Barriers to Digital Services Trade and Export Efficiency of Digital Services

Abstract

An international rules system on digital trade that can satisfy multilateral interest appeal has not been formed yet despite the rapid growth of digital services trade. Using the panel data of 39 countries from 2010 to 2019, this study applied the stochastic frontier gravity model to analyze the influence of five types of barriers to digital services trade on the export efficiency of digital services. The research results show that infrastructure and connectivity barriers had no significant effect on the export efficiency of digital services. The barriers to electronic transactions and other barriers that affected the trade of digital services were conducive to improving the export efficiency of digital services. In contrast, the barriers to payment systems and intellectual property rights had a restraining effect on the export efficiency of digital services. The results also show that most countries were more efficient when exporting digital services to their major trade partner countries, which indicates a demand-driven characteristic. Based on these conclusions, we propose five recommendations that are helpful to enhance the export efficiency of digital services.

1. Introduction

The widespread application of digital technology in the field of international trade has transformed many traditionally non-tradable services into tradable digital services, which is greatly expanding the scope of international trade. Digital technology develops new carriers for traditional services, provides an auxiliary function for trade in services, and breaks geographical limitations between service providers and consumers. It not only makes services more tradeable but also wonderfully enhances the efficiency of services trade. However, existing and emerging trade barriers are likely to deplete the benefits of digitalization in service industries, and thus, harm the sustainability of trade in services [1].

Trade in digital services is a consequence of digitalization in service industries. In order to identify the research scope of this study better, we need to clarify the concept of digital services trade. The United Nations Conference on Trade and Development (UNCTAD) defines the digital delivery service as services that can be delivered remotely through information technology. The digital services trade investigated in this paper refers to the cross-border trade of digital delivery services. Trade in digital services is gradually becoming an important component of international trade. According to the UNCTAD, the proportion of world digital services export in the total export rose from 9.19% in 2006 to 13.42% in 2021. And its proportion in the total services export was up to 62.77% in 2021. Digital services trade can provide an impetus for advances in technology and facilitate global industrial restructuring [2,3]. Trade in digital services, as an emerging engine for world economic growth, plays a noteworthy role in fostering economies in various countries, even in the face of the negative shock caused by COVID-19. At present, the development momentum of digital services trade is strong, which is a new power to promote the recovery of global trade and lead trade growth [4].

Barriers to digital services trade are essential institutional variables influencing the development of digital services trade primarily due to the rule-intensive pattern of digital services trade. An international rules system on digital trade that can satisfy multilateral interest appeal has not been formed yet despite the rapid growth of digital services trade. At the same time, the current international labor division model dominated by global value chains (GVCs) places a strong emphasis on efficiency gains in trade. In this context, exploring how barriers to digital services trade impact the export efficiency of digital services would be a valuable addition to existing theoretical studies. This is also of great benefit to explore regulations of digital trade, which is thus helpful to eliminate inappropriate regulations and boost the sustainability of digital services trade. Improving the export efficiency of digital services from the perspective of institutional opening is another practical significance of our research.

The literature on the trade effect of barriers to digital services trade is relatively scarce, although it has recently been increasing. Existing research is mainly conducted from two perspectives. Some research explores the overall impact of barriers to digital services and concludes that these barriers have negative effects on digital services trade [5]. The other concentrates on the influence of restriction policies on cross-border data flow primarily due to the representativeness of these barriers. It was found that strict restrictions on the cross-border data flow have a negative impact on digital services export, although these restrictions can solve the privacy issues associated with digital services trade [6]. At the same time, when it comes to trade efficiency, scholars mainly focus on doing research in the domain of goods trade, while they rarely pay attention to trade in services. Only a few scholars discussed how economic freedom, government management and other factors impact the efficiency of digital services trade. Hence, there is still a significant research gap regarding how barriers to digital services trade impact the export efficiency of digital services, which needs to be investigated. And it is necessary to conduct empirical research in this domain because theoretical analysis indicates that barriers to digital services trade have both negative and positive effects. Many regulatory policies on trade in digital services are designed to achieve particular social objectives, such as safeguarding health and safety. They will provide substantive benefits to countries if these objectives are achieved. Meanwhile, their restrictive effects on digital services trade also damage economic efficiency and social welfare. Only by doing empirical analysis can we master the net effect of these barriers on export efficiency of digital services. The research results of the net effect provide reasonable and accurate grounds for policy makers to enhance the export efficiency of digital services by modifying domestic regulatory policies. Moreover, empirical research based on the stochastic frontier gravity model makes it possible to quantify and compare the export efficiency of digital services in different countries.

The Organization for Economic Co-operation and Development (OECD) divides barriers to digital services trade into five categories, which include infrastructure and connectivity, electronic transactions, payment systems, intellectual property rights and other barriers. The Digital Services Trade Restrictiveness Index (DSTRI) is obtained by adding up the restrictiveness indices for each type of trade barrier, which makes empirical research on barriers to digital services trade possible. Thus, the research objectives of this paper mainly include analyzing the export efficiency effects of different barriers to digital services trade and comparing digital services export efficiency in different countries. The conclusions of this research can provide meaningful theoretical foundations for governments to refine their policies of trade in digital services and improve the sustainability of digital trade practices. Our main contributions to the existing literature are twofold. First, we focused on five dimensions to study the impact of barriers to digital services trade on the export efficiency of digital services to expand the research on the trade effects of barriers to digital services trade and introduce new perspectives for improving the quality of export of digital services. Second, we measured the export efficiency of digital services for 39 countries and made a comparative analysis of the calculation results. This provides empirical evidence for countries to tap into the export potential of digital services.

The rest of this paper is organized as follows. Section 2 outlines existing studies about the trade effect of barriers to digital services trade, particularly about the impact of these barriers on digital service export. Section 3 discusses our research design, which includes details about the empirical model and data that we used. Section 4 presents the regression results, where we made a detailed analysis of how different barriers influence digital services trade. Section 5 gives an extended analysis of the differences in the export efficiency of digital services between countries. Section 6 summarizes the research conclusions and proposes policy recommendations.

2. Literature Review

Currently, the world’s major economies are no longer satisfied with the increase in trade volume. Instead, they are increasingly concerned about how to tap into the trade potential and improve trade efficiency [7]. Export quality is a better driving force for efficient economic growth than export scale, where export quality can be measured in terms of efficiency [8]. The trade potential is the maximum possible trade under frictionless trade, whereas the trade efficiency refers to the degree to which trade potential is realized. The stochastic frontier gravity model is a popular technique to analyze trade efficiency. In recent years, the literature on trade efficiency or potential has been continuously enriched [9,10,11], but there are few papers concerning digital services trade. At the same time, the global economy has shifted toward the digital economy, with manufacturing companies utilizing digital services trade to achieve digitization and enhance their competitiveness in the global value chain [12]. The digital economy is the name for the new economic form that has emerged after technological progress and economic development reached a higher stage [13]. As an important component of the digital economy, the digital services trade drives the accelerated integration and optimization of the global industrial chain, supply chain and value chain, thus injecting new momentum into economic growth. Therefore, it is imperative to assess the trade efficiency of digital services.

Despite the benefits that digital technology brings to international trade and the economy, its application also causes some new issues, such as privacy disclosure and intellectual property infringement. It is necessary to build a regulatory framework for digital services trade primarily due to its reliance on digital technology [14]. Therefore, appropriate restrictions are requisite for digital services trade. Barriers to digital services trade are adopted by countries, mainly because of privacy protection, industrial policy and national security [15]. For instance, regulations and laws about privacy protection prevent the data of consumers from illegitimate use by enterprises. These domestic policies effectively address the privacy issues that might arise in digital services trade and foster consumer trust in businesses, although they are reckoned to be barriers to digital services trade [16]. And from the perspective of industrial policy, the data localization requirement enables domestic enterprises to have advantages over multinational corporations by attracting more consumers to domestic enterprises [6]. However, existing research also verifies that barriers to digital services trade constrain the prosperity of the world economy and trade activities, especially for the export of digital services. Empirical research conducted by Zhang, X. and Wang, Y. (2022) confirmed that barriers to digital services trade significantly hinder the development of export of digital services [17]. As these barriers increase in their restriction level, domestic enterprises face more challenges when exporting digital services. The deepening of the international division of labor is impeded by barriers to digital services trade because these barriers hamper countries and enterprises from applying digital services as inputs when participating in GVCs, which thus impede the value-added export of digital services [18,19,20].

On the whole, barriers to digital services trade hinder the export of digital services through three channels. Pasadilla, G. O. et al. (2020) summarize the drawbacks of barriers to digital services trade as the increasing costs of export of digital services, which include compliance burden, operation costs and efficiency loss [15]. Stringent restrictions imposed on digital services trade by governments increase costs of enterprises and make processes cumbersome when enterprises export digital services [5]. And enterprises must adapt to multiple regulatory frameworks due to regulatory heterogeneity between countries, which increases the burden of digital services trade [5]. Moreover, barriers to digital services trade result in an increase in market concentration, which can also harm the export of digital services. The increase in trade costs caused by these barriers might only be affordable for big enterprises [15]. In other words, these barriers are formidable market access barriers to small enterprises and weaken market competition. Hindering the innovation of enterprises is another channel by which barriers to digital services trade can present an obstacle to the export of digital services. These restrictions impede the cross-border diffusion of knowledge and technology, increase the costs of innovation and weaken enterprises’ awareness of innovation. As a result, it is hard for enterprises to improve their innovation efficiency and sharpen their competitiveness [5]. Furthermore, some research focused on how restrictions to cross-border data flow affect trade in digital services. Export countries will have difficulty in boosting their trade competitiveness in digital services and their export of digital services will decrease sharply if the cross-border data flow is strictly restricted by them or their trade partner countries [21,22]. As an important approach to restricting cross-border data flow, the data localization requirement is found to reduce the mobility of digital services between countries and hinder the development of trade in digital services [6,18]. The data localization requirement reinforces the monopoly power of domestic enterprises, which thus reduces the variety of digital services available to consumers. And domestic enterprises are likely to have less stimulus to improve the quality of their digital services [6].

3. Research Design

3.1. Model Construction

The gravity model, as a powerful tool for exploring bilateral trade, has a sound theoretical basis [23]. However, the standard gravity model fails to capture the economic distance between countries because geographical distance, which reflects transportation costs, is usually used as a proxy of economic distance in this model, while economic distance also includes institutional differences and other factors. These factors that cannot be captured by geographical distance are likely to comprise the most vital trade distance between countries. Kalirajan, K. (2007) suggested that omitting these factors in the gravity model should lead to incorrect estimates [24]. To cope with this problem, he used a method of estimating a stochastic frontier production function that includes a compound error term, namely, the stochastic frontier gravity model, to measure the combined effects of country-specific factors as the distance between actual trade and potential trade defined by the frontier. This approach has some advantages, especially when information on all factors affecting economic distance is not available. The stochastic frontier gravity model can be written as

$$E{X}_{ijt}=f\left(Z_{ijt};\beta \right)exp\left({\upsilon }_{ijt}-{\mu }_{ijt}\right)$$

(1)

where $E{X}_{ijt}$ represents the actual export of digital services from country i to country j in year t. $Z_{ijt}$ are the determinants of bilateral trade in the gravity model, which include economic scale, geographic distance, language and other natural determinants that remain constant in the short run. $\beta$ is the parameter to be estimated. ${\upsilon }_{ijt}$ and ${\mu }_{ijt}$ represent the random error term and the non-negative inefficiency error term, respectively, and constitute the compound error term together. ${\mu }_{ijt}$ denotes the combined effect of the economic distance factors mentioned above, which result in the difference between actual trade and potential trade. Assume that ${\upsilon }_{ijt}$ follows a normal distribution with a mean of zero, namely, ${\upsilon }_{ijt}~i.i.d.N\left(0,{\sigma }_{\upsilon }^2\right)$, while ${\mu }_{ijt}$ follows a truncated semi-normal distribution, namely, ${\mu }_{ijt}~N^{+} \left({\omega }_{ijt},{\sigma }_{\mu }^2\right)$. The mutual independence of ${\upsilon }_{ijt}$ and ${\mu }_{ijt}$ are assumed as well. ${\omega }_{ijt}$, the mean of ${\mu }_{ijt}$, is set to an exponential function with base e, as shown in Equation (2):

$${\omega }_{ijt}=\mathit{exp}\left({\alpha }_0+\tau S_{ijt}\right)$$

(2)

where ${\alpha }_0$ is a constant. $\tau$ is a parameter to be estimated. $S_{ijt}$ represents variables affecting the inefficiency term of trade. This design makes the heterogeneity of inefficiency terms possible, and thus, enables us to make an analysis of factors that have an impact on trade efficiency.

Values of ${\mu }_{ijt}$ vary from 0 to 1. Economic distance will prevent actual exports from reaching their potential levels, and thus, the export of digital services will fall short of their potential if ${\mu }_{ijt}>$ 0. The export of digital services achieves their potential level under the assumption of no statistical errors if ${\mu }_{ijt}$ is equal to 0. The trade of a certain country is the most efficient if its trade potential is achieved. Therefore, trade efficiency is defined as the ratio of actual trade to potential trade, which is shown in Equation (3):

$$T{E}_{ijt}={\displaystyle \frac{f\left(Z_{ijt};\beta \right)exp\left({\upsilon }_{ijt}-{\mu }_{ijt}\right)}{f\left(Z_{ijt};\beta \right)exp\left({\upsilon }_{ijt}\right)}}=exp\left(-{\mu }_{ijt}\right)=E\left[exp\left(-{\mu }_{ijt}\right)\left|{\epsilon }_{ijt}\right.\right]$$

(3)

A higher value of $T{E}_{ijt}$ means the bilateral trade is more efficient, where the values vary from 0 to 1. Initially, the inefficiency term is assumed to be constant. But scholars propose that the inefficiency term should be assumed to change with time, which is a more reasonable hypothesis. Under the time-varying hypothesis, the structure of ${\mu }_{ijt}$ is described as Equation (4):

$${\mu }_{ijt}=\left\{exp\left[-\eta \left(t-T\right)\right]\right\}{\mu }_{ij}, t=1,2,\cdots ,T$$

(4)

where $\eta$ is the decay parameter and determines how the inefficiency term varies with time. This change is derived from the comprehensive effects of country-specific factors. The inefficiency term will decline over time, and thus, the trade efficiency will have an improvement if $\eta$ > 0. The impediment caused by economic distance, namely, the inefficiency term, will increase over time, and thus, the trade efficiency will be damaged if $\eta$ < 0. The inefficiency term keeps constant with time, and thus, the model is the time-invariant model when $\eta$ is equal to 0. The variation of the inefficiency term is usually defined as follows:

$$\gamma ={\sigma }_{\mu }^2/\left({\sigma }_{\mu }^2+{\sigma }_{\upsilon }^2\right)$$

(5)

where ${\sigma }_{\mu }^2$ and ${\sigma }_{\upsilon }^2$ are the variances of the non-negative inefficiency term and the random error term, respectively. The values of $\gamma$ vary between 0 and 1. All deviations from the trade frontier will result from the random error term if $\gamma$ is equal to 0, while they are derived from the inefficiency term if $\gamma$ is equal to 1. Therefore, the significance of $\gamma$ can be applied to judge whether trade inefficiency exists, namely, whether the stochastic frontier gravity model is applicable.

There are two methods of estimating how variables affect the trade inefficiency term. The first method is to estimate the value of ${\mu }_{ijt}$ and then use it as the explained variable of the regression model, which is called the two-step method. The assumptions of the two-stage method have some drawbacks, which can lead to a bias in its estimation results. These drawbacks are mainly found in the contradiction between two-step assumptions. It is assumed that the mean of the trade efficiency term ${\mu }_{ijt}$ should be a constant in the first stage, while ${\mu }_{ijt}$ is regarded as the explained variable to be estimated in the next stage. Moreover, explanatory variables in the trade inefficiency model are assumed to be mutually independent of those in the stochastic frontier gravity model, which is hardly achievable in practice. The one-step method based on the stochastic frontier gravity model is a great solution to these flaws of the two-step method. We can realize the regression of the trade inefficiency term in the stochastic frontier gravity model by the one-step method, which was thus used to conduct empirical analysis in this study. As shown in Equation (6), the trade inefficiency term can be expressed as the linear function of factors that influence the trade efficiency $k_{ijt}$ and a random error term ${\delta }_{ijt}$:

$${\mu }_{ijt}=\alpha k_{ijt}+{\delta }_{ijt}$$

(6)

According to the above theoretical framework, we constructed the empirical models as follows:

$$\mathit{ln}E{X}_{ijt}={\beta }_0+{\beta }_{1}lnGD{P}_{it}+{\beta }_{2}lnPGD{P}_{it}+{\beta }_{3}lnGD{P}_{jt}+{\beta }_{4}lnPGD{P}_{jt}+{\beta }_{5}lnDI{S}_{ij}+{\beta }_{6}LA{N}_{ij}+{\upsilon }_{ijt}-{\mu }_{ijt}$$

(7)

$${\mu }_{ijt}={\alpha }_0+{\alpha }_{1}IN{F}_{it}+{\alpha }_{2}PA{Y}_{it}+{\alpha }_{3}ET{R}_{it}+{\alpha }_{4}IP{R}_{it}+{\alpha }_{5}OT{H}_{it}+{\delta }_{ijt}$$

(8)

The stochastic frontier gravity model is presented in Equation (7), where $lnGD{P}_{it}$ and $lnGD{P}_{jt}$ represent the logarithm of GDP of export country i and import country j in year t, respectively. These two variables are used to measure the economic scales of export countries and import countries. $lnPGD{P}_{it}$ and $lnPGD{P}_{jt}$ represent the logarithm of GDP per capita of export country i and import country j in year t, respectively. They can reflect the economic development levels of export countries and import countries. $lnDI{S}_{ij}$ is the geographical distance between export country i and import country j, which reflects the cost of digital services trade. $LA{N}_{ij}$ is a dummy variable, whose value is equal to 1 if export country i and import country j have common language, otherwise is 0. Equation (8) presents the trade inefficiency model, where the restrictiveness indices of infrastructure and connectivity ($IN{F}_{it}$), electronic transactions ($ET{R}_{it}$), payment systems ($PA{Y}_{it}$), intellectual property rights ($IP{R}_{it}$) and other barriers ($OT{H}_{it}$) can measure barriers to the digital services trade of export country i in year t from different aspects.

3.2. Data Description

We took the data of 39 countries from 2010 to 2019 as a sample to do empirical research due to the availability of data. The sample countries in this study were Australia, Austria, Belgium, Canada, Chile, China, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, India, Ireland, Israel, Italy, Japan, Korea, Latvia, Lithuania, Luxembourg, Mexico, Netherlands, New Zealand, Norway, Poland, Portugal, Russia, Slovakia, Slovenia, Spain, Sweden, Switzerland, Turkey, the United Kingdom and the United States. The bilateral trade data of the digital services of these countries were sourced from OECD International Trade in Services Statistics (ITSS). However, the OECD ITSS does not provide data on trade in services of China and India. Therefore, the data of import in their trade partner countries were substituted for the data of their exports. And we did something similar for other countries with missing export data. Furthermore, the OECD ITSS does not offer statistical data about export of digital services directly. We identified 10 service industries as digital services according to the definition of digital services proposed by the UNCTAD and calculated the export of digital services of countries by aggregating export of the 10 industries. They included (1) insurance and pension services; (2) financial services; (3) fees for the use of intellectual property rights; (4) telecommunications services; (5) computer services; (6) information services; (7) research and development services; (8) professional and management consulting services; (9) technical trade-related and other business services; and (10) personal, cultural and entertainment services. The data on GDP and GDP per capita are derived from the World Development Indicators of World Bank. The data of bilateral geographic distance and common language were derived from the Gravity of Centre détudes prospectives et d’informations internationals (CEPII). And we obtained the restrictiveness indices of infrastructure and connectivity, electronic transactions, payment systems, intellectual property rights and other barriers from the DSTRI database of OECD. The data of the DSTRI were first published in this database in 2014. Given that the digital services trade policies of countries did not have great changes in the short term, we supplemented the data from 2010 to 2013 with the DSTRI data of 2014.

4. Empirical Analysis

4.1. Stochastic Frontier Gravity Model

The likelihood ratio test (LR test) was used to verify whether the model was applicable and its specification was rational primarily due to the high dependence of stochastic frontier analysis on the functional form of the model. The likelihood ratio statistic LR = −2 [L (H0) − L (H1)] was constructed for testing, where L (H0) and L (H1) are values of the likelihood function under a null hypothesis and alternative hypothesis, respectively. First, we set the null hypothesis that there was no trade inefficiency, namely, $\mu =0$, to test whether the trade inefficiency existed. We then tested whether we should use the time-variant model. Therefore, the null hypothesis of this LR test was that the trade inefficiency term did not change over time, namely, $\eta =0$. Finally, whether the common language variable should be incorporated into the empirical model was tested.

Table 1. Estimation results of stochastic frontier gravity model.

Variable	Time-Invariant Model		Time-Variant Model
	Coefficient	t-Value	Coefficient	t-Value
$lnGD{P}_{it}$	0.760 ***	34.90	0.658 ***	28.62
$lnPGD{P}_{it}$	0.291 ***	8.31	0.193 ***	5.48
$lnGD{P}_{jt}$	0.848 ***	39.63	0.771 ***	34.49
$lnPGD{P}_{jt}$	0.174 ***	5.09	0.0921 **	2.63
$lnDI{S}_{ij}$	−1.179 ***	−34.62	−1.150 ***	−33.08
$LA{N}_{ij}$	1.140 ***	8.09	1.231 ***	8.36
Constant	−7.836 ***	−12.15	−1.367	−0.98
μ	4.267 ***	13.39	6.626 ***	5.28
γ	0.885 ***	197.10	0.897 ***	215.57
η			0.006 ***	5.48
N	12,532		12,532
LR1	538.289		p-value	0.000
LR2	770.481		p-value	0.000
LR3	68.996		p-value	0.000

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

shows the results of the LR tests and estimations for the model. The likelihood ratio statistic LR1 and its p-value showed that there was a trade inefficiency term. Therefore, the stochastic frontier gravity model was suitable for this study. And we chose the time-variant stochastic frontier gravity model to do empirical research because the likelihood ratio statistic LR2 and its p-value showed that the trade inefficiency term varied over time. Whether export countries and import countries have common language is a vital factor that can have an impact on digital services trade. The likelihood statistic LR3 and its p-value suggested that the common languages variable should be included in our model. At the same time, the estimation results of the coefficient of $\mu$ were both positive and significant in the time-variant model and the time-invariant model. This indicates that there were some inefficiency factors that affected the digital services trade as well. The estimates of the coefficient of $\gamma$ were 0.885 and 0.879 in the time-variant model and the time-invariant model, respectively, and both of them were significant. This also shows that incorporating the trade inefficiency term into the stochastic frontier gravity model was helpful for us to acquire consistent estimates of parameters when using the stochastic frontier gravity model. The coefficient of $\eta$ was significantly positive at the level of 1% in the estimation results of the time-variant model, which further proved that the time-variant model was more proper to our study than the time-invariant model.

The regression results of the time-invariant model and the time-variant model show that the GDP and GDP per capita of export countries and import countries significantly boosted the export of digital services in export countries. This indicates that export countries with larger economic scales and higher levels of economic development that had a greater capability of supply and services provided were also more internationally competitive. And import countries with larger economic scales and higher levels of economic development had more demand for digital services, mainly because of their higher levels of consumption. The geographical distance between export countries and import countries ($lnDI{S}_{ij}$) had a significant and negative effect on the export of digital services. Although digital technology brings countries closer and reduces search costs, trust issues and information friction caused by their long distance also impede the development of digital services trade. The estimated coefficient of $LA{N}_{ij}$ was positive and significant. This showed that a common language between export countries and import countries could reduce their communication cost in bilateral trade, and thus, promote trade in digital services.

4.2. Trade Inefficiency Model

The estimation results of the trade inefficiency model by the one-step method are presented in

Table 2. Estimation results of trade inefficiency model.

Stochastic Frontier Gravity Model			Trade Inefficiency Model
Variable	Coefficient	t-Value	Variable	Coefficient	t-Value
$lnGD{P}_{it}$	0.780 ***	100.24	$IN{F}_{it}$	−0.186	−1.09
$lnPGD{P}_{it}$	0.801 ***	44.60	$PA{Y}_{it}$	0.980 ***	3.28
$lnGD{P}_{jt}$	0.860 ***	117.83	$ET{R}_{it}$	−1.765 ***	−5.15
$lnPGD{P}_{jt}$	0.560 ***	38.28	$IP{R}_{it}$	2.904 ***	6.72
$lnDI{S}_{ij}$	−1.072 ***	−92.51	$OT{H}_{it}$	−4.117 ***	−5.42
$LA{N}_{ij}$	0.745 ***	15.89	Constant	0.805 *	2.13
Constant	−21.69 ***	−74.67	N	12,532
LR	354.150		p-value	0.000

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

. The LR test was used to test the rationality of the trade inefficiency model specification. The null hypothesis was that the coefficients of each explanatory variable were simultaneously equal to 0, namely, ${\alpha }_1={\alpha }_2={\alpha }_3={\alpha }_4={\alpha }_5=0$. The likelihood ratio statistic LR and its p-value presented in Table 2 suggest that the null hypothesis should be rejected. Therefore, the trade inefficiency model specified in this study was reasonable.

The coefficient of infrastructure and connectivity ($IN{F}_{it}$) was negative but not significant. This result indicates that infrastructure and connectivity barriers imposed by export countries cannot remarkably improve the export efficiency of digital services. These barriers have an impact on interconnection rules between network operators, which is likely to promote the export efficiency of digital services. Concretely speaking, some infrastructure barriers, such as a mandatory interconnection requirement and interconnection price management, improve infrastructure conditions, and thus, boost the efficiency of digital services suppliers in conducting business [25]. At the same time, appropriate barriers to connectivity reinforce the security of information in the process of digital services trade, which can enhance the trust between consumers and suppliers. But infrastructure and connectivity barriers can also have negative effects on the efficiency of digital services trade. The mandatory interconnection requirement and blocking and restricting the use of communication services increase the costs of telecommunication services suppliers and lower their management efficiency. And then, this may slow the speed at which digital services are delivered by digital services suppliers. Furthermore, too stringent connectivity barriers, such as a data localization requirement and excessive restrictions on cross-border data flow, harm digital services trade efficiency by reducing the utilization of data flow and increasing the construction costs of suppliers [6,26]. The coefficient of infrastructure and connectivity was not significant, which indicates that the positive effects of these barriers on the efficiency of digital services trade were offset by the negative effects to a large extent.

The estimation results show that the coefficient of payment systems ($PA{Y}_{it}$) was significantly positive. This means that the export efficiency of digital services will be damaged if the export country puts trade barriers on payment systems. When discriminatory access conditions for payment and settlement methods are imposed, firms have access to fewer electronic payment methods and the digital payment system must struggle to achieve a scale economy; restrictions on online banking or insurance have the same effect. Therefore, barriers to payment systems increase the costs of cross-border digital payment and hinder the delivery of digital services and realization of digital services trade [27]. Reducing the efficiency of the supervising payment system is another channel for barriers to payment systems to affect the efficiency of digital services trade. The supervision of cross-border payment service suppliers between countries will be inconsistent and these suppliers will face redundant regulations if there is a disparity between domestic payment security standards and international standards. This can reduce the delivery efficiency of digital services as well.

We can conclude that their efficiency of digital services export will be increased if export countries impose barriers to electronic transactions because the coefficient of electronic transactions ($ET{R}_{it}$) was negative and significant at the level of 1%. There are some discrepancies between domestic standards and international standards in contract regulations for cross-border transactions due to the barriers to electronic transactions. These discrepancies increase the costs of making and revising contracts remotely and the uncertainty of long-distance transactions, which thus decrease the efficiency of digital services trade [28]. However, barriers to electronic transactions can fortify the advantages of domestic digital services in terms of export efficiency by boosting the security of business information and reinforcing the price advantage compared with foreign suppliers. Explicit provisions about protecting confidential information in laws and regulations can prevent trade secrets from being disclosed, which is conducive to improving their efficiency of exporting digital services. Moreover, many electronic barriers discriminate against foreign digital services suppliers, such as complex regulatory requirements and cumbersome application processes. These barriers improve the competition advantage of domestic suppliers because these barriers make it easier for foreign suppliers to face delays in the process of digital services trade, and thus, their export efficiency is damaged [29]. The negative coefficient of electronic transactions shows that, in general, the promotion effect of such barriers on digital service trade is obviously greater than the restraining effect.

The coefficient of intellectual property rights ($IP{R}_{it}$) was positive and significant at the level of 1%, which indicates that barriers to intellectual property rights imposed by export countries hurt their efficiency when exporting digital services. Some of the barriers to intellectual property rights can be classified as the lack of enforcement mechanisms to solve intellectual property infringement. The virtuality of networks poses a serious challenge to the enforcement of intellectual property rights [30] because the virtuality of networks makes it possible for digital services suppliers to trade without certification. Consumers cannot inspect services before purchasing or using them. Digital services suppliers need to bear more costs of time and funds primarily due to the lack of enforcement mechanisms, which discourages them from boosting their export efficiency. Another part of these barriers is the discriminatory treatment of foreign suppliers in intellectual property rights. When facing this discriminatory treatment, foreign suppliers are likely to be unwilling to enter the domestic market. Therefore, the domestic digital services market becomes more monopolistic and the innovation efficiency becomes lower, which has a negative effect on the export of digital services.

The coefficient of other barriers ($OT{H}_{it}$) was significantly negative, which means that these barriers could evidently promote the export efficiency of digital services. These barriers improved access conditions for foreign suppliers to conduct business in the domestic market, which can be classified as the market entry barrier. Restrictions on downloads and streaming media and restrictions on online advertising leave foreign suppliers without effective channels to propagandize in the domestic market. And operational requirements and commercial presence requirements that affect cross-border digital trade increase the construction costs and compliance costs of foreign investors. All of these barriers put foreign suppliers at a disadvantage when exporting digital services to the domestic market, and thus, it is easier for domestic suppliers to build market share and achieve a scale economy, which is conducive to improving the export efficiency of digital services. But less competition in the domestic market is likely to reduce the innovation efficiency of domestic suppliers and make it difficult for them to adapt to the complex international market. This can reduce their efficiency of developing business in the international market. The regression result shows that the positive effects of other barriers on the digital services trade efficiency surpassed the negative effects.

4.3. Robustness Test

In order to test whether the baseline regression results were robust in different periods, we divided the sample into three groups according to time intervals and time phases. The regression results of these three groups are presented in

Table 3. Estimation results of robustness test.

Variable	(1)		(2)		(3)		(4)
	Coefficient	t-Value	Coefficient	t-Value	Coefficient	t-Value	Coefficient	t-Value
$IN{F}_{it}$	−0.175	−1.15	−0.066	−0.30	−0.074	−0.59	−0.153	−1.10
$PA{Y}_{it}$	0.806 ***	2.87	1.213 **	2.51	0.649 ***	3.86	0.612 ***	2.74
$ET{R}_{it}$	−1.396 ***	−4.99	−1.477 ***	−4.08	−2.787 ***	−7.40	−1.468 ***	−5.44
$IP{R}_{it}$	2.506 ***	6.44	2.852 ***	4.97	1.472 ***	11.52	2.390 ***	7.38
$OT{H}_{it}$	−3.282 ***	−4.93	−3.567 ***	−3.81	−3.992 ***	−11.29	−3.244 ***	−5.57
Constant	1.144 ***	3.63	1.282 ***	3.73	1.829 ***	10.12	1.153 ***	4.00
N	9,193		3,982		5,211		11,473

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

, where columns (1)–(3) show regression results of the sample from 2013 to 2019, the sample from 2017 to 2019 and the sample from 2013 to 2016. The regression results in these three columns of Table 3 were all consistent with the baseline result shown in Table 2, which shows that the impacts of barriers to digital services trade on the export efficiency of digital services were robust in different periods. In addition, there will be endogenous problems in the model specification, and thus, the estimation results will be biased if there is reverse causality between the explanatory variables and the explained variables in the model. Despite the effect of the export efficiency on digital service trade barriers of exporting countries being relatively limited, we used the data of barriers to digital services trade with a lag of one period to regress in order to weaken the impact of reverse causality. This estimation result, which is shown in column (4) of Table 3, is also consistent with the baseline result, which further manifests the robustness of the estimation results in this study.

5. Extended Analysis

Based on the estimation results in Section 4 and Equation (3), we calculated the export efficiency of digital services of 39 sample countries to their trade partner countries from 2010 to 2019. And the average export efficiency of digital services by each country was calculated, which was applied to make a comparative analysis of the differences in the export efficiency of digital services between countries. The average export efficiency calculated by the arithmetic average method cannot reflect the diverse export values of digital services from a country to different trade partners. Therefore, we used the proportion of digital services exported to different countries in the total exports of the export countries as the weight to calculate the weighted mean of export efficiency. This can be of great help to gain more precise calculation results. Figure 1 shows the arithmetic average, weighted average and rankings of the export efficiency for digital services of each country.

From the perspective of the weighted average export efficiency ranking, the top ten countries were India, Hungary, Ireland, Poland, Japan, China, Latvia, Sweden, Italy and the United States. The United States was the largest exporter of digital services among all sample countries, but its efficiency of export ranked only the tenth among the sample countries. As mentioned above, GDP, GDP per capita and having common language with trade partner countries are all helpful for a country to achieve its export potential. The United States obtained a high export potential in theory, which made its export efficiency lower than the other nine countries. Barriers to electronic transactions and other barriers were shown to have a significant effect on the export efficiency of digital services in Section 4. Some countries with higher levels of protection in these two domains, such as India, Hungary, China and Italy, were more efficient in digital services export. Norway, New Zealand and Chile, who were without any restrictive measures in other barriers, had lower export efficiencies of digital services. And the bottom ten countries in export efficiency had smaller values of DSTRI in electronic transactions, except Chile. Additionally, the empirical results show that barriers to intellectual property rights could significantly hinder the improvement of the export efficiency of digital services. Iceland and Chile had the lowest export efficiency of digital services among all sample countries and China’s export efficiency was evidently lower than that of India, where both are emerging markets, because the former three countries imposed too many barriers to intellectual property rights. The values of DSTRI about intellectual property rights in Iceland and China were both as high as 0.043, which was the largest value among all countries. Chile had a DSTRI value of 0.022 in this field, while sample countries except Iceland, China and Chile have a DSTRI value of zero.

From the perspective of the average export efficiency, the weighted average of the export efficiency of digital services was higher than the arithmetic average in 35 countries. This indicates that most countries were more efficient when exporting to their major digital services trade partner countries. In other words, the exports of digital services in most countries were demand driven. The arithmetic average surpassed the weighted average in Ireland, Luxembourg, Belgium and Canada. This indicates these countries were less efficient when exporting to their major digital services trade partner countries and had greater export potential. Furthermore, the rankings of the weighted average were similar to the rankings of the arithmetic average in most sample countries. But Japan, China, Lithuania and Israel ranked much higher on the weighted average than on the arithmetic average, while the opposite was true in Belgium. This means that Japan, China, Lithuania and Israel were less efficient when exporting to non-main trading partner countries and their export potential needs to be further explored.

6. Discussion and Conclusions

6.1. Discussion

We investigated different trade effects of five kinds of barriers to digital services trade from the research perspective of export efficiency by empirical research. This research had two advantages over existing studies in relevant field. First, we made an analysis of how different types of barriers to digital services trade impact the export efficiency of digital services. Other scholars generally focused on the overall effect of barriers to digital services trade or only do specific research on barriers to cross-border data flow. Therefore, compared with other studies, this study was more detailed and comprehensive, and thus, the conclusions can provide more meaningful recommendations for policy makers. Another virtue of this study was that we undertook research on digital services trade from a novel perspective. Existing studies mainly concentrated on exploring factors that influence the quantity of trade in digital services, while this study highlighted how to improve export efficiency of digital trade services. Nevertheless, there are some shortcomings in our research. For example, we could not attain accurate statistics about digital services trade in different countries, which is likely to have affected empirical results. And the heterogeneity analysis about economic development levels, characteristics of different digital service industries and so on are missing in this study. Hence, it is suggested that relevant research should make progress in the quantification of digital services trade and heterogeneity analysis, which is helpful to acquire more precise empirical results and more detailed recommendations.

6.2. Conclusions

This study used the panel data of 39 countries from 2010 to 2019 and applied the stochastic frontier gravity model to investigate how barriers to digital services influenced the export efficiency of digital services. And based on the regression results, we calculated and analyzed the export efficiency of digital services for sample countries. The main conclusions of this study were as follows. First, the economic scale, level of economic development and common language of import countries and export countries had positive impacts on the export of digital services, while the geographic distance was negatively correlated with the export of digital services. Second, the infrastructure and connectivity barriers of export countries had positive effects on the export efficiency of digital services but the effects were not significant. Both the barriers to electronic transactions and other barriers imposed by export countries could improve the export efficiency of digital services evidently. Barriers to payment systems and intellectual property rights of export countries harm the export efficiency of digital services. Third, most countries were more efficient when exporting digital services to their major trade partner countries, thus showing a demand-driven characteristic.

6.3. Recommendations

The conclusions mentioned above have essential policy implications in the following aspects. (1) Governments should improve the availability of internet infrastructure by relaxing restrictions on the use of communication services and accelerating the construction of internet infrastructure. It is also necessary for governments to ease restrictions on cross-border data flow as much as possible under the prerequisite of ensuring information security. Increasing multinational negotiation about regulations on cross-border data flow should be taken into consideration as well. (2) It is suggested that governments should cultivate the competitiveness of domestic suppliers as soon as possible given that discriminatory barriers to electronic transactions can only work well in the short run. Governments are recommended to actively engage in international negotiations on expanding existing free trade agreements (FTAs) to cover trade in digital services if domestic enterprises have stronger competitiveness. And they should particularly attach importance to reducing discriminatory treatment and integrating rules of cross-border transaction contracts when making international rules for digital services trade. (3) As for the payment systems, governments should prompt the digitalization of financial industries, such as banking and insurance, and diversify methods of payment and settlement under the premise of ensuring payment security. Engaging in international negotiations about the integration of bilateral or multilateral pay payment security standards is of great help to enhance the coordination of cross-border payment security supervision, which should also be advocated. (4) In order to boost the innovation efficiency of digital services, local authorities should improve legislation, refine the enforcement mechanism and reduce discrimination against foreign suppliers in intellectual property rights protection. (5) Operation requirements for foreign suppliers, such as mandatory requirement to use local software and encryption technology and commercial presence requirement, should be properly reduced. And governments should encourage domestic enterprises to innovate independently and improve the quality of digital services to enhance their advantages in export efficiency.