021) The Impact of Regulatory Convergence on Digital Trade: An
Examination Based on WIOD Digital Content Industry. journal of world econmy 44(07):3-28.
https://doi.org/10.19985/j.cnki.cassjwe.2021.07.002
Lusher D, Koskinen J, Robins G (2012) Exponential Random Graph Models for Social Networks:
Theory, Methods, and Applications. Cambridge University Press, Cambridge
Lv Y, Fang R, Wang D (2021) The Topological Structure Characteristics and Influence Mechanism
of the Global Digital Service Trade Network. Quantitative Economics Technical Economics
Research 38(10):128-147. https://doi.org/10.13653/j.cnki.jqte.2021.10.007
Mele A (2017) A Structural Model of Dense Network Formation. Econometrica 85(3):825-850.
https://doi.org/https://doi.org/10.3982/ECTA10400
Meltzer JP (2015) The Internet, Cross-Border Data Flows and International Trade. Asia & the
Pacific Policy Studies 2(1):90-102. https://doi.org/https://doi.org/10.1002/app5.60
Monderer D, Shapley LS (1996) Potential Games. Games and Economic Behavior 14(1):124-143.
https://doi.org/https://doi.org/10.1006/game.1996.0044
Newman MEJ (2006) Finding community structure in networks using the eigenvectors of
matrices. Physical Review E 74(3):036104. https://doi.org/10.1103/PhysRevE.74.036104
OECD (2021) Development Co-operation Report 2021.
Peng D, Ouyang Z (2020) Digital Trade Rules of Main WTO Members and Their Implications for
China. J WTO & China 10:95
Smeets M (2021) Adapting to the digital trade era, challenges and opportunities (WTO).
Spiezia V, Tscheke J (2020) International agreements on cross -border data flows and
international trade. https://doi.org/doi:https://doi.org/10.1787/b9be6cbf-en
Strange R, Chen L, Fleury MTL (2022) Digital Transformation and International Strategies. Journal
of International Management 28(4):100968.
https://doi.org/https://doi.org/10.1016/j.intman.2022.100968
Suh J, Roh J (2023) The effects of digital trade policies on digital trade. The World Economy
n/a(n/a). https://doi.org/https://doi.org/10.1111/twec.13407
UNCTAD (2015) International Trade in ICT Services and ICT-enabled Services. In: UNCTAD
Technical Notes on ICT for Development.
UNCTAD (2017) World Investment Report 2017: investment and the digital economy. In: United
Nations Conference on Trade and Development, United Nations, Geneva. pp 1-56
UNCTAD (2019) Digital Economy Report 2019: Value Creation and Capture: Implications for
Developing Countries. In.,
van der Marel E, Ferracane MF (2021) Do data policy restrictions inhibit trade in services? Review
of World Economics 157(4):727-776. https://doi.org/10.1007/s10290-021-00417-2
Zhan JX (2021) GVC transformation and a new investment landscape in the 2020s: Driving forces,
directions, and a forward-looking research and policy agenda. Journal of International Business
Policy 4(2):206-220. https://doi.org/10.1057/s42214-020-00088-0
Zhou M, Wu G, Xu H (2016) Structure and formation of top networks in international trade,

38 / 38
2001–2010. Social Networks 44:9-21.
https://doi.org/https://doi.org/10.1016/j.socnet.2015.07.006
Zhou N, Chen H (2020) Research on the digital trade effect of American digital trade rules under
the framework of RTAs. the Journal of World Economy 43(10):28-51.
https://doi.org/10.19985/j.cnki.cassjwe.2020.10.003
Zhu K, Yu X, Wei S, Wang Z (2022) Measuring the Activities of Multinational Corporations in
Global Value Chains and Traceability of Their Added Value. Economic Research Journal
57(03):136-154