Revisiting the Leontief Paradox in the Digital Era: Technological Specialization and Sustainable Development of Digital Service Trade

Abstract

To address the new challenges of sustainable international trade under the digital transformation, this study aims to explore the relevance and mechanism of the relationship between technological specialization and the sustainable development of digital service trade (focusing on economic sustainability). Based on panel data from 50 economies from 2006 to 2022, the core hypothesis of “whether technological specialization can enhance the sustainable competitiveness of digital service trade by optimizing the global value chain and industrial structure” is verified. An improved index of technological specialization is proposed, breaking through the limitations of traditional indicators, and for the first time introducing the dimension of “knowledge breadth,” reinterpreting the “Leontief Paradox” in the context of digital trade. The study finds that technological specialization significantly enhances the export of digital services, and the effect is more significant in countries with strict intellectual property protection, latecomers in technology, and the European region. Mechanically, this is achieved through improving the position in the global value chain and upgrading the industrial structure. This provides a theoretical breakthrough to solve the technology–trade paradox in the digital age and offers a path for latecomer economies to reconstruct competitive advantages and achieve sustainable development through technological specialization.

1. Introduction

The ongoing technological revolution and industrial transformation continues to deepen globally. Next-generation information technologies, prominently including cloud computing, blockchain, and artificial intelligence, are profoundly reshaping the competitive landscape of global industries. These technologies constitute a strategic high ground for major economies cultivating new quality productive forces. The synergistic evolution of technology and the digital economy has catalyzed novel business models, such as cross-border e-commerce and cloud services, while simultaneously driving structural transformations within the global trading system. This signifies a trend towards the highly integrated development of economic activity and digitalization. Particularly against the backdrop of pandemic-induced disruption and decline in traditional trade, digital services trade has demonstrated remarkable resilience and counter-cyclical growth, reaching a global scale exceeding USD 3.82 trillion in 2022. This trajectory underscores the strategic imperative of understanding the developmental dynamics governing digital services trade.

Currently, global digital services trade exhibits exponential expansion, with its share of total services trade surging from 30.6% in 2015 to 53.7% in 2022. It has thus emerged as the core trajectory for the reconstruction of the global economic and trade order. Nevertheless, extant research has predominantly focused on examining digital services trade within its relationships to the broader economy, specific industries, and digital infrastructure systems. Typically, studies position it either as a dependent variable to analyze its impacts on economic growth, innovation, and global value chains [1,2,3,4], or as an explanatory variable influenced by digital systems to explore developmental barriers [5,6,7] and the mitigating effects of digital trade rules within Regional Trade Agreements [8,9]. A critical limitation persists; existing scholarship largely overlooks the intrinsic connection and underlying mechanisms linking technological specialization to digital services trade.

Significant scholarly divergence exists regarding the relationship between technological specialization and trade performance. One perspective, grounded in comparative advantage theory, posits that technological specialization fosters specific advantages in particular segments through economies of scale. This enables firms to leverage knowledge accumulation for global value chain positioning and facilitates upgrading towards higher value-added segments. However, initial integration often occurs at the low value-added base of the “smiling curve,” where the impact of technological specialization on trade performance manifests more in export scale expansion than quality enhancement. Research on technology spillover effects further indicates that specialized technological clusters reduce innovation costs via knowledge spillovers and promote the transfer of production factors towards technology- and capital-intensive industries [10]. Empirical analyses corroborate that marketization processes significantly elevate export technological sophistication in capital- and technology-intensive sectors by boosting innovation investment, though impacts on labor-intensive industries remain limited. New growth theory emphasizes innovation capability as the core driver of trade growth and quality upgrading [11,12].

Technological specialization, serving as a crucial vehicle for technological innovation [13], can promote the integration of “manufacturing + services,” empower SMEs to construct vertical industrial internet platforms, optimize industrial structure, and consequently accelerate global services trade development. Conversely, another strand of scholarship cautions against potential risks inherent in technological specialization. Models of ‘increasing returns’ in technological competition reveal that excessive specialization may engender path dependency. Studies by Laursen on catch-up economies like Japan and Finland [14], and Montobbio and Rampa on developing nations [15], both found that high-technology specialization does not invariably translate effectively into export growth. The endogenous growth model constructed by Ran Zheng and Zheng Jianghuai further demonstrates that the economic effects of technological specialization exhibit diminishing marginal returns as regional technological capabilities advance [16].

Given the theoretical controversies surrounding the direct link between technological specialization and trade outcomes, it becomes particularly imperative to re-examine this relationship within the digital economy context by focusing on critical mediating pathways. While existing theories partially support the positive role of technological specialization in enhancing global value chain positioning and promoting industrial structure upgrading, the specific theoretical mechanisms through which these two factors mediate the impact of technological specialization on digital services trade require deeper elucidation. Theoretically, technological specialization fosters the advancement of global value chain positioning by enabling economies to climb towards higher value-added, knowledge-intensive segments of the global production network, particularly those involving digital platform orchestration, complex R&D, and data-driven solutions. This elevated global value chain positioning directly facilitates digital services trade by expanding access to global markets for sophisticated digital services, enhancing bargaining power, and embedding firms within networks where cross-border digital service exchange is paramount. Concurrently, technological specialization drives industrial structure upgrading by shifting resources towards advanced technology-intensive and digital service sectors. This structural transformation establishes a robust domestic foundation for digital services production, fosters specialized clusters of digital service providers, and generates endogenous demand for cross-border digital service linkages through enhanced digitalization of the economy and inter-industry spillovers, thereby systematically boosting the scale and scope of digital services trade. Consequently, to systematically explore the impact mechanisms of technological specialization on digital services trade, this study explicitly incorporates the mediating roles of enhanced global value chain positioning and industrial structure upgrading into its analytical framework.

Furthermore, limitations inherent in prevailing measurement methodologies for technological specialization warrant consideration. Technological specialization reflects comparative advantage in trade. Traditional international trade theory utilized the “Leontief Paradox” to explain the dissonance between comparative advantage and observed trade patterns—specifically, the capital-abundant United States exporting labor-intensive goods. This paradox spurred key theoretical extensions, including human capital stratification [17], the technology gap theory [18], and the product lifecycle theory [19], emphasizing that US export competitiveness stemmed from human capital intensity and technological monopolies. Insights from evolutionary economic geography highlight the close relationship between knowledge breadth [20,21] and technological specialization, constrained by “technological proximity” [22] and “cognitive distance” [23,24]. A region’s technological evolution potential critically depends on the interaction between its existing knowledge breadth and specialization depth. Digital technology, with its inherent permeability and integrative capabilities [25,26], further amplifies this interaction, positioning it as central to understanding modern technological competitiveness. However, conventional indicators of technological specialization fail to effectively capture this interactive relationship between knowledge breadth and specialization depth. Employing such indicators to study their impact on digital services trade risks introducing bias and potentially replicating a form of the “Leontief Paradox” in this domain. Therefore, this study proposes an enhancement to the technological specialization index, expanding it from a single-factor production metric to a composite variable encompassing “knowledge breadth and technological advantage,” aiming to provide a more robust explanation for the development of digital services trade.

This investigation employs quantitative methods and panel data spanning 50 economies from 2006 to 2022 to explore the impact mechanisms of technological specialization on digital services trade. Utilizing a fixed-effects model, it tests the following core hypotheses: H1: Enhanced technological specialization significantly promotes the development of digital services trade. H2: Technological specialization promotes the development of digital services trade by enhancing a nation’s position within the global value chain. H3: Technological specialization promotes the development of digital services trade by advancing industrial structure.

The findings reveal a phenomenon analogous to the “Leontief Paradox” within digital services trade, where conventional technological specialization indicators exhibit structural inconsistency with its development. By transcending the limitations of traditional metrics and pioneering the introduction of the “knowledge breadth” dimension to refine technological specialization indicators, this study reinterprets this paradox within the context of digital trade. The refined measure of technological specialization is empirically demonstrated to actively foster digital services trade development. Further mechanism analysis confirms that improved technological specialization influences digital services trade through the mediating effects of rising GVC status and advanced industrial structure. These findings provide crucial empirical evidence for global policymakers designing innovation and digital services trade strategies, while also offering novel theoretical perspectives for research on emerging markets.

The goal of this study is to develop a novel technological specialization index incorporating the “knowledge breadth” dimension to investigate the intrinsic relationship and mechanisms between technological specialization and the sustainable development of digital services trade with a focus on economic sustainability. Therefore the objectives of our research were to (1) verify that technological specialization can enhance the sustainable competitiveness of digital services trade by optimizing global value chain positioning and upgrading industrial structure, and (2) reinterpret the “New Leontief Paradox” in the context of digital trade. This can result in theoretical breakthroughs and practical pathways for latecomer economies to reconstruct competitive advantages and achieve sustainable development through technological specialization.

2. Methods

2.1. Mechanism Hypothesis

2.1.1. Tech Specialization’s Impact on Digital Service Trade

Technological specialization enhances the development of digital services trade. When a country or enterprise strengthens its comparative technological advantage in specific advanced technologies, it also signifies that its capacity to create, deliver, and support complex digital services surpasses that of other countries or enterprises. Technological specialization facilitates the development of more advanced, efficient, reliable, and distinctive digital service offerings. Services with technological superiority possess greater intrinsic value and competitiveness in the global market. Consequently, they can more effectively meet complex international demands, reduce technical barriers to delivery, and allow for premium pricing based on demonstrable technological advantages, thereby promoting an increase in cross-border transactions. Enhanced technological specialization significantly promotes the development of digital services trade.

2.1.2. Global Value Chain Positioning Transmission Path

Drawing upon innovation diffusion theory and product lifecycle theory, technological specialization within a specific domain fosters distinct technological comparative advantages for an economy by enabling profound knowledge accumulation and cultivating innovation capabilities. Extant research demonstrates that the advantages conferred by technological specialization effectively enhance an economy’s position within global value chains upon integration [27]. This specialization elevates the domestic technological content of exported goods and augments the aggregate value chain standing of industries. Upgraded industrial value chain positioning enables economies to occupy core global value chain segments characterized by higher technological complexity and greater value-added potential, thereby facilitating deeper participation in intricate global value chain networks [28]. Within this dynamic, the comparative advantage stemming from technological specialization empowers economies to lead or deliver high-value global digital services [29]. Concurrently, enhanced value chain positioning expands opportunities and bolsters competitiveness for domestic small and medium-sized enterprises engaging in global value chain and providing digital services. Collectively, these mechanisms advance the overall development of digital services trade. Consequently, technological specialization positively promotes the development of digital services trade by enhancing global value chain positioning.

2.1.3. Industrial Structure Upgrade Transmission Path

The second core path for the development of digital service trade driven by technological specialization originates from its driving role in the upgrading of industrial structure. Based on the endogenous growth theory [30] and the theory of industrial structure evolution, technological specialization is the key engine for the advancement of industries towards sophistication. Technological specialization significantly promotes the inter-industry upgrading in developing countries through the “learning by doing” effect and accelerated technology diffusion [31,32], enhancing the overall economic efficiency of factor allocation and innovation momentum. On the one hand, the upgrading of industrial structure generates significant technological spillover effects through the emergence of new business models such as API economy and cloud platforms, reducing the cross-border delivery costs of digital services and market entry barriers; on the other hand, higher-level industrial forms also require and promote deeper coordination of digital trade rules and policy synergy [33] to meet more complex digital transaction needs. In summary, the upgrading of industrial structure can reduce transaction costs, optimize the policy environment, and effectively promote the scale expansion, structural optimization, and overall competitiveness enhancement of digital service trade. Consequently, technological specialization promotes the development of digital services trade by driving industrial structure upgrading. We draw Figure 1 to clearly illustrate the mechanism of interaction between technological specialization and digital service trade.

2.2. Method of Variable Construction

This paper employs a two-way fixed effects model for regression analysis. The Hausman test results (p < 0.01) reject the null hypothesis that the random-effects model is superior, supporting the choice of the fixed-effects model. Considering the time span of the panel data in this paper, according to the conventional handling methods of panel data analysis [34], it is not mandatory to conduct unit root tests and cointegration tests for “panel data” because the fixed-effects model itself has effectively controlled the unobserved heterogeneity and time trends by incorporating individual fixed effects and time fixed effects, thus greatly mitigating the risk of spurious regression caused by non-stationary sequences. In addition, to enhance the robustness of the model and further reduce the possibility of spurious regression, this paper has taken multiple measures: (1) based on a solid theoretical foundation, the core explanatory variables and control variables are constructed and selected to ensure that the model specification has economic implications; (2) strict multicollinearity tests were conducted (all variance inflation factors VIF are far less than 10), which excluded the influence of high linear correlation between variables; (3) time fixed effects were introduced to capture common macroeconomic shocks or time trends; and (4) the results of subsequent robustness tests (such as replacing core variables, adjusting the sample scope, etc.) were consistent with the conclusions of the baseline regression, further confirming the reliability of the model results.

$$ln{trade}_{ct}={\alpha }_0+{\alpha }_1\times {prespecialization}_{ct}+\alpha \times contro{l}_{ct}+{\mu }_c+{\lambda }_t+{\epsilon }_{ct}$$

(1)

$$ln{trade}_{ct}={\beta }_0+{\beta }_1\times {specialization}_{ct}+\beta \times contro{l}_{ct}+{\mu }_c+{\lambda }_t+{\epsilon }_{ct}$$

(2)

Among them, $ln{trade}_{ct}$ is the digital service trade export volume of country c in period t, which is processed in logarithm. prespecialization_ct is the explicit digital technology specialization index of country c in period t; specialization_ct is the revised explicit digital technology specialization index of country c in period t; control is a series of control variables; α₀ and β₀ are constant terms; α₁, α, β₁, β are the corresponding estimated coefficients; μ_c is the individual effect; λ_t is the year effect; and ε_ct is a random interference term, which represents the influencing factors outside the model. The meanings of other variables in the model are the same as those in Equation (2).

To further verify whether technological specialization plays a mediating role in digital service trade through the two channels of influencing the upgrading of the global value chain and the upgrading of the industrial structure, the following model is constructed:

$${Medium}_{ct}={\mathsf{\gamma }}_0+{\mathsf{\gamma }}_1\times {specialization}_{ct}+\mathsf{\gamma }\times contro{l}_{ct}+{\mu }_c+{\lambda }_t+{\epsilon }_{ct}$$

(3)

In Equation (3), Mediumct is the mediating variable, global value chain position (gvcposition), and industrial structure upgrading degree (upgrad).

2.2.1. Improvement of the Technical Specialization Index

The technical specialization index is functionally specified as follows:

$${prespecialization}_{ct}={\displaystyle \frac{X_i/X_{total}}{W_i/W_{total}}}$$

(4)

Here, X_i represents the number of patents held by a country in technology field i, W_i denotes the global number of patents in that field, X_total is the country’s total patent count, and W_total is the global total patent count. A country or region is considered to have a relative comparative advantage in a new-generation information technology field when its share exceeds the global average.

This article explores the relationship between technology and digital service trade. It separately presents the trend charts of the number of global new-generation information technology patents and the growth of digital service trade exports from 2006 to 2022, the number of new-generation information technology patents in China and the growth of digital service trade exports, as well as the top five and bottom five countries in digital service trade exports from 2006 to 2022 in terms of the specialized index of new-generation information technology. The specific content can be found in the Supplementary Materials.

From the standpoint of knowledge spillover and technology diffusion theory, the conventional technology specialization index demonstrates considerable shortcomings within the framework of the digital economy. Research conducted by Lian Junhua illustrates that international trade in the digital age has surpassed the limitations of individual industry technological advantages, currently defined by digital industry involvement, technological integration, and cross-boundary empowerment. The intricate linkages of knowledge components—especially the amplified technology spillover effects in digital service trade—serve as a principal catalyst for the “New Leontief Paradox.” This issue occurs when real trade flows contradict the comparative advantages suggested by the technological specialization index, despite its apparent technical superiority.

This paper suggests an enhanced technical specialization indicator in answer to this theoretical dilemma. The fundamental innovation resides in the incorporation of knowledge breadth. Firstly, it elucidates that “commodities are knowledge carriers” and correlates the extent of diversity in trade categories with the representation of knowledge breadth; secondly, it presents the global knowledge breadth benchmark and equation of the national knowledge breadth coefficient.

The knowledge breadth coefficient serves a dual regulatory function: (1) positive regulatory mechanism: as the national knowledge breadth aligns with the global knowledge breadth, the spillover effect of technological specialization is exponentially intensified, creating a “specialization–diversification” synergistic enhancement loop; (2) and negative constraint mechanism: with limited knowledge breadth, concentrated technological advantages encounter the “island effect.” The incorporation of the regulatory coefficient mitigates the misleading perception of the specialization index and prevents the actual trade benefits from being much inferior to the computed value of the technical specialization index. As shown in Equation (5):

$$\omega ={\displaystyle \frac{{\mu }_1}{\mu }}$$

(5)

${\mu }_1$ represents the variety of import/export commodity categories for a given country. $\mu$ denotes the total variety of import/export commodity categories across all countries.

As shown in Equation (6), the improved technology specialization index is specified by multiplying before specialization by an adjustment coefficient (ω), as follows:

$${specialization}_i=\mathsf{\omega }{ \times prespecialization}_i$$

(6)

This improvement addresses the content that was not captured before specialization, thereby enhancing the specialization ability to reflect technical advantages. Therefore, the specialization provides a more robust and contextually relevant measurement standard for the specialized models of regions or entities.

2.2.2. Variable Definitions

The dependent variable is the volume of digital service exports (trade), which is quantified using digital delivery service trade data from the United Nations Conference on Trade and Development database and modified logarithmically for analytical purposes. The independent variable is the technology specialization index (prespecialization), and the corrected technology specialization index (specialization). The mediating variable is the position in the global value chain (gvcposition). Based on the decomposition framework of export trade flows by Wang et al. [35] and the decomposition framework of value added, and referring to the approach of Xu Zheng et al. [36], this paper adopts three indicators, namely, export domestic value-added rate, upstream index, and the comprehensive index of global value chain division of labor position constructed by the two, to comprehensively measure the global value chain division of labor position of a country-industry.

DVAR is the ratio of domestic value added from a country-industry’s exports, reflecting the profitability of a country in participating in international trade. According to the decomposition framework of exports by Wang et al. [35], exports can be decomposed into eight categories and 16 items. The domestic value-added rate of country s in exports to country r is

$$DVA{R}_{sr}={\displaystyle \frac{DVA\_FIN+DVA\_INT+DVA\_IN{T}_{rex}+RDV\_G}{E^{sr}}}$$

(7)

where EST represents exports.

Upstream index (GVCpt_pos). The added value decomposition framework is functionally specified as follows:

$$\widehat{V}B\widehat{Y}=\widehat{V}L{\widehat{Y}}^D+\widehat{V}L{\widehat{Y}}^F+\widehat{V}L{A}^{F}L{\widehat{Y}}^D+\widehat{V}L{A}^F\times \left(B\widehat{Y}-L{\widehat{Y}}^D\right)$$

(8)

Among them, $\widehat{V}L{\widehat{Y}}^D$ represents the added value of domestic production and consumption. $\widehat{V}L{\widehat{Y}}^F$ is the added value in the export of final products. $\widehat{V}L{A}^{F}L{\widehat{Y}}^D$ is the added value used by the direct importing country for production and absorbed in the country in simple multinational production activities.$\widehat{V}L{A}^F\times \left(B\widehat{Y}-L{\widehat{Y}}^D\right)$ is the added value used by partner countries to produce export products for other countries in complex multinational production activities.

The horizontal direction-based added value decomposition is functionally specified as follows:

$$V_a^{\prime }=\widehat{V}BY=\underset{V_{-}D}{\underbrace{\widehat{V}L{Y}^D}}+\underset{V_{-}RT}{\underbrace{\widehat{V}L{Y}^F}}+\underset{V_{-}GV{C}_{-}S}{\underbrace{\widehat{V}L{A}^{F}L{Y}^D}}+\underset{V_{-}GV{C}_{-}C}{\underbrace{\widehat{V}L{A}^F\times \left(BY-L{Y}^D\right)}}$$

(9)

Based on the vertical distribution, the final product can be broken down into

$$Y^{\prime }=VB\widehat{Y}=\underset{Y_{-}D}{\underbrace{VL{\widehat{Y}}^D}}+\underset{Y_{-}RT}{\underbrace{VL{\widehat{Y}}^F}}+\underset{Y_{-}GV{C}_{-}S}{\underbrace{VL{A}^{F}L{\widehat{Y}}^D}}+\underset{Y_{-}GV{C}_{-}C}{\underbrace{VL{A}^F\times (B\widehat{Y}-L{\widehat{Y}}^D)}}$$

(10)

The forward participation index of the global value chain is the ratio of the domestic value added reflected in the export of intermediate products to the total value added generated by the national sector in production; $GVCP{t}_f={\displaystyle \frac{V_{G}VC}{\widehat{V}BY}}={\displaystyle \frac{V_{G}V{C}_S}{\widehat{V}BY}}={\displaystyle \frac{V_{G}V{C}_C}{\widehat{V}BY}}$; the backward participation index of the global value chain measures the percentage of value added of a national sector’s total output of final products and services that uses domestic and foreign resources produced across borders $GVCP{t}_b={\displaystyle \frac{Y_{G}VC}{VB\widehat{Y}}}={\displaystyle \frac{Y_{G}V{C}_S}{VB\widehat{Y}}}={\displaystyle \frac{Y_{G}V{C}_C}{VB\widehat{Y}}}$. Wang et al. [35] pointed out that a higher degree of forward participation than backward participation means that participants are more actively involved in upstream production activities. This paper uses the ratio of the forward participation index and the backward participation index to reflect the relative upstream degree and construct the global value chain division of labor position indicator GVCpt_pos as follows:

$${GVC}_{pt\_pos}={\displaystyle \frac{{GVC}_{t\_f}}{{GVC}_{t\_b}}}$$

(11)

The comprehensive index of the division of labor position in the global value chain (gvcposition). The domestic value-added rate only partially measures the position of the division of labor in the global value chain. At the same time, although the global value chain position index can accurately reveal the relative position of each participant in specialized production, which is of great significance for evaluating the position of the international division of labor, this index cannot fully reflect the position of the international division of labor. To this end, this paper draws on the method of Zhang Yanping et al. [37] to construct a comprehensive index gvcposition that takes into account both the added value attribute and the relative upstream attribute to measure the position of the division of labor in the global value chain as follows:

$$gvcposition=DVAR\times {GVC}_{pt\_pos}=DVAR\times {\displaystyle \frac{{GVCP}_{t\_f}}{{GVCP}_{t\_b}}}$$

(12)

The degree of industrial structure upgrading is denoted as “upgrad”. Industrial structure upgrading denotes the systematic transformation of industrial composition in tandem with ongoing economic growth, primarily characterized by a persistent rise in the share of the tertiary sector relative to the primary and secondary sectors. Certain scholars employ the proportion of the tertiary industry as a metric to assess the extent of industrial structure upgrading; however, this approach inadequately captures the dynamic evolution of the upgrading process.

The indicator setting of this paper draws on the method of measuring the upgrading of industrial structure by Fu Linghui [38], First, according to the three-industry division, namely agriculture, industry, and service industry, the output value proportion of each industry is taken as a component in the space vector, thus forming a set of three-dimensional vectors X₀ = (x_1,0,x_2,0,x_3,0). Then, the clamping feet θ₁, θ₂ and θ₃ of X₀ and the vectors X₁ = (x_1,1,x_2,1,x_3,1) = (1,0,0), X₂ = (x_1,2,x_2,2,x_3,2) = (0,1,0), X₃ = (x_1,3,x_2,3,x_3,3) = (0,0,1) of the industry arranged from low level to high level are calculated, respectively.

$${\mathsf{\theta }}_{\mathrm{j}}=\mathrm{a}\mathrm{r}\mathrm{c}\mathrm{c}\mathrm{o}\mathrm{s}\left({\displaystyle \frac{\sum _{\mathrm{i}=1}^3\left({\mathrm{x}}_{\mathrm{i},\mathrm{j}}•{\mathrm{x}}_{\mathrm{i},0}\right)}{\left(\sum _{\mathrm{i}=1}^3{\left({\mathrm{x}}_{\mathrm{i},\mathrm{j}}^2\right)}^{1/2}•\sum _{\mathrm{i}=1}^3{\left({\mathrm{x}}_{\mathrm{i},0}^2\right)}^{1/2}\right)}}\right),\mathrm{j}=\mathrm{1,2},3$$

(13)

Based on this, the equation for calculating industrial structure upgrading (upgrad) is defined as shown in Equation (14):

$$upgrad=\sum _{k=1}^3\sum _{j=1}^k{\theta }_j$$

(14)

Among them, the cumulative process $\sum _{j=1}^k{\theta }_j$, the length of the path for the simulation industry to upgrade from low level to high level. The low-level stage ($\theta$₁) was calculated 3 times in the summation (k = 1, 2, 3), intermediate level ($\theta$₂) was calculated 2 times (k = 2, 3), advanced stage ($\theta$₃) only 1 time (k = 3). Economic meaning: If the industrial structure stagnates at a low-level form (such as 1 small, close to x1), the upgrade value will be smaller; if it successfully evolves to a higher-level form (as $\theta$₁, $\theta$₂ increase, deviation from low/middle-level form), the “upgrad” value significantly increases. Therefore, a higher value indicates a more advanced industrial structure.

Control variables. Per capita GDP indicates a country’s economic progress and influences the demand and supply dynamics of digital service businesses. Controlling for per capita GDP (gdp) isolates the influence of economic development levels on digital technology specialization and trade relationships. The inflation rate measures macroeconomic stability, directly affecting the investment environment and trade costs. Controlling for the inflation rate (inflation) eliminates the impact of economic fluctuations on the research findings. Capital stock represents a nation’s accumulation of physical capital, directly impacting production capacity and technology application. Controlling for capital stock (capital) distinguishes the inherent specialization effect of digital technology from effects driven merely by capital accumulation. Economic structural change reflects resource allocation efficiency and technological advancement. Controlling for structural change (structural) avoids misattributing improvements in resource allocation efficiency to the direct influence of digital technology specialization. Controlling for regulatory quality (government) separates the regulatory effect of the policy environment on digital technology specialization and trade relationships. Controlling for human capital (human) excludes the confounding effect of human capital on the specialization effect of digital technology.

2.2.3. Data Sources and Descriptive Statistics

The explanatory variables are derived from the scale of trade in digitally delivered services as recorded in the United Nations Conference on Trade and Development database, while the original patent data for these variables comes from the Incopat Global Patent Database. Industry selection is based on the “Classification of Strategic Emerging Industries (National Bureau of Statistics Order No. 23)” published by the National Bureau of Statistics, focusing specifically on the “new generation information technology industry” as the core research subject, encompassing next-generation information networks; internet, cloud computing, and big data services; artificial intelligence; electronic core industries; and emerging software and new information technology services. Additional control variables are sourced from the World Bank. The research sample includes Argentina, Australia, Austria, Belgium, Brazil, Bulgaria, Canada, China, Colombia, Croatia, Cyprus, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, India, Indonesia, Ireland, Israel, Italy, Japan, Korea, Latvia, Lithuania, Luxembourg, Mexico, Netherlands, New Zealand, Norway, Poland, Portugal, Romania, Saudi Arabia, Singapore, Slovakia, Slovenia, South Africa, Spain, Sweden, Switzerland, Thailand, Turkey, Ukraine, the United Arab Emirates, the United Kingdom, and the United States.

The descriptive statistical results of this study are shown in

Table 1. Descriptive statistics of main variables.

Variable	Variable Description	SD	Mean	Variance	Min	Max
Trade	Digital service trade volume	4250	6.637	3.239	0.000	10.188
Prespecialization	Specialization index	4250	1.045	0.760	0.000	17.043
Specialization	Improved professionalization index	4250	0.754	0.404	0.000	1.433
gvcposition	Global value chain position	4250	0.001	0.028	−0.041	0.061
upgrad	Industrial structure upgrading	4250	2.392	0.120	2.183	2.646
gdp	Per capita GDP	4250	0.312	0.204	0.056	0.679
Inflation	Inflation rate	4250	3.004	3.252	−1.418	16.332
Capital	Capital stock	4250	4.546	1.531	1.833	8.051
Structural	Economic structure	4250	0.735	0.118	0.439	0.999
Government	Government supervision	4250	0.948	0.715	−1.066	2.252
Human	Human capital	4250	0.647	0.147	0.331	0.933

. These countries were selected primarily due to data availability, ensuring relatively complete and continuous trade and patent data in both the UNCTAD and Incopat databases. Additionally, the sample provides broad economic representation, covering major developed economies, significant emerging markets, and several developing nations with regional representativeness, thereby reflecting the global diversity in digital technology development and digital trade. The selection also considered the level of digital technology development, including leading nations (such as the US, Japan, Korea, and Germany) alongside countries actively developing their digital economies, facilitating the examination of impacts across different developmental stages. Furthermore, geographical distribution was a factor, aiming to cover as many global regions as possible (North America, Europe, Asia, and Latin America). For the small number of missing values, scientific methods—linear interpolation and calculation based on average annual growth rates—were employed to minimize the impact of data gaps on the empirical results.

2.2.4. Multicollinearity Diagnosis

To test for multicollinearity among the independent variables, this paper uses the variance inflation factor for diagnosis. The test results in

Table 2. Collinearity diagnosis.

Variable	Variance Inflation Factor (VIF)	1/VIF
Government	4.830	0.207
gdp	4.560	0.219
Human	3.070	0.326
Capital	3.000	0.333
Structural	2.670	0.375
upgrad	2.070	0.483
gvcposition	1.280	0.783
Inflation	1.210	0.824
Specialization	1.150	0.869
Mean VIF	2.650

show that the VIF values for all independent variables are less than 10. This indicates that there is no severe multicollinearity among the variables, meeting the conditions for subsequent empirical analysis.

3. Results

3.1. Baseline Regression Analysis

3.1.1. Test of the Impact of the Technology Specialization Index on Digital Service Trade

Table 3. Baseline regression results.

Variable	Regression Model
	Trade	Trade	Trade	Trade	Trade	Trade	Trade
Prespecialization	−0.078 *	−0.085 *	−0.088 *	−0.079 *	−0.091 *	−0.094 **	−0.101 **
	(0.047)	(0.047)	(0.047)	(0.048)	(0.047)	(0.047)	(0.048)
gdp		2.908 ***	2.908 ***	5.834 ***	5.517 ***	4.027 ***	3.959 ***
		(0.974)	(0.976)	(1.019)	(1.042)	(1.021)	(1.014)
Inflation			0.031 **	0.036 **	0.027 *	0.039 ***	0.037 **
			(0.014)	(0.014)	(0.014)	(0.014)	(0.014)
Capital				−1.148 ***	−1.442 ***	−1.526 ***	−1.507 ***
				(0.112)	(0.117)	(0.117)	(0.118)
Structural					4.422 ***	4.113 ***	3.688 ***
					(1.046)	(1.016)	(1.053)
Government						1.407 ***	1.417 ***
						(0.194)	(0.194)
Human							3.673 **
							(1.655)
_cons	6.719 ***	5.818 ***	5.729 ***	10.012 ***	8.233 ***	7.940 ***	5.816 ***
	(0.057)	(0.305)	(0.310)	(0.502)	(0.647)	(0.642)	(1.144)
Country fixed	Control	Control	Control	Control	Control	Control	Control
Fixed time	Control	Control	Control	Control	Control	Control	Control
N	4250	4250	4250	4250	4250	4250	4250
R2	0.678	0.679	0.679	0.684	0.685	0.689	0.689

Note: *, **, *** represent the significance levels of 10%, 5%, and 1%, respectively. The data in brackets are robust standard errors.

delineates the mechanism of the technology specialization index (prespecialization) concerning the advancement of digital service trade. Column (1), serving as the baseline model, omits any control variables. The findings indicate that the estimated coefficient is 0.078, statistically significant at the 10% level, which contradicts traditional international trade theory, thereby substantiating the presence of the “Leontief paradox” in digital service trade. In columns (2) through (7), a methodical approach of incrementally incorporating control variables is employed for estimation. The results reveal that the coefficients of the enhanced technology specialization index (prespecialization) regarding the development of digital service trade are consistently negative and statistically significant, further corroborating of the “Leontief paradox” in digital service trade.

3.1.2. Test on the Impact of Technological Specialization on Digital Service Trade

Table 4. Post-improvement technical specialization index regression results.

Variable	Regression Model
	Trade	Trade	Trade	Trade	Trade	Trade	Trade
Specialization	0.278 **	0.274 **	0.275 **	0.286 **	0.265 **	0.245 **	0.244 **
	(0.122)	(0.122)	(0.122)	(0.121)	(0.122)	(0.122)	(0.122)
gdp		2.828 ***	2.826 ***	5.781 ***	5.482 ***	4.007 ***	3.940 ***
		(0.971)	(0.973)	(1.016)	(1.041)	(1.021)	(1.014)
Inflation			0.030 **	0.035 **	0.027 **	0.039 ***	0.036 ***
			(0.014)	(0.014)	(0.014)	(0.014)	(0.014)
Capital				1.158 ***	1.429 ***	1.513 ***	1.495 ***
				(0.112)	(0.117)	(0.117)	(0.118)
Structural					4.094 ***	3.795 ***	3.381 ***
					(1.048)	(1.018)	(1.054)
Government						1.392 ***	1.402 ***
						(0.194)	(0.194)
Human							3.484 **
							(1.649)
_cons	6.428 ***	5.548 ***	5.458 ***	9.773 ***	8.135 ***	7.854 ***	5.841 ***
	(0.098)	(0.313)	(0.318)	(0.502)	(0.648)	(0.644)	(1.142)
Country fixed	Control	Control	Control	Control	Control	Control	Control
Fixed time	Control	Control	Control	Control	Control	Control	Control
N	4250	4250	4250	4250	4250	4250	4250
R2	0.678	0.679	0.679	0.685	0.685	0.689	0.689

Note: **, *** represent the significance levels of 5% and 1%, respectively. The data in brackets are robust standard errors.

delineates the process of the enhanced technical specialization index (specialization) concerning the advancement of digital service commerce. Column (1) represents the baseline model devoid of any control variables. The findings indicate that the estimated coefficient of specialization is 0.278, which is statistically significant at the 5% level. The positive correlation indicates that a 1 percentage point rise in technical specialization will enhance the growth of digital service trade by approximately 28%, demonstrating the pivotal significance of technical specialization in advancing digital service trade. Columns (2) to (7) employ a methodical approach of incrementally including control variables for estimation. The findings indicate that the coefficients of the enhanced technical specialization index (specialization) regarding the advancement of digital service trade are all significantly positive, signifying that the enhanced technical specialization index (specialization) has substantially facilitated the growth of digital service trade.

3.2. Robustness and Endogeneity Test

3.2.1. Robustness Test

To verify the legitimacy and rigor of the findings, the regression results are subjected to robustness testing. The test results are shown in

Table 5. Results of robustness checks using other methods.

Variable	Robustness Test
	Replace the Explained Variable	Remove Outliers	Changing the Estimation Method
L.trade
Specialization	0.234 *	0.249 **	0.039 *
	(0.130)	(0.122)	(0.021)
gdp	4.094 ***	4.736 ***	0.666 ***
	(1.072)	(1.030)	(0.184)
Inflation	0.036 ***	0.036 **	0.003
	(0.014)	(0.014)	(0.002)
Capital	−0.654 ***	−1.500 ***	−0.235 ***
	(0.130)	(0.123)	(0.026)
Structural	0.583	3.718 ***	0.383 *
	(1.097)	(1.087)	(0.205)
Government	0.518 ***	1.345 ***	0.227 ***
	(0.192)	(0.200)	(0.036)
Human	0.412	3.507 **	0.709 **
	(1.707)	(1.643)	(0.314)
_cons	7.483 ***	5.404 ***	1.896 ***
	(1.151)	(1.125)	(0.226)
Country Fixed	control	control	control
Fixed time	control	control	control
N	4250	4250	3995
R2	0.692	0.690

Note: *, **, *** represent the significance levels of 10%, 5%, and 1%, respectively. The data in brackets are robust standard errors.

. Initially, substitute the measurement technique of the dependent variable. The aggregate volume of digital service imports and exports serves as the dependent variable. The directional significance of the core variable (specialization) aligns with the baseline model, hence reinforcing the robustness of the finding. Secondly, remove outliers. The continuous variables, including digital service trade volume and specialization, are winsorized at the upper and lower 1% thresholds to mitigate the influence of outliers. Third, considering the potential issues of intra-group autocorrelation and heteroscedasticity in the static panel model, the panel-corrected standard error (PSE) is employed to cluster the nation and time dimensions, thereby more rigorously managing the error structure. The PSE model confirms the statistical reliability of the baseline data, and the influence of the important variables is unaffected by the adjustment of the error structure.

3.2.2. Endogeneity Treatment

This study employs the following tests to properly address the potential bidirectional causal relationship between digital service commerce and technological specialization, as well as the endogeneity issue arising from the knowledge spillover effect: Initially, differential GMM dynamic estimation. A dynamic panel model is established for the lagged term of the explained variable (L.trade), with both the lagged term (L.trade) and the enhanced technological specialization index (specialization) considered as endogenous variables, while the other control variables are regarded as instrumental variables. The estimation findings from the double fixed-effects model indicate (

Table 6. The text translates to Conduct endogeneity test using GMM and 2SLS methods.

Variable	Endogeneity Test
	GMM	L. Specialization	Neighboring Specialization
L.trade	0.801 ***
	(0.024)
Specialization	0.588 ***	1.036 ***	2.572 ***
	(0.211)	(0.184)	(0.359)
gdp	4.091 ***	−2.602 ***	−1.692 ***
	(0.889)	(0.556)	(0.564)
Inflation	−0.086 ***	−0.022	−0.009
	(0.012)	(0.014)	(0.014)
Capital	0.032	0.329 ***	0.323 ***
	(0.093)	(0.045)	(0.048)
Structural	−4.866 ***	2.272 ***	0.474
	(1.392)	(0.671)	(0.705)
Government	−1.299 ***	−0.340 ***	−0.557 ***
	(0.261)	(0.131)	(0.138)
Human	0.580	6.673 ***	6.459 ***
	(0.956)	(0.572)	(0.550)
AR(2)	0.247
Hansen Test value	0.560
Kleibergen–Paap rk LM statistic		1073.35	390.60
		[0.000]	[0.000]
Kleibergen–Paap rk Wald F statistic		2673.02	495.81
		{16.38}	{16.38}
_cons		−0.312	−0.217
		(0.400)	(0.393)
Country Fixed	Control	Control	Control
Fixed time	Control	Control	Control
N	3750	4000	4250
R2		0.138	0.092

Note: *** represent the significance levels of 1%. The data in brackets are robust standard errors. L. represents the lag term, AR test shows the p-value, Hansen test shows the Chi2 value. The p-value of Kleibergen–Paap rk LM statistic is in square brackets, and the critical value of Stock-Yogo test at the 10% level is in curly brackets.

, column (1)) that the coefficient of the primary explanatory variable is significantly positive at the 1% level. Simultaneously, it successfully passed the AR(2) test and the Hansen test, reinforcing the empirical observation of a “Leontief paradox” in digital service commerce.

Second, two-stage instrumental variable estimate. Two-stage least squares (2SLS) is used for supplementary verification. The lagged one-period core explanatory variable (L. specialization) is designated as the instrumental variable. The selection criteria are as follows: firstly, there exists a strong correlation between the primary explanatory variable and its lagged counterpart; secondly, the lagged version of the primary explanatory variable (L. specialization) exhibits negligible correlation with the disturbance term. According to Moretti [39], regional exogenous technology shocks serve as instrumental factors. This paper employs the spatial mean of the technological specialization index of neighboring countries as an instrumental variable, as the technological framework of geographically adjacent nations is shaped by historical path dependence or natural endowments, and exhibits no direct correlation with the country’s digital service trade exports. Columns (2) and (3) of Table 6 demonstrate that, given both the Kleibergen–Paap LM test and the Wald F statistic significantly reject the weak instrumental variable hypothesis, the specialization coefficient remains resilient and positive.

Upon accounting for endogeneity, the influence of technological specialization on digital service trade is markedly positive. This suggests that technological specialization substantially fosters the advancement of digital service trade. Consequently, the two instrumental variables used in this study are more rational and efficacious, and the regression outcomes align with prior findings, thus confirming the robustness of the key conclusions of this paper.

3.3. Heterogeneity Analysis

3.3.1. Heterogeneity of Intellectual Property Protection Level

As shown in the fourth and fifth columns of

Table 7. Intellectual property protection and the heterogeneity of technological development levels.

Variable	Intellectual Property Protection Level		Level of Technological Development
	Low Protection	High Protection	Underdeveloped Technology	Advanced Technology
Specialization	−0.007	0.430 *	0.279 **	0.197
	(0.121)	(0.232)	(0.129)	(0.373)
gdp	9.910 ***	−1.851	8.401 ***	−9.269 ***
	(1.463)	(1.271)	(1.245)	(1.851)
Inflation	−0.007	−0.048 *	0.035 **	−0.313 ***
	(0.009)	›(0.029)	(0.015)	(0.086)
Capital	−0.781 ***	−1.189 ***	−1.331 ***	−2.494 ***
	(0.147)	(0.193)	(0.111)	(0.861)
Structural	−2.904 ***	6.279 ***	1.010	19.247 ***
	(1.089)	(1.834)	(1.085)	(7.271)
Government	−0.919 ***	3.383 ***	1.205 ***	1.034
	(0.178)	(0.359)	(0.200)	(0.697)
Human	−5.903 ***	11.719 ***	7.856 ***	−1.468
	(1.375)	(2.619)	(1.936)	(3.865)
_cons	14.295 ***	−5.428 **	2.529 **	11.136 **
	(0.817)	(2.357)	(1.248)	(4.967)
Country fixed	Control	Control	Control	Control
Fixed time	Control	Control	Control	Control
N	2130	2105	3315	935
R2	0.722	0.740	0.704	0.555

Note: *, **, *** represent the significance levels of 10%, 5%, and 1%, respectively. The data in brackets are robust standard errors.

, technological specialization significantly promotes digital service trade in countries with high levels of intellectual property protection, but has no significant impact on countries with low protection levels. This difference can be attributed to the impact of the institutional environment on innovation incentives and technological transformation: for countries with high protection levels, strong patent systems, and legal enforcement effectively ensure technological exclusivity, incentivizing enterprises to invest in specialized technology. The resulting technological advantage is more easily transformed into differentiated competitiveness in digital services and higher export value-added (for example, SaaS services protected by specific algorithm patents can achieve a market premium). For countries with low protection levels, weak intellectual property enforcement leads to significant amplification of the spillover effect of technology, with enterprises facing serious “free-riding” risks and imitation behavior. Specialized technology is easily reverse-engineered, weakening its market exclusivity and trade value, and hindering the transformation of technological advantages into trade advantages.

3.3.2. Heterogeneity of Technological Development Level

As shown in the fourth and fifth columns of Table 7, technological specialization significantly promotes the digital service trade in technologically underdeveloped countries, with no significant impact on technologically developed countries. This “catch-up effect” versus “frontier dilemma” may be due to technologically developed countries (such as the US, Japan, Germany, South Korea, etc.): Their technological development has approached or reached the global forefront, and further specialization is prone to fall into the “innovation rigidity” trap and diminishing marginal returns. The highly mature and possibly monopolistic market structures (such as large technology platform ecosystems) also greatly increase the difficulty for new entrants to achieve breakthroughs through specialization.

Technologically underdeveloped countries: Through technological specialization, they can effectively absorb, apply, and improve mature technologies (i.e., the “technology diffusion” effect), achieving “catch-up” in specific fields. For example, Southeast Asian countries have utilized mature e-commerce and payment technologies to rapidly specialize in the development of localized digital service platforms (such as Grab, Gojek), significantly enhancing service efficiency and export capacity.

3.3.3. Regional Heterogeneity

As shown in the fourth and fifth columns of

Table 8. Heterogeneity in regional and economic development levels.

Variable	Different Regions			Digital Infrastructure
	North America	Europe	Asia	Low	Medium	High
Specialization	−0.446 ***	0.238 *	−0.104	−0.026	0.358 **	0.319
	(0.158)	(0.136)	(0.317)	(0.120)	(0.166)	(0.265)
gdp	−1.314	−4.753 **	13.824 ***	4.373 **	−4.151 *	4.534 **
	(0.872)	(1.989)	(1.689)	(1.987)	(2.489)	(2.177)
Inflation	−0.120 ***	0.101 ***	0.178 ***	−0.017 *	0.108 ***	−0.044
	(0.025)	(0.022)	(0.027)	(0.009)	(0.039)	(0.039)
Capital	1.450 ***	−0.704 **	−1.340 ***	−1.306 ***	−0.512	−0.904 ***
	(0.275)	(0.313)	(0.169)	(0.189)	(0.637)	(0.202)
Structural	−2.685 *	7.443 ***	5.217 **	0.928	2.727	19.613 ***
	(1.552)	(1.848)	(2.649)	(1.315)	(1.994)	(2.672)
Government	2.089 ***	0.355	4.797 ***	1.106 ***	0.803 *	−0.922 *
	(0.264)	(0.317)	(0.412)	(0.207)	(0.471)	(0.536)
Human	55.860 ***	−7.021 ***	8.006 ***	7.614 ***	−12.345 ***	8.333 ***
	(2.643)	(2.409)	(2.680)	(1.744)	(3.394)	(3.196)
_cons	−33.939 ***	9.545 ***	−0.618	6.777 ***	15.483 ***	−11.024 ***
	(2.145)	(1.908)	(2.242)	(0.648)	(3.028)	(3.011)
Country fixed	Control	Control	Control	Control	Control	Control
Fixed time	Control	Control	Control	Control	Control	Control
N	765	2550	935	1382	1402	1401
R2	0.932	0.625	0.851	0.928	0.737	0.787

Note: *, **, *** represent the significance levels of 10%, 5%, and 1%, respectively. The data in brackets are robust standard errors.

, technological specialization has a significant positive impact on the digital service trade of European countries. This has a significant negative impact on North American countries, and no significant impact on Asian countries. The differences between regions profoundly reflect the role of institutional coordination, market structure, development stage, and regulatory environment.

First, market structure and innovation dynamics. As a global technological frontier, its technology is highly specialized and often concentrated in a few dominant, highly monopolistic fields (such as operating systems, search engines, and social media). This may lead to a “winner-takes-all” situation and innovation rigidity, with limited space for new entrants. Resource allocation and spillover effects: R&D investment has reached a high level of saturation, and excessive specialization may lead to over-concentration of resources, occupying the innovative space of other emerging technology fields. At the same time, strong technological spillover effects (knowledge dissemination, talent mobility, open-source movements) weaken the direct monopolistic pull of domestic technological advantages on trade. Second, regulatory environment: Strict digital regulatory policies (such as the California Consumer Privacy Act CCPA, ongoing antitrust reviews of technology giants at the federal level) increase compliance costs and may reduce the efficiency of converting technological specialization into trade competitiveness. Regulatory uncertainty also affects enterprises’ long-term investment decisions.

First, the coordination effect of the system is significant. The European Union has greatly reduced the trade barriers between member states in the digital service sector through strong regional coordination mechanisms (such as the General Data Protection Regulation (GDPR) for unified data rules, the GAIA-X plan for building a European independent cloud infrastructure, and the promotion of the digital identity wallet eIDAS 2.0, etc.), creating a vast and rule-based single digital market. This provides a broad internal market transformation space for the results of technological specialization. Second, the industrial specialization model: European companies generally adopt a strategy of deepening their efforts in niche technical fields (such as Dutch ASML in lithography equipment, German SAP in enterprise software, and Swedish Ericsson/Finnish Nokia in communication technology), forming irreplaceable “hidden champions.” This specialization enables them to embed themselves in the mid-to-high-end links of the global value chain, directly driving the growth of exports of high-value-added digital services, rather than relying on low-cost competition.

First, there is a high degree of heterogeneity in the development stage and technological path. The region includes economies at different stages of development, ranging from technological leaders (Japan, South Korea, Singapore) to fast followers (China, India) to beginners (some Southeast Asian countries), with significant differences in their technological specialization directions, target markets, and policy priorities. Second, there is a lack of regional coordination mechanisms: Compared to the European Union, Asia lacks strong regional technical standard coordination and digital governance frameworks (such as data cross-border flow rules, digital tax policies). The regulation of countries is highly fragmented (for example, China’s strict data localization requirements are in tension with Japan’s relatively open rules; there are different standards for digital tax collection among ASEAN member states), which significantly increases the transaction costs and compliance complexity of digital service trade within the region. Third, value chain positioning and competitive models: Many Asian economies are deeply involved in the global value chain, but mainly rely on cost efficiency or scale advantages. Their technological specialization achievements are difficult to effectively transform into cross-regional trade advantages based on technical standards in the absence of regional rule coordination (for example, a specific industry software solution developed in one country may be difficult to deploy and promote in other Asian countries due to data cross-border restrictions).

3.3.4. Digital Infrastructure Level

The quality of digital infrastructure is assessed using the ICT capability index from the United Nations Conference on Trade and Development, which includes fixed-line and mobile phone users, Internet accessibility, and server security. The regression outcomes are presented in columns (4), (5), and (6) of Table 8. The findings indicate that in an economy characterized by a deficiency in digital infrastructure, the establishment of successful digital service export capabilities through technological specialization is challenging. Upon surpassing the fundamental threshold of infrastructure, the facilitative impact of technological specialization on digital service commerce commences to manifest.

In economies with advanced infrastructure, the trade promotion impact of technological specialization is no longer substantial. Low levels of digital infrastructure hinder the scalability of digital services, diminishing the beneficial externalities of technological spillovers. Conversely, economies with advanced infrastructure have typically established comprehensive digital frameworks, resulting in diminished disparities in technological specialization due to infrastructure uniformity. The impact of technological specialization on trade promotion exhibits a marginal decline, rendering it insignificant.

3.4. Mechanism Test Analysis

3.4.1. Analysis of the Global Value Chain Status Mechanism

The impact of technical specialization on the upward climb of the global value chain is verified, as shown in column (1) of

Table 9. Mechanism test results of global value chain status and industrial structure upgrading.

Variable	Mechanism Test
	Gvcposition	Upgrade
Specialization	0.002 **	0.004 **
	(0.001)	(0.002)
gdp	−0.006 ***	0.011 ***
	(0.001)	(0.003)
Inflation	−0.069 ***	−0.073 ***
	(0.012)	(0.022)
Capital	−0.017 ***	−0.007 **
	(0.002)	(0.003)
Structural	−0.031 ***	0.235 ***
	(0.011)	(0.026)
Government	0.104 ***	2.236 ***
	(0.009)	(0.024)
Human	0.002 **	0.004 **
	(0.001)	(0.002)
_cons	0.026 ***	0.040 ***
	(0.006)	(0.013)
Country fixed	control	control
Fixed time	control	control
N	4250	4250
R2	0.778	0.956

Note: **, *** represent the significance levels of 5%, and 1%, respectively. The data in brackets are robust standard errors.

, indicating that technical specialization has a significant positive impact on the upward climb of the global value chain (coefficient: +0.002 **, p < 0.01). It is evident that the position of the global value chain shows a significant mediating effect between technical specialization and digital service trade. This suggests that technical specialization may indirectly affect the scale expansion, structural optimization, and competitiveness enhancement of digital service trade by reshaping the division of labor role of the country in the global value chain and improving negotiation capabilities.

3.4.2. Analysis of the Mechanism of Industrial Structure Upgrading

The impact of technological specialization on industrial structure upgrading is verified, as shown in column (2) of Table 9, indicating that technological specialization has a significant positive effect on the upward movement of the global value chain (coefficient: +0.004 **, p < 0.01). This result supports the mechanism through which technological specialization indirectly promotes the development of digital service trade by driving the industrial structure to a higher level.

Technological specialization serves as a core driver of digital service trade development through dual channels. This dynamic operates primarily by enhancing global value chain positioning. Simultaneously, it accelerates industrial structure upgrading, as confirmed by the two-step test.

4. Discussion

4.1. Comparison with Previous Studies

The relationship between technological specialization and trade performance has long been a subject of debate within academia. Mainstream theories typically emphasize its positive effects. Traditional international trade theory posits that specialization shapes comparative advantage through economies of scale and knowledge accumulation. Technology spillover theory indicates that specialized clusters can reduce innovation costs [10]. New growth theory, furthermore, views specialization as an endogenous driver of innovation capability, propelling trade growth and product quality enhancement [11,12].

However, other research reveals nonlinear impacts. Models of increasing returns suggest that excessive specialization can lead to path dependency. Empirical studies show that high-tech specialization does not necessarily lead to export growth [14,15]. The model by Ran Zheng and Zheng Jianghuai further predicts that its economic effects diminish as the technological level rises [40]. Crucially, these studies predominantly explore the relationship between technological specialization and trade performance based on traditional trade frameworks and have not been grounded in the context of contemporary international trade.

Significantly diverging from prior research, this paper re-examines the impact of technological specialization on trade performance from the perspectives of new context, new object, and new mechanisms, thereby seeking to explain the “new Leontief paradox.” First, this paper focuses on a new context. The rise of digital services trade constitutes a fundamentally new trade environment, profoundly altering traditional trade forms and patterns, and presenting novel challenges and opportunities.

Second, the research object shifts to digital services trade itself. Its characteristics—intangibility, immediacy, cross-border nature, and high value-added—result in trade flows and factor distributions that differ significantly from traditional goods trade. Furthermore, when investigating the “Leontief paradox,” the scope expands from a traditional focus on single nations to a global perspective. This enhances universality and applicability, facilitates international comparison and cooperation, helps reveal diverse influencing factors, and promotes theoretical innovation.

Third, this paper incorporates new measurement factors. The technology-driven, data-centric nature of digital services trade dictates that the mismatch between its trade flows and factor distribution is intrinsically linked to knowledge and technology. Incorporating this new measurement factor not only challenges traditional international trade theory but also provides a fresh perspective for understanding digital services trade.

4.2. Research Implications

The formation mechanism of competitive advantage in digital service trade does not rely on superiority in a single technological domain, but rather stems from the alignment between knowledge integration capability and the degree of technological specialization. This coordination between technological breadth and depth constitutes a new source of national competitive advantage in the digital era and offers a fresh theoretical perspective for unraveling the Leontief Paradox. Therefore, developing digital trade requires breakthroughs in both technological sophistication and knowledge breadth. This involves driving innovation and breakthroughs in digital technologies like big data, cloud computing, and artificial intelligence, while simultaneously building interdisciplinary knowledge maps. It necessitates promoting the extensive application and deep integration of high-tech in the service trade sector, catalyzing new business models and formats, strengthening the cultivation of interdisciplinary talent, and achieving a dynamic balance between technological height and knowledge width.

A country’s position in the global value chain and its industrial structure upgrading are critical pathways driving the development of digital service trade. The development of digital service trade lies in the deep integration of digitalization and industrialization, rather than following a linear path of industrialization first followed by digitalization [41]. Therefore, it is crucial to break through the traditional specialization paradigm, encourage enterprises to increase investment in digital technology R&D, promote the deep integration of industrial digitalization and digital industrialization, enhance the added value and competitiveness of digital service products, and convert technological advantages into leadership in trade rule-making.

In the European region, due to institutional advantages in technical standards coordination and cross-border data flows, the marginal effect of technological specialization on trade exceeds that of other regions, validating the amplifying role of regional integration on technology spillovers. Conversely, the driving effect of technological specialization on digital service trade is statistically stronger for technologically latecomer countries than for technologically advanced countries, confirming the potential for leapfrogging development. Therefore, for developing countries, the focus should be on achieving deep specialization in 1–2 core digital technologies. They should strengthen international R&D collaboration, actively participate in open-source communities and patent cross-licensing to expand knowledge breadth, while simultaneously promoting the alignment and mutual recognition of domestic digital service trade standards with international standards to reduce barriers.

5. Conclusions

This study confirms that technological specialization is a key driver for enhancing the competitiveness of digital services exports and promoting their economically sustainable development. Based on an improved measurement methodology, the research finds that technological specialization effectively elevates a country’s position in the global division of labor and drives industrial upgrading, thereby strengthening the sustainable competitiveness of digital services trade. This positive impact is particularly pronounced in countries with strict intellectual property protection, technologically lagging nations, and within the European region (a summary of key results is provided in Appendix A). These findings offer new insights for resolving the tensions between technological advancement and trade development in the digital era, and chart a path for latecomer economies to build new competitive advantages and achieve sustainable development through deepening technological specialization. Although this study advances the understanding of the relationship between technological specialization and digital trade, its limitations should be noted: the scope is limited to national-level data, lacking details on micro-level enterprise innovation behavior; measurement relies on patent data, making it difficult to fully capture intangible technological elements in digital services like algorithms and data architecture; differences in digital trade rules across countries may interfere with accurately measuring the effects of rule-making capacity; furthermore, the analysis does not systematically demonstrate how the degree of industrial specialization and the heterogeneity of national economic structures lead to differential outcomes of technological specialization on digital services exports.

In response to these limitations, future research can (1) construct more micro-level datasets to track the connections between corporate research and development, patents, and trade, (2) develop new indicators to better measure important intangible digital technology, (3) more deeply explore how digital platform ecosystems change the governance model of global value chains, and (4) conduct comparative analyses across distinct groups of countries (e.g., grouped by development stage, dominant industry specialization, or regional economic bloc) to investigate how national economic structures moderate the relationship between technological specialization and digital trade competitiveness. This comparative approach would help validate the generalizability of our findings and identify context-specific pathways. These directions will help to understand more comprehensively and deeply the issue of technological specialization in digital trade, providing more precise references for policy-making.