Do Digital Trade Rules Matter? Empirical Evidence from TAPED

Abstract

This study aims to examine (1) the impact of digital trade rules on digital services exports, (2) the heterogeneity of this impact in different trade categories, (3) the impact of digital trade rules on the growth of services exports through reducing trade costs, and (4) whether this impact is different in national income levels. In order to test the hypotheses, this study uses panel data from the TAPED database encompassing 143 economies between 2005 and 2019, employing a differences-in-differences model as the analytical framework. Our empirical results yield several key findings. Firstly, digital trade rules have a significant role in promoting the export of digital services. Secondly, the impact of trade rules on different types of trade and national income levels is significantly different. Thirdly, the mechanism analysis results show that trade costs have an intermediary role, and digital trade rules can promote the export of digital services by reducing trade costs. Finally, compared to “high-income countries”, the establishment of digital trade rules brings greater benefits to the digital services trade of low- and middle-income countries. Digital trade has brought significant changes to globalization development. Few studies address the role of digital trade rules in regulating the development of digital trade, and some researchers suggest reconstructing the global trade rules to meet the growth of digital trade. Nevertheless, research on how digital trade rules affect digital services exports is still sparse. This study contributes to the literature by revealing the importance of digital trade rules and provides a reference to promote institutional openness in the field of digital trade.

1. Introduction

Digital trade has brought significant changes to the globalization development model and has expanded the breadth and depth of international trade, acting as a potent engine for promoting global economic growth and recovery. According to UNCTAD’s statistical data, the export volume of global digital trade, with digital service trade in exports as the core, has increased from USD 2.53 trillion in 2015 to USD 3.81 trillion in 2021. China’s digital service trade has also grown rapidly, with the volume of digital service trade increasing from USD 200 billion in 2015 to USD 294.76 billion in 2020, constituting 44.5% of service trade. Thus, digital service trade refers to the trade of digital products and services delivered through the cross-border transmission of information and communication networks. It is characterized by the digital knowledge and information of key elements as the core content and the transmission and completion of transactions with the help of the modern information network [1]. Nevertheless, digital trade faces numerous challenges, with digital trade barriers being one of the major impediments to the rapid development of digital trade. In response, many countries have implemented digital trade agreements with digital trade rules at the core or added digital trade rules to traditional free trade agreements to regulate the growth of digital trade. Developed countries such as the United States, the European Union, and Japan, and emerging economies such as China, Russia, and India have integrated into the “circle of friends” of digital trade, promoting its growth and building bilateral or multilateral digital trade agreements with other nations. In June 2020, Singapore, New Zealand, and Chile signed the Digital Economy Partnership Agreement (DEPA), becoming the world’s first “pure digital,” modular, and non-binding trade agreement. In November 2020, the Regional Comprehensive Economic Partnership Agreement (RCEP), signed by the 10 ASEAN countries and 15 Asia-Pacific countries, including China, Japan, South Korea, Australia, and New Zealand, added chapters related to digital trade rules. Nevertheless, the question arises whether digital trade rules such as these can play a similar role in safeguarding and promoting free trade as those of the Free Trade Agreement (FTA) and World Trade Organization (WTO) agreements.

Digital trade rules have garnered considerable academic attention, with the scholarly community dividing trade rules into various models, such as the “American style template” represented by the United States, the “European style template” represented by the European Union, the “Chinese style template” represented by China, and the “South Pacific template” represented by Australia and New Zealand. Scholars have researched the content and related clauses of these templates, with the focus of American-style digital service trade rules being “cross-border data free flow”, non-mandatory localization of data storage, and source code protection [2,3]. The core of European digital service trade rules lies in intellectual property protection, the audiovisual sector, and privacy protection [4,5]. The “Chinese style template” represented by China regards data as a national strategic asset, setting high standards for the protection of personal data privacy, corporate data confidentiality, and national data security [6]. However, the digital trade rules under these templates are essentially at the national or regional level, resulting in “isolated island” and “overlapping” fragmentation patterns, which, to some extent, even form the root of digital trade barriers [7,8].

There is a paucity of empirical studies on coordinating bilateral digital trade rules to regulate the development of digital trade [9]. As a fundamental institutional arrangement for digital trade, bilateral digital trade rules may play an important role in regulating the development of digital trade. Sheng and Gao [10] called for negotiations and consultations on digital trade rules that are not yet included in the multilateral trade system to respond to the demands of global digital governance Ma and Shen [11] proposed strengthening the reconstruction of global trade rules to address the development of digital trade. Many researchers address the type of FTA and digital trade rules. However, research on the impact of digital trade rules on digital trade is still lacking. In order to address this gap, this article aims to utilize a double difference model and draw on the TAPED database from the University of Lucerne [12], which covers digital service trade rules, to investigate the influence of digital service trade rules on digital service exports, thus providing empirical evidence for the comprehensive promotion of digital service trade rules. This article is innovative in two ways compared to prior research: First, it highlights the varied effects of digital trade regulations on the exports of digital services by nations (regions) with various service industries and levels of economic development. Additionally, it identifies the intermediary routes of reducing cost by which digital trade agreements provide the promotion effect on exports of digital services. This study provides a reference for promoting institutional openness in the field of digital trade, which has a promoting effect on sustainable economic development.

2. Literature Review

Digital trade rules are the fundamental institutional arrangements for regulating the development of digital trade. The research topics closely related to this article mainly include the study of free trade agreement (FTA) trade rules and their effects, the study of national and regional digital trade rules, and the trade effects of digital trade rules.

2.1. Research on FTA Trade Rules and Their Effects

The core of multilateral trade rules is WTO rules. The evolution path of the new generation of international trade rules is based on regional trade rules [13]. The United States has established a framework system for leading new global trade rules through regional trade agreements such as the Trans-Pacific Partnership Agreement (TPP), the Transatlantic Trade and Investment Partnership (TTIP), and the United States–Mexico–Canada Agreement (USMCA), as well as the European Union through TTIP, the EU–Canada FTA, and the EU–Japan FTA. However, there is still controversy over whether regional economic integration is a stepping stone or a stumbling block to multilateral trade liberalization. Bhagwati even believed that FTA was a cancer of free trade [14]. Much literature has also begun to study the intra and intra-domain effects of FTA trade rules [15,16]. Dai et al. [15] found that FTA has a significant trade transfer effect, as it not only transfers trade from non-member countries but also from within member countries. Anderson et al. [17] estimated the impact of the FTA agreement on manufacturing-level trade from 1990 to 2002 and found that some countries gained over 5% of profits, while others lost less than 0.3% of profits, resulting in a 0.9% increase in global manufacturing trade efficiency. Baier and Bergstrand [18] found that the FTA agreement doubled bilateral trade among member countries 10 years later. Xi L. [16] found that regional service trade agreements can effectively drive the growth of member countries’ service exports. In addition to the aforementioned trade transfer and trade creation effects, the FTA agreement also has a transmission effect, meaning that a country signing the FTA agreement will incentive it to sign new agreements with other countries [19].

2.2. Research on Regional Digital Trade Rules

Various national and regional digital trade regulations have been developed as a result of the WTO’s international rules in this area being insufficient. Digital trade rules refer to the inclusion of digital trade (e-commerce) rules or provisions in global, regional, and bilateral trade agreements. Geiger and Schotten’s study presented a global overview of bilateral digital trade agreements, revealing the scope and characteristics of agreements between different countries [20]. Reimer and Farley focused their research on digital trade regulations in regional trade agreements, analyzing recent trends [21]. National and regional digital trade rules also lead to heterogeneity and differentiation of rules, resulting in different rules such as the “American template”, “European template”, “Chinese template”, and “South Pacific template”. Discriminatory and uncoordinated issues may arise between different rules, which is also one of the reasons for the emergence of digital trade barriers [22]. The research on regional and bilateral digital trade rules mainly includes two major fields: textual research and quantitative research [23]. The research on rule texts focuses on analyzing the current status of rules [24]; analyzing key issues [25]; and organizing and analyzing specific chapter texts of FTA, such as “digital trade” [26]. The quantitative analysis mainly involves measuring the regulatory intensity and degree of heterogeneity of digital trade rules. The OECD has constructed a Digital Services Trade Restriction Index (DSTRI) to measure the degree of restrictions on digital trade rules in various countries [27,28]. The European Centre for International Politics and Economics (ECIPE) has also constructed a Digital Trade Restriction Index (DTRI) to measure the degree of restrictions on digital trade rules. In terms of measuring the degree of heterogeneity in bilateral digital trade rules, the OECD has also constructed a Digital Services Trade Regulatory Heterogeneity Index (DSTRIH) [29].

2.3. Research on the Trade Effects of Digital Trade Rules

With the continuous deepening of research on digital trade rules, scholars have also begun to study the trade effects generated by digital trade rules. The relevant research mainly focuses on the impact of digital trade rules on the export of digital services and trade costs. Duval et al. examined the impact of digital trade facilitation clauses on trade costs. The study believes that the implementation of digital trade facilitation clauses can reduce the trade costs of countries in the Asia Pacific region by more than 26% [30]. Mitchell and Mishra [31] believe that certain provisions in RTAs (especially the e-commerce chapter) can promote digital trade integration: provisions on non-discrimination and prohibition of data localization reduce digital trade barriers; the regulations on paperless transactions and electronic signatures can promote electronic transactions required for digital trade; the domestic regulatory framework for digital trade is supported by regulations on data protection, consumer protection, intellectual property, and competition in the digital industry. Sun et al. [32] found through research on the impact of digital trade rules in RTA on service trade exports that it significantly promotes service trade among contracting parties and has a stronger promoting effect on knowledge-intensive service trade between developed and developing economies. Li Y. [33] constructed an indicator of the depth of digital trade rules and found that the improvement of the depth of digital trade rules, as well as the four types of clauses: market access, trade facilitation, consumer protection, and dispute resolution, has a significant promoting effect on the total exports among members. The depth of service trade rules has a significant positive moderating effect on the relationship between digital trade rules and value chain trade. Yu et al. constructed a clause heterogeneity index based on the rule attributes of different types of digital trade clauses in regional trade agreements. This study found that the great improvement of RTA digital trade rules significantly promotes the export of digital services by participating countries, and different types of clauses have heterogeneous effects [34].

In summary, the types of FTA and digital trade rules, and the impacts of digital trade rules, have been the subject of much research by academicians. However, existing research has not conducted systematic research on the impact of bilateral digital service trade rules on digital service exports. Compared with existing research, the innovation of this article has two aspects: Firstly, it reveals the heterogeneous impact of bilateral digital trade rules on digital service exports of countries (regions) with different service sectors and economic development levels. Secondly, it reveals the intermediary channels through which bilateral digital trade rules generate the promotion effect of digital service exports. This study provides sufficient empirical evidence support for a deep understanding and promotion of the construction of bilateral digital trade rules and provides a useful reference for China to promote institutional openness in the field of digital service trade.

3. Theoretical Mechanism and Research Hypotheses

3.1. Trade Creation Effect

Viner [35] first proposed the concept of the trade creation effect, which refers to the cancellation of tariff and non-tariff barriers between signatory countries after signing the FTA, expanding the scale of bilateral trade and improving welfare levels. After the signing of bilateral or regional FTA agreements, the FTA network forms a “circle of friends” effect, which specifically includes institutional improvement and innovation effects, thereby promoting exports [36]. The digitization of trade has greatly increased the scale of global trade [37,38]. The development of the Internet, information technology, and the deepening of network popularization have also provided a more convenient environment and higher transaction efficiency for digital service trade, improving the development level of digital service trade. Qi and Qiang [8] found that the FTA digital trade rules have a fusion effect, which can alleviate the adverse impact of bilateral regulatory heterogeneity on digital service exports and generate trade promotion. In particular, when the depth level of the FTA digital trade rules and provisions increases, the trade promotion effect brought by the regulatory integration effect is stronger. Therefore, we propose Hypothesis 1:

Hypothesis 1.

The bilateral digital trade rules have had a significant promoting effect on the digital service exports of member countries.

3.2. Intermediary Effects of Trade Costs

Trade costs refer to all costs that must be paid to obtain goods, including search costs, transportation costs, coordination costs, and contract costs, in addition to the cost of producing goods. At present, the trade cost measurement method of Anderson and Wincoop is mostly used in academic circles to measure the trade cost [39]. This method improves the traditional gravity model, adds multilateral trade costs and other factors of both trade sides, and measures bilateral trade costs from the perspective of economic scale and relative trade costs. Novy [40] further improved the gravity model, set the trade cost to be jointly determined by domestic trade and bilateral trade, and divided tradable goods and non-tradable goods. Much literature has studied the relationship between FTA and trade costs [41,42,43]. The signing of the FTA can help overcome the conversion costs caused by differences in trade rules, increase information transparency, and effectively reduce the fixed and variable trade costs paid by enterprises for exports [42]. Zhao and Du [43] found that the decrease in bilateral trade costs is one of the mediating variables for FTA to promote the growth of Chinese enterprises’ export trade. Similarly, the signing of digital trade agreements promotes bilateral digital service trade by reducing trade costs, reducing restrictions on the free flow of cross-border data, and reducing communication and search costs. The electronic commerce chapter clauses in digital trade agreements can promote the integration of digital trade, reduce transaction costs, and reduce digital trade barriers through provisions on non-discrimination and prohibition of data localization. The regulations on paperless transactions and electronic signatures can promote electronic transactions required for digital trade. Thus, we propose hypothesis 2:

Hypothesis 2.

Digital trade rules promote the growth of digital service exports by reducing trade costs.

3.3. Heterogeneity Effect of Digital Trade

At present, digital service trade is developing rapidly, and the trade effects of different digital trade departments on signing digital trade agreements have heterogeneity [37]. According to the definition of digital service trade USBEA [44], digital service trade is divided into six categories: “financial services”; “telecommunications, computer and information services”; “intellectual property fees”; “insurance and pension services”; “Personal, cultural and entertainment services”; and “other business services”. Qi and Li [45] distinguished digital services into three categories for heterogeneity analysis: digital interactive services, digital empowering basic services, and data element-driven services when analyzing the value chain effect of heterogeneity in digital service trade regulation. Zhou and Li divided digital intellectual property into five subcategories for research: “Digital Content Copyright Protection”, “Non-Mandatory Localization of Source Code”, “Trade Secret Protection in Computers”, “Electronic Trademark System”, and “Responsibility of Network Service Providers” [9]. Therefore, we propose hypothesis 3:

Hypothesis 3.

The impact of digital trade rules on the export of digital services is heterogeneous.

The 2022 World Trade Report pointed out that the service trade reached USD 1.6 trillion, slightly higher than the level before the COVID-19 epidemic. The export of middle-income countries increased by about 30% compared with the same period in 2020, while the export of high-income countries increased by 15%. Compared with high-income countries, the service trade of middle-income countries developed more rapidly and had great potential. The greater the difference in the level of economic development between economies, the more likely it is to generate differences in digital trade rules. Countries with different levels of economic development have different demands on digital trade, and the impact of signing digital trade rules in RTAs on digital service trade varies. Therefore, we propose hypothesis 4:

Hypothesis 4.

Compared to “high-income countries”, the establishment of digital trade rules brings greater benefits to the digital services trade of low- and middle-income countries.

4. Variable Setting and Model Construction

4.1. Model Construction

In order to study whether the establishment of digital trade rules can help promote the export of digital services, based on the TAPED database and the WTO database, this article uses 143 countries from 2005 to 2019 as research samples. In view of the different time points when countries and different trading partners formulate digital trade rules, in order to study the impact of digital trade rules on the export of digital services, this article constructs a multi-period DID model. The corresponding panel measurement model is shown in Equation (1).

$$\ln Y_i{}_{jt}={\beta }_0+{\beta }_{1}Digita{l}_{ijt}\times tim{e}_{ijt}+{\beta }_2{Z}_{ijt}+u_j+u_t+{\epsilon }_{ijt}$$

(1)

Among them, subscripts represent the country (region) and trading partner country (region), respectively, representing time. The explanatory variable is the logarithm of bilateral export trade flows between economies in the digital trade sector during the period.

Based on the definition of digital trade as defined by USBEA [44] (“potential digitizable service trade PICTE”), and comparing the service sector classification of the OECD bilateral service trade database, this article includes the following OECD service trade sectors into the statistical category of digital trade flows, which are classified as “financial services”; “telecommunications, computer and information services”; “intellectual property fees”; “insurance and pension services”; “Personal, cultural and entertainment services”; and “other business services”; Digital_ijt is a virtual variable of a free trade agreement, that is, a policy processing variable; time_ijt is a dummy variable during the policy implementation period; Z_ijt is a set of control variables, including the per capita GDP of signatory countries (LnGDP_i, LnGDP_j) converted in 2015 constant US dollars, weighted distance (LnDist), whether it is bordering on each other (Contig), whether it shares the official common language (Comlang), whether it has a colonial relationship (Colony) and Internet penetration (Int_i, Int_j); and u_j, u_t and ε_ijt represent individual fixed effects, time fixed effects, and residuals, respectively.

4.2. Variable Interpretation

4.2.1. Explained Variable

The explanatory variable of this empirical model is the bilateral digital service exports of signatory countries. The data is sourced from the WTO database, with a time of 2005–2019 and a total of 15 years of sample period. LnY_ijt is the logarithm of bilateral export trade flows between economies i and j in the digital trade sector during the period t. Based on the definition of digital trade as defined by USBEA [44], (“potential digitizable service trade PICTE”) and comparing the service sector classification of the OECD bilateral service trade database, this article includes the following OECD service trade sectors into the statistical category of digital trade flows, which are classified as “financial services”; “telecommunications, computer and information services”; “intellectual property fees”; “insurance and pension services”; “Personal, cultural, and entertainment services”; and “other business services”.

4.2.2. Core Explanatory Variables

The TAPED database developed by Mira Burri and Rodrigo Polanco of the University of Lucerne School of Law is the most detailed and complete database for quantifying the text of digital trade rules in regional trade agreements. There are 123 digital trade clauses involved in this article, and 20 have not been signed. The signed years and countries are shown in

Table 1. Countries signing digital trade agreements in various years.

Time	Country
2005	UAE, Bahrain, Kuwait, Oman, Saudi Arabia, Qatar, Macedonia, Croatia, Algeria, Denmark, Norway, Portugal, Austria, Bulgaria, Latvia, Lithuania, Indonesia, Romania, Ukraine, Pakistan, Colombia, Nicaragua, Venezuela, Argentina, Paraguay, Tunisia
2006	United States, Japan, Nicaragua, Panama, Lebanon, China, Morocco, Bolivia
2007	Albania, Sweden, Chile, Kosovo, Dominica, Syria, Egypt
2008	New Zealand, El Salvador, Guatemala, Honduras, Bosnia, Montenegro, Afghanistan, Azerbaijan, Iran, Kazakhstan, Kyrgyzstan, Pakistan, Tajikistan, Turkmenistan, Uzbekistan
2009	Singapore, Canada, Switzerland, Australia, Peru
2010	Türkiye, Malaysia, Thailand, Philippines, Brunei, Vietnam, Laos, Myanmar, Cambodia, India
2011	France, Italy, Netherlands, Belgium, Germany, United Kingdom, Ireland, Greece, Spain, Estonia, Finland, Malta, Cyprus, Poland, Slovakia, Slovenia, Czech Republic
2012	Jordan, South Korea, Comoros, Madagascar, Mauritius, Seychelles, Zambia, Zimbabwe, Hong Kong
2013	Costa Rica, Serbia, Mexico
2014	Moldova, Cameroon, Iceland
2015	Armenia, Belarus, Russia
2016	Cote d’Ivoire, Georgia, Mongolia, South Africa, Angola, Botswana, Lesotho, Malawi, Mozambique, Namibia, Eswatini, Tanzania, Congo, Ghana
2018	Uruguay, Sri Lanka
Not signed	North Korea, Nepal, Bhutan, Bangladesh, Maldives, Iraq, Israel, Hungary, Libya, Brazil, Sudan, Ethiopia, Somalia, Djibouti, Kenya, Uganda, Rwanda, Chad, Guinea, Jamaica

Source: The author compiled the data based on the TAPED database. https://www.unilu.ch (accessed on 12 April 2023).

. Digital and time are policy processing variables and dummy variables during policy implementation, respectively. If both parties have signed digital trade rules, the value of Digital is 1; otherwise, it is 0. If the digital trade rule signed by both parties takes effect in year t, the value of time after the rule takes effect is 1, and the value before it takes effect is 0.

4.2.3. Control Variables

By combining the research on digital trade by Liu and Chen [46,47], the control variables selected in this article include:

(a)

Domestic and trading partner countries’ gross domestic product (LnGDP_i, LnGDP_j): These data are converted using the constant price of US dollars in 2015 and are in logarithmic form. GDP is an essential key variable in the gravity model. Therefore, this article introduces the GDP of domestic and trading partners as the control variable. The data are sourced from the World Bank database. These two variables, respectively, represent the level of economic development of import and export economies.

(b)

Potential factors that affect the cost of trade between economies: Weighted distance (LnDist) is an indicator that does not change at any time, representing the weighted distance (in kilometers) between the export economy i and the import economy j, expressed in logarithmic form. It is considered a trade cost in the traditional gravitational model. As the distance increases, the cost of trade will rise, which may have a negative impact on trade. The indicator data are from the CEPII database.

(c)

Whether it borders (Contig): Whether a country borders its trading partners will have a significant impact on digital trade. Whether the two countries are contiguous not only represents the spatial distance between them but also reflects the cultural distance. Referring to the practice of González and Ferencz [37],, this article sets the value as 1 if the two countries border; otherwise, it is 0.

(d)

Whether there is a common language (Comlang): Common language can reduce the communication costs of micro participants in digital trade, such as enterprises and users, thereby promoting the development of digital trade. If the common official languages of the two countries are the same, the value is 1; otherwise, it is 0.

(e)

Whether there is a colonial relationship (Colony): The colonial relationship can reflect the institutional distance between the two countries, and the existence of a colonial relationship between the two countries may affect the development of digital trade between countries. If the two countries have a colonial relationship, the value is 1; otherwise, it is 0. All the above control variables belong to the gravitational model variables, and the data are sourced from the CEPII Gravity database.

(f)

Internet level of domestic and trading partner countries (Int_i Int_j): The level of the Internet may affect digital trade through the transmission media that affect it. This article refers to the practice ofGonzález and Ferencz [37], adding the Internet level of the country and trading partner countries as a control variable. The selected indicator for Internet level is “the percentage of individual users using the Internet”, with data from the International Telecommunication Union.

4.3. Descriptive Statistics

The descriptive statistics of the main variable are shown in

Table 2. Descriptive statistics.

Variable	N	Mean	Standard Deviation	Min	Max
Digital	2145	0.860	0.347	0	1
time	2145	0.595	0.491	0	1
LnY	2140	2.629	2.285	0	8.570
LnGDPi	2081	8.758	1.388	5.667	11.63
LnGDPj	2115	9.670	1.175	6.710	11.63
Lndist	2070	8.224	1.072	5.050	9.856
Contig	2145	0.154	0.361	0	1
Comlang	2145	0.273	0.445	0	1
Colony	2130	0.106	0.307	0	1
Inti	2072	43.01	30.15	0.0652	99.70
Intj	2137	61.68	27.55	0.238	99.70

Source: TAPED database. https://www.unilu.ch (accessed on 12 April 2023), WTO database. https://stats.wto.org/en (accessed on 12 April 2023), The World Bank database. https://data.worldbank.org.cn/ (accessed on 12 April 2023), The CEPII database. http://cepii.fr/CEPII/en/bdd_modele/bdd_modele.asp (accessed on 12 April 2023), The International Telecommunication Union. https://www.itu.int (accessed on 12 April 2023).

.

4.4. Applicability Test of Double Difference Method

In order to test the applicability of the difference-in-differences model (DID), this article needs to confirm whether the control group in the DID model is the counterfactual of the experimental combination, that is, whether the parallel trend assumption is valid. This article takes countries (regions) that signed digital trade rules during 2005–2019 as an experimental group, including Norway, Japan, China, Portugal, the United States, Switzerland, and the United Kingdom. Countries (regions) that did not sign digital trade rules during 2005–2019 as a control group, including Nepal, Maldives, Brazil, Jamaica, etc., to investigate the differences in digital services exports between the two.

To more intuitively reflect the changes in the export volume of digital services in the experimental group and the control group during the sample period, this paper uses curves to depict the changing trend of the average digital services exports in the experimental group and the control group, as shown in Figure 1. After 2005, the export volume of digital services trade in the experimental group and the control group showed an overall upward trend, and the growth trend of the export volume of digital services in the experimental group was significantly stronger than that in the control group. Parallelism assumes that there is no systematic difference in the export volume of digital services between countries that have signed digital trade rules and countries that have not signed digital trade rules, or even if there is a difference, the difference is fixed, that is, the development trend of the two countries is consistent.

5. Model Regression

5.1. Staggered DID Benchmark Regression

The benchmark regression results are shown in

Table 3. Staggered DID benchmark regression.

Variable	(1)	(2)	(3)	(4)
Digital × time	1.310 *** (13.00)	0.620 *** (6.62)	0.697 *** (7.33)	0.576 *** (6.38)
LnGDPi		0.928 *** (20.84)	0.774 *** (10.81)	0.784 *** (10.50)
LnGDPj		−0.264 ** (−2.26)	−0.279 ** (−2.44)	0.270 ** (2.09)
Inti			0.008 *** (2.67)	0.012 *** (3.98)
Intj			0.015 *** (3.98)	0.011 *** (3.30)
Lndist				0.182 *** (2.92)
Contig				2.012 *** (9.79)
Comlang				−0.936 *** (−5.30)
Colony				−0.047 (−0.26)
Constant	1.849 *** (26.18)	−3.240 *** (−2.72)	−3.061 ** (−2.49)	−9.851 *** (−6.96)
Fe	Control	Control	Control	Control
Te	Control	Control	Control	Control
observed value	2140	2049	1993	1948
R2	0.567	0.695	0.702	0.727

Note: **, and *** represent significant levels of 5% and 1%, respectively, with t-values in parentheses.

. Columns (1)–(3) are the regression results after successively adding various control variables. Column (1) shows that the establishment of digital trade rules has significantly promoted the export of digital services by signatory countries and trading partner countries without adding control variables. The regression results in columns (2) to (3) further illustrate that after adding various control variables, the positive effects of signing digital trade agreements on digital trade between signatory countries and trading partners are still significant, indicating the importance of digital trade rules.

5.2. Endogeneity Issue

Instrumental variable method: In order to solve the endogenous problem caused by the possible omission of important explanatory variables, this article treats the core explanatory variable with a lag of one period as an instrumental variable. Column (1) of

Table 4. Endogeneity and robustness analysis.

Variable	(1)	(2)	(3)
Digital × time	0.681 *** (6.22)	0.599 *** (3.14)	0.576 *** (6.38)
Digital × time × WTO			0.592 *** (6.33)
Anderson canon. corr. LM statistic	0.000
Cragg–Donald Wald F statistic	2912.414 [16.38]
Control variable	Control	Control	Control
Fe	Control	Control	Control
Te	Control	Control	Control
Observed value	1819	522	1948
Centered R2	0.727
R2		0.726	0.727

*** represent significant levels of 1%, with t-values in parentheses.

shows the regression results of the instrumental variable method. The P value of the Anderson canon. corr. LM statistic is less than 0.1, rejecting the assumption that the instrumental variable is not sufficiently identified. In addition, the Cragg–Donald Wald F statistics are all greater than the critical value of 16.38, rejecting the original assumption of weak instrumental variables. The above test indicates that the research in this article is reliable and also means that the regression results of the original explanatory variable and the instrumental variable are similar, and the endogenous problem of the model is not obvious.

5.3. Robust Test

5.3.1. Randomly Sampled Placebo Test

In order to further eliminate the impact of possible missing variables on the validity of experimental results, an effective solution is to adopt a placebo test. If there are systematic deviations and other missing factors in the regression model in this article, the results obtained through random sampling are similar to the benchmark regression results; otherwise, it indicates that the benchmark regression results in this article are true and effective. Based on this, this article conducted a placebo test using a random sampling method. First, the time of the policy impact was set randomly from the sample period but maintained a time gap of at least one year before and after the implementation of the policy; that is, the time of policy implementation was set randomly from 2005 to 2019. A virtual multiple difference estimator f_Digital × f_time was then obtained, and the regression coefficient of the estimator β was extracted. The above process was randomly performed 500 times to obtain β. Regarding the estimated kernel density (as shown in Figure 2), where the vertical line is the true multiplier coefficient α 1, the kernel density distribution is quite different from that of the virtual regression coefficient, which can exclude the possibility that the regression results are affected by other unobservable factors. It can be seen that the regression coefficients follow the normal distribution and are concentrated near the zero point, and most of the regression coefficients have P values greater than 0.1. In summary, the placebo test shows that the benchmark regression in this article is true and effective.

5.3.2. Sample Interval Division Problem

After both parties formulate digital trade rules, the manifestation of their policy effects often takes a period of time. Considering this and taking into account the issue of sample utilization, this article used the method of Lin and Bao [48] to regress the sample using a sample interval division method with an interval of 3 years; that is, the sample time dimensions are 2005, 2009, 2013, and 2017, as shown in column (2) of Table 4. After dividing the interval for regression, the size of the regression coefficient of the core explanatory variable has improved and remains significant. This indicates that the promotion effect of signing digital trade rules on the volume of digital service trade is not affected by the sample interval division method, confirming the robustness of the results.

5.3.3. The Impact of the WTO

Whether a country (region) is a member of the WTO determines whether it applies the most favored nation tariff treatment of the WTO, which, to some extent, interferes with the effect of formulating digital trade rules on the export of digital services. The difference-in-differences model estimation method does not eliminate the impact of policies other than digital trade rules, such as whether it is a WTO member, which may lead to bias in the estimation results. Therefore, this article uses the triple difference method to conduct triple difference regression by adding the dummy variable, whether the trading partner is a WTO member and the cross product of Digital, and the WTO dummy variable to the benchmark regression model. Column (3) of Table 4 shows that after excluding the impact of WTO on the effects of establishing digital trade rules on digital services exports, the regression coefficient of digital services exports on whether to sign digital trade rules is still significantly positive, indicating that the benchmark regression results are robust.

6. Heterogeneity Analysis

6.1. Differences in Categories of Digital Trade

According to the definition of digital service trade (USBEA) [44], this section divides digital service trade into six categories: “insurance and pension services”; “finance services”; “charges for the use of intellectual property”; “telecommunications, computer, and information services”; “other business services”; and “personal, cultural, and recreational services” to conduct empirical testing. Columns (1) to (6) in

Table 5. Heterogeneity of digital trade subsectors.

Variable	(1) Insurance	(2) Finance	(3) Intellectual Property	(4) Telecommunications	(5) Other Business Services	(6) Personal, Cultural, and Recreational Services
Digital × time	0.633 *** (8.63)	0.646 *** (9.25)	0.271 *** (3.44)	0.500 *** (6.31)	0.641 *** (6.97)	0.241 *** (4.42)
Control variable	Control	Control	Control	Control	Control	Control
Fe	Control	Control	Control	Control	Control	Control
Te	Control	Control	Control	Control	Control	Control
Observed value	1948	1948	1948	1948	1948	1948
R2	0.667	0.750	0.730	0.697	0.750	0.583

*** represent significant levels of 1%, with t-values in parentheses.

, respectively, present the estimated results of the impact of establishing digital trade rules under the six categories of digital trade mentioned above. Different sectors in service trade have a highly differentiated nature, so will the same development of digital trade rules have a differentiated impact on these six sectors? The results show that there are significant differences in the effects of digital trade rules on different sectors of digital trade.

As shown in Table 5, digital trade rules have a positive and significant promotion effect on “insurance and pension services”; “finance services”; “charges for the use of intellectual property”; “telecommunications, computer, and information services”; “other business services”; and “personal, cultural, and recreational services”. Among them, the coefficients of “other business services”; “finance services”; “insurance and pension services”; and “telecommunications, computer, and information services” are larger, and the impact is more significant. The coefficients of the “charges for the use of intellectual property” and “personal, cultural, and recreational services” sectors are smaller, and the effect is smaller than the above four departments. The possible reason for the above heterogeneity lies in the differences in the sensitivity of various types of digital services to digital trade rules. In summary, the main reasons for these differentiation phenomena are as follows:

After the establishment of digital trade rules, the volume of trade in digital services in the “finance services”; “other business services”; “insurance and pension services”; and “telecommunications, computer, and information services” sectors has significantly increased. In particular, research and advisory services in “other business services” often have to overcome the challenges posed by geographical barriers in traditional trade to provide face-to-face services. After the establishment of digital trade rules, while ensuring the flow of data, service providers can reach more potential customers based on online platforms, provide insurance and consulting services, and expand their overseas markets. The “telecommunications, computer, and information services” sector itself is an emerging form of trade, which is relatively sensitive, and trading partners are more susceptible to the impact of digital trade rules, thereby increasing the volume of trade. With the development of artificial intelligence, 5G, and blockchain, the digital delivery capabilities of the “finance services” and “insurance and pension services” sectors have gradually increased. Digital technology and digital services have spawned a large number of new trade formats and models, injecting new impetus into the economic growth of the “finance services” and “insurance and pension services” sectors and promoting the development of traditional financial industries and emerging financial technologies.

Due to differences in the degree of protection of “charges for the use of intellectual property” among countries, in some countries without strict intellectual property protection, digital trade rules have little impact on relevant industries in the long run, and individual differences in the “personal, cultural, and recreational services” sector are large, with consumption habits varying from country to country, resulting in a small coefficient of impact.

6.2. Country Income Heterogeneity

The impact of digital trade rules under the condition of distinguishing income heterogeneity among countries was analyzed. According to the World Bank’s 2018 [49] income classification criteria, countries that establish digital trade rules are divided into high–middle-income countries and low-income countries to explore the impact of establishing digital trade rules among countries with different income levels on digital services trade. The regression results after group matching are shown in

Table 6. Country income heterogeneity survey.

Variable	(1)	(2)	(3)	(4)
Exporting country	high-income countries	high-income countries	Low- and middle-income countries	Low- and middle-income countries
Importing country	high-income countries	Low- and middle-income countries	high-income countries	Low- and middle-income countries
Digital × time	−0.362 ** (−2.22)	−0.232 (−1.30)	0.562 *** (4.98)	0.303 *** (2.72)
Control variable	Control	control	control	control
Fe	Control	control	control	control
Te	Control	control	control	control
Observed value	494	225	683	546
R2	0.845	0.813	0.774	0.835

**, and *** represent significant levels of 5%, and 1%, respectively, with t-values in parentheses.

.

From Table 6, it can be found that under the differentiation of country income heterogeneity, the difference in the impact effect categories of digital trade rules is more obvious. The group where the exporting country is a low- and middle-income country, the importing country is a high-income country, and both low- and middle-income countries have a significant positive impact at the statistical level of 1%. When both countries are high-income countries, there is a significant negative impact. The group of exporting countries as high-income countries and importing countries as low- and middle-income countries are not significant. This result indicates that in the grouping of country income heterogeneity, when the exporting country is a low- and middle-income country, the establishment of digital trade rules can better promote trade between the two countries, indicating that middle-income countries should pay attention to the positive effects of digital trade rules, deeply participate in international cooperation such as the digital economy, strengthen cooperation with allies, and face the free trade circle of friends with a more positive attitude. When both countries are high-income countries, there is a negative impact. When the exporting country is a high-income country and an importing country is a group of low- and middle-income countries, the impact is not significant. High-income countries should adopt new paths and models when participating in regional and global digital trade.

7. Mechanism Analysis Based on Trade Costs

This section conducts a mechanism test based on trade costs. Before testing the mechanism, the cost of trade is first measured. By using the methods of Novy D. [40] for reference, the trade cost calculation formula was constructed as follows:

$$COS{T}_{ijt}={\left(\frac{x_{ii}{x}_{jj}}{x_{ij}{x}_{ji}}\right)}^{\frac{1}{2\left(\sigma -1\right)}}-1$$

(2)

where COST_ijt represents the trade cost of exports from country i to country j in year t. x_ij represents the total amount of digital trade exported by country i to country j, and x_ji represents the total amount of digital trade exported by country j to country i. x_ii and x_jj represent the total amount of domestic digital trade in countries i and j, respectively, and are obtained by subtracting the total amount of exports from the total amount of digital trade in that country. Since the total amount of digital trade in various countries cannot be directly obtained, this article follows the assumption of Anderson J.E. [39] and Novy D. [40] that only 80% of the total amount of industry trade belongs to the tradable share, and infers the corresponding total amount of digital trade. The elasticity in Equation (2) is set to 7 according to the method of Anderson J.E. [39].

Table 7. Mechanism analysis of trade costs.

Variable	(1)	(2)	(3)
	A (2005–2008)	B (2009–2012)	C (2005–2019)
Digital × time	−0.007 * (−1.87)	−0.151 *** (−3.27)	−0.123 *** (−4.93)
Control variable	Control	Control	control
Fe	Control	Control	control
Te	Control	Control	control
Observed value	405	434	1621
R2	0.295	0.323	0.251

* and *** represent significant levels of 5% and 1%, respectively, with t-values in parentheses.

shows the impact of digital trade rules on trade costs over three periods in terms of time dimension. Sample A uses a sample from 2005 to 2008, which is in the early stages of digital trade rulemaking. Regression results show that digital trade rules have had a significant negative impact on trade costs during this period. Sample B uses the sample from 2009 to 2012. As more and more digital trade rules are formulated, the impact of signing digital trade rules on trade costs is becoming increasingly significant, and the absolute value of the coefficient begins to increase. From a longer time interval, using the 15-year difference between 2005 and 2019, the regression result for sample C is consistent with the previous two stages; that is, digital trade rules have a significant negative impact on trade costs, and the impact is a significant in different time intervals.

8. Discussion

As a result of digital trade, which has significantly enhanced the breadth and depth of international trade and acted as a potent driver of global economic growth and recovery, the globalization development model has undergone significant change. As a result, digital trade rules have garnered considerable academic attention. However, the regulations governing digital trade under these templates are mostly at the national or regional level, leading to an “isolated island” and “overlapping” fragmentation pattern that, in some ways, even serves as the foundation for barriers to digital trade [7,8]. Few empirical studies have been conducted on the topic of coordinating bilateral digital trade rules to regulate the growth of this industry [9]. As a fundamental institutional arrangement for digital trade, bilateral digital trade rules may play an important role in regulating the development of digital trade. Many researchers address the type of free trade agreement (FTA) and digital trade rules; however, research on the impact of bilateral digital trade rules on digital trade is still lacking. In order to address this gap, this article aims to utilize a double difference model and draw on the TAPED database from the University of Lucerne [12], which covers digital service trade rules, to investigate the influence of bilateral digital service trade rules on digital service exports, thus providing empirical evidence for the comprehensive promotion of bilateral digital service trade rules.

The empirical results show that the establishment of bilateral digital trade rules significantly promotes the export of digital services between the two sides. This finding is consistent with research [30,31,32,33,34,42] that demonstrates the benefits of bilateral digital trade rules for increasing the export of digital services. Among these, Duval et al. [30] looked at the effect of trade cost on digital trade facilitation. According to the study, implementing digital trade facilitation clauses might lower trade costs for countries in the Asia Pacific by more than 26 percent.

According to Mitchell et al. [31], certain features in regional trade agreements (RTAs) might encourage the integration of digital trade: they lower obstacles to digital trade by prohibiting data localization and prohibiting discrimination; regulations on electronic signatures and paperless transactions may encourage the use of these technologies in digital trade; regulations on data protection, consumer protection, intellectual property, and competition in the digital economy support the domestic regulatory framework for digital trade. Sun [32] discovered through a study on the effects of digital trade regulations in RTA on service trade exports that it considerably promotes service trade among contracting parties and has a stronger promotional effect on knowledge-intensive service trade between developed and developing economies. Li [33] constructed an indicator of the depth of digital trade rules and found that the improvement of the depth of digital trade rules, as well as the four types of clauses: market access, trade facilitation, consumer protection, and dispute resolution, has a significant promoting effect on the total exports among members. Peng [34] constructed a clause heterogeneity index based on the rule attributes of different types of digital trade clauses in regional trade agreements. This study found that the great improvement of RTA digital trade rules significantly promotes the export of digital services by participating countries, and different types of clauses have heterogeneous effects. The result of mechanism analysis shows that digital trade rules reduce trade costs and boost the export of service trade. Establishing digital trade rules can significantly reduce the cost of digital trade for both sides of the trade. We need to provide a reference for this. As FTAs can effectively lower the fixed and variable trade expenses that businesses pay for exports by reducing conversion costs brought on by disparities in trade laws [42]. Zhao and Du [43] claim that FTA lowers the expenses of bilateral trade while fostering the expansion of Chinese exports to China. Similar to this, the signing of digital trade agreements encourages bilateral digital service trade by lowering trade costs, lowering barriers to the free flow of cross-border data, and lowering communication and search costs, which leads to the expansion of export of service trade. Our results confirm this and show that digital trade regulations such as the FTAs reduce trade costs and enhance the export of service trade.

The digital trade rules have a positive and significant promoting effect on “insurance and pension services”; “finance services”; “charges for the use of intellectual property”; “telecommunications, computer, and information services”; “other business services”; and “personal, cultural, and recreational services”, while the coefficients in the “charges for the use of intellectual property” and “personal, cultural, and recreational services” sectors are smaller, but still significant. This supports the notion of González and Ferencz that the trade effects of different digital trade departments on signing digital trade agreements have heterogeneity [44]. The findings demonstrate that compared to “high-income countries”, the establishment of digital trade rules brings greater benefits to the digital services trade of low- and middle-income countries. This can be related to the reason that the service trade of low- and middle-income countries that contains great potential developed more rapidly than the developed countries [50]. As a result of signing digital trade rules, low-income and middle-income countries may receive access to more markets (they find purchasers around the globe when they become part of the online digital economy) [51]. Elms give the greater potential of the availability of services to export to the untapped markets. The service sector in Asia has the potential to become a new engine of economic growth for developing Asia, which has traditionally relied on export-oriented manufacturing to power its growth [51]. Konstantakopoulou investigated the static and dynamic relationship between exports and economic growth in the southern Euro-zone countries, and the findings indicate that bidirectional Granger causality is predominant in Spain and Greece. Unidirectional causality from exports to economic growth is found in Portugal [52].

9. Conclusions, Limitations, and Future Research

This article analyzes whether the establishment of bilateral digital trade rules can help promote the export of digital services from both sides, demonstrating its importance. It also empirically tests the theoretical proposition using panel data on digital trade from 143 economies worldwide from 2005 to 2019. The empirical results show that the establishment of bilateral digital trade rules significantly promotes the export of digital services between the two sides. The heterogeneity test found that bilateral digital trade rules have a positive and significant promoting effect on “insurance and pension services”; “finance services”; “charges for the use of intellectual property”; “telecommunications, computer, and information services”; “other business services”; and “personal, cultural, and recreational services”, while the coefficients in the “charges for the use of intellectual property” and “personal, cultural, and recreational services” sectors are smaller, but still significant. At the country level, when the exporting country is a low- and middle-income country, digital trade rules have a significant positive impact on both sides of the trade in digital services exports. When both countries are high-income countries, there is a significant negative impact. The group of exporting countries as high-income countries and importing countries as low- and middle-income countries is not significant. The result of mechanism analysis shows that trade cost is an intermediary variable between digital trade rules and the export of digital services, which can play a partial intermediary effect. Establishing digital trade rules can significantly reduce the cost of digital trade for both sides of the trade. This article adds to the literature on the connection between digital trade regulations and the service exports trade and provides practical implications.

Given the above conclusions, this article suggests the following: First, after joining the Digital Economy Partnership Agreement, developing countries, particularly China, should actively participate in and promote the signing of digital trade rules, explore new ways of cooperation in the digital economy, further expand open cooperation in the digital economy, and strengthen cooperation between all parties in this field. Secondly, the formulation of digital trade rules in different digital trade sectors presents differentiated trade effects. For the two sectors with less impact, namely “charges for the use of intellectual property” and “personal, cultural, and recreational services”, developing countries, including China, should give priority to formulating and improving intellectual property protection measures, promoting innovation, and escorting the development of digital trade in intellectual property related sectors. Thirdly, economies with different levels of economic development have different sensitivities to digital trade rules. When an economy China participates in foreign economic cooperation, it should consider relevant paths and models based on the economic development and position demands of different economies. Finally, China and other economies should accelerate the construction of domestic legal and regulatory governance capabilities in the field of digital trade, improve the domestic institutional environment, promote the smooth implementation of digital trade rules, and thus safeguard the rapid development of digital trade.

This study also has some limitations, which present new research perspectives for other researchers in the same field. Firstly, the data in this study only collected the export volume of digital services from the WTO database, and some countries with smaller economies did not have relevant data. This raises the issue of the scalability of this study. Therefore, future research can include more country samples in the data set to obtain broader research conclusions. Secondly, the selected six categories of digital trade represent a proxy variable for digital trade. Personal, cultural, and recreational services may include all types of economic activities rather than just digital ones. Therefore, in future research, this classification can be further refined to explore the impact of digital trade rules on digital service export better.