1. Introduction
Digital transformation is progressing through the expansion of digital trade and the spread of COVID-19 pandemic has accelerated this trend.Footnote 1 This trend requires demands for updated international rules that fit in with this progress.Footnote 2 At multilateral trade fora, the ‘Declaration on Global Electronic Commerce’Footnote 3 adopted in 1998 had already promised to continue the current practice of not imposing customs duties on electronic transmissions. However, other than the ‘Moratorium on Customs Duties on Electronic Transmissions’ in the Declaration, more visible progress has never been made. Delays in rule-making at the multilateral level have induced negotiators to develop digital trade rules under the Regional Trade Agreements (RTAs). This began with a few clauses such as the ‘paperless trading’ provision,Footnote 4 but soon it has evolved into a separate chapter for digital trade, and recently into a stand-alone digital trade agreement covering various issues including artificial intelligence (AI) and data innovation.Footnote 5 Along this course, the goals of the digital trade rules in the RTAs have expanded from promoting digital trade to enabling trusted data flows, building trust in digital systems, and so on. The goal for promoting digital trade especially has materialized into certain principles against unjustified trade barriers.Footnote 6 In general, the fundamental trade norms in the multilateral trade fora include principles on non-discrimination, market access, fair trade, and solutions to conflicts between free trade and social value,Footnote 7 and it is easy to find that the recent digital trade agreements also adopt similar principles.Footnote 8 Among them, this study concentrates on the ‘non-discrimination principles' (denoted by the principles hereinafter, to avoid confusion),Footnote 9 prohibiting discrimination against ‘like’ objects for the purpose of providing a fair and free trade environment in the context of digital trade.
Discussion on the principles in digital trade is not a new topic in itself. The Seventh Session of the WTO Ministerial Conference in 2009 had already decided to begin discussions on the principles by ‘Work Program on Electronic Commerce’.Footnote 10 However, with the development of digital technology, discriminatory measures in digital trade are taking more diverse forms over time, such as placing additional requirements on overseas digital service providers (e.g. data localization, additional license or fee, local presence) and unforeseen blocking or delaying the supply of overseas digital services for specific purposes (e.g. social order, cultural or political inappropriateness).Footnote 11 Responding to this, the importance of the principles has recently been reconfirmed in that ‘addressing discriminatory practice’ was set forth as the main goal in the ‘Digital Economy’ category under Pillar I (Trade) of IPEF (Indo-Pacific Economic Framework) in September 2022.Footnote 12
Even if the principles are stipulated in the international agreements, it is still not clear how they will be applied in practice. One of the most fundamental obstacles comes from the fact that it is difficult to determine what the discriminatory treatment itself means without having a common definition of ‘digital trade’. To make it worse, it is even harder to specify the subjects of the principles in digital trade. The principles usually materialize with ‘market access’, which is referred to as the ‘ability of foreign suppliers to compete in the national markets without encountering discriminatory, excessively burdensome or restrictive conditions’.Footnote 13 However, commitments on market access of digital trade are not mentioned or at best are unclear in most RTAs. What does it then mean to have the principles in the agreement?
To address these issues, we first investigate thoses associated with the unclearness of the subject of the principles under the existing trade norms and introduce a theoretical framework on the incentives behind introducing the principles. We then empirically test, using our framework, whether the inclusion of the principles in a trade agreement actually helps to promote digital trade flow and what their roles are. Though one may argue that it is clear that the principles help to promote digital trade, how these principles are actually applied and implemented in practice is by no means a simple issue. As will be discussed later, the effect of the principles on digital trade flow may be more effective even in a situation where the subject of the principles is not completely clear as the principles are more densely stipulated in more areas. We also briefly discuss how the principles are stipulated in the RTAs to overcome such unclearness and which form of the principles can be clearer solution at a current stage of a norm development process.
This study is related to three different strands of literature on the principles. First, a majority of studies focus on how or whether the existing WTO rules should be changed or mutatis mutandis applied in digital trade. Thus, the principles are covered as a part of other existing WTO rules. For instance, these types of studies focus on issues such as the gap between the GATS and the digital economy.Footnote 14 Their main interests lie in reviewing the applicability of WTO rules in digital trade and do not focus on the principles in detail.
Second, some studies analyze the principles in the very specific context of digital trade. Most of them so far have been embodied mainly in the field of digital taxes. They investigate whether a specific country's unilateral imposition of digital taxes may violate the principle under WTO. Discussions have continued with conflicting views on whether this does violate the non-discrimination principleFootnote 15 or not,Footnote 16 for example, studies on the EU's digital tax and the ‘Equalization Levy’Footnote 17 imposed by India. Our study differs from them in that it deals with the more general aspect of the principles rather than just digital tax.
Third, recently a few narrowly tailored studies on the principles have been undertaken in policy research institutes. For example recent trend changes in the principles, such as the range of interpretations on ‘like products’ or applications on their exceptions under the digital trade agreement by the US since the US makes a constant effort to recognize only very exceptional excuses for appling the principles.Footnote 18 Rather than dealing with the particular digital trade-related provisions in a specific group of trade agreements, we investigate the most fundamental principle provisions of digital trade across all existing agreements.
In sum, only a few studies have dealt with the principles to be applied ‘throughout digital trade’, not limited to specific areas such as digital taxes. In addition, there has been no research on the principles ‘in connection with their expansion effect on digital trade as a whole’.Footnote 19 From this point of view, this study contributes to the literature in that it directly addresses a very fundamental question on the principles, whether and how they may cause the effect of promoting digital trade. To our best knowledge, this is the first study to discuss the meaning and role of the principles and to empirically analyze how the principles affect digital trade particularly.
The rest of this study proceeds as follows. Section 2 summarizes the issues related to the principles in trade agreements and provides a theoretical framework to explain them. Section 3 introduces the digital trade-related data we use and empirically analyses the relationship between the principles and digital trade across trade agreements. Based on the results from previous sections, Section 4 discusses additional considerations of the empirical analysis and derives policy implications. Finally, we conclude in Section 5.
2. Issues and Theoretical Framework
2.1 Issues on the Definition of Digital Trade
To begin with, it cannot be denied that a wide range of international organizations played a critical role in continuously discussing digital trade and its potential rules at the international level. According to the ‘Work Program on Electronic Commerce’Footnote 20 adopted by the WTO General Council on 25 September 1998, the term ‘electronic commerce’ is understood to mean the production, distribution, marketing, sale, or delivery of goods and services by electronic means. Moreover, IMF made an attempt to define ‘digital trade’ as all international trade flows that are either digitally ordered, digital intermediary platform-enabled, or digitally delivered.Footnote 21 In a similar sense, the OECD stipulates digital trade as ‘all trade that is digitally ordered and/or digitally delivered’.Footnote 22
However, there is no consensus among countries with regard to the definition of ‘digital trade’, as the process of regulating digital trade within a determined scope is highly connected with each country's national interest and domestic industries. First of all, in the case of the US, the US International Trade Commission (USITC) tends to broadly define digital trade.Footnote 23 We anticipate the US to easily regulate the online sales of tangible goods in conformity with the rules of trade-in-goods, while the US intends to concentrate on establishing new digital rules to regulate intangible goods and related services. Second, the EU's definition on digital tradeFootnote 24 is not sufficient to evaluate the EU's position on digital trade because it is too broad and vague. Meanwhile, China is reluctant to accept the term of ‘digital trade’ and adheres to the term ‘e-commerce’ instead.Footnote 25 As outlined above, only a few FTAs contain a clear-cut definition of ‘e-commerce’ or ‘digital trade’.Footnote 26 Accordingly, it is unclear what is digitally traded and furthermore, whether the digital products are classified as goods or services.
Although a variety of views on the definition of digital trade are presented as aforementioned, a core issue from the promotion of the digital trade perspective is whether the target of digital trade is subject to ‘goods’, ‘services’, or ‘tertium quid’.Footnote 27 The principles under trade-in-goods provide that foreign products should be accorded no less favorable treatment than domestic products, and the principles under trade-in-services, in a similar manner, provide that foreign services and service suppliers should be accorded no less favorable treatment than domestic services and service suppliers. In addition, the principles in digital trade stipulate that no less favorable treatment should be granted to other ‘like’ digital products, along with the definition of ‘digital products’. Meanwhile, digital trade rules usually set up ‘measures affecting trade by electronic transmission’ as their scope of application. This is similar to the scope of cross-border trade in services, which is defined as ‘measures affecting cross-border trade in services by foreign service suppliers’. However, unlike trade-in-services which expressly specify the modes of supply from Mode 1 to Mode 4, it is complicated to answer correctly what the subject of market access is in digital trade and what the target of the principles is, due to the controversy over the term ‘trade by electronic transmission’. However, since the term ‘electronic transmission’ is generally defined as the transfer of digital products using ‘electromagnetic or photonic means’, digital trade is highly likely to be closer to trade-in-services rather than trade-in-goods considering the dichotomy under the current WTO system. Still, digital trade is not clearly distinguished from trade-in-services because the subject of digital trade is somewhat ambiguous. Furthermore, the recent dichotomy between goods and services is unclear because newly released smart items using digital technology cannot be clearly considered as goods nor their embedded services.Footnote 28 In addition, it is highly likely that countries are reluctant to categorize digital trade as either trade-in-goods or trade-in-services. The types or forms of exceptions clauses in RTAs are able to support this theory. For instance, in accordance with the Korea–US FTA,Footnote 29 e-commerce is, at a glance, likely to be categorized as trade-in-services because it quotes GATS Article 14 (Exceptions) other than GATT. Whereas, its footnote explicitly states that ‘Article 23.1 is without prejudice to whether digital products should be classified as goods or services’.Footnote 30 Moreover, dominant phenomena such as the advent of new digital business models, package of goods and services, and servicification effect might make the market access more complicated and critical.Footnote 31 To make it worse, digital technology, which enables the trade of bundling of goods and services, will challenge the traditional market access commitment under the clear-cut divided WTO world, such as trade-in-goods and trade-in-services.Footnote 32 As previously stated, disagreements on the definition of digital trade and difficulties in the classification of digital trade produce manifold problems, especially how to apply the principles to digital trade.
2.2 Issues on the Principles in Digital Trade
The traditional meaning of the principles in international trade implies no discrimination between ‘like’ products according to the nationality of the products. The principles under WTO are twofold: the most-favored-nation (MFN) principle, which prohibits a country from discriminating between other countries, and the national treatment (NT) principle. The NT principle under the WTO requires that no less unfavorable treatment should be imposed under the concept of ‘like product’ or ‘direct competition or alternative product’. In this regard, there is not much controversy over the interpretation of ‘no less unfavorable treatment’; contrariwise, there have been disputes on ‘like product’. The MFN principle is also a concept to guarantee fair conditions of competition by preventing intervention in the market to favor or unfavorably manipulate the products of a specific country. In the case of digital products, it is even more difficult to distinguish the country of origin among foreign digital products because it is harder to determine the country of origin than for goods or services and furthermore, a method on country of origin has not been established under digital trade. As will be shown later, in practice the MFN principle for digital trade has overall so far been secondary to the NT principle in that there is no RTA which stipulates the MFN principle only without the NT.
The GATT and WTO's appellate body and panels have relied on a traditional interpretation of ‘like product’ on a case-by-case basis.Footnote 33 The basic criteria for determining ‘likeness’ are stipulated in the Report of the Working Party on Border Tax Adjustments,Footnote 34 which includes the following: (i) the product's end uses in a given market or the product's objectives in a particular market, (ii) consumers’ tastes and habits, (iii) the product's properties, and (iv) the nature and quality of the product. In other words, its criteria include physical characteristics or the physical identity of two products: functional likeness or end-uses, tariff classification such as harmonized system code (HS Code), and consumer tastes and habits. The principles share exactly the same hurdle as aforementioned, that is, ‘likeness’. It is even worse for grouping ‘like digital products’ since it is intangible and wirelessly transmitted.
In the case of physical products, it is possible to distinguish whether they are in the same or a different group according to their appearance and propensity, which can be distinguished with the naked eye. However, in the case of digital products, which are intangible, their appearance or propensity cannot be uniformly determined. The consumption behavior and final consumption purpose of digital products are extremely diverse. Many FTAs provide for the non-discrimination principle under chapters on trade-in-goods and trade-in-services respectively, and the interpretation of these provisions follows the above WTO jurisprudence which applies to goods under the GATT or services under the GATS. However, when it comes to the principles specific to digital products, there is a debate on how to interpret them. Furthermore, many FTAs reserve decisions on whether digital products are goods or services by adding such a footnote under Korea–US FTA, ‘the definition of digital products should not be understood to reflect a Party's view on whether trade in digital should be categorized as trade-in-services or trade-in-goods. Some commentators try to indicate that it is unavoidable to consider additionally new criteria such as skills, size of the company, business areas, experience and knowledge, number of employees, type of assets, and technology equipment in the case of intangible services.Footnote 35 Of course, even though digital products such as online games and videos are likely to be standardized in a very basic sense, it is not always straightforward to classify digital products or digital services into existing CPC code.Footnote 36 Digital products have no form and thus can often be individually tailored to all individual consumers with excellent reproducibility and process ability. In sum, it is not easy to distinguish whether digital products optimized for individuals are ‘like products or not’.
Therefore, there is an obvious limit to responding to the discriminatory measures in digital trade by making use of the principles developed solely for trade-in-goods and trade-in-services. Thus, in the long run, a separate and additional principle on non-discrimination specialized for digital trade should be contemplated. In this sense, there is room for ambiguity in the application of the principles. Against such a backdrop, to be sure of providing a level playing field, it can be more effective for the time being to stipulate that the principles apply to both trade-in-services and digital trade than to stipulate that they apply to only one of them.
2.3 Theoretical Framework on the Principles in Digital Trade
Despite the ambiguity in the subject of the principles, as discussed, then why do many trade agreements include clauses on the principles? What is the theoretical reasoning that these provisions will help facilitate digital trade? There can be at least two possible theoretical explanations for the incentives to include the principles.
First, the inclusion of the principles in digital trade agreements may help mitigate trade policy uncertainty. The emergence of new objects of digital trade causes uncertainty as to whether they are goods or services, and if so, which classification code they fall under. Even if the new object of digital trade is classified as a service for instance, most of the FTA negotiations on service concessions are often based on the 1998 CPC code.Footnote 37 Therefore, it is highly unlikely that concessions on new digital services will be accepted under ‘the positive list approach’, under which the subject sectors of liberalization, conditions, and restrictions are explicitly inscribed on the list. Even when concessions are granted with ‘a negative list approach’ under which liberalization is committed for all those that are not included in the schedule of reservations, it is not certain whether the other country will accept the interpretation as including new trade objects that could not be foreseen at the time of the negotiation. Such a situation implies additional compliance costs for exporting countries and institutional uncertainty for exporters, which in turn impedes trade.
Defining a new area and newly including the principles are expected to have the effect of mitigating this uncertainty. Although the principles may be partially redundant with the existence of the principles under trade-in-goods and trade-in-services chapters, it could still be helpful to have a clause on the principles specific to digital trade as long as there is novelty to which market participants believe that the principles under other chapters may not apply. Specifically, to raise an issue against the discriminatory measures of the importing country, it should be possible to specify on what legal basis the measures are problematic. However, due to the ambiguity of the definition of digital trade itself, there may be uncertainties in whether the clause on the principles in either trade-in-goods or trade-in-services is applicable to the case or an additional clause on the principles is needed to cover the case. Such uncertainty hinders companies from entering foreign markets. Explicit inclusion of the principles on digital trade under the trade agreements can mitigate such uncertainty, though not completely, by improving the credibility of the importing country not to take discriminatory measures. This is particularly important for exporters when they must decide on costly irreversible investments, such as adopting a technology, producing a new product, or selling in a new market even if applied trade barriers are currently low or do not existFootnote 38. Related examples are requirements for server localizationFootnote 39 or local presence,Footnote 40 and unforeseen control over international data flow (e.g. purposed data transmission delay, website blockingFootnote 41). For example, imposing a server localization requirement on multinational digital platform companies doing business in a specific country that requires them to keep their server locally would impose a much greater burden on foreign companies than on domestic companies, which are more likely to already have servers in this country. In this respect, server localization is directly related to the principles. When policy uncertainty over cross-border data transfer is significant, it discourages foreign firms’ market entry decision and investment, and subsequently hinders the flow of digital trade across borders.
Second, having the principles can help to mitigate trade costs incurred by the characteristics of digital technology. One of the key objectives of trade agreements between countries is trade liberalization, and digital trade is no exception. Even if two countries have agreed to abolish a cross-border measure, the importing country can restore it through the imposition of discriminatory domestic measures against the imported goods. And the principle devised to solve such a problem is national treatment.Footnote 42 It is also argued that the same logic is maintained in digital trade as well.Footnote 43 In digital trade, discriminatory treatment by itself creates substantial trade costs for foreign firms. Trade liberalization generally aims at reducing trade costs and it is more relevant for digital trade which is often characterized by the reduction of various costs (e.g. search costs, transportation costs, replication costs.Footnote 44 In a situation where any other transaction costs are dramatically reduced due to digital technology, additional costs incurred due to the discriminative measure of the importing country can be significant enough to undermine the foreign producers’ accessibility to the importing country's market. That is to say, as the absolute cost level related to digital trade decreases, the relative importance of the cost due to the discriminatory measures becomes decisive. In this sense, the non-discrimination treatment becomes a critical factor for market access when other transaction costs get relatively smaller. After entering the market, foreign firms may encounter additional trade costs in the form of de facto discriminatory feesFootnote 45 or license costs.Footnote 46 Even when restrictions on cross-border data movement are expected, the existence of these measures or other restrictive technology requirements could incur additional costs to foreign firms in terms of the deterioration of their service quality, especially when the service is operated through well-integrated globally distributed data networks.
Due to these underlying functions of the principles, we can expect the following phenomena to be observed. That is, when the effect of alleviating policy uncertainty and that of reducing trade costs exist, we can expect that a country has the incentive to include the non-discrimination principle in its trade agreement anticipating an increase in digital trade flow. For such an effect on trade flow of the principles, we call it the trade promotion effect. Notice that it is very challenging to directly empirically measure how much the uncertainty has been reduced and how much trade cost has been alleviated due to the principles. To overcome this, instead we measure indirectly how effective the introduction of a principle is in mitigating trade barriers, assuming that the uncertainty and trade costs are embodied as trade barriers. For such an indirect effect of mitigating the uncertainty and trade costs, which materialized as alleviating trade barriers, we call it ‘the trade barrier mitigation effect’. Finally, these effects are expected to be stronger when the principles are more clearly stipulated, even if they overlap, than when the principles are marginally added to the existing terms. We denote it as the acceleration effect by rule clarification. Based on this framework, we can establish empirical hypotheses as follows.
H 1.1 (Trade promotion effect) A pair of countries that have a trade agreement containing non-discrimination clauses on digital trade are more likely to experience more digital trade flows.
H 1.2 (Trade barrier mitigation effect) Having the non-discrimination principle clauses on digital trade mitigates the negative impact on digital trade caused by digital trade barriers.
H 2: (Acceleration effect) This tendency is even more pronounced as the principles are stipulated (albeit overlapping) more clearly in the agreement.
3. Empirical Analysis
3.1 Data Description
3.1.1 Digital Trade Flow
To test the empirical hypotheses, we use various digital trade related datasets. First, to measure the digital trade flow between two countries, we use the cross-border trade in services (which is known as ‘Mode 1’ trade in GATS) in a sector which heavily depends on digital trade. Specifically, based on the bilateral service trade flow information, we apply a proportional allocation method to the modes of supply.Footnote 47 For the bilateral service trade flow, we use the database established by the OECD–WTO, BaTiS (Balanced Trade in Services Database) which covers the period from 2005 to 2019. For the share of Mode 1 trade information by country-sector, we use TiSMoS (Trade in services data by mode of supply).Footnote 48 As the representative sectors closely related to digital trade, we focus on the cross-border ICT (Information, Computer, and Telecommunication) service sector in this study. Suh and Roh show that the Mode 1 trade in ICT-enabled service sectors, the most digital trade-intensive service sector, can be a reasonable proxy for digital trade. For more discussion on the validity and significance of this approach.Footnote 49
3.1.2 Non-Discrimination Principles in Digital Trade Agreements
To identify the trade agreement that includes the non-discrimination principles related to digital trade, we use the TAPED (Trade Agreements Provisions on Electronic-commerce and Data) database, which covers the trade agreements signed since 2000.Footnote 50 To match the period with the digital trade flow data, we focus on the active agreements in force until 2019. During the period, there were 97 trade agreements containing at least more than one digital trade-related provision (in short, digital trade agreements). Of these, 73 agreements include the principles either in a separate digital chapter or in a service chapter.Footnote 51
In the separate digital chapter, the principles usually appear in the article titled either ‘digital products’ or ‘non-discriminatory treatment of digital products’. In the service chapter, the principles are stipulated under the article such as ‘market access’, ‘national treatment’, ‘national treatment, and most favored nation treatment’. For its simplicity, we call the former case D-chapter and the latter S-chapter. Table 1 summarizes the distribution of agreements depending on whether and how they include the principles.
Frequency table of the principles for digital trade
Notes: aMost of them include both ‘National Treatment’ (NT) and ‘Most Favored Nation Treatment’ (MFN) but three which mention NT only: India–Singapore ECA, Korea–Singapore FTA, and Central AmericaKorea FTA.
bFollowed by GATS 16.17, it is typical that national treatment is mentioned only if market access is allowed in an agreement on trade in service. In contrast, MFN is based on a negative system that unconditionally applied to all services regardless of whether or not market access is allowed, and only items to which MFN does not apply are listed separately. Therefore, in our study, whether the trade agreements include the principles in the S-chapter is determined by whether the service and investment chapters include ‘market access’ and ‘national treatment’ in the trade agreements. Since our study focuses on the ICT sector, we only consider the principles for computer, telecommunication, and related services sectors.
Source: calculated by the authors based on the TAPED.
There are 24 agreements that do not include the principles in any chapters, even when they contain digital trade related provisions. 43 agreements do not include the principles in the D-chapter, but do so in the S-chapter. As mentioned above, there is no case that includes the principles in the D-chapter but not in the S-chapter. Finally, 30 agreements contain the principles in both chapters.
From this, we can observe two interesting properties in the mode of inclusion, overlap and direction. First, a significant number of agreements contains the principles in both chapters, seemingly overlapping. Second, every agreement with the principles in the digital trade chapter contains them in the service trade chapter as well.Footnote 52 Based on these two properties, we can develop a simple index of how an agreement includes the principles, and call it ND (Non-Discrimination). That is, depending on whether the principles are included in any chapter, we can assign the value of 1 if they are included and 0 otherwise. According to the direction property, this is actually a classification method depending on whether it is an S-chapter or not. We can further develop this index by disaggregating it into two sub-indexes, ND both and ND S-only. ND both has 1 if the principles included are overlapping (or in the D-chapter due to direction property in our sample) and 0 otherwise. ND S-only has 1 if they are only in the S-chapter and 0 otherwise. Note that we can restore ND with these two sub-indexes, that is, ND = ND both + ND S-only. In the sense that overlapping inclusion will work in a manner that reduces uncertainty about the applicability of the principles, such a disaggregated index allows us to observe the intensity of how an agreement includes the principles.
Consequently, we focus on the disaggregated approach initially and then discuss later whether our main results change depending on whether we combine, or separate, the indexes.
3.1.3 Digital Trade Barriers
Discriminatory treatments are materialized in the form of various kinds of domestic measures. To capture this, we use the OECD digital services trade restrictiveness index (DSTRI), which is an overall index to measure how a country has a digital trade restrictive environment. It is a composite index that is based on all the domestic laws or regulations recognized as trade barriers to digital trade through scoring, weighting, and aggregation.Footnote 53 Its value ranges between 0 and 1, and the higher value implies a more restrictive digital trade environment. The original dataset covers the period from 2014 to 2021 for 74 countries, but we use the dataset up to 2019 to cope with the digital trade flow data.
According to the policy areas, DSTRI can be broken down into five sub-indexes, namely ‘Class 1 (Infrastructure and connectivity)’, ‘Class 2 (Electronic transactions)’, ‘Class 3 (Payment systems)’, ‘Class 4 (Intellectual property rights)’, and ‘Class 5 (Other barriers)’.Footnote 54 To scrutinize digital trade barriers in depth, we pay our attention to Class 1, especially. It is not only because it is the most important policy area in constructing the overall DSTRI, in which the highest weight of 0.55 is assigned to it, but also because it is relevant to the discriminatory treatment. As discussed before, most discriminatory perceived digital trade barriers are related to cross-border data flow restriction or data localization issues that are categorized into Class 1. For example, the policies recognized as discriminatory measures by the USTR NTE report (such as ‘Article 37 of the Cybersecurity Law’ of China, and ‘Article 16 of Act on Establishment, Management of Spatial Data’ of Korea) are listed in the regulatory measures consisting of Class 1.Footnote 55 To distinguish this from the overall index, DSTRI, we denote it as ‘DSTRI 1’ as a variable name in our empirical analysis.
3.2 Empirical Model Specification
The main interest of our empirical analysis is to test how the inclusion of the principles in a trade agreement affects digital trade flows. To see this, we introduce a panel gravity model with the following regression equation.
$${\rm I}{\rm M}_{ij, t} = \alpha + \beta _1ND_{ij, t} + \beta _2DSTRI_{ij, t} + \beta _3ND_{ij, t} \times DSTRI_{ij, t} + \beta _4X_{ij} + I_{it} + I_{\,jt} + \varepsilon _{ij, t}$$The dependent variable (IMij,t) is the imports of digital trade from country i to country j in year t. Our key independent variable, ND, captures whether the pair of countries, i and j, have a trade agreement with a non-discrimination clause. This is sometimes broken down into two sub-index variables, ‘ND both’ and ‘ND S-only’, as explained before. It is expected that its coefficient β 1 will be positive if there exists the trade promotion effect of the principles. When we use sub-indexes instead of ND as a whole, we expect that the coefficient of ND both is greater than that of ND S-only, that is, the acceleration effect.
As the main control variable, we use DSTRI but in a bilateral form, DSTRI ij,t = DSTRI i,t × DSTRI j,t. The reason why we do not directly use each country's DSTRI is to control the time-varying unobservable multilateral resistance which is required to be theoretically consistent.Footnote 56 To fulfill this, we introduce the country-year fixed effect model, in which I it and I jt are the dummy variables for importer-year and exporter-year, respectively. Under these, all the country-year varying variables are absorbed by these fixed effect variables, including DSTRI i,t or DSTRI j,t. To overcome this issue as well as to consider the DSTRI aspect, we use their interaction term and interpret it as the restrictiveness of the regulatory environment the pair of countries have between them. We expect that its coefficient β 2 will be negative, which implies that a more digital trade-restrictive regulatory environment will dampen the digital trade flow. Furthermore, to estimate the marginal effect of ND given DSTRI, we also introduce an interaction term between two, ND ij,t × DSTRI ijt. If the sign of its coefficient β 3 is positive, it will mean that the ND inclusion has the mitigation effect against the negative effect from trade barriers on the digital trade flow. Having a greater magnitude will imply a greater mitigation effect against a more restrictive regulatory environment. Finally, the other typical time-invariant bilateral trade cost variables in the gravity model of international trade literature are denoted by X ij. This includes ‘distance’, ‘contiguity’, and ‘common language’ and is sourced from the CEPII gravity data.
Taking logarithms is often used in a gravity model, but it makes zero trade flows drop out of the estimations. In that zero trade flows can be informative if they are due to prohibitive trade costs, useful information can be lost when zero trade flows drop out of the estimations. To handle this issue, we use the Poisson Pseudo Maximum Likelihood (PPML) estimation, which enables us to avoid dropping zero trade flows out of the estimations because it is applied to the levels of trade.Footnote 57
3.3 Empirical Results
Our main empirical results are presented in Tables 2 and 3. We first examine whether (H1.1, Trade promotion effect) holds or not.Footnote 58 In Table 2, all the coefficients of ND have a positive sign and they are statistically significant and this implies that once the principles are included in a trade agreement, they are helpful to promote the digital trade flow.Footnote 59 As expected, the coefficients of digital trade barriers have a statistically significant negative sign in columns (2) and (3), implying that digital trade barriers basically impede digital trade flows. In columns (3), we can see that a more discriminatory trade barrier (DSTRI1) has a stronger negative effect than overall barriers (DSTRI) in column (2).
Empirical results of ND in any chapters
Note: *denotes statistical significance at the 10%, ** at the 5%, ***, and at the 1% level.
Empirical results of ND overlap
Note: *denotes statistical significance at the 10%, ** at the 5%, *** and at the 1% level.
To check whether DSTRI is a proper measurement for digital trade barriers, we also consider OECD STRI for the relevant sectors, computer and telecommunications. In columns (4) and (5), the signs of their coefficients are negative but statistically insignificant. This may imply that DSTRI is certainly a more targeted index to digital trade compared to a typical STRI measurement.Footnote 60 As the distance decreases and a common language is used between the parties, the digital trade flows also increase statistically significantly. However, ‘contiguity’ is not statistically significant, meaning that we could not find evidence that sharing the same border between two countries increases digital trade flows.
Next, we conduct a similar analysis with the finer measures of ND, ND S-only and ND both, and the results are summarized in Table 3. Similar to the previous results, it confirms the positive effect of ND (for both variables) while the negative effect of DSTRI (as well as DSTRI1) on digital trade flow. Other trade cost variables still show the expected signs as well. However, there are also several new observations. First, the coefficient of ‘ND both’ is larger than that of ‘ND S-only’ in our baseline model, column (1), which implies that defining ND in both chapters affects digital trade to a greater extent than defining ND in the service chapter only. However, this does not appear in columns (2) and (4), which consider only the direct effect of ND on the digital trade flow under the existence of DSTRI. Second, when we take the indirect effect of ND against DSTRI into account as well, we can see how clarification on ND mitigates the negative effect of digital trade barriers on the digital trade flow (H1.3, Trade barrier mitigation effect). Columns (3) and (5) show that the interaction term between ND both and DSTRI as well as that between ND both and DSTRI1 are positive and statistically significant, respectively. It means that ND both has a greater trade barrier mitigation effect on digital trade and it is more effective under a more restrictive digital trade environment. Column (5) shows that this tendency is more apparent against more discriminatory barriers, DSTRI1. ND S-only has a positive sign with a greater magnitude coefficient than ND both at a glance but such a seemingly advantageous effect disappears when we consider its indirect effect against DSTRI. With a negative sign for its interaction term with DSTRI, ND S-only easily loses its positive role to help the digital trade flow against digital trade barriers. It suggests that stipulating the principles clearly, even if they overlap, can help to easing digital trade barriers (H2, Acceleration effect).Footnote 61
4. Discussion
In this section, we consider other relevant aspects of the results from our above analysis. First, in that the empirical results in Section 3 focused only on the effects of the principles, we further discuss cases when other related provisions of the agreement are also considered. Also, we discuss what policy implications can be based on the issues and empirical results from Sections 2 and 3.
Indeed, besides the principles, there can be other provisions in the trade agreement which may affect digital trade flows. Since the purpose of this study is not which clause affects digital trade flows more (or the most) than the others, we do not consider the effect of every single component of the trade agreement here. Instead, we further examine the effect of other provisions which are expected to influence the digital trade flow as well as being closely associated with ND, that is provisions on data flow and no data localization. To identify the trade agreement that includes the provisions on data flow and no data localization, we also use the TAPED.Footnote 62
Table 4 below shows the effect on digital trade flows when ‘data flow provision’ and ‘no data localization provision’ are separate and together with ND. While the sign of ND is positive and statistically significant, those of other provisions are mixed but all are insignificant. In that there are not many observations yet,Footnote 63 it may be too early to conclude that these provisions do not affect the digital trade flow. In any case, it confirms that including the principles has a relatively stable and substantial effect.Footnote 64
Empirical results of provisions on data flow and data localization
Note: *denotes statistical significance at the 10%, ** at the 5%, *** at the 1% level.
We found that the principles in trade agreements have the effect on digital trade flow. This effect works well even in a situation where the subject of the principles is not completely clear, and the more densely stipulated the principles are in more areas, even if they are partly overlapped, the more effective they are. However, such an ad hoc and overlapping approach has limitations in its effectiveness in that it inevitably produces a new loophole as the technology keeps developing. Even when the principles are partly overlapped, the emergence of new digital trade could open up another new dimension of ambiguity in which field the principles could be applied in actual cases as long as we do not have a clear definition and scope of digital trade. A representative example must be the recent debate over tech giants’ in-app billing systems.Footnote 65 In 2021, the Amendment Bill to the current Telecommunications Business Act, the world's first attempt to prohibit large application (app) market operators from allowing third party payment systems, was announced in Korea, and similar bills have appeared in other countries lately. However, despite the opinion that in-app payments should be regulated in terms of a fair competition perspective, controversy over the violations of the principles from the international economic law perspective has cropped up. The US criticized this Amendment Bill because the Korean Government appears to specifically target US service providers.Footnote 66 The US has raised complaints on these discriminatory measures, but it seems unclear as to from which Chapter under the Korea-US FTA the US would like to quote the principles.
Even though explicit and independent provisions of the principles are stipulated under Chapter 15 (Electronic Commerce)Footnote 67 of the Korea-US FTA, the US seems to choose the principles under Chapter 12 (Trade in Services) by priority. Korea should abide by the principles under Chapter 12 (Trade in Services), avoiding discriminatory treatment towards US application market operators. The Amendment Bill adopts non-biased terms to ban app market operators from enforcing app developers to use in-app purchasing systems and thus, this Bill does not seem to violate the de-jure non-discrimination principle. However, based on the fact that the market share of the US based multinational tech-giants is over 80%,Footnote 68 we cannot exclude any single possibility of a de-facto violation. The Korea–US FTA, a legal base on the aforementioned debate, contains the principles not only under Chapter 12 but also under Chapter 15. Nevertheless, it is likely not to precisely quote which principles between Chapter 12 and Chapter 15 should be applied.Footnote 69 It is also difficult to clarify whether ‘digital products’ defined in the principles of Chapter 15 is applied to the in-app purchasing system. To make matters worse, even though the principles of Chapter 12 are quoted at the moment, it is still vague whether to apply this clause to the newest services such as application market operators.
This shows that the principles under trade-in-services alone are not enough to foster fair competition for digital goods or services produced by the ever-changing digital technology, and that separate non-discrimination principles much more tailored and detailed for digital trade are absolutely necessary. Of course, as aforementioned, the current non-discrimination principles in digital trade are far from complete. It is expected that adding the clearer definition and scope of digital trade can enhance the effectiveness of the principles on digital trade by clearing up the gray areas that are not completely covered by the principles under trade-in-services only.
5. Conclusion
Despite the common perceptions in many countries of the importance of digital trade, it is still controversial among countries whether the subject of digital trade is classified as ‘goods’, ‘services’, or ‘tertium quid’. The absence of a clear definition of digital trade makes the principles of digital trade a non-trivial issue. This is because it is hard to determine what a discriminatory measure is without defining ‘likeness’, a core factor of the traditional interpretation on the principles. And ‘likeness’ is a non-separable concept from what digital trade really is. A temporary solution might be to apply the traditional classification approach to digital trade. However, there might be a loophole whereby the principles in trade-in-services alone cannot completely respond to discriminatory practices in digital trade. After explaining how the principles in digital trade may promote digital trade and mitigate related trade costs even when the subject of the principles is completely clarified, we found the following empirical analysis results. First, the inclusion of the principles in trade agreements is more likely to increase digital trade flows. This is true whether the principles are set out independently for digital trade or as a part of the trade-in-services chapter, or both. Second, though digital trade barriers decrease digital trade flows, this negative effect can be mitigated when the principles are more clearly defined in the trade agreements, even if there are overlapping parts between the chapters.
With the progress of the digital economy and its irreversibility, it is expected that more countries will pursue digital trade liberalization in the near future. Accordingly, the principles for digital trade will also be specified and generalized. Nevertheless, the reality is that a number of countries are still concerned about a rapid opening up through the application of the principles. In this sense, the deepening of the digital economy may exacerbate these differences in positions between countries. In this study, we only highlighted the most fundamental aspect of the principles, that is, their trade promotion effect. There may be another side to the effects of the principles making a group of countries to be reserved. A more balanced study is thus needed, which will also address the other side and suggest how to handle it in future studies.
Funding
This work was supported by a Yeungnam University Research Grant.
Appendix
Under the structure of correlations between the elements of an agreement, we use the principal component analysis method to examine the effect of a specific element. To proceed with this method, we take the following three-step approach. That is, we (1) construct a variable which is closely related to the principles, one reflecting the commitment to ‘data flow’ or ‘no data localization’ requirement, (2) calculate the principal components, which are independent vectors that linearly combine the provision variables of the digital trade agreement classified on the TAPED, and then (3) estimate how principal components representing the key provisions affect the digital trade flow.
STEP 1. Constructing free data flow provision variable
As explained in footnote 61, variables directly related to the commitment to data flow or no data localization requirement, are TAPED #1.28.1 and #1.28.4. Considering the relationship between these clauses, ‘Data flow only’ is defined as giving 1 if only TAPED #1.28.1 is greater than 1 and 0 otherwise. ‘No data localization only’ is defined as giving 1 if only TAPED #1.28.4 is greater than 1 and 0 otherwise. ‘Data flow & No data localization’ is defined as giving 1 if both TAPED #1.28.1 and #1.28.4 are greater than 1 and 0 otherwise. The cross-frequency among the principles, data flow, and no data localization is shown in Table A1 below.
Cross-frequency among key provisions in DTAs
STEP 2. Calculating the principal components with provisions in digital trade agreements
For the principal component analysis, 43 variables classified as ecommerce data flow related clauses in the TAPED are used, excluding non-essential or redundant variables such as dummy of digital trade provision inclusion, and dummies of US type or EU type, etc.Footnote 70 For more details on the variable information, please refer to the Codebook Big Data Trade Agreements.Footnote 71 below summarizes the top 10 provisions which consist of each principal component.
Top provisions for each principal component
We can observe that the non-discrimination principle clause is highly correlated with the first principal component (pc1). So, we may interpret pc1 as a proxy variable of the non-discrimination principle. A free data flow related clause is also included in pc1, but its ranking (26th) is not so high. There is no principal component that contains the non-discrimination clause variable within the top 20 or even the top 30. It is similar to the free data flow clause. The only exception is the subordinated principal component (PC5) which contain it in the 16th place. This does not necessarily imply that the free data flow is not important, but it may be because there are not many observations that contain the related clauses yet.
STEP 3. Estimating the effect of principle components on digital trade flow
Table A3 below shows the results of estimating the effects of digital trade flows as in the previous analysis, using the principle components calculated above as explanatory variables. It shows that pc1 has a significant positive effect on digital trade flow. Though this result does not tell us much about how exactly the principle interacts with other provisions (incl. free data flow ones), at least it confirms that the non-discrimination principle still matters even after considering other provisions together at the same time.
Empirical results of data flow provisions grouped by principal components
